⚙️  Running in WANDB offline mode
⚙️  Running in WANDB offline mode
⚙️  Running in WANDB offline mode
⚙️  Running in WANDB offline mode
Loading vision encoder from pretrained_models/videollama3_7b_local ...
`torch_dtype` is deprecated! Use `dtype` instead!
`torch_dtype` is deprecated! Use `dtype` instead!
`torch_dtype` is deprecated! Use `dtype` instead!
`torch_dtype` is deprecated! Use `dtype` instead!
`torch_dtype` is deprecated! Use `dtype` instead!
`torch_dtype` is deprecated! Use `dtype` instead!
`torch_dtype` is deprecated! Use `dtype` instead!
`torch_dtype` is deprecated! Use `dtype` instead!
Vision encoder loaded. hidden_size=1152
The tokenizer you are loading from 'pretrained_models/videollama3_7b_local' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
Compressor type=local_attn_conv layers=8 heads=8 HW=16×16
Decoder layers=3 (=log2 max_output_frames) max_output_frames=8
The tokenizer you are loading from 'pretrained_models/videollama3_7b_local' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
The tokenizer you are loading from 'pretrained_models/videollama3_7b_local' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
Force set image size to be 448
The tokenizer you are loading from 'pretrained_models/videollama3_7b_local' with an incorrect regex pattern: https://huggingface.co/mistralai/Mistral-Small-3.1-24B-Instruct-2503/discussions/84#69121093e8b480e709447d5e. This will lead to incorrect tokenization. You should set the `fix_mistral_regex=True` flag when loading this tokenizer to fix this issue.
Dropped 79/977423 samples without precomputed feature under /share/dataset/vision_feat_cache/
Dropped 79/977423 samples without precomputed feature under /share/dataset/vision_feat_cache/
Dropped 79/977423 samples without precomputed feature under /share/dataset/vision_feat_cache/
Dataset size: 977344
[2026-08-19 17:20:35,031] [INFO] [real_accelerator.py:219:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2026-08-19 17:20:35,032] [INFO] [real_accelerator.py:219:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2026-08-19 17:20:35,064] [INFO] [real_accelerator.py:219:get_accelerator] Setting ds_accelerator to cuda (auto detect)
Dropped 79/977423 samples without precomputed feature under /share/dataset/vision_feat_cache/
[2026-08-19 17:20:59,820] [INFO] [real_accelerator.py:219:get_accelerator] Setting ds_accelerator to cuda (auto detect)

  0%|          | 0/19090 [00:00<?, ?it/s][rank0]:[W819 17:21:10.467170055 reducer.cpp:1431] Warning: find_unused_parameters=True was specified in DDP constructor, but did not find any unused parameters in the forward pass. This flag results in an extra traversal of the autograd graph every iteration,  which can adversely affect performance. If your model indeed never has any unused parameters in the forward pass, consider turning this flag off. Note that this warning may be a false positive if your model has flow control causing later iterations to have unused parameters. (function operator())
[rank1]:[W819 17:21:10.557646050 reducer.cpp:1431] Warning: find_unused_parameters=True was specified in DDP constructor, but did not find any unused parameters in the forward pass. This flag results in an extra traversal of the autograd graph every iteration,  which can adversely affect performance. If your model indeed never has any unused parameters in the forward pass, consider turning this flag off. Note that this warning may be a false positive if your model has flow control causing later iterations to have unused parameters. (function operator())
[rank3]:[W819 17:21:11.822234572 reducer.cpp:1431] Warning: find_unused_parameters=True was specified in DDP constructor, but did not find any unused parameters in the forward pass. This flag results in an extra traversal of the autograd graph every iteration,  which can adversely affect performance. If your model indeed never has any unused parameters in the forward pass, consider turning this flag off. Note that this warning may be a false positive if your model has flow control causing later iterations to have unused parameters. (function operator())
[rank2]:[W819 17:21:11.887444284 reducer.cpp:1431] Warning: find_unused_parameters=True was specified in DDP constructor, but did not find any unused parameters in the forward pass. This flag results in an extra traversal of the autograd graph every iteration,  which can adversely affect performance. If your model indeed never has any unused parameters in the forward pass, consider turning this flag off. Note that this warning may be a false positive if your model has flow control causing later iterations to have unused parameters. (function operator())

  0%|          | 1/19090 [00:09<52:33:33,  9.91s/it]
  0%|          | 2/19090 [00:11<25:41:20,  4.84s/it]
  0%|          | 3/19090 [00:13<20:30:19,  3.87s/it]
  0%|          | 4/19090 [00:15<15:06:29,  2.85s/it]
  0%|          | 5/19090 [00:16<12:07:18,  2.29s/it]
  0%|          | 6/19090 [00:18<12:27:53,  2.35s/it]
  0%|          | 7/19090 [00:20<10:36:09,  2.00s/it]
  0%|          | 8/19090 [00:21<9:23:16,  1.77s/it] 
  0%|          | 9/19090 [00:24<10:39:38,  2.01s/it]
  0%|          | 10/19090 [00:25<9:28:14,  1.79s/it]
                                                    
{'loss': 1.701, 'grad_norm': 24.286043167114258, 'learning_rate': 2.3572551073860661e-07, 'recon_loss': 1.701093378663063, 'recon_static': 1.6747213035821915, 'recon_motion': 0.02637208136729896, 'recon_norm': 44.28108882904053, 'recon_norm_std': 0.3493720218539238, 'target_norm': 1.6199289381504058, 'recon_cos': 0.03164615710265935, 'gradnorm/compressor.query': 4.384210972018565, 'gradnorm/compressor.layer0': 15.500445451795944, 'gradnorm/compressor.layer1': 10.367514242213511, 'gradnorm/compressor.layer2': 7.839357643281581, 'gradnorm/compressor.layer3': 6.356850012682401, 'gradnorm/compressor.layer4': 5.294329636511956, 'gradnorm/compressor.layer5': 4.555697451813601, 'gradnorm/compressor.layer6': 3.977954793137131, 'gradnorm/compressor.layer7': 3.747631342178668, 'gradnorm/decoder.layer0': 4.169176464772051, 'gradnorm/decoder.layer1': 3.727283719886082, 'gradnorm/decoder.layer2': 3.323927403737173, 'gradnorm/max_param': 11.933223724365234, 'epoch': 0.01}

  0%|          | 10/19090 [00:25<9:28:14,  1.79s/it]
  0%|          | 11/19090 [00:26<8:39:20,  1.63s/it]
  0%|          | 12/19090 [00:29<10:22:12,  1.96s/it]
  0%|          | 13/19090 [00:30<9:17:01,  1.75s/it] 
  0%|          | 14/19090 [00:31<8:31:32,  1.61s/it]
  0%|          | 15/19090 [00:35<11:11:16,  2.11s/it]
  0%|          | 16/19090 [00:36<9:53:11,  1.87s/it] 
  0%|          | 17/19090 [00:37<8:58:27,  1.69s/it]
  0%|          | 18/19090 [00:40<10:27:08,  1.97s/it]
  0%|          | 19/19090 [00:41<9:22:46,  1.77s/it] 
  0%|          | 20/19090 [00:42<8:37:42,  1.63s/it]
                                                    
{'loss': 1.3843, 'grad_norm': 17.940492630004883, 'learning_rate': 4.97642744892614e-07, 'recon_loss': 1.3846725136041642, 'recon_static': 1.3576122015714644, 'recon_motion': 0.02706027473323047, 'recon_norm': 39.9105728149414, 'recon_norm_std': 0.33621060475707054, 'target_norm': 1.6221267253160476, 'recon_cos': 0.033688435610383746, 'gradnorm/compressor.query': 3.134334052019024, 'gradnorm/compressor.layer0': 10.957650940586772, 'gradnorm/compressor.layer1': 7.496895211030749, 'gradnorm/compressor.layer2': 5.817154200251353, 'gradnorm/compressor.layer3': 4.847535850388399, 'gradnorm/compressor.layer4': 4.129915025774403, 'gradnorm/compressor.layer5': 3.640431673213217, 'gradnorm/compressor.layer6': 3.2164731166765153, 'gradnorm/compressor.layer7': 3.0585635050987054, 'gradnorm/decoder.layer0': 3.4338595446932105, 'gradnorm/decoder.layer1': 3.0633333590413763, 'gradnorm/decoder.layer2': 2.6909053430484287, 'gradnorm/max_param': 8.419015884399414, 'epoch': 0.01}

  0%|          | 20/19090 [00:42<8:37:42,  1.63s/it]
  0%|          | 21/19090 [00:45<10:01:46,  1.89s/it]
  0%|          | 22/19090 [00:46<9:05:08,  1.72s/it] 
  0%|          | 23/19090 [00:48<8:25:07,  1.59s/it]
  0%|          | 24/19090 [00:50<10:19:48,  1.95s/it]
  0%|          | 25/19090 [00:52<9:16:58,  1.75s/it] 
  0%|          | 26/19090 [00:53<8:33:39,  1.62s/it]
  0%|          | 27/19090 [00:56<10:36:08,  2.00s/it]
  0%|          | 28/19090 [00:57<9:27:41,  1.79s/it] 
  0%|          | 29/19090 [00:58<8:39:25,  1.64s/it]
  0%|          | 30/19090 [01:01<10:06:51,  1.91s/it]
                                                     
{'loss': 0.9077, 'grad_norm': 10.898565292358398, 'learning_rate': 7.595599790466213e-07, 'recon_loss': 0.9076116055250167, 'recon_static': 0.8785084098577499, 'recon_motion': 0.02910318234935403, 'recon_norm': 32.26834349632263, 'recon_norm_std': 0.3858411990106106, 'target_norm': 1.616932186484337, 'recon_cos': 0.03643485633656383, 'gradnorm/compressor.query': 1.7918337884480788, 'gradnorm/compressor.layer0': 6.432204651855351, 'gradnorm/compressor.layer1': 4.388321350251288, 'gradnorm/compressor.layer2': 3.4788270582792724, 'gradnorm/compressor.layer3': 2.929728840854153, 'gradnorm/compressor.layer4': 2.565834295213384, 'gradnorm/compressor.layer5': 2.3022056535585294, 'gradnorm/compressor.layer6': 2.149593305934965, 'gradnorm/compressor.layer7': 2.0810890570836182, 'gradnorm/decoder.layer0': 2.3547584892562994, 'gradnorm/decoder.layer1': 2.1207311049769366, 'gradnorm/decoder.layer2': 1.8368846641512147, 'gradnorm/max_param': 4.9796142578125, 'epoch': 0.02}

  0%|          | 30/19090 [01:01<10:06:51,  1.91s/it]
  0%|          | 31/19090 [01:02<9:11:29,  1.74s/it] 
  0%|          | 32/19090 [01:04<8:31:37,  1.61s/it]
  0%|          | 33/19090 [01:06<9:50:53,  1.86s/it]
  0%|          | 34/19090 [01:07<8:56:17,  1.69s/it]
  0%|          | 35/19090 [01:09<8:18:18,  1.57s/it]
  0%|          | 36/19090 [01:11<10:03:33,  1.90s/it]
  0%|          | 37/19090 [01:13<9:05:00,  1.72s/it] 
  0%|          | 38/19090 [01:14<8:23:58,  1.59s/it]
  0%|          | 39/19090 [01:17<10:15:32,  1.94s/it]
  0%|          | 40/19090 [01:18<9:13:07,  1.74s/it] 
                                                    
{'loss': 0.5604, 'grad_norm': 6.61223030090332, 'learning_rate': 1.0214772132006288e-06, 'recon_loss': 0.5603357776999474, 'recon_static': 0.5285362310707569, 'recon_motion': 0.03179955608211458, 'recon_norm': 25.34688882827759, 'recon_norm_std': 0.622382552921772, 'target_norm': 1.6215011954307557, 'recon_cos': 0.03534060176461935, 'gradnorm/compressor.query': 1.0907708603348159, 'gradnorm/compressor.layer0': 3.8895658100688992, 'gradnorm/compressor.layer1': 2.7158952830918763, 'gradnorm/compressor.layer2': 2.1242506164427772, 'gradnorm/compressor.layer3': 1.7712452652463444, 'gradnorm/compressor.layer4': 1.5062791212123887, 'gradnorm/compressor.layer5': 1.4106403058953678, 'gradnorm/compressor.layer6': 1.3177237327590579, 'gradnorm/compressor.layer7': 1.21431605417351, 'gradnorm/decoder.layer0': 1.4088886512959384, 'gradnorm/decoder.layer1': 1.283974936519942, 'gradnorm/decoder.layer2': 1.1286917617998757, 'gradnorm/max_param': 2.986100435256958, 'epoch': 0.02}

  0%|          | 40/19090 [01:18<9:13:07,  1.74s/it]
  0%|          | 41/19090 [01:19<8:30:09,  1.61s/it]
  0%|          | 42/19090 [01:22<10:08:51,  1.92s/it]
  0%|          | 43/19090 [01:23<9:08:29,  1.73s/it] 
  0%|          | 44/19090 [01:24<8:27:23,  1.60s/it]
  0%|          | 45/19090 [01:27<10:08:17,  1.92s/it]
  0%|          | 46/19090 [01:28<9:07:41,  1.73s/it] 
  0%|          | 47/19090 [01:30<8:25:16,  1.59s/it]
  0%|          | 48/19090 [01:32<10:20:31,  1.96s/it]
  0%|          | 49/19090 [01:34<9:16:05,  1.75s/it] 
  0%|          | 50/19090 [01:35<8:31:13,  1.61s/it]
                                                    
{'loss': 0.3801, 'grad_norm': 4.0086588859558105, 'learning_rate': 1.2833944473546361e-06, 'recon_loss': 0.38026320561766624, 'recon_static': 0.3472249649465084, 'recon_motion': 0.03303823219612241, 'recon_norm': 20.88285074234009, 'recon_norm_std': 0.8057055205106736, 'target_norm': 1.622039020061493, 'recon_cos': 0.038296852819621564, 'gradnorm/compressor.query': 0.7406821393855578, 'gradnorm/compressor.layer0': 2.4438036081039103, 'gradnorm/compressor.layer1': 1.626325946650091, 'gradnorm/compressor.layer2': 1.2819187465960744, 'gradnorm/compressor.layer3': 1.0537602622549482, 'gradnorm/compressor.layer4': 0.9213903090463138, 'gradnorm/compressor.layer5': 0.8290576361532568, 'gradnorm/compressor.layer6': 0.7427905794663797, 'gradnorm/compressor.layer7': 0.68398090843137, 'gradnorm/decoder.layer0': 0.7774005658266655, 'gradnorm/decoder.layer1': 0.7132016692207815, 'gradnorm/decoder.layer2': 0.6935608691942042, 'gradnorm/max_param': 1.8473336696624756, 'epoch': 0.03}

  0%|          | 50/19090 [01:35<8:31:13,  1.61s/it]
  0%|          | 51/19090 [01:38<10:00:37,  1.89s/it]
  0%|          | 52/19090 [01:39<9:02:22,  1.71s/it] 
  0%|          | 53/19090 [01:40<8:21:56,  1.58s/it]
  0%|          | 54/19090 [01:43<10:06:20,  1.91s/it]
  0%|          | 55/19090 [01:44<9:06:20,  1.72s/it] 
  0%|          | 56/19090 [01:45<8:24:37,  1.59s/it]
  0%|          | 57/19090 [01:48<10:12:41,  1.93s/it]
  0%|          | 58/19090 [01:49<9:10:49,  1.74s/it] 
  0%|          | 59/19090 [01:51<8:27:27,  1.60s/it]
  0%|          | 60/19090 [01:54<10:48:56,  2.05s/it]
                                                     
{'loss': 0.2711, 'grad_norm': 2.4367995262145996, 'learning_rate': 1.5453116815086435e-06, 'recon_loss': 0.27072983868420125, 'recon_static': 0.2381476119160652, 'recon_motion': 0.03258223617449403, 'recon_norm': 17.60934524536133, 'recon_norm_std': 0.7646239906549454, 'target_norm': 1.6202936917543411, 'recon_cos': 0.04157971115782857, 'gradnorm/compressor.query': 0.4790131143196244, 'gradnorm/compressor.layer0': 1.5110316723546893, 'gradnorm/compressor.layer1': 0.978354800593694, 'gradnorm/compressor.layer2': 0.7515712412367468, 'gradnorm/compressor.layer3': 0.6221790548045537, 'gradnorm/compressor.layer4': 0.535939166008053, 'gradnorm/compressor.layer5': 0.48768153173052325, 'gradnorm/compressor.layer6': 0.44275642769561824, 'gradnorm/compressor.layer7': 0.40607177052565546, 'gradnorm/decoder.layer0': 0.45491758469987853, 'gradnorm/decoder.layer1': 0.4370314609607546, 'gradnorm/decoder.layer2': 0.48191151669845494, 'gradnorm/max_param': 1.138709306716919, 'epoch': 0.03}

  0%|          | 60/19090 [01:54<10:48:56,  2.05s/it]
  0%|          | 61/19090 [01:55<9:36:31,  1.82s/it] 
  0%|          | 62/19090 [01:56<8:46:21,  1.66s/it]
  0%|          | 63/19090 [01:59<10:25:59,  1.97s/it]
  0%|          | 64/19090 [02:00<9:21:15,  1.77s/it] 
  0%|          | 65/19090 [02:02<8:37:28,  1.63s/it]
  0%|          | 66/19090 [02:05<10:46:27,  2.04s/it]
  0%|          | 67/19090 [02:06<9:38:13,  1.82s/it] 
  0%|          | 68/19090 [02:07<8:50:29,  1.67s/it]
  0%|          | 69/19090 [02:10<10:13:43,  1.94s/it]
  0%|          | 70/19090 [02:11<9:15:21,  1.75s/it] 
                                                    
{'loss': 0.1941, 'grad_norm': 1.760119915008545, 'learning_rate': 1.8072289156626506e-06, 'recon_loss': 0.19336296655237675, 'recon_static': 0.1640384502708912, 'recon_motion': 0.029324513114988803, 'recon_norm': 14.869684410095214, 'recon_norm_std': 0.6504198826849461, 'target_norm': 1.617485484480858, 'recon_cos': 0.04404091266915202, 'gradnorm/compressor.query': 0.34531828370580037, 'gradnorm/compressor.layer0': 1.0833257100252767, 'gradnorm/compressor.layer1': 0.7264045625709304, 'gradnorm/compressor.layer2': 0.5512762077744916, 'gradnorm/compressor.layer3': 0.45156321542016614, 'gradnorm/compressor.layer4': 0.3938772238205842, 'gradnorm/compressor.layer5': 0.3460656867263784, 'gradnorm/compressor.layer6': 0.3084243236884975, 'gradnorm/compressor.layer7': 0.27853516433894493, 'gradnorm/decoder.layer0': 0.31852417173467307, 'gradnorm/decoder.layer1': 0.302632160774071, 'gradnorm/decoder.layer2': 0.3592387853463454, 'gradnorm/max_param': 0.812593400478363, 'epoch': 0.04}

  0%|          | 70/19090 [02:11<9:15:21,  1.75s/it]
  0%|          | 71/19090 [02:12<8:34:52,  1.62s/it]
  0%|          | 72/19090 [02:15<10:16:36,  1.95s/it]
  0%|          | 73/19090 [02:16<9:14:41,  1.75s/it] 
  0%|          | 74/19090 [02:18<8:31:05,  1.61s/it]
  0%|          | 75/19090 [02:20<10:12:19,  1.93s/it]
  0%|          | 76/19090 [02:22<9:10:32,  1.74s/it] 
  0%|          | 77/19090 [02:23<8:27:15,  1.60s/it]
  0%|          | 78/19090 [02:25<9:46:44,  1.85s/it]
  0%|          | 79/19090 [02:27<8:53:08,  1.68s/it]
  0%|          | 80/19090 [02:28<8:15:30,  1.56s/it]
                                                    
{'loss': 0.1399, 'grad_norm': 1.2163437604904175, 'learning_rate': 2.069146149816658e-06, 'recon_loss': 0.1397917265072465, 'recon_static': 0.11495002266019583, 'recon_motion': 0.0248417011462152, 'recon_norm': 12.647602915763855, 'recon_norm_std': 0.5336506783962249, 'target_norm': 1.619560831785202, 'recon_cos': 0.05645017214119434, 'gradnorm/compressor.query': 0.23894949529733817, 'gradnorm/compressor.layer0': 0.7121123363540017, 'gradnorm/compressor.layer1': 0.487068210627559, 'gradnorm/compressor.layer2': 0.3917043080233473, 'gradnorm/compressor.layer3': 0.32304807522053763, 'gradnorm/compressor.layer4': 0.292865289857652, 'gradnorm/compressor.layer5': 0.2645912953302669, 'gradnorm/compressor.layer6': 0.2368973426548234, 'gradnorm/compressor.layer7': 0.21608815898268488, 'gradnorm/decoder.layer0': 0.23760775455663063, 'gradnorm/decoder.layer1': 0.20422788297491093, 'gradnorm/decoder.layer2': 0.2519999320147438, 'gradnorm/max_param': 0.5314485430717468, 'epoch': 0.04}

  0%|          | 80/19090 [02:28<8:15:30,  1.56s/it]
  0%|          | 81/19090 [02:30<9:41:12,  1.83s/it]
  0%|          | 82/19090 [02:32<8:48:33,  1.67s/it]
  0%|          | 83/19090 [02:33<8:11:45,  1.55s/it]
  0%|          | 84/19090 [02:36<9:54:01,  1.88s/it]
  0%|          | 85/19090 [02:37<8:57:26,  1.70s/it]
  0%|          | 86/19090 [02:38<8:17:57,  1.57s/it]
  0%|          | 87/19090 [02:41<9:47:54,  1.86s/it]
  0%|          | 88/19090 [02:42<9:21:36,  1.77s/it]
  0%|          | 89/19090 [02:44<8:35:12,  1.63s/it]
  0%|          | 90/19090 [02:46<9:55:09,  1.88s/it]
                                                    
{'loss': 0.1028, 'grad_norm': 1.1555908918380737, 'learning_rate': 2.3310633839706653e-06, 'recon_loss': 0.10293984487652778, 'recon_static': 0.08283213526010513, 'recon_motion': 0.020107712224125864, 'recon_norm': 10.856172704696656, 'recon_norm_std': 0.4829840548336506, 'target_norm': 1.6203633427619935, 'recon_cos': 0.0717698922380805, 'gradnorm/compressor.query': 0.19762421110829376, 'gradnorm/compressor.layer0': 0.645068643968505, 'gradnorm/compressor.layer1': 0.48004302105236457, 'gradnorm/compressor.layer2': 0.38983920215102125, 'gradnorm/compressor.layer3': 0.3342359053001736, 'gradnorm/compressor.layer4': 0.29860710247932204, 'gradnorm/compressor.layer5': 0.26613642040759533, 'gradnorm/compressor.layer6': 0.2352274413086233, 'gradnorm/compressor.layer7': 0.21222050216337945, 'gradnorm/decoder.layer0': 0.23056160139403553, 'gradnorm/decoder.layer1': 0.1859014536273661, 'gradnorm/decoder.layer2': 0.19496040078182705, 'gradnorm/max_param': 0.49306511878967285, 'epoch': 0.05}

  0%|          | 90/19090 [02:46<9:55:09,  1.88s/it]
  0%|          | 91/19090 [02:48<9:38:19,  1.83s/it]
  0%|          | 92/19090 [02:49<8:46:57,  1.66s/it]
  0%|          | 93/19090 [02:51<9:39:24,  1.83s/it]
  0%|          | 94/19090 [02:53<9:08:06,  1.73s/it]
  0%|          | 95/19090 [02:54<8:25:26,  1.60s/it]
  1%|          | 96/19090 [02:57<9:55:32,  1.88s/it]
  1%|          | 97/19090 [02:58<8:59:13,  1.70s/it]
  1%|          | 98/19090 [02:59<8:19:49,  1.58s/it]
  1%|          | 99/19090 [03:02<10:18:56,  1.96s/it]
  1%|          | 100/19090 [03:03<9:16:15,  1.76s/it]
                                                     
{'loss': 0.0766, 'grad_norm': 0.8975359797477722, 'learning_rate': 2.592980618124673e-06, 'recon_loss': 0.07665263991802931, 'recon_static': 0.06096348986029625, 'recon_motion': 0.01568915331736207, 'recon_norm': 9.367672204971313, 'recon_norm_std': 0.4579809859395027, 'target_norm': 1.6198723286390304, 'recon_cos': 0.08613492045551538, 'gradnorm/compressor.query': 0.15250972118817938, 'gradnorm/compressor.layer0': 0.48884927897431835, 'gradnorm/compressor.layer1': 0.38131215495638077, 'gradnorm/compressor.layer2': 0.29983898053692115, 'gradnorm/compressor.layer3': 0.2550006608627208, 'gradnorm/compressor.layer4': 0.2297241823916032, 'gradnorm/compressor.layer5': 0.20847256994543176, 'gradnorm/compressor.layer6': 0.190700700065451, 'gradnorm/compressor.layer7': 0.17180782869537822, 'gradnorm/decoder.layer0': 0.1917980815593115, 'gradnorm/decoder.layer1': 0.15271646804423877, 'gradnorm/decoder.layer2': 0.14416228702770562, 'gradnorm/max_param': 0.3759218156337738, 'epoch': 0.05}

  1%|          | 100/19090 [03:03<9:16:15,  1.76s/it]
  1%|          | 101/19090 [03:05<8:32:54,  1.62s/it]
  1%|          | 102/19090 [03:07<9:57:26,  1.89s/it]
  1%|          | 103/19090 [03:08<8:59:32,  1.70s/it]
  1%|          | 104/19090 [03:10<8:19:28,  1.58s/it]
  1%|          | 105/19090 [03:12<9:58:08,  1.89s/it]
  1%|          | 106/19090 [03:14<9:21:53,  1.78s/it]
  1%|          | 107/19090 [03:15<8:34:54,  1.63s/it]
  1%|          | 108/19090 [03:18<10:09:00,  1.92s/it]
  1%|          | 109/19090 [03:19<9:18:23,  1.77s/it] 
  1%|          | 110/19090 [03:20<8:32:56,  1.62s/it]
                                                     
{'loss': 0.0582, 'grad_norm': 0.6669182181358337, 'learning_rate': 2.85489785227868e-06, 'recon_loss': 0.05821698177605868, 'recon_static': 0.04598997626453638, 'recon_motion': 0.012227006792090832, 'recon_norm': 8.159266912937165, 'recon_norm_std': 0.44194013103842733, 'target_norm': 1.6191533118486405, 'recon_cos': 0.09801654741168023, 'gradnorm/compressor.query': 0.11370840599887819, 'gradnorm/compressor.layer0': 0.35831163482987416, 'gradnorm/compressor.layer1': 0.2767232885478558, 'gradnorm/compressor.layer2': 0.22295848511376012, 'gradnorm/compressor.layer3': 0.19275437698515505, 'gradnorm/compressor.layer4': 0.17226953091161548, 'gradnorm/compressor.layer5': 0.158312906628059, 'gradnorm/compressor.layer6': 0.1458281185432579, 'gradnorm/compressor.layer7': 0.13061741622100725, 'gradnorm/decoder.layer0': 0.14603804939141865, 'gradnorm/decoder.layer1': 0.11949067595873585, 'gradnorm/decoder.layer2': 0.10654041434074663, 'gradnorm/max_param': 0.27664583921432495, 'epoch': 0.06}

  1%|          | 110/19090 [03:20<8:32:56,  1.62s/it]
  1%|          | 111/19090 [03:23<10:07:13,  1.92s/it]
  1%|          | 112/19090 [03:25<9:42:20,  1.84s/it] 
  1%|          | 113/19090 [03:26<8:49:20,  1.67s/it]
  1%|          | 114/19090 [03:28<9:53:25,  1.88s/it]
  1%|          | 115/19090 [03:30<9:16:04,  1.76s/it]
  1%|          | 116/19090 [03:31<8:30:59,  1.62s/it]
  1%|          | 117/19090 [03:34<10:02:39,  1.91s/it]
  1%|          | 118/19090 [03:35<9:03:26,  1.72s/it] 
  1%|          | 119/19090 [03:36<8:22:06,  1.59s/it]
  1%|          | 120/19090 [03:39<10:14:00,  1.94s/it]
                                                      
{'loss': 0.046, 'grad_norm': 0.48884522914886475, 'learning_rate': 3.1168150864326876e-06, 'recon_loss': 0.04595917649567127, 'recon_static': 0.036142801959067584, 'recon_motion': 0.009816373325884342, 'recon_norm': 7.242431509494781, 'recon_norm_std': 0.4353673130273819, 'target_norm': 1.6178334683179856, 'recon_cos': 0.10815145261585712, 'gradnorm/compressor.query': 0.08946152943272151, 'gradnorm/compressor.layer0': 0.25814138011823085, 'gradnorm/compressor.layer1': 0.19793736930953668, 'gradnorm/compressor.layer2': 0.15994906398875544, 'gradnorm/compressor.layer3': 0.14052262155791478, 'gradnorm/compressor.layer4': 0.12689576687239842, 'gradnorm/compressor.layer5': 0.11790173945227685, 'gradnorm/compressor.layer6': 0.10832820369729156, 'gradnorm/compressor.layer7': 0.09934599969073973, 'gradnorm/decoder.layer0': 0.11287701765635508, 'gradnorm/decoder.layer1': 0.09461905582678912, 'gradnorm/decoder.layer2': 0.08073308349408408, 'gradnorm/max_param': 0.19677728414535522, 'epoch': 0.06}

  1%|          | 120/19090 [03:39<10:14:00,  1.94s/it]
  1%|          | 121/19090 [03:40<9:13:46,  1.75s/it] 
  1%|          | 122/19090 [03:42<8:31:29,  1.62s/it]
  1%|          | 123/19090 [03:44<10:08:32,  1.93s/it]
  1%|          | 124/19090 [03:46<9:08:43,  1.74s/it] 
  1%|          | 125/19090 [03:47<8:27:37,  1.61s/it]
  1%|          | 126/19090 [03:49<9:45:15,  1.85s/it]
  1%|          | 127/19090 [03:51<9:16:32,  1.76s/it]
  1%|          | 128/19090 [03:52<8:32:04,  1.62s/it]
  1%|          | 129/19090 [03:55<9:48:56,  1.86s/it]
  1%|          | 130/19090 [03:56<8:59:53,  1.71s/it]
                                                     
{'loss': 0.038, 'grad_norm': 0.34373143315315247, 'learning_rate': 3.3787323205866943e-06, 'recon_loss': 0.03799993498250842, 'recon_static': 0.029799864813685416, 'recon_motion': 0.008200070261955262, 'recon_norm': 6.578157818317413, 'recon_norm_std': 0.4308108314871788, 'target_norm': 1.6203296065330506, 'recon_cos': 0.11755663733929396, 'gradnorm/compressor.query': 0.06507690176667114, 'gradnorm/compressor.layer0': 0.1782291463427938, 'gradnorm/compressor.layer1': 0.13937956929735373, 'gradnorm/compressor.layer2': 0.11397766753731849, 'gradnorm/compressor.layer3': 0.09965704263537589, 'gradnorm/compressor.layer4': 0.08912807464011022, 'gradnorm/compressor.layer5': 0.08196039204883675, 'gradnorm/compressor.layer6': 0.07531884499588112, 'gradnorm/compressor.layer7': 0.07062638607395122, 'gradnorm/decoder.layer0': 0.0792784097387913, 'gradnorm/decoder.layer1': 0.0676453389813848, 'gradnorm/decoder.layer2': 0.060152064447196084, 'gradnorm/max_param': 0.13354533910751343, 'epoch': 0.07}

  1%|          | 130/19090 [03:56<8:59:53,  1.71s/it]
  1%|          | 131/19090 [03:57<8:20:51,  1.59s/it]
  1%|          | 132/19090 [04:00<10:11:30,  1.94s/it]
  1%|          | 133/19090 [04:01<9:20:50,  1.78s/it] 
  1%|          | 134/19090 [04:03<8:36:04,  1.63s/it]
  1%|          | 135/19090 [04:05<10:01:08,  1.90s/it]
  1%|          | 136/19090 [04:07<9:19:58,  1.77s/it] 
  1%|          | 137/19090 [04:08<8:34:31,  1.63s/it]
  1%|          | 138/19090 [04:10<9:57:46,  1.89s/it]
  1%|          | 139/19090 [04:12<9:32:15,  1.81s/it]
  1%|          | 140/19090 [04:13<8:44:34,  1.66s/it]
                                                     
{'loss': 0.0326, 'grad_norm': 0.2816027104854584, 'learning_rate': 3.6406495547407023e-06, 'recon_loss': 0.03253911724314094, 'recon_static': 0.025469468208029866, 'recon_motion': 0.007069648860488087, 'recon_norm': 6.082175135612488, 'recon_norm_std': 0.433873225748539, 'target_norm': 1.6196665108203887, 'recon_cos': 0.1272183170542121, 'gradnorm/compressor.query': 0.05673162368741907, 'gradnorm/compressor.layer0': 0.14487527164799102, 'gradnorm/compressor.layer1': 0.11103638268349539, 'gradnorm/compressor.layer2': 0.09027540132968975, 'gradnorm/compressor.layer3': 0.07997662228754121, 'gradnorm/compressor.layer4': 0.07524121728722516, 'gradnorm/compressor.layer5': 0.06923224526420285, 'gradnorm/compressor.layer6': 0.06437260482091398, 'gradnorm/compressor.layer7': 0.059259616740099344, 'gradnorm/decoder.layer0': 0.06701716785851748, 'gradnorm/decoder.layer1': 0.05681986442222977, 'gradnorm/decoder.layer2': 0.0488681700529183, 'gradnorm/max_param': 0.10596249997615814, 'epoch': 0.07}

  1%|          | 140/19090 [04:13<8:44:34,  1.66s/it]
  1%|          | 141/19090 [04:16<9:30:50,  1.81s/it]
  1%|          | 142/19090 [04:18<10:17:01,  1.95s/it]
  1%|          | 143/19090 [04:19<9:14:54,  1.76s/it] 
  1%|          | 144/19090 [04:21<8:49:33,  1.68s/it]
  1%|          | 145/19090 [04:23<9:55:15,  1.89s/it]
  1%|          | 146/19090 [04:24<8:59:26,  1.71s/it]
  1%|          | 147/19090 [04:26<8:22:19,  1.59s/it]
  1%|          | 148/19090 [04:28<9:54:18,  1.88s/it]
  1%|          | 149/19090 [04:29<8:58:41,  1.71s/it]
  1%|          | 150/19090 [04:31<8:33:49,  1.63s/it]
                                                     
{'loss': 0.0285, 'grad_norm': 0.22905349731445312, 'learning_rate': 3.902566788894709e-06, 'recon_loss': 0.02854767329990864, 'recon_static': 0.022299451241269707, 'recon_motion': 0.006248222768772393, 'recon_norm': 5.690397548675537, 'recon_norm_std': 0.4291048772633076, 'target_norm': 1.6209303468465805, 'recon_cos': 0.13481748700141907, 'gradnorm/compressor.query': 0.04886474190072759, 'gradnorm/compressor.layer0': 0.11885177664717736, 'gradnorm/compressor.layer1': 0.08864684087630467, 'gradnorm/compressor.layer2': 0.0733228874668938, 'gradnorm/compressor.layer3': 0.0652475419433324, 'gradnorm/compressor.layer4': 0.06061082585170027, 'gradnorm/compressor.layer5': 0.055671183231053634, 'gradnorm/compressor.layer6': 0.05166973636329296, 'gradnorm/compressor.layer7': 0.047277133797616944, 'gradnorm/decoder.layer0': 0.055454062982216445, 'gradnorm/decoder.layer1': 0.045955027027094426, 'gradnorm/decoder.layer2': 0.03994223178765863, 'gradnorm/max_param': 0.08465172350406647, 'epoch': 0.08}

  1%|          | 150/19090 [04:31<8:33:49,  1.63s/it]
  1%|          | 151/19090 [04:34<10:35:04,  2.01s/it]
  1%|          | 152/19090 [04:35<9:27:43,  1.80s/it] 
  1%|          | 153/19090 [04:37<8:49:09,  1.68s/it]
  1%|          | 154/19090 [04:38<9:14:36,  1.76s/it]
  1%|          | 155/19090 [04:40<8:30:37,  1.62s/it]
  1%|          | 156/19090 [04:41<8:00:00,  1.52s/it]
  1%|          | 157/19090 [04:44<9:43:39,  1.85s/it]
  1%|          | 158/19090 [04:45<8:51:16,  1.68s/it]
  1%|          | 159/19090 [04:46<8:15:03,  1.57s/it]
  1%|          | 160/19090 [04:49<10:14:01,  1.95s/it]
                                                      
{'loss': 0.0254, 'grad_norm': 0.2251899540424347, 'learning_rate': 4.164484023048717e-06, 'recon_loss': 0.025484414817765356, 'recon_static': 0.019862109841778875, 'recon_motion': 0.005622304882854224, 'recon_norm': 5.37079621553421, 'recon_norm_std': 0.4301928721368313, 'target_norm': 1.6245403051376344, 'recon_cos': 0.14283457323908805, 'gradnorm/compressor.query': 0.042875908937478, 'gradnorm/compressor.layer0': 0.11795403597185974, 'gradnorm/compressor.layer1': 0.09077251879024188, 'gradnorm/compressor.layer2': 0.0748532149028274, 'gradnorm/compressor.layer3': 0.06404219020499417, 'gradnorm/compressor.layer4': 0.059007771380246155, 'gradnorm/compressor.layer5': 0.05406247751560736, 'gradnorm/compressor.layer6': 0.04948769697684057, 'gradnorm/compressor.layer7': 0.04578796239864675, 'gradnorm/decoder.layer0': 0.052776792032795436, 'gradnorm/decoder.layer1': 0.044152556606506264, 'gradnorm/decoder.layer2': 0.03646754339376601, 'gradnorm/max_param': 0.08845120668411255, 'epoch': 0.08}

  1%|          | 160/19090 [04:49<10:14:01,  1.95s/it]
  1%|          | 161/19090 [04:50<9:12:31,  1.75s/it] 
  1%|          | 162/19090 [04:52<8:39:51,  1.65s/it]
  1%|          | 163/19090 [04:54<9:59:26,  1.90s/it]
  1%|          | 164/19090 [04:56<9:01:15,  1.72s/it]
  1%|          | 165/19090 [04:57<8:39:42,  1.65s/it]
  1%|          | 166/19090 [04:59<9:25:17,  1.79s/it]
  1%|          | 167/19090 [05:00<8:37:04,  1.64s/it]
  1%|          | 168/19090 [05:03<9:26:26,  1.80s/it]
  1%|          | 169/19090 [05:04<9:30:53,  1.81s/it]
  1%|          | 170/19090 [05:06<8:41:43,  1.65s/it]
                                                     
{'loss': 0.0229, 'grad_norm': 0.20350725948810577, 'learning_rate': 4.426401257202724e-06, 'recon_loss': 0.022983462223783134, 'recon_static': 0.017889212910085918, 'recon_motion': 0.0050942506757564844, 'recon_norm': 5.096869587898254, 'recon_norm_std': 0.4327198274433613, 'target_norm': 1.6201691806316376, 'recon_cos': 0.15086820386350155, 'gradnorm/compressor.query': 0.038849606623243646, 'gradnorm/compressor.layer0': 0.10014233871204274, 'gradnorm/compressor.layer1': 0.08071152016087195, 'gradnorm/compressor.layer2': 0.06762032528014694, 'gradnorm/compressor.layer3': 0.05937668529741332, 'gradnorm/compressor.layer4': 0.05525463512340396, 'gradnorm/compressor.layer5': 0.05154080376987543, 'gradnorm/compressor.layer6': 0.04735965567407783, 'gradnorm/compressor.layer7': 0.04401438238942856, 'gradnorm/decoder.layer0': 0.049356354689260626, 'gradnorm/decoder.layer1': 0.04203444118244269, 'gradnorm/decoder.layer2': 0.034238873056105165, 'gradnorm/max_param': 0.07384934276342392, 'epoch': 0.09}

  1%|          | 170/19090 [05:06<8:41:43,  1.65s/it]
  1%|          | 171/19090 [05:08<9:26:05,  1.80s/it]
  1%|          | 172/19090 [05:10<9:25:27,  1.79s/it]
  1%|          | 173/19090 [05:11<8:37:08,  1.64s/it]
  1%|          | 174/19090 [05:13<9:22:40,  1.78s/it]
  1%|          | 175/19090 [05:15<9:23:39,  1.79s/it]
  1%|          | 176/19090 [05:16<8:36:40,  1.64s/it]
  1%|          | 177/19090 [05:18<9:12:28,  1.75s/it]
  1%|          | 178/19090 [05:20<9:49:52,  1.87s/it]
  1%|          | 179/19090 [05:22<8:54:55,  1.70s/it]
  1%|          | 180/19090 [05:23<9:07:06,  1.74s/it]
                                                     
{'loss': 0.0209, 'grad_norm': 0.18282219767570496, 'learning_rate': 4.688318491356731e-06, 'recon_loss': 0.020867264177650214, 'recon_static': 0.01619417271576822, 'recon_motion': 0.004673091974109411, 'recon_norm': 4.852641868591308, 'recon_norm_std': 0.4308890886604786, 'target_norm': 1.620098939538002, 'recon_cos': 0.15862636789679527, 'gradnorm/compressor.query': 0.036018692367473015, 'gradnorm/compressor.layer0': 0.09191651360080272, 'gradnorm/compressor.layer1': 0.07294357142618184, 'gradnorm/compressor.layer2': 0.059692981537544654, 'gradnorm/compressor.layer3': 0.05293580060404802, 'gradnorm/compressor.layer4': 0.048327661192931866, 'gradnorm/compressor.layer5': 0.04478341252259169, 'gradnorm/compressor.layer6': 0.041519491691383886, 'gradnorm/compressor.layer7': 0.03843958967477384, 'gradnorm/decoder.layer0': 0.043967155485159504, 'gradnorm/decoder.layer1': 0.03862659755112415, 'gradnorm/decoder.layer2': 0.03200058384987616, 'gradnorm/max_param': 0.0674431100487709, 'epoch': 0.09}

  1%|          | 180/19090 [05:23<9:07:06,  1.74s/it]
  1%|          | 181/19090 [05:26<9:49:08,  1.87s/it]
  1%|          | 182/19090 [05:27<8:53:30,  1.69s/it]
  1%|          | 183/19090 [05:29<9:11:25,  1.75s/it]
  1%|          | 184/19090 [05:31<9:45:26,  1.86s/it]
  1%|          | 185/19090 [05:32<8:51:20,  1.69s/it]
  1%|          | 186/19090 [05:34<9:19:35,  1.78s/it]
  1%|          | 187/19090 [05:36<9:25:11,  1.79s/it]
  1%|          | 188/19090 [05:37<8:37:00,  1.64s/it]
  1%|          | 189/19090 [05:40<9:52:55,  1.88s/it]
  1%|          | 190/19090 [05:41<9:10:33,  1.75s/it]
                                                     
{'loss': 0.0191, 'grad_norm': 0.17308731377124786, 'learning_rate': 4.950235725510738e-06, 'recon_loss': 0.019079682184383273, 'recon_static': 0.014750572806224226, 'recon_motion': 0.004329108772799373, 'recon_norm': 4.635569989681244, 'recon_norm_std': 0.43219587951898575, 'target_norm': 1.6216096520423888, 'recon_cos': 0.16644803062081337, 'gradnorm/compressor.query': 0.031858466081623744, 'gradnorm/compressor.layer0': 0.08380028598372145, 'gradnorm/compressor.layer1': 0.06826230036606452, 'gradnorm/compressor.layer2': 0.056134272860336586, 'gradnorm/compressor.layer3': 0.05073685264147401, 'gradnorm/compressor.layer4': 0.04702464199060436, 'gradnorm/compressor.layer5': 0.04447223220006511, 'gradnorm/compressor.layer6': 0.041290497479725095, 'gradnorm/compressor.layer7': 0.03813597711404445, 'gradnorm/decoder.layer0': 0.04356844555377752, 'gradnorm/decoder.layer1': 0.037220828309193675, 'gradnorm/decoder.layer2': 0.03007969465123383, 'gradnorm/max_param': 0.0626436099410057, 'epoch': 0.1}

  1%|          | 190/19090 [05:41<9:10:33,  1.75s/it]
  1%|          | 191/19090 [05:42<8:27:49,  1.61s/it]
  1%|          | 192/19090 [05:45<9:30:20,  1.81s/it]
  1%|          | 193/19090 [05:46<8:59:23,  1.71s/it]
  1%|          | 194/19090 [05:48<8:19:05,  1.58s/it]
  1%|          | 195/19090 [05:50<9:12:19,  1.75s/it]
  1%|          | 196/19090 [05:52<10:00:34,  1.91s/it]
  1%|          | 197/19090 [05:53<9:01:50,  1.72s/it] 
  1%|          | 198/19090 [05:55<8:40:41,  1.65s/it]
  1%|          | 199/19090 [05:57<9:51:09,  1.88s/it]
  1%|          | 200/19090 [05:58<8:55:25,  1.70s/it]
                                                     
{'loss': 0.0176, 'grad_norm': 0.18296851217746735, 'learning_rate': 5.212152959664746e-06, 'recon_loss': 0.017586683854460716, 'recon_static': 0.013573371665552258, 'recon_motion': 0.004013312037568539, 'recon_norm': 4.447495901584626, 'recon_norm_std': 0.4348138518631458, 'target_norm': 1.6180385053157806, 'recon_cos': 0.17396380752325058, 'gradnorm/compressor.query': 0.029683720119045184, 'gradnorm/compressor.layer0': 0.08611011238436547, 'gradnorm/compressor.layer1': 0.07317742027202882, 'gradnorm/compressor.layer2': 0.062139340360679654, 'gradnorm/compressor.layer3': 0.05507594122852398, 'gradnorm/compressor.layer4': 0.050591023108483936, 'gradnorm/compressor.layer5': 0.046453919170711, 'gradnorm/compressor.layer6': 0.04302440636747298, 'gradnorm/compressor.layer7': 0.04015578724264245, 'gradnorm/decoder.layer0': 0.04620246936183263, 'gradnorm/decoder.layer1': 0.040056954433538704, 'gradnorm/decoder.layer2': 0.0318052923258153, 'gradnorm/max_param': 0.06596127152442932, 'epoch': 0.1}

  1%|          | 200/19090 [05:58<8:55:25,  1.70s/it]
  1%|          | 201/19090 [06:00<8:32:42,  1.63s/it]
  1%|          | 202/19090 [06:02<10:01:23,  1.91s/it]
  1%|          | 203/19090 [06:04<9:02:30,  1.72s/it] 
  1%|          | 204/19090 [06:05<8:43:35,  1.66s/it]
  1%|          | 205/19090 [06:08<10:02:25,  1.91s/it]
  1%|          | 206/19090 [06:09<9:03:12,  1.73s/it] 
  1%|          | 207/19090 [06:11<8:40:17,  1.65s/it]
  1%|          | 208/19090 [06:13<10:07:19,  1.93s/it]
  1%|          | 209/19090 [06:14<9:06:16,  1.74s/it] 
  1%|          | 210/19090 [06:16<8:41:26,  1.66s/it]
                                                     
{'loss': 0.0163, 'grad_norm': 0.18608807027339935, 'learning_rate': 5.4740701938187535e-06, 'recon_loss': 0.01622429348062724, 'recon_static': 0.012460750504396855, 'recon_motion': 0.00376354327891022, 'recon_norm': 4.268199062347412, 'recon_norm_std': 0.43181034103035926, 'target_norm': 1.6193503618240357, 'recon_cos': 0.1816437728703022, 'gradnorm/compressor.query': 0.028681107273539504, 'gradnorm/compressor.layer0': 0.08868151960983228, 'gradnorm/compressor.layer1': 0.07549377756971754, 'gradnorm/compressor.layer2': 0.0634317863703001, 'gradnorm/compressor.layer3': 0.057048765850153955, 'gradnorm/compressor.layer4': 0.051729248370079056, 'gradnorm/compressor.layer5': 0.04715926186737157, 'gradnorm/compressor.layer6': 0.04303261362072926, 'gradnorm/compressor.layer7': 0.04028665787881266, 'gradnorm/decoder.layer0': 0.04664085940041739, 'gradnorm/decoder.layer1': 0.03848148770687584, 'gradnorm/decoder.layer2': 0.030550965165800498, 'gradnorm/max_param': 0.0692717507481575, 'epoch': 0.11}

  1%|          | 210/19090 [06:16<8:41:26,  1.66s/it]
  1%|          | 211/19090 [06:18<9:44:53,  1.86s/it]
  1%|          | 212/19090 [06:19<8:50:29,  1.69s/it]
  1%|          | 213/19090 [06:21<8:27:08,  1.61s/it]
  1%|          | 214/19090 [06:24<10:27:02,  1.99s/it]
  1%|          | 215/19090 [06:25<9:20:24,  1.78s/it] 
  1%|          | 216/19090 [06:26<8:33:45,  1.63s/it]
  1%|          | 217/19090 [06:29<10:16:10,  1.96s/it]
  1%|          | 218/19090 [06:30<9:12:33,  1.76s/it] 
  1%|          | 219/19090 [06:32<8:28:32,  1.62s/it]
  1%|          | 220/19090 [06:34<9:33:42,  1.82s/it]
                                                     
{'loss': 0.0151, 'grad_norm': 0.1868661493062973, 'learning_rate': 5.735987427972761e-06, 'recon_loss': 0.015067835920490324, 'recon_static': 0.01152347270399332, 'recon_motion': 0.0035443633038084953, 'recon_norm': 4.108575177192688, 'recon_norm_std': 0.43101300448179247, 'target_norm': 1.6200387209653855, 'recon_cos': 0.1884414106607437, 'gradnorm/compressor.query': 0.025247626230004414, 'gradnorm/compressor.layer0': 0.08698180323400366, 'gradnorm/compressor.layer1': 0.07582365427607438, 'gradnorm/compressor.layer2': 0.06276585985984479, 'gradnorm/compressor.layer3': 0.05617556713687599, 'gradnorm/compressor.layer4': 0.052071062023237136, 'gradnorm/compressor.layer5': 0.04903673432411432, 'gradnorm/compressor.layer6': 0.04567337402532352, 'gradnorm/compressor.layer7': 0.0416613998598566, 'gradnorm/decoder.layer0': 0.04840611481216718, 'gradnorm/decoder.layer1': 0.03999719669765322, 'gradnorm/decoder.layer2': 0.03147298819639237, 'gradnorm/max_param': 0.06923500448465347, 'epoch': 0.12}

  1%|          | 220/19090 [06:34<9:33:42,  1.82s/it]
  1%|          | 221/19090 [06:35<8:42:43,  1.66s/it]
  1%|          | 222/19090 [06:37<8:34:16,  1.64s/it]
  1%|          | 223/19090 [06:40<10:14:34,  1.95s/it]
  1%|          | 224/19090 [06:41<9:11:13,  1.75s/it] 
  1%|          | 225/19090 [06:42<8:26:49,  1.61s/it]
  1%|          | 226/19090 [06:45<10:07:57,  1.93s/it]
  1%|          | 227/19090 [06:46<9:05:50,  1.74s/it] 
  1%|          | 228/19090 [06:47<8:22:39,  1.60s/it]
  1%|          | 229/19090 [06:50<10:51:48,  2.07s/it]
  1%|          | 230/19090 [06:52<9:37:58,  1.84s/it] 
                                                     
{'loss': 0.0141, 'grad_norm': 0.20786404609680176, 'learning_rate': 5.997904662126768e-06, 'recon_loss': 0.014041510527022183, 'recon_static': 0.01070579383522272, 'recon_motion': 0.003335716767469421, 'recon_norm': 3.964230555295944, 'recon_norm_std': 0.4315131865441799, 'target_norm': 1.6184274524450302, 'recon_cos': 0.19648276306688786, 'gradnorm/compressor.query': 0.02505007800205611, 'gradnorm/compressor.layer0': 0.09643204115277842, 'gradnorm/compressor.layer1': 0.08705393915351992, 'gradnorm/compressor.layer2': 0.07091792708353402, 'gradnorm/compressor.layer3': 0.0628718364383741, 'gradnorm/compressor.layer4': 0.05835495010578065, 'gradnorm/compressor.layer5': 0.05389898068023981, 'gradnorm/compressor.layer6': 0.049344260309605266, 'gradnorm/compressor.layer7': 0.045139121734852024, 'gradnorm/decoder.layer0': 0.05230576013757191, 'gradnorm/decoder.layer1': 0.044689523465188015, 'gradnorm/decoder.layer2': 0.034703911027632976, 'gradnorm/max_param': 0.07889953255653381, 'epoch': 0.12}

  1%|          | 230/19090 [06:52<9:37:58,  1.84s/it]
  1%|          | 231/19090 [06:53<8:46:12,  1.67s/it]
  1%|          | 232/19090 [06:56<10:59:16,  2.10s/it]
  1%|          | 233/19090 [06:57<9:42:25,  1.85s/it] 
  1%|          | 234/19090 [06:59<8:48:17,  1.68s/it]
  1%|          | 235/19090 [07:01<10:18:10,  1.97s/it]
  1%|          | 236/19090 [07:03<9:13:15,  1.76s/it] 
  1%|          | 237/19090 [07:04<8:27:46,  1.62s/it]
  1%|          | 238/19090 [07:07<10:44:23,  2.05s/it]
  1%|▏         | 239/19090 [07:08<9:32:29,  1.82s/it] 
  1%|▏         | 240/19090 [07:10<8:42:05,  1.66s/it]
                                                     
{'loss': 0.0132, 'grad_norm': 0.41257205605506897, 'learning_rate': 6.259821896280775e-06, 'recon_loss': 0.013254799693822861, 'recon_static': 0.010082131996750832, 'recon_motion': 0.0031726678076665848, 'recon_norm': 3.8455439031124117, 'recon_norm_std': 0.43215591087937355, 'target_norm': 1.6226247429847718, 'recon_cos': 0.20178887583315372, 'gradnorm/compressor.query': 0.030788090543038465, 'gradnorm/compressor.layer0': 0.1848730601582304, 'gradnorm/compressor.layer1': 0.1734302983206437, 'gradnorm/compressor.layer2': 0.1424525047193343, 'gradnorm/compressor.layer3': 0.1266726329602454, 'gradnorm/compressor.layer4': 0.11813754974412018, 'gradnorm/compressor.layer5': 0.1096602237578776, 'gradnorm/compressor.layer6': 0.10301123398214747, 'gradnorm/compressor.layer7': 0.0939916288362264, 'gradnorm/decoder.layer0': 0.10653443061890047, 'gradnorm/decoder.layer1': 0.08750769855913132, 'gradnorm/decoder.layer2': 0.06510175541545081, 'gradnorm/max_param': 0.1546684056520462, 'epoch': 0.13}

  1%|▏         | 240/19090 [07:10<8:42:05,  1.66s/it]
  1%|▏         | 241/19090 [07:12<10:38:41,  2.03s/it]
  1%|▏         | 242/19090 [07:14<9:27:42,  1.81s/it] 
  1%|▏         | 243/19090 [07:15<8:37:45,  1.65s/it]
  1%|▏         | 244/19090 [07:18<10:16:06,  1.96s/it]
  1%|▏         | 245/19090 [07:19<9:11:33,  1.76s/it] 
  1%|▏         | 246/19090 [07:20<8:26:48,  1.61s/it]
  1%|▏         | 247/19090 [07:23<10:12:29,  1.95s/it]
  1%|▏         | 248/19090 [07:24<9:09:09,  1.75s/it] 
  1%|▏         | 249/19090 [07:26<8:24:58,  1.61s/it]
  1%|▏         | 250/19090 [07:28<10:09:55,  1.94s/it]
                                                      
{'loss': 0.013, 'grad_norm': 0.7761999368667603, 'learning_rate': 6.521739130434783e-06, 'recon_loss': 0.012983942590653896, 'recon_static': 0.009974054200574756, 'recon_motion': 0.0030098888149950654, 'recon_norm': 3.808888280391693, 'recon_norm_std': 0.4185283422470093, 'target_norm': 1.6189368546009064, 'recon_cos': 0.20446280017495155, 'gradnorm/compressor.query': 0.05175455550976306, 'gradnorm/compressor.layer0': 0.34530930798121123, 'gradnorm/compressor.layer1': 0.3286738573671498, 'gradnorm/compressor.layer2': 0.2737636120124932, 'gradnorm/compressor.layer3': 0.24192526679422927, 'gradnorm/compressor.layer4': 0.22650539566449004, 'gradnorm/compressor.layer5': 0.2081534542931996, 'gradnorm/compressor.layer6': 0.18773805883748457, 'gradnorm/compressor.layer7': 0.17186763595931084, 'gradnorm/decoder.layer0': 0.19402693138409743, 'gradnorm/decoder.layer1': 0.16426782021896733, 'gradnorm/decoder.layer2': 0.12256209303887054, 'gradnorm/max_param': 0.29316866397857666, 'epoch': 0.13}

  1%|▏         | 250/19090 [07:28<10:09:55,  1.94s/it]
  1%|▏         | 251/19090 [07:30<9:07:22,  1.74s/it] 
  1%|▏         | 252/19090 [07:31<8:24:57,  1.61s/it]
  1%|▏         | 253/19090 [07:33<9:58:47,  1.91s/it]
  1%|▏         | 254/19090 [07:35<9:01:16,  1.72s/it]
  1%|▏         | 255/19090 [07:36<8:19:14,  1.59s/it]
  1%|▏         | 256/19090 [07:39<10:01:00,  1.91s/it]
  1%|▏         | 257/19090 [07:40<9:01:19,  1.72s/it] 
  1%|▏         | 258/19090 [07:41<8:18:39,  1.59s/it]
  1%|▏         | 259/19090 [07:44<10:09:12,  1.94s/it]
  1%|▏         | 260/19090 [07:45<9:06:36,  1.74s/it] 
                                                     
{'loss': 0.0121, 'grad_norm': 0.4533209800720215, 'learning_rate': 6.783656364588791e-06, 'recon_loss': 0.012094021844677626, 'recon_static': 0.00920216771773994, 'recon_motion': 0.0028918541793245822, 'recon_norm': 3.670008087158203, 'recon_norm_std': 0.41996529921889303, 'target_norm': 1.622133269906044, 'recon_cos': 0.212205808237195, 'gradnorm/compressor.query': 0.03293506142636803, 'gradnorm/compressor.layer0': 0.20192558794416568, 'gradnorm/compressor.layer1': 0.19429823734985363, 'gradnorm/compressor.layer2': 0.15891912700725228, 'gradnorm/compressor.layer3': 0.14091902697424244, 'gradnorm/compressor.layer4': 0.1321717682623914, 'gradnorm/compressor.layer5': 0.1204154219949418, 'gradnorm/compressor.layer6': 0.10946100549013062, 'gradnorm/compressor.layer7': 0.09993327063896795, 'gradnorm/decoder.layer0': 0.11267880032852721, 'gradnorm/decoder.layer1': 0.09442020802665288, 'gradnorm/decoder.layer2': 0.07229578187980541, 'gradnorm/max_param': 0.17049112915992737, 'epoch': 0.14}

  1%|▏         | 260/19090 [07:45<9:06:36,  1.74s/it]
  1%|▏         | 261/19090 [07:47<8:23:02,  1.60s/it]
  1%|▏         | 262/19090 [07:49<9:44:32,  1.86s/it]
  1%|▏         | 263/19090 [07:50<8:49:13,  1.69s/it]
  1%|▏         | 264/19090 [07:52<8:10:37,  1.56s/it]
  1%|▏         | 265/19090 [07:54<10:16:21,  1.96s/it]
  1%|▏         | 266/19090 [07:56<9:11:41,  1.76s/it] 
  1%|▏         | 267/19090 [07:57<8:26:43,  1.62s/it]
  1%|▏         | 268/19090 [08:00<9:56:36,  1.90s/it]
  1%|▏         | 269/19090 [08:01<8:57:43,  1.71s/it]
  1%|▏         | 270/19090 [08:02<8:16:49,  1.58s/it]
                                                     
{'loss': 0.0114, 'grad_norm': 0.8058778643608093, 'learning_rate': 7.045573598742798e-06, 'recon_loss': 0.011396259628236293, 'recon_static': 0.008635213505476714, 'recon_motion': 0.0027610457735136153, 'recon_norm': 3.561768025159836, 'recon_norm_std': 0.4183576852083206, 'target_norm': 1.6191132962703705, 'recon_cos': 0.21940031424164771, 'gradnorm/compressor.query': 0.0490494361837928, 'gradnorm/compressor.layer0': 0.35983584751027065, 'gradnorm/compressor.layer1': 0.34573178792114767, 'gradnorm/compressor.layer2': 0.2800216075104239, 'gradnorm/compressor.layer3': 0.24521989684251777, 'gradnorm/compressor.layer4': 0.23021522413846815, 'gradnorm/compressor.layer5': 0.21533959131096128, 'gradnorm/compressor.layer6': 0.1983323667272954, 'gradnorm/compressor.layer7': 0.18135233530510858, 'gradnorm/decoder.layer0': 0.2067360843186736, 'gradnorm/decoder.layer1': 0.17000685134573224, 'gradnorm/decoder.layer2': 0.12743375470222282, 'gradnorm/max_param': 0.3037785291671753, 'epoch': 0.14}

  1%|▏         | 270/19090 [08:02<8:16:49,  1.58s/it]
  1%|▏         | 271/19090 [08:05<9:58:10,  1.91s/it]
  1%|▏         | 272/19090 [08:06<8:59:24,  1.72s/it]
  1%|▏         | 273/19090 [08:07<8:20:29,  1.60s/it]
  1%|▏         | 274/19090 [08:10<10:03:45,  1.93s/it]
  1%|▏         | 275/19090 [08:11<9:04:02,  1.73s/it] 
  1%|▏         | 276/19090 [08:13<8:20:34,  1.60s/it]
  1%|▏         | 277/19090 [08:16<10:23:17,  1.99s/it]
  1%|▏         | 278/19090 [08:17<9:16:17,  1.77s/it] 
  1%|▏         | 279/19090 [08:18<8:29:59,  1.63s/it]
  1%|▏         | 280/19090 [08:21<10:17:54,  1.97s/it]
                                                      
{'loss': 0.0108, 'grad_norm': 0.6238892674446106, 'learning_rate': 7.307490832896804e-06, 'recon_loss': 0.010773346573114396, 'recon_static': 0.008103782567195595, 'recon_motion': 0.002669564407551661, 'recon_norm': 3.459783601760864, 'recon_norm_std': 0.41532966420054435, 'target_norm': 1.6233017086982726, 'recon_cos': 0.22689442858099937, 'gradnorm/compressor.query': 0.039252743286969545, 'gradnorm/compressor.layer0': 0.27495219454526865, 'gradnorm/compressor.layer1': 0.2637979841651277, 'gradnorm/compressor.layer2': 0.22085987702276538, 'gradnorm/compressor.layer3': 0.19650595339827348, 'gradnorm/compressor.layer4': 0.17816591796711742, 'gradnorm/compressor.layer5': 0.16446431658758426, 'gradnorm/compressor.layer6': 0.15175302613318908, 'gradnorm/compressor.layer7': 0.14086426143030706, 'gradnorm/decoder.layer0': 0.16051796446329789, 'gradnorm/decoder.layer1': 0.13236575161071165, 'gradnorm/decoder.layer2': 0.10082267475427587, 'gradnorm/max_param': 0.23429173231124878, 'epoch': 0.15}

  1%|▏         | 280/19090 [08:21<10:17:54,  1.97s/it]
  1%|▏         | 281/19090 [08:22<9:12:51,  1.76s/it] 
  1%|▏         | 282/19090 [08:23<8:27:25,  1.62s/it]
  1%|▏         | 283/19090 [08:26<9:44:47,  1.87s/it]
  1%|▏         | 284/19090 [08:27<8:50:04,  1.69s/it]
  1%|▏         | 285/19090 [08:29<8:12:18,  1.57s/it]
  1%|▏         | 286/19090 [08:31<9:56:41,  1.90s/it]
  2%|▏         | 287/19090 [08:32<8:59:00,  1.72s/it]
  2%|▏         | 288/19090 [08:34<8:19:58,  1.60s/it]
  2%|▏         | 289/19090 [08:37<10:10:58,  1.95s/it]
  2%|▏         | 290/19090 [08:38<9:08:44,  1.75s/it] 
                                                     
{'loss': 0.0105, 'grad_norm': 0.5598253607749939, 'learning_rate': 7.569408067050812e-06, 'recon_loss': 0.010513508063741028, 'recon_static': 0.007944856595713645, 'recon_motion': 0.0025686518114525826, 'recon_norm': 3.4184248626232145, 'recon_norm_std': 0.4073697663843632, 'target_norm': 1.6186415523290634, 'recon_cos': 0.22964383512735367, 'gradnorm/compressor.query': 0.03254869960046499, 'gradnorm/compressor.layer0': 0.2444886892653249, 'gradnorm/compressor.layer1': 0.24165904163931973, 'gradnorm/compressor.layer2': 0.19599317197802946, 'gradnorm/compressor.layer3': 0.1740005687924932, 'gradnorm/compressor.layer4': 0.15991263935496325, 'gradnorm/compressor.layer5': 0.1469472438602788, 'gradnorm/compressor.layer6': 0.13646803756477238, 'gradnorm/compressor.layer7': 0.12514841686219794, 'gradnorm/decoder.layer0': 0.1437438574805439, 'gradnorm/decoder.layer1': 0.12141415292898382, 'gradnorm/decoder.layer2': 0.09288034302775823, 'gradnorm/max_param': 0.20994046330451965, 'epoch': 0.15}

  2%|▏         | 290/19090 [08:38<9:08:44,  1.75s/it]
  2%|▏         | 291/19090 [08:39<8:24:26,  1.61s/it]
  2%|▏         | 292/19090 [08:42<10:04:20,  1.93s/it]
  2%|▏         | 293/19090 [08:43<9:03:07,  1.73s/it] 
  2%|▏         | 294/19090 [08:44<8:32:52,  1.64s/it]
  2%|▏         | 295/19090 [08:47<10:22:33,  1.99s/it]
  2%|▏         | 296/19090 [08:49<9:15:39,  1.77s/it] 
  2%|▏         | 297/19090 [08:50<8:28:55,  1.62s/it]
  2%|▏         | 298/19090 [08:53<10:20:20,  1.98s/it]
  2%|▏         | 299/19090 [08:54<9:14:29,  1.77s/it] 
  2%|▏         | 300/19090 [08:55<8:28:30,  1.62s/it]
                                                     
{'loss': 0.0096, 'grad_norm': 0.3730476200580597, 'learning_rate': 7.83132530120482e-06, 'recon_loss': 0.009607310895808041, 'recon_static': 0.007122510450426489, 'recon_motion': 0.0024848004686646163, 'recon_norm': 3.2632432162761686, 'recon_norm_std': 0.40918243303894997, 'target_norm': 1.618670493364334, 'recon_cos': 0.24132027104496956, 'gradnorm/compressor.query': 0.02391600043980532, 'gradnorm/compressor.layer0': 0.16437371005977217, 'gradnorm/compressor.layer1': 0.1588274387376692, 'gradnorm/compressor.layer2': 0.12921578639136458, 'gradnorm/compressor.layer3': 0.11591036979751526, 'gradnorm/compressor.layer4': 0.10567046673135423, 'gradnorm/compressor.layer5': 0.09640366232380755, 'gradnorm/compressor.layer6': 0.09150348610308259, 'gradnorm/compressor.layer7': 0.08413829117516183, 'gradnorm/decoder.layer0': 0.09784887865996332, 'gradnorm/decoder.layer1': 0.08218901087093793, 'gradnorm/decoder.layer2': 0.06304169024333925, 'gradnorm/max_param': 0.13938389718532562, 'epoch': 0.16}

  2%|▏         | 300/19090 [08:55<8:28:30,  1.62s/it]
  2%|▏         | 301/19090 [08:58<9:35:42,  1.84s/it]
  2%|▏         | 302/19090 [08:59<8:42:48,  1.67s/it]
  2%|▏         | 303/19090 [09:00<8:05:51,  1.55s/it]
  2%|▏         | 304/19090 [09:03<9:58:34,  1.91s/it]
  2%|▏         | 305/19090 [09:04<8:58:44,  1.72s/it]
  2%|▏         | 306/19090 [09:05<8:17:04,  1.59s/it]
  2%|▏         | 307/19090 [09:08<10:07:53,  1.94s/it]
  2%|▏         | 308/19090 [09:09<9:05:19,  1.74s/it] 
  2%|▏         | 309/19090 [09:11<8:21:35,  1.60s/it]
  2%|▏         | 310/19090 [09:13<9:40:56,  1.86s/it]
                                                     
{'loss': 0.0091, 'grad_norm': 0.4629223942756653, 'learning_rate': 8.093242535358828e-06, 'recon_loss': 0.009121210873126983, 'recon_static': 0.0067143382271751765, 'recon_motion': 0.0024068725237157197, 'recon_norm': 3.1761346399784087, 'recon_norm_std': 0.40755102038383484, 'target_norm': 1.6201568961143493, 'recon_cos': 0.24797684215009214, 'gradnorm/compressor.query': 0.027126256196757115, 'gradnorm/compressor.layer0': 0.20353532574973002, 'gradnorm/compressor.layer1': 0.19664003300113553, 'gradnorm/compressor.layer2': 0.16232426315305173, 'gradnorm/compressor.layer3': 0.14340155731897458, 'gradnorm/compressor.layer4': 0.13271405273830533, 'gradnorm/compressor.layer5': 0.12188224037377964, 'gradnorm/compressor.layer6': 0.1131819585858143, 'gradnorm/compressor.layer7': 0.10319636252726266, 'gradnorm/decoder.layer0': 0.11996334582799054, 'gradnorm/decoder.layer1': 0.10092749519485901, 'gradnorm/decoder.layer2': 0.07777340869816812, 'gradnorm/max_param': 0.17541924118995667, 'epoch': 0.16}

  2%|▏         | 310/19090 [09:13<9:40:56,  1.86s/it]
  2%|▏         | 311/19090 [09:14<8:46:47,  1.68s/it]
  2%|▏         | 312/19090 [09:16<8:08:40,  1.56s/it]
  2%|▏         | 313/19090 [09:18<9:58:23,  1.91s/it]
  2%|▏         | 314/19090 [09:20<8:58:40,  1.72s/it]
  2%|▏         | 315/19090 [09:21<8:16:54,  1.59s/it]
  2%|▏         | 316/19090 [09:24<10:05:17,  1.93s/it]
  2%|▏         | 317/19090 [09:25<9:03:42,  1.74s/it] 
  2%|▏         | 318/19090 [09:26<8:21:23,  1.60s/it]
  2%|▏         | 319/19090 [09:29<10:00:46,  1.92s/it]
  2%|▏         | 320/19090 [09:30<9:01:19,  1.73s/it] 
                                                     
{'loss': 0.0094, 'grad_norm': 0.6534315347671509, 'learning_rate': 8.355159769512835e-06, 'recon_loss': 0.009389563463628292, 'recon_static': 0.0070537174819037315, 'recon_motion': 0.0023358458769507707, 'recon_norm': 3.226546919345856, 'recon_norm_std': 0.39106479063630106, 'target_norm': 1.6207309693098069, 'recon_cos': 0.24422664120793341, 'gradnorm/compressor.query': 0.03423516505488443, 'gradnorm/compressor.layer0': 0.2792063613090065, 'gradnorm/compressor.layer1': 0.27960452228778676, 'gradnorm/compressor.layer2': 0.23220028451863928, 'gradnorm/compressor.layer3': 0.2073500535836115, 'gradnorm/compressor.layer4': 0.18711356235987217, 'gradnorm/compressor.layer5': 0.17041972005070471, 'gradnorm/compressor.layer6': 0.15918172940307798, 'gradnorm/compressor.layer7': 0.14768674752483224, 'gradnorm/decoder.layer0': 0.16920923554996314, 'gradnorm/decoder.layer1': 0.1434691362357087, 'gradnorm/decoder.layer2': 0.11106372271188342, 'gradnorm/max_param': 0.24062831699848175, 'epoch': 0.17}

  2%|▏         | 320/19090 [09:30<9:01:19,  1.73s/it]
  2%|▏         | 321/19090 [09:32<8:19:06,  1.60s/it]
  2%|▏         | 322/19090 [09:35<10:41:32,  2.05s/it]
  2%|▏         | 323/19090 [09:36<9:32:33,  1.83s/it] 
  2%|▏         | 324/19090 [09:37<8:42:15,  1.67s/it]
  2%|▏         | 325/19090 [09:40<10:36:14,  2.03s/it]
  2%|▏         | 326/19090 [09:41<9:27:51,  1.82s/it] 
  2%|▏         | 327/19090 [09:43<8:38:06,  1.66s/it]
  2%|▏         | 328/19090 [09:45<9:43:37,  1.87s/it]
  2%|▏         | 329/19090 [09:46<8:49:53,  1.69s/it]
  2%|▏         | 330/19090 [09:48<8:12:30,  1.58s/it]
                                                     
{'loss': 0.0089, 'grad_norm': 0.8285995125770569, 'learning_rate': 8.61707700366684e-06, 'recon_loss': 0.008925492409616709, 'recon_static': 0.006653199414722621, 'recon_motion': 0.00227229303563945, 'recon_norm': 3.1439432084560393, 'recon_norm_std': 0.3878687210381031, 'target_norm': 1.618455982208252, 'recon_cos': 0.250659603998065, 'gradnorm/compressor.query': 0.04321358028227918, 'gradnorm/compressor.layer0': 0.3572581807488566, 'gradnorm/compressor.layer1': 0.3463164885063551, 'gradnorm/compressor.layer2': 0.289944874276029, 'gradnorm/compressor.layer3': 0.25787651827740143, 'gradnorm/compressor.layer4': 0.2378503076248032, 'gradnorm/compressor.layer5': 0.2208080042552251, 'gradnorm/compressor.layer6': 0.20633812704979537, 'gradnorm/compressor.layer7': 0.18864252758622818, 'gradnorm/decoder.layer0': 0.2196291983311171, 'gradnorm/decoder.layer1': 0.18622519331102788, 'gradnorm/decoder.layer2': 0.14200594807813458, 'gradnorm/max_param': 0.3070749044418335, 'epoch': 0.17}

  2%|▏         | 330/19090 [09:48<8:12:30,  1.58s/it]
  2%|▏         | 331/19090 [09:50<9:37:14,  1.85s/it]
  2%|▏         | 332/19090 [09:51<8:46:53,  1.69s/it]
  2%|▏         | 333/19090 [09:53<8:11:10,  1.57s/it]
  2%|▏         | 334/19090 [09:56<10:32:22,  2.02s/it]
  2%|▏         | 335/19090 [09:57<9:26:15,  1.81s/it] 
  2%|▏         | 336/19090 [09:59<8:40:57,  1.67s/it]
  2%|▏         | 337/19090 [10:01<10:32:26,  2.02s/it]
  2%|▏         | 338/19090 [10:03<9:23:07,  1.80s/it] 
  2%|▏         | 339/19090 [10:04<8:36:08,  1.65s/it]
  2%|▏         | 340/19090 [10:07<10:42:56,  2.06s/it]
                                                      
{'loss': 0.0084, 'grad_norm': 0.49333614110946655, 'learning_rate': 8.878994237820849e-06, 'recon_loss': 0.008407536381855606, 'recon_static': 0.006184420047793538, 'recon_motion': 0.002223116555251181, 'recon_norm': 3.047042053937912, 'recon_norm_std': 0.388303891569376, 'target_norm': 1.6210826605558395, 'recon_cos': 0.2597226031124592, 'gradnorm/compressor.query': 0.024265857516566885, 'gradnorm/compressor.layer0': 0.20815842955690522, 'gradnorm/compressor.layer1': 0.20631593712281143, 'gradnorm/compressor.layer2': 0.17028977041197557, 'gradnorm/compressor.layer3': 0.15393713771871914, 'gradnorm/compressor.layer4': 0.14105567944457045, 'gradnorm/compressor.layer5': 0.1310364891072022, 'gradnorm/compressor.layer6': 0.12488052126823283, 'gradnorm/compressor.layer7': 0.11590519925543982, 'gradnorm/decoder.layer0': 0.13284315428670926, 'gradnorm/decoder.layer1': 0.11324088840982968, 'gradnorm/decoder.layer2': 0.0873673598106914, 'gradnorm/max_param': 0.1795036941766739, 'epoch': 0.18}

  2%|▏         | 340/19090 [10:07<10:42:56,  2.06s/it]
  2%|▏         | 341/19090 [10:08<9:32:39,  1.83s/it] 
  2%|▏         | 342/19090 [10:10<8:42:35,  1.67s/it]
  2%|▏         | 343/19090 [10:12<10:38:17,  2.04s/it]
  2%|▏         | 344/19090 [10:14<9:28:05,  1.82s/it] 
  2%|▏         | 345/19090 [10:15<8:39:14,  1.66s/it]
  2%|▏         | 346/19090 [10:18<10:10:15,  1.95s/it]
  2%|▏         | 347/19090 [10:19<9:08:27,  1.76s/it] 
  2%|▏         | 348/19090 [10:20<8:25:44,  1.62s/it]
  2%|▏         | 349/19090 [10:23<10:33:16,  2.03s/it]
  2%|▏         | 350/19090 [10:25<9:25:39,  1.81s/it] 
                                                     
{'loss': 0.0078, 'grad_norm': 0.582398533821106, 'learning_rate': 9.140911471974857e-06, 'recon_loss': 0.007847530581057072, 'recon_static': 0.005688727681990713, 'recon_motion': 0.0021588030620478095, 'recon_norm': 2.9414144456386566, 'recon_norm_std': 0.39021539464592936, 'target_norm': 1.6208380460739136, 'recon_cos': 0.26989974826574326, 'gradnorm/compressor.query': 0.028787270631564325, 'gradnorm/compressor.layer0': 0.25018040287282567, 'gradnorm/compressor.layer1': 0.2435018728422131, 'gradnorm/compressor.layer2': 0.2000699185554427, 'gradnorm/compressor.layer3': 0.1801787068612779, 'gradnorm/compressor.layer4': 0.16673562490906169, 'gradnorm/compressor.layer5': 0.15413975864752769, 'gradnorm/compressor.layer6': 0.1470606720944933, 'gradnorm/compressor.layer7': 0.13628967256592617, 'gradnorm/decoder.layer0': 0.15500736832409676, 'gradnorm/decoder.layer1': 0.13238856028767146, 'gradnorm/decoder.layer2': 0.10328602519372601, 'gradnorm/max_param': 0.21647211909294128, 'epoch': 0.18}

  2%|▏         | 350/19090 [10:25<9:25:39,  1.81s/it]
  2%|▏         | 351/19090 [10:26<8:36:18,  1.65s/it]
  2%|▏         | 352/19090 [10:29<10:11:20,  1.96s/it]
  2%|▏         | 353/19090 [10:30<9:08:38,  1.76s/it] 
  2%|▏         | 354/19090 [10:31<8:24:11,  1.61s/it]
  2%|▏         | 355/19090 [10:34<10:19:52,  1.99s/it]
  2%|▏         | 356/19090 [10:35<9:20:52,  1.80s/it] 
  2%|▏         | 357/19090 [10:37<8:33:02,  1.64s/it]
  2%|▏         | 358/19090 [10:39<9:47:23,  1.88s/it]
  2%|▏         | 359/19090 [10:40<8:53:00,  1.71s/it]
  2%|▏         | 360/19090 [10:42<8:13:23,  1.58s/it]
                                                     
{'loss': 0.0083, 'grad_norm': 1.0180628299713135, 'learning_rate': 9.402828706128863e-06, 'recon_loss': 0.008248287974856794, 'recon_static': 0.006133933854289353, 'recon_motion': 0.0021143540448974817, 'recon_norm': 3.021497493982315, 'recon_norm_std': 0.3697299838066101, 'target_norm': 1.620090940594673, 'recon_cos': 0.26219919249415397, 'gradnorm/compressor.query': 0.04712548484982864, 'gradnorm/compressor.layer0': 0.4301696499789185, 'gradnorm/compressor.layer1': 0.4322469948663601, 'gradnorm/compressor.layer2': 0.35515324770476175, 'gradnorm/compressor.layer3': 0.31743523971660204, 'gradnorm/compressor.layer4': 0.2938219109232863, 'gradnorm/compressor.layer5': 0.2742366964222284, 'gradnorm/compressor.layer6': 0.2527960052107453, 'gradnorm/compressor.layer7': 0.2345753052217469, 'gradnorm/decoder.layer0': 0.2682320616235646, 'gradnorm/decoder.layer1': 0.2265509076276879, 'gradnorm/decoder.layer2': 0.17806318281209488, 'gradnorm/max_param': 0.36997100710868835, 'epoch': 0.19}

  2%|▏         | 360/19090 [10:42<8:13:23,  1.58s/it]
  2%|▏         | 361/19090 [10:44<9:22:54,  1.80s/it]
  2%|▏         | 362/19090 [10:45<8:33:50,  1.65s/it]
  2%|▏         | 363/19090 [10:47<7:59:53,  1.54s/it]
  2%|▏         | 364/19090 [10:49<9:39:50,  1.86s/it]
  2%|▏         | 365/19090 [10:50<8:46:13,  1.69s/it]
  2%|▏         | 366/19090 [10:52<8:08:11,  1.56s/it]
  2%|▏         | 367/19090 [10:54<9:40:42,  1.86s/it]
  2%|▏         | 368/19090 [10:56<8:46:39,  1.69s/it]
  2%|▏         | 369/19090 [10:57<8:08:56,  1.57s/it]
  2%|▏         | 370/19090 [11:00<10:03:44,  1.94s/it]
                                                      
{'loss': 0.0076, 'grad_norm': 0.5055372714996338, 'learning_rate': 9.664745940282871e-06, 'recon_loss': 0.007656662072986364, 'recon_static': 0.005592665786389261, 'recon_motion': 0.0020639965194277467, 'recon_norm': 2.906933057308197, 'recon_norm_std': 0.37103068009018897, 'target_norm': 1.619677147269249, 'recon_cos': 0.27394427433609964, 'gradnorm/compressor.query': 0.02492788118003723, 'gradnorm/compressor.layer0': 0.214007388280957, 'gradnorm/compressor.layer1': 0.21403592564269425, 'gradnorm/compressor.layer2': 0.17636899176231277, 'gradnorm/compressor.layer3': 0.1568785007530742, 'gradnorm/compressor.layer4': 0.14503905938700973, 'gradnorm/compressor.layer5': 0.1349471293561027, 'gradnorm/compressor.layer6': 0.12523047984319008, 'gradnorm/compressor.layer7': 0.11707034215077129, 'gradnorm/decoder.layer0': 0.13283380870417935, 'gradnorm/decoder.layer1': 0.11390497470529029, 'gradnorm/decoder.layer2': 0.09145886504504606, 'gradnorm/max_param': 0.1853741854429245, 'epoch': 0.19}

  2%|▏         | 370/19090 [11:00<10:03:44,  1.94s/it]
  2%|▏         | 371/19090 [11:01<9:03:13,  1.74s/it] 
  2%|▏         | 372/19090 [11:02<8:20:27,  1.60s/it]
  2%|▏         | 373/19090 [11:05<9:59:22,  1.92s/it]
  2%|▏         | 374/19090 [11:06<8:59:35,  1.73s/it]
  2%|▏         | 375/19090 [11:07<8:17:47,  1.60s/it]
  2%|▏         | 376/19090 [11:10<10:17:16,  1.98s/it]
  2%|▏         | 377/19090 [11:12<9:11:52,  1.77s/it] 
  2%|▏         | 378/19090 [11:13<8:26:29,  1.62s/it]
  2%|▏         | 379/19090 [11:16<10:23:40,  2.00s/it]
  2%|▏         | 380/19090 [11:17<9:16:29,  1.78s/it] 
                                                     
{'loss': 0.0074, 'grad_norm': 0.9194788932800293, 'learning_rate': 9.926663174436879e-06, 'recon_loss': 0.007355747336987406, 'recon_static': 0.005336683592759073, 'recon_motion': 0.0020190640294458716, 'recon_norm': 2.84810129404068, 'recon_norm_std': 0.3684804320335388, 'target_norm': 1.6196309059858323, 'recon_cos': 0.2794608399271965, 'gradnorm/compressor.query': 0.04285537463047359, 'gradnorm/compressor.layer0': 0.3944868497131378, 'gradnorm/compressor.layer1': 0.3880522115765536, 'gradnorm/compressor.layer2': 0.3154983918093095, 'gradnorm/compressor.layer3': 0.2881520340953112, 'gradnorm/compressor.layer4': 0.2645767368160158, 'gradnorm/compressor.layer5': 0.244378782460143, 'gradnorm/compressor.layer6': 0.22915596377059516, 'gradnorm/compressor.layer7': 0.2106344377118185, 'gradnorm/decoder.layer0': 0.24122295272492897, 'gradnorm/decoder.layer1': 0.2073592879992439, 'gradnorm/decoder.layer2': 0.16444482986297454, 'gradnorm/max_param': 0.33967262506484985, 'epoch': 0.2}

  2%|▏         | 380/19090 [11:17<9:16:29,  1.78s/it]
  2%|▏         | 381/19090 [11:18<8:29:40,  1.63s/it]
  2%|▏         | 382/19090 [11:21<10:38:42,  2.05s/it]
  2%|▏         | 383/19090 [11:23<9:27:09,  1.82s/it] 
  2%|▏         | 384/19090 [11:24<8:36:48,  1.66s/it]
  2%|▏         | 385/19090 [11:26<9:48:08,  1.89s/it]
  2%|▏         | 386/19090 [11:28<8:51:27,  1.70s/it]
  2%|▏         | 387/19090 [11:29<8:12:00,  1.58s/it]
  2%|▏         | 388/19090 [11:32<10:11:32,  1.96s/it]
  2%|▏         | 389/19090 [11:33<9:08:42,  1.76s/it] 
  2%|▏         | 390/19090 [11:34<8:25:09,  1.62s/it]
                                                     
{'loss': 0.0077, 'grad_norm': 0.7012185454368591, 'learning_rate': 1.0188580408590885e-05, 'recon_loss': 0.0076405227882787585, 'recon_static': 0.0056738143786787985, 'recon_motion': 0.0019667084387037905, 'recon_norm': 2.9117112040519713, 'recon_norm_std': 0.3504774607717991, 'target_norm': 1.616256907582283, 'recon_cos': 0.2751661278307438, 'gradnorm/compressor.query': 0.032665576775059164, 'gradnorm/compressor.layer0': 0.28623970821153044, 'gradnorm/compressor.layer1': 0.2916515749478271, 'gradnorm/compressor.layer2': 0.24662705567484908, 'gradnorm/compressor.layer3': 0.22226221789277836, 'gradnorm/compressor.layer4': 0.2050841466837052, 'gradnorm/compressor.layer5': 0.18976579996715787, 'gradnorm/compressor.layer6': 0.1782586361871516, 'gradnorm/compressor.layer7': 0.16328359378610952, 'gradnorm/decoder.layer0': 0.1872837886348625, 'gradnorm/decoder.layer1': 0.15962885456278067, 'gradnorm/decoder.layer2': 0.12789170232702401, 'gradnorm/max_param': 0.246298685669899, 'epoch': 0.2}

  2%|▏         | 390/19090 [11:34<8:25:09,  1.62s/it]
  2%|▏         | 391/19090 [11:37<10:04:58,  1.94s/it]
  2%|▏         | 392/19090 [11:38<9:03:41,  1.74s/it] 
  2%|▏         | 393/19090 [11:40<8:21:08,  1.61s/it]
  2%|▏         | 394/19090 [11:42<10:03:12,  1.94s/it]
  2%|▏         | 395/19090 [11:44<9:02:36,  1.74s/it] 
  2%|▏         | 396/19090 [11:45<8:20:59,  1.61s/it]
  2%|▏         | 397/19090 [11:48<10:18:00,  1.98s/it]
  2%|▏         | 398/19090 [11:49<9:13:15,  1.78s/it] 
  2%|▏         | 399/19090 [11:50<8:27:45,  1.63s/it]
  2%|▏         | 400/19090 [11:53<10:27:50,  2.02s/it]
                                                      
{'loss': 0.0067, 'grad_norm': 0.4204210638999939, 'learning_rate': 1.0450497642744893e-05, 'recon_loss': 0.006716339348349721, 'recon_static': 0.004782220639754087, 'recon_motion': 0.001934118603821844, 'recon_norm': 2.7181858479976655, 'recon_norm_std': 0.36292324811220167, 'target_norm': 1.6194305330514909, 'recon_cos': 0.29377845078706744, 'gradnorm/compressor.query': 0.019265808191176802, 'gradnorm/compressor.layer0': 0.1777948339566219, 'gradnorm/compressor.layer1': 0.17701133966221183, 'gradnorm/compressor.layer2': 0.14354589655941213, 'gradnorm/compressor.layer3': 0.1290937490594846, 'gradnorm/compressor.layer4': 0.12147714311226805, 'gradnorm/compressor.layer5': 0.11336229154961126, 'gradnorm/compressor.layer6': 0.10614971799857927, 'gradnorm/compressor.layer7': 0.09771504539577212, 'gradnorm/decoder.layer0': 0.1120422572749817, 'gradnorm/decoder.layer1': 0.09548433264848621, 'gradnorm/decoder.layer2': 0.07793356567905439, 'gradnorm/max_param': 0.1538352072238922, 'epoch': 0.21}

  2%|▏         | 400/19090 [11:53<10:27:50,  2.02s/it]
  2%|▏         | 401/19090 [11:54<9:19:39,  1.80s/it] 
  2%|▏         | 402/19090 [11:56<8:32:05,  1.64s/it]
  2%|▏         | 403/19090 [11:59<10:18:49,  1.99s/it]
  2%|▏         | 404/19090 [12:00<9:13:07,  1.78s/it] 
  2%|▏         | 405/19090 [12:01<8:27:23,  1.63s/it]
  2%|▏         | 406/19090 [12:04<10:17:31,  1.98s/it]
  2%|▏         | 407/19090 [12:05<9:12:29,  1.77s/it] 
  2%|▏         | 408/19090 [12:07<8:26:44,  1.63s/it]
  2%|▏         | 409/19090 [12:09<10:20:05,  1.99s/it]
  2%|▏         | 410/19090 [12:11<9:14:09,  1.78s/it] 
                                                     
{'loss': 0.0066, 'grad_norm': 0.7714824676513672, 'learning_rate': 1.0712414876898901e-05, 'recon_loss': 0.006571946246549487, 'recon_static': 0.004662299202755094, 'recon_motion': 0.0019096470321528613, 'recon_norm': 2.686742001771927, 'recon_norm_std': 0.35733439922332766, 'target_norm': 1.6235876858234406, 'recon_cos': 0.2972400918602943, 'gradnorm/compressor.query': 0.03334560223601402, 'gradnorm/compressor.layer0': 0.33072522178559993, 'gradnorm/compressor.layer1': 0.3313991139033459, 'gradnorm/compressor.layer2': 0.26988670848751395, 'gradnorm/compressor.layer3': 0.23765018414011296, 'gradnorm/compressor.layer4': 0.22064393017688594, 'gradnorm/compressor.layer5': 0.2044141739010849, 'gradnorm/compressor.layer6': 0.18815455823909763, 'gradnorm/compressor.layer7': 0.17595098565761333, 'gradnorm/decoder.layer0': 0.20149679702562992, 'gradnorm/decoder.layer1': 0.17061277765821187, 'gradnorm/decoder.layer2': 0.13788661113440762, 'gradnorm/max_param': 0.2888764441013336, 'epoch': 0.21}

  2%|▏         | 410/19090 [12:11<9:14:09,  1.78s/it]
  2%|▏         | 411/19090 [12:12<8:27:53,  1.63s/it]
  2%|▏         | 412/19090 [12:15<10:13:16,  1.97s/it]
  2%|▏         | 413/19090 [12:16<9:09:36,  1.77s/it] 
  2%|▏         | 414/19090 [12:17<8:24:24,  1.62s/it]
  2%|▏         | 415/19090 [12:20<10:17:09,  1.98s/it]
  2%|▏         | 416/19090 [12:21<9:11:34,  1.77s/it] 
  2%|▏         | 417/19090 [12:23<8:26:37,  1.63s/it]
  2%|▏         | 418/19090 [12:25<10:06:59,  1.95s/it]
  2%|▏         | 419/19090 [12:27<9:05:09,  1.75s/it] 
  2%|▏         | 420/19090 [12:28<8:21:09,  1.61s/it]
                                                     
{'loss': 0.0072, 'grad_norm': 0.7720927000045776, 'learning_rate': 1.0974332111052908e-05, 'recon_loss': 0.00722406153799966, 'recon_static': 0.0053478068090043965, 'recon_motion': 0.0018762547755613924, 'recon_norm': 2.8283242702484133, 'recon_norm_std': 0.33279369324445723, 'target_norm': 1.6197682708501815, 'recon_cos': 0.28286911100149154, 'gradnorm/compressor.query': 0.03336075398288481, 'gradnorm/compressor.layer0': 0.3301249927902076, 'gradnorm/compressor.layer1': 0.3237417918162592, 'gradnorm/compressor.layer2': 0.26902319529126306, 'gradnorm/compressor.layer3': 0.24081455511557248, 'gradnorm/compressor.layer4': 0.22062850635973408, 'gradnorm/compressor.layer5': 0.20383128838151449, 'gradnorm/compressor.layer6': 0.19248190133671142, 'gradnorm/compressor.layer7': 0.17593893819597184, 'gradnorm/decoder.layer0': 0.20373543384474915, 'gradnorm/decoder.layer1': 0.1741522702789648, 'gradnorm/decoder.layer2': 0.14419689682433584, 'gradnorm/max_param': 0.28722333908081055, 'epoch': 0.22}

  2%|▏         | 420/19090 [12:28<8:21:09,  1.61s/it]
  2%|▏         | 421/19090 [12:31<10:12:54,  1.97s/it]
  2%|▏         | 422/19090 [12:32<9:08:27,  1.76s/it] 
  2%|▏         | 423/19090 [12:33<8:23:56,  1.62s/it]
  2%|▏         | 424/19090 [12:36<10:07:57,  1.95s/it]
  2%|▏         | 425/19090 [12:37<9:05:12,  1.75s/it] 
  2%|▏         | 426/19090 [12:39<8:20:57,  1.61s/it]
  2%|▏         | 427/19090 [12:42<10:23:43,  2.01s/it]
  2%|▏         | 428/19090 [12:43<9:16:00,  1.79s/it] 
  2%|▏         | 429/19090 [12:44<8:45:31,  1.69s/it]
  2%|▏         | 430/19090 [12:47<10:13:41,  1.97s/it]
                                                      
{'loss': 0.0068, 'grad_norm': 0.633567750453949, 'learning_rate': 1.1236249345206914e-05, 'recon_loss': 0.006800626846961677, 'recon_static': 0.0049559876206330955, 'recon_motion': 0.001844639316550456, 'recon_norm': 2.73991973400116, 'recon_norm_std': 0.33391167744994166, 'target_norm': 1.621579998731613, 'recon_cos': 0.291666054725647, 'gradnorm/compressor.query': 0.02648245376887716, 'gradnorm/compressor.layer0': 0.2642222981860446, 'gradnorm/compressor.layer1': 0.2672419141272369, 'gradnorm/compressor.layer2': 0.22066619867961168, 'gradnorm/compressor.layer3': 0.1962889391391044, 'gradnorm/compressor.layer4': 0.1815488234669104, 'gradnorm/compressor.layer5': 0.17033472550462272, 'gradnorm/compressor.layer6': 0.1566450251581986, 'gradnorm/compressor.layer7': 0.14594032983262203, 'gradnorm/decoder.layer0': 0.17020340868499795, 'gradnorm/decoder.layer1': 0.1452154704672367, 'gradnorm/decoder.layer2': 0.11993883041028605, 'gradnorm/max_param': 0.2302749902009964, 'epoch': 0.23}

  2%|▏         | 430/19090 [12:47<10:13:41,  1.97s/it]
  2%|▏         | 431/19090 [12:48<9:13:46,  1.78s/it] 
  2%|▏         | 432/19090 [12:50<8:28:22,  1.63s/it]
  2%|▏         | 433/19090 [12:52<9:32:14,  1.84s/it]
  2%|▏         | 434/19090 [12:53<8:40:05,  1.67s/it]
  2%|▏         | 435/19090 [12:55<8:25:10,  1.62s/it]
  2%|▏         | 436/19090 [12:57<9:52:45,  1.91s/it]
  2%|▏         | 437/19090 [12:58<8:54:11,  1.72s/it]
  2%|▏         | 438/19090 [13:00<8:39:52,  1.67s/it]
  2%|▏         | 439/19090 [13:03<10:06:22,  1.95s/it]
  2%|▏         | 440/19090 [13:04<9:04:06,  1.75s/it] 
                                                     
{'loss': 0.006, 'grad_norm': 0.4544029235839844, 'learning_rate': 1.1498166579360922e-05, 'recon_loss': 0.005992153217084706, 'recon_static': 0.004184163454920053, 'recon_motion': 0.0018079897796269505, 'recon_norm': 2.567637640237808, 'recon_norm_std': 0.3432423330843449, 'target_norm': 1.6208462089300155, 'recon_cos': 0.314274363219738, 'gradnorm/compressor.query': 0.018943907990452834, 'gradnorm/compressor.layer0': 0.18452551969308495, 'gradnorm/compressor.layer1': 0.19093895330984187, 'gradnorm/compressor.layer2': 0.15534854368020487, 'gradnorm/compressor.layer3': 0.14069552772255986, 'gradnorm/compressor.layer4': 0.13199174645523212, 'gradnorm/compressor.layer5': 0.12138099979177247, 'gradnorm/compressor.layer6': 0.11445850694911334, 'gradnorm/compressor.layer7': 0.10742676463702977, 'gradnorm/decoder.layer0': 0.12504549986533503, 'gradnorm/decoder.layer1': 0.10698063133757524, 'gradnorm/decoder.layer2': 0.08837265709127906, 'gradnorm/max_param': 0.16039898991584778, 'epoch': 0.23}

  2%|▏         | 440/19090 [13:04<9:04:06,  1.75s/it]
  2%|▏         | 441/19090 [13:05<8:20:37,  1.61s/it]
  2%|▏         | 442/19090 [13:08<10:13:22,  1.97s/it]
  2%|▏         | 443/19090 [13:09<9:08:36,  1.77s/it] 
  2%|▏         | 444/19090 [13:11<8:23:14,  1.62s/it]
  2%|▏         | 445/19090 [13:13<10:11:54,  1.97s/it]
  2%|▏         | 446/19090 [13:15<9:07:31,  1.76s/it] 
  2%|▏         | 447/19090 [13:16<8:22:34,  1.62s/it]
  2%|▏         | 448/19090 [13:19<10:23:46,  2.01s/it]
  2%|▏         | 449/19090 [13:20<9:15:44,  1.79s/it] 
  2%|▏         | 450/19090 [13:21<8:28:52,  1.64s/it]
                                                     
{'loss': 0.0063, 'grad_norm': 0.9435873627662659, 'learning_rate': 1.176008381351493e-05, 'recon_loss': 0.006268958188593387, 'recon_static': 0.004488669813144952, 'recon_motion': 0.0017802885005949066, 'recon_norm': 2.628937989473343, 'recon_norm_std': 0.33007543683052065, 'target_norm': 1.6202621191740036, 'recon_cos': 0.30671968013048173, 'gradnorm/compressor.query': 0.039740860674193704, 'gradnorm/compressor.layer0': 0.3995876541275346, 'gradnorm/compressor.layer1': 0.4052402528711968, 'gradnorm/compressor.layer2': 0.32938706553254876, 'gradnorm/compressor.layer3': 0.294387320303197, 'gradnorm/compressor.layer4': 0.2676093229497682, 'gradnorm/compressor.layer5': 0.24921557923694554, 'gradnorm/compressor.layer6': 0.23013688294088305, 'gradnorm/compressor.layer7': 0.21341216223187334, 'gradnorm/decoder.layer0': 0.24728144042192018, 'gradnorm/decoder.layer1': 0.21278786585283965, 'gradnorm/decoder.layer2': 0.17626815224739487, 'gradnorm/max_param': 0.3471584618091583, 'epoch': 0.24}

  2%|▏         | 450/19090 [13:21<8:28:52,  1.64s/it]
  2%|▏         | 451/19090 [13:24<10:10:08,  1.96s/it]
  2%|▏         | 452/19090 [13:25<9:06:39,  1.76s/it] 
  2%|▏         | 453/19090 [13:27<8:22:32,  1.62s/it]
  2%|▏         | 454/19090 [13:29<10:05:20,  1.95s/it]
  2%|▏         | 455/19090 [13:31<9:04:24,  1.75s/it] 
  2%|▏         | 456/19090 [13:32<8:21:36,  1.62s/it]
  2%|▏         | 457/19090 [13:35<9:59:19,  1.93s/it]
  2%|▏         | 458/19090 [13:36<9:00:25,  1.74s/it]
  2%|▏         | 459/19090 [13:37<8:18:15,  1.60s/it]
  2%|▏         | 460/19090 [13:40<10:03:20,  1.94s/it]
                                                      
{'loss': 0.0066, 'grad_norm': 0.6835563778877258, 'learning_rate': 1.2022001047668938e-05, 'recon_loss': 0.00657901733648032, 'recon_static': 0.004827121109701693, 'recon_motion': 0.001751896255882457, 'recon_norm': 2.697435140609741, 'recon_norm_std': 0.31455749794840815, 'target_norm': 1.6196452379226685, 'recon_cos': 0.29723177179694177, 'gradnorm/compressor.query': 0.029680635417803542, 'gradnorm/compressor.layer0': 0.28190237783536753, 'gradnorm/compressor.layer1': 0.28691105902692254, 'gradnorm/compressor.layer2': 0.23630179896991635, 'gradnorm/compressor.layer3': 0.212317114205763, 'gradnorm/compressor.layer4': 0.1961867222243414, 'gradnorm/compressor.layer5': 0.18191413386275726, 'gradnorm/compressor.layer6': 0.16994087069214062, 'gradnorm/compressor.layer7': 0.16039997838169115, 'gradnorm/decoder.layer0': 0.18436978057973655, 'gradnorm/decoder.layer1': 0.15994622252509172, 'gradnorm/decoder.layer2': 0.13378825500690109, 'gradnorm/max_param': 0.2449178546667099, 'epoch': 0.24}

  2%|▏         | 460/19090 [13:40<10:03:20,  1.94s/it]
  2%|▏         | 461/19090 [13:41<9:02:38,  1.75s/it] 
  2%|▏         | 462/19090 [13:43<8:19:21,  1.61s/it]
  2%|▏         | 463/19090 [13:46<10:30:18,  2.03s/it]
  2%|▏         | 464/19090 [13:47<9:21:49,  1.81s/it] 
  2%|▏         | 465/19090 [13:48<8:33:33,  1.65s/it]
  2%|▏         | 466/19090 [13:51<10:25:02,  2.01s/it]
  2%|▏         | 467/19090 [13:52<9:17:33,  1.80s/it] 
  2%|▏         | 468/19090 [13:54<8:29:52,  1.64s/it]
  2%|▏         | 469/19090 [13:56<9:36:49,  1.86s/it]
  2%|▏         | 470/19090 [13:57<8:43:36,  1.69s/it]
                                                     
{'loss': 0.0056, 'grad_norm': 0.4357492923736572, 'learning_rate': 1.2283918281822944e-05, 'recon_loss': 0.00557468100450933, 'recon_static': 0.003854319854872301, 'recon_motion': 0.0017203612689627335, 'recon_norm': 2.4756413102149963, 'recon_norm_std': 0.32964227348566055, 'target_norm': 1.6188421368598938, 'recon_cos': 0.3265958957374096, 'gradnorm/compressor.query': 0.01807217087149323, 'gradnorm/compressor.layer0': 0.18017183211632676, 'gradnorm/compressor.layer1': 0.18286587107252142, 'gradnorm/compressor.layer2': 0.15030105202237523, 'gradnorm/compressor.layer3': 0.13546362087303127, 'gradnorm/compressor.layer4': 0.12476324966548649, 'gradnorm/compressor.layer5': 0.1162835022824307, 'gradnorm/compressor.layer6': 0.10832594260198082, 'gradnorm/compressor.layer7': 0.10200619978622137, 'gradnorm/decoder.layer0': 0.11724839217948238, 'gradnorm/decoder.layer1': 0.101488567916144, 'gradnorm/decoder.layer2': 0.08621846943937878, 'gradnorm/max_param': 0.15585599839687347, 'epoch': 0.25}

  2%|▏         | 470/19090 [13:57<8:43:36,  1.69s/it]
  2%|▏         | 471/19090 [13:59<8:06:04,  1.57s/it]
  2%|▏         | 472/19090 [14:01<9:49:16,  1.90s/it]
  2%|▏         | 473/19090 [14:03<8:52:02,  1.71s/it]
  2%|▏         | 474/19090 [14:04<8:12:26,  1.59s/it]
  2%|▏         | 475/19090 [14:06<9:55:02,  1.92s/it]
  2%|▏         | 476/19090 [14:08<8:56:06,  1.73s/it]
  2%|▏         | 477/19090 [14:09<8:18:50,  1.61s/it]
  3%|▎         | 478/19090 [14:12<9:48:04,  1.90s/it]
  3%|▎         | 479/19090 [14:13<8:50:57,  1.71s/it]
  3%|▎         | 480/19090 [14:14<8:25:17,  1.63s/it]
                                                     
{'loss': 0.0056, 'grad_norm': 0.9934445023536682, 'learning_rate': 1.2545835515976954e-05, 'recon_loss': 0.005596033635083586, 'recon_static': 0.003885376686230302, 'recon_motion': 0.0017106569663155823, 'recon_norm': 2.4758536875247956, 'recon_norm_std': 0.32139159068465234, 'target_norm': 1.6222581565380096, 'recon_cos': 0.3258670084178448, 'gradnorm/compressor.query': 0.03664441796447839, 'gradnorm/compressor.layer0': 0.3982608440943876, 'gradnorm/compressor.layer1': 0.416948463595435, 'gradnorm/compressor.layer2': 0.34755348444047945, 'gradnorm/compressor.layer3': 0.3114079347977375, 'gradnorm/compressor.layer4': 0.28695137191639924, 'gradnorm/compressor.layer5': 0.2696179236773149, 'gradnorm/compressor.layer6': 0.25223128171549525, 'gradnorm/compressor.layer7': 0.2333088786786578, 'gradnorm/decoder.layer0': 0.2674883253232673, 'gradnorm/decoder.layer1': 0.2301514672185766, 'gradnorm/decoder.layer2': 0.19430593517792943, 'gradnorm/max_param': 0.3471047282218933, 'epoch': 0.25}

  3%|▎         | 480/19090 [14:14<8:25:17,  1.63s/it]
  3%|▎         | 481/19090 [14:17<9:44:40,  1.89s/it]
  3%|▎         | 482/19090 [14:18<8:48:23,  1.70s/it]
  3%|▎         | 483/19090 [14:20<9:08:56,  1.77s/it]
  3%|▎         | 484/19090 [14:22<9:55:52,  1.92s/it]
  3%|▎         | 485/19090 [14:24<8:56:19,  1.73s/it]
  3%|▎         | 486/19090 [14:25<8:47:49,  1.70s/it]
  3%|▎         | 487/19090 [14:28<10:12:12,  1.97s/it]
  3%|▎         | 488/19090 [14:29<9:09:06,  1.77s/it] 
  3%|▎         | 489/19090 [14:30<8:24:08,  1.63s/it]
  3%|▎         | 490/19090 [14:33<9:51:54,  1.91s/it]
                                                     
{'loss': 0.006, 'grad_norm': 0.5314133167266846, 'learning_rate': 1.2807752750130958e-05, 'recon_loss': 0.006002677220385522, 'recon_static': 0.0043196846905630085, 'recon_motion': 0.0016829926462378353, 'recon_norm': 2.570494592189789, 'recon_norm_std': 0.3051811009645462, 'target_norm': 1.6211038917303084, 'recon_cos': 0.3152059853076935, 'gradnorm/compressor.query': 0.019659968252853984, 'gradnorm/compressor.layer0': 0.21129289463860904, 'gradnorm/compressor.layer1': 0.2189444205872526, 'gradnorm/compressor.layer2': 0.18336003219298178, 'gradnorm/compressor.layer3': 0.16551744099480278, 'gradnorm/compressor.layer4': 0.1523753463551072, 'gradnorm/compressor.layer5': 0.14497074115633185, 'gradnorm/compressor.layer6': 0.13529884886213175, 'gradnorm/compressor.layer7': 0.125902368446881, 'gradnorm/decoder.layer0': 0.1476310986083619, 'gradnorm/decoder.layer1': 0.12870522198410767, 'gradnorm/decoder.layer2': 0.10799738435260035, 'gradnorm/max_param': 0.18330036103725433, 'epoch': 0.26}

  3%|▎         | 490/19090 [14:33<9:51:54,  1.91s/it]
  3%|▎         | 491/19090 [14:34<8:54:21,  1.72s/it]
  3%|▎         | 492/19090 [14:36<8:25:31,  1.63s/it]
  3%|▎         | 493/19090 [14:38<10:02:17,  1.94s/it]
  3%|▎         | 494/19090 [14:40<9:01:05,  1.75s/it] 
  3%|▎         | 495/19090 [14:41<8:17:57,  1.61s/it]
  3%|▎         | 496/19090 [14:44<11:02:44,  2.14s/it]
  3%|▎         | 497/19090 [14:46<9:42:58,  1.88s/it] 
  3%|▎         | 498/19090 [14:47<8:47:41,  1.70s/it]
  3%|▎         | 499/19090 [14:49<9:34:34,  1.85s/it]
  3%|▎         | 500/19090 [14:50<8:42:18,  1.69s/it]
                                                     
{'loss': 0.0061, 'grad_norm': 0.6681753396987915, 'learning_rate': 1.3069669984284966e-05, 'recon_loss': 0.006047975108958781, 'recon_static': 0.004384714789921418, 'recon_motion': 0.0016632603423204274, 'recon_norm': 2.5864190995693206, 'recon_norm_std': 0.2960774749517441, 'target_norm': 1.6215242326259613, 'recon_cos': 0.3133209504187107, 'gradnorm/compressor.query': 0.025617016596862444, 'gradnorm/compressor.layer0': 0.2746310750491647, 'gradnorm/compressor.layer1': 0.2782313422162401, 'gradnorm/compressor.layer2': 0.22799693850398473, 'gradnorm/compressor.layer3': 0.20513142392689937, 'gradnorm/compressor.layer4': 0.19046715587859514, 'gradnorm/compressor.layer5': 0.17878558991660118, 'gradnorm/compressor.layer6': 0.17014541391805435, 'gradnorm/compressor.layer7': 0.1586139724785904, 'gradnorm/decoder.layer0': 0.1828279146395866, 'gradnorm/decoder.layer1': 0.15795071210234263, 'gradnorm/decoder.layer2': 0.13483812621004734, 'gradnorm/max_param': 0.24100743234157562, 'epoch': 0.26}

  3%|▎         | 500/19090 [14:50<8:42:18,  1.69s/it]
  3%|▎         | 501/19090 [14:56<14:56:09,  2.89s/it]
  3%|▎         | 502/19090 [14:57<12:26:26,  2.41s/it]
  3%|▎         | 503/19090 [14:59<10:42:28,  2.07s/it]
  3%|▎         | 504/19090 [15:00<9:29:24,  1.84s/it] 
  3%|▎         | 505/19090 [15:01<8:38:08,  1.67s/it]
  3%|▎         | 506/19090 [15:03<8:02:28,  1.56s/it]
  3%|▎         | 507/19090 [15:04<7:37:50,  1.48s/it]
  3%|▎         | 508/19090 [15:05<7:19:59,  1.42s/it]
  3%|▎         | 509/19090 [15:06<7:07:40,  1.38s/it]
  3%|▎         | 510/19090 [15:08<7:00:03,  1.36s/it]
                                                     
{'loss': 0.0062, 'grad_norm': 0.830612063407898, 'learning_rate': 1.3331587218438973e-05, 'recon_loss': 0.00623255823738873, 'recon_static': 0.004587756446562707, 'recon_motion': 0.0016448016307549552, 'recon_norm': 2.6296012103557587, 'recon_norm_std': 0.2847331836819649, 'target_norm': 1.621987760066986, 'recon_cos': 0.30732705593109133, 'gradnorm/compressor.query': 0.0316731481529122, 'gradnorm/compressor.layer0': 0.3425161477445118, 'gradnorm/compressor.layer1': 0.34555593687004427, 'gradnorm/compressor.layer2': 0.287952198084824, 'gradnorm/compressor.layer3': 0.2578345058181684, 'gradnorm/compressor.layer4': 0.23753010282978465, 'gradnorm/compressor.layer5': 0.22297388141718047, 'gradnorm/compressor.layer6': 0.20909932798620154, 'gradnorm/compressor.layer7': 0.19272008851429145, 'gradnorm/decoder.layer0': 0.2243355948345763, 'gradnorm/decoder.layer1': 0.1953042314686531, 'gradnorm/decoder.layer2': 0.16528811651728065, 'gradnorm/max_param': 0.2979784607887268, 'epoch': 0.27}

  3%|▎         | 510/19090 [15:08<7:00:03,  1.36s/it]
  3%|▎         | 511/19090 [15:09<6:54:26,  1.34s/it]
  3%|▎         | 512/19090 [15:10<6:50:03,  1.32s/it]
  3%|▎         | 513/19090 [15:12<6:51:26,  1.33s/it]
  3%|▎         | 514/19090 [15:13<7:08:46,  1.38s/it]
  3%|▎         | 515/19090 [15:15<7:21:18,  1.43s/it]
  3%|▎         | 516/19090 [15:17<8:34:33,  1.66s/it]
  3%|▎         | 517/19090 [15:19<8:30:30,  1.65s/it]
  3%|▎         | 518/19090 [15:20<8:25:02,  1.63s/it]
  3%|▎         | 519/19090 [15:23<9:48:41,  1.90s/it]
  3%|▎         | 520/19090 [15:24<8:59:25,  1.74s/it]
                                                     
{'loss': 0.0057, 'grad_norm': 0.7645646929740906, 'learning_rate': 1.359350445259298e-05, 'recon_loss': 0.005668779776897281, 'recon_static': 0.00405683318967931, 'recon_motion': 0.001611946424236521, 'recon_norm': 2.504912185668945, 'recon_norm_std': 0.29134612530469894, 'target_norm': 1.6194012224674226, 'recon_cos': 0.32333201691508295, 'gradnorm/compressor.query': 0.027311695844167688, 'gradnorm/compressor.layer0': 0.31306591053173266, 'gradnorm/compressor.layer1': 0.31879495883345776, 'gradnorm/compressor.layer2': 0.257330154263628, 'gradnorm/compressor.layer3': 0.23551995833728012, 'gradnorm/compressor.layer4': 0.21826530299988006, 'gradnorm/compressor.layer5': 0.2082806441118681, 'gradnorm/compressor.layer6': 0.19515262736309413, 'gradnorm/compressor.layer7': 0.1810388955283536, 'gradnorm/decoder.layer0': 0.2076703589348679, 'gradnorm/decoder.layer1': 0.18287340844948696, 'gradnorm/decoder.layer2': 0.1550305824526852, 'gradnorm/max_param': 0.2745981216430664, 'epoch': 0.27}

  3%|▎         | 520/19090 [15:24<8:59:25,  1.74s/it]
  3%|▎         | 521/19090 [15:25<8:22:26,  1.62s/it]
  3%|▎         | 522/19090 [15:28<9:50:27,  1.91s/it]
  3%|▎         | 523/19090 [15:29<8:52:30,  1.72s/it]
  3%|▎         | 524/19090 [15:31<8:19:22,  1.61s/it]
  3%|▎         | 525/19090 [15:33<10:14:23,  1.99s/it]
  3%|▎         | 526/19090 [15:35<9:09:05,  1.77s/it] 
  3%|▎         | 527/19090 [15:36<8:51:56,  1.72s/it]
  3%|▎         | 528/19090 [15:39<9:58:03,  1.93s/it]
  3%|▎         | 529/19090 [15:40<8:57:27,  1.74s/it]
  3%|▎         | 530/19090 [15:42<8:56:21,  1.73s/it]
                                                     
{'loss': 0.0052, 'grad_norm': 0.509499728679657, 'learning_rate': 1.3855421686746989e-05, 'recon_loss': 0.005220999987795949, 'recon_static': 0.0036398479423951358, 'recon_motion': 0.0015811518707778305, 'recon_norm': 2.4007875978946687, 'recon_norm_std': 0.2946502223610878, 'target_norm': 1.6176275879144668, 'recon_cos': 0.33877367377281187, 'gradnorm/compressor.query': 0.01736073042366094, 'gradnorm/compressor.layer0': 0.20384545205432164, 'gradnorm/compressor.layer1': 0.21254009296354628, 'gradnorm/compressor.layer2': 0.1766437932033612, 'gradnorm/compressor.layer3': 0.1603615770453606, 'gradnorm/compressor.layer4': 0.1463909296164314, 'gradnorm/compressor.layer5': 0.13824129416378417, 'gradnorm/compressor.layer6': 0.12955651131271928, 'gradnorm/compressor.layer7': 0.11804860361829303, 'gradnorm/decoder.layer0': 0.13720838199914454, 'gradnorm/decoder.layer1': 0.12053843036624066, 'gradnorm/decoder.layer2': 0.10500712388996837, 'gradnorm/max_param': 0.17843621969223022, 'epoch': 0.28}

  3%|▎         | 530/19090 [15:42<8:56:21,  1.73s/it]
  3%|▎         | 531/19090 [15:44<9:55:08,  1.92s/it]
  3%|▎         | 532/19090 [15:46<9:12:18,  1.79s/it]
  3%|▎         | 533/19090 [15:47<8:36:49,  1.67s/it]
  3%|▎         | 534/19090 [15:50<10:04:15,  1.95s/it]
  3%|▎         | 535/19090 [15:51<9:39:57,  1.88s/it] 
  3%|▎         | 536/19090 [15:53<8:45:32,  1.70s/it]
  3%|▎         | 537/19090 [15:55<9:39:22,  1.87s/it]
  3%|▎         | 538/19090 [15:57<9:20:46,  1.81s/it]
  3%|▎         | 539/19090 [15:58<8:31:33,  1.65s/it]
  3%|▎         | 540/19090 [16:00<9:33:05,  1.85s/it]
                                                     
{'loss': 0.0048, 'grad_norm': 0.5142649412155151, 'learning_rate': 1.4117338920900997e-05, 'recon_loss': 0.00480803003301844, 'recon_static': 0.003218286688206717, 'recon_motion': 0.0015897432313067839, 'recon_norm': 2.2961816668510435, 'recon_norm_std': 0.2983628161251545, 'target_norm': 1.6242370367050172, 'recon_cos': 0.3535385116934776, 'gradnorm/compressor.query': 0.018286535891430943, 'gradnorm/compressor.layer0': 0.21100377878256246, 'gradnorm/compressor.layer1': 0.21618987500324532, 'gradnorm/compressor.layer2': 0.18006502794810358, 'gradnorm/compressor.layer3': 0.15892835679908485, 'gradnorm/compressor.layer4': 0.1467683074870101, 'gradnorm/compressor.layer5': 0.1367628162757627, 'gradnorm/compressor.layer6': 0.12859378861047807, 'gradnorm/compressor.layer7': 0.11913342923181032, 'gradnorm/decoder.layer0': 0.13753905798216762, 'gradnorm/decoder.layer1': 0.1199153157468865, 'gradnorm/decoder.layer2': 0.10450331013743726, 'gradnorm/max_param': 0.18445256352424622, 'epoch': 0.28}

  3%|▎         | 540/19090 [16:00<9:33:05,  1.85s/it]
  3%|▎         | 541/19090 [16:02<9:22:10,  1.82s/it]
  3%|▎         | 542/19090 [16:03<8:32:16,  1.66s/it]
  3%|▎         | 543/19090 [16:05<9:33:08,  1.85s/it]
  3%|▎         | 544/19090 [16:07<9:12:09,  1.79s/it]
  3%|▎         | 545/19090 [16:08<8:25:29,  1.64s/it]
  3%|▎         | 546/19090 [16:11<9:32:32,  1.85s/it]
  3%|▎         | 547/19090 [16:12<8:54:11,  1.73s/it]
  3%|▎         | 548/19090 [16:13<8:13:26,  1.60s/it]
  3%|▎         | 549/19090 [16:16<9:39:23,  1.87s/it]
  3%|▎         | 550/19090 [16:18<9:11:46,  1.79s/it]
                                                     
{'loss': 0.0049, 'grad_norm': 0.6563799381256104, 'learning_rate': 1.4379256155055005e-05, 'recon_loss': 0.004866247170139104, 'recon_static': 0.003310417215107009, 'recon_motion': 0.0015558298240648583, 'recon_norm': 2.316133463382721, 'recon_norm_std': 0.29052467569708823, 'target_norm': 1.6196858137845993, 'recon_cos': 0.3515759833157063, 'gradnorm/compressor.query': 0.02217927452234016, 'gradnorm/compressor.layer0': 0.26280480977984916, 'gradnorm/compressor.layer1': 0.28173964316635597, 'gradnorm/compressor.layer2': 0.22707828642255978, 'gradnorm/compressor.layer3': 0.2041012430872996, 'gradnorm/compressor.layer4': 0.18956948669052495, 'gradnorm/compressor.layer5': 0.17330326715253697, 'gradnorm/compressor.layer6': 0.160776443585695, 'gradnorm/compressor.layer7': 0.15278965542299772, 'gradnorm/decoder.layer0': 0.1775279680568831, 'gradnorm/decoder.layer1': 0.1553136245736716, 'gradnorm/decoder.layer2': 0.1336949846054922, 'gradnorm/max_param': 0.2317528873682022, 'epoch': 0.29}

  3%|▎         | 550/19090 [16:18<9:11:46,  1.79s/it]
  3%|▎         | 551/19090 [16:19<8:25:30,  1.64s/it]
  3%|▎         | 552/19090 [16:21<9:52:56,  1.92s/it]
  3%|▎         | 553/19090 [16:23<8:55:32,  1.73s/it]
  3%|▎         | 554/19090 [16:24<8:14:21,  1.60s/it]
  3%|▎         | 555/19090 [16:27<9:45:07,  1.89s/it]
  3%|▎         | 556/19090 [16:28<9:11:36,  1.79s/it]
  3%|▎         | 557/19090 [16:29<8:25:57,  1.64s/it]
  3%|▎         | 558/19090 [16:32<9:33:35,  1.86s/it]
  3%|▎         | 559/19090 [16:33<9:16:36,  1.80s/it]
  3%|▎         | 560/19090 [16:35<8:28:51,  1.65s/it]
                                                     
{'loss': 0.0062, 'grad_norm': 0.9133386015892029, 'learning_rate': 1.4641173389209013e-05, 'recon_loss': 0.006217625108547509, 'recon_static': 0.004680082434788346, 'recon_motion': 0.0015375425689853727, 'recon_norm': 2.6330974459648133, 'recon_norm_std': 0.2548270050436258, 'target_norm': 1.6211664915084838, 'recon_cos': 0.3088222213089466, 'gradnorm/compressor.query': 0.03265280320946046, 'gradnorm/compressor.layer0': 0.3778972956849601, 'gradnorm/compressor.layer1': 0.38422194666888154, 'gradnorm/compressor.layer2': 0.3100433586162275, 'gradnorm/compressor.layer3': 0.2767639547136208, 'gradnorm/compressor.layer4': 0.25812140656249105, 'gradnorm/compressor.layer5': 0.23957942444611824, 'gradnorm/compressor.layer6': 0.22885588231758186, 'gradnorm/compressor.layer7': 0.21804515497615148, 'gradnorm/decoder.layer0': 0.2507126010091768, 'gradnorm/decoder.layer1': 0.2187121903577275, 'gradnorm/decoder.layer2': 0.18795654227108757, 'gradnorm/max_param': 0.33229029178619385, 'epoch': 0.29}

  3%|▎         | 560/19090 [16:35<8:28:51,  1.65s/it]
  3%|▎         | 561/19090 [16:37<9:49:19,  1.91s/it]
  3%|▎         | 562/19090 [16:39<9:18:30,  1.81s/it]
  3%|▎         | 563/19090 [16:40<8:29:59,  1.65s/it]
  3%|▎         | 564/19090 [16:42<9:22:15,  1.82s/it]
  3%|▎         | 565/19090 [16:44<8:40:44,  1.69s/it]
  3%|▎         | 566/19090 [16:45<8:03:54,  1.57s/it]
  3%|▎         | 567/19090 [16:48<10:04:39,  1.96s/it]
  3%|▎         | 568/19090 [16:49<9:03:46,  1.76s/it] 
  3%|▎         | 569/19090 [16:51<8:33:51,  1.66s/it]
  3%|▎         | 570/19090 [16:53<10:03:59,  1.96s/it]
                                                      
{'loss': 0.0059, 'grad_norm': 0.7544755339622498, 'learning_rate': 1.4903090623363017e-05, 'recon_loss': 0.005891912104561925, 'recon_static': 0.0043678924033883956, 'recon_motion': 0.0015240197826642543, 'recon_norm': 2.563798028230667, 'recon_norm_std': 0.25509249567985537, 'target_norm': 1.6182936489582063, 'recon_cos': 0.3181811347603798, 'gradnorm/compressor.query': 0.02481658628565967, 'gradnorm/compressor.layer0': 0.2973859931778711, 'gradnorm/compressor.layer1': 0.31467947306618943, 'gradnorm/compressor.layer2': 0.26422823383042615, 'gradnorm/compressor.layer3': 0.23456484735047714, 'gradnorm/compressor.layer4': 0.2176523256735415, 'gradnorm/compressor.layer5': 0.20213689172179192, 'gradnorm/compressor.layer6': 0.18998463783462455, 'gradnorm/compressor.layer7': 0.17794919917198948, 'gradnorm/decoder.layer0': 0.2059750425621597, 'gradnorm/decoder.layer1': 0.1801117987603282, 'gradnorm/decoder.layer2': 0.15956406902112588, 'gradnorm/max_param': 0.26069512963294983, 'epoch': 0.3}

  3%|▎         | 570/19090 [16:53<10:03:59,  1.96s/it]
  3%|▎         | 571/19090 [16:55<9:01:52,  1.76s/it] 
  3%|▎         | 572/19090 [16:56<8:18:35,  1.62s/it]
  3%|▎         | 573/19090 [16:59<10:11:55,  1.98s/it]
  3%|▎         | 574/19090 [17:00<9:08:18,  1.78s/it] 
  3%|▎         | 575/19090 [17:01<8:23:46,  1.63s/it]
  3%|▎         | 576/19090 [17:04<10:22:16,  2.02s/it]
  3%|▎         | 577/19090 [17:05<9:14:41,  1.80s/it] 
  3%|▎         | 578/19090 [17:07<8:27:24,  1.64s/it]
  3%|▎         | 579/19090 [17:10<10:21:32,  2.01s/it]
  3%|▎         | 580/19090 [17:11<9:14:14,  1.80s/it] 
                                                     
{'loss': 0.0049, 'grad_norm': 0.4959809482097626, 'learning_rate': 1.5165007857517025e-05, 'recon_loss': 0.004869608161970973, 'recon_static': 0.003363473410718143, 'recon_motion': 0.001506134759983979, 'recon_norm': 2.318976551294327, 'recon_norm_std': 0.27344827726483345, 'target_norm': 1.6202606976032257, 'recon_cos': 0.35158301666378977, 'gradnorm/compressor.query': 0.017513627032361565, 'gradnorm/compressor.layer0': 0.20232324827749334, 'gradnorm/compressor.layer1': 0.20818779550770172, 'gradnorm/compressor.layer2': 0.17000325764311697, 'gradnorm/compressor.layer3': 0.15279539723437008, 'gradnorm/compressor.layer4': 0.1423856598892477, 'gradnorm/compressor.layer5': 0.13258481845582396, 'gradnorm/compressor.layer6': 0.12300332373445076, 'gradnorm/compressor.layer7': 0.11568813372386388, 'gradnorm/decoder.layer0': 0.1340027376720837, 'gradnorm/decoder.layer1': 0.11756961829757279, 'gradnorm/decoder.layer2': 0.10497924449118089, 'gradnorm/max_param': 0.17730972170829773, 'epoch': 0.3}

  3%|▎         | 580/19090 [17:11<9:14:14,  1.80s/it]
  3%|▎         | 581/19090 [17:12<8:27:44,  1.65s/it]
  3%|▎         | 582/19090 [17:15<10:25:03,  2.03s/it]
  3%|▎         | 583/19090 [17:16<9:16:43,  1.80s/it] 
  3%|▎         | 584/19090 [17:18<8:28:54,  1.65s/it]
  3%|▎         | 585/19090 [17:20<9:15:43,  1.80s/it]
  3%|▎         | 586/19090 [17:21<8:28:17,  1.65s/it]
  3%|▎         | 587/19090 [17:22<7:55:14,  1.54s/it]
  3%|▎         | 588/19090 [17:25<9:09:08,  1.78s/it]
  3%|▎         | 589/19090 [17:26<8:30:02,  1.65s/it]
  3%|▎         | 590/19090 [17:28<8:03:45,  1.57s/it]
                                                     
{'loss': 0.0043, 'grad_norm': 0.41141477227211, 'learning_rate': 1.5426925091671033e-05, 'recon_loss': 0.004331397404894233, 'recon_static': 0.002833225985523313, 'recon_motion': 0.001498171323328279, 'recon_norm': 2.1799804627895356, 'recon_norm_std': 0.28059526681900027, 'target_norm': 1.6218852043151855, 'recon_cos': 0.3742730624973774, 'gradnorm/compressor.query': 0.013043321868769673, 'gradnorm/compressor.layer0': 0.15901448740237856, 'gradnorm/compressor.layer1': 0.1710411783560668, 'gradnorm/compressor.layer2': 0.14208486227105968, 'gradnorm/compressor.layer3': 0.12808063570563175, 'gradnorm/compressor.layer4': 0.11864197061665596, 'gradnorm/compressor.layer5': 0.11067182350749921, 'gradnorm/compressor.layer6': 0.1055265903816676, 'gradnorm/compressor.layer7': 0.09893844952672887, 'gradnorm/decoder.layer0': 0.11393928960471325, 'gradnorm/decoder.layer1': 0.09968367778541787, 'gradnorm/decoder.layer2': 0.08826480792540725, 'gradnorm/max_param': 0.14060702919960022, 'epoch': 0.31}

  3%|▎         | 590/19090 [17:28<8:03:45,  1.57s/it]
  3%|▎         | 591/19090 [17:30<9:38:29,  1.88s/it]
  3%|▎         | 592/19090 [17:31<8:49:54,  1.72s/it]
  3%|▎         | 593/19090 [17:33<8:25:09,  1.64s/it]
  3%|▎         | 594/19090 [17:36<9:59:01,  1.94s/it]
  3%|▎         | 595/19090 [17:37<8:58:22,  1.75s/it]
  3%|▎         | 596/19090 [17:38<8:28:30,  1.65s/it]
  3%|▎         | 597/19090 [17:41<9:35:55,  1.87s/it]
  3%|▎         | 598/19090 [17:42<8:41:27,  1.69s/it]
  3%|▎         | 599/19090 [17:44<8:30:23,  1.66s/it]
  3%|▎         | 600/19090 [17:46<9:34:52,  1.87s/it]
                                                     
{'loss': 0.005, 'grad_norm': 0.8077011108398438, 'learning_rate': 1.568884232582504e-05, 'recon_loss': 0.0049664625898003575, 'recon_static': 0.0034971036191564055, 'recon_motion': 0.0014693588309455663, 'recon_norm': 2.346997302770615, 'recon_norm_std': 0.25899222008883954, 'target_norm': 1.6194131404161454, 'recon_cos': 0.3497369736433029, 'gradnorm/compressor.query': 0.026861801267833937, 'gradnorm/compressor.layer0': 0.3099746830854404, 'gradnorm/compressor.layer1': 0.334617672411432, 'gradnorm/compressor.layer2': 0.27808280313973033, 'gradnorm/compressor.layer3': 0.25205089830693655, 'gradnorm/compressor.layer4': 0.23404096005162003, 'gradnorm/compressor.layer5': 0.21652002098255085, 'gradnorm/compressor.layer6': 0.20374424699548174, 'gradnorm/compressor.layer7': 0.1952014798298325, 'gradnorm/decoder.layer0': 0.22694529882649833, 'gradnorm/decoder.layer1': 0.1992689611699613, 'gradnorm/decoder.layer2': 0.1739619468601092, 'gradnorm/max_param': 0.2691056430339813, 'epoch': 0.31}

  3%|▎         | 600/19090 [17:46<9:34:52,  1.87s/it]
  3%|▎         | 601/19090 [17:47<8:40:43,  1.69s/it]
  3%|▎         | 602/19090 [17:49<8:49:41,  1.72s/it]
  3%|▎         | 603/19090 [17:51<9:24:29,  1.83s/it]
  3%|▎         | 604/19090 [17:52<8:37:11,  1.68s/it]
  3%|▎         | 605/19090 [17:54<8:59:29,  1.75s/it]
  3%|▎         | 606/19090 [17:57<9:51:12,  1.92s/it]
  3%|▎         | 607/19090 [17:58<8:52:12,  1.73s/it]
  3%|▎         | 608/19090 [18:00<8:49:36,  1.72s/it]
  3%|▎         | 609/19090 [18:02<9:53:48,  1.93s/it]
  3%|▎         | 610/19090 [18:03<8:54:20,  1.73s/it]
                                                     
{'loss': 0.0051, 'grad_norm': 0.7436301708221436, 'learning_rate': 1.595075955997905e-05, 'recon_loss': 0.005065561493393034, 'recon_static': 0.0036037293553818016, 'recon_motion': 0.0014618321874877438, 'recon_norm': 2.370232766866684, 'recon_norm_std': 0.2503715332597494, 'target_norm': 1.61987144947052, 'recon_cos': 0.3434178024530411, 'gradnorm/compressor.query': 0.023695310108267705, 'gradnorm/compressor.layer0': 0.28899770732237523, 'gradnorm/compressor.layer1': 0.3111250671073991, 'gradnorm/compressor.layer2': 0.2561778496141216, 'gradnorm/compressor.layer3': 0.22956797383265917, 'gradnorm/compressor.layer4': 0.21497422504620303, 'gradnorm/compressor.layer5': 0.20148714601312986, 'gradnorm/compressor.layer6': 0.18950748646851956, 'gradnorm/compressor.layer7': 0.1765313621087781, 'gradnorm/decoder.layer0': 0.2062370614126055, 'gradnorm/decoder.layer1': 0.1793189311537935, 'gradnorm/decoder.layer2': 0.1587478302634639, 'gradnorm/max_param': 0.253758043050766, 'epoch': 0.32}

  3%|▎         | 610/19090 [18:03<8:54:20,  1.73s/it]
  3%|▎         | 611/19090 [18:05<8:36:40,  1.68s/it]
  3%|▎         | 612/19090 [18:07<9:43:01,  1.89s/it]
  3%|▎         | 613/19090 [18:09<9:25:05,  1.83s/it]
  3%|▎         | 614/19090 [18:10<9:01:22,  1.76s/it]
  3%|▎         | 615/19090 [18:12<8:48:31,  1.72s/it]
  3%|▎         | 616/19090 [18:14<9:12:39,  1.79s/it]
  3%|▎         | 617/19090 [18:16<8:56:02,  1.74s/it]
  3%|▎         | 618/19090 [18:17<8:41:55,  1.70s/it]
  3%|▎         | 619/19090 [18:19<9:09:19,  1.78s/it]
  3%|▎         | 620/19090 [18:21<9:30:30,  1.85s/it]
                                                     
{'loss': 0.005, 'grad_norm': 0.7921046018600464, 'learning_rate': 1.6212676794133055e-05, 'recon_loss': 0.0049922815291211006, 'recon_static': 0.003552431729622185, 'recon_motion': 0.0014398499217350036, 'recon_norm': 2.3582712411880493, 'recon_norm_std': 0.24689044877886773, 'target_norm': 1.6187352180480956, 'recon_cos': 0.34776159673929213, 'gradnorm/compressor.query': 0.025787012673557617, 'gradnorm/compressor.layer0': 0.30754593523530505, 'gradnorm/compressor.layer1': 0.32692999307892234, 'gradnorm/compressor.layer2': 0.272754961743141, 'gradnorm/compressor.layer3': 0.24410720596468136, 'gradnorm/compressor.layer4': 0.23031710630440425, 'gradnorm/compressor.layer5': 0.2145973960580055, 'gradnorm/compressor.layer6': 0.20296705047036842, 'gradnorm/compressor.layer7': 0.18947779111448515, 'gradnorm/decoder.layer0': 0.22007872440905155, 'gradnorm/decoder.layer1': 0.1918495261339823, 'gradnorm/decoder.layer2': 0.17284988143879315, 'gradnorm/max_param': 0.2680850625038147, 'epoch': 0.32}

  3%|▎         | 620/19090 [18:21<9:30:30,  1.85s/it]
  3%|▎         | 621/19090 [18:23<9:00:10,  1.75s/it]
  3%|▎         | 622/19090 [18:25<9:16:22,  1.81s/it]
  3%|▎         | 623/19090 [18:26<8:59:57,  1.75s/it]
  3%|▎         | 624/19090 [18:28<8:51:35,  1.73s/it]
  3%|▎         | 625/19090 [18:30<9:24:28,  1.83s/it]
  3%|▎         | 626/19090 [18:32<9:08:43,  1.78s/it]
  3%|▎         | 627/19090 [18:33<8:37:41,  1.68s/it]
  3%|▎         | 628/19090 [18:35<9:26:22,  1.84s/it]
  3%|▎         | 629/19090 [18:37<9:05:18,  1.77s/it]
  3%|▎         | 630/19090 [18:39<9:04:39,  1.77s/it]
                                                     
{'loss': 0.0055, 'grad_norm': 0.7063626646995544, 'learning_rate': 1.6474594028287062e-05, 'recon_loss': 0.005509793607052416, 'recon_static': 0.004066412633983418, 'recon_motion': 0.001443381022545509, 'recon_norm': 2.479669612646103, 'recon_norm_std': 0.23120535984635354, 'target_norm': 1.6205312788486481, 'recon_cos': 0.3304882079362869, 'gradnorm/compressor.query': 0.022812798013112298, 'gradnorm/compressor.layer0': 0.2804046483225277, 'gradnorm/compressor.layer1': 0.29010025817365837, 'gradnorm/compressor.layer2': 0.2421770478115843, 'gradnorm/compressor.layer3': 0.22095908454771157, 'gradnorm/compressor.layer4': 0.2056449595072002, 'gradnorm/compressor.layer5': 0.189368248636617, 'gradnorm/compressor.layer6': 0.1782414152123221, 'gradnorm/compressor.layer7': 0.165878014471724, 'gradnorm/decoder.layer0': 0.192687775615083, 'gradnorm/decoder.layer1': 0.1724907779042526, 'gradnorm/decoder.layer2': 0.1544435564414355, 'gradnorm/max_param': 0.2453048974275589, 'epoch': 0.33}

  3%|▎         | 630/19090 [18:39<9:04:39,  1.77s/it]
  3%|▎         | 631/19090 [18:40<8:47:12,  1.71s/it]
  3%|▎         | 632/19090 [18:42<8:56:40,  1.74s/it]
  3%|▎         | 633/19090 [18:44<9:11:42,  1.79s/it]
  3%|▎         | 634/19090 [18:46<9:07:30,  1.78s/it]
  3%|▎         | 635/19090 [18:48<9:07:26,  1.78s/it]
  3%|▎         | 636/19090 [18:49<9:02:57,  1.77s/it]
  3%|▎         | 637/19090 [18:51<8:24:02,  1.64s/it]
  3%|▎         | 638/19090 [18:53<9:03:31,  1.77s/it]
  3%|▎         | 639/19090 [18:55<9:02:13,  1.76s/it]
  3%|▎         | 640/19090 [18:56<8:43:24,  1.70s/it]
                                                     
{'loss': 0.0043, 'grad_norm': 0.46233993768692017, 'learning_rate': 1.6736511262441068e-05, 'recon_loss': 0.004300809430424124, 'recon_static': 0.002876702678622678, 'recon_motion': 0.00142410661210306, 'recon_norm': 2.1787264943122864, 'recon_norm_std': 0.25395271368324757, 'target_norm': 1.6213187634944917, 'recon_cos': 0.3756864756345749, 'gradnorm/compressor.query': 0.014096508568483906, 'gradnorm/compressor.layer0': 0.18149476706761297, 'gradnorm/compressor.layer1': 0.19064615124472484, 'gradnorm/compressor.layer2': 0.15755339233127885, 'gradnorm/compressor.layer3': 0.14113331028913592, 'gradnorm/compressor.layer4': 0.13307448101306196, 'gradnorm/compressor.layer5': 0.12406832716154295, 'gradnorm/compressor.layer6': 0.11784713188671019, 'gradnorm/compressor.layer7': 0.11089295530591844, 'gradnorm/decoder.layer0': 0.12958763396643574, 'gradnorm/decoder.layer1': 0.11416904963041954, 'gradnorm/decoder.layer2': 0.10206394019456892, 'gradnorm/max_param': 0.16010494530200958, 'epoch': 0.34}

  3%|▎         | 640/19090 [18:56<8:43:24,  1.70s/it]
  3%|▎         | 641/19090 [18:58<9:23:17,  1.83s/it]
  3%|▎         | 642/19090 [19:00<8:33:09,  1.67s/it]
  3%|▎         | 643/19090 [19:01<8:54:05,  1.74s/it]
  3%|▎         | 644/19090 [19:04<9:50:31,  1.92s/it]
  3%|▎         | 645/19090 [19:05<8:51:53,  1.73s/it]
  3%|▎         | 646/19090 [19:07<8:57:12,  1.75s/it]
  3%|▎         | 647/19090 [19:09<9:22:07,  1.83s/it]
  3%|▎         | 648/19090 [19:10<8:34:39,  1.67s/it]
  3%|▎         | 649/19090 [19:12<8:56:13,  1.74s/it]
  3%|▎         | 650/19090 [19:14<9:23:33,  1.83s/it]
                                                     
{'loss': 0.004, 'grad_norm': 0.4125291109085083, 'learning_rate': 1.6998428496595074e-05, 'recon_loss': 0.0040081208455376325, 'recon_static': 0.002594193781260401, 'recon_motion': 0.0014139270759187638, 'recon_norm': 2.10166939496994, 'recon_norm_std': 0.25638961531221865, 'target_norm': 1.6216048330068589, 'recon_cos': 0.3909804843366146, 'gradnorm/compressor.query': 0.012316974610332057, 'gradnorm/compressor.layer0': 0.15482291353767208, 'gradnorm/compressor.layer1': 0.1691320577922939, 'gradnorm/compressor.layer2': 0.1406294967674148, 'gradnorm/compressor.layer3': 0.1283976245038528, 'gradnorm/compressor.layer4': 0.1199606635457932, 'gradnorm/compressor.layer5': 0.11269131327760823, 'gradnorm/compressor.layer6': 0.10646898548871626, 'gradnorm/compressor.layer7': 0.10077981873907983, 'gradnorm/decoder.layer0': 0.11674713204606563, 'gradnorm/decoder.layer1': 0.102288758085365, 'gradnorm/decoder.layer2': 0.09227762746628018, 'gradnorm/max_param': 0.1361839324235916, 'epoch': 0.34}

  3%|▎         | 650/19090 [19:14<9:23:33,  1.83s/it]
  3%|▎         | 651/19090 [19:15<8:33:36,  1.67s/it]
  3%|▎         | 652/19090 [19:17<9:09:26,  1.79s/it]
  3%|▎         | 653/19090 [19:19<9:09:22,  1.79s/it]
  3%|▎         | 654/19090 [19:21<8:30:32,  1.66s/it]
  3%|▎         | 655/19090 [19:23<9:09:29,  1.79s/it]
  3%|▎         | 656/19090 [19:25<9:19:42,  1.82s/it]
  3%|▎         | 657/19090 [19:26<8:30:29,  1.66s/it]
  3%|▎         | 658/19090 [19:28<9:19:52,  1.82s/it]
  3%|▎         | 659/19090 [19:30<8:52:08,  1.73s/it]
  3%|▎         | 660/19090 [19:31<8:13:16,  1.61s/it]
                                                     
{'loss': 0.0047, 'grad_norm': 1.0147236585617065, 'learning_rate': 1.7260345730749084e-05, 'recon_loss': 0.004700293327914551, 'recon_static': 0.003300645365379751, 'recon_motion': 0.0013996477442560717, 'recon_norm': 2.2792086005210876, 'recon_norm_std': 0.23618910349905492, 'target_norm': 1.6226333379745483, 'recon_cos': 0.3633159019052982, 'gradnorm/compressor.query': 0.03083650169087169, 'gradnorm/compressor.layer0': 0.3936160586381266, 'gradnorm/compressor.layer1': 0.41919714910055567, 'gradnorm/compressor.layer2': 0.3428539062120871, 'gradnorm/compressor.layer3': 0.31068098423382856, 'gradnorm/compressor.layer4': 0.2959150303383864, 'gradnorm/compressor.layer5': 0.27587925733695856, 'gradnorm/compressor.layer6': 0.2591434715052558, 'gradnorm/compressor.layer7': 0.24323496876289616, 'gradnorm/decoder.layer0': 0.2830670786340111, 'gradnorm/decoder.layer1': 0.2502385134180571, 'gradnorm/decoder.layer2': 0.22636660788908672, 'gradnorm/max_param': 0.34496375918388367, 'epoch': 0.35}

  3%|▎         | 660/19090 [19:31<8:13:16,  1.61s/it]
  3%|▎         | 661/19090 [19:34<9:40:33,  1.89s/it]
  3%|▎         | 662/19090 [19:35<9:00:33,  1.76s/it]
  3%|▎         | 663/19090 [19:36<8:25:28,  1.65s/it]
  3%|▎         | 664/19090 [19:39<9:36:39,  1.88s/it]
  3%|▎         | 665/19090 [19:40<9:03:21,  1.77s/it]
  3%|▎         | 666/19090 [19:42<8:43:05,  1.70s/it]
  3%|▎         | 667/19090 [19:44<9:50:12,  1.92s/it]
  3%|▎         | 668/19090 [19:46<8:51:25,  1.73s/it]
  4%|▎         | 669/19090 [19:47<8:30:37,  1.66s/it]
  4%|▎         | 670/19090 [19:49<9:39:56,  1.89s/it]
                                                     
{'loss': 0.0059, 'grad_norm': 1.0338674783706665, 'learning_rate': 1.752226296490309e-05, 'recon_loss': 0.005876747716683894, 'recon_static': 0.004481809242861345, 'recon_motion': 0.0013949383865110576, 'recon_norm': 2.567687749862671, 'recon_norm_std': 0.206332154199481, 'target_norm': 1.6223618596792222, 'recon_cos': 0.3194658160209656, 'gradnorm/compressor.query': 0.0308936107231634, 'gradnorm/compressor.layer0': 0.3973898859274783, 'gradnorm/compressor.layer1': 0.4188379003305133, 'gradnorm/compressor.layer2': 0.35068715581569393, 'gradnorm/compressor.layer3': 0.32375874115399783, 'gradnorm/compressor.layer4': 0.30414057820062795, 'gradnorm/compressor.layer5': 0.28085643994916004, 'gradnorm/compressor.layer6': 0.26564209546445694, 'gradnorm/compressor.layer7': 0.2501083824541725, 'gradnorm/decoder.layer0': 0.29045361777498824, 'gradnorm/decoder.layer1': 0.25387235561967675, 'gradnorm/decoder.layer2': 0.23122756307886622, 'gradnorm/max_param': 0.3501526713371277, 'epoch': 0.35}

  4%|▎         | 670/19090 [19:49<9:39:56,  1.89s/it]
  4%|▎         | 671/19090 [19:51<8:53:28,  1.74s/it]
  4%|▎         | 672/19090 [19:53<8:55:39,  1.74s/it]
  4%|▎         | 673/19090 [19:55<9:52:42,  1.93s/it]
  4%|▎         | 674/19090 [19:56<9:02:36,  1.77s/it]
  4%|▎         | 675/19090 [19:58<8:42:52,  1.70s/it]
  4%|▎         | 676/19090 [20:00<9:29:10,  1.85s/it]
  4%|▎         | 677/19090 [20:02<8:59:40,  1.76s/it]
  4%|▎         | 678/19090 [20:03<9:06:54,  1.78s/it]
  4%|▎         | 679/19090 [20:05<9:04:58,  1.78s/it]
  4%|▎         | 680/19090 [20:07<9:07:23,  1.78s/it]
                                                     
{'loss': 0.0051, 'grad_norm': 0.7056247591972351, 'learning_rate': 1.77841801990571e-05, 'recon_loss': 0.005079988541547209, 'recon_static': 0.003708338941214606, 'recon_motion': 0.0013716497051063925, 'recon_norm': 2.381274914741516, 'recon_norm_std': 0.21851951591670513, 'target_norm': 1.618733423948288, 'recon_cos': 0.34465928450226785, 'gradnorm/compressor.query': 0.020019026784821224, 'gradnorm/compressor.layer0': 0.26869849776099003, 'gradnorm/compressor.layer1': 0.2897119957814781, 'gradnorm/compressor.layer2': 0.24158940758809866, 'gradnorm/compressor.layer3': 0.21942143765211508, 'gradnorm/compressor.layer4': 0.20497970944391639, 'gradnorm/compressor.layer5': 0.19255329459478968, 'gradnorm/compressor.layer6': 0.17983623237043786, 'gradnorm/compressor.layer7': 0.16876323899683068, 'gradnorm/decoder.layer0': 0.19606518670376308, 'gradnorm/decoder.layer1': 0.174467021723473, 'gradnorm/decoder.layer2': 0.16145135431177957, 'gradnorm/max_param': 0.23728185892105103, 'epoch': 0.36}

  4%|▎         | 680/19090 [20:07<9:07:23,  1.78s/it]
  4%|▎         | 681/19090 [20:09<8:55:15,  1.74s/it]
  4%|▎         | 682/19090 [20:10<8:59:15,  1.76s/it]
  4%|▎         | 683/19090 [20:12<9:04:55,  1.78s/it]
  4%|▎         | 684/19090 [20:14<9:16:34,  1.81s/it]
  4%|▎         | 685/19090 [20:16<8:45:17,  1.71s/it]
  4%|▎         | 686/19090 [20:18<8:54:09,  1.74s/it]
  4%|▎         | 687/19090 [20:20<9:21:55,  1.83s/it]
  4%|▎         | 688/19090 [20:21<8:38:30,  1.69s/it]
  4%|▎         | 689/19090 [20:23<8:37:32,  1.69s/it]
  4%|▎         | 690/19090 [20:25<9:46:40,  1.91s/it]
                                                     
{'loss': 0.0044, 'grad_norm': 0.6102712750434875, 'learning_rate': 1.8046097433211106e-05, 'recon_loss': 0.004384393850341439, 'recon_static': 0.0030226159607991574, 'recon_motion': 0.0013617779652122409, 'recon_norm': 2.2135357916355134, 'recon_norm_std': 0.226913283765316, 'target_norm': 1.6187241673469543, 'recon_cos': 0.3740746721625328, 'gradnorm/compressor.query': 0.01727981649911914, 'gradnorm/compressor.layer0': 0.2322923636033989, 'gradnorm/compressor.layer1': 0.24839470722716744, 'gradnorm/compressor.layer2': 0.20738583914451736, 'gradnorm/compressor.layer3': 0.18954121616134165, 'gradnorm/compressor.layer4': 0.1767197562711012, 'gradnorm/compressor.layer5': 0.16592997978432925, 'gradnorm/compressor.layer6': 0.15619353823955598, 'gradnorm/compressor.layer7': 0.1476774520890497, 'gradnorm/decoder.layer0': 0.17170518917898386, 'gradnorm/decoder.layer1': 0.15298314895699638, 'gradnorm/decoder.layer2': 0.14029900859804004, 'gradnorm/max_param': 0.20400752127170563, 'epoch': 0.36}

  4%|▎         | 690/19090 [20:25<9:46:40,  1.91s/it]
  4%|▎         | 691/19090 [20:26<8:49:29,  1.73s/it]
  4%|▎         | 692/19090 [20:28<8:56:47,  1.75s/it]
  4%|▎         | 693/19090 [20:30<9:44:44,  1.91s/it]
  4%|▎         | 694/19090 [20:32<8:47:42,  1.72s/it]
  4%|▎         | 695/19090 [20:33<8:52:30,  1.74s/it]
  4%|▎         | 696/19090 [20:35<9:14:26,  1.81s/it]
  4%|▎         | 697/19090 [20:37<8:26:34,  1.65s/it]
  4%|▎         | 698/19090 [20:39<9:18:01,  1.82s/it]
  4%|▎         | 699/19090 [20:41<9:28:41,  1.86s/it]
  4%|▎         | 700/19090 [20:42<8:37:00,  1.69s/it]
                                                     
{'loss': 0.004, 'grad_norm': 0.530336856842041, 'learning_rate': 1.8308014667365116e-05, 'recon_loss': 0.003988809656584635, 'recon_static': 0.0026337967021390794, 'recon_motion': 0.0013550129893701523, 'recon_norm': 2.1042532682418824, 'recon_norm_std': 0.23311429657042027, 'target_norm': 1.6212222546339035, 'recon_cos': 0.3926235310733318, 'gradnorm/compressor.query': 0.015318711399984627, 'gradnorm/compressor.layer0': 0.19787494110834303, 'gradnorm/compressor.layer1': 0.2187885948602507, 'gradnorm/compressor.layer2': 0.18216958499329292, 'gradnorm/compressor.layer3': 0.16448120774453628, 'gradnorm/compressor.layer4': 0.1533757431518253, 'gradnorm/compressor.layer5': 0.14327827103158555, 'gradnorm/compressor.layer6': 0.13529139388269068, 'gradnorm/compressor.layer7': 0.12799885750797133, 'gradnorm/decoder.layer0': 0.1491518836969801, 'gradnorm/decoder.layer1': 0.13286732495733097, 'gradnorm/decoder.layer2': 0.12296151045969882, 'gradnorm/max_param': 0.17562653124332428, 'epoch': 0.37}

  4%|▎         | 700/19090 [20:42<8:37:00,  1.69s/it]
  4%|▎         | 701/19090 [20:44<9:06:57,  1.78s/it]
  4%|▎         | 702/19090 [20:46<9:19:45,  1.83s/it]
  4%|▎         | 703/19090 [20:47<8:31:51,  1.67s/it]
  4%|▎         | 704/19090 [20:49<8:53:01,  1.74s/it]
  4%|▎         | 705/19090 [20:52<9:45:11,  1.91s/it]
  4%|▎         | 706/19090 [20:53<8:48:16,  1.72s/it]
  4%|▎         | 707/19090 [20:55<8:43:27,  1.71s/it]
  4%|▎         | 708/19090 [20:57<9:43:27,  1.90s/it]
  4%|▎         | 709/19090 [20:58<8:46:28,  1.72s/it]
  4%|▎         | 710/19090 [21:00<8:40:18,  1.70s/it]
                                                     
{'loss': 0.0045, 'grad_norm': 0.8619454503059387, 'learning_rate': 1.8569931901519122e-05, 'recon_loss': 0.004476913053076714, 'recon_static': 0.003135309449862689, 'recon_motion': 0.0013416035973932594, 'recon_norm': 2.233866494894028, 'recon_norm_std': 0.21704675815999508, 'target_norm': 1.6189791560173035, 'recon_cos': 0.3699790246784687, 'gradnorm/compressor.query': 0.02477266401519769, 'gradnorm/compressor.layer0': 0.3318517199575038, 'gradnorm/compressor.layer1': 0.361879515710304, 'gradnorm/compressor.layer2': 0.2914192166140963, 'gradnorm/compressor.layer3': 0.264617467665833, 'gradnorm/compressor.layer4': 0.24374193062018779, 'gradnorm/compressor.layer5': 0.23086077274990208, 'gradnorm/compressor.layer6': 0.21728278169513165, 'gradnorm/compressor.layer7': 0.2050900501235433, 'gradnorm/decoder.layer0': 0.24153060426155604, 'gradnorm/decoder.layer1': 0.21648185628019795, 'gradnorm/decoder.layer2': 0.1977986697172718, 'gradnorm/max_param': 0.2921433448791504, 'epoch': 0.37}

  4%|▎         | 710/19090 [21:00<8:40:18,  1.70s/it]
  4%|▎         | 711/19090 [21:02<9:09:09,  1.79s/it]
  4%|▎         | 712/19090 [21:03<8:22:35,  1.64s/it]
  4%|▎         | 713/19090 [21:05<8:58:16,  1.76s/it]
  4%|▎         | 714/19090 [21:07<9:24:22,  1.84s/it]
  4%|▎         | 715/19090 [21:09<8:33:10,  1.68s/it]
  4%|▍         | 716/19090 [21:10<8:50:07,  1.73s/it]
  4%|▍         | 717/19090 [21:13<9:31:15,  1.87s/it]
  4%|▍         | 718/19090 [21:14<8:38:17,  1.69s/it]
  4%|▍         | 719/19090 [21:16<8:54:22,  1.75s/it]
  4%|▍         | 720/19090 [21:18<9:35:20,  1.88s/it]
                                                     
{'loss': 0.0051, 'grad_norm': 0.6613144278526306, 'learning_rate': 1.883184913567313e-05, 'recon_loss': 0.00509809140348807, 'recon_static': 0.0037578201561700554, 'recon_motion': 0.0013402711279923095, 'recon_norm': 2.391470330953598, 'recon_norm_std': 0.1994759812951088, 'target_norm': 1.6218071907758713, 'recon_cos': 0.34574121683835984, 'gradnorm/compressor.query': 0.019386178158575473, 'gradnorm/compressor.layer0': 0.25089259542772524, 'gradnorm/compressor.layer1': 0.2736077868326363, 'gradnorm/compressor.layer2': 0.2244077357901835, 'gradnorm/compressor.layer3': 0.20096187371051205, 'gradnorm/compressor.layer4': 0.1907902263401074, 'gradnorm/compressor.layer5': 0.18108032316670422, 'gradnorm/compressor.layer6': 0.16917278983887998, 'gradnorm/compressor.layer7': 0.16011612668678304, 'gradnorm/decoder.layer0': 0.1848579970227755, 'gradnorm/decoder.layer1': 0.1644158021975772, 'gradnorm/decoder.layer2': 0.15412800638139335, 'gradnorm/max_param': 0.22088471055030823, 'epoch': 0.38}

  4%|▍         | 720/19090 [21:18<9:35:20,  1.88s/it]
  4%|▍         | 721/19090 [21:19<8:40:45,  1.70s/it]
  4%|▍         | 722/19090 [21:21<9:21:25,  1.83s/it]
  4%|▍         | 723/19090 [21:23<9:04:19,  1.78s/it]
  4%|▍         | 724/19090 [21:24<8:19:08,  1.63s/it]
  4%|▍         | 725/19090 [21:27<9:26:20,  1.85s/it]
  4%|▍         | 726/19090 [21:28<8:59:24,  1.76s/it]
  4%|▍         | 727/19090 [21:30<8:22:57,  1.64s/it]
  4%|▍         | 728/19090 [21:32<9:04:48,  1.78s/it]
  4%|▍         | 729/19090 [21:34<9:20:24,  1.83s/it]
  4%|▍         | 730/19090 [21:35<8:30:58,  1.67s/it]
                                                     
{'loss': 0.0043, 'grad_norm': 0.5863144993782043, 'learning_rate': 1.9093766369827135e-05, 'recon_loss': 0.00426555925514549, 'recon_static': 0.0029502216493710876, 'recon_motion': 0.0013153376174159348, 'recon_norm': 2.1814743876457214, 'recon_norm_std': 0.21327480748295785, 'target_norm': 1.6187451362609864, 'recon_cos': 0.3779761277139187, 'gradnorm/compressor.query': 0.015968110419846915, 'gradnorm/compressor.layer0': 0.21650636815248783, 'gradnorm/compressor.layer1': 0.23747544757771924, 'gradnorm/compressor.layer2': 0.2002887972765132, 'gradnorm/compressor.layer3': 0.183260056407731, 'gradnorm/compressor.layer4': 0.1709577348560178, 'gradnorm/compressor.layer5': 0.1611051246667695, 'gradnorm/compressor.layer6': 0.1499658505985576, 'gradnorm/compressor.layer7': 0.14193569093086345, 'gradnorm/decoder.layer0': 0.16638205983740895, 'gradnorm/decoder.layer1': 0.14804794618123426, 'gradnorm/decoder.layer2': 0.13828282786401233, 'gradnorm/max_param': 0.19248200953006744, 'epoch': 0.38}

  4%|▍         | 730/19090 [21:35<8:30:58,  1.67s/it]
  4%|▍         | 731/19090 [21:37<9:24:18,  1.84s/it]
  4%|▍         | 732/19090 [21:39<9:27:07,  1.85s/it]
  4%|▍         | 733/19090 [21:40<8:35:05,  1.68s/it]
  4%|▍         | 734/19090 [21:42<8:37:49,  1.69s/it]
  4%|▍         | 735/19090 [21:45<9:52:14,  1.94s/it]
  4%|▍         | 736/19090 [21:46<8:53:06,  1.74s/it]
  4%|▍         | 737/19090 [21:47<8:29:37,  1.67s/it]
  4%|▍         | 738/19090 [21:50<9:49:38,  1.93s/it]
  4%|▍         | 739/19090 [21:51<8:51:16,  1.74s/it]
  4%|▍         | 740/19090 [21:53<8:31:42,  1.67s/it]
                                                     
{'loss': 0.004, 'grad_norm': 0.6906322240829468, 'learning_rate': 1.935568360398114e-05, 'recon_loss': 0.004001736611826345, 'recon_static': 0.002695417549693957, 'recon_motion': 0.0013063190621323884, 'recon_norm': 2.1139739215373994, 'recon_norm_std': 0.21495537273585796, 'target_norm': 1.6152804225683213, 'recon_cos': 0.3926819883286953, 'gradnorm/compressor.query': 0.019749667368291608, 'gradnorm/compressor.layer0': 0.2606203601844959, 'gradnorm/compressor.layer1': 0.28380296256986925, 'gradnorm/compressor.layer2': 0.23479425760299033, 'gradnorm/compressor.layer3': 0.21601665420700936, 'gradnorm/compressor.layer4': 0.20034549011500588, 'gradnorm/compressor.layer5': 0.1871169668102918, 'gradnorm/compressor.layer6': 0.1768543022197333, 'gradnorm/compressor.layer7': 0.16484700827971663, 'gradnorm/decoder.layer0': 0.19388745185296133, 'gradnorm/decoder.layer1': 0.1708658841327462, 'gradnorm/decoder.layer2': 0.16103942939451127, 'gradnorm/max_param': 0.2309388965368271, 'epoch': 0.39}

  4%|▍         | 740/19090 [21:53<8:31:42,  1.67s/it]
  4%|▍         | 741/19090 [21:55<9:31:52,  1.87s/it]
  4%|▍         | 742/19090 [21:56<8:38:55,  1.70s/it]
  4%|▍         | 743/19090 [21:58<8:23:04,  1.65s/it]
  4%|▍         | 744/19090 [22:00<9:03:06,  1.78s/it]
  4%|▍         | 745/19090 [22:01<8:18:19,  1.63s/it]
  4%|▍         | 746/19090 [22:03<8:49:27,  1.73s/it]
  4%|▍         | 747/19090 [22:05<9:23:35,  1.84s/it]
  4%|▍         | 748/19090 [22:07<8:32:24,  1.68s/it]
  4%|▍         | 749/19090 [22:09<8:58:32,  1.76s/it]
  4%|▍         | 750/19090 [22:11<9:51:44,  1.94s/it]
                                                     
{'loss': 0.0048, 'grad_norm': 0.5113673806190491, 'learning_rate': 1.961760083813515e-05, 'recon_loss': 0.004819052817765623, 'recon_static': 0.003513512632343918, 'recon_motion': 0.0013055401534074918, 'recon_norm': 2.3164360165596007, 'recon_norm_std': 0.19550214633345603, 'target_norm': 1.6184937298297881, 'recon_cos': 0.35977688655257223, 'gradnorm/compressor.query': 0.014023281667024158, 'gradnorm/compressor.layer0': 0.189120728613428, 'gradnorm/compressor.layer1': 0.2055679828646766, 'gradnorm/compressor.layer2': 0.17194968464939764, 'gradnorm/compressor.layer3': 0.16064015994168365, 'gradnorm/compressor.layer4': 0.1486590325967382, 'gradnorm/compressor.layer5': 0.14017216968388463, 'gradnorm/compressor.layer6': 0.13307819016963254, 'gradnorm/compressor.layer7': 0.12485403986504885, 'gradnorm/decoder.layer0': 0.1461930620791552, 'gradnorm/decoder.layer1': 0.12835498904479853, 'gradnorm/decoder.layer2': 0.1225137894639735, 'gradnorm/max_param': 0.1673402637243271, 'epoch': 0.39}

  4%|▍         | 750/19090 [22:11<9:51:44,  1.94s/it]
  4%|▍         | 751/19090 [22:12<8:52:26,  1.74s/it]
  4%|▍         | 752/19090 [22:14<8:51:08,  1.74s/it]
  4%|▍         | 753/19090 [22:16<9:05:59,  1.79s/it]
  4%|▍         | 754/19090 [22:17<8:20:58,  1.64s/it]
  4%|▍         | 755/19090 [22:19<9:08:11,  1.79s/it]
  4%|▍         | 756/19090 [22:21<9:16:27,  1.82s/it]
  4%|▍         | 757/19090 [22:22<8:28:48,  1.67s/it]
  4%|▍         | 758/19090 [22:24<9:04:37,  1.78s/it]
  4%|▍         | 759/19090 [22:26<9:16:59,  1.82s/it]
  4%|▍         | 760/19090 [22:28<8:28:06,  1.66s/it]
                                                     
{'loss': 0.0041, 'grad_norm': 0.5730670094490051, 'learning_rate': 1.9879518072289157e-05, 'recon_loss': 0.004138036834774539, 'recon_static': 0.0028393773653078826, 'recon_motion': 0.001298659446183592, 'recon_norm': 2.149099987745285, 'recon_norm_std': 0.20464542731642724, 'target_norm': 1.6206265270709992, 'recon_cos': 0.3856558263301849, 'gradnorm/compressor.query': 0.014325899918405577, 'gradnorm/compressor.layer0': 0.20771653641413254, 'gradnorm/compressor.layer1': 0.23035881685707998, 'gradnorm/compressor.layer2': 0.190671757138541, 'gradnorm/compressor.layer3': 0.17927344964944547, 'gradnorm/compressor.layer4': 0.16808847108527666, 'gradnorm/compressor.layer5': 0.1577214632328118, 'gradnorm/compressor.layer6': 0.15127236257111196, 'gradnorm/compressor.layer7': 0.14206916291529836, 'gradnorm/decoder.layer0': 0.16408387680659547, 'gradnorm/decoder.layer1': 0.14615669475452678, 'gradnorm/decoder.layer2': 0.13788548992202396, 'gradnorm/max_param': 0.1836485117673874, 'epoch': 0.4}

  4%|▍         | 760/19090 [22:28<8:28:06,  1.66s/it]
  4%|▍         | 761/19090 [22:30<9:05:41,  1.79s/it]
  4%|▍         | 762/19090 [22:31<8:46:59,  1.73s/it]
  4%|▍         | 763/19090 [22:33<8:06:47,  1.59s/it]
  4%|▍         | 764/19090 [22:35<8:52:22,  1.74s/it]
  4%|▍         | 765/19090 [22:37<9:15:37,  1.82s/it]
  4%|▍         | 766/19090 [22:38<8:27:09,  1.66s/it]
  4%|▍         | 767/19090 [22:40<9:18:52,  1.83s/it]
  4%|▍         | 768/19090 [22:42<9:11:19,  1.81s/it]
  4%|▍         | 769/19090 [22:44<8:46:30,  1.72s/it]
  4%|▍         | 770/19090 [22:45<8:56:30,  1.76s/it]
                                                     
{'loss': 0.0043, 'grad_norm': 0.6620303988456726, 'learning_rate': 2.0141435306443167e-05, 'recon_loss': 0.004257889091968536, 'recon_static': 0.0029818051378242673, 'recon_motion': 0.0012760840239934624, 'recon_norm': 2.187349909543991, 'recon_norm_std': 0.19855579398572445, 'target_norm': 1.6189071744680406, 'recon_cos': 0.3800992503762245, 'gradnorm/compressor.query': 0.017670737615234557, 'gradnorm/compressor.layer0': 0.2492692295359757, 'gradnorm/compressor.layer1': 0.2756969155455206, 'gradnorm/compressor.layer2': 0.22616137149120003, 'gradnorm/compressor.layer3': 0.20416055401747166, 'gradnorm/compressor.layer4': 0.18977799041966395, 'gradnorm/compressor.layer5': 0.17757912027572667, 'gradnorm/compressor.layer6': 0.16728118517650115, 'gradnorm/compressor.layer7': 0.15654272435217406, 'gradnorm/decoder.layer0': 0.18544156951154628, 'gradnorm/decoder.layer1': 0.1660947779285421, 'gradnorm/decoder.layer2': 0.1580597557536323, 'gradnorm/max_param': 0.21987316012382507, 'epoch': 0.4}

  4%|▍         | 770/19090 [22:45<8:56:30,  1.76s/it]
  4%|▍         | 771/19090 [22:47<9:01:59,  1.78s/it]
  4%|▍         | 772/19090 [22:48<8:17:13,  1.63s/it]
  4%|▍         | 773/19090 [22:50<8:49:16,  1.73s/it]
  4%|▍         | 774/19090 [22:52<8:57:26,  1.76s/it]
  4%|▍         | 775/19090 [22:54<8:38:44,  1.70s/it]
  4%|▍         | 776/19090 [22:56<8:43:31,  1.72s/it]
  4%|▍         | 777/19090 [22:58<9:10:06,  1.80s/it]
  4%|▍         | 778/19090 [22:59<9:04:09,  1.78s/it]
  4%|▍         | 779/19090 [23:01<8:34:20,  1.69s/it]
  4%|▍         | 780/19090 [23:03<9:37:43,  1.89s/it]
                                                     
{'loss': 0.0042, 'grad_norm': 0.6680054068565369, 'learning_rate': 2.0403352540597173e-05, 'recon_loss': 0.004247011686675251, 'recon_static': 0.002972711401525885, 'recon_motion': 0.0012743002764182166, 'recon_norm': 2.1835361182689668, 'recon_norm_std': 0.19480583481490613, 'target_norm': 1.620311725139618, 'recon_cos': 0.3813444174826145, 'gradnorm/compressor.query': 0.01914925897504242, 'gradnorm/compressor.layer0': 0.25984044539138895, 'gradnorm/compressor.layer1': 0.2776229554426952, 'gradnorm/compressor.layer2': 0.22556657419042392, 'gradnorm/compressor.layer3': 0.2012918262821255, 'gradnorm/compressor.layer4': 0.18900902795863989, 'gradnorm/compressor.layer5': 0.17781960885464407, 'gradnorm/compressor.layer6': 0.170381111795957, 'gradnorm/compressor.layer7': 0.16036126924005753, 'gradnorm/decoder.layer0': 0.18724997062326362, 'gradnorm/decoder.layer1': 0.1670490748009327, 'gradnorm/decoder.layer2': 0.15753290884653845, 'gradnorm/max_param': 0.229253888130188, 'epoch': 0.41}

  4%|▍         | 780/19090 [23:03<9:37:43,  1.89s/it]
  4%|▍         | 781/19090 [23:05<8:58:45,  1.77s/it]
  4%|▍         | 782/19090 [23:06<8:21:56,  1.64s/it]
  4%|▍         | 783/19090 [23:08<9:36:28,  1.89s/it]
  4%|▍         | 784/19090 [23:10<8:42:00,  1.71s/it]
  4%|▍         | 785/19090 [23:11<8:30:29,  1.67s/it]
  4%|▍         | 786/19090 [23:14<9:27:36,  1.86s/it]
  4%|▍         | 787/19090 [23:15<8:35:46,  1.69s/it]
  4%|▍         | 788/19090 [23:17<8:51:09,  1.74s/it]
  4%|▍         | 789/19090 [23:19<9:48:11,  1.93s/it]
  4%|▍         | 790/19090 [23:20<8:49:54,  1.74s/it]
                                                     
{'loss': 0.0048, 'grad_norm': 0.8083521723747253, 'learning_rate': 2.066526977475118e-05, 'recon_loss': 0.004801928880624473, 'recon_static': 0.003531745116924867, 'recon_motion': 0.0012701837404165417, 'recon_norm': 2.325653165578842, 'recon_norm_std': 0.1799461092799902, 'target_norm': 1.6182465553283691, 'recon_cos': 0.35692291185259817, 'gradnorm/compressor.query': 0.021706106116944932, 'gradnorm/compressor.layer0': 0.30197461682186455, 'gradnorm/compressor.layer1': 0.3328992493270347, 'gradnorm/compressor.layer2': 0.2748571599277214, 'gradnorm/compressor.layer3': 0.24713036029869986, 'gradnorm/compressor.layer4': 0.23104648591654584, 'gradnorm/compressor.layer5': 0.21804257513362793, 'gradnorm/compressor.layer6': 0.20715129718680517, 'gradnorm/compressor.layer7': 0.19482567196258613, 'gradnorm/decoder.layer0': 0.22777293224013723, 'gradnorm/decoder.layer1': 0.20375821560695884, 'gradnorm/decoder.layer2': 0.19783823546462528, 'gradnorm/max_param': 0.2656075060367584, 'epoch': 0.41}

  4%|▍         | 790/19090 [23:20<8:49:54,  1.74s/it]
  4%|▍         | 791/19090 [23:22<8:23:15,  1.65s/it]
  4%|▍         | 792/19090 [23:24<9:17:24,  1.83s/it]
  4%|▍         | 793/19090 [23:25<8:28:11,  1.67s/it]
  4%|▍         | 794/19090 [23:27<8:31:55,  1.68s/it]
  4%|▍         | 795/19090 [23:29<9:27:01,  1.86s/it]
  4%|▍         | 796/19090 [23:31<8:34:42,  1.69s/it]
  4%|▍         | 797/19090 [23:32<8:29:45,  1.67s/it]
  4%|▍         | 798/19090 [23:35<9:49:56,  1.94s/it]
  4%|▍         | 799/19090 [23:36<8:50:48,  1.74s/it]
  4%|▍         | 800/19090 [23:38<8:24:26,  1.65s/it]
                                                     
{'loss': 0.0049, 'grad_norm': 0.5934702157974243, 'learning_rate': 2.0927187008905186e-05, 'recon_loss': 0.004935350094456226, 'recon_static': 0.0036583484092261643, 'recon_motion': 0.0012770018656738102, 'recon_norm': 2.354364514350891, 'recon_norm_std': 0.1756556775420904, 'target_norm': 1.6224102258682251, 'recon_cos': 0.3520515874028206, 'gradnorm/compressor.query': 0.015285222046623002, 'gradnorm/compressor.layer0': 0.20842938394487587, 'gradnorm/compressor.layer1': 0.2445165562709167, 'gradnorm/compressor.layer2': 0.20423990334170078, 'gradnorm/compressor.layer3': 0.1852977284595879, 'gradnorm/compressor.layer4': 0.17293385524135002, 'gradnorm/compressor.layer5': 0.16041600283468876, 'gradnorm/compressor.layer6': 0.1520261204296221, 'gradnorm/compressor.layer7': 0.14369659805407217, 'gradnorm/decoder.layer0': 0.17021002398794807, 'gradnorm/decoder.layer1': 0.15109909681309658, 'gradnorm/decoder.layer2': 0.14659628725712473, 'gradnorm/max_param': 0.18397192656993866, 'epoch': 0.42}

  4%|▍         | 800/19090 [23:38<8:24:26,  1.65s/it]
  4%|▍         | 801/19090 [23:40<9:10:16,  1.81s/it]
  4%|▍         | 802/19090 [23:41<8:23:09,  1.65s/it]
  4%|▍         | 803/19090 [23:43<8:32:46,  1.68s/it]
  4%|▍         | 804/19090 [23:45<9:27:43,  1.86s/it]
  4%|▍         | 805/19090 [23:46<8:35:06,  1.69s/it]
  4%|▍         | 806/19090 [23:48<8:14:09,  1.62s/it]
  4%|▍         | 807/19090 [23:50<9:19:49,  1.84s/it]
  4%|▍         | 808/19090 [23:51<8:29:33,  1.67s/it]
  4%|▍         | 809/19090 [23:53<8:46:52,  1.73s/it]
  4%|▍         | 810/19090 [23:56<9:33:10,  1.88s/it]
                                                     
{'loss': 0.0039, 'grad_norm': 0.5861880779266357, 'learning_rate': 2.1189104243059192e-05, 'recon_loss': 0.003942503040889278, 'recon_static': 0.002685963350813836, 'recon_motion': 0.0012565397686557845, 'recon_norm': 2.1046468198299406, 'recon_norm_std': 0.1906812261790037, 'target_norm': 1.6204991459846496, 'recon_cos': 0.3966280668973923, 'gradnorm/compressor.query': 0.01637663175862414, 'gradnorm/compressor.layer0': 0.21743368151700843, 'gradnorm/compressor.layer1': 0.24204290595902705, 'gradnorm/compressor.layer2': 0.20071088879077492, 'gradnorm/compressor.layer3': 0.1807655770968396, 'gradnorm/compressor.layer4': 0.16713400167526188, 'gradnorm/compressor.layer5': 0.15629690777481683, 'gradnorm/compressor.layer6': 0.14932878092451812, 'gradnorm/compressor.layer7': 0.14141821336457686, 'gradnorm/decoder.layer0': 0.16499373699533396, 'gradnorm/decoder.layer1': 0.14840897486810425, 'gradnorm/decoder.layer2': 0.14358349595385395, 'gradnorm/max_param': 0.19097772240638733, 'epoch': 0.42}

  4%|▍         | 810/19090 [23:56<9:33:10,  1.88s/it]
  4%|▍         | 811/19090 [23:57<8:38:54,  1.70s/it]
  4%|▍         | 812/19090 [23:59<8:41:05,  1.71s/it]
  4%|▍         | 813/19090 [24:01<9:23:13,  1.85s/it]
  4%|▍         | 814/19090 [24:02<8:32:30,  1.68s/it]
  4%|▍         | 815/19090 [24:04<8:19:12,  1.64s/it]
  4%|▍         | 816/19090 [24:06<9:04:26,  1.79s/it]
  4%|▍         | 817/19090 [24:07<8:36:09,  1.69s/it]
  4%|▍         | 818/19090 [24:09<8:51:44,  1.75s/it]
  4%|▍         | 819/19090 [24:11<8:46:41,  1.73s/it]
  4%|▍         | 820/19090 [24:13<9:05:39,  1.79s/it]
                                                     
{'loss': 0.004, 'grad_norm': 0.5459645390510559, 'learning_rate': 2.14510214772132e-05, 'recon_loss': 0.003965342236915604, 'recon_static': 0.0027129630267154425, 'recon_motion': 0.0012523791432613508, 'recon_norm': 2.1106323540210723, 'recon_norm_std': 0.18713828213512898, 'target_norm': 1.6225243717432023, 'recon_cos': 0.3956740289926529, 'gradnorm/compressor.query': 0.015373566010922624, 'gradnorm/compressor.layer0': 0.2004768160195567, 'gradnorm/compressor.layer1': 0.22260779974642422, 'gradnorm/compressor.layer2': 0.18781797943754353, 'gradnorm/compressor.layer3': 0.16966941319102516, 'gradnorm/compressor.layer4': 0.1576170780696749, 'gradnorm/compressor.layer5': 0.14723691913495987, 'gradnorm/compressor.layer6': 0.13996272300754575, 'gradnorm/compressor.layer7': 0.1305625952062447, 'gradnorm/decoder.layer0': 0.15299899453957858, 'gradnorm/decoder.layer1': 0.13886457137636057, 'gradnorm/decoder.layer2': 0.1348420009632213, 'gradnorm/max_param': 0.17638950049877167, 'epoch': 0.43}

  4%|▍         | 820/19090 [24:13<9:05:39,  1.79s/it]
  4%|▍         | 821/19090 [24:14<8:19:48,  1.64s/it]
  4%|▍         | 822/19090 [24:16<9:15:01,  1.82s/it]
  4%|▍         | 823/19090 [24:18<9:25:27,  1.86s/it]
  4%|▍         | 824/19090 [24:19<8:33:24,  1.69s/it]
  4%|▍         | 825/19090 [24:21<9:02:55,  1.78s/it]
  4%|▍         | 826/19090 [24:23<9:04:58,  1.79s/it]
  4%|▍         | 827/19090 [24:25<8:26:57,  1.67s/it]
  4%|▍         | 828/19090 [24:27<9:08:57,  1.80s/it]
  4%|▍         | 829/19090 [24:29<9:22:38,  1.85s/it]
  4%|▍         | 830/19090 [24:30<8:31:45,  1.68s/it]
                                                     
{'loss': 0.0051, 'grad_norm': 0.8291902542114258, 'learning_rate': 2.1712938711367208e-05, 'recon_loss': 0.005103173397947103, 'recon_static': 0.003862583648879081, 'recon_motion': 0.0012405895977281034, 'recon_norm': 2.3993375062942506, 'recon_norm_std': 0.16356462500989438, 'target_norm': 1.6171759217977524, 'recon_cos': 0.34620783403515815, 'gradnorm/compressor.query': 0.021859506166167195, 'gradnorm/compressor.layer0': 0.3004177299543303, 'gradnorm/compressor.layer1': 0.33818383256780976, 'gradnorm/compressor.layer2': 0.28474337591633464, 'gradnorm/compressor.layer3': 0.2545480832176671, 'gradnorm/compressor.layer4': 0.23851440310672833, 'gradnorm/compressor.layer5': 0.22476758775062913, 'gradnorm/compressor.layer6': 0.2111810673171183, 'gradnorm/compressor.layer7': 0.20067031148128997, 'gradnorm/decoder.layer0': 0.2349246097680805, 'gradnorm/decoder.layer1': 0.21381386218379989, 'gradnorm/decoder.layer2': 0.2083373436939168, 'gradnorm/max_param': 0.2664943039417267, 'epoch': 0.43}

  4%|▍         | 830/19090 [24:30<8:31:45,  1.68s/it]
  4%|▍         | 831/19090 [24:32<8:47:21,  1.73s/it]
  4%|▍         | 832/19090 [24:34<9:16:30,  1.83s/it]
  4%|▍         | 833/19090 [24:35<8:34:35,  1.69s/it]
  4%|▍         | 834/19090 [24:37<8:54:27,  1.76s/it]
  4%|▍         | 835/19090 [24:39<9:35:42,  1.89s/it]
  4%|▍         | 836/19090 [24:41<8:40:26,  1.71s/it]
  4%|▍         | 837/19090 [24:43<9:14:31,  1.82s/it]
  4%|▍         | 838/19090 [24:44<8:44:56,  1.73s/it]
  4%|▍         | 839/19090 [24:46<8:04:56,  1.59s/it]
  4%|▍         | 840/19090 [24:48<9:18:53,  1.84s/it]
                                                     
{'loss': 0.0044, 'grad_norm': 0.49639636278152466, 'learning_rate': 2.1974855945521218e-05, 'recon_loss': 0.004437822994077578, 'recon_static': 0.0032000447681639344, 'recon_motion': 0.0012377782870316878, 'recon_norm': 2.2338032603263853, 'recon_norm_std': 0.1725398238748312, 'target_norm': 1.6203036248683929, 'recon_cos': 0.3756293699145317, 'gradnorm/compressor.query': 0.011630240615285148, 'gradnorm/compressor.layer0': 0.17170854657919044, 'gradnorm/compressor.layer1': 0.19800473541740587, 'gradnorm/compressor.layer2': 0.16841310121796593, 'gradnorm/compressor.layer3': 0.15627999613845683, 'gradnorm/compressor.layer4': 0.1460124807187631, 'gradnorm/compressor.layer5': 0.13733652730574575, 'gradnorm/compressor.layer6': 0.12974552556961058, 'gradnorm/compressor.layer7': 0.12284888505671349, 'gradnorm/decoder.layer0': 0.14318097055051485, 'gradnorm/decoder.layer1': 0.12705857958893102, 'gradnorm/decoder.layer2': 0.12667511062752754, 'gradnorm/max_param': 0.15251502394676208, 'epoch': 0.44}

  4%|▍         | 840/19090 [24:48<9:18:53,  1.84s/it]
  4%|▍         | 841/19090 [24:50<9:01:54,  1.78s/it]
  4%|▍         | 842/19090 [24:51<8:16:56,  1.63s/it]
  4%|▍         | 843/19090 [24:53<9:18:15,  1.84s/it]
  4%|▍         | 844/19090 [24:55<9:01:45,  1.78s/it]
  4%|▍         | 845/19090 [24:56<8:16:49,  1.63s/it]
  4%|▍         | 846/19090 [24:59<9:37:16,  1.90s/it]
  4%|▍         | 847/19090 [25:00<8:58:19,  1.77s/it]
  4%|▍         | 848/19090 [25:01<8:14:23,  1.63s/it]
  4%|▍         | 849/19090 [25:04<9:12:29,  1.82s/it]
  4%|▍         | 850/19090 [25:05<8:47:42,  1.74s/it]
                                                     
{'loss': 0.0032, 'grad_norm': 0.2822414040565491, 'learning_rate': 2.2236773179675224e-05, 'recon_loss': 0.003243551729246974, 'recon_static': 0.0020094715378945693, 'recon_motion': 0.0012340800138190388, 'recon_norm': 1.9028236538171768, 'recon_norm_std': 0.1938475463539362, 'target_norm': 1.6216253399848939, 'recon_cos': 0.4397291444242001, 'gradnorm/compressor.query': 0.007357976029763356, 'gradnorm/compressor.layer0': 0.10187914893998748, 'gradnorm/compressor.layer1': 0.11503058533449288, 'gradnorm/compressor.layer2': 0.09663937431395794, 'gradnorm/compressor.layer3': 0.0872156851213591, 'gradnorm/compressor.layer4': 0.0810821211016277, 'gradnorm/compressor.layer5': 0.07602340031678695, 'gradnorm/compressor.layer6': 0.0722453565784228, 'gradnorm/compressor.layer7': 0.06790328120203523, 'gradnorm/decoder.layer0': 0.0802348974855343, 'gradnorm/decoder.layer1': 0.07235389550791912, 'gradnorm/decoder.layer2': 0.07205615744590192, 'gradnorm/max_param': 0.08986196666955948, 'epoch': 0.45}

  4%|▍         | 850/19090 [25:05<8:47:42,  1.74s/it]
  4%|▍         | 851/19090 [25:07<8:07:07,  1.60s/it]
  4%|▍         | 852/19090 [25:09<9:08:33,  1.80s/it]
  4%|▍         | 853/19090 [25:10<8:32:50,  1.69s/it]
  4%|▍         | 854/19090 [25:12<8:06:31,  1.60s/it]
  4%|▍         | 855/19090 [25:14<8:37:34,  1.70s/it]
  4%|▍         | 856/19090 [25:15<8:33:15,  1.69s/it]
  4%|▍         | 857/19090 [25:17<8:03:06,  1.59s/it]
  4%|▍         | 858/19090 [25:19<8:55:22,  1.76s/it]
  4%|▍         | 859/19090 [25:20<8:14:28,  1.63s/it]
  5%|▍         | 860/19090 [25:22<8:15:07,  1.63s/it]
                                                     
{'loss': 0.0031, 'grad_norm': 0.5640645027160645, 'learning_rate': 2.2498690413829234e-05, 'recon_loss': 0.00314503816771321, 'recon_static': 0.0019511341612087562, 'recon_motion': 0.0011939039715798572, 'recon_norm': 1.8795652598142625, 'recon_norm_std': 0.19722753763198853, 'target_norm': 1.6166459232568742, 'recon_cos': 0.4487206719815731, 'gradnorm/compressor.query': 0.01567661989905021, 'gradnorm/compressor.layer0': 0.20484060429093404, 'gradnorm/compressor.layer1': 0.23089554433957427, 'gradnorm/compressor.layer2': 0.1929847382474181, 'gradnorm/compressor.layer3': 0.1751032522555123, 'gradnorm/compressor.layer4': 0.1645661638354038, 'gradnorm/compressor.layer5': 0.15132294285083536, 'gradnorm/compressor.layer6': 0.1432033118118258, 'gradnorm/compressor.layer7': 0.1349894627348492, 'gradnorm/decoder.layer0': 0.1593619514247337, 'gradnorm/decoder.layer1': 0.1433153687791798, 'gradnorm/decoder.layer2': 0.1417773175479304, 'gradnorm/max_param': 0.17881423234939575, 'epoch': 0.45}

  5%|▍         | 860/19090 [25:22<8:15:07,  1.63s/it]
  5%|▍         | 861/19090 [25:24<9:16:36,  1.83s/it]
  5%|▍         | 862/19090 [25:25<8:26:57,  1.67s/it]
  5%|▍         | 863/19090 [25:27<8:04:02,  1.59s/it]
  5%|▍         | 864/19090 [25:29<8:49:51,  1.74s/it]
  5%|▍         | 865/19090 [25:30<8:08:31,  1.61s/it]
  5%|▍         | 866/19090 [25:32<8:07:31,  1.61s/it]
  5%|▍         | 867/19090 [25:34<8:30:53,  1.68s/it]
  5%|▍         | 868/19090 [25:35<7:54:46,  1.56s/it]
  5%|▍         | 869/19090 [25:37<8:55:07,  1.76s/it]
  5%|▍         | 870/19090 [25:38<8:14:09,  1.63s/it]
                                                     
{'loss': 0.0052, 'grad_norm': 0.9633708000183105, 'learning_rate': 2.276060764798324e-05, 'recon_loss': 0.00519180903211236, 'recon_static': 0.00397689591627568, 'recon_motion': 0.0012149130983743817, 'recon_norm': 2.4174056172370912, 'recon_norm_std': 0.15354580134153367, 'target_norm': 1.6199387550354003, 'recon_cos': 0.3441619448363781, 'gradnorm/compressor.query': 0.023772583685563013, 'gradnorm/compressor.layer0': 0.3394176289005222, 'gradnorm/compressor.layer1': 0.3965486371057447, 'gradnorm/compressor.layer2': 0.32947053233260404, 'gradnorm/compressor.layer3': 0.2966180869284339, 'gradnorm/compressor.layer4': 0.27898529652934984, 'gradnorm/compressor.layer5': 0.25930849164805575, 'gradnorm/compressor.layer6': 0.24823445263809352, 'gradnorm/compressor.layer7': 0.2348149224060662, 'gradnorm/decoder.layer0': 0.2738755531975762, 'gradnorm/decoder.layer1': 0.24965296440813153, 'gradnorm/decoder.layer2': 0.2436460903183222, 'gradnorm/max_param': 0.2979315221309662, 'epoch': 0.46}

  5%|▍         | 870/19090 [25:38<8:14:09,  1.63s/it]
  5%|▍         | 871/19090 [25:40<7:43:20,  1.53s/it]
  5%|▍         | 872/19090 [25:43<9:46:52,  1.93s/it]
  5%|▍         | 873/19090 [25:44<8:47:47,  1.74s/it]
  5%|▍         | 874/19090 [25:45<8:06:45,  1.60s/it]
  5%|▍         | 875/19090 [25:48<9:34:23,  1.89s/it]
  5%|▍         | 876/19090 [25:49<8:39:15,  1.71s/it]
  5%|▍         | 877/19090 [25:50<8:11:19,  1.62s/it]
  5%|▍         | 878/19090 [25:53<9:28:48,  1.87s/it]
  5%|▍         | 879/19090 [25:54<8:40:02,  1.71s/it]
  5%|▍         | 880/19090 [25:56<8:08:27,  1.61s/it]
                                                     
{'loss': 0.005, 'grad_norm': 0.7606122493743896, 'learning_rate': 2.3022524882137246e-05, 'recon_loss': 0.00502836313098669, 'recon_static': 0.00381359857856296, 'recon_motion': 0.0012147645436925813, 'recon_norm': 2.387856423854828, 'recon_norm_std': 0.15182085298001766, 'target_norm': 1.6205290585756302, 'recon_cos': 0.3530580297112465, 'gradnorm/compressor.query': 0.018360720874275253, 'gradnorm/compressor.layer0': 0.2850131106707265, 'gradnorm/compressor.layer1': 0.3096448285164543, 'gradnorm/compressor.layer2': 0.2578666659010192, 'gradnorm/compressor.layer3': 0.2305999640958548, 'gradnorm/compressor.layer4': 0.21788293283972063, 'gradnorm/compressor.layer5': 0.2051903553815447, 'gradnorm/compressor.layer6': 0.19259023376837592, 'gradnorm/compressor.layer7': 0.18195390859183733, 'gradnorm/decoder.layer0': 0.21396098409816494, 'gradnorm/decoder.layer1': 0.19525972607024883, 'gradnorm/decoder.layer2': 0.1942422535577863, 'gradnorm/max_param': 0.2515225410461426, 'epoch': 0.46}

  5%|▍         | 880/19090 [25:56<8:08:27,  1.61s/it]
  5%|▍         | 881/19090 [25:58<9:37:39,  1.90s/it]
  5%|▍         | 882/19090 [25:59<8:46:31,  1.74s/it]
  5%|▍         | 883/19090 [26:01<8:27:21,  1.67s/it]
  5%|▍         | 884/19090 [26:04<9:44:50,  1.93s/it]
  5%|▍         | 885/19090 [26:05<8:46:48,  1.74s/it]
  5%|▍         | 886/19090 [26:06<8:20:41,  1.65s/it]
  5%|▍         | 887/19090 [26:09<9:45:54,  1.93s/it]
  5%|▍         | 888/19090 [26:10<8:56:23,  1.77s/it]
  5%|▍         | 889/19090 [26:12<8:36:38,  1.70s/it]
  5%|▍         | 890/19090 [26:14<9:11:56,  1.82s/it]
                                                     
{'loss': 0.0047, 'grad_norm': 0.6381377577781677, 'learning_rate': 2.3284442116291252e-05, 'recon_loss': 0.004674020269885659, 'recon_static': 0.0034703933633863926, 'recon_motion': 0.0012036268220981584, 'recon_norm': 2.299701821804047, 'recon_norm_std': 0.1551931884139776, 'target_norm': 1.6195020735263825, 'recon_cos': 0.36393469125032424, 'gradnorm/compressor.query': 0.014161724381927896, 'gradnorm/compressor.layer0': 0.22698571698274, 'gradnorm/compressor.layer1': 0.25754047260913826, 'gradnorm/compressor.layer2': 0.2177828542957067, 'gradnorm/compressor.layer3': 0.194454827674777, 'gradnorm/compressor.layer4': 0.18439988437119373, 'gradnorm/compressor.layer5': 0.17058119620304618, 'gradnorm/compressor.layer6': 0.16322565660940572, 'gradnorm/compressor.layer7': 0.15416006469337204, 'gradnorm/decoder.layer0': 0.18439517718972623, 'gradnorm/decoder.layer1': 0.16790842481135115, 'gradnorm/decoder.layer2': 0.16799868826978484, 'gradnorm/max_param': 0.20159608125686646, 'epoch': 0.47}

  5%|▍         | 890/19090 [26:14<9:11:56,  1.82s/it]
  5%|▍         | 891/19090 [26:16<9:01:15,  1.78s/it]
  5%|▍         | 892/19090 [26:17<8:48:27,  1.74s/it]
  5%|▍         | 893/19090 [26:19<9:33:03,  1.89s/it]
  5%|▍         | 894/19090 [26:21<8:56:06,  1.77s/it]
  5%|▍         | 895/19090 [26:22<8:33:30,  1.69s/it]
  5%|▍         | 896/19090 [26:25<9:42:48,  1.92s/it]
  5%|▍         | 897/19090 [26:26<8:45:34,  1.73s/it]
  5%|▍         | 898/19090 [26:28<9:24:52,  1.86s/it]
  5%|▍         | 899/19090 [26:30<9:18:12,  1.84s/it]
  5%|▍         | 900/19090 [26:31<8:28:01,  1.68s/it]
                                                     
{'loss': 0.0037, 'grad_norm': 0.5045722723007202, 'learning_rate': 2.354635935044526e-05, 'recon_loss': 0.0037230325222481044, 'recon_static': 0.0025279083813074974, 'recon_motion': 0.0011951241933275015, 'recon_norm': 2.0495047867298126, 'recon_norm_std': 0.16885759979486464, 'target_norm': 1.619431671500206, 'recon_cos': 0.40951836109161377, 'gradnorm/compressor.query': 0.011919859571175322, 'gradnorm/compressor.layer0': 0.1772449693039886, 'gradnorm/compressor.layer1': 0.20153246663217733, 'gradnorm/compressor.layer2': 0.17106815630077474, 'gradnorm/compressor.layer3': 0.1553551860556508, 'gradnorm/compressor.layer4': 0.14545138116066372, 'gradnorm/compressor.layer5': 0.13850655573906742, 'gradnorm/compressor.layer6': 0.1323903330368952, 'gradnorm/compressor.layer7': 0.12382703910014557, 'gradnorm/decoder.layer0': 0.14487556092768342, 'gradnorm/decoder.layer1': 0.13214152910902746, 'gradnorm/decoder.layer2': 0.13179390555147508, 'gradnorm/max_param': 0.15601897239685059, 'epoch': 0.47}

  5%|▍         | 900/19090 [26:31<8:28:01,  1.68s/it]
  5%|▍         | 901/19090 [26:34<9:03:29,  1.79s/it]
  5%|▍         | 902/19090 [26:36<9:22:30,  1.86s/it]
  5%|▍         | 903/19090 [26:37<8:31:58,  1.69s/it]
  5%|▍         | 904/19090 [26:39<9:29:17,  1.88s/it]
  5%|▍         | 905/19090 [26:41<8:52:01,  1.76s/it]
  5%|▍         | 906/19090 [26:42<8:16:42,  1.64s/it]
  5%|▍         | 907/19090 [26:45<9:55:29,  1.97s/it]
  5%|▍         | 908/19090 [26:46<8:54:39,  1.76s/it]
  5%|▍         | 909/19090 [26:47<8:24:48,  1.67s/it]
  5%|▍         | 910/19090 [26:50<9:34:48,  1.90s/it]
                                                     
{'loss': 0.0036, 'grad_norm': 0.6309776306152344, 'learning_rate': 2.380827658459927e-05, 'recon_loss': 0.0035973799647763373, 'recon_static': 0.0024115763255394993, 'recon_motion': 0.0011858036305056885, 'recon_norm': 2.016166239976883, 'recon_norm_std': 0.16847056709229946, 'target_norm': 1.6184831738471985, 'recon_cos': 0.41655826270580293, 'gradnorm/compressor.query': 0.015008293140435619, 'gradnorm/compressor.layer0': 0.2263486362521314, 'gradnorm/compressor.layer1': 0.2552434044050259, 'gradnorm/compressor.layer2': 0.2096320215191176, 'gradnorm/compressor.layer3': 0.19441346198892315, 'gradnorm/compressor.layer4': 0.1812045184160231, 'gradnorm/compressor.layer5': 0.17396918477592868, 'gradnorm/compressor.layer6': 0.16524121358260804, 'gradnorm/compressor.layer7': 0.1535249304438466, 'gradnorm/decoder.layer0': 0.18031700409630433, 'gradnorm/decoder.layer1': 0.1628261141109729, 'gradnorm/decoder.layer2': 0.16364900957319783, 'gradnorm/max_param': 0.2011122703552246, 'epoch': 0.48}

  5%|▍         | 910/19090 [26:50<9:34:48,  1.90s/it]
  5%|▍         | 911/19090 [26:51<8:40:32,  1.72s/it]
  5%|▍         | 912/19090 [26:53<8:13:45,  1.63s/it]
  5%|▍         | 913/19090 [26:55<9:45:21,  1.93s/it]
  5%|▍         | 914/19090 [26:57<9:00:39,  1.78s/it]
  5%|▍         | 915/19090 [26:58<8:16:31,  1.64s/it]
  5%|▍         | 916/19090 [27:00<9:33:26,  1.89s/it]
  5%|▍         | 917/19090 [27:02<8:39:02,  1.71s/it]
  5%|▍         | 918/19090 [27:03<8:21:43,  1.66s/it]
  5%|▍         | 919/19090 [27:06<9:29:10,  1.88s/it]
  5%|▍         | 920/19090 [27:07<8:46:43,  1.74s/it]
                                                     
{'loss': 0.0052, 'grad_norm': 0.870644211769104, 'learning_rate': 2.4070193818753275e-05, 'recon_loss': 0.005199618684127927, 'recon_static': 0.004003255511634052, 'recon_motion': 0.0011963633121922612, 'recon_norm': 2.4302254021167755, 'recon_norm_std': 0.1404357250779867, 'target_norm': 1.6222750008106233, 'recon_cos': 0.3465157553553581, 'gradnorm/compressor.query': 0.02098681763401606, 'gradnorm/compressor.layer0': 0.3112368087194625, 'gradnorm/compressor.layer1': 0.3461173964402649, 'gradnorm/compressor.layer2': 0.2925775204538942, 'gradnorm/compressor.layer3': 0.2682511157095283, 'gradnorm/compressor.layer4': 0.25073154527312225, 'gradnorm/compressor.layer5': 0.23580329722421914, 'gradnorm/compressor.layer6': 0.2267005269767196, 'gradnorm/compressor.layer7': 0.2120567404001116, 'gradnorm/decoder.layer0': 0.2538616434675547, 'gradnorm/decoder.layer1': 0.2270757845985896, 'gradnorm/decoder.layer2': 0.22916186483005196, 'gradnorm/max_param': 0.2751009166240692, 'epoch': 0.48}

  5%|▍         | 920/19090 [27:07<8:46:43,  1.74s/it]
  5%|▍         | 921/19090 [27:08<8:13:34,  1.63s/it]
  5%|▍         | 922/19090 [27:11<9:22:38,  1.86s/it]
  5%|▍         | 923/19090 [27:12<8:50:46,  1.75s/it]
  5%|▍         | 924/19090 [27:14<8:15:52,  1.64s/it]
  5%|▍         | 925/19090 [27:16<9:39:49,  1.92s/it]
  5%|▍         | 926/19090 [27:18<8:43:27,  1.73s/it]
  5%|▍         | 927/19090 [27:19<8:04:39,  1.60s/it]
  5%|▍         | 928/19090 [27:22<9:41:36,  1.92s/it]
  5%|▍         | 929/19090 [27:23<8:48:18,  1.75s/it]
  5%|▍         | 930/19090 [27:24<8:07:39,  1.61s/it]
                                                     
{'loss': 0.0043, 'grad_norm': 0.5888200402259827, 'learning_rate': 2.4332111052907284e-05, 'recon_loss': 0.004293155722552911, 'recon_static': 0.0031002268893644213, 'recon_motion': 0.001192928821546957, 'recon_norm': 2.2035933792591096, 'recon_norm_std': 0.1506177343428135, 'target_norm': 1.6224630802869797, 'recon_cos': 0.380317884683609, 'gradnorm/compressor.query': 0.01317506311070009, 'gradnorm/compressor.layer0': 0.19411107573339514, 'gradnorm/compressor.layer1': 0.23043072111962032, 'gradnorm/compressor.layer2': 0.19896307601735086, 'gradnorm/compressor.layer3': 0.18138963524683693, 'gradnorm/compressor.layer4': 0.17205856555932686, 'gradnorm/compressor.layer5': 0.16351146917807777, 'gradnorm/compressor.layer6': 0.15608542835403252, 'gradnorm/compressor.layer7': 0.14826560733838076, 'gradnorm/decoder.layer0': 0.17275759195436077, 'gradnorm/decoder.layer1': 0.15740669191467704, 'gradnorm/decoder.layer2': 0.16112714237293715, 'gradnorm/max_param': 0.1716463565826416, 'epoch': 0.49}

  5%|▍         | 930/19090 [27:24<8:07:39,  1.61s/it]
  5%|▍         | 931/19090 [27:27<9:43:17,  1.93s/it]
  5%|▍         | 932/19090 [27:28<8:46:15,  1.74s/it]
  5%|▍         | 933/19090 [27:29<8:05:44,  1.61s/it]
  5%|▍         | 934/19090 [27:33<10:30:05,  2.08s/it]
  5%|▍         | 935/19090 [27:34<9:17:52,  1.84s/it] 
  5%|▍         | 936/19090 [27:35<8:27:28,  1.68s/it]
  5%|▍         | 937/19090 [27:38<10:05:31,  2.00s/it]
  5%|▍         | 938/19090 [27:39<9:00:44,  1.79s/it] 
  5%|▍         | 939/19090 [27:41<8:15:19,  1.64s/it]
  5%|▍         | 940/19090 [27:43<10:00:44,  1.99s/it]
                                                      
{'loss': 0.0035, 'grad_norm': 0.49534493684768677, 'learning_rate': 2.459402828706129e-05, 'recon_loss': 0.003529491415247321, 'recon_static': 0.0023588714248035103, 'recon_motion': 0.0011706200486514718, 'recon_norm': 2.0001103460788725, 'recon_norm_std': 0.1619059257209301, 'target_norm': 1.618562823534012, 'recon_cos': 0.4214186996221542, 'gradnorm/compressor.query': 0.01155767406132141, 'gradnorm/compressor.layer0': 0.17751442221022604, 'gradnorm/compressor.layer1': 0.1978940678962275, 'gradnorm/compressor.layer2': 0.16403957047765108, 'gradnorm/compressor.layer3': 0.15016448990703607, 'gradnorm/compressor.layer4': 0.14157244284240406, 'gradnorm/compressor.layer5': 0.13506411929353598, 'gradnorm/compressor.layer6': 0.12781212041840465, 'gradnorm/compressor.layer7': 0.12199141813921781, 'gradnorm/decoder.layer0': 0.1444520668973075, 'gradnorm/decoder.layer1': 0.13090234022660435, 'gradnorm/decoder.layer2': 0.13265182087831517, 'gradnorm/max_param': 0.15738926827907562, 'epoch': 0.49}

  5%|▍         | 940/19090 [27:43<10:00:44,  1.99s/it]
  5%|▍         | 941/19090 [27:45<8:58:23,  1.78s/it] 
  5%|▍         | 942/19090 [27:46<8:14:28,  1.63s/it]
  5%|▍         | 943/19090 [27:48<9:31:12,  1.89s/it]
  5%|▍         | 944/19090 [27:50<8:36:24,  1.71s/it]
  5%|▍         | 945/19090 [27:51<7:58:52,  1.58s/it]
  5%|▍         | 946/19090 [27:54<10:05:59,  2.00s/it]
  5%|▍         | 947/19090 [27:55<9:03:13,  1.80s/it] 
  5%|▍         | 948/19090 [27:57<8:17:36,  1.65s/it]
  5%|▍         | 949/19090 [27:59<9:24:28,  1.87s/it]
  5%|▍         | 950/19090 [28:00<8:32:27,  1.70s/it]
                                                     
{'loss': 0.004, 'grad_norm': 0.734473705291748, 'learning_rate': 2.4855945521215297e-05, 'recon_loss': 0.004014547349652275, 'recon_static': 0.0028355555434245614, 'recon_motion': 0.001178991692722775, 'recon_norm': 2.1300437331199644, 'recon_norm_std': 0.1501333687454462, 'target_norm': 1.6206599235534669, 'recon_cos': 0.3954513676464558, 'gradnorm/compressor.query': 0.017474867113682818, 'gradnorm/compressor.layer0': 0.2537461729888588, 'gradnorm/compressor.layer1': 0.292311197504319, 'gradnorm/compressor.layer2': 0.24649603071025014, 'gradnorm/compressor.layer3': 0.22568414045108356, 'gradnorm/compressor.layer4': 0.2143349778932698, 'gradnorm/compressor.layer5': 0.20209684035954228, 'gradnorm/compressor.layer6': 0.19261257374207721, 'gradnorm/compressor.layer7': 0.1816926697033912, 'gradnorm/decoder.layer0': 0.21343988031872427, 'gradnorm/decoder.layer1': 0.19239224251034878, 'gradnorm/decoder.layer2': 0.19533000867567168, 'gradnorm/max_param': 0.22355155646800995, 'epoch': 0.5}

  5%|▍         | 950/19090 [28:00<8:32:27,  1.70s/it]
  5%|▍         | 951/19090 [28:02<7:55:25,  1.57s/it]
  5%|▍         | 952/19090 [28:04<9:34:43,  1.90s/it]
  5%|▍         | 953/19090 [28:06<8:39:46,  1.72s/it]
  5%|▍         | 954/19090 [28:07<8:01:40,  1.59s/it]
  5%|▌         | 955/19090 [28:10<10:12:00,  2.02s/it]
  5%|▌         | 956/19090 [28:11<9:05:29,  1.80s/it] 
  5%|▌         | 957/19090 [28:12<8:19:02,  1.65s/it]
  5%|▌         | 958/19090 [28:15<9:47:34,  1.94s/it]
  5%|▌         | 959/19090 [28:16<8:48:37,  1.75s/it]
  5%|▌         | 960/19090 [28:18<8:07:26,  1.61s/it]
                                                     
{'loss': 0.0046, 'grad_norm': 0.6939722895622253, 'learning_rate': 2.5117862755369303e-05, 'recon_loss': 0.004614369833143428, 'recon_static': 0.0034459869842976333, 'recon_motion': 0.0011683828051900491, 'recon_norm': 2.285911411046982, 'recon_norm_std': 0.1401877213269472, 'target_norm': 1.6199496269226075, 'recon_cos': 0.36719044148921964, 'gradnorm/compressor.query': 0.015720233671363992, 'gradnorm/compressor.layer0': 0.23945100730469848, 'gradnorm/compressor.layer1': 0.2725968244222462, 'gradnorm/compressor.layer2': 0.23308649325476344, 'gradnorm/compressor.layer3': 0.21304061400638835, 'gradnorm/compressor.layer4': 0.20224184835814227, 'gradnorm/compressor.layer5': 0.1899743725722261, 'gradnorm/compressor.layer6': 0.1797437787033828, 'gradnorm/compressor.layer7': 0.1700530203802707, 'gradnorm/decoder.layer0': 0.20466047174057617, 'gradnorm/decoder.layer1': 0.1848980954976561, 'gradnorm/decoder.layer2': 0.1889388685091701, 'gradnorm/max_param': 0.21239608526229858, 'epoch': 0.5}

  5%|▌         | 960/19090 [28:18<8:07:26,  1.61s/it]
  5%|▌         | 961/19090 [28:21<10:07:30,  2.01s/it]
  5%|▌         | 962/19090 [28:22<9:02:23,  1.80s/it] 
  5%|▌         | 963/19090 [28:23<8:18:41,  1.65s/it]
  5%|▌         | 964/19090 [28:26<9:27:48,  1.88s/it]
  5%|▌         | 965/19090 [28:27<8:34:04,  1.70s/it]
  5%|▌         | 966/19090 [28:28<7:56:47,  1.58s/it]
  5%|▌         | 967/19090 [28:31<10:01:27,  1.99s/it]
  5%|▌         | 968/19090 [28:32<8:58:41,  1.78s/it] 
  5%|▌         | 969/19090 [28:34<8:13:43,  1.63s/it]
  5%|▌         | 970/19090 [28:36<9:38:52,  1.92s/it]
                                                     
{'loss': 0.0044, 'grad_norm': 0.7324543595314026, 'learning_rate': 2.5379779989523313e-05, 'recon_loss': 0.004437978425994515, 'recon_static': 0.003269721841206774, 'recon_motion': 0.0011682566633680834, 'recon_norm': 2.246434098482132, 'recon_norm_std': 0.140269036591053, 'target_norm': 1.6241568446159362, 'recon_cos': 0.37553356066346166, 'gradnorm/compressor.query': 0.016090277304669297, 'gradnorm/compressor.layer0': 0.24849637671444563, 'gradnorm/compressor.layer1': 0.29386234062930755, 'gradnorm/compressor.layer2': 0.2455553416978028, 'gradnorm/compressor.layer3': 0.2239224336742416, 'gradnorm/compressor.layer4': 0.2109983409453655, 'gradnorm/compressor.layer5': 0.20121332315649934, 'gradnorm/compressor.layer6': 0.18962455151227647, 'gradnorm/compressor.layer7': 0.17887419659189557, 'gradnorm/decoder.layer0': 0.214266408784354, 'gradnorm/decoder.layer1': 0.19540889268702039, 'gradnorm/decoder.layer2': 0.20157150278020544, 'gradnorm/max_param': 0.22128377854824066, 'epoch': 0.51}

  5%|▌         | 970/19090 [28:36<9:38:52,  1.92s/it]
  5%|▌         | 971/19090 [28:38<8:41:59,  1.73s/it]
  5%|▌         | 972/19090 [28:39<8:02:42,  1.60s/it]
  5%|▌         | 973/19090 [28:42<9:34:18,  1.90s/it]
  5%|▌         | 974/19090 [28:43<8:38:21,  1.72s/it]
  5%|▌         | 975/19090 [28:44<8:00:19,  1.59s/it]
  5%|▌         | 976/19090 [28:47<9:50:22,  1.96s/it]
  5%|▌         | 977/19090 [28:48<8:49:49,  1.76s/it]
  5%|▌         | 978/19090 [28:49<8:07:46,  1.62s/it]
  5%|▌         | 979/19090 [28:52<9:30:14,  1.89s/it]
  5%|▌         | 980/19090 [28:53<8:36:32,  1.71s/it]
                                                     
{'loss': 0.004, 'grad_norm': 0.5262927412986755, 'learning_rate': 2.564169722367732e-05, 'recon_loss': 0.003977036173455417, 'recon_static': 0.002822138712508604, 'recon_motion': 0.0011548975540790706, 'recon_norm': 2.1222596019506454, 'recon_norm_std': 0.14560083523392678, 'target_norm': 1.6189518213272094, 'recon_cos': 0.39616029486060145, 'gradnorm/compressor.query': 0.01097047874495171, 'gradnorm/compressor.layer0': 0.17468062319040673, 'gradnorm/compressor.layer1': 0.2120889917348288, 'gradnorm/compressor.layer2': 0.18014499225495842, 'gradnorm/compressor.layer3': 0.16391078996325592, 'gradnorm/compressor.layer4': 0.15176070344348716, 'gradnorm/compressor.layer5': 0.1429505774408268, 'gradnorm/compressor.layer6': 0.13633896667086437, 'gradnorm/compressor.layer7': 0.12922658277294308, 'gradnorm/decoder.layer0': 0.15203458029617162, 'gradnorm/decoder.layer1': 0.13916400435631773, 'gradnorm/decoder.layer2': 0.14427245012357456, 'gradnorm/max_param': 0.15507127344608307, 'epoch': 0.51}

  5%|▌         | 980/19090 [28:53<8:36:32,  1.71s/it]
  5%|▌         | 981/19090 [28:55<7:58:03,  1.58s/it]
  5%|▌         | 982/19090 [28:57<9:35:50,  1.91s/it]
  5%|▌         | 983/19090 [28:59<8:39:24,  1.72s/it]
  5%|▌         | 984/19090 [29:00<8:00:23,  1.59s/it]
  5%|▌         | 985/19090 [29:03<9:55:03,  1.97s/it]
  5%|▌         | 986/19090 [29:04<8:52:53,  1.77s/it]
  5%|▌         | 987/19090 [29:05<8:09:20,  1.62s/it]
  5%|▌         | 988/19090 [29:08<9:25:13,  1.87s/it]
  5%|▌         | 989/19090 [29:09<8:32:03,  1.70s/it]
  5%|▌         | 990/19090 [29:10<7:55:20,  1.58s/it]
                                                     
{'loss': 0.0037, 'grad_norm': 0.645161509513855, 'learning_rate': 2.5903614457831325e-05, 'recon_loss': 0.0036612671276088802, 'recon_static': 0.0025163511687424033, 'recon_motion': 0.0011449158482719213, 'recon_norm': 2.040153893828392, 'recon_norm_std': 0.1484463233500719, 'target_norm': 1.6188434481620788, 'recon_cos': 0.4142552249133587, 'gradnorm/compressor.query': 0.014662159512718551, 'gradnorm/compressor.layer0': 0.22427566450620356, 'gradnorm/compressor.layer1': 0.2531006872015569, 'gradnorm/compressor.layer2': 0.21817785610487997, 'gradnorm/compressor.layer3': 0.1979031505905217, 'gradnorm/compressor.layer4': 0.18521717292522283, 'gradnorm/compressor.layer5': 0.17518987253816512, 'gradnorm/compressor.layer6': 0.1687884015523042, 'gradnorm/compressor.layer7': 0.16033603881147232, 'gradnorm/decoder.layer0': 0.18932291845945193, 'gradnorm/decoder.layer1': 0.17033007251132484, 'gradnorm/decoder.layer2': 0.17561903992650038, 'gradnorm/max_param': 0.19875767827033997, 'epoch': 0.52}

  5%|▌         | 990/19090 [29:10<7:55:20,  1.58s/it]
  5%|▌         | 991/19090 [29:13<9:35:52,  1.91s/it]
  5%|▌         | 992/19090 [29:14<8:39:26,  1.72s/it]
  5%|▌         | 993/19090 [29:16<7:59:57,  1.59s/it]
  5%|▌         | 994/19090 [29:19<10:08:21,  2.02s/it]
  5%|▌         | 995/19090 [29:20<9:02:18,  1.80s/it] 
  5%|▌         | 996/19090 [29:21<8:15:35,  1.64s/it]
  5%|▌         | 997/19090 [29:24<9:29:57,  1.89s/it]
  5%|▌         | 998/19090 [29:25<8:35:41,  1.71s/it]
  5%|▌         | 999/19090 [29:26<7:57:20,  1.58s/it]
  5%|▌         | 1000/19090 [29:29<9:38:00,  1.92s/it]
                                                      
{'loss': 0.0046, 'grad_norm': 0.8034644722938538, 'learning_rate': 2.6165531691985335e-05, 'recon_loss': 0.004631714324932546, 'recon_static': 0.0034850470721721647, 'recon_motion': 0.0011466671741800383, 'recon_norm': 2.295436590909958, 'recon_norm_std': 0.13227651882916688, 'target_norm': 1.6212353706359863, 'recon_cos': 0.36849204525351525, 'gradnorm/compressor.query': 0.01851288028812174, 'gradnorm/compressor.layer0': 0.2807543811668555, 'gradnorm/compressor.layer1': 0.31870469865327683, 'gradnorm/compressor.layer2': 0.26887484799907935, 'gradnorm/compressor.layer3': 0.24581555885508713, 'gradnorm/compressor.layer4': 0.23139828393272224, 'gradnorm/compressor.layer5': 0.218726293096667, 'gradnorm/compressor.layer6': 0.20939086882394373, 'gradnorm/compressor.layer7': 0.19673600247685294, 'gradnorm/decoder.layer0': 0.23251345777530177, 'gradnorm/decoder.layer1': 0.21302504189460306, 'gradnorm/decoder.layer2': 0.22072524407198624, 'gradnorm/max_param': 0.24951085448265076, 'epoch': 0.52}

  5%|▌         | 1000/19090 [29:29<9:38:00,  1.92s/it]
  5%|▌         | 1001/19090 [29:32<11:53:07,  2.37s/it]
  5%|▌         | 1002/19090 [29:34<10:15:56,  2.04s/it]
  5%|▌         | 1003/19090 [29:35<9:07:55,  1.82s/it] 
  5%|▌         | 1004/19090 [29:36<8:19:28,  1.66s/it]
  5%|▌         | 1005/19090 [29:37<7:45:50,  1.55s/it]
  5%|▌         | 1006/19090 [29:39<7:21:49,  1.47s/it]
  5%|▌         | 1007/19090 [29:40<7:05:36,  1.41s/it]
  5%|▌         | 1008/19090 [29:41<6:54:16,  1.37s/it]
  5%|▌         | 1009/19090 [29:43<7:24:40,  1.48s/it]
  5%|▌         | 1010/19090 [29:44<7:08:41,  1.42s/it]
                                                      
{'loss': 0.0045, 'grad_norm': 0.6741877198219299, 'learning_rate': 2.642744892613934e-05, 'recon_loss': 0.004552368994336575, 'recon_static': 0.003406585706397891, 'recon_motion': 0.0011457833752501756, 'recon_norm': 2.2767486393451692, 'recon_norm_std': 0.13220101445913315, 'target_norm': 1.6226594597101212, 'recon_cos': 0.37124133110046387, 'gradnorm/compressor.query': 0.01593500371099167, 'gradnorm/compressor.layer0': 0.23518446376015587, 'gradnorm/compressor.layer1': 0.26411708403070294, 'gradnorm/compressor.layer2': 0.22037140214420572, 'gradnorm/compressor.layer3': 0.20323074779322337, 'gradnorm/compressor.layer4': 0.19398284349209524, 'gradnorm/compressor.layer5': 0.1841548199011457, 'gradnorm/compressor.layer6': 0.17843751888976522, 'gradnorm/compressor.layer7': 0.1672134550839279, 'gradnorm/decoder.layer0': 0.19993463617869223, 'gradnorm/decoder.layer1': 0.18113575779747837, 'gradnorm/decoder.layer2': 0.18765024285147017, 'gradnorm/max_param': 0.20839013159275055, 'epoch': 0.53}

  5%|▌         | 1010/19090 [29:44<7:08:41,  1.42s/it]
  5%|▌         | 1011/19090 [29:46<6:56:26,  1.38s/it]
  5%|▌         | 1012/19090 [29:48<8:31:17,  1.70s/it]
  5%|▌         | 1013/19090 [29:49<7:54:47,  1.58s/it]
  5%|▌         | 1014/19090 [29:51<7:28:59,  1.49s/it]
  5%|▌         | 1015/19090 [29:53<9:13:44,  1.84s/it]
  5%|▌         | 1016/19090 [29:55<8:24:10,  1.67s/it]
  5%|▌         | 1017/19090 [29:56<7:49:41,  1.56s/it]
  5%|▌         | 1018/19090 [29:58<9:06:25,  1.81s/it]
  5%|▌         | 1019/19090 [30:00<8:20:06,  1.66s/it]
  5%|▌         | 1020/19090 [30:01<7:47:15,  1.55s/it]
                                                      
{'loss': 0.0034, 'grad_norm': 0.4869479537010193, 'learning_rate': 2.668936616029335e-05, 'recon_loss': 0.0034149222890846433, 'recon_static': 0.0022799782891524956, 'recon_motion': 0.0011349440639605746, 'recon_norm': 1.9672072798013687, 'recon_norm_std': 0.14843041375279425, 'target_norm': 1.6210762530565261, 'recon_cos': 0.4317970260977745, 'gradnorm/compressor.query': 0.011084138400734326, 'gradnorm/compressor.layer0': 0.16517532524487089, 'gradnorm/compressor.layer1': 0.19186949627279865, 'gradnorm/compressor.layer2': 0.16070335841590794, 'gradnorm/compressor.layer3': 0.1498514952761133, 'gradnorm/compressor.layer4': 0.14093349976582603, 'gradnorm/compressor.layer5': 0.13209867082133955, 'gradnorm/compressor.layer6': 0.12836877439032396, 'gradnorm/compressor.layer7': 0.12106939040019524, 'gradnorm/decoder.layer0': 0.14443623151556428, 'gradnorm/decoder.layer1': 0.12982994810718818, 'gradnorm/decoder.layer2': 0.13583566342468398, 'gradnorm/max_param': 0.14639580249786377, 'epoch': 0.53}

  5%|▌         | 1020/19090 [30:01<7:47:15,  1.55s/it]
  5%|▌         | 1021/19090 [30:03<9:23:33,  1.87s/it]
  5%|▌         | 1022/19090 [30:05<8:31:33,  1.70s/it]
  5%|▌         | 1023/19090 [30:06<7:54:31,  1.58s/it]
  5%|▌         | 1024/19090 [30:09<9:55:52,  1.98s/it]
  5%|▌         | 1025/19090 [30:10<8:54:01,  1.77s/it]
  5%|▌         | 1026/19090 [30:12<8:10:33,  1.63s/it]
  5%|▌         | 1027/19090 [30:14<9:38:05,  1.92s/it]
  5%|▌         | 1028/19090 [30:15<8:40:49,  1.73s/it]
  5%|▌         | 1029/19090 [30:17<8:01:56,  1.60s/it]
  5%|▌         | 1030/19090 [30:19<9:25:23,  1.88s/it]
                                                      
{'loss': 0.0046, 'grad_norm': 0.7465716004371643, 'learning_rate': 2.6951283394447357e-05, 'recon_loss': 0.004581837181467563, 'recon_static': 0.0034452672116458416, 'recon_motion': 0.001136569958180189, 'recon_norm': 2.2779167085886, 'recon_norm_std': 0.12839566487818957, 'target_norm': 1.6205471187829972, 'recon_cos': 0.3704223521053791, 'gradnorm/compressor.query': 0.016417311873082405, 'gradnorm/compressor.layer0': 0.26062456256729116, 'gradnorm/compressor.layer1': 0.298589586281803, 'gradnorm/compressor.layer2': 0.25341654282832066, 'gradnorm/compressor.layer3': 0.22864346311089434, 'gradnorm/compressor.layer4': 0.21259779432902937, 'gradnorm/compressor.layer5': 0.1994111964125353, 'gradnorm/compressor.layer6': 0.19176640006236423, 'gradnorm/compressor.layer7': 0.1831137693431075, 'gradnorm/decoder.layer0': 0.21398200693592434, 'gradnorm/decoder.layer1': 0.19801909949800967, 'gradnorm/decoder.layer2': 0.20797431926879556, 'gradnorm/max_param': 0.23059780895709991, 'epoch': 0.54}

  5%|▌         | 1030/19090 [30:19<9:25:23,  1.88s/it]
  5%|▌         | 1031/19090 [30:21<8:32:36,  1.70s/it]
  5%|▌         | 1032/19090 [30:22<7:56:08,  1.58s/it]
  5%|▌         | 1033/19090 [30:24<9:13:58,  1.84s/it]
  5%|▌         | 1034/19090 [30:26<8:24:41,  1.68s/it]
  5%|▌         | 1035/19090 [30:27<7:50:10,  1.56s/it]
  5%|▌         | 1036/19090 [30:30<9:36:40,  1.92s/it]
  5%|▌         | 1037/19090 [30:31<8:40:09,  1.73s/it]
  5%|▌         | 1038/19090 [30:32<8:00:35,  1.60s/it]
  5%|▌         | 1039/19090 [30:35<9:28:10,  1.89s/it]
  5%|▌         | 1040/19090 [30:36<8:33:45,  1.71s/it]
                                                      
{'loss': 0.005, 'grad_norm': 0.6034703850746155, 'learning_rate': 2.7213200628601364e-05, 'recon_loss': 0.005050516169285401, 'recon_static': 0.003916873841080815, 'recon_motion': 0.0011336422932799906, 'recon_norm': 2.3929964184761046, 'recon_norm_std': 0.12237784508615732, 'target_norm': 1.621988758444786, 'recon_cos': 0.3579264871776104, 'gradnorm/compressor.query': 0.01300190920071677, 'gradnorm/compressor.layer0': 0.19864454919279645, 'gradnorm/compressor.layer1': 0.23510384086504058, 'gradnorm/compressor.layer2': 0.20169497505689032, 'gradnorm/compressor.layer3': 0.18512054453194782, 'gradnorm/compressor.layer4': 0.1761830981839037, 'gradnorm/compressor.layer5': 0.16554559477616662, 'gradnorm/compressor.layer6': 0.15840458111308126, 'gradnorm/compressor.layer7': 0.15016149430341075, 'gradnorm/decoder.layer0': 0.17821248691757058, 'gradnorm/decoder.layer1': 0.16337006767886003, 'gradnorm/decoder.layer2': 0.1725912396778344, 'gradnorm/max_param': 0.17573626339435577, 'epoch': 0.54}

  5%|▌         | 1040/19090 [30:36<8:33:45,  1.71s/it]
  5%|▌         | 1041/19090 [30:37<8:00:56,  1.60s/it]
  5%|▌         | 1042/19090 [30:40<9:29:48,  1.89s/it]
  5%|▌         | 1043/19090 [30:41<8:34:36,  1.71s/it]
  5%|▌         | 1044/19090 [30:43<7:55:59,  1.58s/it]
  5%|▌         | 1045/19090 [30:45<9:17:12,  1.85s/it]
  5%|▌         | 1046/19090 [30:46<8:26:06,  1.68s/it]
  5%|▌         | 1047/19090 [30:48<7:50:28,  1.56s/it]
  5%|▌         | 1048/19090 [30:50<9:05:58,  1.82s/it]
  5%|▌         | 1049/19090 [30:51<8:17:50,  1.66s/it]
  6%|▌         | 1050/19090 [30:53<7:44:28,  1.54s/it]
                                                      
{'loss': 0.0037, 'grad_norm': 0.5206547975540161, 'learning_rate': 2.7475117862755373e-05, 'recon_loss': 0.003660975961247459, 'recon_static': 0.002529149304609746, 'recon_motion': 0.0011318267264869065, 'recon_norm': 2.0380362540483477, 'recon_norm_std': 0.13847847133874894, 'target_norm': 1.6223209261894227, 'recon_cos': 0.41355526372790335, 'gradnorm/compressor.query': 0.011643714011890035, 'gradnorm/compressor.layer0': 0.1767855353770789, 'gradnorm/compressor.layer1': 0.20463851859529109, 'gradnorm/compressor.layer2': 0.17462953391398053, 'gradnorm/compressor.layer3': 0.15966439899338752, 'gradnorm/compressor.layer4': 0.15193177418882597, 'gradnorm/compressor.layer5': 0.1411316011674064, 'gradnorm/compressor.layer6': 0.13458332917884647, 'gradnorm/compressor.layer7': 0.12836576350700743, 'gradnorm/decoder.layer0': 0.152988286928459, 'gradnorm/decoder.layer1': 0.1387472485149197, 'gradnorm/decoder.layer2': 0.14689995624682461, 'gradnorm/max_param': 0.1571006178855896, 'epoch': 0.55}

  6%|▌         | 1050/19090 [30:53<7:44:28,  1.54s/it]
  6%|▌         | 1051/19090 [30:56<9:53:53,  1.98s/it]
  6%|▌         | 1052/19090 [30:57<8:51:21,  1.77s/it]
  6%|▌         | 1053/19090 [30:58<8:07:49,  1.62s/it]
  6%|▌         | 1054/19090 [31:01<9:23:00,  1.87s/it]
  6%|▌         | 1055/19090 [31:02<8:29:33,  1.70s/it]
  6%|▌         | 1056/19090 [31:03<7:52:36,  1.57s/it]
  6%|▌         | 1057/19090 [31:06<10:06:04,  2.02s/it]
  6%|▌         | 1058/19090 [31:08<9:00:37,  1.80s/it] 
  6%|▌         | 1059/19090 [31:09<8:15:02,  1.65s/it]
  6%|▌         | 1060/19090 [31:12<9:54:25,  1.98s/it]
                                                      
{'loss': 0.0046, 'grad_norm': 0.8260498046875, 'learning_rate': 2.773703509690938e-05, 'recon_loss': 0.004561583494069055, 'recon_static': 0.0034471843391656876, 'recon_motion': 0.0011143993149744346, 'recon_norm': 2.2782796740531923, 'recon_norm_std': 0.12337316982448102, 'target_norm': 1.6168030768632888, 'recon_cos': 0.3749754451215267, 'gradnorm/compressor.query': 0.017798460942106756, 'gradnorm/compressor.layer0': 0.27909343486080845, 'gradnorm/compressor.layer1': 0.32466319082200806, 'gradnorm/compressor.layer2': 0.27343769073479673, 'gradnorm/compressor.layer3': 0.2549932686300607, 'gradnorm/compressor.layer4': 0.23986129945051698, 'gradnorm/compressor.layer5': 0.2258268036476076, 'gradnorm/compressor.layer6': 0.2156557949162518, 'gradnorm/compressor.layer7': 0.20301467393587613, 'gradnorm/decoder.layer0': 0.2419406340259832, 'gradnorm/decoder.layer1': 0.21999419972782264, 'gradnorm/decoder.layer2': 0.23614860052773962, 'gradnorm/max_param': 0.24801628291606903, 'epoch': 0.56}

  6%|▌         | 1060/19090 [31:12<9:54:25,  1.98s/it]
  6%|▌         | 1061/19090 [31:13<8:52:07,  1.77s/it]
  6%|▌         | 1062/19090 [31:14<8:08:15,  1.62s/it]
  6%|▌         | 1063/19090 [31:17<10:02:42,  2.01s/it]
  6%|▌         | 1064/19090 [31:18<8:57:51,  1.79s/it] 
  6%|▌         | 1065/19090 [31:20<8:12:05,  1.64s/it]
  6%|▌         | 1066/19090 [31:22<9:51:09,  1.97s/it]
  6%|▌         | 1067/19090 [31:24<8:49:29,  1.76s/it]
  6%|▌         | 1068/19090 [31:25<8:06:22,  1.62s/it]
  6%|▌         | 1069/19090 [31:28<10:02:37,  2.01s/it]
  6%|▌         | 1070/19090 [31:29<8:57:43,  1.79s/it] 
                                                      
{'loss': 0.0046, 'grad_norm': 0.5540702939033508, 'learning_rate': 2.7998952331063383e-05, 'recon_loss': 0.004575816862052306, 'recon_static': 0.0034635005693417044, 'recon_motion': 0.0011123163247248157, 'recon_norm': 2.2825992882251738, 'recon_norm_std': 0.12195501197129488, 'target_norm': 1.6201844334602356, 'recon_cos': 0.3749809294939041, 'gradnorm/compressor.query': 0.011395695396708132, 'gradnorm/compressor.layer0': 0.1859774562635541, 'gradnorm/compressor.layer1': 0.22121567414062843, 'gradnorm/compressor.layer2': 0.18358896127502228, 'gradnorm/compressor.layer3': 0.16757175193352278, 'gradnorm/compressor.layer4': 0.15742917954983532, 'gradnorm/compressor.layer5': 0.1487885202191146, 'gradnorm/compressor.layer6': 0.14266532034518167, 'gradnorm/compressor.layer7': 0.13781534874279153, 'gradnorm/decoder.layer0': 0.16421566920380493, 'gradnorm/decoder.layer1': 0.14940483713030528, 'gradnorm/decoder.layer2': 0.16123843343783795, 'gradnorm/max_param': 0.1650111824274063, 'epoch': 0.56}

  6%|▌         | 1070/19090 [31:29<8:57:43,  1.79s/it]
  6%|▌         | 1071/19090 [31:30<8:12:48,  1.64s/it]
  6%|▌         | 1072/19090 [31:33<9:47:19,  1.96s/it]
  6%|▌         | 1073/19090 [31:34<8:46:50,  1.75s/it]
  6%|▌         | 1074/19090 [31:36<8:04:58,  1.62s/it]
  6%|▌         | 1075/19090 [31:38<9:44:38,  1.95s/it]
  6%|▌         | 1076/19090 [31:40<8:44:51,  1.75s/it]
  6%|▌         | 1077/19090 [31:41<8:03:07,  1.61s/it]
  6%|▌         | 1078/19090 [31:44<9:37:58,  1.93s/it]
  6%|▌         | 1079/19090 [31:45<8:40:03,  1.73s/it]
  6%|▌         | 1080/19090 [31:46<7:59:41,  1.60s/it]
                                                      
{'loss': 0.0033, 'grad_norm': 0.4289793372154236, 'learning_rate': 2.826086956521739e-05, 'recon_loss': 0.0032675445370841773, 'recon_static': 0.0021444724407047033, 'recon_motion': 0.0011230721371248364, 'recon_norm': 1.9240202933549881, 'recon_norm_std': 0.14005980491638184, 'target_norm': 1.6234857439994812, 'recon_cos': 0.43989442735910417, 'gradnorm/compressor.query': 0.009234850276114336, 'gradnorm/compressor.layer0': 0.14564885593831572, 'gradnorm/compressor.layer1': 0.1690144094388362, 'gradnorm/compressor.layer2': 0.1447941026608489, 'gradnorm/compressor.layer3': 0.13112981805913643, 'gradnorm/compressor.layer4': 0.12307065022580296, 'gradnorm/compressor.layer5': 0.11804572003858006, 'gradnorm/compressor.layer6': 0.11120932916537439, 'gradnorm/compressor.layer7': 0.10594199552956192, 'gradnorm/decoder.layer0': 0.12302976822390735, 'gradnorm/decoder.layer1': 0.11282957399560704, 'gradnorm/decoder.layer2': 0.12433395037200194, 'gradnorm/max_param': 0.12992075085639954, 'epoch': 0.57}

  6%|▌         | 1080/19090 [31:46<7:59:41,  1.60s/it]
  6%|▌         | 1081/19090 [31:48<8:55:06,  1.78s/it]
  6%|▌         | 1082/19090 [31:50<8:10:14,  1.63s/it]
  6%|▌         | 1083/19090 [31:51<7:38:44,  1.53s/it]
  6%|▌         | 1084/19090 [31:54<9:16:05,  1.85s/it]
  6%|▌         | 1085/19090 [31:55<8:24:43,  1.68s/it]
  6%|▌         | 1086/19090 [31:56<7:48:34,  1.56s/it]
  6%|▌         | 1087/19090 [31:59<9:29:56,  1.90s/it]
  6%|▌         | 1088/19090 [32:00<8:35:19,  1.72s/it]
  6%|▌         | 1089/19090 [32:01<7:58:16,  1.59s/it]
  6%|▌         | 1090/19090 [32:04<9:42:56,  1.94s/it]
                                                      
{'loss': 0.0033, 'grad_norm': 0.5618371367454529, 'learning_rate': 2.85227867993714e-05, 'recon_loss': 0.0033373151265550403, 'recon_static': 0.002228825745987706, 'recon_motion': 0.001108489534817636, 'recon_norm': 1.9488086462020875, 'recon_norm_std': 0.13634572215378285, 'target_norm': 1.620220085978508, 'recon_cos': 0.43483448699116706, 'gradnorm/compressor.query': 0.012300908789498959, 'gradnorm/compressor.layer0': 0.19451483921551432, 'gradnorm/compressor.layer1': 0.22145358078439423, 'gradnorm/compressor.layer2': 0.1891787597141992, 'gradnorm/compressor.layer3': 0.17060395072519313, 'gradnorm/compressor.layer4': 0.1610278915052826, 'gradnorm/compressor.layer5': 0.152658382154898, 'gradnorm/compressor.layer6': 0.1448255583893442, 'gradnorm/compressor.layer7': 0.13841723851371532, 'gradnorm/decoder.layer0': 0.1628180433618277, 'gradnorm/decoder.layer1': 0.149305459987871, 'gradnorm/decoder.layer2': 0.159898195316584, 'gradnorm/max_param': 0.17195990681648254, 'epoch': 0.57}

  6%|▌         | 1090/19090 [32:04<9:42:56,  1.94s/it]
  6%|▌         | 1091/19090 [32:05<8:44:36,  1.75s/it]
  6%|▌         | 1092/19090 [32:07<8:03:47,  1.61s/it]
  6%|▌         | 1093/19090 [32:10<9:48:49,  1.96s/it]
  6%|▌         | 1094/19090 [32:11<8:48:23,  1.76s/it]
  6%|▌         | 1095/19090 [32:12<8:06:12,  1.62s/it]
  6%|▌         | 1096/19090 [32:15<9:44:17,  1.95s/it]
  6%|▌         | 1097/19090 [32:16<8:46:29,  1.76s/it]
  6%|▌         | 1098/19090 [32:17<8:04:26,  1.62s/it]
  6%|▌         | 1099/19090 [32:20<8:49:44,  1.77s/it]
  6%|▌         | 1100/19090 [32:21<8:07:20,  1.63s/it]
                                                      
{'loss': 0.0033, 'grad_norm': 0.47431185841560364, 'learning_rate': 2.8784704033525405e-05, 'recon_loss': 0.0033292595064267514, 'recon_static': 0.0022240220801904797, 'recon_motion': 0.0011052373854909092, 'recon_norm': 1.950474774837494, 'recon_norm_std': 0.13429748751223086, 'target_norm': 1.62032231092453, 'recon_cos': 0.43601375445723534, 'gradnorm/compressor.query': 0.009851151948218013, 'gradnorm/compressor.layer0': 0.15922408859616788, 'gradnorm/compressor.layer1': 0.19117759164448045, 'gradnorm/compressor.layer2': 0.15849575012132175, 'gradnorm/compressor.layer3': 0.14382136480802363, 'gradnorm/compressor.layer4': 0.13751332451459017, 'gradnorm/compressor.layer5': 0.12912546649268822, 'gradnorm/compressor.layer6': 0.12115279803820471, 'gradnorm/compressor.layer7': 0.1157376766516708, 'gradnorm/decoder.layer0': 0.13740313081712474, 'gradnorm/decoder.layer1': 0.12625696336550254, 'gradnorm/decoder.layer2': 0.13671526904443132, 'gradnorm/max_param': 0.14125896990299225, 'epoch': 0.58}

  6%|▌         | 1100/19090 [32:21<8:07:20,  1.63s/it]
  6%|▌         | 1101/19090 [32:22<7:37:45,  1.53s/it]
  6%|▌         | 1102/19090 [32:25<9:16:07,  1.85s/it]
  6%|▌         | 1103/19090 [32:26<8:25:18,  1.69s/it]
  6%|▌         | 1104/19090 [32:27<7:50:00,  1.57s/it]
  6%|▌         | 1105/19090 [32:30<9:25:56,  1.89s/it]
  6%|▌         | 1106/19090 [32:31<8:32:15,  1.71s/it]
  6%|▌         | 1107/19090 [32:33<7:54:24,  1.58s/it]
  6%|▌         | 1108/19090 [32:35<9:52:11,  1.98s/it]
  6%|▌         | 1109/19090 [32:37<8:50:18,  1.77s/it]
  6%|▌         | 1110/19090 [32:38<8:07:30,  1.63s/it]
                                                      
{'loss': 0.0046, 'grad_norm': 1.0386103391647339, 'learning_rate': 2.9046621267679414e-05, 'recon_loss': 0.004594146349700168, 'recon_static': 0.003494401107309386, 'recon_motion': 0.0010997452336596325, 'recon_norm': 2.2738191902637483, 'recon_norm_std': 0.11945281997323036, 'target_norm': 1.6186159431934357, 'recon_cos': 0.38130092471837995, 'gradnorm/compressor.query': 0.021803361992805855, 'gradnorm/compressor.layer0': 0.3572421012656575, 'gradnorm/compressor.layer1': 0.4118252176812548, 'gradnorm/compressor.layer2': 0.34304635684018325, 'gradnorm/compressor.layer3': 0.31147005108040765, 'gradnorm/compressor.layer4': 0.2953338902916152, 'gradnorm/compressor.layer5': 0.28284656835813926, 'gradnorm/compressor.layer6': 0.26929672376220637, 'gradnorm/compressor.layer7': 0.25454515621725937, 'gradnorm/decoder.layer0': 0.30192372473813583, 'gradnorm/decoder.layer1': 0.28048762999082405, 'gradnorm/decoder.layer2': 0.30323287774639857, 'gradnorm/max_param': 0.3177575170993805, 'epoch': 0.58}

  6%|▌         | 1110/19090 [32:38<8:07:30,  1.63s/it]
  6%|▌         | 1111/19090 [32:41<10:00:57,  2.01s/it]
  6%|▌         | 1112/19090 [32:42<8:57:14,  1.79s/it] 
  6%|▌         | 1113/19090 [32:44<8:11:58,  1.64s/it]
  6%|▌         | 1114/19090 [32:46<9:53:00,  1.98s/it]
  6%|▌         | 1115/19090 [32:48<8:51:53,  1.78s/it]
  6%|▌         | 1116/19090 [32:49<8:07:59,  1.63s/it]
  6%|▌         | 1117/19090 [32:52<9:48:33,  1.96s/it]
  6%|▌         | 1118/19090 [32:53<8:47:25,  1.76s/it]
  6%|▌         | 1119/19090 [32:54<8:05:08,  1.62s/it]
  6%|▌         | 1120/19090 [32:57<9:53:04,  1.98s/it]
                                                      
{'loss': 0.0056, 'grad_norm': 0.8146750330924988, 'learning_rate': 2.930853850183342e-05, 'recon_loss': 0.005576370924245566, 'recon_static': 0.004481041291728616, 'recon_motion': 0.0010953296965453773, 'recon_norm': 2.5284769952297212, 'recon_norm_std': 0.10596249233931303, 'target_norm': 1.619821161031723, 'recon_cos': 0.33659840300679206, 'gradnorm/compressor.query': 0.017662432277123392, 'gradnorm/compressor.layer0': 0.27807046484177395, 'gradnorm/compressor.layer1': 0.3193316211467055, 'gradnorm/compressor.layer2': 0.27086777957082364, 'gradnorm/compressor.layer3': 0.24826142651478675, 'gradnorm/compressor.layer4': 0.23337688905209863, 'gradnorm/compressor.layer5': 0.21954742758932977, 'gradnorm/compressor.layer6': 0.20755430042481887, 'gradnorm/compressor.layer7': 0.19618894386550106, 'gradnorm/decoder.layer0': 0.23837271482685554, 'gradnorm/decoder.layer1': 0.22103585000673792, 'gradnorm/decoder.layer2': 0.24325174642771014, 'gradnorm/max_param': 0.24646908044815063, 'epoch': 0.59}

  6%|▌         | 1120/19090 [32:57<9:53:04,  1.98s/it]
  6%|▌         | 1121/19090 [32:58<8:50:05,  1.77s/it]
  6%|▌         | 1122/19090 [33:00<8:06:08,  1.62s/it]
  6%|▌         | 1123/19090 [33:02<9:44:56,  1.95s/it]
  6%|▌         | 1124/19090 [33:04<8:46:05,  1.76s/it]
  6%|▌         | 1125/19090 [33:05<8:04:26,  1.62s/it]
  6%|▌         | 1126/19090 [33:08<9:36:21,  1.93s/it]
  6%|▌         | 1127/19090 [33:09<8:39:35,  1.74s/it]
  6%|▌         | 1128/19090 [33:10<8:00:08,  1.60s/it]
  6%|▌         | 1129/19090 [33:13<9:34:14,  1.92s/it]
  6%|▌         | 1130/19090 [33:14<8:38:23,  1.73s/it]
                                                      
{'loss': 0.005, 'grad_norm': 0.6403241157531738, 'learning_rate': 2.957045573598743e-05, 'recon_loss': 0.004988544515799731, 'recon_static': 0.003881324332905933, 'recon_motion': 0.0011072201770730317, 'recon_norm': 2.388418608903885, 'recon_norm_std': 0.10895995739847422, 'target_norm': 1.6206080466508865, 'recon_cos': 0.3559421479701996, 'gradnorm/compressor.query': 0.012821515870163846, 'gradnorm/compressor.layer0': 0.20932627723073655, 'gradnorm/compressor.layer1': 0.25318676171516497, 'gradnorm/compressor.layer2': 0.21253652609328455, 'gradnorm/compressor.layer3': 0.19652709003679367, 'gradnorm/compressor.layer4': 0.18571740395258596, 'gradnorm/compressor.layer5': 0.175464434693792, 'gradnorm/compressor.layer6': 0.16766704587671533, 'gradnorm/compressor.layer7': 0.1584536723059834, 'gradnorm/decoder.layer0': 0.1863907646756205, 'gradnorm/decoder.layer1': 0.17022428787303504, 'gradnorm/decoder.layer2': 0.18952176728391856, 'gradnorm/max_param': 0.18660633265972137, 'epoch': 0.59}

  6%|▌         | 1130/19090 [33:14<8:38:23,  1.73s/it]
  6%|▌         | 1131/19090 [33:15<8:00:57,  1.61s/it]
  6%|▌         | 1132/19090 [33:18<9:47:22,  1.96s/it]
  6%|▌         | 1133/19090 [33:19<8:48:36,  1.77s/it]
  6%|▌         | 1134/19090 [33:21<8:05:52,  1.62s/it]
  6%|▌         | 1135/19090 [33:24<10:02:39,  2.01s/it]
  6%|▌         | 1136/19090 [33:25<8:57:43,  1.80s/it] 
  6%|▌         | 1137/19090 [33:26<8:12:21,  1.65s/it]
  6%|▌         | 1138/19090 [33:29<10:04:48,  2.02s/it]
  6%|▌         | 1139/19090 [33:30<8:59:04,  1.80s/it] 
  6%|▌         | 1140/19090 [33:32<8:13:02,  1.65s/it]
                                                      
{'loss': 0.004, 'grad_norm': 0.5459286570549011, 'learning_rate': 2.9832372970141437e-05, 'recon_loss': 0.00398095251293853, 'recon_static': 0.0028952549793757497, 'recon_motion': 0.001085697431699373, 'recon_norm': 2.136503356695175, 'recon_norm_std': 0.1191799497231841, 'target_norm': 1.6176097095012665, 'recon_cos': 0.4001080483198166, 'gradnorm/compressor.query': 0.011449615417062539, 'gradnorm/compressor.layer0': 0.18498998693805976, 'gradnorm/compressor.layer1': 0.21544308728818343, 'gradnorm/compressor.layer2': 0.17791531334108443, 'gradnorm/compressor.layer3': 0.16194127455809826, 'gradnorm/compressor.layer4': 0.15536133059955373, 'gradnorm/compressor.layer5': 0.14757198926813272, 'gradnorm/compressor.layer6': 0.14207575095641256, 'gradnorm/compressor.layer7': 0.13434535991549715, 'gradnorm/decoder.layer0': 0.1627468993613805, 'gradnorm/decoder.layer1': 0.14895909671543378, 'gradnorm/decoder.layer2': 0.1629937762413529, 'gradnorm/max_param': 0.16520555317401886, 'epoch': 0.6}

  6%|▌         | 1140/19090 [33:32<8:13:02,  1.65s/it]
  6%|▌         | 1141/19090 [33:34<9:29:35,  1.90s/it]
  6%|▌         | 1142/19090 [33:36<8:42:18,  1.75s/it]
  6%|▌         | 1143/19090 [33:37<8:02:53,  1.61s/it]
  6%|▌         | 1144/19090 [33:39<9:06:04,  1.83s/it]
  6%|▌         | 1145/19090 [33:41<8:33:25,  1.72s/it]
  6%|▌         | 1146/19090 [33:42<7:55:07,  1.59s/it]
  6%|▌         | 1147/19090 [33:44<9:08:40,  1.83s/it]
  6%|▌         | 1148/19090 [33:46<8:43:21,  1.75s/it]
  6%|▌         | 1149/19090 [33:47<8:01:51,  1.61s/it]
  6%|▌         | 1150/19090 [33:50<9:17:28,  1.86s/it]
                                                      
{'loss': 0.0041, 'grad_norm': 0.6080405116081238, 'learning_rate': 3.0094290204295443e-05, 'recon_loss': 0.004074449121253565, 'recon_static': 0.00298317878623493, 'recon_motion': 0.0010912705271039157, 'recon_norm': 2.1615473508834837, 'recon_norm_std': 0.11572504602372646, 'target_norm': 1.618518790602684, 'recon_cos': 0.39364896938204763, 'gradnorm/compressor.query': 0.011992201489609689, 'gradnorm/compressor.layer0': 0.19716063623269967, 'gradnorm/compressor.layer1': 0.23601936631581982, 'gradnorm/compressor.layer2': 0.20254682822132403, 'gradnorm/compressor.layer3': 0.18608411009680917, 'gradnorm/compressor.layer4': 0.17712420221026579, 'gradnorm/compressor.layer5': 0.16675206259140993, 'gradnorm/compressor.layer6': 0.15926100403256996, 'gradnorm/compressor.layer7': 0.14821232481109078, 'gradnorm/decoder.layer0': 0.17902393759670737, 'gradnorm/decoder.layer1': 0.16558994808501046, 'gradnorm/decoder.layer2': 0.1826462748000664, 'gradnorm/max_param': 0.17591972649097443, 'epoch': 0.6}

  6%|▌         | 1150/19090 [33:50<9:17:28,  1.86s/it]
  6%|▌         | 1151/19090 [33:51<8:55:35,  1.79s/it]
  6%|▌         | 1152/19090 [33:53<8:10:22,  1.64s/it]
  6%|▌         | 1153/19090 [33:55<9:31:50,  1.91s/it]
  6%|▌         | 1154/19090 [33:57<8:41:33,  1.74s/it]
  6%|▌         | 1155/19090 [33:58<8:00:36,  1.61s/it]
  6%|▌         | 1156/19090 [34:00<9:26:39,  1.90s/it]
  6%|▌         | 1157/19090 [34:02<8:58:55,  1.80s/it]
  6%|▌         | 1158/19090 [34:03<8:13:46,  1.65s/it]
  6%|▌         | 1159/19090 [34:06<9:14:35,  1.86s/it]
  6%|▌         | 1160/19090 [34:07<9:04:32,  1.82s/it]
                                                      
{'loss': 0.0034, 'grad_norm': 0.4315149486064911, 'learning_rate': 3.0356207438449453e-05, 'recon_loss': 0.003349428391084075, 'recon_static': 0.002267515449784696, 'recon_motion': 0.0010819130460731686, 'recon_norm': 1.9586215525865556, 'recon_norm_std': 0.12686086259782314, 'target_norm': 1.6184880495071412, 'recon_cos': 0.43659273609519006, 'gradnorm/compressor.query': 0.009431038023107423, 'gradnorm/compressor.layer0': 0.14700145375942347, 'gradnorm/compressor.layer1': 0.1671440092455439, 'gradnorm/compressor.layer2': 0.14181395419961007, 'gradnorm/compressor.layer3': 0.13044486655274268, 'gradnorm/compressor.layer4': 0.12218423867690378, 'gradnorm/compressor.layer5': 0.11553360590726618, 'gradnorm/compressor.layer6': 0.11147378711505791, 'gradnorm/compressor.layer7': 0.10598775532463732, 'gradnorm/decoder.layer0': 0.12801260117673122, 'gradnorm/decoder.layer1': 0.11853938750281118, 'gradnorm/decoder.layer2': 0.13064065133331657, 'gradnorm/max_param': 0.13072054088115692, 'epoch': 0.61}

  6%|▌         | 1160/19090 [34:07<9:04:32,  1.82s/it]
  6%|▌         | 1161/19090 [34:09<8:17:39,  1.67s/it]
  6%|▌         | 1162/19090 [34:11<9:19:00,  1.87s/it]
  6%|▌         | 1163/19090 [34:12<8:42:55,  1.75s/it]
  6%|▌         | 1164/19090 [34:14<8:02:07,  1.61s/it]
  6%|▌         | 1165/19090 [34:16<9:37:24,  1.93s/it]
  6%|▌         | 1166/19090 [34:18<9:08:27,  1.84s/it]
  6%|▌         | 1167/19090 [34:19<8:19:37,  1.67s/it]
  6%|▌         | 1168/19090 [34:22<10:00:43,  2.01s/it]
  6%|▌         | 1169/19090 [34:23<8:55:47,  1.79s/it] 
  6%|▌         | 1170/19090 [34:25<8:10:53,  1.64s/it]
                                                      
{'loss': 0.0038, 'grad_norm': 0.7319920063018799, 'learning_rate': 3.061812467260346e-05, 'recon_loss': 0.0037863172066863625, 'recon_static': 0.0027045811468269674, 'recon_motion': 0.001081736187916249, 'recon_norm': 2.078027981519699, 'recon_norm_std': 0.11838831752538681, 'target_norm': 1.6205422013998032, 'recon_cos': 0.4124633602797985, 'gradnorm/compressor.query': 0.014649678217477908, 'gradnorm/compressor.layer0': 0.23465542706182568, 'gradnorm/compressor.layer1': 0.28624760795410054, 'gradnorm/compressor.layer2': 0.24255890977468622, 'gradnorm/compressor.layer3': 0.223242690118339, 'gradnorm/compressor.layer4': 0.2101425804599168, 'gradnorm/compressor.layer5': 0.19836489633310694, 'gradnorm/compressor.layer6': 0.19181468748832212, 'gradnorm/compressor.layer7': 0.18268623678597423, 'gradnorm/decoder.layer0': 0.21825710180805719, 'gradnorm/decoder.layer1': 0.19978860620720948, 'gradnorm/decoder.layer2': 0.22067477457709817, 'gradnorm/max_param': 0.20815108716487885, 'epoch': 0.61}

  6%|▌         | 1170/19090 [34:25<8:10:53,  1.64s/it]
  6%|▌         | 1171/19090 [34:28<10:15:44,  2.06s/it]
  6%|▌         | 1172/19090 [34:29<9:06:48,  1.83s/it] 
  6%|▌         | 1173/19090 [34:30<8:18:58,  1.67s/it]
  6%|▌         | 1174/19090 [34:33<9:40:10,  1.94s/it]
  6%|▌         | 1175/19090 [34:34<8:42:48,  1.75s/it]
  6%|▌         | 1176/19090 [34:36<8:02:04,  1.61s/it]
  6%|▌         | 1177/19090 [34:38<9:47:07,  1.97s/it]
  6%|▌         | 1178/19090 [34:40<8:46:27,  1.76s/it]
  6%|▌         | 1179/19090 [34:41<8:04:14,  1.62s/it]
  6%|▌         | 1180/19090 [34:44<10:00:52,  2.01s/it]
                                                       
{'loss': 0.0042, 'grad_norm': 0.6070747375488281, 'learning_rate': 3.0880041906757465e-05, 'recon_loss': 0.004161867569200695, 'recon_static': 0.0030753925908356904, 'recon_motion': 0.001086474914336577, 'recon_norm': 2.1831868469715117, 'recon_norm_std': 0.11265429630875587, 'target_norm': 1.6211834996938705, 'recon_cos': 0.3899339817464352, 'gradnorm/compressor.query': 0.012387183734794152, 'gradnorm/compressor.layer0': 0.2032488023838781, 'gradnorm/compressor.layer1': 0.2388407283174063, 'gradnorm/compressor.layer2': 0.20152346432636317, 'gradnorm/compressor.layer3': 0.18824956002000354, 'gradnorm/compressor.layer4': 0.17337376437266008, 'gradnorm/compressor.layer5': 0.16408450115366904, 'gradnorm/compressor.layer6': 0.15572951172202507, 'gradnorm/compressor.layer7': 0.14746300272726953, 'gradnorm/decoder.layer0': 0.1779754596547266, 'gradnorm/decoder.layer1': 0.16138597887338865, 'gradnorm/decoder.layer2': 0.1824329997720073, 'gradnorm/max_param': 0.1807820349931717, 'epoch': 0.62}

  6%|▌         | 1180/19090 [34:44<10:00:52,  2.01s/it]
  6%|▌         | 1181/19090 [34:45<8:57:07,  1.80s/it] 
  6%|▌         | 1182/19090 [34:46<8:11:57,  1.65s/it]
  6%|▌         | 1183/19090 [34:49<9:35:03,  1.93s/it]
  6%|▌         | 1184/19090 [34:50<8:37:45,  1.73s/it]
  6%|▌         | 1185/19090 [34:52<7:58:00,  1.60s/it]
  6%|▌         | 1186/19090 [34:54<9:23:10,  1.89s/it]
  6%|▌         | 1187/19090 [34:55<8:29:44,  1.71s/it]
  6%|▌         | 1188/19090 [34:57<7:52:28,  1.58s/it]
  6%|▌         | 1189/19090 [34:59<9:14:04,  1.86s/it]
  6%|▌         | 1190/19090 [35:01<8:36:12,  1.73s/it]
                                                      
{'loss': 0.0055, 'grad_norm': 0.9237999320030212, 'learning_rate': 3.1141959140911475e-05, 'recon_loss': 0.0054766438901424404, 'recon_static': 0.004384543490596115, 'recon_motion': 0.0010921003035036848, 'recon_norm': 2.491362726688385, 'recon_norm_std': 0.10012854970991611, 'target_norm': 1.6228890538215637, 'recon_cos': 0.34503071904182436, 'gradnorm/compressor.query': 0.019117996161692474, 'gradnorm/compressor.layer0': 0.31125500154115215, 'gradnorm/compressor.layer1': 0.36433706367443774, 'gradnorm/compressor.layer2': 0.3023821596888365, 'gradnorm/compressor.layer3': 0.2769227526411426, 'gradnorm/compressor.layer4': 0.26126785723787355, 'gradnorm/compressor.layer5': 0.24612773933019597, 'gradnorm/compressor.layer6': 0.23531221717776948, 'gradnorm/compressor.layer7': 0.22741138166713423, 'gradnorm/decoder.layer0': 0.2742663491608487, 'gradnorm/decoder.layer1': 0.2541576461415937, 'gradnorm/decoder.layer2': 0.2820510318401812, 'gradnorm/max_param': 0.2759329676628113, 'epoch': 0.62}

  6%|▌         | 1190/19090 [35:01<8:36:12,  1.73s/it]
  6%|▌         | 1191/19090 [35:02<7:57:38,  1.60s/it]
  6%|▌         | 1192/19090 [35:04<9:18:33,  1.87s/it]
  6%|▌         | 1193/19090 [35:06<8:27:37,  1.70s/it]
  6%|▋         | 1194/19090 [35:07<7:50:44,  1.58s/it]
  6%|▋         | 1195/19090 [35:10<9:15:48,  1.86s/it]
  6%|▋         | 1196/19090 [35:11<8:58:08,  1.80s/it]
  6%|▋         | 1197/19090 [35:13<8:12:37,  1.65s/it]
  6%|▋         | 1198/19090 [35:15<9:26:56,  1.90s/it]
  6%|▋         | 1199/19090 [35:16<8:37:39,  1.74s/it]
  6%|▋         | 1200/19090 [35:18<7:58:39,  1.61s/it]
                                                      
{'loss': 0.0064, 'grad_norm': 0.7463095188140869, 'learning_rate': 3.1403876375065485e-05, 'recon_loss': 0.006451920734252781, 'recon_static': 0.005368529917905107, 'recon_motion': 0.0010833908061613328, 'recon_norm': 2.719929224252701, 'recon_norm_std': 0.09194218292832375, 'target_norm': 1.6219389766454697, 'recon_cos': 0.3152636229991913, 'gradnorm/compressor.query': 0.014019644067769517, 'gradnorm/compressor.layer0': 0.2398952556191259, 'gradnorm/compressor.layer1': 0.2895295005041421, 'gradnorm/compressor.layer2': 0.2478192103128025, 'gradnorm/compressor.layer3': 0.22888956692672605, 'gradnorm/compressor.layer4': 0.21366858296761204, 'gradnorm/compressor.layer5': 0.2019866443316805, 'gradnorm/compressor.layer6': 0.19409743966591048, 'gradnorm/compressor.layer7': 0.18697265137254307, 'gradnorm/decoder.layer0': 0.22070633202307324, 'gradnorm/decoder.layer1': 0.2050746463298945, 'gradnorm/decoder.layer2': 0.22747988606710182, 'gradnorm/max_param': 0.2128056138753891, 'epoch': 0.63}

  6%|▋         | 1200/19090 [35:18<7:58:39,  1.61s/it]
  6%|▋         | 1201/19090 [35:20<9:32:18,  1.92s/it]
  6%|▋         | 1202/19090 [35:22<9:00:44,  1.81s/it]
  6%|▋         | 1203/19090 [35:23<8:14:44,  1.66s/it]
  6%|▋         | 1204/19090 [35:26<9:19:55,  1.88s/it]
  6%|▋         | 1205/19090 [35:27<9:17:43,  1.87s/it]
  6%|▋         | 1206/19090 [35:29<8:25:51,  1.70s/it]
  6%|▋         | 1207/19090 [35:31<8:59:47,  1.81s/it]
  6%|▋         | 1208/19090 [35:33<9:00:29,  1.81s/it]
  6%|▋         | 1209/19090 [35:34<8:14:06,  1.66s/it]
  6%|▋         | 1210/19090 [35:36<9:11:27,  1.85s/it]
                                                      
{'loss': 0.004, 'grad_norm': 0.48371651768684387, 'learning_rate': 3.166579360921949e-05, 'recon_loss': 0.00402970599825494, 'recon_static': 0.002963572507724166, 'recon_motion': 0.0010661335225449875, 'recon_norm': 2.1435933291912077, 'recon_norm_std': 0.11313146390020848, 'target_norm': 1.620101049542427, 'recon_cos': 0.40154671892523763, 'gradnorm/compressor.query': 0.00969596583069874, 'gradnorm/compressor.layer0': 0.1508136966363802, 'gradnorm/compressor.layer1': 0.18664234425649873, 'gradnorm/compressor.layer2': 0.16008249771155575, 'gradnorm/compressor.layer3': 0.14550614196023207, 'gradnorm/compressor.layer4': 0.14017085413932598, 'gradnorm/compressor.layer5': 0.13210256952660743, 'gradnorm/compressor.layer6': 0.12588642651309784, 'gradnorm/compressor.layer7': 0.12023313477702549, 'gradnorm/decoder.layer0': 0.14420577724504782, 'gradnorm/decoder.layer1': 0.13343505943130132, 'gradnorm/decoder.layer2': 0.15331831402296192, 'gradnorm/max_param': 0.13439098000526428, 'epoch': 0.63}

  6%|▋         | 1210/19090 [35:36<9:11:27,  1.85s/it]
  6%|▋         | 1211/19090 [35:38<8:57:02,  1.80s/it]
  6%|▋         | 1212/19090 [35:39<8:10:45,  1.65s/it]
  6%|▋         | 1213/19090 [35:41<9:02:49,  1.82s/it]
  6%|▋         | 1214/19090 [35:44<9:31:39,  1.92s/it]
  6%|▋         | 1215/19090 [35:45<8:35:28,  1.73s/it]
  6%|▋         | 1216/19090 [35:47<8:48:52,  1.78s/it]
  6%|▋         | 1217/19090 [35:49<9:39:33,  1.95s/it]
  6%|▋         | 1218/19090 [35:50<8:41:23,  1.75s/it]
  6%|▋         | 1219/19090 [35:52<8:36:04,  1.73s/it]
  6%|▋         | 1220/19090 [35:54<9:15:47,  1.87s/it]
                                                      
{'loss': 0.0031, 'grad_norm': 0.49582499265670776, 'learning_rate': 3.192771084337349e-05, 'recon_loss': 0.003056900796946138, 'recon_static': 0.0019932364171836524, 'recon_motion': 0.0010636642982717603, 'recon_norm': 1.874099913239479, 'recon_norm_std': 0.12300604414194823, 'target_norm': 1.618334373831749, 'recon_cos': 0.456379459053278, 'gradnorm/compressor.query': 0.009951786291594787, 'gradnorm/compressor.layer0': 0.16317902266961926, 'gradnorm/compressor.layer1': 0.19090115797564383, 'gradnorm/compressor.layer2': 0.16282378616130946, 'gradnorm/compressor.layer3': 0.14827990950369407, 'gradnorm/compressor.layer4': 0.1415841848299348, 'gradnorm/compressor.layer5': 0.13504035561769775, 'gradnorm/compressor.layer6': 0.12816721581251284, 'gradnorm/compressor.layer7': 0.12181383273030277, 'gradnorm/decoder.layer0': 0.146575251046977, 'gradnorm/decoder.layer1': 0.1363429558783519, 'gradnorm/decoder.layer2': 0.1566942515579096, 'gradnorm/max_param': 0.14467085897922516, 'epoch': 0.64}

  6%|▋         | 1220/19090 [35:54<9:15:47,  1.87s/it]
  6%|▋         | 1221/19090 [35:56<8:25:46,  1.70s/it]
  6%|▋         | 1222/19090 [35:57<8:35:01,  1.73s/it]
  6%|▋         | 1223/19090 [35:59<8:56:23,  1.80s/it]
  6%|▋         | 1224/19090 [36:01<8:12:30,  1.65s/it]
  6%|▋         | 1225/19090 [36:03<9:02:45,  1.82s/it]
  6%|▋         | 1226/19090 [36:04<8:36:45,  1.74s/it]
  6%|▋         | 1227/19090 [36:06<7:57:49,  1.60s/it]
  6%|▋         | 1228/19090 [36:08<9:27:45,  1.91s/it]
  6%|▋         | 1229/19090 [36:10<8:33:08,  1.72s/it]
  6%|▋         | 1230/19090 [36:11<7:54:46,  1.59s/it]
                                                      
{'loss': 0.0049, 'grad_norm': 0.753669023513794, 'learning_rate': 3.21896280775275e-05, 'recon_loss': 0.004865316854557022, 'recon_static': 0.003791809146059677, 'recon_motion': 0.0010735078452853487, 'recon_norm': 2.3580446600914002, 'recon_norm_std': 0.10100686829537153, 'target_norm': 1.6201349169015884, 'recon_cos': 0.3638701759278774, 'gradnorm/compressor.query': 0.014374632212770947, 'gradnorm/compressor.layer0': 0.24649334058461364, 'gradnorm/compressor.layer1': 0.29570557506357587, 'gradnorm/compressor.layer2': 0.25045865424662417, 'gradnorm/compressor.layer3': 0.22753979924291434, 'gradnorm/compressor.layer4': 0.2133647992601497, 'gradnorm/compressor.layer5': 0.20161504910657593, 'gradnorm/compressor.layer6': 0.19522841550929276, 'gradnorm/compressor.layer7': 0.18575639439752378, 'gradnorm/decoder.layer0': 0.22190241610695266, 'gradnorm/decoder.layer1': 0.20564061182406246, 'gradnorm/decoder.layer2': 0.2340604340063573, 'gradnorm/max_param': 0.2184624969959259, 'epoch': 0.64}

  6%|▋         | 1230/19090 [36:11<7:54:46,  1.59s/it]
  6%|▋         | 1231/19090 [36:14<9:50:31,  1.98s/it]
  6%|▋         | 1232/19090 [36:15<8:48:35,  1.78s/it]
  6%|▋         | 1233/19090 [36:16<8:06:47,  1.64s/it]
  6%|▋         | 1234/19090 [36:20<10:21:26,  2.09s/it]
  6%|▋         | 1235/19090 [36:21<9:10:24,  1.85s/it] 
  6%|▋         | 1236/19090 [36:22<8:20:03,  1.68s/it]
  6%|▋         | 1237/19090 [36:25<9:43:38,  1.96s/it]
  6%|▋         | 1238/19090 [36:26<8:43:05,  1.76s/it]
  6%|▋         | 1239/19090 [36:27<8:01:40,  1.62s/it]
  6%|▋         | 1240/19090 [36:30<9:42:44,  1.96s/it]
                                                      
{'loss': 0.0042, 'grad_norm': 0.5544041395187378, 'learning_rate': 3.245154531168151e-05, 'recon_loss': 0.0041506454697810115, 'recon_static': 0.0030770608456805347, 'recon_motion': 0.0010735846706666052, 'recon_norm': 2.1793375968933106, 'recon_norm_std': 0.10585251525044441, 'target_norm': 1.619637778401375, 'recon_cos': 0.3929289698600769, 'gradnorm/compressor.query': 0.010899327766894394, 'gradnorm/compressor.layer0': 0.17765867276917593, 'gradnorm/compressor.layer1': 0.21273976851850324, 'gradnorm/compressor.layer2': 0.18377164581830469, 'gradnorm/compressor.layer3': 0.1661798226160165, 'gradnorm/compressor.layer4': 0.15913825870902668, 'gradnorm/compressor.layer5': 0.15074066198901212, 'gradnorm/compressor.layer6': 0.14256996627875113, 'gradnorm/compressor.layer7': 0.13748287831257094, 'gradnorm/decoder.layer0': 0.1647951422795835, 'gradnorm/decoder.layer1': 0.15265278161207507, 'gradnorm/decoder.layer2': 0.17692029450554628, 'gradnorm/max_param': 0.1571178138256073, 'epoch': 0.65}

  6%|▋         | 1240/19090 [36:30<9:42:44,  1.96s/it]
  7%|▋         | 1241/19090 [36:31<8:42:58,  1.76s/it]
  7%|▋         | 1242/19090 [36:33<8:01:31,  1.62s/it]
  7%|▋         | 1243/19090 [36:35<9:30:42,  1.92s/it]
  7%|▋         | 1244/19090 [36:37<8:35:29,  1.73s/it]
  7%|▋         | 1245/19090 [36:38<7:55:18,  1.60s/it]
  7%|▋         | 1246/19090 [36:40<9:14:05,  1.86s/it]
  7%|▋         | 1247/19090 [36:42<8:22:21,  1.69s/it]
  7%|▋         | 1248/19090 [36:43<7:46:26,  1.57s/it]
  7%|▋         | 1249/19090 [36:46<9:36:04,  1.94s/it]
  7%|▋         | 1250/19090 [36:47<8:37:40,  1.74s/it]
                                                      
{'loss': 0.0038, 'grad_norm': 0.5680374503135681, 'learning_rate': 3.271346254583551e-05, 'recon_loss': 0.0038196725014131515, 'recon_static': 0.0027630945201963185, 'recon_motion': 0.0010565780321485363, 'recon_norm': 2.0993997275829317, 'recon_norm_std': 0.10950050633400679, 'target_norm': 1.617936447262764, 'recon_cos': 0.409575417637825, 'gradnorm/compressor.query': 0.011144240295686557, 'gradnorm/compressor.layer0': 0.18237100418764834, 'gradnorm/compressor.layer1': 0.22166830193273743, 'gradnorm/compressor.layer2': 0.18872702959320825, 'gradnorm/compressor.layer3': 0.17367954011579428, 'gradnorm/compressor.layer4': 0.1642023013158974, 'gradnorm/compressor.layer5': 0.15412097271996802, 'gradnorm/compressor.layer6': 0.14631827792280158, 'gradnorm/compressor.layer7': 0.1388709425921496, 'gradnorm/decoder.layer0': 0.16542857101045655, 'gradnorm/decoder.layer1': 0.15545987985696805, 'gradnorm/decoder.layer2': 0.17725566701546014, 'gradnorm/max_param': 0.16259777545928955, 'epoch': 0.65}

  7%|▋         | 1250/19090 [36:47<8:37:40,  1.74s/it]
  7%|▋         | 1251/19090 [36:48<7:57:07,  1.60s/it]
  7%|▋         | 1252/19090 [36:51<9:47:31,  1.98s/it]
  7%|▋         | 1253/19090 [36:52<8:46:02,  1.77s/it]
  7%|▋         | 1254/19090 [36:54<8:02:46,  1.62s/it]
  7%|▋         | 1255/19090 [36:57<9:52:58,  1.99s/it]
  7%|▋         | 1256/19090 [36:58<8:49:54,  1.78s/it]
  7%|▋         | 1257/19090 [36:59<8:05:56,  1.63s/it]
  7%|▋         | 1258/19090 [37:02<9:35:44,  1.94s/it]
  7%|▋         | 1259/19090 [37:03<8:37:57,  1.74s/it]
  7%|▋         | 1260/19090 [37:04<7:56:59,  1.61s/it]
                                                      
{'loss': 0.0041, 'grad_norm': 0.5755442976951599, 'learning_rate': 3.297537977998952e-05, 'recon_loss': 0.004072324896696955, 'recon_static': 0.003006820875452831, 'recon_motion': 0.001065504019788932, 'recon_norm': 2.1630237817764284, 'recon_norm_std': 0.10614091418683529, 'target_norm': 1.622839131951332, 'recon_cos': 0.3955969147384167, 'gradnorm/compressor.query': 0.011126010420858314, 'gradnorm/compressor.layer0': 0.18613293098373884, 'gradnorm/compressor.layer1': 0.22790002638053825, 'gradnorm/compressor.layer2': 0.1909130612886553, 'gradnorm/compressor.layer3': 0.17514908817632707, 'gradnorm/compressor.layer4': 0.16376460937492254, 'gradnorm/compressor.layer5': 0.15381074328661348, 'gradnorm/compressor.layer6': 0.1461767017254682, 'gradnorm/compressor.layer7': 0.14032636165116266, 'gradnorm/decoder.layer0': 0.1687634155892389, 'gradnorm/decoder.layer1': 0.1575841509811519, 'gradnorm/decoder.layer2': 0.18046894362468158, 'gradnorm/max_param': 0.16644220054149628, 'epoch': 0.66}

  7%|▋         | 1260/19090 [37:04<7:56:59,  1.61s/it]
  7%|▋         | 1261/19090 [37:07<10:00:26,  2.02s/it]
  7%|▋         | 1262/19090 [37:09<8:54:18,  1.80s/it] 
  7%|▋         | 1263/19090 [37:10<8:07:59,  1.64s/it]
  7%|▋         | 1264/19090 [37:13<9:37:30,  1.94s/it]
  7%|▋         | 1265/19090 [37:14<8:38:18,  1.74s/it]
  7%|▋         | 1266/19090 [37:15<7:57:34,  1.61s/it]
  7%|▋         | 1267/19090 [37:18<10:13:59,  2.07s/it]
  7%|▋         | 1268/19090 [37:19<9:04:09,  1.83s/it] 
  7%|▋         | 1269/19090 [37:21<8:15:22,  1.67s/it]
  7%|▋         | 1270/19090 [37:24<9:55:48,  2.01s/it]
                                                      
{'loss': 0.0041, 'grad_norm': 0.7400206327438354, 'learning_rate': 3.323729701414353e-05, 'recon_loss': 0.0041332797438371925, 'recon_static': 0.0030719697184395044, 'recon_motion': 0.0010613099249894731, 'recon_norm': 2.180220681428909, 'recon_norm_std': 0.10467743538320065, 'target_norm': 1.6201214432716369, 'recon_cos': 0.3932824872434139, 'gradnorm/compressor.query': 0.013976185576133844, 'gradnorm/compressor.layer0': 0.23543616709219203, 'gradnorm/compressor.layer1': 0.2834850092293483, 'gradnorm/compressor.layer2': 0.24311093435566422, 'gradnorm/compressor.layer3': 0.2230512645309551, 'gradnorm/compressor.layer4': 0.21619146891730961, 'gradnorm/compressor.layer5': 0.2029142203141099, 'gradnorm/compressor.layer6': 0.19248659461892473, 'gradnorm/compressor.layer7': 0.1832046748667149, 'gradnorm/decoder.layer0': 0.21791042415735432, 'gradnorm/decoder.layer1': 0.20455048212040264, 'gradnorm/decoder.layer2': 0.23437720933508652, 'gradnorm/max_param': 0.2085885852575302, 'epoch': 0.67}

  7%|▋         | 1270/19090 [37:24<9:55:48,  2.01s/it]
  7%|▋         | 1271/19090 [37:25<8:51:50,  1.79s/it]
  7%|▋         | 1272/19090 [37:26<8:06:36,  1.64s/it]
  7%|▋         | 1273/19090 [37:29<9:37:49,  1.95s/it]
  7%|▋         | 1274/19090 [37:30<8:40:01,  1.75s/it]
  7%|▋         | 1275/19090 [37:31<7:58:00,  1.61s/it]
  7%|▋         | 1276/19090 [37:34<9:34:09,  1.93s/it]
  7%|▋         | 1277/19090 [37:35<8:36:15,  1.74s/it]
  7%|▋         | 1278/19090 [37:37<7:55:19,  1.60s/it]
  7%|▋         | 1279/19090 [37:39<9:45:34,  1.97s/it]
  7%|▋         | 1280/19090 [37:41<8:44:55,  1.77s/it]
                                                      
{'loss': 0.007, 'grad_norm': 0.9833950996398926, 'learning_rate': 3.349921424829754e-05, 'recon_loss': 0.006995706353336573, 'recon_static': 0.005918743729125708, 'recon_motion': 0.0010769627493573352, 'recon_norm': 2.8334212839603423, 'recon_norm_std': 0.08198962174355984, 'target_norm': 1.6222247302532196, 'recon_cos': 0.3011888012290001, 'gradnorm/compressor.query': 0.019687264905577966, 'gradnorm/compressor.layer0': 0.32831903805732165, 'gradnorm/compressor.layer1': 0.38756982035891036, 'gradnorm/compressor.layer2': 0.3233870789484198, 'gradnorm/compressor.layer3': 0.29767341914684253, 'gradnorm/compressor.layer4': 0.27984103298944957, 'gradnorm/compressor.layer5': 0.2666097786444308, 'gradnorm/compressor.layer6': 0.24878157295481135, 'gradnorm/compressor.layer7': 0.23702022787127824, 'gradnorm/decoder.layer0': 0.2871762541643507, 'gradnorm/decoder.layer1': 0.26390892957972373, 'gradnorm/decoder.layer2': 0.30968541794201165, 'gradnorm/max_param': 0.2919008433818817, 'epoch': 0.67}

  7%|▋         | 1280/19090 [37:41<8:44:55,  1.77s/it]
  7%|▋         | 1281/19090 [37:42<8:02:57,  1.63s/it]
  7%|▋         | 1282/19090 [37:45<10:01:19,  2.03s/it]
  7%|▋         | 1283/19090 [37:46<8:55:50,  1.81s/it] 
  7%|▋         | 1284/19090 [37:48<8:10:07,  1.65s/it]
  7%|▋         | 1285/19090 [37:50<9:40:46,  1.96s/it]
  7%|▋         | 1286/19090 [37:52<8:41:45,  1.76s/it]
  7%|▋         | 1287/19090 [37:53<8:01:13,  1.62s/it]
  7%|▋         | 1288/19090 [37:55<9:23:16,  1.90s/it]
  7%|▋         | 1289/19090 [37:57<8:29:03,  1.72s/it]
  7%|▋         | 1290/19090 [37:58<7:52:05,  1.59s/it]
                                                      
{'loss': 0.0056, 'grad_norm': 0.674756646156311, 'learning_rate': 3.3761131482451545e-05, 'recon_loss': 0.00563808383885771, 'recon_static': 0.004575929418206215, 'recon_motion': 0.0010621546272886917, 'recon_norm': 2.538835698366165, 'recon_norm_std': 0.0910150794312358, 'target_norm': 1.6210645973682403, 'recon_cos': 0.3410701438784599, 'gradnorm/compressor.query': 0.012921923670992758, 'gradnorm/compressor.layer0': 0.21542192170160676, 'gradnorm/compressor.layer1': 0.2546867405698722, 'gradnorm/compressor.layer2': 0.21708621229256705, 'gradnorm/compressor.layer3': 0.20086010597834897, 'gradnorm/compressor.layer4': 0.19174212550303824, 'gradnorm/compressor.layer5': 0.18375475904893235, 'gradnorm/compressor.layer6': 0.17541611683726127, 'gradnorm/compressor.layer7': 0.1671840279655368, 'gradnorm/decoder.layer0': 0.20091019434519958, 'gradnorm/decoder.layer1': 0.19133120642190432, 'gradnorm/decoder.layer2': 0.2239681129020698, 'gradnorm/max_param': 0.19084186851978302, 'epoch': 0.68}

  7%|▋         | 1290/19090 [37:58<7:52:05,  1.59s/it]
  7%|▋         | 1291/19090 [38:01<9:48:03,  1.98s/it]
  7%|▋         | 1292/19090 [38:02<8:46:08,  1.77s/it]
  7%|▋         | 1293/19090 [38:03<8:03:32,  1.63s/it]
  7%|▋         | 1294/19090 [38:06<9:44:18,  1.97s/it]
  7%|▋         | 1295/19090 [38:08<8:44:09,  1.77s/it]
  7%|▋         | 1296/19090 [38:09<8:01:35,  1.62s/it]
  7%|▋         | 1297/19090 [38:11<9:27:40,  1.91s/it]
  7%|▋         | 1298/19090 [38:13<8:32:36,  1.73s/it]
  7%|▋         | 1299/19090 [38:14<7:55:08,  1.60s/it]
  7%|▋         | 1300/19090 [38:17<9:41:00,  1.96s/it]
                                                      
{'loss': 0.0035, 'grad_norm': 0.5012180209159851, 'learning_rate': 3.4023048716605554e-05, 'recon_loss': 0.003463277011178434, 'recon_static': 0.002417071961099282, 'recon_motion': 0.001046205117017962, 'recon_norm': 1.998044815659523, 'recon_norm_std': 0.11043328903615475, 'target_norm': 1.619982573390007, 'recon_cos': 0.4310580648481846, 'gradnorm/compressor.query': 0.009358843728349525, 'gradnorm/compressor.layer0': 0.15951597929937694, 'gradnorm/compressor.layer1': 0.192874956695888, 'gradnorm/compressor.layer2': 0.16349027956676304, 'gradnorm/compressor.layer3': 0.1504590324777628, 'gradnorm/compressor.layer4': 0.14437485300609817, 'gradnorm/compressor.layer5': 0.13490953225522967, 'gradnorm/compressor.layer6': 0.1288860984425446, 'gradnorm/compressor.layer7': 0.12356437367748979, 'gradnorm/decoder.layer0': 0.14951944686281732, 'gradnorm/decoder.layer1': 0.13956436881246156, 'gradnorm/decoder.layer2': 0.1623578293306157, 'gradnorm/max_param': 0.14183752238750458, 'epoch': 0.68}

  7%|▋         | 1300/19090 [38:17<9:41:00,  1.96s/it]
  7%|▋         | 1301/19090 [38:18<8:42:25,  1.76s/it]
  7%|▋         | 1302/19090 [38:19<8:00:20,  1.62s/it]
  7%|▋         | 1303/19090 [38:22<9:19:20,  1.89s/it]
  7%|▋         | 1304/19090 [38:23<8:26:21,  1.71s/it]
  7%|▋         | 1305/19090 [38:25<7:49:31,  1.58s/it]
  7%|▋         | 1306/19090 [38:27<9:23:16,  1.90s/it]
  7%|▋         | 1307/19090 [38:28<8:29:11,  1.72s/it]
  7%|▋         | 1308/19090 [38:30<7:53:12,  1.60s/it]
  7%|▋         | 1309/19090 [38:33<9:43:19,  1.97s/it]
  7%|▋         | 1310/19090 [38:34<8:43:40,  1.77s/it]
                                                      
{'loss': 0.0038, 'grad_norm': 0.6364181637763977, 'learning_rate': 3.4284965950759564e-05, 'recon_loss': 0.003822045534616336, 'recon_static': 0.002766884834272787, 'recon_motion': 0.0010551606304943562, 'recon_norm': 2.0931721717119216, 'recon_norm_std': 0.10686777103692294, 'target_norm': 1.6224446564912796, 'recon_cos': 0.4093180298805237, 'gradnorm/compressor.query': 0.011690043764720664, 'gradnorm/compressor.layer0': 0.20374344249130966, 'gradnorm/compressor.layer1': 0.24996546417348664, 'gradnorm/compressor.layer2': 0.20731896076192255, 'gradnorm/compressor.layer3': 0.18988458916280845, 'gradnorm/compressor.layer4': 0.18046046977017838, 'gradnorm/compressor.layer5': 0.16922761230826425, 'gradnorm/compressor.layer6': 0.1636178143608625, 'gradnorm/compressor.layer7': 0.15595160840948824, 'gradnorm/decoder.layer0': 0.18974101422510567, 'gradnorm/decoder.layer1': 0.17687012070911542, 'gradnorm/decoder.layer2': 0.20561996813243733, 'gradnorm/max_param': 0.18212193250656128, 'epoch': 0.69}

  7%|▋         | 1310/19090 [38:34<8:43:40,  1.77s/it]
  7%|▋         | 1311/19090 [38:35<8:02:06,  1.63s/it]
  7%|▋         | 1312/19090 [38:38<9:21:08,  1.89s/it]
  7%|▋         | 1313/19090 [38:39<8:27:53,  1.71s/it]
  7%|▋         | 1314/19090 [38:40<7:50:43,  1.59s/it]
  7%|▋         | 1315/19090 [38:43<9:33:41,  1.94s/it]
  7%|▋         | 1316/19090 [38:44<8:37:53,  1.75s/it]
  7%|▋         | 1317/19090 [38:46<7:59:40,  1.62s/it]
  7%|▋         | 1318/19090 [38:48<9:04:17,  1.84s/it]
  7%|▋         | 1319/19090 [38:49<8:15:52,  1.67s/it]
  7%|▋         | 1320/19090 [38:51<7:42:50,  1.56s/it]
                                                      
{'loss': 0.0044, 'grad_norm': 0.7363038659095764, 'learning_rate': 3.454688318491357e-05, 'recon_loss': 0.004433639970375225, 'recon_static': 0.003388114564586431, 'recon_motion': 0.0010455255571287126, 'recon_norm': 2.261129456758499, 'recon_norm_std': 0.09823453929275275, 'target_norm': 1.6196242719888687, 'recon_cos': 0.38128271475434306, 'gradnorm/compressor.query': 0.013928870696723102, 'gradnorm/compressor.layer0': 0.23105437826102063, 'gradnorm/compressor.layer1': 0.28735919386083436, 'gradnorm/compressor.layer2': 0.24170005875258468, 'gradnorm/compressor.layer3': 0.223462468866247, 'gradnorm/compressor.layer4': 0.21137156683425143, 'gradnorm/compressor.layer5': 0.199371606489976, 'gradnorm/compressor.layer6': 0.1899280298225105, 'gradnorm/compressor.layer7': 0.18144150556465402, 'gradnorm/decoder.layer0': 0.21886918533559896, 'gradnorm/decoder.layer1': 0.2014133060157466, 'gradnorm/decoder.layer2': 0.23609717567932548, 'gradnorm/max_param': 0.2058621495962143, 'epoch': 0.69}

  7%|▋         | 1320/19090 [38:51<7:42:50,  1.56s/it]
  7%|▋         | 1321/19090 [38:54<9:54:23,  2.01s/it]
  7%|▋         | 1322/19090 [38:55<8:51:08,  1.79s/it]
  7%|▋         | 1323/19090 [38:56<8:06:53,  1.64s/it]
  7%|▋         | 1324/19090 [38:59<10:01:46,  2.03s/it]
  7%|▋         | 1325/19090 [39:00<8:55:47,  1.81s/it] 
  7%|▋         | 1326/19090 [39:02<8:09:45,  1.65s/it]
  7%|▋         | 1327/19090 [39:05<9:58:16,  2.02s/it]
  7%|▋         | 1328/19090 [39:06<8:53:12,  1.80s/it]
  7%|▋         | 1329/19090 [39:07<8:08:10,  1.65s/it]
  7%|▋         | 1330/19090 [39:10<9:29:25,  1.92s/it]
                                                      
{'loss': 0.0057, 'grad_norm': 0.778155505657196, 'learning_rate': 3.4808800419067577e-05, 'recon_loss': 0.00567223783582449, 'recon_static': 0.004619122203439474, 'recon_motion': 0.0010531156411161646, 'recon_norm': 2.555164271593094, 'recon_norm_std': 0.08931224755942821, 'target_norm': 1.6219132035970687, 'recon_cos': 0.3358225546777248, 'gradnorm/compressor.query': 0.014561332566946531, 'gradnorm/compressor.layer0': 0.2474247497256685, 'gradnorm/compressor.layer1': 0.3000891885701756, 'gradnorm/compressor.layer2': 0.2596808703984439, 'gradnorm/compressor.layer3': 0.2343635556287927, 'gradnorm/compressor.layer4': 0.22170186877084638, 'gradnorm/compressor.layer5': 0.20765990053847388, 'gradnorm/compressor.layer6': 0.20312594450217517, 'gradnorm/compressor.layer7': 0.1918372342907281, 'gradnorm/decoder.layer0': 0.22923400640909816, 'gradnorm/decoder.layer1': 0.21228069461124838, 'gradnorm/decoder.layer2': 0.25245526273908026, 'gradnorm/max_param': 0.22065259516239166, 'epoch': 0.7}

  7%|▋         | 1330/19090 [39:10<9:29:25,  1.92s/it]
  7%|▋         | 1331/19090 [39:11<8:33:23,  1.73s/it]
  7%|▋         | 1332/19090 [39:12<7:54:20,  1.60s/it]
  7%|▋         | 1333/19090 [39:15<9:28:01,  1.92s/it]
  7%|▋         | 1334/19090 [39:16<8:32:48,  1.73s/it]
  7%|▋         | 1335/19090 [39:18<7:54:35,  1.60s/it]
  7%|▋         | 1336/19090 [39:20<8:38:52,  1.75s/it]
  7%|▋         | 1337/19090 [39:21<7:59:01,  1.62s/it]
  7%|▋         | 1338/19090 [39:22<7:30:47,  1.52s/it]
  7%|▋         | 1339/19090 [39:25<9:31:12,  1.93s/it]
  7%|▋         | 1340/19090 [39:27<8:35:24,  1.74s/it]
                                                      
{'loss': 0.0062, 'grad_norm': 0.7552667260169983, 'learning_rate': 3.5070717653221586e-05, 'recon_loss': 0.006228282395750284, 'recon_static': 0.005185846413951367, 'recon_motion': 0.0010424359061289579, 'recon_norm': 2.6753057301044465, 'recon_norm_std': 0.08551903888583183, 'target_norm': 1.6184355825185777, 'recon_cos': 0.323114537447691, 'gradnorm/compressor.query': 0.013710803146057248, 'gradnorm/compressor.layer0': 0.24151933700187722, 'gradnorm/compressor.layer1': 0.2870597926293924, 'gradnorm/compressor.layer2': 0.24227961972486578, 'gradnorm/compressor.layer3': 0.2229690195414806, 'gradnorm/compressor.layer4': 0.21534126936630082, 'gradnorm/compressor.layer5': 0.20508648991554881, 'gradnorm/compressor.layer6': 0.19569002857664095, 'gradnorm/compressor.layer7': 0.18652537479692757, 'gradnorm/decoder.layer0': 0.22827496041491938, 'gradnorm/decoder.layer1': 0.21214212723241058, 'gradnorm/decoder.layer2': 0.24991594331981573, 'gradnorm/max_param': 0.2154691368341446, 'epoch': 0.7}

  7%|▋         | 1340/19090 [39:27<8:35:24,  1.74s/it]
  7%|▋         | 1341/19090 [39:28<7:55:59,  1.61s/it]
  7%|▋         | 1342/19090 [39:30<9:12:18,  1.87s/it]
  7%|▋         | 1343/19090 [39:32<8:21:37,  1.70s/it]
  7%|▋         | 1344/19090 [39:33<7:47:50,  1.58s/it]
  7%|▋         | 1345/19090 [39:36<9:32:13,  1.93s/it]
  7%|▋         | 1346/19090 [39:37<8:36:32,  1.75s/it]
  7%|▋         | 1347/19090 [39:38<8:00:06,  1.62s/it]
  7%|▋         | 1348/19090 [39:41<10:08:38,  2.06s/it]
  7%|▋         | 1349/19090 [39:43<9:01:27,  1.83s/it] 
  7%|▋         | 1350/19090 [39:44<8:15:33,  1.68s/it]
                                                      
{'loss': 0.005, 'grad_norm': 0.7043073177337646, 'learning_rate': 3.5332634887375596e-05, 'recon_loss': 0.0049722045194357635, 'recon_static': 0.003924078476848081, 'recon_motion': 0.0010481261153472588, 'recon_norm': 2.3933006167411803, 'recon_norm_std': 0.09199483022093773, 'target_norm': 1.6215553402900695, 'recon_cos': 0.35836979672312735, 'gradnorm/compressor.query': 0.012703281546969177, 'gradnorm/compressor.layer0': 0.21351107905517877, 'gradnorm/compressor.layer1': 0.27166519729727856, 'gradnorm/compressor.layer2': 0.23274107259772914, 'gradnorm/compressor.layer3': 0.2121766305170417, 'gradnorm/compressor.layer4': 0.20007917245265222, 'gradnorm/compressor.layer5': 0.1887345401483068, 'gradnorm/compressor.layer6': 0.18007266187345483, 'gradnorm/compressor.layer7': 0.17310708227807653, 'gradnorm/decoder.layer0': 0.21047143309992092, 'gradnorm/decoder.layer1': 0.19739219646240502, 'gradnorm/decoder.layer2': 0.23665286837541502, 'gradnorm/max_param': 0.1898050159215927, 'epoch': 0.71}

  7%|▋         | 1350/19090 [39:44<8:15:33,  1.68s/it]
  7%|▋         | 1351/19090 [39:47<9:42:09,  1.97s/it]
  7%|▋         | 1352/19090 [39:48<8:42:49,  1.77s/it]
  7%|▋         | 1353/19090 [39:49<8:01:07,  1.63s/it]
  7%|▋         | 1354/19090 [39:52<9:30:41,  1.93s/it]
  7%|▋         | 1355/19090 [39:53<8:34:17,  1.74s/it]
  7%|▋         | 1356/19090 [39:54<7:54:42,  1.61s/it]
  7%|▋         | 1357/19090 [39:57<9:26:48,  1.92s/it]
  7%|▋         | 1358/19090 [39:58<8:30:57,  1.73s/it]
  7%|▋         | 1359/19090 [40:00<7:52:23,  1.60s/it]
  7%|▋         | 1360/19090 [40:03<9:43:08,  1.97s/it]
                                                      
{'loss': 0.0042, 'grad_norm': 0.5459851026535034, 'learning_rate': 3.55945521215296e-05, 'recon_loss': 0.004170341388089582, 'recon_static': 0.003133284149225801, 'recon_motion': 0.001037057297071442, 'recon_norm': 2.1937713027000427, 'recon_norm_std': 0.09808938167989253, 'target_norm': 1.6172897070646286, 'recon_cos': 0.3931089036166668, 'gradnorm/compressor.query': 0.00944072759696294, 'gradnorm/compressor.layer0': 0.1623131264375945, 'gradnorm/compressor.layer1': 0.20533048452286246, 'gradnorm/compressor.layer2': 0.17670319171654308, 'gradnorm/compressor.layer3': 0.1640552905178936, 'gradnorm/compressor.layer4': 0.15661323169898428, 'gradnorm/compressor.layer5': 0.14943198177618452, 'gradnorm/compressor.layer6': 0.144268544611582, 'gradnorm/compressor.layer7': 0.13708451254848614, 'gradnorm/decoder.layer0': 0.16606842211248452, 'gradnorm/decoder.layer1': 0.1538229193903679, 'gradnorm/decoder.layer2': 0.18347453243959083, 'gradnorm/max_param': 0.14473029971122742, 'epoch': 0.71}

  7%|▋         | 1360/19090 [40:03<9:43:08,  1.97s/it]
  7%|▋         | 1361/19090 [40:04<8:43:04,  1.77s/it]
  7%|▋         | 1362/19090 [40:05<8:00:26,  1.63s/it]
  7%|▋         | 1363/19090 [40:08<9:24:09,  1.91s/it]
  7%|▋         | 1364/19090 [40:09<8:29:11,  1.72s/it]
  7%|▋         | 1365/19090 [40:10<7:51:05,  1.59s/it]
  7%|▋         | 1366/19090 [40:13<9:20:58,  1.90s/it]
  7%|▋         | 1367/19090 [40:14<8:27:21,  1.72s/it]
  7%|▋         | 1368/19090 [40:15<7:49:00,  1.59s/it]
  7%|▋         | 1369/19090 [40:18<9:29:57,  1.93s/it]
  7%|▋         | 1370/19090 [40:20<8:33:31,  1.74s/it]
                                                      
{'loss': 0.0043, 'grad_norm': 0.598405122756958, 'learning_rate': 3.585646935568361e-05, 'recon_loss': 0.004267817939398811, 'recon_static': 0.003216553828679025, 'recon_motion': 0.0010512640816159546, 'recon_norm': 2.2159342169761658, 'recon_norm_std': 0.09546761736273765, 'target_norm': 1.621813178062439, 'recon_cos': 0.38758178874850274, 'gradnorm/compressor.query': 0.010921305056473148, 'gradnorm/compressor.layer0': 0.18930109550692595, 'gradnorm/compressor.layer1': 0.2325349260318263, 'gradnorm/compressor.layer2': 0.19674037652370743, 'gradnorm/compressor.layer3': 0.17923415048841412, 'gradnorm/compressor.layer4': 0.16862817998712615, 'gradnorm/compressor.layer5': 0.15912037897142592, 'gradnorm/compressor.layer6': 0.1533452092749048, 'gradnorm/compressor.layer7': 0.14590806763309955, 'gradnorm/decoder.layer0': 0.17659030796599, 'gradnorm/decoder.layer1': 0.16615123259183026, 'gradnorm/decoder.layer2': 0.1997029646853505, 'gradnorm/max_param': 0.1685769110918045, 'epoch': 0.72}

  7%|▋         | 1370/19090 [40:20<8:33:31,  1.74s/it]
  7%|▋         | 1371/19090 [40:21<7:53:31,  1.60s/it]
  7%|▋         | 1372/19090 [40:23<9:20:32,  1.90s/it]
  7%|▋         | 1373/19090 [40:25<8:26:45,  1.72s/it]
  7%|▋         | 1374/19090 [40:26<7:51:53,  1.60s/it]
  7%|▋         | 1375/19090 [40:28<9:05:48,  1.85s/it]
  7%|▋         | 1376/19090 [40:30<8:16:28,  1.68s/it]
  7%|▋         | 1377/19090 [40:31<7:42:08,  1.57s/it]
  7%|▋         | 1378/19090 [40:33<8:50:23,  1.80s/it]
  7%|▋         | 1379/19090 [40:35<8:13:54,  1.67s/it]
  7%|▋         | 1380/19090 [40:36<7:40:17,  1.56s/it]
                                                      
{'loss': 0.0053, 'grad_norm': 0.9065115451812744, 'learning_rate': 3.611838658983761e-05, 'recon_loss': 0.005289898777846247, 'recon_static': 0.004245778708718717, 'recon_motion': 0.0010441200487548486, 'recon_norm': 2.456381505727768, 'recon_norm_std': 0.08944546356797219, 'target_norm': 1.6201784282922744, 'recon_cos': 0.35160511657595633, 'gradnorm/compressor.query': 0.017267333253386932, 'gradnorm/compressor.layer0': 0.2899897626599048, 'gradnorm/compressor.layer1': 0.3479952605879812, 'gradnorm/compressor.layer2': 0.2947737340673634, 'gradnorm/compressor.layer3': 0.27283094433356275, 'gradnorm/compressor.layer4': 0.25593533735095153, 'gradnorm/compressor.layer5': 0.24448276197479527, 'gradnorm/compressor.layer6': 0.23413956261012905, 'gradnorm/compressor.layer7': 0.22117156556598627, 'gradnorm/decoder.layer0': 0.27077371907093667, 'gradnorm/decoder.layer1': 0.2471300889631096, 'gradnorm/decoder.layer2': 0.30255144806857187, 'gradnorm/max_param': 0.25840213894844055, 'epoch': 0.72}

  7%|▋         | 1380/19090 [40:36<7:40:17,  1.56s/it]
  7%|▋         | 1381/19090 [40:38<8:57:01,  1.82s/it]
  7%|▋         | 1382/19090 [40:40<8:22:21,  1.70s/it]
  7%|▋         | 1383/19090 [40:41<7:46:18,  1.58s/it]
  7%|▋         | 1384/19090 [40:44<9:39:29,  1.96s/it]
  7%|▋         | 1385/19090 [40:45<8:40:01,  1.76s/it]
  7%|▋         | 1386/19090 [40:47<7:58:50,  1.62s/it]
  7%|▋         | 1387/19090 [40:49<9:44:58,  1.98s/it]
  7%|▋         | 1388/19090 [40:51<8:42:56,  1.77s/it]
  7%|▋         | 1389/19090 [40:52<7:59:50,  1.63s/it]
  7%|▋         | 1390/19090 [40:55<10:10:47,  2.07s/it]
                                                       
{'loss': 0.0066, 'grad_norm': 0.6523158550262451, 'learning_rate': 3.638030382399162e-05, 'recon_loss': 0.006605258747003973, 'recon_static': 0.005562719417503104, 'recon_motion': 0.0010425393367768266, 'recon_norm': 2.7532978951931, 'recon_norm_std': 0.0804345715790987, 'target_norm': 1.620199254155159, 'recon_cos': 0.3158704608678818, 'gradnorm/compressor.query': 0.011671025606057388, 'gradnorm/compressor.layer0': 0.20441322623908645, 'gradnorm/compressor.layer1': 0.24652466059372125, 'gradnorm/compressor.layer2': 0.21490119686234976, 'gradnorm/compressor.layer3': 0.1946618400153221, 'gradnorm/compressor.layer4': 0.18533279717884393, 'gradnorm/compressor.layer5': 0.1773162681863602, 'gradnorm/compressor.layer6': 0.1697850224349485, 'gradnorm/compressor.layer7': 0.15818190137940133, 'gradnorm/decoder.layer0': 0.19421040491726016, 'gradnorm/decoder.layer1': 0.18146811260941131, 'gradnorm/decoder.layer2': 0.2201333413208947, 'gradnorm/max_param': 0.18179461359977722, 'epoch': 0.73}

  7%|▋         | 1390/19090 [40:55<10:10:47,  2.07s/it]
  7%|▋         | 1391/19090 [40:56<9:01:55,  1.84s/it] 
  7%|▋         | 1392/19090 [40:58<8:13:24,  1.67s/it]
  7%|▋         | 1393/19090 [41:01<10:14:34,  2.08s/it]
  7%|▋         | 1394/19090 [41:02<9:03:39,  1.84s/it] 
  7%|▋         | 1395/19090 [41:03<8:15:03,  1.68s/it]
  7%|▋         | 1396/19090 [41:06<9:35:15,  1.95s/it]
  7%|▋         | 1397/19090 [41:07<8:36:28,  1.75s/it]
  7%|▋         | 1398/19090 [41:08<7:55:23,  1.61s/it]
  7%|▋         | 1399/19090 [41:11<9:21:57,  1.91s/it]
  7%|▋         | 1400/19090 [41:12<8:27:24,  1.72s/it]
                                                      
{'loss': 0.0055, 'grad_norm': 0.8457407355308533, 'learning_rate': 3.6642221058145624e-05, 'recon_loss': 0.005530137452296913, 'recon_static': 0.004489117325283587, 'recon_motion': 0.00104102015029639, 'recon_norm': 2.5286915063858033, 'recon_norm_std': 0.08508742433041334, 'target_norm': 1.6199699223041535, 'recon_cos': 0.3435759410262108, 'gradnorm/compressor.query': 0.015243498770649752, 'gradnorm/compressor.layer0': 0.2640725376546307, 'gradnorm/compressor.layer1': 0.31974214854901833, 'gradnorm/compressor.layer2': 0.27386574588777624, 'gradnorm/compressor.layer3': 0.25372464948312917, 'gradnorm/compressor.layer4': 0.24067556760209277, 'gradnorm/compressor.layer5': 0.2290799842162141, 'gradnorm/compressor.layer6': 0.22045933078134555, 'gradnorm/compressor.layer7': 0.2096970874197773, 'gradnorm/decoder.layer0': 0.2521899832516996, 'gradnorm/decoder.layer1': 0.2359640927517596, 'gradnorm/decoder.layer2': 0.2853781607449071, 'gradnorm/max_param': 0.23543934524059296, 'epoch': 0.73}

  7%|▋         | 1400/19090 [41:12<8:27:24,  1.72s/it]
  7%|▋         | 1401/19090 [41:14<7:50:08,  1.59s/it]
  7%|▋         | 1402/19090 [41:17<9:41:34,  1.97s/it]
  7%|▋         | 1403/19090 [41:18<8:41:17,  1.77s/it]
  7%|▋         | 1404/19090 [41:19<7:59:52,  1.63s/it]
  7%|▋         | 1405/19090 [41:22<9:35:22,  1.95s/it]
  7%|▋         | 1406/19090 [41:23<8:36:33,  1.75s/it]
  7%|▋         | 1407/19090 [41:24<7:55:11,  1.61s/it]
  7%|▋         | 1408/19090 [41:27<9:36:10,  1.96s/it]
  7%|▋         | 1409/19090 [41:28<8:38:05,  1.76s/it]
  7%|▋         | 1410/19090 [41:30<7:57:25,  1.62s/it]
                                                      
{'loss': 0.0051, 'grad_norm': 0.6172212958335876, 'learning_rate': 3.6904138292299634e-05, 'recon_loss': 0.005060994869563728, 'recon_static': 0.004027387272799388, 'recon_motion': 0.001033607532735914, 'recon_norm': 2.4158828020095826, 'recon_norm_std': 0.09005902502685785, 'target_norm': 1.6192341476678849, 'recon_cos': 0.35739707201719284, 'gradnorm/compressor.query': 0.011010350679787456, 'gradnorm/compressor.layer0': 0.1883187992594575, 'gradnorm/compressor.layer1': 0.23693494681117505, 'gradnorm/compressor.layer2': 0.2005978421965903, 'gradnorm/compressor.layer3': 0.1840518046284917, 'gradnorm/compressor.layer4': 0.17389649214706793, 'gradnorm/compressor.layer5': 0.16464192387739268, 'gradnorm/compressor.layer6': 0.15876050765075372, 'gradnorm/compressor.layer7': 0.1514723320296029, 'gradnorm/decoder.layer0': 0.1842678862117347, 'gradnorm/decoder.layer1': 0.1754668921680557, 'gradnorm/decoder.layer2': 0.21186139731219517, 'gradnorm/max_param': 0.1676645576953888, 'epoch': 0.74}

  7%|▋         | 1410/19090 [41:30<7:57:25,  1.62s/it]
  7%|▋         | 1411/19090 [41:33<9:40:24,  1.97s/it]
  7%|▋         | 1412/19090 [41:34<8:41:01,  1.77s/it]
  7%|▋         | 1413/19090 [41:35<7:59:22,  1.63s/it]
  7%|▋         | 1414/19090 [41:38<9:23:16,  1.91s/it]
  7%|▋         | 1415/19090 [41:39<8:28:41,  1.73s/it]
  7%|▋         | 1416/19090 [41:40<7:49:56,  1.60s/it]
  7%|▋         | 1417/19090 [41:43<9:12:01,  1.87s/it]
  7%|▋         | 1418/19090 [41:44<8:19:50,  1.70s/it]
  7%|▋         | 1419/19090 [41:45<7:43:32,  1.57s/it]
  7%|▋         | 1420/19090 [41:48<9:18:06,  1.90s/it]
                                                      
{'loss': 0.0043, 'grad_norm': 0.6705796122550964, 'learning_rate': 3.716605552645364e-05, 'recon_loss': 0.00429275311762467, 'recon_static': 0.0032581039355136453, 'recon_motion': 0.0010346491559175774, 'recon_norm': 2.2252093315124513, 'recon_norm_std': 0.09590803179889917, 'target_norm': 1.6209630072116852, 'recon_cos': 0.3860219232738018, 'gradnorm/compressor.query': 0.012170370005251542, 'gradnorm/compressor.layer0': 0.20632751987632866, 'gradnorm/compressor.layer1': 0.25947263155568456, 'gradnorm/compressor.layer2': 0.21899645409696353, 'gradnorm/compressor.layer3': 0.20293361560885811, 'gradnorm/compressor.layer4': 0.19034965985445373, 'gradnorm/compressor.layer5': 0.18056075776891337, 'gradnorm/compressor.layer6': 0.171458449391673, 'gradnorm/compressor.layer7': 0.16251033007359436, 'gradnorm/decoder.layer0': 0.19896109131371104, 'gradnorm/decoder.layer1': 0.18756335896654733, 'gradnorm/decoder.layer2': 0.2260131918350618, 'gradnorm/max_param': 0.18425093591213226, 'epoch': 0.74}

  7%|▋         | 1420/19090 [41:48<9:18:06,  1.90s/it]
  7%|▋         | 1421/19090 [41:49<8:24:35,  1.71s/it]
  7%|▋         | 1422/19090 [41:51<7:46:42,  1.58s/it]
  7%|▋         | 1423/19090 [41:53<9:34:47,  1.95s/it]
  7%|▋         | 1424/19090 [41:55<8:36:14,  1.75s/it]
  7%|▋         | 1425/19090 [41:56<7:55:55,  1.62s/it]
  7%|▋         | 1426/19090 [41:59<9:29:06,  1.93s/it]
  7%|▋         | 1427/19090 [42:00<8:32:22,  1.74s/it]
  7%|▋         | 1428/19090 [42:01<8:09:34,  1.66s/it]
  7%|▋         | 1429/19090 [42:04<9:24:43,  1.92s/it]
  7%|▋         | 1430/19090 [42:05<8:30:41,  1.74s/it]
                                                      
{'loss': 0.0064, 'grad_norm': 0.8788084387779236, 'learning_rate': 3.7427972760607646e-05, 'recon_loss': 0.00639954200014472, 'recon_static': 0.005359479028265923, 'recon_motion': 0.0010400629311334343, 'recon_norm': 2.7159339427947997, 'recon_norm_std': 0.0798434080556035, 'target_norm': 1.6210068255662917, 'recon_cos': 0.3178526572883129, 'gradnorm/compressor.query': 0.01569256386909723, 'gradnorm/compressor.layer0': 0.27053644187169346, 'gradnorm/compressor.layer1': 0.3397911074852255, 'gradnorm/compressor.layer2': 0.28521528024946347, 'gradnorm/compressor.layer3': 0.2613090468258119, 'gradnorm/compressor.layer4': 0.24691003568938716, 'gradnorm/compressor.layer5': 0.23436089314286254, 'gradnorm/compressor.layer6': 0.22624432485533705, 'gradnorm/compressor.layer7': 0.21460744693752382, 'gradnorm/decoder.layer0': 0.2660877880519409, 'gradnorm/decoder.layer1': 0.2435820866203334, 'gradnorm/decoder.layer2': 0.300852915660588, 'gradnorm/max_param': 0.24158340692520142, 'epoch': 0.75}

  7%|▋         | 1430/19090 [42:05<8:30:41,  1.74s/it]
  7%|▋         | 1431/19090 [42:07<7:56:19,  1.62s/it]
  8%|▊         | 1432/19090 [42:09<9:18:37,  1.90s/it]
  8%|▊         | 1433/19090 [42:10<8:26:12,  1.72s/it]
  8%|▊         | 1434/19090 [42:12<8:28:37,  1.73s/it]
  8%|▊         | 1435/19090 [42:14<9:08:45,  1.86s/it]
  8%|▊         | 1436/19090 [42:16<8:17:11,  1.69s/it]
  8%|▊         | 1437/19090 [42:17<8:17:02,  1.69s/it]
  8%|▊         | 1438/19090 [42:20<9:31:33,  1.94s/it]
  8%|▊         | 1439/19090 [42:21<8:34:14,  1.75s/it]
  8%|▊         | 1440/19090 [42:23<8:07:42,  1.66s/it]
                                                      
{'loss': 0.0064, 'grad_norm': 0.8025550842285156, 'learning_rate': 3.7689889994761656e-05, 'recon_loss': 0.006385443545877933, 'recon_static': 0.005345211550593376, 'recon_motion': 0.0010402319545391946, 'recon_norm': 2.7112515568733215, 'recon_norm_std': 0.08043378796428442, 'target_norm': 1.6219436943531036, 'recon_cos': 0.3165335915982723, 'gradnorm/compressor.query': 0.012803780627001006, 'gradnorm/compressor.layer0': 0.2375057216632689, 'gradnorm/compressor.layer1': 0.2959137084847633, 'gradnorm/compressor.layer2': 0.25738012296516444, 'gradnorm/compressor.layer3': 0.2390302003594801, 'gradnorm/compressor.layer4': 0.2274615765706249, 'gradnorm/compressor.layer5': 0.21938787000078402, 'gradnorm/compressor.layer6': 0.20939105563023047, 'gradnorm/compressor.layer7': 0.1979836624508405, 'gradnorm/decoder.layer0': 0.24551235883656367, 'gradnorm/decoder.layer1': 0.2332071054780252, 'gradnorm/decoder.layer2': 0.2810748800164299, 'gradnorm/max_param': 0.2123315930366516, 'epoch': 0.75}

  8%|▊         | 1440/19090 [42:23<8:07:42,  1.66s/it]
  8%|▊         | 1441/19090 [42:25<9:39:42,  1.97s/it]
  8%|▊         | 1442/19090 [42:27<8:39:55,  1.77s/it]
  8%|▊         | 1443/19090 [42:28<8:37:59,  1.76s/it]
  8%|▊         | 1444/19090 [42:30<8:57:50,  1.83s/it]
  8%|▊         | 1445/19090 [42:32<8:27:52,  1.73s/it]
  8%|▊         | 1446/19090 [42:34<8:34:32,  1.75s/it]
  8%|▊         | 1447/19090 [42:35<8:40:34,  1.77s/it]
  8%|▊         | 1448/19090 [42:37<8:55:42,  1.82s/it]
  8%|▊         | 1449/19090 [42:39<8:32:15,  1.74s/it]
  8%|▊         | 1450/19090 [42:40<8:04:15,  1.65s/it]
                                                      
{'loss': 0.0066, 'grad_norm': 0.8680099248886108, 'learning_rate': 3.7951807228915666e-05, 'recon_loss': 0.006608967564534396, 'recon_static': 0.005579537490848452, 'recon_motion': 0.0010294299354427494, 'recon_norm': 2.7656255304813384, 'recon_norm_std': 0.07913757245987654, 'target_norm': 1.6193051517009736, 'recon_cos': 0.3147197440266609, 'gradnorm/compressor.query': 0.01585736663232938, 'gradnorm/compressor.layer0': 0.27324697123573943, 'gradnorm/compressor.layer1': 0.32213622298431427, 'gradnorm/compressor.layer2': 0.2803514750760758, 'gradnorm/compressor.layer3': 0.2558430964991693, 'gradnorm/compressor.layer4': 0.24621061547470277, 'gradnorm/compressor.layer5': 0.23475622326500434, 'gradnorm/compressor.layer6': 0.22515027802679122, 'gradnorm/compressor.layer7': 0.21412456823841042, 'gradnorm/decoder.layer0': 0.25960894570557663, 'gradnorm/decoder.layer1': 0.24582475008096893, 'gradnorm/decoder.layer2': 0.30100279241356526, 'gradnorm/max_param': 0.24317514896392822, 'epoch': 0.76}

  8%|▊         | 1450/19090 [42:40<8:04:15,  1.65s/it]
  8%|▊         | 1451/19090 [42:43<9:14:56,  1.89s/it]
  8%|▊         | 1452/19090 [42:44<8:32:39,  1.74s/it]
  8%|▊         | 1453/19090 [42:46<7:53:56,  1.61s/it]
  8%|▊         | 1454/19090 [42:48<9:15:45,  1.89s/it]
  8%|▊         | 1455/19090 [42:50<8:39:12,  1.77s/it]
  8%|▊         | 1456/19090 [42:51<8:01:22,  1.64s/it]
  8%|▊         | 1457/19090 [42:54<9:27:04,  1.93s/it]
  8%|▊         | 1458/19090 [42:55<9:03:50,  1.85s/it]
  8%|▊         | 1459/19090 [42:57<8:43:09,  1.78s/it]
  8%|▊         | 1460/19090 [42:59<9:09:53,  1.87s/it]
                                                      
{'loss': 0.0055, 'grad_norm': 0.6155405640602112, 'learning_rate': 3.8213724463069675e-05, 'recon_loss': 0.0055414574220776554, 'recon_static': 0.004511196602834389, 'recon_motion': 0.0010302608643542044, 'recon_norm': 2.5222962737083434, 'recon_norm_std': 0.08502570111304522, 'target_norm': 1.6202088385820388, 'recon_cos': 0.3427284970879555, 'gradnorm/compressor.query': 0.010949758754935519, 'gradnorm/compressor.layer0': 0.18561288763595757, 'gradnorm/compressor.layer1': 0.2303625201432525, 'gradnorm/compressor.layer2': 0.19750468244011726, 'gradnorm/compressor.layer3': 0.1861204417421816, 'gradnorm/compressor.layer4': 0.17273797327794305, 'gradnorm/compressor.layer5': 0.16453666513267848, 'gradnorm/compressor.layer6': 0.1602659721876551, 'gradnorm/compressor.layer7': 0.15247442680849405, 'gradnorm/decoder.layer0': 0.18599631437180195, 'gradnorm/decoder.layer1': 0.17416293336340305, 'gradnorm/decoder.layer2': 0.21640391434603745, 'gradnorm/max_param': 0.16655877232551575, 'epoch': 0.76}

  8%|▊         | 1460/19090 [42:59<9:09:53,  1.87s/it]
  8%|▊         | 1461/19090 [43:00<8:18:46,  1.70s/it]
  8%|▊         | 1462/19090 [43:02<8:28:35,  1.73s/it]
  8%|▊         | 1463/19090 [43:04<9:06:58,  1.86s/it]
  8%|▊         | 1464/19090 [43:05<8:16:51,  1.69s/it]
  8%|▊         | 1465/19090 [43:07<8:23:44,  1.71s/it]
  8%|▊         | 1466/19090 [43:09<9:06:08,  1.86s/it]
  8%|▊         | 1467/19090 [43:11<8:24:44,  1.72s/it]
  8%|▊         | 1468/19090 [43:12<8:05:18,  1.65s/it]
  8%|▊         | 1469/19090 [43:15<9:05:52,  1.86s/it]
  8%|▊         | 1470/19090 [43:16<8:16:01,  1.69s/it]
                                                      
{'loss': 0.0044, 'grad_norm': 0.6639971137046814, 'learning_rate': 3.847564169722368e-05, 'recon_loss': 0.004436450020875782, 'recon_static': 0.003404115611920133, 'recon_motion': 0.0010323343187337742, 'recon_norm': 2.2643991351127624, 'recon_norm_std': 0.09348534885793924, 'target_norm': 1.623032009601593, 'recon_cos': 0.38099348470568656, 'gradnorm/compressor.query': 0.011419204353458346, 'gradnorm/compressor.layer0': 0.20146411668181222, 'gradnorm/compressor.layer1': 0.2520709834081759, 'gradnorm/compressor.layer2': 0.2149087980806327, 'gradnorm/compressor.layer3': 0.19416902553131016, 'gradnorm/compressor.layer4': 0.1859670799832943, 'gradnorm/compressor.layer5': 0.17868978865092802, 'gradnorm/compressor.layer6': 0.17142842537939004, 'gradnorm/compressor.layer7': 0.1627945724252707, 'gradnorm/decoder.layer0': 0.20224821236024323, 'gradnorm/decoder.layer1': 0.18944055967839674, 'gradnorm/decoder.layer2': 0.23141110646305244, 'gradnorm/max_param': 0.18022577464580536, 'epoch': 0.77}

  8%|▊         | 1470/19090 [43:16<8:16:01,  1.69s/it]
  8%|▊         | 1471/19090 [43:18<8:19:16,  1.70s/it]
  8%|▊         | 1472/19090 [43:20<8:45:58,  1.79s/it]
  8%|▊         | 1473/19090 [43:21<8:19:40,  1.70s/it]
  8%|▊         | 1474/19090 [43:23<8:33:00,  1.75s/it]
  8%|▊         | 1475/19090 [43:25<8:50:49,  1.81s/it]
  8%|▊         | 1476/19090 [43:26<8:06:37,  1.66s/it]
  8%|▊         | 1477/19090 [43:28<8:31:57,  1.74s/it]
  8%|▊         | 1478/19090 [43:30<9:07:11,  1.86s/it]
  8%|▊         | 1479/19090 [43:32<8:16:18,  1.69s/it]
  8%|▊         | 1480/19090 [43:34<9:16:03,  1.89s/it]
                                                      
{'loss': 0.0071, 'grad_norm': 0.9252257943153381, 'learning_rate': 3.873755893137769e-05, 'recon_loss': 0.007077370828483254, 'recon_static': 0.006040007632691413, 'recon_motion': 0.0010373632845585234, 'recon_norm': 2.8508124589920043, 'recon_norm_std': 0.07772080786526203, 'target_norm': 1.6222381919622422, 'recon_cos': 0.3049896441400051, 'gradnorm/compressor.query': 0.015713114501266272, 'gradnorm/compressor.layer0': 0.27918608008442714, 'gradnorm/compressor.layer1': 0.3487344481476969, 'gradnorm/compressor.layer2': 0.29809172906499487, 'gradnorm/compressor.layer3': 0.2767700251843981, 'gradnorm/compressor.layer4': 0.2631577601165804, 'gradnorm/compressor.layer5': 0.2517042631341066, 'gradnorm/compressor.layer6': 0.23886949595773027, 'gradnorm/compressor.layer7': 0.227725439730626, 'gradnorm/decoder.layer0': 0.27742485418629553, 'gradnorm/decoder.layer1': 0.2575949126382451, 'gradnorm/decoder.layer2': 0.32546998071827005, 'gradnorm/max_param': 0.2491787075996399, 'epoch': 0.78}

  8%|▊         | 1480/19090 [43:34<9:16:03,  1.89s/it]
  8%|▊         | 1481/19090 [43:36<8:59:58,  1.84s/it]
  8%|▊         | 1482/19090 [43:37<8:10:51,  1.67s/it]
  8%|▊         | 1483/19090 [43:39<9:11:52,  1.88s/it]
  8%|▊         | 1484/19090 [43:41<8:49:24,  1.80s/it]
  8%|▊         | 1485/19090 [43:42<8:04:18,  1.65s/it]
  8%|▊         | 1486/19090 [43:45<9:16:18,  1.90s/it]
  8%|▊         | 1487/19090 [43:46<8:49:51,  1.81s/it]
  8%|▊         | 1488/19090 [43:48<8:03:32,  1.65s/it]
  8%|▊         | 1489/19090 [43:50<9:11:16,  1.88s/it]
  8%|▊         | 1490/19090 [43:52<8:51:24,  1.81s/it]
                                                      
{'loss': 0.006, 'grad_norm': 0.7218790650367737, 'learning_rate': 3.89994761655317e-05, 'recon_loss': 0.006008335098158568, 'recon_static': 0.004984969121869654, 'recon_motion': 0.0010233659850200638, 'recon_norm': 2.6385974884033203, 'recon_norm_std': 0.08264822121709585, 'target_norm': 1.617264524102211, 'recon_cos': 0.33144270107150076, 'gradnorm/compressor.query': 0.012341228979138794, 'gradnorm/compressor.layer0': 0.21657725542616477, 'gradnorm/compressor.layer1': 0.26913078245815686, 'gradnorm/compressor.layer2': 0.23215372982344062, 'gradnorm/compressor.layer3': 0.21436382790083627, 'gradnorm/compressor.layer4': 0.2035412741110331, 'gradnorm/compressor.layer5': 0.19299858804665587, 'gradnorm/compressor.layer6': 0.18653075718123588, 'gradnorm/compressor.layer7': 0.17961451872785542, 'gradnorm/decoder.layer0': 0.21726561040511316, 'gradnorm/decoder.layer1': 0.20634265862915338, 'gradnorm/decoder.layer2': 0.2574072600689079, 'gradnorm/max_param': 0.19338537752628326, 'epoch': 0.78}

  8%|▊         | 1490/19090 [43:52<8:51:24,  1.81s/it]
  8%|▊         | 1491/19090 [43:53<8:21:11,  1.71s/it]
  8%|▊         | 1492/19090 [43:55<8:59:41,  1.84s/it]
  8%|▊         | 1493/19090 [43:57<8:28:33,  1.73s/it]
  8%|▊         | 1494/19090 [43:58<7:50:33,  1.60s/it]
  8%|▊         | 1495/19090 [44:00<8:56:09,  1.83s/it]
  8%|▊         | 1496/19090 [44:03<9:14:15,  1.89s/it]
  8%|▊         | 1497/19090 [44:04<8:21:43,  1.71s/it]
  8%|▊         | 1498/19090 [44:06<8:39:24,  1.77s/it]
  8%|▊         | 1499/19090 [44:08<9:06:54,  1.87s/it]
  8%|▊         | 1500/19090 [44:09<8:16:50,  1.69s/it]
                                                      
{'loss': 0.0049, 'grad_norm': 0.5996567606925964, 'learning_rate': 3.92613933996857e-05, 'recon_loss': 0.004930301499553025, 'recon_static': 0.003893779718782753, 'recon_motion': 0.0010365217123762705, 'recon_norm': 2.3787149786949158, 'recon_norm_std': 0.08915890604257584, 'target_norm': 1.623704206943512, 'recon_cos': 0.3612747721374035, 'gradnorm/compressor.query': 0.01062283741434485, 'gradnorm/compressor.layer0': 0.17695553927427218, 'gradnorm/compressor.layer1': 0.2251020087692248, 'gradnorm/compressor.layer2': 0.19359775260439063, 'gradnorm/compressor.layer3': 0.17956098995775438, 'gradnorm/compressor.layer4': 0.16990537124814886, 'gradnorm/compressor.layer5': 0.1581213527850627, 'gradnorm/compressor.layer6': 0.15542512358248986, 'gradnorm/compressor.layer7': 0.14577700529021143, 'gradnorm/decoder.layer0': 0.1814236318982694, 'gradnorm/decoder.layer1': 0.17157967076770336, 'gradnorm/decoder.layer2': 0.21479462148580772, 'gradnorm/max_param': 0.15735787153244019, 'epoch': 0.79}

  8%|▊         | 1500/19090 [44:09<8:16:50,  1.69s/it]
  8%|▊         | 1501/19090 [44:13<11:35:02,  2.37s/it]
  8%|▊         | 1502/19090 [44:14<9:59:55,  2.05s/it] 
  8%|▊         | 1503/19090 [44:16<8:53:13,  1.82s/it]
  8%|▊         | 1504/19090 [44:17<8:06:47,  1.66s/it]
  8%|▊         | 1505/19090 [44:18<7:33:34,  1.55s/it]
  8%|▊         | 1506/19090 [44:19<7:10:09,  1.47s/it]
  8%|▊         | 1507/19090 [44:21<7:00:27,  1.43s/it]
  8%|▊         | 1508/19090 [44:23<8:32:30,  1.75s/it]
  8%|▊         | 1509/19090 [44:25<7:52:23,  1.61s/it]
  8%|▊         | 1510/19090 [44:26<7:27:54,  1.53s/it]
                                                      
{'loss': 0.0062, 'grad_norm': 1.0447163581848145, 'learning_rate': 3.952331063383971e-05, 'recon_loss': 0.006156505097169429, 'recon_static': 0.005124359583714977, 'recon_motion': 0.0010321457055397331, 'recon_norm': 2.6532158613204957, 'recon_norm_std': 0.08249881379306316, 'target_norm': 1.6202955365180969, 'recon_cos': 0.3278053220361471, 'gradnorm/compressor.query': 0.01855751301558999, 'gradnorm/compressor.layer0': 0.3223616948127712, 'gradnorm/compressor.layer1': 0.3944824112982925, 'gradnorm/compressor.layer2': 0.3399527692772337, 'gradnorm/compressor.layer3': 0.30826627280216384, 'gradnorm/compressor.layer4': 0.2945701699888246, 'gradnorm/compressor.layer5': 0.2762967077098448, 'gradnorm/compressor.layer6': 0.2675473411282867, 'gradnorm/compressor.layer7': 0.2525748337431822, 'gradnorm/decoder.layer0': 0.30889531381994356, 'gradnorm/decoder.layer1': 0.29659757702140654, 'gradnorm/decoder.layer2': 0.37294108703330475, 'gradnorm/max_param': 0.2871221899986267, 'epoch': 0.79}

  8%|▊         | 1510/19090 [44:26<7:27:54,  1.53s/it]
  8%|▊         | 1511/19090 [44:28<8:23:15,  1.72s/it]
  8%|▊         | 1512/19090 [44:29<7:45:35,  1.59s/it]
  8%|▊         | 1513/19090 [44:31<8:00:33,  1.64s/it]
  8%|▊         | 1514/19090 [44:33<8:52:24,  1.82s/it]
  8%|▊         | 1515/19090 [44:35<8:06:05,  1.66s/it]
  8%|▊         | 1516/19090 [44:37<8:31:56,  1.75s/it]
  8%|▊         | 1517/19090 [44:39<9:07:47,  1.87s/it]
  8%|▊         | 1518/19090 [44:40<8:16:24,  1.70s/it]
  8%|▊         | 1519/19090 [44:42<8:41:52,  1.78s/it]
  8%|▊         | 1520/19090 [44:44<9:01:58,  1.85s/it]
                                                      
{'loss': 0.0082, 'grad_norm': 1.0163358449935913, 'learning_rate': 3.978522786799372e-05, 'recon_loss': 0.008165398391429334, 'recon_static': 0.007123395556118339, 'recon_motion': 0.001042002961912658, 'recon_norm': 3.0711337625980377, 'recon_norm_std': 0.07206937074661254, 'target_norm': 1.6219095438718796, 'recon_cos': 0.2820118859410286, 'gradnorm/compressor.query': 0.01835607193975384, 'gradnorm/compressor.layer0': 0.3033973448650065, 'gradnorm/compressor.layer1': 0.3774318521136759, 'gradnorm/compressor.layer2': 0.3269970557848766, 'gradnorm/compressor.layer3': 0.30233039974338893, 'gradnorm/compressor.layer4': 0.2830066262293372, 'gradnorm/compressor.layer5': 0.270265394858336, 'gradnorm/compressor.layer6': 0.261435127304987, 'gradnorm/compressor.layer7': 0.2485417588112003, 'gradnorm/decoder.layer0': 0.3093567602001906, 'gradnorm/decoder.layer1': 0.29379306538404476, 'gradnorm/decoder.layer2': 0.36673229489954107, 'gradnorm/max_param': 0.2694046199321747, 'epoch': 0.8}

  8%|▊         | 1520/19090 [44:44<9:01:58,  1.85s/it]
  8%|▊         | 1521/19090 [44:45<8:15:17,  1.69s/it]
  8%|▊         | 1522/19090 [44:47<8:40:32,  1.78s/it]
  8%|▊         | 1523/19090 [44:49<8:59:11,  1.84s/it]
  8%|▊         | 1524/19090 [44:51<8:28:04,  1.74s/it]
  8%|▊         | 1525/19090 [44:53<8:36:32,  1.76s/it]
  8%|▊         | 1526/19090 [44:55<8:50:04,  1.81s/it]
  8%|▊         | 1527/19090 [44:56<8:38:59,  1.77s/it]
  8%|▊         | 1528/19090 [44:58<8:17:56,  1.70s/it]
  8%|▊         | 1529/19090 [45:00<9:23:54,  1.93s/it]
  8%|▊         | 1530/19090 [45:02<8:46:36,  1.80s/it]
                                                      
{'loss': 0.0081, 'grad_norm': 0.9404222369194031, 'learning_rate': 4.004714510214772e-05, 'recon_loss': 0.008088011748623104, 'recon_static': 0.007064879359677434, 'recon_motion': 0.0010231325475615449, 'recon_norm': 3.0602294385433195, 'recon_norm_std': 0.07360847909003496, 'target_norm': 1.6213877230882645, 'recon_cos': 0.28513662740588186, 'gradnorm/compressor.query': 0.015814369389299473, 'gradnorm/compressor.layer0': 0.27865330368200714, 'gradnorm/compressor.layer1': 0.35061550812614545, 'gradnorm/compressor.layer2': 0.30068451700691395, 'gradnorm/compressor.layer3': 0.28088044683066954, 'gradnorm/compressor.layer4': 0.2670158269723953, 'gradnorm/compressor.layer5': 0.2513761493593987, 'gradnorm/compressor.layer6': 0.2407678248371615, 'gradnorm/compressor.layer7': 0.23131742042321626, 'gradnorm/decoder.layer0': 0.28487390100184284, 'gradnorm/decoder.layer1': 0.266716635326367, 'gradnorm/decoder.layer2': 0.34059470068185255, 'gradnorm/max_param': 0.24894241988658905, 'epoch': 0.8}

  8%|▊         | 1530/19090 [45:02<8:46:36,  1.80s/it]
  8%|▊         | 1531/19090 [45:03<8:02:36,  1.65s/it]
  8%|▊         | 1532/19090 [45:06<9:22:35,  1.92s/it]
  8%|▊         | 1533/19090 [45:07<8:38:27,  1.77s/it]
  8%|▊         | 1534/19090 [45:09<8:22:57,  1.72s/it]
  8%|▊         | 1535/19090 [45:11<9:16:21,  1.90s/it]
  8%|▊         | 1536/19090 [45:12<8:29:55,  1.74s/it]
  8%|▊         | 1537/19090 [45:14<8:34:42,  1.76s/it]
  8%|▊         | 1538/19090 [45:16<9:17:57,  1.91s/it]
  8%|▊         | 1539/19090 [45:18<8:27:16,  1.73s/it]
  8%|▊         | 1540/19090 [45:20<8:48:45,  1.81s/it]
                                                      
{'loss': 0.0058, 'grad_norm': 0.6687783598899841, 'learning_rate': 4.0309062336301726e-05, 'recon_loss': 0.005757872003596276, 'recon_static': 0.004728598694782704, 'recon_motion': 0.0010292731094523333, 'recon_norm': 2.5765860617160796, 'recon_norm_std': 0.08364072013646365, 'target_norm': 1.6217761516571045, 'recon_cos': 0.33696088194847107, 'gradnorm/compressor.query': 0.011115129757893069, 'gradnorm/compressor.layer0': 0.19550612388895305, 'gradnorm/compressor.layer1': 0.24988743363771163, 'gradnorm/compressor.layer2': 0.21144657164938196, 'gradnorm/compressor.layer3': 0.1966873701578202, 'gradnorm/compressor.layer4': 0.18867128818333828, 'gradnorm/compressor.layer5': 0.1782086301536071, 'gradnorm/compressor.layer6': 0.17195582115177543, 'gradnorm/compressor.layer7': 0.16469579470373732, 'gradnorm/decoder.layer0': 0.2047034246496231, 'gradnorm/decoder.layer1': 0.19171169782568848, 'gradnorm/decoder.layer2': 0.24571002077211404, 'gradnorm/max_param': 0.17492467164993286, 'epoch': 0.81}

  8%|▊         | 1540/19090 [45:20<8:48:45,  1.81s/it]
  8%|▊         | 1541/19090 [45:22<9:10:36,  1.88s/it]
  8%|▊         | 1542/19090 [45:24<9:00:30,  1.85s/it]
  8%|▊         | 1543/19090 [45:25<8:27:05,  1.73s/it]
  8%|▊         | 1544/19090 [45:27<9:23:40,  1.93s/it]
  8%|▊         | 1545/19090 [45:29<9:04:41,  1.86s/it]
  8%|▊         | 1546/19090 [45:30<8:21:28,  1.72s/it]
  8%|▊         | 1547/19090 [45:33<9:20:07,  1.92s/it]
  8%|▊         | 1548/19090 [45:34<8:46:24,  1.80s/it]
  8%|▊         | 1549/19090 [45:36<8:04:44,  1.66s/it]
  8%|▊         | 1550/19090 [45:38<9:06:52,  1.87s/it]
                                                      
{'loss': 0.004, 'grad_norm': 0.5899640321731567, 'learning_rate': 4.0570979570455735e-05, 'recon_loss': 0.004000317846657708, 'recon_static': 0.0029762820166070016, 'recon_motion': 0.00102403574710479, 'recon_norm': 2.159656453132629, 'recon_norm_std': 0.0949087891727686, 'target_norm': 1.6196869343519211, 'recon_cos': 0.4045159675180912, 'gradnorm/compressor.query': 0.009844512745516532, 'gradnorm/compressor.layer0': 0.17602770270788076, 'gradnorm/compressor.layer1': 0.2189882292529095, 'gradnorm/compressor.layer2': 0.1875937446055363, 'gradnorm/compressor.layer3': 0.17297260442951673, 'gradnorm/compressor.layer4': 0.1662338953714939, 'gradnorm/compressor.layer5': 0.1575599772207083, 'gradnorm/compressor.layer6': 0.15111769141350978, 'gradnorm/compressor.layer7': 0.14631815698687695, 'gradnorm/decoder.layer0': 0.1793216110652476, 'gradnorm/decoder.layer1': 0.16908064146419194, 'gradnorm/decoder.layer2': 0.21554396703071318, 'gradnorm/max_param': 0.15649458765983582, 'epoch': 0.81}

  8%|▊         | 1550/19090 [45:38<9:06:52,  1.87s/it]
  8%|▊         | 1551/19090 [45:40<8:51:31,  1.82s/it]
  8%|▊         | 1552/19090 [45:41<8:17:36,  1.70s/it]
  8%|▊         | 1553/19090 [45:43<9:05:32,  1.87s/it]
  8%|▊         | 1554/19090 [45:45<8:17:35,  1.70s/it]
  8%|▊         | 1555/19090 [45:46<8:01:47,  1.65s/it]
  8%|▊         | 1556/19090 [45:49<8:57:32,  1.84s/it]
  8%|▊         | 1557/19090 [45:50<8:10:42,  1.68s/it]
  8%|▊         | 1558/19090 [45:52<9:03:01,  1.86s/it]
  8%|▊         | 1559/19090 [45:54<9:34:09,  1.97s/it]
  8%|▊         | 1560/19090 [45:56<8:35:27,  1.76s/it]
                                                      
{'loss': 0.0061, 'grad_norm': 0.9132096767425537, 'learning_rate': 4.0832896804609745e-05, 'recon_loss': 0.006045287498272956, 'recon_static': 0.005024777865037322, 'recon_motion': 0.001020509487716481, 'recon_norm': 2.6182720720767976, 'recon_norm_std': 0.08407679218798876, 'target_norm': 1.617364653944969, 'recon_cos': 0.33533674143254755, 'gradnorm/compressor.query': 0.015147951499753856, 'gradnorm/compressor.layer0': 0.26210356647633065, 'gradnorm/compressor.layer1': 0.3388647799447582, 'gradnorm/compressor.layer2': 0.29706292729474426, 'gradnorm/compressor.layer3': 0.27128626778730586, 'gradnorm/compressor.layer4': 0.25835373449267623, 'gradnorm/compressor.layer5': 0.24631519799108778, 'gradnorm/compressor.layer6': 0.23584367421200672, 'gradnorm/compressor.layer7': 0.22306288012052403, 'gradnorm/decoder.layer0': 0.2765753268915014, 'gradnorm/decoder.layer1': 0.2590644378957753, 'gradnorm/decoder.layer2': 0.33458458357268994, 'gradnorm/max_param': 0.233361154794693, 'epoch': 0.82}

  8%|▊         | 1560/19090 [45:56<8:35:27,  1.76s/it]
  8%|▊         | 1561/19090 [45:57<8:36:44,  1.77s/it]
  8%|▊         | 1562/19090 [45:59<8:57:16,  1.84s/it]
  8%|▊         | 1563/19090 [46:01<8:26:16,  1.73s/it]
  8%|▊         | 1564/19090 [46:03<8:22:39,  1.72s/it]
  8%|▊         | 1565/19090 [46:05<9:34:05,  1.97s/it]
  8%|▊         | 1566/19090 [46:06<8:35:08,  1.76s/it]
  8%|▊         | 1567/19090 [46:08<8:21:23,  1.72s/it]
  8%|▊         | 1568/19090 [46:10<9:02:15,  1.86s/it]
  8%|▊         | 1569/19090 [46:12<8:14:13,  1.69s/it]
  8%|▊         | 1570/19090 [46:13<8:26:21,  1.73s/it]
                                                      
{'loss': 0.0075, 'grad_norm': 0.8190680146217346, 'learning_rate': 4.1094814038763755e-05, 'recon_loss': 0.007471084536518902, 'recon_static': 0.006440714991185814, 'recon_motion': 0.001030369552609045, 'recon_norm': 2.9353818237781524, 'recon_norm_std': 0.07474693562835455, 'target_norm': 1.6206011742353439, 'recon_cos': 0.2955674953758717, 'gradnorm/compressor.query': 0.01392728470562005, 'gradnorm/compressor.layer0': 0.24429833468155737, 'gradnorm/compressor.layer1': 0.30559618260158233, 'gradnorm/compressor.layer2': 0.2607430018723018, 'gradnorm/compressor.layer3': 0.2396563189018092, 'gradnorm/compressor.layer4': 0.2289042957730248, 'gradnorm/compressor.layer5': 0.21674696029799786, 'gradnorm/compressor.layer6': 0.2122868454076543, 'gradnorm/compressor.layer7': 0.20170214125901395, 'gradnorm/decoder.layer0': 0.2451381294453654, 'gradnorm/decoder.layer1': 0.2345062126986812, 'gradnorm/decoder.layer2': 0.3032719545939603, 'gradnorm/max_param': 0.21889448165893555, 'epoch': 0.82}

  8%|▊         | 1570/19090 [46:13<8:26:21,  1.73s/it]
  8%|▊         | 1571/19090 [46:15<8:23:25,  1.72s/it]
  8%|▊         | 1572/19090 [46:17<8:10:45,  1.68s/it]
  8%|▊         | 1573/19090 [46:19<8:30:52,  1.75s/it]
  8%|▊         | 1574/19090 [46:20<8:17:27,  1.70s/it]
  8%|▊         | 1575/19090 [46:22<8:03:34,  1.66s/it]
  8%|▊         | 1576/19090 [46:24<8:57:18,  1.84s/it]
  8%|▊         | 1577/19090 [46:25<8:24:02,  1.73s/it]
  8%|▊         | 1578/19090 [46:27<8:33:58,  1.76s/it]
  8%|▊         | 1579/19090 [46:29<8:47:11,  1.81s/it]
  8%|▊         | 1580/19090 [46:31<8:02:56,  1.65s/it]
                                                      
{'loss': 0.007, 'grad_norm': 0.850554883480072, 'learning_rate': 4.135673127291776e-05, 'recon_loss': 0.006964943394996226, 'recon_static': 0.005931940977461636, 'recon_motion': 0.0010330024611903354, 'recon_norm': 2.837442696094513, 'recon_norm_std': 0.07757434360682965, 'target_norm': 1.6215296417474747, 'recon_cos': 0.3057492561638355, 'gradnorm/compressor.query': 0.014349820788156473, 'gradnorm/compressor.layer0': 0.2508298541590259, 'gradnorm/compressor.layer1': 0.3164916453094972, 'gradnorm/compressor.layer2': 0.27076600075303336, 'gradnorm/compressor.layer3': 0.25086141237983706, 'gradnorm/compressor.layer4': 0.23836064970062606, 'gradnorm/compressor.layer5': 0.22455788727122852, 'gradnorm/compressor.layer6': 0.21993599976735, 'gradnorm/compressor.layer7': 0.2102929528633991, 'gradnorm/decoder.layer0': 0.25797743289967373, 'gradnorm/decoder.layer1': 0.2423370883795893, 'gradnorm/decoder.layer2': 0.31431308732022467, 'gradnorm/max_param': 0.22380712628364563, 'epoch': 0.83}

  8%|▊         | 1580/19090 [46:31<8:02:56,  1.65s/it]
  8%|▊         | 1581/19090 [46:33<8:47:38,  1.81s/it]
  8%|▊         | 1582/19090 [46:35<8:55:10,  1.83s/it]
  8%|▊         | 1583/19090 [46:36<8:11:00,  1.68s/it]
  8%|▊         | 1584/19090 [46:38<8:54:48,  1.83s/it]
  8%|▊         | 1585/19090 [46:40<9:24:10,  1.93s/it]
  8%|▊         | 1586/19090 [46:42<8:28:24,  1.74s/it]
  8%|▊         | 1587/19090 [46:44<8:59:02,  1.85s/it]
  8%|▊         | 1588/19090 [46:46<9:19:04,  1.92s/it]
  8%|▊         | 1589/19090 [46:47<8:25:08,  1.73s/it]
  8%|▊         | 1590/19090 [46:49<8:58:16,  1.85s/it]
                                                      
{'loss': 0.0075, 'grad_norm': 0.868924081325531, 'learning_rate': 4.161864850707177e-05, 'recon_loss': 0.007481170271057635, 'recon_static': 0.0064610149594955145, 'recon_motion': 0.0010201553130173124, 'recon_norm': 2.9469701945781708, 'recon_norm_std': 0.07405050806701183, 'target_norm': 1.6191504687070846, 'recon_cos': 0.2976131282746792, 'gradnorm/compressor.query': 0.014617649645287747, 'gradnorm/compressor.layer0': 0.25787739948334243, 'gradnorm/compressor.layer1': 0.3261395969523614, 'gradnorm/compressor.layer2': 0.27799232297183857, 'gradnorm/compressor.layer3': 0.255695727480542, 'gradnorm/compressor.layer4': 0.24307200975959345, 'gradnorm/compressor.layer5': 0.23082227586320167, 'gradnorm/compressor.layer6': 0.2218094332620858, 'gradnorm/compressor.layer7': 0.21222300357457013, 'gradnorm/decoder.layer0': 0.259832818081777, 'gradnorm/decoder.layer1': 0.24793295609762905, 'gradnorm/decoder.layer2': 0.3220291541677774, 'gradnorm/max_param': 0.22941036522388458, 'epoch': 0.83}

  8%|▊         | 1590/19090 [46:49<8:58:16,  1.85s/it]
  8%|▊         | 1591/19090 [46:51<8:37:44,  1.78s/it]
  8%|▊         | 1592/19090 [46:53<8:46:35,  1.81s/it]
  8%|▊         | 1593/19090 [46:54<8:33:18,  1.76s/it]
  8%|▊         | 1594/19090 [46:56<8:41:38,  1.79s/it]
  8%|▊         | 1595/19090 [46:58<8:35:09,  1.77s/it]
  8%|▊         | 1596/19090 [47:00<8:29:18,  1.75s/it]
  8%|▊         | 1597/19090 [47:02<9:20:45,  1.92s/it]
  8%|▊         | 1598/19090 [47:03<8:28:30,  1.74s/it]
  8%|▊         | 1599/19090 [47:05<8:27:34,  1.74s/it]
  8%|▊         | 1600/19090 [47:07<9:09:26,  1.88s/it]
                                                      
{'loss': 0.0072, 'grad_norm': 0.8851515054702759, 'learning_rate': 4.188056574122578e-05, 'recon_loss': 0.007143671426456421, 'recon_static': 0.006126042979303747, 'recon_motion': 0.0010176284733461217, 'recon_norm': 2.8796239733695983, 'recon_norm_std': 0.07550684791058301, 'target_norm': 1.6180877059698104, 'recon_cos': 0.30445015206933024, 'gradnorm/compressor.query': 0.013696035575730911, 'gradnorm/compressor.layer0': 0.2555723400250689, 'gradnorm/compressor.layer1': 0.32086547942353727, 'gradnorm/compressor.layer2': 0.285109176408217, 'gradnorm/compressor.layer3': 0.2637311939341687, 'gradnorm/compressor.layer4': 0.24845810840709953, 'gradnorm/compressor.layer5': 0.23619586617036425, 'gradnorm/compressor.layer6': 0.22791467206931826, 'gradnorm/compressor.layer7': 0.21570213672725175, 'gradnorm/decoder.layer0': 0.26924691890327945, 'gradnorm/decoder.layer1': 0.2553098907944367, 'gradnorm/decoder.layer2': 0.33244981749167146, 'gradnorm/max_param': 0.22771738469600677, 'epoch': 0.84}

  8%|▊         | 1600/19090 [47:07<9:09:26,  1.88s/it]
  8%|▊         | 1601/19090 [47:08<8:18:23,  1.71s/it]
  8%|▊         | 1602/19090 [47:10<8:37:14,  1.77s/it]
  8%|▊         | 1603/19090 [47:12<8:53:01,  1.83s/it]
  8%|▊         | 1604/19090 [47:14<8:05:32,  1.67s/it]
  8%|▊         | 1605/19090 [47:16<9:02:00,  1.86s/it]
  8%|▊         | 1606/19090 [47:18<9:13:18,  1.90s/it]
  8%|▊         | 1607/19090 [47:19<8:20:24,  1.72s/it]
  8%|▊         | 1608/19090 [47:21<8:41:34,  1.79s/it]
  8%|▊         | 1609/19090 [47:23<9:04:17,  1.87s/it]
  8%|▊         | 1610/19090 [47:25<8:14:09,  1.70s/it]
                                                      
{'loss': 0.0052, 'grad_norm': 0.5931895971298218, 'learning_rate': 4.214248297537978e-05, 'recon_loss': 0.005231712234672159, 'recon_static': 0.004203555913409218, 'recon_motion': 0.0010281561277224681, 'recon_norm': 2.454428517818451, 'recon_norm_std': 0.08606570865958929, 'target_norm': 1.6230373859405518, 'recon_cos': 0.3542172133922577, 'gradnorm/compressor.query': 0.009777580334189623, 'gradnorm/compressor.layer0': 0.1710646447809279, 'gradnorm/compressor.layer1': 0.21833166072807808, 'gradnorm/compressor.layer2': 0.18928554831642783, 'gradnorm/compressor.layer3': 0.17738495528676934, 'gradnorm/compressor.layer4': 0.16664171528806054, 'gradnorm/compressor.layer5': 0.1585027365429471, 'gradnorm/compressor.layer6': 0.15298583971712712, 'gradnorm/compressor.layer7': 0.1442697969690918, 'gradnorm/decoder.layer0': 0.17778577166228623, 'gradnorm/decoder.layer1': 0.17055637458979284, 'gradnorm/decoder.layer2': 0.22303307585779303, 'gradnorm/max_param': 0.1547829955816269, 'epoch': 0.84}

  8%|▊         | 1610/19090 [47:25<8:14:09,  1.70s/it]
  8%|▊         | 1611/19090 [47:27<8:40:22,  1.79s/it]
  8%|▊         | 1612/19090 [47:29<8:59:15,  1.85s/it]
  8%|▊         | 1613/19090 [47:30<8:12:04,  1.69s/it]
  8%|▊         | 1614/19090 [47:32<8:22:27,  1.73s/it]
  8%|▊         | 1615/19090 [47:34<8:48:54,  1.82s/it]
  8%|▊         | 1616/19090 [47:35<8:03:28,  1.66s/it]
  8%|▊         | 1617/19090 [47:37<8:53:17,  1.83s/it]
  8%|▊         | 1618/19090 [47:39<8:39:04,  1.78s/it]
  8%|▊         | 1619/19090 [47:40<8:18:11,  1.71s/it]
  8%|▊         | 1620/19090 [47:43<8:57:35,  1.85s/it]
                                                      
{'loss': 0.0047, 'grad_norm': 0.6302547454833984, 'learning_rate': 4.240440020953379e-05, 'recon_loss': 0.004649769875686616, 'recon_static': 0.0036329234833829107, 'recon_motion': 0.001016846280253958, 'recon_norm': 2.323430836200714, 'recon_norm_std': 0.08902338240295649, 'target_norm': 1.6183728128671646, 'recon_cos': 0.3736615151166916, 'gradnorm/compressor.query': 0.010329414123649006, 'gradnorm/compressor.layer0': 0.18117115920411933, 'gradnorm/compressor.layer1': 0.23234659466416216, 'gradnorm/compressor.layer2': 0.2012685247300776, 'gradnorm/compressor.layer3': 0.18451339611083697, 'gradnorm/compressor.layer4': 0.17660437821521113, 'gradnorm/compressor.layer5': 0.166708677679404, 'gradnorm/compressor.layer6': 0.1622392582214021, 'gradnorm/compressor.layer7': 0.1534591155841002, 'gradnorm/decoder.layer0': 0.1910890948897487, 'gradnorm/decoder.layer1': 0.18271124554368068, 'gradnorm/decoder.layer2': 0.23874998890292556, 'gradnorm/max_param': 0.16456839442253113, 'epoch': 0.85}

  8%|▊         | 1620/19090 [47:43<8:57:35,  1.85s/it]
  8%|▊         | 1621/19090 [47:44<8:25:02,  1.73s/it]
  8%|▊         | 1622/19090 [47:45<7:45:26,  1.60s/it]
  9%|▊         | 1623/19090 [47:48<9:20:27,  1.93s/it]
  9%|▊         | 1624/19090 [47:49<8:24:31,  1.73s/it]
  9%|▊         | 1625/19090 [47:51<8:01:40,  1.65s/it]
  9%|▊         | 1626/19090 [47:54<9:42:59,  2.00s/it]
  9%|▊         | 1627/19090 [47:55<8:40:42,  1.79s/it]
  9%|▊         | 1628/19090 [47:56<7:59:10,  1.65s/it]
  9%|▊         | 1629/19090 [47:59<9:40:11,  1.99s/it]
  9%|▊         | 1630/19090 [48:00<8:39:22,  1.78s/it]
                                                      
{'loss': 0.007, 'grad_norm': 0.8086510300636292, 'learning_rate': 4.26663174436878e-05, 'recon_loss': 0.006992693478241563, 'recon_static': 0.0059674702177289875, 'recon_motion': 0.0010252232139464467, 'recon_norm': 2.833495002985001, 'recon_norm_std': 0.07640040088444948, 'target_norm': 1.6192906737327575, 'recon_cos': 0.30904633551836014, 'gradnorm/compressor.query': 0.012467311935372141, 'gradnorm/compressor.layer0': 0.23678502926951372, 'gradnorm/compressor.layer1': 0.30168943648964475, 'gradnorm/compressor.layer2': 0.25636358302458323, 'gradnorm/compressor.layer3': 0.2384766730980769, 'gradnorm/compressor.layer4': 0.22631398909710193, 'gradnorm/compressor.layer5': 0.21213331175285183, 'gradnorm/compressor.layer6': 0.20608439841466336, 'gradnorm/compressor.layer7': 0.19646079109578726, 'gradnorm/decoder.layer0': 0.24574131176134137, 'gradnorm/decoder.layer1': 0.23255006476733353, 'gradnorm/decoder.layer2': 0.30385768816196224, 'gradnorm/max_param': 0.21117183566093445, 'epoch': 0.85}

  9%|▊         | 1630/19090 [48:00<8:39:22,  1.78s/it]
  9%|▊         | 1631/19090 [48:02<7:55:55,  1.64s/it]
  9%|▊         | 1632/19090 [48:04<9:26:40,  1.95s/it]
  9%|▊         | 1633/19090 [48:06<8:29:45,  1.75s/it]
  9%|▊         | 1634/19090 [48:07<7:50:11,  1.62s/it]
  9%|▊         | 1635/19090 [48:09<9:09:54,  1.89s/it]
  9%|▊         | 1636/19090 [48:11<8:17:59,  1.71s/it]
  9%|▊         | 1637/19090 [48:12<8:10:08,  1.68s/it]
  9%|▊         | 1638/19090 [48:14<8:15:30,  1.70s/it]
  9%|▊         | 1639/19090 [48:15<7:40:01,  1.58s/it]
  9%|▊         | 1640/19090 [48:17<8:22:43,  1.73s/it]
                                                      
{'loss': 0.0076, 'grad_norm': 0.8782606720924377, 'learning_rate': 4.29282346778418e-05, 'recon_loss': 0.007623210491146892, 'recon_static': 0.0065841099712997675, 'recon_motion': 0.0010391003859695047, 'recon_norm': 2.959501326084137, 'recon_norm_std': 0.07368098990991712, 'target_norm': 1.6228038102388382, 'recon_cos': 0.29944765493273734, 'gradnorm/compressor.query': 0.01463053921153292, 'gradnorm/compressor.layer0': 0.25330466381670924, 'gradnorm/compressor.layer1': 0.32052669109231313, 'gradnorm/compressor.layer2': 0.27734027108173076, 'gradnorm/compressor.layer3': 0.2579048164851993, 'gradnorm/compressor.layer4': 0.2435822089705247, 'gradnorm/compressor.layer5': 0.2327585266581707, 'gradnorm/compressor.layer6': 0.2248641517931779, 'gradnorm/compressor.layer7': 0.2181308362447824, 'gradnorm/decoder.layer0': 0.26635854075727416, 'gradnorm/decoder.layer1': 0.25558200390642766, 'gradnorm/decoder.layer2': 0.33596401331758324, 'gradnorm/max_param': 0.23112329840660095, 'epoch': 0.86}

  9%|▊         | 1640/19090 [48:17<8:22:43,  1.73s/it]
  9%|▊         | 1641/19090 [48:19<8:11:31,  1.69s/it]
  9%|▊         | 1642/19090 [48:20<7:37:04,  1.57s/it]
  9%|▊         | 1643/19090 [48:22<8:22:36,  1.73s/it]
  9%|▊         | 1644/19090 [48:24<8:01:10,  1.65s/it]
  9%|▊         | 1645/19090 [48:25<7:29:23,  1.55s/it]
  9%|▊         | 1646/19090 [48:28<8:55:44,  1.84s/it]
  9%|▊         | 1647/19090 [48:29<8:32:52,  1.76s/it]
  9%|▊         | 1648/19090 [48:31<7:52:50,  1.63s/it]
  9%|▊         | 1649/19090 [48:33<8:31:26,  1.76s/it]
  9%|▊         | 1650/19090 [48:34<8:21:01,  1.72s/it]
                                                      
{'loss': 0.0078, 'grad_norm': 0.8077167868614197, 'learning_rate': 4.319015191199581e-05, 'recon_loss': 0.0077916212263517085, 'recon_static': 0.0067763517494313415, 'recon_motion': 0.0010152695162105375, 'recon_norm': 2.997623246908188, 'recon_norm_std': 0.07206480503082276, 'target_norm': 1.6181314706802368, 'recon_cos': 0.2895446725189686, 'gradnorm/compressor.query': 0.012170325167023446, 'gradnorm/compressor.layer0': 0.23062245852147778, 'gradnorm/compressor.layer1': 0.30049558177530833, 'gradnorm/compressor.layer2': 0.25809861692584923, 'gradnorm/compressor.layer3': 0.23611098157810573, 'gradnorm/compressor.layer4': 0.2251687092877835, 'gradnorm/compressor.layer5': 0.2127806790826438, 'gradnorm/compressor.layer6': 0.20592764241847006, 'gradnorm/compressor.layer7': 0.19688467501527593, 'gradnorm/decoder.layer0': 0.24732742953360848, 'gradnorm/decoder.layer1': 0.2307502998823323, 'gradnorm/decoder.layer2': 0.3079737821658482, 'gradnorm/max_param': 0.21365907788276672, 'epoch': 0.86}

  9%|▊         | 1650/19090 [48:34<8:21:01,  1.72s/it]
  9%|▊         | 1651/19090 [48:36<7:44:59,  1.60s/it]
  9%|▊         | 1652/19090 [48:37<7:59:53,  1.65s/it]
  9%|▊         | 1653/19090 [48:39<8:28:08,  1.75s/it]
  9%|▊         | 1654/19090 [48:41<7:48:31,  1.61s/it]
  9%|▊         | 1655/19090 [48:42<7:38:36,  1.58s/it]
  9%|▊         | 1656/19090 [48:44<8:24:26,  1.74s/it]
  9%|▊         | 1657/19090 [48:46<7:46:29,  1.61s/it]
  9%|▊         | 1658/19090 [48:47<7:33:14,  1.56s/it]
  9%|▊         | 1659/19090 [48:49<8:34:24,  1.77s/it]
  9%|▊         | 1660/19090 [48:51<7:53:11,  1.63s/it]
                                                      
{'loss': 0.0071, 'grad_norm': 0.8531242609024048, 'learning_rate': 4.345206914614982e-05, 'recon_loss': 0.007112350407987833, 'recon_static': 0.006084769463632256, 'recon_motion': 0.0010275807420839556, 'recon_norm': 2.872275185585022, 'recon_norm_std': 0.07487734220921993, 'target_norm': 1.6214859336614609, 'recon_cos': 0.30515243411064147, 'gradnorm/compressor.query': 0.015234230782822697, 'gradnorm/compressor.layer0': 0.26087642457442384, 'gradnorm/compressor.layer1': 0.3233391308121352, 'gradnorm/compressor.layer2': 0.2710099500258256, 'gradnorm/compressor.layer3': 0.24865053717520486, 'gradnorm/compressor.layer4': 0.23471956356569357, 'gradnorm/compressor.layer5': 0.2230077278269348, 'gradnorm/compressor.layer6': 0.21762177185368164, 'gradnorm/compressor.layer7': 0.20367883409171578, 'gradnorm/decoder.layer0': 0.25308804359978676, 'gradnorm/decoder.layer1': 0.24007730078563483, 'gradnorm/decoder.layer2': 0.3230326569829064, 'gradnorm/max_param': 0.2336931824684143, 'epoch': 0.87}

  9%|▊         | 1660/19090 [48:51<7:53:11,  1.63s/it]
  9%|▊         | 1661/19090 [48:52<7:23:18,  1.53s/it]
  9%|▊         | 1662/19090 [48:54<8:28:37,  1.75s/it]
  9%|▊         | 1663/19090 [48:55<7:48:16,  1.61s/it]
  9%|▊         | 1664/19090 [48:57<7:24:12,  1.53s/it]
  9%|▊         | 1665/19090 [48:59<8:34:46,  1.77s/it]
  9%|▊         | 1666/19090 [49:00<7:52:40,  1.63s/it]
  9%|▊         | 1667/19090 [49:02<7:23:42,  1.53s/it]
  9%|▊         | 1668/19090 [49:04<8:43:18,  1.80s/it]
  9%|▊         | 1669/19090 [49:05<7:59:18,  1.65s/it]
  9%|▊         | 1670/19090 [49:07<7:41:39,  1.59s/it]
                                                      
{'loss': 0.0072, 'grad_norm': 0.8522263765335083, 'learning_rate': 4.371398638030383e-05, 'recon_loss': 0.007147719489876181, 'recon_static': 0.006125053041614592, 'recon_motion': 0.0010226663434877992, 'recon_norm': 2.875196176767349, 'recon_norm_std': 0.07301907371729613, 'target_norm': 1.6189590483903884, 'recon_cos': 0.3041079707443714, 'gradnorm/compressor.query': 0.013664193422184848, 'gradnorm/compressor.layer0': 0.2465149438827708, 'gradnorm/compressor.layer1': 0.31549169680842964, 'gradnorm/compressor.layer2': 0.2713879475766502, 'gradnorm/compressor.layer3': 0.24625102573026234, 'gradnorm/compressor.layer4': 0.2351725994015565, 'gradnorm/compressor.layer5': 0.22085240472158463, 'gradnorm/compressor.layer6': 0.2179287668698618, 'gradnorm/compressor.layer7': 0.20702238783672292, 'gradnorm/decoder.layer0': 0.2591796231072122, 'gradnorm/decoder.layer1': 0.2478398187258444, 'gradnorm/decoder.layer2': 0.3293563015351611, 'gradnorm/max_param': 0.22547270357608795, 'epoch': 0.87}

  9%|▊         | 1670/19090 [49:07<7:41:39,  1.59s/it]
  9%|▉         | 1671/19090 [49:09<8:54:20,  1.84s/it]
  9%|▉         | 1672/19090 [49:11<8:06:29,  1.68s/it]
  9%|▉         | 1673/19090 [49:12<7:34:03,  1.56s/it]
  9%|▉         | 1674/19090 [49:14<8:18:41,  1.72s/it]
  9%|▉         | 1675/19090 [49:15<7:42:48,  1.59s/it]
  9%|▉         | 1676/19090 [49:17<7:17:03,  1.51s/it]
  9%|▉         | 1677/19090 [49:19<7:56:33,  1.64s/it]
  9%|▉         | 1678/19090 [49:20<7:26:26,  1.54s/it]
  9%|▉         | 1679/19090 [49:21<7:05:41,  1.47s/it]
  9%|▉         | 1680/19090 [49:23<7:34:46,  1.57s/it]
                                                      
{'loss': 0.0056, 'grad_norm': 0.5983381271362305, 'learning_rate': 4.3975903614457834e-05, 'recon_loss': 0.00563068320043385, 'recon_static': 0.004615046002436429, 'recon_motion': 0.001015637356613297, 'recon_norm': 2.553889054059982, 'recon_norm_std': 0.08216332029551268, 'target_norm': 1.619292649626732, 'recon_cos': 0.3425391919910908, 'gradnorm/compressor.query': 0.010500552050404198, 'gradnorm/compressor.layer0': 0.17330454614720617, 'gradnorm/compressor.layer1': 0.219702511636735, 'gradnorm/compressor.layer2': 0.18919141134056863, 'gradnorm/compressor.layer3': 0.17360161907685587, 'gradnorm/compressor.layer4': 0.16538681546645417, 'gradnorm/compressor.layer5': 0.1583353714330738, 'gradnorm/compressor.layer6': 0.15153684030189585, 'gradnorm/compressor.layer7': 0.14714485042565456, 'gradnorm/decoder.layer0': 0.18151051978751875, 'gradnorm/decoder.layer1': 0.17359385616597964, 'gradnorm/decoder.layer2': 0.23139768021792345, 'gradnorm/max_param': 0.16090098023414612, 'epoch': 0.88}

  9%|▉         | 1680/19090 [49:23<7:34:46,  1.57s/it]
  9%|▉         | 1681/19090 [49:24<7:11:02,  1.49s/it]
  9%|▉         | 1682/19090 [49:26<6:55:21,  1.43s/it]
  9%|▉         | 1683/19090 [49:27<7:15:35,  1.50s/it]
  9%|▉         | 1684/19090 [49:28<6:56:58,  1.44s/it]
  9%|▉         | 1685/19090 [49:30<6:43:46,  1.39s/it]
  9%|▉         | 1686/19090 [49:32<8:01:01,  1.66s/it]
  9%|▉         | 1687/19090 [49:33<7:30:07,  1.55s/it]
  9%|▉         | 1688/19090 [49:35<7:07:11,  1.47s/it]
  9%|▉         | 1689/19090 [49:36<7:09:09,  1.48s/it]
  9%|▉         | 1690/19090 [49:37<6:53:52,  1.43s/it]
                                                      
{'loss': 0.0038, 'grad_norm': 0.4694773256778717, 'learning_rate': 4.423782084861184e-05, 'recon_loss': 0.0037878559436649085, 'recon_static': 0.002766324981348589, 'recon_motion': 0.001021530969592277, 'recon_norm': 2.099092134833336, 'recon_norm_std': 0.0950358621776104, 'target_norm': 1.6233947902917862, 'recon_cos': 0.4169685423374176, 'gradnorm/compressor.query': 0.007247577960097631, 'gradnorm/compressor.layer0': 0.12997709165294472, 'gradnorm/compressor.layer1': 0.17077803593897983, 'gradnorm/compressor.layer2': 0.14791218133515044, 'gradnorm/compressor.layer3': 0.13761323146239313, 'gradnorm/compressor.layer4': 0.13111495211426072, 'gradnorm/compressor.layer5': 0.12593667189796745, 'gradnorm/compressor.layer6': 0.12308370705014914, 'gradnorm/compressor.layer7': 0.11602878319331866, 'gradnorm/decoder.layer0': 0.14377989587683004, 'gradnorm/decoder.layer1': 0.1357960971291626, 'gradnorm/decoder.layer2': 0.18084556123120832, 'gradnorm/max_param': 0.12667648494243622, 'epoch': 0.89}

  9%|▉         | 1690/19090 [49:37<6:53:52,  1.43s/it]
  9%|▉         | 1691/19090 [49:39<6:43:20,  1.39s/it]
  9%|▉         | 1692/19090 [49:41<7:29:29,  1.55s/it]
  9%|▉         | 1693/19090 [49:42<7:06:32,  1.47s/it]
  9%|▉         | 1694/19090 [49:43<6:50:39,  1.42s/it]
  9%|▉         | 1695/19090 [49:45<7:18:21,  1.51s/it]
  9%|▉         | 1696/19090 [49:46<6:59:06,  1.45s/it]
  9%|▉         | 1697/19090 [49:48<6:45:55,  1.40s/it]
  9%|▉         | 1698/19090 [49:50<7:36:46,  1.58s/it]
  9%|▉         | 1699/19090 [49:51<7:12:08,  1.49s/it]
  9%|▉         | 1700/19090 [49:52<7:01:12,  1.45s/it]
                                                      
{'loss': 0.0055, 'grad_norm': 0.91298508644104, 'learning_rate': 4.4499738082765847e-05, 'recon_loss': 0.005486457818187773, 'recon_static': 0.004471530526643619, 'recon_motion': 0.0010149272231501528, 'recon_norm': 2.5011270344257355, 'recon_norm_std': 0.08342980146408081, 'target_norm': 1.6192772835493088, 'recon_cos': 0.35534561425447464, 'gradnorm/compressor.query': 0.016102919353529123, 'gradnorm/compressor.layer0': 0.27131828881208475, 'gradnorm/compressor.layer1': 0.33941678378322226, 'gradnorm/compressor.layer2': 0.2943442542831875, 'gradnorm/compressor.layer3': 0.2661852118193819, 'gradnorm/compressor.layer4': 0.2532057121307112, 'gradnorm/compressor.layer5': 0.23687789747886717, 'gradnorm/compressor.layer6': 0.2335655539856286, 'gradnorm/compressor.layer7': 0.22017803011303833, 'gradnorm/decoder.layer0': 0.27447657025947175, 'gradnorm/decoder.layer1': 0.2598850721349604, 'gradnorm/decoder.layer2': 0.3476188296634951, 'gradnorm/max_param': 0.2423066347837448, 'epoch': 0.89}

  9%|▉         | 1700/19090 [49:52<7:01:12,  1.45s/it]
  9%|▉         | 1701/19090 [49:54<7:51:04,  1.63s/it]
  9%|▉         | 1702/19090 [49:56<7:21:56,  1.53s/it]
  9%|▉         | 1703/19090 [49:57<7:21:28,  1.52s/it]
  9%|▉         | 1704/19090 [49:59<7:47:43,  1.61s/it]
  9%|▉         | 1705/19090 [50:00<7:19:34,  1.52s/it]
  9%|▉         | 1706/19090 [50:02<7:26:32,  1.54s/it]
  9%|▉         | 1707/19090 [50:04<8:12:17,  1.70s/it]
  9%|▉         | 1708/19090 [50:05<7:37:30,  1.58s/it]
  9%|▉         | 1709/19090 [50:07<8:17:16,  1.72s/it]
  9%|▉         | 1710/19090 [50:09<8:45:09,  1.81s/it]
                                                      
{'loss': 0.0084, 'grad_norm': 0.9349996447563171, 'learning_rate': 4.4761655316919856e-05, 'recon_loss': 0.008382525458000601, 'recon_static': 0.007364391977898776, 'recon_motion': 0.0010181333986110986, 'recon_norm': 3.118748462200165, 'recon_norm_std': 0.06827010633423924, 'target_norm': 1.6203245908021926, 'recon_cos': 0.28260780647397044, 'gradnorm/compressor.query': 0.014448885883986858, 'gradnorm/compressor.layer0': 0.2653100165710323, 'gradnorm/compressor.layer1': 0.3393910560884412, 'gradnorm/compressor.layer2': 0.2927064357076112, 'gradnorm/compressor.layer3': 0.2740066599314939, 'gradnorm/compressor.layer4': 0.25983777872588953, 'gradnorm/compressor.layer5': 0.24653808655400453, 'gradnorm/compressor.layer6': 0.24070807793909485, 'gradnorm/compressor.layer7': 0.22912969164511546, 'gradnorm/decoder.layer0': 0.2870080989831336, 'gradnorm/decoder.layer1': 0.2703872990617212, 'gradnorm/decoder.layer2': 0.3653255287699122, 'gradnorm/max_param': 0.2534755766391754, 'epoch': 0.9}

  9%|▉         | 1710/19090 [50:09<8:45:09,  1.81s/it]
  9%|▉         | 1711/19090 [50:11<7:59:38,  1.66s/it]
  9%|▉         | 1712/19090 [50:13<8:36:25,  1.78s/it]
  9%|▉         | 1713/19090 [50:15<8:55:51,  1.85s/it]
  9%|▉         | 1714/19090 [50:16<8:06:49,  1.68s/it]
  9%|▉         | 1715/19090 [50:18<8:37:01,  1.79s/it]
  9%|▉         | 1716/19090 [50:20<9:04:11,  1.88s/it]
  9%|▉         | 1717/19090 [50:21<8:14:29,  1.71s/it]
  9%|▉         | 1718/19090 [50:23<8:33:30,  1.77s/it]
  9%|▉         | 1719/19090 [50:25<9:05:08,  1.88s/it]
  9%|▉         | 1720/19090 [50:27<8:13:40,  1.71s/it]
                                                      
{'loss': 0.0076, 'grad_norm': 0.8290544152259827, 'learning_rate': 4.502357255107386e-05, 'recon_loss': 0.007598778232932091, 'recon_static': 0.00658723630476743, 'recon_motion': 0.0010115420998772606, 'recon_norm': 2.964151167869568, 'recon_norm_std': 0.07117770006880164, 'target_norm': 1.6184319883584977, 'recon_cos': 0.2963232632726431, 'gradnorm/compressor.query': 0.012195677245341008, 'gradnorm/compressor.layer0': 0.23420814296735262, 'gradnorm/compressor.layer1': 0.30244822360501034, 'gradnorm/compressor.layer2': 0.2595027224199453, 'gradnorm/compressor.layer3': 0.23654273826166453, 'gradnorm/compressor.layer4': 0.22715641148774393, 'gradnorm/compressor.layer5': 0.21942302506940067, 'gradnorm/compressor.layer6': 0.21434393745288474, 'gradnorm/compressor.layer7': 0.20470727359104798, 'gradnorm/decoder.layer0': 0.2567484311331683, 'gradnorm/decoder.layer1': 0.2417545370896129, 'gradnorm/decoder.layer2': 0.3249615009136523, 'gradnorm/max_param': 0.22526663541793823, 'epoch': 0.9}

  9%|▉         | 1720/19090 [50:27<8:13:40,  1.71s/it]
  9%|▉         | 1721/19090 [50:28<8:10:36,  1.69s/it]
  9%|▉         | 1722/19090 [50:31<9:11:45,  1.91s/it]
  9%|▉         | 1723/19090 [50:32<8:19:24,  1.73s/it]
  9%|▉         | 1724/19090 [50:34<8:08:42,  1.69s/it]
  9%|▉         | 1725/19090 [50:36<9:10:53,  1.90s/it]
  9%|▉         | 1726/19090 [50:37<8:17:21,  1.72s/it]
  9%|▉         | 1727/19090 [50:39<8:22:57,  1.74s/it]
  9%|▉         | 1728/19090 [50:41<9:13:30,  1.91s/it]
  9%|▉         | 1729/19090 [50:43<8:18:55,  1.72s/it]
  9%|▉         | 1730/19090 [50:44<8:21:46,  1.73s/it]
                                                      
{'loss': 0.0068, 'grad_norm': 0.6716527938842773, 'learning_rate': 4.528548978522787e-05, 'recon_loss': 0.006831493985373527, 'recon_static': 0.005818635318428278, 'recon_motion': 0.001012858623289503, 'recon_norm': 2.8107935309410097, 'recon_norm_std': 0.0750375159084797, 'target_norm': 1.6214685142040253, 'recon_cos': 0.31153691485524176, 'gradnorm/compressor.query': 0.011261187269904278, 'gradnorm/compressor.layer0': 0.19570335380556053, 'gradnorm/compressor.layer1': 0.24709332023516653, 'gradnorm/compressor.layer2': 0.21060928665318898, 'gradnorm/compressor.layer3': 0.19257736051148494, 'gradnorm/compressor.layer4': 0.1848569390306762, 'gradnorm/compressor.layer5': 0.1759147613849658, 'gradnorm/compressor.layer6': 0.16969865079987165, 'gradnorm/compressor.layer7': 0.16225349959506583, 'gradnorm/decoder.layer0': 0.20165527928992294, 'gradnorm/decoder.layer1': 0.19706797376290294, 'gradnorm/decoder.layer2': 0.26542590196841953, 'gradnorm/max_param': 0.18492305278778076, 'epoch': 0.91}

  9%|▉         | 1730/19090 [50:44<8:21:46,  1.73s/it]
  9%|▉         | 1731/19090 [50:47<9:00:02,  1.87s/it]
  9%|▉         | 1732/19090 [50:48<8:09:53,  1.69s/it]
  9%|▉         | 1733/19090 [50:50<8:23:52,  1.74s/it]
  9%|▉         | 1734/19090 [50:52<9:09:32,  1.90s/it]
  9%|▉         | 1735/19090 [50:53<8:16:33,  1.72s/it]
  9%|▉         | 1736/19090 [50:55<8:34:13,  1.78s/it]
  9%|▉         | 1737/19090 [50:57<9:05:27,  1.89s/it]
  9%|▉         | 1738/19090 [50:59<8:13:47,  1.71s/it]
  9%|▉         | 1739/19090 [51:01<8:42:58,  1.81s/it]
  9%|▉         | 1740/19090 [51:03<8:42:46,  1.81s/it]
                                                      
{'loss': 0.0045, 'grad_norm': 0.5429263710975647, 'learning_rate': 4.554740701938188e-05, 'recon_loss': 0.004533331573475153, 'recon_static': 0.0035222662205342203, 'recon_motion': 0.0010110652467119508, 'recon_norm': 2.3018763840198515, 'recon_norm_std': 0.0858758568763733, 'target_norm': 1.6190639317035675, 'recon_cos': 0.3825895495712757, 'gradnorm/compressor.query': 0.008682519978946712, 'gradnorm/compressor.layer0': 0.15866905054934033, 'gradnorm/compressor.layer1': 0.20067409854896498, 'gradnorm/compressor.layer2': 0.17370960395494986, 'gradnorm/compressor.layer3': 0.15906786355004723, 'gradnorm/compressor.layer4': 0.15004017485603763, 'gradnorm/compressor.layer5': 0.1410818890035566, 'gradnorm/compressor.layer6': 0.1357967280860135, 'gradnorm/compressor.layer7': 0.13001571395241998, 'gradnorm/decoder.layer0': 0.1624314945504449, 'gradnorm/decoder.layer1': 0.15676917606462087, 'gradnorm/decoder.layer2': 0.21224646274362216, 'gradnorm/max_param': 0.14944200217723846, 'epoch': 0.91}

  9%|▉         | 1740/19090 [51:03<8:42:46,  1.81s/it]
  9%|▉         | 1741/19090 [51:04<7:57:45,  1.65s/it]
  9%|▉         | 1742/19090 [51:06<8:57:48,  1.86s/it]
  9%|▉         | 1743/19090 [51:08<8:48:51,  1.83s/it]
  9%|▉         | 1744/19090 [51:09<8:01:40,  1.67s/it]
  9%|▉         | 1745/19090 [51:12<9:04:28,  1.88s/it]
  9%|▉         | 1746/19090 [51:14<9:04:45,  1.88s/it]
  9%|▉         | 1747/19090 [51:15<8:13:28,  1.71s/it]
  9%|▉         | 1748/19090 [51:17<8:45:50,  1.82s/it]
  9%|▉         | 1749/19090 [51:19<8:30:47,  1.77s/it]
  9%|▉         | 1750/19090 [51:20<7:50:11,  1.63s/it]
                                                      
{'loss': 0.0053, 'grad_norm': 0.8092369437217712, 'learning_rate': 4.580932425353588e-05, 'recon_loss': 0.005259762820787728, 'recon_static': 0.004247576941270381, 'recon_motion': 0.0010121859872015194, 'recon_norm': 2.4612443864345552, 'recon_norm_std': 0.08164395969361067, 'target_norm': 1.6211893528699874, 'recon_cos': 0.3570724785327911, 'gradnorm/compressor.query': 0.013096159062815506, 'gradnorm/compressor.layer0': 0.23069759903544276, 'gradnorm/compressor.layer1': 0.29924163382251695, 'gradnorm/compressor.layer2': 0.2598468682153064, 'gradnorm/compressor.layer3': 0.23673594572855294, 'gradnorm/compressor.layer4': 0.22444924160990032, 'gradnorm/compressor.layer5': 0.21127164829939452, 'gradnorm/compressor.layer6': 0.20070284266171728, 'gradnorm/compressor.layer7': 0.19478443284466024, 'gradnorm/decoder.layer0': 0.24204711536988757, 'gradnorm/decoder.layer1': 0.2337358581503053, 'gradnorm/decoder.layer2': 0.3192282095435059, 'gradnorm/max_param': 0.2226167619228363, 'epoch': 0.92}

  9%|▉         | 1750/19090 [51:20<7:50:11,  1.63s/it]
  9%|▉         | 1751/19090 [51:22<9:17:41,  1.93s/it]
  9%|▉         | 1752/19090 [51:24<8:40:35,  1.80s/it]
  9%|▉         | 1753/19090 [51:25<7:56:27,  1.65s/it]
  9%|▉         | 1754/19090 [51:28<9:00:46,  1.87s/it]
  9%|▉         | 1755/19090 [51:30<8:58:47,  1.86s/it]
  9%|▉         | 1756/19090 [51:31<8:09:21,  1.69s/it]
  9%|▉         | 1757/19090 [51:32<8:01:56,  1.67s/it]
  9%|▉         | 1758/19090 [51:34<8:18:33,  1.73s/it]
  9%|▉         | 1759/19090 [51:36<7:40:31,  1.59s/it]
  9%|▉         | 1760/19090 [51:37<7:34:54,  1.57s/it]
                                                      
{'loss': 0.0077, 'grad_norm': 0.8616753816604614, 'learning_rate': 4.607124148768989e-05, 'recon_loss': 0.007653951377142221, 'recon_static': 0.006637368141673506, 'recon_motion': 0.0010165831845370121, 'recon_norm': 2.969923365116119, 'recon_norm_std': 0.07070051124319435, 'target_norm': 1.621892336010933, 'recon_cos': 0.296407176554203, 'gradnorm/compressor.query': 0.014564730471268111, 'gradnorm/compressor.layer0': 0.24488039471602102, 'gradnorm/compressor.layer1': 0.314432830629537, 'gradnorm/compressor.layer2': 0.2764515966908831, 'gradnorm/compressor.layer3': 0.251615519889384, 'gradnorm/compressor.layer4': 0.2385091161105312, 'gradnorm/compressor.layer5': 0.22610073881545278, 'gradnorm/compressor.layer6': 0.21812207493282715, 'gradnorm/compressor.layer7': 0.20694514049307008, 'gradnorm/decoder.layer0': 0.2599544412559299, 'gradnorm/decoder.layer1': 0.24749555669273604, 'gradnorm/decoder.layer2': 0.3412623339998611, 'gradnorm/max_param': 0.24093177914619446, 'epoch': 0.92}

  9%|▉         | 1760/19090 [51:37<7:34:54,  1.57s/it]
  9%|▉         | 1761/19090 [51:39<8:36:31,  1.79s/it]
  9%|▉         | 1762/19090 [51:41<7:53:22,  1.64s/it]
  9%|▉         | 1763/19090 [51:42<7:31:43,  1.56s/it]
  9%|▉         | 1764/19090 [51:44<8:38:10,  1.79s/it]
  9%|▉         | 1765/19090 [51:46<7:54:03,  1.64s/it]
  9%|▉         | 1766/19090 [51:47<7:43:35,  1.61s/it]
  9%|▉         | 1767/19090 [51:50<9:01:43,  1.88s/it]
  9%|▉         | 1768/19090 [51:51<8:10:14,  1.70s/it]
  9%|▉         | 1769/19090 [51:52<7:41:56,  1.60s/it]
  9%|▉         | 1770/19090 [51:55<9:23:57,  1.95s/it]
                                                      
{'loss': 0.0062, 'grad_norm': 0.9691984057426453, 'learning_rate': 4.63331587218439e-05, 'recon_loss': 0.006181984313298017, 'recon_static': 0.005159524013288319, 'recon_motion': 0.0010224603640381247, 'recon_norm': 2.667466050386429, 'recon_norm_std': 0.07588318455964327, 'target_norm': 1.6241127997636795, 'recon_cos': 0.33083341717720033, 'gradnorm/compressor.query': 0.015767261032661774, 'gradnorm/compressor.layer0': 0.28142466684470646, 'gradnorm/compressor.layer1': 0.3611747416143839, 'gradnorm/compressor.layer2': 0.3114350411733152, 'gradnorm/compressor.layer3': 0.2826714095251578, 'gradnorm/compressor.layer4': 0.2670756167121786, 'gradnorm/compressor.layer5': 0.25427451203865487, 'gradnorm/compressor.layer6': 0.24349774977644958, 'gradnorm/compressor.layer7': 0.23254651646431687, 'gradnorm/decoder.layer0': 0.28630653029294895, 'gradnorm/decoder.layer1': 0.27785851162989395, 'gradnorm/decoder.layer2': 0.37950515607113083, 'gradnorm/max_param': 0.26689231395721436, 'epoch': 0.93}

  9%|▉         | 1770/19090 [51:55<9:23:57,  1.95s/it]
  9%|▉         | 1771/19090 [51:56<8:25:55,  1.75s/it]
  9%|▉         | 1772/19090 [51:58<8:16:51,  1.72s/it]
  9%|▉         | 1773/19090 [52:01<9:45:38,  2.03s/it]
  9%|▉         | 1774/19090 [52:02<8:41:05,  1.81s/it]
  9%|▉         | 1775/19090 [52:03<7:55:41,  1.65s/it]
  9%|▉         | 1776/19090 [52:06<9:35:02,  1.99s/it]
  9%|▉         | 1777/19090 [52:07<8:34:22,  1.78s/it]
  9%|▉         | 1778/19090 [52:09<7:52:08,  1.64s/it]
  9%|▉         | 1779/19090 [52:11<9:02:34,  1.88s/it]
  9%|▉         | 1780/19090 [52:12<8:10:40,  1.70s/it]
                                                      
{'loss': 0.0083, 'grad_norm': 0.9488059878349304, 'learning_rate': 4.659507595599791e-05, 'recon_loss': 0.008276367979124188, 'recon_static': 0.007259587408043444, 'recon_motion': 0.0010167806831304915, 'recon_norm': 3.094023126363754, 'recon_norm_std': 0.06739205848425626, 'target_norm': 1.6214823782444001, 'recon_cos': 0.2847621016204357, 'gradnorm/compressor.query': 0.016005016875539117, 'gradnorm/compressor.layer0': 0.2713224765311564, 'gradnorm/compressor.layer1': 0.3465301298532228, 'gradnorm/compressor.layer2': 0.3009638046541204, 'gradnorm/compressor.layer3': 0.27294457829169716, 'gradnorm/compressor.layer4': 0.2587050679488642, 'gradnorm/compressor.layer5': 0.2491393399257778, 'gradnorm/compressor.layer6': 0.24033679949677744, 'gradnorm/compressor.layer7': 0.23126958461258612, 'gradnorm/decoder.layer0': 0.2821643721226588, 'gradnorm/decoder.layer1': 0.2737349527177518, 'gradnorm/decoder.layer2': 0.38263022696069376, 'gradnorm/max_param': 0.26655077934265137, 'epoch': 0.93}

  9%|▉         | 1780/19090 [52:13<8:10:40,  1.70s/it]
  9%|▉         | 1781/19090 [52:14<7:34:42,  1.58s/it]
  9%|▉         | 1782/19090 [52:17<9:33:40,  1.99s/it]
  9%|▉         | 1783/19090 [52:18<8:32:22,  1.78s/it]
  9%|▉         | 1784/19090 [52:19<7:49:45,  1.63s/it]
  9%|▉         | 1785/19090 [52:22<9:18:27,  1.94s/it]
  9%|▉         | 1786/19090 [52:23<8:22:00,  1.74s/it]
  9%|▉         | 1787/19090 [52:25<7:43:06,  1.61s/it]
  9%|▉         | 1788/19090 [52:27<9:01:24,  1.88s/it]
  9%|▉         | 1789/19090 [52:28<8:10:08,  1.70s/it]
  9%|▉         | 1790/19090 [52:30<7:49:04,  1.63s/it]
                                                      
{'loss': 0.008, 'grad_norm': 0.7289273142814636, 'learning_rate': 4.685699319015191e-05, 'recon_loss': 0.007987681846134364, 'recon_static': 0.00697844180976972, 'recon_motion': 0.0010092399636050687, 'recon_norm': 3.0381712913513184, 'recon_norm_std': 0.06919097881764173, 'target_norm': 1.619903063774109, 'recon_cos': 0.2902486599981785, 'gradnorm/compressor.query': 0.011254954467139875, 'gradnorm/compressor.layer0': 0.2037010461119325, 'gradnorm/compressor.layer1': 0.26842125091844354, 'gradnorm/compressor.layer2': 0.23224658125823888, 'gradnorm/compressor.layer3': 0.21079176707572328, 'gradnorm/compressor.layer4': 0.1981064279919109, 'gradnorm/compressor.layer5': 0.19053635177551626, 'gradnorm/compressor.layer6': 0.18359900527899684, 'gradnorm/compressor.layer7': 0.17821280047194762, 'gradnorm/decoder.layer0': 0.2212008964742659, 'gradnorm/decoder.layer1': 0.21112018198068433, 'gradnorm/decoder.layer2': 0.291156875330446, 'gradnorm/max_param': 0.20548982918262482, 'epoch': 0.94}

  9%|▉         | 1790/19090 [52:30<7:49:04,  1.63s/it]
  9%|▉         | 1791/19090 [52:32<9:18:42,  1.94s/it]
  9%|▉         | 1792/19090 [52:34<8:22:32,  1.74s/it]
  9%|▉         | 1793/19090 [52:35<7:49:08,  1.63s/it]
  9%|▉         | 1794/19090 [52:38<9:51:30,  2.05s/it]
  9%|▉         | 1795/19090 [52:39<8:45:21,  1.82s/it]
  9%|▉         | 1796/19090 [52:41<7:59:06,  1.66s/it]
  9%|▉         | 1797/19090 [52:43<9:20:39,  1.95s/it]
  9%|▉         | 1798/19090 [52:45<8:23:45,  1.75s/it]
  9%|▉         | 1799/19090 [52:46<7:44:21,  1.61s/it]
  9%|▉         | 1800/19090 [52:49<9:43:09,  2.02s/it]
                                                      
{'loss': 0.0044, 'grad_norm': 0.4916464388370514, 'learning_rate': 4.711891042430592e-05, 'recon_loss': 0.004373040597420186, 'recon_static': 0.003376341384137049, 'recon_motion': 0.000996699144889135, 'recon_norm': 2.2583174228668215, 'recon_norm_std': 0.08712056092917919, 'target_norm': 1.6170290470123292, 'recon_cos': 0.38985720872879026, 'gradnorm/compressor.query': 0.0074445422460911875, 'gradnorm/compressor.layer0': 0.13937954925160428, 'gradnorm/compressor.layer1': 0.17786995464013397, 'gradnorm/compressor.layer2': 0.15484207743290884, 'gradnorm/compressor.layer3': 0.14311654185618994, 'gradnorm/compressor.layer4': 0.13505265863469884, 'gradnorm/compressor.layer5': 0.1287329476947929, 'gradnorm/compressor.layer6': 0.12221781021167477, 'gradnorm/compressor.layer7': 0.11828132596007192, 'gradnorm/decoder.layer0': 0.14914692578162309, 'gradnorm/decoder.layer1': 0.14437008584110206, 'gradnorm/decoder.layer2': 0.19827109628746578, 'gradnorm/max_param': 0.14019106328487396, 'epoch': 0.94}

  9%|▉         | 1800/19090 [52:49<9:43:09,  2.02s/it]
  9%|▉         | 1801/19090 [52:50<8:40:22,  1.81s/it]
  9%|▉         | 1802/19090 [52:51<7:55:57,  1.65s/it]
  9%|▉         | 1803/19090 [52:54<9:30:17,  1.98s/it]
  9%|▉         | 1804/19090 [52:56<8:30:46,  1.77s/it]
  9%|▉         | 1805/19090 [52:57<7:49:07,  1.63s/it]
  9%|▉         | 1806/19090 [52:59<9:15:13,  1.93s/it]
  9%|▉         | 1807/19090 [53:01<8:20:14,  1.74s/it]
  9%|▉         | 1808/19090 [53:02<7:41:39,  1.60s/it]
  9%|▉         | 1809/19090 [53:05<9:18:48,  1.94s/it]
  9%|▉         | 1810/19090 [53:06<8:22:20,  1.74s/it]
                                                      
{'loss': 0.0046, 'grad_norm': 0.740195095539093, 'learning_rate': 4.738082765845993e-05, 'recon_loss': 0.004587557940976694, 'recon_static': 0.0035731512994971125, 'recon_motion': 0.0010144065861823037, 'recon_norm': 2.3006952941417693, 'recon_norm_std': 0.08441485296934843, 'target_norm': 1.621701866388321, 'recon_cos': 0.3831874839961529, 'gradnorm/compressor.query': 0.011299944635351606, 'gradnorm/compressor.layer0': 0.2166048092680458, 'gradnorm/compressor.layer1': 0.27263458078402714, 'gradnorm/compressor.layer2': 0.236995008309426, 'gradnorm/compressor.layer3': 0.21655964114626478, 'gradnorm/compressor.layer4': 0.20325761830678782, 'gradnorm/compressor.layer5': 0.1932350973074285, 'gradnorm/compressor.layer6': 0.18571949006840952, 'gradnorm/compressor.layer7': 0.17394046153892886, 'gradnorm/decoder.layer0': 0.22129760281276736, 'gradnorm/decoder.layer1': 0.21495542224905118, 'gradnorm/decoder.layer2': 0.29132484802465386, 'gradnorm/max_param': 0.20606280863285065, 'epoch': 0.95}

  9%|▉         | 1810/19090 [53:06<8:22:20,  1.74s/it]
  9%|▉         | 1811/19090 [53:07<7:42:35,  1.61s/it]
  9%|▉         | 1812/19090 [53:10<9:35:39,  2.00s/it]
  9%|▉         | 1813/19090 [53:12<8:34:00,  1.79s/it]
 10%|▉         | 1814/19090 [53:13<7:51:05,  1.64s/it]
 10%|▉         | 1815/19090 [53:16<9:32:48,  1.99s/it]
 10%|▉         | 1816/19090 [53:17<8:33:03,  1.78s/it]
 10%|▉         | 1817/19090 [53:18<7:50:44,  1.64s/it]
 10%|▉         | 1818/19090 [53:21<9:42:51,  2.02s/it]
 10%|▉         | 1819/19090 [53:22<8:39:07,  1.80s/it]
 10%|▉         | 1820/19090 [53:24<7:55:03,  1.65s/it]
                                                      
{'loss': 0.007, 'grad_norm': 0.6542767286300659, 'learning_rate': 4.7642744892613936e-05, 'recon_loss': 0.006964779656846077, 'recon_static': 0.005957611184567213, 'recon_motion': 0.0010071683529531584, 'recon_norm': 2.82689887881279, 'recon_norm_std': 0.07280856249853969, 'target_norm': 1.619996392726898, 'recon_cos': 0.3164789404720068, 'gradnorm/compressor.query': 0.010082689045284648, 'gradnorm/compressor.layer0': 0.1879315476453339, 'gradnorm/compressor.layer1': 0.24062174293245825, 'gradnorm/compressor.layer2': 0.20883265897464887, 'gradnorm/compressor.layer3': 0.1901108150538939, 'gradnorm/compressor.layer4': 0.1801030185233373, 'gradnorm/compressor.layer5': 0.16862270666914264, 'gradnorm/compressor.layer6': 0.16361295326858513, 'gradnorm/compressor.layer7': 0.1560738763004652, 'gradnorm/decoder.layer0': 0.19850116406839288, 'gradnorm/decoder.layer1': 0.19061441509376223, 'gradnorm/decoder.layer2': 0.2593656486526019, 'gradnorm/max_param': 0.18358214199543, 'epoch': 0.95}

 10%|▉         | 1820/19090 [53:24<7:55:03,  1.65s/it]
 10%|▉         | 1821/19090 [53:27<9:43:39,  2.03s/it]
 10%|▉         | 1822/19090 [53:28<8:40:28,  1.81s/it]
 10%|▉         | 1823/19090 [53:29<7:55:17,  1.65s/it]
 10%|▉         | 1824/19090 [53:32<9:19:13,  1.94s/it]
 10%|▉         | 1825/19090 [53:33<8:22:50,  1.75s/it]
 10%|▉         | 1826/19090 [53:34<7:43:24,  1.61s/it]
 10%|▉         | 1827/19090 [53:37<9:39:48,  2.02s/it]
 10%|▉         | 1828/19090 [53:39<8:37:34,  1.80s/it]
 10%|▉         | 1829/19090 [53:40<7:53:53,  1.65s/it]
 10%|▉         | 1830/19090 [53:43<9:33:59,  2.00s/it]
                                                      
{'loss': 0.0053, 'grad_norm': 0.6582821011543274, 'learning_rate': 4.7904662126767945e-05, 'recon_loss': 0.005343899352010339, 'recon_static': 0.004347045620670542, 'recon_motion': 0.0009968537895474582, 'recon_norm': 2.4899144947528837, 'recon_norm_std': 0.08052012212574482, 'target_norm': 1.6188431590795518, 'recon_cos': 0.3529520682990551, 'gradnorm/compressor.query': 0.010166228689618612, 'gradnorm/compressor.layer0': 0.18876542798835547, 'gradnorm/compressor.layer1': 0.2397423256385677, 'gradnorm/compressor.layer2': 0.2083160373342772, 'gradnorm/compressor.layer3': 0.19156212201477893, 'gradnorm/compressor.layer4': 0.1814661726442578, 'gradnorm/compressor.layer5': 0.17055960171299922, 'gradnorm/compressor.layer6': 0.1660496508566515, 'gradnorm/compressor.layer7': 0.15828505502064327, 'gradnorm/decoder.layer0': 0.20090626338852485, 'gradnorm/decoder.layer1': 0.19106857525429222, 'gradnorm/decoder.layer2': 0.2617515998835159, 'gradnorm/max_param': 0.18660002946853638, 'epoch': 0.96}

 10%|▉         | 1830/19090 [53:43<9:33:59,  2.00s/it]
 10%|▉         | 1831/19090 [53:44<8:33:23,  1.78s/it]
 10%|▉         | 1832/19090 [53:45<7:51:31,  1.64s/it]
 10%|▉         | 1833/19090 [53:48<9:31:10,  1.99s/it]
 10%|▉         | 1834/19090 [53:49<8:31:51,  1.78s/it]
 10%|▉         | 1835/19090 [53:51<7:50:05,  1.63s/it]
 10%|▉         | 1836/19090 [53:54<9:54:39,  2.07s/it]
 10%|▉         | 1837/19090 [53:55<8:48:10,  1.84s/it]
 10%|▉         | 1838/19090 [53:56<8:00:18,  1.67s/it]
 10%|▉         | 1839/19090 [53:59<9:01:08,  1.88s/it]
 10%|▉         | 1840/19090 [54:00<8:09:54,  1.70s/it]
                                                      
{'loss': 0.0072, 'grad_norm': 1.0248216390609741, 'learning_rate': 4.816657936092195e-05, 'recon_loss': 0.007208865415304899, 'recon_static': 0.006207203166559339, 'recon_motion': 0.0010016624437412247, 'recon_norm': 2.8704719066619875, 'recon_norm_std': 0.07163248937577009, 'target_norm': 1.6198089748620987, 'recon_cos': 0.31019061170518397, 'gradnorm/compressor.query': 0.015953157005969155, 'gradnorm/compressor.layer0': 0.29036669737498166, 'gradnorm/compressor.layer1': 0.3807020400368652, 'gradnorm/compressor.layer2': 0.32718011503438293, 'gradnorm/compressor.layer3': 0.29779055861816744, 'gradnorm/compressor.layer4': 0.281185593118029, 'gradnorm/compressor.layer5': 0.2701694148415333, 'gradnorm/compressor.layer6': 0.25504122667275186, 'gradnorm/compressor.layer7': 0.24749890572855093, 'gradnorm/decoder.layer0': 0.31049051313143416, 'gradnorm/decoder.layer1': 0.293721490856329, 'gradnorm/decoder.layer2': 0.4049650374674421, 'gradnorm/max_param': 0.28592249751091003, 'epoch': 0.96}

 10%|▉         | 1840/19090 [54:00<8:09:54,  1.70s/it]
 10%|▉         | 1841/19090 [54:01<7:33:49,  1.58s/it]
 10%|▉         | 1842/19090 [54:03<8:22:13,  1.75s/it]
 10%|▉         | 1843/19090 [54:05<7:42:21,  1.61s/it]
 10%|▉         | 1844/19090 [54:06<7:14:58,  1.51s/it]
 10%|▉         | 1845/19090 [54:08<7:48:21,  1.63s/it]
 10%|▉         | 1846/19090 [54:09<7:18:49,  1.53s/it]
 10%|▉         | 1847/19090 [54:11<6:58:22,  1.46s/it]
 10%|▉         | 1848/19090 [54:13<8:08:06,  1.70s/it]
 10%|▉         | 1849/19090 [54:14<7:34:09,  1.58s/it]
 10%|▉         | 1850/19090 [54:15<7:08:48,  1.49s/it]
                                                      
{'loss': 0.0075, 'grad_norm': 0.6853473782539368, 'learning_rate': 4.842849659507596e-05, 'recon_loss': 0.007550908997654915, 'recon_static': 0.00654271476669237, 'recon_motion': 0.0010081944084959105, 'recon_norm': 2.9517111480236053, 'recon_norm_std': 0.06856528380885721, 'target_norm': 1.6213591188192367, 'recon_cos': 0.30286188423633575, 'gradnorm/compressor.query': 0.010771863150724113, 'gradnorm/compressor.layer0': 0.19698263397439883, 'gradnorm/compressor.layer1': 0.25310134953813895, 'gradnorm/compressor.layer2': 0.21656360620433565, 'gradnorm/compressor.layer3': 0.20014729738128514, 'gradnorm/compressor.layer4': 0.19057088654500126, 'gradnorm/compressor.layer5': 0.17912488429857074, 'gradnorm/compressor.layer6': 0.16833664835857987, 'gradnorm/compressor.layer7': 0.16486304111506156, 'gradnorm/decoder.layer0': 0.20424583118321235, 'gradnorm/decoder.layer1': 0.19676746125880182, 'gradnorm/decoder.layer2': 0.27460876019817354, 'gradnorm/max_param': 0.19545291364192963, 'epoch': 0.97}

 10%|▉         | 1850/19090 [54:15<7:08:48,  1.49s/it]
 10%|▉         | 1851/19090 [54:17<7:41:38,  1.61s/it]
 10%|▉         | 1852/19090 [54:19<7:14:00,  1.51s/it]
 10%|▉         | 1853/19090 [54:20<6:54:29,  1.44s/it]
 10%|▉         | 1854/19090 [54:22<8:29:46,  1.77s/it]
 10%|▉         | 1855/19090 [54:24<7:47:21,  1.63s/it]
 10%|▉         | 1856/19090 [54:25<7:18:41,  1.53s/it]
 10%|▉         | 1857/19090 [54:28<8:45:11,  1.83s/it]
 10%|▉         | 1858/19090 [54:29<7:58:29,  1.67s/it]
 10%|▉         | 1859/19090 [54:30<7:26:28,  1.55s/it]
 10%|▉         | 1860/19090 [54:33<9:12:14,  1.92s/it]
                                                      
{'loss': 0.0059, 'grad_norm': 0.7974567413330078, 'learning_rate': 4.869041382922996e-05, 'recon_loss': 0.0059193450142629445, 'recon_static': 0.004922729893587529, 'recon_motion': 0.000996615126496181, 'recon_norm': 2.619467169046402, 'recon_norm_std': 0.07719946298748255, 'target_norm': 1.6190918147563935, 'recon_cos': 0.3333749048411846, 'gradnorm/compressor.query': 0.012619893206321075, 'gradnorm/compressor.layer0': 0.23078984183225754, 'gradnorm/compressor.layer1': 0.2956183464438694, 'gradnorm/compressor.layer2': 0.24798517886652774, 'gradnorm/compressor.layer3': 0.23177284693468325, 'gradnorm/compressor.layer4': 0.2157344042997022, 'gradnorm/compressor.layer5': 0.2070846758590434, 'gradnorm/compressor.layer6': 0.20050478020520499, 'gradnorm/compressor.layer7': 0.18878211321394225, 'gradnorm/decoder.layer0': 0.23717099415036075, 'gradnorm/decoder.layer1': 0.23016200422926486, 'gradnorm/decoder.layer2': 0.32468081490369044, 'gradnorm/max_param': 0.22729802131652832, 'epoch': 0.97}

 10%|▉         | 1860/19090 [54:33<9:12:14,  1.92s/it]
 10%|▉         | 1861/19090 [54:34<8:16:58,  1.73s/it]
 10%|▉         | 1862/19090 [54:35<7:38:31,  1.60s/it]
 10%|▉         | 1863/19090 [54:38<9:20:31,  1.95s/it]
 10%|▉         | 1864/19090 [54:39<8:22:48,  1.75s/it]
 10%|▉         | 1865/19090 [54:41<7:42:45,  1.61s/it]
 10%|▉         | 1866/19090 [54:44<9:18:11,  1.94s/it]
 10%|▉         | 1867/19090 [54:45<8:22:24,  1.75s/it]
 10%|▉         | 1868/19090 [54:46<7:42:14,  1.61s/it]
 10%|▉         | 1869/19090 [54:49<8:52:26,  1.86s/it]
 10%|▉         | 1870/19090 [54:50<8:03:32,  1.68s/it]
                                                      
{'loss': 0.0066, 'grad_norm': 0.5671477913856506, 'learning_rate': 4.895233106338397e-05, 'recon_loss': 0.006584987696260214, 'recon_static': 0.005576842912705615, 'recon_motion': 0.001008144773368258, 'recon_norm': 2.7543406665325163, 'recon_norm_std': 0.07214508298784494, 'target_norm': 1.6215386897325517, 'recon_cos': 0.3220288783311844, 'gradnorm/compressor.query': 0.008516222151299942, 'gradnorm/compressor.layer0': 0.15392228420599322, 'gradnorm/compressor.layer1': 0.20584433788414253, 'gradnorm/compressor.layer2': 0.17536972936695777, 'gradnorm/compressor.layer3': 0.1652157137187561, 'gradnorm/compressor.layer4': 0.1560777966954812, 'gradnorm/compressor.layer5': 0.1497999270104261, 'gradnorm/compressor.layer6': 0.14283580226815634, 'gradnorm/compressor.layer7': 0.1374345772789378, 'gradnorm/decoder.layer0': 0.1717906549879465, 'gradnorm/decoder.layer1': 0.16715625043074914, 'gradnorm/decoder.layer2': 0.23221856529383525, 'gradnorm/max_param': 0.16393820941448212, 'epoch': 0.98}

 10%|▉         | 1870/19090 [54:50<8:03:32,  1.68s/it]
 10%|▉         | 1871/19090 [54:51<7:30:26,  1.57s/it]
 10%|▉         | 1872/19090 [54:54<9:27:09,  1.98s/it]
 10%|▉         | 1873/19090 [54:55<8:27:37,  1.77s/it]
 10%|▉         | 1874/19090 [54:57<7:45:49,  1.62s/it]
 10%|▉         | 1875/19090 [54:59<8:53:48,  1.86s/it]
 10%|▉         | 1876/19090 [55:00<8:04:15,  1.69s/it]
 10%|▉         | 1877/19090 [55:02<7:30:38,  1.57s/it]
 10%|▉         | 1878/19090 [55:04<9:10:12,  1.92s/it]
 10%|▉         | 1879/19090 [55:06<8:15:25,  1.73s/it]
 10%|▉         | 1880/19090 [55:07<7:38:33,  1.60s/it]
                                                      
{'loss': 0.0045, 'grad_norm': 0.5781188011169434, 'learning_rate': 4.921424829753798e-05, 'recon_loss': 0.004515587736386805, 'recon_static': 0.003533572278684005, 'recon_motion': 0.000982015419867821, 'recon_norm': 2.3012316465377807, 'recon_norm_std': 0.08493779394775629, 'target_norm': 1.6193513929843903, 'recon_cos': 0.3840003624558449, 'gradnorm/compressor.query': 0.009697332982816645, 'gradnorm/compressor.layer0': 0.16479576393194878, 'gradnorm/compressor.layer1': 0.2106404774098459, 'gradnorm/compressor.layer2': 0.18452831577678194, 'gradnorm/compressor.layer3': 0.16711673219240789, 'gradnorm/compressor.layer4': 0.15756886106540663, 'gradnorm/compressor.layer5': 0.15076294550697422, 'gradnorm/compressor.layer6': 0.14436327348910313, 'gradnorm/compressor.layer7': 0.13949543233711825, 'gradnorm/decoder.layer0': 0.17367830141904406, 'gradnorm/decoder.layer1': 0.16738252027701972, 'gradnorm/decoder.layer2': 0.23363076302720584, 'gradnorm/max_param': 0.16625447571277618, 'epoch': 0.98}

 10%|▉         | 1880/19090 [55:07<7:38:33,  1.60s/it]
 10%|▉         | 1881/19090 [55:10<9:06:21,  1.90s/it]
 10%|▉         | 1882/19090 [55:11<8:13:03,  1.72s/it]
 10%|▉         | 1883/19090 [55:12<7:36:48,  1.59s/it]
 10%|▉         | 1884/19090 [55:15<8:59:00,  1.88s/it]
 10%|▉         | 1885/19090 [55:16<8:07:33,  1.70s/it]
 10%|▉         | 1886/19090 [55:17<7:32:43,  1.58s/it]
 10%|▉         | 1887/19090 [55:20<9:11:10,  1.92s/it]
 10%|▉         | 1888/19090 [55:21<8:16:38,  1.73s/it]
 10%|▉         | 1889/19090 [55:23<7:37:52,  1.60s/it]
 10%|▉         | 1890/19090 [55:25<9:13:23,  1.93s/it]
                                                      
{'loss': 0.004, 'grad_norm': 0.5216410160064697, 'learning_rate': 4.947616553169199e-05, 'recon_loss': 0.003970065974863246, 'recon_static': 0.002971271250862628, 'recon_motion': 0.0009987948724301532, 'recon_norm': 2.1573527812957765, 'recon_norm_std': 0.08654892649501562, 'target_norm': 1.6223774522542953, 'recon_cos': 0.41022649481892587, 'gradnorm/compressor.query': 0.007918723953735314, 'gradnorm/compressor.layer0': 0.15002546317817742, 'gradnorm/compressor.layer1': 0.19254720025598873, 'gradnorm/compressor.layer2': 0.16436946059092647, 'gradnorm/compressor.layer3': 0.1497883751752509, 'gradnorm/compressor.layer4': 0.1409010429553884, 'gradnorm/compressor.layer5': 0.1341692888542184, 'gradnorm/compressor.layer6': 0.12929266032469092, 'gradnorm/compressor.layer7': 0.12379669127425985, 'gradnorm/decoder.layer0': 0.1599399571522563, 'gradnorm/decoder.layer1': 0.15261842345886215, 'gradnorm/decoder.layer2': 0.2103349503782077, 'gradnorm/max_param': 0.14945480227470398, 'epoch': 0.99}

 10%|▉         | 1890/19090 [55:25<9:13:23,  1.93s/it]
 10%|▉         | 1891/19090 [55:27<8:18:31,  1.74s/it]
 10%|▉         | 1892/19090 [55:28<7:39:09,  1.60s/it]
 10%|▉         | 1893/19090 [55:31<9:16:25,  1.94s/it]
 10%|▉         | 1894/19090 [55:32<8:20:23,  1.75s/it]
 10%|▉         | 1895/19090 [55:33<7:41:32,  1.61s/it]
 10%|▉         | 1896/19090 [55:36<9:20:41,  1.96s/it]
 10%|▉         | 1897/19090 [55:37<8:23:39,  1.76s/it]
 10%|▉         | 1898/19090 [55:38<7:43:17,  1.62s/it]
 10%|▉         | 1899/19090 [55:42<9:51:02,  2.06s/it]
 10%|▉         | 1900/19090 [55:43<8:43:52,  1.83s/it]
                                                      
{'loss': 0.006, 'grad_norm': 0.9714177250862122, 'learning_rate': 4.973808276584599e-05, 'recon_loss': 0.005989252391736954, 'recon_static': 0.004987224005162716, 'recon_motion': 0.0010020282832556404, 'recon_norm': 2.6117190599441527, 'recon_norm_std': 0.07443682989105582, 'target_norm': 1.620579907298088, 'recon_cos': 0.341611360386014, 'gradnorm/compressor.query': 0.015249037010481208, 'gradnorm/compressor.layer0': 0.27106330656670913, 'gradnorm/compressor.layer1': 0.3562940511740059, 'gradnorm/compressor.layer2': 0.3029150024042145, 'gradnorm/compressor.layer3': 0.2823284189762226, 'gradnorm/compressor.layer4': 0.2654875511889613, 'gradnorm/compressor.layer5': 0.2505584380595381, 'gradnorm/compressor.layer6': 0.24267076913120986, 'gradnorm/compressor.layer7': 0.23499906425594136, 'gradnorm/decoder.layer0': 0.2919433323231232, 'gradnorm/decoder.layer1': 0.28256577430756846, 'gradnorm/decoder.layer2': 0.3988497788905305, 'gradnorm/max_param': 0.2806822955608368, 'epoch': 1.0}

 10%|▉         | 1900/19090 [55:43<8:43:52,  1.83s/it]
 10%|▉         | 1901/19090 [55:44<7:57:39,  1.67s/it]
 10%|▉         | 1902/19090 [55:47<9:41:18,  2.03s/it]
 10%|▉         | 1903/19090 [55:48<8:42:29,  1.82s/it]
 10%|▉         | 1904/19090 [55:50<7:56:56,  1.67s/it]
 10%|▉         | 1905/19090 [55:52<8:45:04,  1.83s/it]
 10%|▉         | 1906/19090 [55:53<7:56:20,  1.66s/it]
 10%|▉         | 1907/19090 [55:54<7:22:17,  1.54s/it]
 10%|▉         | 1908/19090 [55:56<6:58:24,  1.46s/it]
 10%|█         | 1909/19090 [55:57<7:24:29,  1.55s/it]
 10%|█         | 1910/19090 [56:05<15:50:28,  3.32s/it]
                                                       
{'loss': 0.0089, 'grad_norm': 0.8938829302787781, 'learning_rate': 5e-05, 'recon_loss': 0.008872801205143332, 'recon_static': 0.007858840248081833, 'recon_motion': 0.0010139609832549468, 'recon_norm': 3.2088368773460387, 'recon_norm_std': 0.061974899377673864, 'target_norm': 1.6221486806869507, 'recon_cos': 0.27504044957458973, 'gradnorm/compressor.query': 0.013442788030507642, 'gradnorm/compressor.layer0': 0.2551168047333046, 'gradnorm/compressor.layer1': 0.32806080235575996, 'gradnorm/compressor.layer2': 0.2771643743576969, 'gradnorm/compressor.layer3': 0.2569784631991387, 'gradnorm/compressor.layer4': 0.2412931325383015, 'gradnorm/compressor.layer5': 0.23214642045802292, 'gradnorm/compressor.layer6': 0.2248446186306036, 'gradnorm/compressor.layer7': 0.21803234943103403, 'gradnorm/decoder.layer0': 0.27028897795014434, 'gradnorm/decoder.layer1': 0.2603785804872286, 'gradnorm/decoder.layer2': 0.36361912427753235, 'gradnorm/max_param': 0.25970715284347534, 'epoch': 1.0}

 10%|█         | 1910/19090 [56:05<15:50:28,  3.32s/it]
 10%|█         | 1911/19090 [56:06<12:56:19,  2.71s/it]
 10%|█         | 1912/19090 [56:09<13:19:43,  2.79s/it]
 10%|█         | 1913/19090 [56:10<11:11:40,  2.35s/it]
 10%|█         | 1914/19090 [56:12<9:44:14,  2.04s/it] 
 10%|█         | 1915/19090 [56:13<8:44:57,  1.83s/it]
 10%|█         | 1916/19090 [56:14<7:59:17,  1.67s/it]
 10%|█         | 1917/19090 [56:16<7:27:01,  1.56s/it]
 10%|█         | 1918/19090 [56:18<8:03:14,  1.69s/it]
 10%|█         | 1919/19090 [56:19<7:28:43,  1.57s/it]
 10%|█         | 1920/19090 [56:20<7:05:26,  1.49s/it]
                                                      
{'loss': 0.0076, 'grad_norm': 0.6992345452308655, 'learning_rate': 4.999995820611324e-05, 'recon_loss': 0.007624845765531063, 'recon_static': 0.00662979717599228, 'recon_motion': 0.0009950486564775929, 'recon_norm': 2.9722672283649443, 'recon_norm_std': 0.06811873987317085, 'target_norm': 1.6205770611763, 'recon_cos': 0.2986399568617344, 'gradnorm/compressor.query': 0.009894737926253068, 'gradnorm/compressor.layer0': 0.19345145489485008, 'gradnorm/compressor.layer1': 0.2537329889571236, 'gradnorm/compressor.layer2': 0.21798598176939443, 'gradnorm/compressor.layer3': 0.20405240356681392, 'gradnorm/compressor.layer4': 0.18874231686017848, 'gradnorm/compressor.layer5': 0.1813714791849393, 'gradnorm/compressor.layer6': 0.17554424520745882, 'gradnorm/compressor.layer7': 0.16842600771742808, 'gradnorm/decoder.layer0': 0.21255889034487147, 'gradnorm/decoder.layer1': 0.20524824443373008, 'gradnorm/decoder.layer2': 0.2878564340689363, 'gradnorm/max_param': 0.20393617451190948, 'epoch': 1.01}

 10%|█         | 1920/19090 [56:20<7:05:26,  1.49s/it]
 10%|█         | 1921/19090 [56:22<6:54:10,  1.45s/it]
 10%|█         | 1922/19090 [56:23<6:40:29,  1.40s/it]
 10%|█         | 1923/19090 [56:24<6:31:11,  1.37s/it]
 10%|█         | 1924/19090 [56:26<7:19:35,  1.54s/it]
 10%|█         | 1925/19090 [56:27<6:58:14,  1.46s/it]
 10%|█         | 1926/19090 [56:29<6:43:34,  1.41s/it]
 10%|█         | 1927/19090 [56:31<7:21:05,  1.54s/it]
 10%|█         | 1928/19090 [56:32<6:59:20,  1.47s/it]
 10%|█         | 1929/19090 [56:33<6:44:01,  1.41s/it]
 10%|█         | 1930/19090 [56:35<7:43:30,  1.62s/it]
                                                      
{'loss': 0.004, 'grad_norm': 0.363989919424057, 'learning_rate': 4.999983282459268e-05, 'recon_loss': 0.00399698379333131, 'recon_static': 0.0030091318709310143, 'recon_motion': 0.000987851939862594, 'recon_norm': 2.151553365588188, 'recon_norm_std': 0.08731775842607022, 'target_norm': 1.620811602473259, 'recon_cos': 0.4139733150601387, 'gradnorm/compressor.query': 0.0053374538490577545, 'gradnorm/compressor.layer0': 0.10042268968295912, 'gradnorm/compressor.layer1': 0.13095058991409284, 'gradnorm/compressor.layer2': 0.11184576640239295, 'gradnorm/compressor.layer3': 0.10449639427772421, 'gradnorm/compressor.layer4': 0.09873096850803567, 'gradnorm/compressor.layer5': 0.09422947541043693, 'gradnorm/compressor.layer6': 0.09142252500032742, 'gradnorm/compressor.layer7': 0.08834420624157384, 'gradnorm/decoder.layer0': 0.11077087205772926, 'gradnorm/decoder.layer1': 0.10728411510968626, 'gradnorm/decoder.layer2': 0.152409294895826, 'gradnorm/max_param': 0.10875193774700165, 'epoch': 1.01}

 10%|█         | 1930/19090 [56:35<7:43:30,  1.62s/it]
 10%|█         | 1931/19090 [56:37<7:15:13,  1.52s/it]
 10%|█         | 1932/19090 [56:38<6:55:33,  1.45s/it]
 10%|█         | 1933/19090 [56:40<7:48:12,  1.64s/it]
 10%|█         | 1934/19090 [56:41<7:17:58,  1.53s/it]
 10%|█         | 1935/19090 [56:43<6:57:36,  1.46s/it]
 10%|█         | 1936/19090 [56:44<7:25:45,  1.56s/it]
 10%|█         | 1937/19090 [56:46<7:47:08,  1.63s/it]
 10%|█         | 1938/19090 [56:47<7:17:33,  1.53s/it]
 10%|█         | 1939/19090 [56:49<7:54:57,  1.66s/it]
 10%|█         | 1940/19090 [56:51<7:34:14,  1.59s/it]
                                                      
{'loss': 0.003, 'grad_norm': 0.46540141105651855, 'learning_rate': 4.9999623855857544e-05, 'recon_loss': 0.0029630188713781536, 'recon_static': 0.0019729873485630376, 'recon_motion': 0.0009900314718834125, 'recon_norm': 1.8692025005817414, 'recon_norm_std': 0.09650330469012261, 'target_norm': 1.6206729710102081, 'recon_cos': 0.4721168138086796, 'gradnorm/compressor.query': 0.0066676838050072264, 'gradnorm/compressor.layer0': 0.1285232214146442, 'gradnorm/compressor.layer1': 0.1706157471607226, 'gradnorm/compressor.layer2': 0.14623436255281225, 'gradnorm/compressor.layer3': 0.13460940137963637, 'gradnorm/compressor.layer4': 0.12635026992450876, 'gradnorm/compressor.layer5': 0.12154308522602124, 'gradnorm/compressor.layer6': 0.11572168647585111, 'gradnorm/compressor.layer7': 0.1111939357682257, 'gradnorm/decoder.layer0': 0.14136446472438274, 'gradnorm/decoder.layer1': 0.13645488616966986, 'gradnorm/decoder.layer2': 0.19059204614055622, 'gradnorm/max_param': 0.13398201763629913, 'epoch': 1.02}

 10%|█         | 1940/19090 [56:51<7:34:14,  1.59s/it]
 10%|█         | 1941/19090 [56:52<7:08:49,  1.50s/it]
 10%|█         | 1942/19090 [56:54<7:52:27,  1.65s/it]
 10%|█         | 1943/19090 [56:56<7:54:02,  1.66s/it]
 10%|█         | 1944/19090 [56:57<7:22:31,  1.55s/it]
 10%|█         | 1945/19090 [56:59<7:58:14,  1.67s/it]
 10%|█         | 1946/19090 [57:01<8:06:35,  1.70s/it]
 10%|█         | 1947/19090 [57:02<7:31:03,  1.58s/it]
 10%|█         | 1948/19090 [57:04<7:41:00,  1.61s/it]
 10%|█         | 1949/19090 [57:06<8:09:29,  1.71s/it]
 10%|█         | 1950/19090 [57:07<7:33:55,  1.59s/it]
                                                      
{'loss': 0.0042, 'grad_norm': 0.5306325554847717, 'learning_rate': 4.999933130060651e-05, 'recon_loss': 0.004182048141956329, 'recon_static': 0.0031872148101683706, 'recon_motion': 0.0009948333987267687, 'recon_norm': 2.2129343271255495, 'recon_norm_std': 0.08360285609960556, 'target_norm': 1.6212269127368928, 'recon_cos': 0.39941536709666253, 'gradnorm/compressor.query': 0.008146233090689008, 'gradnorm/compressor.layer0': 0.14874590657716621, 'gradnorm/compressor.layer1': 0.19210916993082527, 'gradnorm/compressor.layer2': 0.16556428817720825, 'gradnorm/compressor.layer3': 0.15583904583128244, 'gradnorm/compressor.layer4': 0.14617960059761897, 'gradnorm/compressor.layer5': 0.13941602774464343, 'gradnorm/compressor.layer6': 0.13402371121064235, 'gradnorm/compressor.layer7': 0.12812605921377254, 'gradnorm/decoder.layer0': 0.15943561060596073, 'gradnorm/decoder.layer1': 0.15489027142453096, 'gradnorm/decoder.layer2': 0.2148982759122088, 'gradnorm/max_param': 0.1534433215856552, 'epoch': 1.02}

 10%|█         | 1950/19090 [57:07<7:33:55,  1.59s/it]
 10%|█         | 1951/19090 [57:09<8:16:35,  1.74s/it]
 10%|█         | 1952/19090 [57:11<8:07:47,  1.71s/it]
 10%|█         | 1953/19090 [57:12<7:31:43,  1.58s/it]
 10%|█         | 1954/19090 [57:14<7:54:08,  1.66s/it]
 10%|█         | 1955/19090 [57:15<7:30:54,  1.58s/it]
 10%|█         | 1956/19090 [57:17<7:07:09,  1.50s/it]
 10%|█         | 1957/19090 [57:19<7:49:55,  1.65s/it]
 10%|█         | 1958/19090 [57:20<8:09:01,  1.71s/it]
 10%|█         | 1959/19090 [57:22<7:33:53,  1.59s/it]
 10%|█         | 1960/19090 [57:24<8:15:30,  1.74s/it]
                                                      
{'loss': 0.0051, 'grad_norm': 0.6951536536216736, 'learning_rate': 4.999895515981776e-05, 'recon_loss': 0.005126988317351788, 'recon_static': 0.004137861245544628, 'recon_motion': 0.0009891271343803964, 'recon_norm': 2.4465246737003326, 'recon_norm_std': 0.07742785755544901, 'target_norm': 1.6190916866064071, 'recon_cos': 0.36009508892893793, 'gradnorm/compressor.query': 0.009629469405539333, 'gradnorm/compressor.layer0': 0.18745844102275452, 'gradnorm/compressor.layer1': 0.25134876127368444, 'gradnorm/compressor.layer2': 0.21573279837463716, 'gradnorm/compressor.layer3': 0.2026210181346862, 'gradnorm/compressor.layer4': 0.19349285296676363, 'gradnorm/compressor.layer5': 0.18289675197579908, 'gradnorm/compressor.layer6': 0.17452445023796673, 'gradnorm/compressor.layer7': 0.16797661762885113, 'gradnorm/decoder.layer0': 0.2100629425662735, 'gradnorm/decoder.layer1': 0.202514712820566, 'gradnorm/decoder.layer2': 0.2872781970831751, 'gradnorm/max_param': 0.20284877717494965, 'epoch': 1.03}

 10%|█         | 1960/19090 [57:24<8:15:30,  1.74s/it]
 10%|█         | 1961/19090 [57:25<8:09:02,  1.71s/it]
 10%|█         | 1962/19090 [57:27<7:34:29,  1.59s/it]
 10%|█         | 1963/19090 [57:29<8:13:04,  1.73s/it]
 10%|█         | 1964/19090 [57:31<8:20:37,  1.75s/it]
 10%|█         | 1965/19090 [57:32<7:40:51,  1.61s/it]
 10%|█         | 1966/19090 [57:34<8:16:56,  1.74s/it]
 10%|█         | 1967/19090 [57:35<7:49:34,  1.65s/it]
 10%|█         | 1968/19090 [57:37<7:19:15,  1.54s/it]
 10%|█         | 1969/19090 [57:39<8:55:39,  1.88s/it]
 10%|█         | 1970/19090 [57:41<8:05:24,  1.70s/it]
                                                      
{'loss': 0.0053, 'grad_norm': 0.6089839935302734, 'learning_rate': 4.999849543474891e-05, 'recon_loss': 0.0052760865073651075, 'recon_static': 0.0042804486176464705, 'recon_motion': 0.0009956381691154093, 'recon_norm': 2.479567086696625, 'recon_norm_std': 0.07680573984980583, 'target_norm': 1.6216303050518035, 'recon_cos': 0.35464033856987953, 'gradnorm/compressor.query': 0.009060539223476953, 'gradnorm/compressor.layer0': 0.1687390164933452, 'gradnorm/compressor.layer1': 0.22193256524641722, 'gradnorm/compressor.layer2': 0.19275055027671667, 'gradnorm/compressor.layer3': 0.17889975906657568, 'gradnorm/compressor.layer4': 0.1658435996140998, 'gradnorm/compressor.layer5': 0.1566910657743238, 'gradnorm/compressor.layer6': 0.151756880179233, 'gradnorm/compressor.layer7': 0.14555042718104225, 'gradnorm/decoder.layer0': 0.18288906279250716, 'gradnorm/decoder.layer1': 0.17742695264046837, 'gradnorm/decoder.layer2': 0.2503889158193542, 'gradnorm/max_param': 0.17806130647659302, 'epoch': 1.03}

 10%|█         | 1970/19090 [57:41<8:05:24,  1.70s/it]
 10%|█         | 1971/19090 [57:42<7:31:14,  1.58s/it]
 10%|█         | 1972/19090 [57:44<8:06:44,  1.71s/it]
 10%|█         | 1973/19090 [57:45<7:31:21,  1.58s/it]
 10%|█         | 1974/19090 [57:47<7:05:53,  1.49s/it]
 10%|█         | 1975/19090 [57:49<8:48:23,  1.85s/it]
 10%|█         | 1976/19090 [57:51<8:00:28,  1.68s/it]
 10%|█         | 1977/19090 [57:52<7:26:44,  1.57s/it]
 10%|█         | 1978/19090 [57:54<8:30:30,  1.79s/it]
 10%|█         | 1979/19090 [57:55<7:47:32,  1.64s/it]
 10%|█         | 1980/19090 [57:57<7:17:54,  1.54s/it]
                                                      
{'loss': 0.0046, 'grad_norm': 0.5280138254165649, 'learning_rate': 4.9997952126937056e-05, 'recon_loss': 0.004595200100447982, 'recon_static': 0.003603551781270653, 'recon_motion': 0.0009916481401887722, 'recon_norm': 2.3187572717666627, 'recon_norm_std': 0.08065557386726141, 'target_norm': 1.621171772480011, 'recon_cos': 0.3806312926113605, 'gradnorm/compressor.query': 0.008111779502362663, 'gradnorm/compressor.layer0': 0.1496387487713903, 'gradnorm/compressor.layer1': 0.1960761780483718, 'gradnorm/compressor.layer2': 0.16498331674669398, 'gradnorm/compressor.layer3': 0.15301671919367685, 'gradnorm/compressor.layer4': 0.1436892158078977, 'gradnorm/compressor.layer5': 0.13729647100399373, 'gradnorm/compressor.layer6': 0.13112101091830086, 'gradnorm/compressor.layer7': 0.12709493053264123, 'gradnorm/decoder.layer0': 0.1553192131176645, 'gradnorm/decoder.layer1': 0.15318692838071135, 'gradnorm/decoder.layer2': 0.21639890486167762, 'gradnorm/max_param': 0.15439286828041077, 'epoch': 1.04}

 10%|█         | 1980/19090 [57:57<7:17:54,  1.54s/it]
 10%|█         | 1981/19090 [57:59<8:19:04,  1.75s/it]
 10%|█         | 1982/19090 [58:00<7:39:22,  1.61s/it]
 10%|█         | 1983/19090 [58:02<7:11:45,  1.51s/it]
 10%|█         | 1984/19090 [58:04<8:52:39,  1.87s/it]
 10%|█         | 1985/19090 [58:06<8:03:48,  1.70s/it]
 10%|█         | 1986/19090 [58:07<7:29:31,  1.58s/it]
 10%|█         | 1987/19090 [58:09<8:40:21,  1.83s/it]
 10%|█         | 1988/19090 [58:11<7:54:50,  1.67s/it]
 10%|█         | 1989/19090 [58:12<7:23:23,  1.56s/it]
 10%|█         | 1990/19090 [58:14<8:57:36,  1.89s/it]
                                                      
{'loss': 0.0034, 'grad_norm': 0.3780016601085663, 'learning_rate': 4.999732523819876e-05, 'recon_loss': 0.0033710108429659156, 'recon_static': 0.002391010127030313, 'recon_motion': 0.0009800006708246655, 'recon_norm': 1.996364802122116, 'recon_norm_std': 0.09057793635874986, 'target_norm': 1.618210217356682, 'recon_cos': 0.44548539742827414, 'gradnorm/compressor.query': 0.006091549886251272, 'gradnorm/compressor.layer0': 0.10670071767179758, 'gradnorm/compressor.layer1': 0.1381514886015321, 'gradnorm/compressor.layer2': 0.11601419785132333, 'gradnorm/compressor.layer3': 0.10852653220654217, 'gradnorm/compressor.layer4': 0.10058381717261844, 'gradnorm/compressor.layer5': 0.09721291044275379, 'gradnorm/compressor.layer6': 0.0952893528034304, 'gradnorm/compressor.layer7': 0.09108859603788312, 'gradnorm/decoder.layer0': 0.11482384259388942, 'gradnorm/decoder.layer1': 0.11058539117228777, 'gradnorm/decoder.layer2': 0.15742195616878565, 'gradnorm/max_param': 0.11132194846868515, 'epoch': 1.04}

 10%|█         | 1990/19090 [58:14<8:57:36,  1.89s/it]
 10%|█         | 1991/19090 [58:16<8:07:11,  1.71s/it]
 10%|█         | 1992/19090 [58:17<7:32:04,  1.59s/it]
 10%|█         | 1993/19090 [58:19<8:31:31,  1.80s/it]
 10%|█         | 1994/19090 [58:21<7:48:28,  1.64s/it]
 10%|█         | 1995/19090 [58:22<7:18:17,  1.54s/it]
 10%|█         | 1996/19090 [58:24<8:07:39,  1.71s/it]
 10%|█         | 1997/19090 [58:25<7:31:27,  1.58s/it]
 10%|█         | 1998/19090 [58:27<7:05:54,  1.50s/it]
 10%|█         | 1999/19090 [58:29<8:46:29,  1.85s/it]
 10%|█         | 2000/19090 [58:31<7:59:16,  1.68s/it]
                                                      
{'loss': 0.004, 'grad_norm': 0.6483995914459229, 'learning_rate': 4.9996614770630027e-05, 'recon_loss': 0.0040287198033183815, 'recon_static': 0.003041116168606095, 'recon_motion': 0.0009876037438516505, 'recon_norm': 2.1680755406618117, 'recon_norm_std': 0.0854924051091075, 'target_norm': 1.6199645847082138, 'recon_cos': 0.4128062859177589, 'gradnorm/compressor.query': 0.010628054998431618, 'gradnorm/compressor.layer0': 0.1857617589545224, 'gradnorm/compressor.layer1': 0.23792490344200407, 'gradnorm/compressor.layer2': 0.20437282806082332, 'gradnorm/compressor.layer3': 0.18960728223327103, 'gradnorm/compressor.layer4': 0.17434457075900525, 'gradnorm/compressor.layer5': 0.1656305784509601, 'gradnorm/compressor.layer6': 0.16376063414252418, 'gradnorm/compressor.layer7': 0.1541981683155348, 'gradnorm/decoder.layer0': 0.1935254453093138, 'gradnorm/decoder.layer1': 0.18663696509022673, 'gradnorm/decoder.layer2': 0.2660162092474125, 'gradnorm/max_param': 0.18939495086669922, 'epoch': 1.05}

 10%|█         | 2000/19090 [58:31<7:59:16,  1.68s/it]
 10%|█         | 2001/19090 [58:35<11:29:18,  2.42s/it]
 10%|█         | 2002/19090 [58:36<9:53:32,  2.08s/it] 
 10%|█         | 2003/19090 [58:37<8:45:38,  1.85s/it]
 10%|█         | 2004/19090 [58:39<7:58:17,  1.68s/it]
 11%|█         | 2005/19090 [58:40<7:24:49,  1.56s/it]
 11%|█         | 2006/19090 [58:41<7:01:32,  1.48s/it]
 11%|█         | 2007/19090 [58:42<6:45:25,  1.42s/it]
 11%|█         | 2008/19090 [58:44<6:33:56,  1.38s/it]
 11%|█         | 2009/19090 [58:45<6:25:37,  1.35s/it]
 11%|█         | 2010/19090 [58:46<6:20:32,  1.34s/it]
                                                      
{'loss': 0.0047, 'grad_norm': 0.553121030330658, 'learning_rate': 4.999582072660631e-05, 'recon_loss': 0.004686699633020907, 'recon_static': 0.003688940341817215, 'recon_motion': 0.000997759247547947, 'recon_norm': 2.339384877681732, 'recon_norm_std': 0.0804733557626605, 'target_norm': 1.6233405351638794, 'recon_cos': 0.37649751231074335, 'gradnorm/compressor.query': 0.009036788760398467, 'gradnorm/compressor.layer0': 0.15963321272514644, 'gradnorm/compressor.layer1': 0.20178129423264174, 'gradnorm/compressor.layer2': 0.17264426996776774, 'gradnorm/compressor.layer3': 0.1608678604581721, 'gradnorm/compressor.layer4': 0.1495649784837643, 'gradnorm/compressor.layer5': 0.1446250276219334, 'gradnorm/compressor.layer6': 0.1381620046512501, 'gradnorm/compressor.layer7': 0.13281735243067175, 'gradnorm/decoder.layer0': 0.16413197862746579, 'gradnorm/decoder.layer1': 0.1617228887553594, 'gradnorm/decoder.layer2': 0.2255424193314442, 'gradnorm/max_param': 0.16091220080852509, 'epoch': 1.05}

 11%|█         | 2010/19090 [58:46<6:20:32,  1.34s/it]
 11%|█         | 2011/19090 [58:48<6:17:28,  1.33s/it]
 11%|█         | 2012/19090 [58:49<6:14:48,  1.32s/it]
 11%|█         | 2013/19090 [58:50<6:12:32,  1.31s/it]
 11%|█         | 2014/19090 [58:52<6:10:56,  1.30s/it]
 11%|█         | 2015/19090 [58:53<6:46:53,  1.43s/it]
 11%|█         | 2016/19090 [58:55<7:07:43,  1.50s/it]
 11%|█         | 2017/19090 [58:57<7:13:57,  1.53s/it]
 11%|█         | 2018/19090 [58:58<7:00:37,  1.48s/it]
 11%|█         | 2019/19090 [59:00<7:58:35,  1.68s/it]
 11%|█         | 2020/19090 [59:02<7:53:19,  1.66s/it]
                                                      
{'loss': 0.0049, 'grad_norm': 0.5665603280067444, 'learning_rate': 4.999494310878252e-05, 'recon_loss': 0.004935332189779729, 'recon_static': 0.003945857367943973, 'recon_motion': 0.0009894747447106056, 'recon_norm': 2.401772993803024, 'recon_norm_std': 0.07772896755486727, 'target_norm': 1.621592602133751, 'recon_cos': 0.3672033794224262, 'gradnorm/compressor.query': 0.009074065938440658, 'gradnorm/compressor.layer0': 0.1618217709100402, 'gradnorm/compressor.layer1': 0.21044823630827184, 'gradnorm/compressor.layer2': 0.17637070264724553, 'gradnorm/compressor.layer3': 0.16337449135661022, 'gradnorm/compressor.layer4': 0.15284259747429624, 'gradnorm/compressor.layer5': 0.14746856038177775, 'gradnorm/compressor.layer6': 0.14124757652819975, 'gradnorm/compressor.layer7': 0.13411257257775136, 'gradnorm/decoder.layer0': 0.16998340325823338, 'gradnorm/decoder.layer1': 0.16440898207675017, 'gradnorm/decoder.layer2': 0.23165947224717276, 'gradnorm/max_param': 0.1659356951713562, 'epoch': 1.06}

 11%|█         | 2020/19090 [59:02<7:53:19,  1.66s/it]
 11%|█         | 2021/19090 [59:04<8:10:21,  1.72s/it]
 11%|█         | 2022/19090 [59:05<8:11:59,  1.73s/it]
 11%|█         | 2023/19090 [59:07<7:44:53,  1.63s/it]
 11%|█         | 2024/19090 [59:08<7:59:27,  1.69s/it]
 11%|█         | 2025/19090 [59:10<8:19:25,  1.76s/it]
 11%|█         | 2026/19090 [59:12<7:40:50,  1.62s/it]
 11%|█         | 2027/19090 [59:14<8:26:04,  1.78s/it]
 11%|█         | 2028/19090 [59:15<7:53:24,  1.66s/it]
 11%|█         | 2029/19090 [59:17<8:19:38,  1.76s/it]
 11%|█         | 2030/19090 [59:19<9:01:02,  1.90s/it]
                                                      
{'loss': 0.0038, 'grad_norm': 0.4640311598777771, 'learning_rate': 4.999398192009296e-05, 'recon_loss': 0.0037795413110870866, 'recon_static': 0.0027984953369013963, 'recon_motion': 0.0009810459290747532, 'recon_norm': 2.1095454394817352, 'recon_norm_std': 0.08470442667603492, 'target_norm': 1.6190698444843292, 'recon_cos': 0.419640788435936, 'gradnorm/compressor.query': 0.007335687638279053, 'gradnorm/compressor.layer0': 0.13406487735116476, 'gradnorm/compressor.layer1': 0.16948952974555365, 'gradnorm/compressor.layer2': 0.14532036196041992, 'gradnorm/compressor.layer3': 0.13476033131250573, 'gradnorm/compressor.layer4': 0.1277937494458456, 'gradnorm/compressor.layer5': 0.11805063509932844, 'gradnorm/compressor.layer6': 0.11586324786904699, 'gradnorm/compressor.layer7': 0.11020308185572762, 'gradnorm/decoder.layer0': 0.13959581544089752, 'gradnorm/decoder.layer1': 0.13483396124423427, 'gradnorm/decoder.layer2': 0.18918053197741155, 'gradnorm/max_param': 0.13649319112300873, 'epoch': 1.06}

 11%|█         | 2030/19090 [59:19<9:01:02,  1.90s/it]
 11%|█         | 2031/19090 [59:21<8:08:51,  1.72s/it]
 11%|█         | 2032/19090 [59:22<8:01:54,  1.70s/it]
 11%|█         | 2033/19090 [59:25<8:53:57,  1.88s/it]
 11%|█         | 2034/19090 [59:26<8:03:31,  1.70s/it]
 11%|█         | 2035/19090 [59:28<8:04:46,  1.71s/it]
 11%|█         | 2036/19090 [59:30<8:44:28,  1.85s/it]
 11%|█         | 2037/19090 [59:31<7:57:23,  1.68s/it]
 11%|█         | 2038/19090 [59:33<7:43:07,  1.63s/it]
 11%|█         | 2039/19090 [59:35<8:10:24,  1.73s/it]
 11%|█         | 2040/19090 [59:36<7:33:27,  1.60s/it]
                                                      
{'loss': 0.0034, 'grad_norm': 0.4261852502822876, 'learning_rate': 4.9992937163751386e-05, 'recon_loss': 0.0033734492666553707, 'recon_static': 0.002394854580052197, 'recon_motion': 0.0009785947055206633, 'recon_norm': 2.0005746692419053, 'recon_norm_std': 0.0904499577358365, 'target_norm': 1.6193236768245698, 'recon_cos': 0.44471599012613294, 'gradnorm/compressor.query': 0.006532605303153819, 'gradnorm/compressor.layer0': 0.12350651254403545, 'gradnorm/compressor.layer1': 0.15597139195662058, 'gradnorm/compressor.layer2': 0.13568231236369063, 'gradnorm/compressor.layer3': 0.12529095672633153, 'gradnorm/compressor.layer4': 0.11549147922262819, 'gradnorm/compressor.layer5': 0.11047281507326408, 'gradnorm/compressor.layer6': 0.10475074562129448, 'gradnorm/compressor.layer7': 0.1014509964453351, 'gradnorm/decoder.layer0': 0.12560014896459917, 'gradnorm/decoder.layer1': 0.12282292103416227, 'gradnorm/decoder.layer2': 0.17388022638510983, 'gradnorm/max_param': 0.12518778443336487, 'epoch': 1.07}

 11%|█         | 2040/19090 [59:36<7:33:27,  1.60s/it]
 11%|█         | 2041/19090 [59:37<7:11:47,  1.52s/it]
 11%|█         | 2042/19090 [59:39<8:11:39,  1.73s/it]
 11%|█         | 2043/19090 [59:41<7:34:27,  1.60s/it]
 11%|█         | 2044/19090 [59:42<7:07:45,  1.51s/it]
 11%|█         | 2045/19090 [59:44<8:07:28,  1.72s/it]
 11%|█         | 2046/19090 [59:46<7:30:54,  1.59s/it]
 11%|█         | 2047/19090 [59:47<7:13:04,  1.52s/it]
 11%|█         | 2048/19090 [59:49<8:04:51,  1.71s/it]
 11%|█         | 2049/19090 [59:50<7:29:18,  1.58s/it]
 11%|█         | 2050/19090 [59:52<7:28:58,  1.58s/it]
                                                      
{'loss': 0.0036, 'grad_norm': 0.6354886889457703, 'learning_rate': 4.999180884325095e-05, 'recon_loss': 0.0036381874524522573, 'recon_static': 0.002656381856650114, 'recon_motion': 0.0009818056772928686, 'recon_norm': 2.0634476810693743, 'recon_norm_std': 0.08693546615540981, 'target_norm': 1.619748768210411, 'recon_cos': 0.4331608586013317, 'gradnorm/compressor.query': 0.009971392453115513, 'gradnorm/compressor.layer0': 0.18172397546971375, 'gradnorm/compressor.layer1': 0.23332247435613857, 'gradnorm/compressor.layer2': 0.19936019887364534, 'gradnorm/compressor.layer3': 0.18686719244974467, 'gradnorm/compressor.layer4': 0.1718944679418776, 'gradnorm/compressor.layer5': 0.16407905790243404, 'gradnorm/compressor.layer6': 0.15884495862215445, 'gradnorm/compressor.layer7': 0.1495447024011315, 'gradnorm/decoder.layer0': 0.19282193100553194, 'gradnorm/decoder.layer1': 0.18246321893767156, 'gradnorm/decoder.layer2': 0.25900104405770796, 'gradnorm/max_param': 0.18498875200748444, 'epoch': 1.07}

 11%|█         | 2050/19090 [59:52<7:28:58,  1.58s/it]
 11%|█         | 2051/19090 [59:54<8:37:25,  1.82s/it]
 11%|█         | 2052/19090 [59:56<7:52:07,  1.66s/it]
 11%|█         | 2053/19090 [59:57<7:20:24,  1.55s/it]
 11%|█         | 2054/19090 [59:59<8:48:07,  1.86s/it]
 11%|█         | 2055/19090 [1:00:01<7:59:29,  1.69s/it]
 11%|█         | 2056/19090 [1:00:02<7:25:31,  1.57s/it]
 11%|█         | 2057/19090 [1:00:04<8:17:34,  1.75s/it]
 11%|█         | 2058/19090 [1:00:06<7:38:27,  1.62s/it]
 11%|█         | 2059/19090 [1:00:07<7:11:02,  1.52s/it]
 11%|█         | 2060/19090 [1:00:09<7:32:04,  1.59s/it]
                                                        
{'loss': 0.0055, 'grad_norm': 0.5847112536430359, 'learning_rate': 4.999059696236421e-05, 'recon_loss': 0.005540471954736859, 'recon_static': 0.00455931129399687, 'recon_motion': 0.0009811604730202816, 'recon_norm': 2.546119827032089, 'recon_norm_std': 0.07332851309329272, 'target_norm': 1.618327757716179, 'recon_cos': 0.34811695143580434, 'gradnorm/compressor.query': 0.009008144081618303, 'gradnorm/compressor.layer0': 0.1638524550134106, 'gradnorm/compressor.layer1': 0.2147180102600271, 'gradnorm/compressor.layer2': 0.1827671434917117, 'gradnorm/compressor.layer3': 0.16917599380924794, 'gradnorm/compressor.layer4': 0.15902764132328184, 'gradnorm/compressor.layer5': 0.14922765560437798, 'gradnorm/compressor.layer6': 0.14646107170394695, 'gradnorm/compressor.layer7': 0.13909849842128005, 'gradnorm/decoder.layer0': 0.1741488851211461, 'gradnorm/decoder.layer1': 0.17255436946618344, 'gradnorm/decoder.layer2': 0.24159780359080976, 'gradnorm/max_param': 0.17290998995304108, 'epoch': 1.08}

 11%|█         | 2060/19090 [1:00:09<7:32:04,  1.59s/it]
 11%|█         | 2061/19090 [1:00:10<7:06:28,  1.50s/it]
 11%|█         | 2062/19090 [1:00:11<6:50:35,  1.45s/it]
 11%|█         | 2063/19090 [1:00:14<8:02:41,  1.70s/it]
 11%|█         | 2064/19090 [1:00:15<7:27:34,  1.58s/it]
 11%|█         | 2065/19090 [1:00:16<7:18:12,  1.54s/it]
 11%|█         | 2066/19090 [1:00:18<8:07:52,  1.72s/it]
 11%|█         | 2067/19090 [1:00:20<7:31:32,  1.59s/it]
 11%|█         | 2068/19090 [1:00:22<7:51:45,  1.66s/it]
 11%|█         | 2069/19090 [1:00:24<8:22:36,  1.77s/it]
 11%|█         | 2070/19090 [1:00:25<7:42:04,  1.63s/it]
                                                        
{'loss': 0.005, 'grad_norm': 0.6788329482078552, 'learning_rate': 4.9989301525143096e-05, 'recon_loss': 0.005019687360618264, 'recon_static': 0.004042528523132205, 'recon_motion': 0.000977158961177338, 'recon_norm': 2.424771952629089, 'recon_norm_std': 0.07678338214755058, 'target_norm': 1.6179156243801116, 'recon_cos': 0.36622608453035355, 'gradnorm/compressor.query': 0.010590237548807141, 'gradnorm/compressor.layer0': 0.19581077911650838, 'gradnorm/compressor.layer1': 0.24954049528619532, 'gradnorm/compressor.layer2': 0.211419790416531, 'gradnorm/compressor.layer3': 0.19987744473065003, 'gradnorm/compressor.layer4': 0.1815925452475168, 'gradnorm/compressor.layer5': 0.1721455547003961, 'gradnorm/compressor.layer6': 0.16900900373372765, 'gradnorm/compressor.layer7': 0.16130208458765474, 'gradnorm/decoder.layer0': 0.20305259037523665, 'gradnorm/decoder.layer1': 0.19787591067012417, 'gradnorm/decoder.layer2': 0.2790390371730663, 'gradnorm/max_param': 0.1982564479112625, 'epoch': 1.08}

 11%|█         | 2070/19090 [1:00:25<7:42:04,  1.63s/it]
 11%|█         | 2071/19090 [1:00:27<8:20:37,  1.76s/it]
 11%|█         | 2072/19090 [1:00:29<8:37:32,  1.82s/it]
 11%|█         | 2073/19090 [1:00:30<7:51:32,  1.66s/it]
 11%|█         | 2074/19090 [1:00:32<8:00:25,  1.69s/it]
 11%|█         | 2075/19090 [1:00:34<9:11:32,  1.94s/it]
 11%|█         | 2076/19090 [1:00:36<8:15:16,  1.75s/it]
 11%|█         | 2077/19090 [1:00:38<9:37:32,  2.04s/it]
 11%|█         | 2078/19090 [1:00:40<8:33:40,  1.81s/it]
 11%|█         | 2079/19090 [1:00:41<7:49:03,  1.65s/it]
 11%|█         | 2080/19090 [1:00:42<7:24:43,  1.57s/it]
                                                        
{'loss': 0.0051, 'grad_norm': 0.5658428072929382, 'learning_rate': 4.998792253591892e-05, 'recon_loss': 0.0051246895571239294, 'recon_static': 0.004134388518286869, 'recon_motion': 0.0009903010723064654, 'recon_norm': 2.4449737071990967, 'recon_norm_std': 0.07544584963470698, 'target_norm': 1.6214729100465775, 'recon_cos': 0.36372680887579917, 'gradnorm/compressor.query': 0.008747167790818312, 'gradnorm/compressor.layer0': 0.1599260629870551, 'gradnorm/compressor.layer1': 0.20737120769233958, 'gradnorm/compressor.layer2': 0.17663526238387506, 'gradnorm/compressor.layer3': 0.16103173177674068, 'gradnorm/compressor.layer4': 0.15169908376456637, 'gradnorm/compressor.layer5': 0.14474406554808986, 'gradnorm/compressor.layer6': 0.14247856191152686, 'gradnorm/compressor.layer7': 0.13506211271411422, 'gradnorm/decoder.layer0': 0.1689874398072183, 'gradnorm/decoder.layer1': 0.16470759585860645, 'gradnorm/decoder.layer2': 0.23706637688646515, 'gradnorm/max_param': 0.16732379794120789, 'epoch': 1.09}

 11%|█         | 2080/19090 [1:00:42<7:24:43,  1.57s/it]
 11%|█         | 2081/19090 [1:00:45<8:44:13,  1.85s/it]
 11%|█         | 2082/19090 [1:00:46<7:56:40,  1.68s/it]
 11%|█         | 2083/19090 [1:00:48<7:29:33,  1.59s/it]
 11%|█         | 2084/19090 [1:00:50<8:14:14,  1.74s/it]
 11%|█         | 2085/19090 [1:00:51<7:35:42,  1.61s/it]
 11%|█         | 2086/19090 [1:00:53<8:11:52,  1.74s/it]
 11%|█         | 2087/19090 [1:00:55<8:44:30,  1.85s/it]
 11%|█         | 2088/19090 [1:00:56<7:56:24,  1.68s/it]
 11%|█         | 2089/19090 [1:00:58<8:10:44,  1.73s/it]
 11%|█         | 2090/19090 [1:01:00<8:12:36,  1.74s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.4205307960510254, 'learning_rate': 4.998645999930234e-05, 'recon_loss': 0.003483005438465625, 'recon_static': 0.002506297058425844, 'recon_motion': 0.0009767084047780372, 'recon_norm': 2.028249529004097, 'recon_norm_std': 0.08779294528067112, 'target_norm': 1.619832769036293, 'recon_cos': 0.43846984952688217, 'gradnorm/compressor.query': 0.006019781445720923, 'gradnorm/compressor.layer0': 0.11763284224308214, 'gradnorm/compressor.layer1': 0.1532089959128017, 'gradnorm/compressor.layer2': 0.13103874211871172, 'gradnorm/compressor.layer3': 0.12102718215585376, 'gradnorm/compressor.layer4': 0.11223966955489928, 'gradnorm/compressor.layer5': 0.1082162227690776, 'gradnorm/compressor.layer6': 0.10462641784431916, 'gradnorm/compressor.layer7': 0.09961157684136553, 'gradnorm/decoder.layer0': 0.1278236910996774, 'gradnorm/decoder.layer1': 0.1242779650590957, 'gradnorm/decoder.layer2': 0.17526874534255424, 'gradnorm/max_param': 0.12511363625526428, 'epoch': 1.09}

 11%|█         | 2090/19090 [1:01:00<8:12:36,  1.74s/it]
 11%|█         | 2091/19090 [1:01:01<7:38:30,  1.62s/it]
 11%|█         | 2092/19090 [1:01:03<8:15:52,  1.75s/it]
 11%|█         | 2093/19090 [1:01:05<8:01:54,  1.70s/it]
 11%|█         | 2094/19090 [1:01:06<7:35:02,  1.61s/it]
 11%|█         | 2095/19090 [1:01:09<8:41:45,  1.84s/it]
 11%|█         | 2096/19090 [1:01:11<8:47:27,  1.86s/it]
 11%|█         | 2097/19090 [1:01:12<7:58:33,  1.69s/it]
 11%|█         | 2098/19090 [1:01:14<8:29:12,  1.80s/it]
 11%|█         | 2099/19090 [1:01:16<8:19:43,  1.76s/it]
 11%|█         | 2100/19090 [1:01:17<7:39:44,  1.62s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.42278897762298584, 'learning_rate': 4.9984913920183375e-05, 'recon_loss': 0.0030962239543441683, 'recon_static': 0.0021108089829795064, 'recon_motion': 0.0009854149873717688, 'recon_norm': 1.9153349816799163, 'recon_norm_std': 0.09033709708601237, 'target_norm': 1.620646396279335, 'recon_cos': 0.4626803815364838, 'gradnorm/compressor.query': 0.006855889775112227, 'gradnorm/compressor.layer0': 0.12110572417338436, 'gradnorm/compressor.layer1': 0.15902098836303608, 'gradnorm/compressor.layer2': 0.13272223490609888, 'gradnorm/compressor.layer3': 0.12259489259009933, 'gradnorm/compressor.layer4': 0.11271350918635652, 'gradnorm/compressor.layer5': 0.10836006464804049, 'gradnorm/compressor.layer6': 0.10285693806191357, 'gradnorm/compressor.layer7': 0.09803168133496967, 'gradnorm/decoder.layer0': 0.1250876402586243, 'gradnorm/decoder.layer1': 0.12198967368265595, 'gradnorm/decoder.layer2': 0.1758981157283394, 'gradnorm/max_param': 0.12543267011642456, 'epoch': 1.1}

 11%|█         | 2100/19090 [1:01:17<7:39:44,  1.62s/it]
 11%|█         | 2101/19090 [1:01:19<8:26:14,  1.79s/it]
 11%|█         | 2102/19090 [1:01:21<8:23:13,  1.78s/it]
 11%|█         | 2103/19090 [1:01:22<7:41:27,  1.63s/it]
 11%|█         | 2104/19090 [1:01:24<8:35:32,  1.82s/it]
 11%|█         | 2105/19090 [1:01:26<8:29:36,  1.80s/it]
 11%|█         | 2106/19090 [1:01:28<7:55:09,  1.68s/it]
 11%|█         | 2107/19090 [1:01:30<8:47:17,  1.86s/it]
 11%|█         | 2108/19090 [1:01:31<8:19:45,  1.77s/it]
 11%|█         | 2109/19090 [1:01:34<8:57:05,  1.90s/it]
 11%|█         | 2110/19090 [1:01:35<8:46:06,  1.86s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 0.611661970615387, 'learning_rate': 4.998328430373135e-05, 'recon_loss': 0.0041223668784368785, 'recon_static': 0.003139597363770008, 'recon_motion': 0.0009827694826526566, 'recon_norm': 2.199930891394615, 'recon_norm_std': 0.0821217842400074, 'target_norm': 1.6205647885799408, 'recon_cos': 0.40549895018339155, 'gradnorm/compressor.query': 0.010259708256010103, 'gradnorm/compressor.layer0': 0.17594406746655486, 'gradnorm/compressor.layer1': 0.22921530070241689, 'gradnorm/compressor.layer2': 0.18904157554285708, 'gradnorm/compressor.layer3': 0.17754009653593952, 'gradnorm/compressor.layer4': 0.16430739449360685, 'gradnorm/compressor.layer5': 0.1555440691508541, 'gradnorm/compressor.layer6': 0.14942189115073246, 'gradnorm/compressor.layer7': 0.14439054031873205, 'gradnorm/decoder.layer0': 0.18270670894616967, 'gradnorm/decoder.layer1': 0.17643387231904661, 'gradnorm/decoder.layer2': 0.2537613969038058, 'gradnorm/max_param': 0.18069662153720856, 'epoch': 1.11}

 11%|█         | 2110/19090 [1:01:35<8:46:06,  1.86s/it]
 11%|█         | 2111/19090 [1:01:37<8:10:23,  1.73s/it]
 11%|█         | 2112/19090 [1:01:38<7:32:44,  1.60s/it]
 11%|█         | 2113/19090 [1:01:40<8:27:47,  1.79s/it]
 11%|█         | 2114/19090 [1:01:43<9:46:41,  2.07s/it]
 11%|█         | 2115/19090 [1:01:44<8:39:44,  1.84s/it]
 11%|█         | 2116/19090 [1:01:46<8:10:49,  1.73s/it]
 11%|█         | 2117/19090 [1:01:48<8:05:23,  1.72s/it]
 11%|█         | 2118/19090 [1:01:49<7:29:07,  1.59s/it]
 11%|█         | 2119/19090 [1:01:51<8:30:01,  1.80s/it]
 11%|█         | 2120/19090 [1:01:53<8:37:24,  1.83s/it]
                                                        
{'loss': 0.0054, 'grad_norm': 0.5473328828811646, 'learning_rate': 4.99815711553949e-05, 'recon_loss': 0.0053601314662955705, 'recon_static': 0.004391975386533886, 'recon_motion': 0.0009681561961770058, 'recon_norm': 2.5040409684181215, 'recon_norm_std': 0.07594071896746754, 'target_norm': 1.6196491509675979, 'recon_cos': 0.35764521434903146, 'gradnorm/compressor.query': 0.00926631229278084, 'gradnorm/compressor.layer0': 0.16063416514242979, 'gradnorm/compressor.layer1': 0.2023995631799159, 'gradnorm/compressor.layer2': 0.17048629655577585, 'gradnorm/compressor.layer3': 0.15876993437134074, 'gradnorm/compressor.layer4': 0.14576265560964907, 'gradnorm/compressor.layer5': 0.13915420653356744, 'gradnorm/compressor.layer6': 0.13495212586891975, 'gradnorm/compressor.layer7': 0.1293089673706111, 'gradnorm/decoder.layer0': 0.16193501160939178, 'gradnorm/decoder.layer1': 0.15779729288311975, 'gradnorm/decoder.layer2': 0.22754909018494124, 'gradnorm/max_param': 0.16342931985855103, 'epoch': 1.11}

 11%|█         | 2120/19090 [1:01:53<8:37:24,  1.83s/it]
 11%|█         | 2121/19090 [1:01:54<7:51:20,  1.67s/it]
 11%|█         | 2122/19090 [1:01:57<8:37:14,  1.83s/it]
 11%|█         | 2123/19090 [1:01:58<8:44:47,  1.86s/it]
 11%|█         | 2124/19090 [1:02:00<7:56:41,  1.69s/it]
 11%|█         | 2125/19090 [1:02:02<9:00:21,  1.91s/it]
 11%|█         | 2126/19090 [1:02:04<8:09:39,  1.73s/it]
 11%|█         | 2127/19090 [1:02:05<7:31:42,  1.60s/it]
 11%|█         | 2128/19090 [1:02:07<8:54:04,  1.89s/it]
 11%|█         | 2129/19090 [1:02:10<9:30:04,  2.02s/it]
 11%|█         | 2130/19090 [1:02:11<8:28:01,  1.80s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.4480099678039551, 'learning_rate': 4.9979774480901974e-05, 'recon_loss': 0.003664135152939707, 'recon_static': 0.0026810459268745033, 'recon_motion': 0.0009830891751335003, 'recon_norm': 2.0790109425783156, 'recon_norm_std': 0.08582813367247581, 'target_norm': 1.6210096538066865, 'recon_cos': 0.4274984411895275, 'gradnorm/compressor.query': 0.007202524240032007, 'gradnorm/compressor.layer0': 0.1273399109642015, 'gradnorm/compressor.layer1': 0.1658850773658656, 'gradnorm/compressor.layer2': 0.13971253237992276, 'gradnorm/compressor.layer3': 0.13046839651185535, 'gradnorm/compressor.layer4': 0.1202274723376745, 'gradnorm/compressor.layer5': 0.11531869179686602, 'gradnorm/compressor.layer6': 0.11062532411409301, 'gradnorm/compressor.layer7': 0.1052000765070551, 'gradnorm/decoder.layer0': 0.13141037106713993, 'gradnorm/decoder.layer1': 0.12937264887198363, 'gradnorm/decoder.layer2': 0.1879417560026984, 'gradnorm/max_param': 0.13474543392658234, 'epoch': 1.12}

 11%|█         | 2130/19090 [1:02:11<8:28:01,  1.80s/it]
 11%|█         | 2131/19090 [1:02:13<8:58:45,  1.91s/it]
 11%|█         | 2132/19090 [1:02:14<8:06:08,  1.72s/it]
 11%|█         | 2133/19090 [1:02:16<7:29:43,  1.59s/it]
 11%|█         | 2134/19090 [1:02:18<9:04:44,  1.93s/it]
 11%|█         | 2135/19090 [1:02:20<8:10:20,  1.74s/it]
 11%|█         | 2136/19090 [1:02:21<7:36:37,  1.62s/it]
 11%|█         | 2137/19090 [1:02:24<9:00:59,  1.91s/it]
 11%|█         | 2138/19090 [1:02:25<8:22:56,  1.78s/it]
 11%|█         | 2139/19090 [1:02:26<7:41:25,  1.63s/it]
 11%|█         | 2140/19090 [1:02:29<8:46:47,  1.86s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.45856329798698425, 'learning_rate': 4.997789428625975e-05, 'recon_loss': 0.0031948366900905966, 'recon_static': 0.002219262073049322, 'recon_motion': 0.0009755746126757003, 'recon_norm': 1.9449996322393417, 'recon_norm_std': 0.09131186064332723, 'target_norm': 1.6199517667293548, 'recon_cos': 0.4556849770247936, 'gradnorm/compressor.query': 0.007704399019883626, 'gradnorm/compressor.layer0': 0.13372226077608465, 'gradnorm/compressor.layer1': 0.17266069531651634, 'gradnorm/compressor.layer2': 0.14394157849959174, 'gradnorm/compressor.layer3': 0.1336931385961832, 'gradnorm/compressor.layer4': 0.1223785400010116, 'gradnorm/compressor.layer5': 0.11662383417945134, 'gradnorm/compressor.layer6': 0.11262274360910297, 'gradnorm/compressor.layer7': 0.10688779886750653, 'gradnorm/decoder.layer0': 0.13426510841805633, 'gradnorm/decoder.layer1': 0.13111576896817959, 'gradnorm/decoder.layer2': 0.18981462577054017, 'gradnorm/max_param': 0.13632556796073914, 'epoch': 1.12}

 11%|█         | 2140/19090 [1:02:29<8:46:47,  1.86s/it]
 11%|█         | 2141/19090 [1:02:30<8:27:39,  1.80s/it]
 11%|█         | 2142/19090 [1:02:32<7:44:10,  1.64s/it]
 11%|█         | 2143/19090 [1:02:35<9:27:03,  2.01s/it]
 11%|█         | 2144/19090 [1:02:36<8:27:31,  1.80s/it]
 11%|█         | 2145/19090 [1:02:37<7:44:27,  1.64s/it]
 11%|█         | 2146/19090 [1:02:39<8:38:02,  1.83s/it]
 11%|█         | 2147/19090 [1:02:41<8:14:29,  1.75s/it]
 11%|█▏        | 2148/19090 [1:02:42<7:35:03,  1.61s/it]
 11%|█▏        | 2149/19090 [1:02:45<9:17:00,  1.97s/it]
 11%|█▏        | 2150/19090 [1:02:46<8:18:56,  1.77s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.5914032459259033, 'learning_rate': 4.99759305777547e-05, 'recon_loss': 0.003934979648329317, 'recon_static': 0.0029591195518150924, 'recon_motion': 0.0009758601008797996, 'recon_norm': 2.1567612409591677, 'recon_norm_std': 0.08298271093517542, 'target_norm': 1.6207191556692124, 'recon_cos': 0.4142161875963211, 'gradnorm/compressor.query': 0.009144632195217425, 'gradnorm/compressor.layer0': 0.16435595788276586, 'gradnorm/compressor.layer1': 0.21514606013993617, 'gradnorm/compressor.layer2': 0.1854580013853108, 'gradnorm/compressor.layer3': 0.1709051457859596, 'gradnorm/compressor.layer4': 0.15857049297001674, 'gradnorm/compressor.layer5': 0.15128209583366728, 'gradnorm/compressor.layer6': 0.14551823853896545, 'gradnorm/compressor.layer7': 0.14083954551452538, 'gradnorm/decoder.layer0': 0.17918546660247167, 'gradnorm/decoder.layer1': 0.1737403219258288, 'gradnorm/decoder.layer2': 0.24786064335726327, 'gradnorm/max_param': 0.17443856596946716, 'epoch': 1.13}

 11%|█▏        | 2150/19090 [1:02:46<8:18:56,  1.77s/it]
 11%|█▏        | 2151/19090 [1:02:48<7:38:05,  1.62s/it]
 11%|█▏        | 2152/19090 [1:02:50<9:14:37,  1.96s/it]
 11%|█▏        | 2153/19090 [1:02:52<8:26:19,  1.79s/it]
 11%|█▏        | 2154/19090 [1:02:53<7:47:39,  1.66s/it]
 11%|█▏        | 2155/19090 [1:02:56<9:15:37,  1.97s/it]
 11%|█▏        | 2156/19090 [1:02:57<8:17:36,  1.76s/it]
 11%|█▏        | 2157/19090 [1:02:58<7:37:20,  1.62s/it]
 11%|█▏        | 2158/19090 [1:03:01<9:30:14,  2.02s/it]
 11%|█▏        | 2159/19090 [1:03:03<8:27:49,  1.80s/it]
 11%|█▏        | 2160/19090 [1:03:04<7:44:18,  1.65s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.561617374420166, 'learning_rate': 4.997388336195249e-05, 'recon_loss': 0.004622764652594924, 'recon_static': 0.003630467632319778, 'recon_motion': 0.0009922970915795305, 'recon_norm': 2.3276865482330322, 'recon_norm_std': 0.07772169411182403, 'target_norm': 1.6220266848802567, 'recon_cos': 0.3813182607293129, 'gradnorm/compressor.query': 0.009160518733113902, 'gradnorm/compressor.layer0': 0.15829670457809142, 'gradnorm/compressor.layer1': 0.2056607643725191, 'gradnorm/compressor.layer2': 0.17643769397887088, 'gradnorm/compressor.layer3': 0.161164910558468, 'gradnorm/compressor.layer4': 0.15280908650016242, 'gradnorm/compressor.layer5': 0.14214087418711227, 'gradnorm/compressor.layer6': 0.14045804762672118, 'gradnorm/compressor.layer7': 0.1346191633167047, 'gradnorm/decoder.layer0': 0.16650627783375635, 'gradnorm/decoder.layer1': 0.16313818705629446, 'gradnorm/decoder.layer2': 0.23479297243808434, 'gradnorm/max_param': 0.16781429946422577, 'epoch': 1.13}

 11%|█▏        | 2160/19090 [1:03:04<7:44:18,  1.65s/it]
 11%|█▏        | 2161/19090 [1:03:06<8:54:12,  1.89s/it]
 11%|█▏        | 2162/19090 [1:03:08<8:04:24,  1.72s/it]
 11%|█▏        | 2163/19090 [1:03:09<7:28:07,  1.59s/it]
 11%|█▏        | 2164/19090 [1:03:12<8:51:24,  1.88s/it]
 11%|█▏        | 2165/19090 [1:03:13<8:06:52,  1.73s/it]
 11%|█▏        | 2166/19090 [1:03:14<7:45:03,  1.65s/it]
 11%|█▏        | 2167/19090 [1:03:17<8:41:23,  1.85s/it]
 11%|█▏        | 2168/19090 [1:03:18<8:30:44,  1.81s/it]
 11%|█▏        | 2169/19090 [1:03:20<7:47:54,  1.66s/it]
 11%|█▏        | 2170/19090 [1:03:22<8:38:52,  1.84s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.4772559404373169, 'learning_rate': 4.997175264569801e-05, 'recon_loss': 0.004244048398686573, 'recon_static': 0.003264832525746897, 'recon_motion': 0.000979215763800312, 'recon_norm': 2.233056050539017, 'recon_norm_std': 0.08311065286397934, 'target_norm': 1.6224436730146408, 'recon_cos': 0.3989832177758217, 'gradnorm/compressor.query': 0.007065386007911442, 'gradnorm/compressor.layer0': 0.12977856202376564, 'gradnorm/compressor.layer1': 0.17619573152487467, 'gradnorm/compressor.layer2': 0.1500890668547046, 'gradnorm/compressor.layer3': 0.13834374518832845, 'gradnorm/compressor.layer4': 0.12840141073602065, 'gradnorm/compressor.layer5': 0.12181175314790273, 'gradnorm/compressor.layer6': 0.11830052858035656, 'gradnorm/compressor.layer7': 0.11275502594696316, 'gradnorm/decoder.layer0': 0.14295995872859193, 'gradnorm/decoder.layer1': 0.13905278798810974, 'gradnorm/decoder.layer2': 0.2009059018108171, 'gradnorm/max_param': 0.14322061836719513, 'epoch': 1.14}

 11%|█▏        | 2170/19090 [1:03:22<8:38:52,  1.84s/it]
 11%|█▏        | 2171/19090 [1:03:24<8:58:56,  1.91s/it]
 11%|█▏        | 2172/19090 [1:03:25<8:06:04,  1.72s/it]
 11%|█▏        | 2173/19090 [1:03:27<8:27:57,  1.80s/it]
 11%|█▏        | 2174/19090 [1:03:29<8:47:09,  1.87s/it]
 11%|█▏        | 2175/19090 [1:03:31<7:58:01,  1.70s/it]
 11%|█▏        | 2176/19090 [1:03:33<8:26:14,  1.80s/it]
 11%|█▏        | 2177/19090 [1:03:35<9:27:26,  2.01s/it]
 11%|█▏        | 2178/19090 [1:03:37<8:27:36,  1.80s/it]
 11%|█▏        | 2179/19090 [1:03:38<8:01:38,  1.71s/it]
 11%|█▏        | 2180/19090 [1:03:41<9:24:32,  2.00s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.4451137185096741, 'learning_rate': 4.996953843611535e-05, 'recon_loss': 0.0038654149859212337, 'recon_static': 0.002887647057650611, 'recon_motion': 0.0009777678758837282, 'recon_norm': 2.135043501853943, 'recon_norm_std': 0.08465845324099064, 'target_norm': 1.62175615131855, 'recon_cos': 0.4169190667569637, 'gradnorm/compressor.query': 0.006750145076956918, 'gradnorm/compressor.layer0': 0.12353287570341051, 'gradnorm/compressor.layer1': 0.16680467685615277, 'gradnorm/compressor.layer2': 0.13999751949880201, 'gradnorm/compressor.layer3': 0.12933209910410137, 'gradnorm/compressor.layer4': 0.12078525872135353, 'gradnorm/compressor.layer5': 0.11357950541032244, 'gradnorm/compressor.layer6': 0.10924519772581066, 'gradnorm/compressor.layer7': 0.10452551176660511, 'gradnorm/decoder.layer0': 0.13221241919017426, 'gradnorm/decoder.layer1': 0.1271450938826728, 'gradnorm/decoder.layer2': 0.1862120200642446, 'gradnorm/max_param': 0.1335749328136444, 'epoch': 1.14}

 11%|█▏        | 2180/19090 [1:03:41<9:24:32,  2.00s/it]
 11%|█▏        | 2181/19090 [1:03:42<8:24:52,  1.79s/it]
 11%|█▏        | 2182/19090 [1:03:43<7:42:32,  1.64s/it]
 11%|█▏        | 2183/19090 [1:03:46<9:17:13,  1.98s/it]
 11%|█▏        | 2184/19090 [1:03:47<8:19:11,  1.77s/it]
 11%|█▏        | 2185/19090 [1:03:49<7:38:51,  1.63s/it]
 11%|█▏        | 2186/19090 [1:03:51<9:04:12,  1.93s/it]
 11%|█▏        | 2187/19090 [1:03:53<8:11:09,  1.74s/it]
 11%|█▏        | 2188/19090 [1:03:54<7:32:57,  1.61s/it]
 11%|█▏        | 2189/19090 [1:03:57<9:27:21,  2.01s/it]
 11%|█▏        | 2190/19090 [1:03:58<8:26:55,  1.80s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.5402565002441406, 'learning_rate': 4.9967240740607724e-05, 'recon_loss': 0.0033173287811223418, 'recon_static': 0.002337090065702796, 'recon_motion': 0.0009802386732189917, 'recon_norm': 1.982008957862854, 'recon_norm_std': 0.08777173236012459, 'target_norm': 1.620263585448265, 'recon_cos': 0.4504100620746613, 'gradnorm/compressor.query': 0.008102531486176425, 'gradnorm/compressor.layer0': 0.15432758267168223, 'gradnorm/compressor.layer1': 0.202778529906257, 'gradnorm/compressor.layer2': 0.17084545949217053, 'gradnorm/compressor.layer3': 0.1574886400519717, 'gradnorm/compressor.layer4': 0.14577444340719498, 'gradnorm/compressor.layer5': 0.13712305506089334, 'gradnorm/compressor.layer6': 0.13154735770263265, 'gradnorm/compressor.layer7': 0.12740974471440977, 'gradnorm/decoder.layer0': 0.16007675547314157, 'gradnorm/decoder.layer1': 0.15525559856698434, 'gradnorm/decoder.layer2': 0.22263136942199963, 'gradnorm/max_param': 0.16024677455425262, 'epoch': 1.15}

 11%|█▏        | 2190/19090 [1:03:58<8:26:55,  1.80s/it]
 11%|█▏        | 2191/19090 [1:04:00<7:46:11,  1.66s/it]
 11%|█▏        | 2192/19090 [1:04:02<9:06:12,  1.94s/it]
 11%|█▏        | 2193/19090 [1:04:03<8:11:30,  1.75s/it]
 11%|█▏        | 2194/19090 [1:04:05<7:32:54,  1.61s/it]
 11%|█▏        | 2195/19090 [1:04:07<9:13:10,  1.96s/it]
 12%|█▏        | 2196/19090 [1:04:09<8:16:28,  1.76s/it]
 12%|█▏        | 2197/19090 [1:04:10<7:36:41,  1.62s/it]
 12%|█▏        | 2198/19090 [1:04:13<9:05:25,  1.94s/it]
 12%|█▏        | 2199/19090 [1:04:14<8:10:37,  1.74s/it]
 12%|█▏        | 2200/19090 [1:04:15<7:32:20,  1.61s/it]
                                                        
{'loss': 0.0052, 'grad_norm': 0.6970215439796448, 'learning_rate': 4.99648595668575e-05, 'recon_loss': 0.00520926509052515, 'recon_static': 0.004228991363197565, 'recon_motion': 0.0009802738422877155, 'recon_norm': 2.468387168645859, 'recon_norm_std': 0.07694283835589885, 'target_norm': 1.6235406816005706, 'recon_cos': 0.359283934533596, 'gradnorm/compressor.query': 0.010439531889257917, 'gradnorm/compressor.layer0': 0.19928833737212082, 'gradnorm/compressor.layer1': 0.258322024418321, 'gradnorm/compressor.layer2': 0.21920358939845278, 'gradnorm/compressor.layer3': 0.2030627540435329, 'gradnorm/compressor.layer4': 0.1894956717940475, 'gradnorm/compressor.layer5': 0.1759177260949363, 'gradnorm/compressor.layer6': 0.16834099131537247, 'gradnorm/compressor.layer7': 0.1625441786885187, 'gradnorm/decoder.layer0': 0.20555830552184753, 'gradnorm/decoder.layer1': 0.20243698755678174, 'gradnorm/decoder.layer2': 0.29180368826596176, 'gradnorm/max_param': 0.2085346132516861, 'epoch': 1.15}

 12%|█▏        | 2200/19090 [1:04:15<7:32:20,  1.61s/it]
 12%|█▏        | 2201/19090 [1:04:18<9:10:45,  1.96s/it]
 12%|█▏        | 2202/19090 [1:04:19<8:14:15,  1.76s/it]
 12%|█▏        | 2203/19090 [1:04:21<7:34:38,  1.62s/it]
 12%|█▏        | 2204/19090 [1:04:24<9:41:25,  2.07s/it]
 12%|█▏        | 2205/19090 [1:04:25<8:35:48,  1.83s/it]
 12%|█▏        | 2206/19090 [1:04:26<7:50:32,  1.67s/it]
 12%|█▏        | 2207/19090 [1:04:29<9:12:51,  1.96s/it]
 12%|█▏        | 2208/19090 [1:04:30<8:15:49,  1.76s/it]
 12%|█▏        | 2209/19090 [1:04:32<7:35:52,  1.62s/it]
 12%|█▏        | 2210/19090 [1:04:34<9:17:55,  1.98s/it]
                                                        
{'loss': 0.0048, 'grad_norm': 0.4789436459541321, 'learning_rate': 4.996239492282618e-05, 'recon_loss': 0.004770437750266865, 'recon_static': 0.0037927259341813623, 'recon_motion': 0.000977711920859292, 'recon_norm': 2.362161415815353, 'recon_norm_std': 0.0774860461242497, 'target_norm': 1.6213555932044983, 'recon_cos': 0.37955913692712784, 'gradnorm/compressor.query': 0.0074901483670242845, 'gradnorm/compressor.layer0': 0.1339761650862114, 'gradnorm/compressor.layer1': 0.17586507917632263, 'gradnorm/compressor.layer2': 0.14951108761684728, 'gradnorm/compressor.layer3': 0.13722925922448975, 'gradnorm/compressor.layer4': 0.12769612255755539, 'gradnorm/compressor.layer5': 0.12019197337856015, 'gradnorm/compressor.layer6': 0.11793215907319274, 'gradnorm/compressor.layer7': 0.1137435580554292, 'gradnorm/decoder.layer0': 0.14396747615168679, 'gradnorm/decoder.layer1': 0.14011443356353845, 'gradnorm/decoder.layer2': 0.2032502503475651, 'gradnorm/max_param': 0.14468544721603394, 'epoch': 1.16}

 12%|█▏        | 2210/19090 [1:04:34<9:17:55,  1.98s/it]
 12%|█▏        | 2211/19090 [1:04:36<8:19:16,  1.77s/it]
 12%|█▏        | 2212/19090 [1:04:37<7:37:56,  1.63s/it]
 12%|█▏        | 2213/19090 [1:04:40<9:03:00,  1.93s/it]
 12%|█▏        | 2214/19090 [1:04:41<8:09:46,  1.74s/it]
 12%|█▏        | 2215/19090 [1:04:42<7:31:38,  1.61s/it]
 12%|█▏        | 2216/19090 [1:04:45<8:41:04,  1.85s/it]
 12%|█▏        | 2217/19090 [1:04:46<7:53:52,  1.69s/it]
 12%|█▏        | 2218/19090 [1:04:47<7:20:56,  1.57s/it]
 12%|█▏        | 2219/19090 [1:04:50<8:43:53,  1.86s/it]
 12%|█▏        | 2220/19090 [1:04:51<7:56:14,  1.69s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.35441887378692627, 'learning_rate': 4.9959846816754306e-05, 'recon_loss': 0.0030099971976596863, 'recon_static': 0.0020315583155024798, 'recon_motion': 0.000978438902529888, 'recon_norm': 1.8899072617292405, 'recon_norm_std': 0.09042485170066357, 'target_norm': 1.6214294582605362, 'recon_cos': 0.4716122753918171, 'gradnorm/compressor.query': 0.005284707037740242, 'gradnorm/compressor.layer0': 0.09782488567799907, 'gradnorm/compressor.layer1': 0.12755943182414545, 'gradnorm/compressor.layer2': 0.11033295584257646, 'gradnorm/compressor.layer3': 0.10148181828368327, 'gradnorm/compressor.layer4': 0.09663416050635272, 'gradnorm/compressor.layer5': 0.09131597805253241, 'gradnorm/compressor.layer6': 0.08776685954487655, 'gradnorm/compressor.layer7': 0.08437001602146475, 'gradnorm/decoder.layer0': 0.10663664971093005, 'gradnorm/decoder.layer1': 0.10450613517011191, 'gradnorm/decoder.layer2': 0.14993591429717434, 'gradnorm/max_param': 0.1081906408071518, 'epoch': 1.16}

 12%|█▏        | 2220/19090 [1:04:51<7:56:14,  1.69s/it]
 12%|█▏        | 2221/19090 [1:04:52<7:23:10,  1.58s/it]
 12%|█▏        | 2222/19090 [1:04:55<8:22:32,  1.79s/it]
 12%|█▏        | 2223/19090 [1:04:56<7:41:23,  1.64s/it]
 12%|█▏        | 2224/19090 [1:04:57<7:12:05,  1.54s/it]
 12%|█▏        | 2225/19090 [1:04:59<7:38:42,  1.63s/it]
 12%|█▏        | 2226/19090 [1:05:00<7:09:54,  1.53s/it]
 12%|█▏        | 2227/19090 [1:05:02<6:49:25,  1.46s/it]
 12%|█▏        | 2228/19090 [1:05:04<7:30:03,  1.60s/it]
 12%|█▏        | 2229/19090 [1:05:05<7:05:10,  1.51s/it]
 12%|█▏        | 2230/19090 [1:05:06<6:46:42,  1.45s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.4898272752761841, 'learning_rate': 4.995721525716152e-05, 'recon_loss': 0.0032174921419937164, 'recon_static': 0.0022402535629225894, 'recon_motion': 0.0009772385674295947, 'recon_norm': 1.9512549877166747, 'recon_norm_std': 0.08889936059713363, 'target_norm': 1.6211127549409867, 'recon_cos': 0.4568890169262886, 'gradnorm/compressor.query': 0.0076715658074817845, 'gradnorm/compressor.layer0': 0.13612750137454924, 'gradnorm/compressor.layer1': 0.18183435343084142, 'gradnorm/compressor.layer2': 0.1550134245188868, 'gradnorm/compressor.layer3': 0.14427255986370674, 'gradnorm/compressor.layer4': 0.13283768583359934, 'gradnorm/compressor.layer5': 0.12413545866047342, 'gradnorm/compressor.layer6': 0.12056799163468132, 'gradnorm/compressor.layer7': 0.11567643204121661, 'gradnorm/decoder.layer0': 0.145124904483859, 'gradnorm/decoder.layer1': 0.140628960341742, 'gradnorm/decoder.layer2': 0.20401673630415176, 'gradnorm/max_param': 0.14688429236412048, 'epoch': 1.17}

 12%|█▏        | 2230/19090 [1:05:06<6:46:42,  1.45s/it]
 12%|█▏        | 2231/19090 [1:05:08<7:36:49,  1.63s/it]
 12%|█▏        | 2232/19090 [1:05:10<7:08:36,  1.53s/it]
 12%|█▏        | 2233/19090 [1:05:11<6:48:55,  1.46s/it]
 12%|█▏        | 2234/19090 [1:05:13<7:54:38,  1.69s/it]
 12%|█▏        | 2235/19090 [1:05:14<7:20:35,  1.57s/it]
 12%|█▏        | 2236/19090 [1:05:16<6:56:49,  1.48s/it]
 12%|█▏        | 2237/19090 [1:05:18<7:28:09,  1.60s/it]
 12%|█▏        | 2238/19090 [1:05:19<7:02:07,  1.50s/it]
 12%|█▏        | 2239/19090 [1:05:20<6:43:50,  1.44s/it]
 12%|█▏        | 2240/19090 [1:05:22<7:37:54,  1.63s/it]
                                                        
{'loss': 0.004, 'grad_norm': 0.5079662799835205, 'learning_rate': 4.995450025284645e-05, 'recon_loss': 0.00396266405005008, 'recon_static': 0.0029848216974642126, 'recon_motion': 0.0009778424006071872, 'recon_norm': 2.1624291837215424, 'recon_norm_std': 0.08284850232303143, 'target_norm': 1.6224990040063858, 'recon_cos': 0.411693187057972, 'gradnorm/compressor.query': 0.00859861280345809, 'gradnorm/compressor.layer0': 0.14589348826765472, 'gradnorm/compressor.layer1': 0.18868848518248327, 'gradnorm/compressor.layer2': 0.15943461172836662, 'gradnorm/compressor.layer3': 0.14636065026398928, 'gradnorm/compressor.layer4': 0.13680237528637818, 'gradnorm/compressor.layer5': 0.1289374357127197, 'gradnorm/compressor.layer6': 0.12488049143739367, 'gradnorm/compressor.layer7': 0.11929596092433518, 'gradnorm/decoder.layer0': 0.15029543799639278, 'gradnorm/decoder.layer1': 0.14664477149901223, 'gradnorm/decoder.layer2': 0.21205398229688263, 'gradnorm/max_param': 0.15313395857810974, 'epoch': 1.17}

 12%|█▏        | 2240/19090 [1:05:22<7:37:54,  1.63s/it]
 12%|█▏        | 2241/19090 [1:05:23<7:09:07,  1.53s/it]
 12%|█▏        | 2242/19090 [1:05:25<6:49:11,  1.46s/it]
 12%|█▏        | 2243/19090 [1:05:27<7:56:01,  1.70s/it]
 12%|█▏        | 2244/19090 [1:05:28<7:21:38,  1.57s/it]
 12%|█▏        | 2245/19090 [1:05:30<6:58:21,  1.49s/it]
 12%|█▏        | 2246/19090 [1:05:32<7:37:57,  1.63s/it]
 12%|█▏        | 2247/19090 [1:05:33<7:09:44,  1.53s/it]
 12%|█▏        | 2248/19090 [1:05:34<6:51:05,  1.46s/it]
 12%|█▏        | 2249/19090 [1:05:36<7:57:44,  1.70s/it]
 12%|█▏        | 2250/19090 [1:05:38<7:23:44,  1.58s/it]
                                                        
{'loss': 0.004, 'grad_norm': 0.4660396873950958, 'learning_rate': 4.995170181288675e-05, 'recon_loss': 0.003997894568601623, 'recon_static': 0.0030250809330027552, 'recon_motion': 0.0009728136370540597, 'recon_norm': 2.175213432312012, 'recon_norm_std': 0.08403697423636913, 'target_norm': 1.6219899952411652, 'recon_cos': 0.41183682903647423, 'gradnorm/compressor.query': 0.007445047521272748, 'gradnorm/compressor.layer0': 0.13182053652628858, 'gradnorm/compressor.layer1': 0.17328909010044438, 'gradnorm/compressor.layer2': 0.14593121672343246, 'gradnorm/compressor.layer3': 0.13531138304013715, 'gradnorm/compressor.layer4': 0.12454936636434699, 'gradnorm/compressor.layer5': 0.11825147264669497, 'gradnorm/compressor.layer6': 0.113744974555888, 'gradnorm/compressor.layer7': 0.1107360589217081, 'gradnorm/decoder.layer0': 0.13881419494586864, 'gradnorm/decoder.layer1': 0.13497753344901597, 'gradnorm/decoder.layer2': 0.19478714860344537, 'gradnorm/max_param': 0.14040859043598175, 'epoch': 1.18}

 12%|█▏        | 2250/19090 [1:05:38<7:23:44,  1.58s/it]
 12%|█▏        | 2251/19090 [1:05:39<7:00:26,  1.50s/it]
 12%|█▏        | 2252/19090 [1:05:41<7:25:23,  1.59s/it]
 12%|█▏        | 2253/19090 [1:05:42<7:00:58,  1.50s/it]
 12%|█▏        | 2254/19090 [1:05:43<6:44:27,  1.44s/it]
 12%|█▏        | 2255/19090 [1:05:46<7:37:50,  1.63s/it]
 12%|█▏        | 2256/19090 [1:05:47<7:10:10,  1.53s/it]
 12%|█▏        | 2257/19090 [1:05:48<6:50:08,  1.46s/it]
 12%|█▏        | 2258/19090 [1:05:50<7:30:23,  1.61s/it]
 12%|█▏        | 2259/19090 [1:05:51<7:04:09,  1.51s/it]
 12%|█▏        | 2260/19090 [1:05:53<6:45:32,  1.45s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 0.49196892976760864, 'learning_rate': 4.994881994663904e-05, 'recon_loss': 0.004126183356856927, 'recon_static': 0.0031583835429046304, 'recon_motion': 0.0009677998707047664, 'recon_norm': 2.2103094816207887, 'recon_norm_std': 0.08029820676892996, 'target_norm': 1.6181032836437226, 'recon_cos': 0.40531815886497496, 'gradnorm/compressor.query': 0.007623314819268232, 'gradnorm/compressor.layer0': 0.14090742784229074, 'gradnorm/compressor.layer1': 0.18228313675859922, 'gradnorm/compressor.layer2': 0.15472121978989012, 'gradnorm/compressor.layer3': 0.14269200441006974, 'gradnorm/compressor.layer4': 0.13343848636807593, 'gradnorm/compressor.layer5': 0.12268184832100981, 'gradnorm/compressor.layer6': 0.1197062300862051, 'gradnorm/compressor.layer7': 0.11651066187339812, 'gradnorm/decoder.layer0': 0.1468446560722335, 'gradnorm/decoder.layer1': 0.14238212778897474, 'gradnorm/decoder.layer2': 0.20493076116133682, 'gradnorm/max_param': 0.1484198123216629, 'epoch': 1.18}

 12%|█▏        | 2260/19090 [1:05:53<6:45:32,  1.45s/it]
 12%|█▏        | 2261/19090 [1:05:55<7:41:00,  1.64s/it]
 12%|█▏        | 2262/19090 [1:05:56<7:11:33,  1.54s/it]
 12%|█▏        | 2263/19090 [1:05:57<6:50:27,  1.46s/it]
 12%|█▏        | 2264/19090 [1:06:00<8:08:53,  1.74s/it]
 12%|█▏        | 2265/19090 [1:06:01<7:30:37,  1.61s/it]
 12%|█▏        | 2266/19090 [1:06:02<7:04:05,  1.51s/it]
 12%|█▏        | 2267/19090 [1:06:05<8:51:17,  1.89s/it]
 12%|█▏        | 2268/19090 [1:06:06<7:59:50,  1.71s/it]
 12%|█▏        | 2269/19090 [1:06:08<7:24:06,  1.58s/it]
 12%|█▏        | 2270/19090 [1:06:11<9:11:30,  1.97s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.4526140093803406, 'learning_rate': 4.9945854663738866e-05, 'recon_loss': 0.004524723551003263, 'recon_static': 0.0035544421465601773, 'recon_motion': 0.0009702814073534683, 'recon_norm': 2.3058824062347414, 'recon_norm_std': 0.07987476419657469, 'target_norm': 1.621737429499626, 'recon_cos': 0.38832602351903917, 'gradnorm/compressor.query': 0.00622942078278198, 'gradnorm/compressor.layer0': 0.12179810883054297, 'gradnorm/compressor.layer1': 0.16477294210155377, 'gradnorm/compressor.layer2': 0.13905756330835659, 'gradnorm/compressor.layer3': 0.13175029798805155, 'gradnorm/compressor.layer4': 0.12129138769233781, 'gradnorm/compressor.layer5': 0.11519169825371645, 'gradnorm/compressor.layer6': 0.11094657867517137, 'gradnorm/compressor.layer7': 0.10683385561042608, 'gradnorm/decoder.layer0': 0.1366394221545713, 'gradnorm/decoder.layer1': 0.13432234270190876, 'gradnorm/decoder.layer2': 0.1942276772912493, 'gradnorm/max_param': 0.13867664337158203, 'epoch': 1.19}

 12%|█▏        | 2270/19090 [1:06:11<9:11:30,  1.97s/it]
 12%|█▏        | 2271/19090 [1:06:12<8:14:17,  1.76s/it]
 12%|█▏        | 2272/19090 [1:06:13<7:34:11,  1.62s/it]
 12%|█▏        | 2273/19090 [1:06:16<9:19:21,  2.00s/it]
 12%|█▏        | 2274/19090 [1:06:17<8:20:00,  1.78s/it]
 12%|█▏        | 2275/19090 [1:06:19<7:38:02,  1.63s/it]
 12%|█▏        | 2276/19090 [1:06:21<8:42:33,  1.86s/it]
 12%|█▏        | 2277/19090 [1:06:22<7:53:52,  1.69s/it]
 12%|█▏        | 2278/19090 [1:06:24<7:19:45,  1.57s/it]
 12%|█▏        | 2279/19090 [1:06:27<9:19:49,  2.00s/it]
 12%|█▏        | 2280/19090 [1:06:28<8:20:11,  1.79s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.3544645607471466, 'learning_rate': 4.994280597410068e-05, 'recon_loss': 0.002979434939334169, 'recon_static': 0.0020104526716750114, 'recon_motion': 0.0009689821425126865, 'recon_norm': 1.8868094831705093, 'recon_norm_std': 0.09207503739744424, 'target_norm': 1.6206835865974427, 'recon_cos': 0.4742085367441177, 'gradnorm/compressor.query': 0.005417608157101887, 'gradnorm/compressor.layer0': 0.09894511381325877, 'gradnorm/compressor.layer1': 0.1307137661357026, 'gradnorm/compressor.layer2': 0.11152921467930968, 'gradnorm/compressor.layer3': 0.1018834041934854, 'gradnorm/compressor.layer4': 0.09413638065789816, 'gradnorm/compressor.layer5': 0.0901775482546617, 'gradnorm/compressor.layer6': 0.08503940127055126, 'gradnorm/compressor.layer7': 0.08315725898058872, 'gradnorm/decoder.layer0': 0.10531796608869066, 'gradnorm/decoder.layer1': 0.10433390920674857, 'gradnorm/decoder.layer2': 0.15100544188179682, 'gradnorm/max_param': 0.108808234333992, 'epoch': 1.19}

 12%|█▏        | 2280/19090 [1:06:28<8:20:11,  1.79s/it]
 12%|█▏        | 2281/19090 [1:06:29<7:38:28,  1.64s/it]
 12%|█▏        | 2282/19090 [1:06:32<9:07:55,  1.96s/it]
 12%|█▏        | 2283/19090 [1:06:33<8:11:24,  1.75s/it]
 12%|█▏        | 2284/19090 [1:06:34<7:31:44,  1.61s/it]
 12%|█▏        | 2285/19090 [1:06:37<9:00:48,  1.93s/it]
 12%|█▏        | 2286/19090 [1:06:38<8:06:11,  1.74s/it]
 12%|█▏        | 2287/19090 [1:06:40<7:28:01,  1.60s/it]
 12%|█▏        | 2288/19090 [1:06:42<8:43:29,  1.87s/it]
 12%|█▏        | 2289/19090 [1:06:43<7:53:55,  1.69s/it]
 12%|█▏        | 2290/19090 [1:06:45<7:19:37,  1.57s/it]
                                                        
{'loss': 0.0029, 'grad_norm': 0.37993741035461426, 'learning_rate': 4.993967388791782e-05, 'recon_loss': 0.0028574254421982912, 'recon_static': 0.0018879710551118478, 'recon_motion': 0.0009694544161902741, 'recon_norm': 1.850078147649765, 'recon_norm_std': 0.09267149977385998, 'target_norm': 1.6205794602632522, 'recon_cos': 0.48371452167630197, 'gradnorm/compressor.query': 0.005781622994740717, 'gradnorm/compressor.layer0': 0.10642409331883346, 'gradnorm/compressor.layer1': 0.1420022360120269, 'gradnorm/compressor.layer2': 0.11958121302709823, 'gradnorm/compressor.layer3': 0.11116739851141404, 'gradnorm/compressor.layer4': 0.10164685597478346, 'gradnorm/compressor.layer5': 0.09601735571317185, 'gradnorm/compressor.layer6': 0.0917837356741235, 'gradnorm/compressor.layer7': 0.08911914007355809, 'gradnorm/decoder.layer0': 0.11232753604161966, 'gradnorm/decoder.layer1': 0.10923198733655234, 'gradnorm/decoder.layer2': 0.1603228992952236, 'gradnorm/max_param': 0.11589814722537994, 'epoch': 1.2}

 12%|█▏        | 2290/19090 [1:06:45<7:19:37,  1.57s/it]
 12%|█▏        | 2291/19090 [1:06:48<9:07:16,  1.95s/it]
 12%|█▏        | 2292/19090 [1:06:49<8:10:43,  1.75s/it]
 12%|█▏        | 2293/19090 [1:06:50<7:31:35,  1.61s/it]
 12%|█▏        | 2294/19090 [1:06:53<8:59:35,  1.93s/it]
 12%|█▏        | 2295/19090 [1:06:54<8:05:15,  1.73s/it]
 12%|█▏        | 2296/19090 [1:06:55<7:27:43,  1.60s/it]
 12%|█▏        | 2297/19090 [1:06:58<8:45:15,  1.88s/it]
 12%|█▏        | 2298/19090 [1:06:59<7:58:23,  1.71s/it]
 12%|█▏        | 2299/19090 [1:07:00<7:25:42,  1.59s/it]
 12%|█▏        | 2300/19090 [1:07:03<8:21:58,  1.79s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.5704843997955322, 'learning_rate': 4.993645841566245e-05, 'recon_loss': 0.0034550664538983257, 'recon_static': 0.002482396984123625, 'recon_motion': 0.000972669420298189, 'recon_norm': 2.0232249021530153, 'recon_norm_std': 0.08752759713679552, 'target_norm': 1.621299621462822, 'recon_cos': 0.44586414024233817, 'gradnorm/compressor.query': 0.009087843935138626, 'gradnorm/compressor.layer0': 0.15539569959606195, 'gradnorm/compressor.layer1': 0.2098746455094716, 'gradnorm/compressor.layer2': 0.17735484756499742, 'gradnorm/compressor.layer3': 0.16419717392885252, 'gradnorm/compressor.layer4': 0.15266023065497802, 'gradnorm/compressor.layer5': 0.14526331284422542, 'gradnorm/compressor.layer6': 0.14217059800014395, 'gradnorm/compressor.layer7': 0.13517728074124954, 'gradnorm/decoder.layer0': 0.17082378391774822, 'gradnorm/decoder.layer1': 0.16606419918553478, 'gradnorm/decoder.layer2': 0.2431238133467715, 'gradnorm/max_param': 0.17163331806659698, 'epoch': 1.2}

 12%|█▏        | 2300/19090 [1:07:03<8:21:58,  1.79s/it]
 12%|█▏        | 2301/19090 [1:07:04<7:39:44,  1.64s/it]
 12%|█▏        | 2302/19090 [1:07:05<7:11:16,  1.54s/it]
 12%|█▏        | 2303/19090 [1:07:08<9:12:56,  1.98s/it]
 12%|█▏        | 2304/19090 [1:07:10<8:14:37,  1.77s/it]
 12%|█▏        | 2305/19090 [1:07:11<7:34:04,  1.62s/it]
 12%|█▏        | 2306/19090 [1:07:13<8:46:21,  1.88s/it]
 12%|█▏        | 2307/19090 [1:07:15<7:56:25,  1.70s/it]
 12%|█▏        | 2308/19090 [1:07:16<7:21:43,  1.58s/it]
 12%|█▏        | 2309/19090 [1:07:19<9:11:04,  1.97s/it]
 12%|█▏        | 2310/19090 [1:07:20<8:14:20,  1.77s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.48011812567710876, 'learning_rate': 4.993315956808552e-05, 'recon_loss': 0.00437315926537849, 'recon_static': 0.003403759654611349, 'recon_motion': 0.0009693994230474346, 'recon_norm': 2.267791521549225, 'recon_norm_std': 0.0801308911293745, 'target_norm': 1.6203931361436843, 'recon_cos': 0.39331220984458926, 'gradnorm/compressor.query': 0.0072361644903013835, 'gradnorm/compressor.layer0': 0.13172330636892096, 'gradnorm/compressor.layer1': 0.17594734398115527, 'gradnorm/compressor.layer2': 0.14738631728853324, 'gradnorm/compressor.layer3': 0.1377570034378256, 'gradnorm/compressor.layer4': 0.12782452898617663, 'gradnorm/compressor.layer5': 0.12395190631676578, 'gradnorm/compressor.layer6': 0.11834766767533064, 'gradnorm/compressor.layer7': 0.11437528635598904, 'gradnorm/decoder.layer0': 0.14579388320456588, 'gradnorm/decoder.layer1': 0.14160120733808504, 'gradnorm/decoder.layer2': 0.20335936780439343, 'gradnorm/max_param': 0.14673534035682678, 'epoch': 1.21}

 12%|█▏        | 2310/19090 [1:07:20<8:14:20,  1.77s/it]
 12%|█▏        | 2311/19090 [1:07:21<7:34:27,  1.63s/it]
 12%|█▏        | 2312/19090 [1:07:24<9:01:37,  1.94s/it]
 12%|█▏        | 2313/19090 [1:07:25<8:07:57,  1.75s/it]
 12%|█▏        | 2314/19090 [1:07:27<7:30:37,  1.61s/it]
 12%|█▏        | 2315/19090 [1:07:29<8:38:21,  1.85s/it]
 12%|█▏        | 2316/19090 [1:07:30<7:50:55,  1.68s/it]
 12%|█▏        | 2317/19090 [1:07:32<7:18:50,  1.57s/it]
 12%|█▏        | 2318/19090 [1:07:35<9:09:46,  1.97s/it]
 12%|█▏        | 2319/19090 [1:07:36<8:12:59,  1.76s/it]
 12%|█▏        | 2320/19090 [1:07:37<7:34:05,  1.62s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.4017801582813263, 'learning_rate': 4.992977735621678e-05, 'recon_loss': 0.0034981289762072267, 'recon_static': 0.0025175992399454118, 'recon_motion': 0.000980529711523559, 'recon_norm': 2.037967413663864, 'recon_norm_std': 0.08334070667624474, 'target_norm': 1.61985442340374, 'recon_cos': 0.4381882421672344, 'gradnorm/compressor.query': 0.006595745603789194, 'gradnorm/compressor.layer0': 0.1121985265432023, 'gradnorm/compressor.layer1': 0.14800142556829723, 'gradnorm/compressor.layer2': 0.12385893973804239, 'gradnorm/compressor.layer3': 0.11434601778089237, 'gradnorm/compressor.layer4': 0.10909420823148783, 'gradnorm/compressor.layer5': 0.10284455522890237, 'gradnorm/compressor.layer6': 0.09861466741314859, 'gradnorm/compressor.layer7': 0.09409417118230362, 'gradnorm/decoder.layer0': 0.11937388661125004, 'gradnorm/decoder.layer1': 0.1167309889383364, 'gradnorm/decoder.layer2': 0.17187624085585296, 'gradnorm/max_param': 0.12368794530630112, 'epoch': 1.22}

 12%|█▏        | 2320/19090 [1:07:37<7:34:05,  1.62s/it]
 12%|█▏        | 2321/19090 [1:07:40<9:23:18,  2.02s/it]
 12%|█▏        | 2322/19090 [1:07:41<8:22:01,  1.80s/it]
 12%|█▏        | 2323/19090 [1:07:43<7:39:22,  1.64s/it]
 12%|█▏        | 2324/19090 [1:07:45<9:12:04,  1.98s/it]
 12%|█▏        | 2325/19090 [1:07:47<8:14:13,  1.77s/it]
 12%|█▏        | 2326/19090 [1:07:48<7:33:53,  1.62s/it]
 12%|█▏        | 2327/19090 [1:07:51<9:06:11,  1.95s/it]
 12%|█▏        | 2328/19090 [1:07:52<8:10:53,  1.76s/it]
 12%|█▏        | 2329/19090 [1:07:53<7:32:00,  1.62s/it]
 12%|█▏        | 2330/19090 [1:07:56<8:58:27,  1.93s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.4411800503730774, 'learning_rate': 4.992631179136468e-05, 'recon_loss': 0.003032484173309058, 'recon_static': 0.0020691085635917263, 'recon_motion': 0.0009633757625124417, 'recon_norm': 1.905216470360756, 'recon_norm_std': 0.0919180877506733, 'target_norm': 1.6207457512617112, 'recon_cos': 0.47087180987000465, 'gradnorm/compressor.query': 0.006408989990181578, 'gradnorm/compressor.layer0': 0.1244624956857275, 'gradnorm/compressor.layer1': 0.1611101307944183, 'gradnorm/compressor.layer2': 0.13659473653029952, 'gradnorm/compressor.layer3': 0.1289611757188079, 'gradnorm/compressor.layer4': 0.11990804219689867, 'gradnorm/compressor.layer5': 0.11276559373660379, 'gradnorm/compressor.layer6': 0.10760366922619302, 'gradnorm/compressor.layer7': 0.10401812979994565, 'gradnorm/decoder.layer0': 0.13039800816152672, 'gradnorm/decoder.layer1': 0.12878979855975875, 'gradnorm/decoder.layer2': 0.18649256785326287, 'gradnorm/max_param': 0.13407503068447113, 'epoch': 1.22}

 12%|█▏        | 2330/19090 [1:07:56<8:58:27,  1.93s/it]
 12%|█▏        | 2331/19090 [1:07:57<8:04:57,  1.74s/it]
 12%|█▏        | 2332/19090 [1:07:59<7:28:01,  1.60s/it]
 12%|█▏        | 2333/19090 [1:08:01<8:43:28,  1.87s/it]
 12%|█▏        | 2334/19090 [1:08:02<7:54:04,  1.70s/it]
 12%|█▏        | 2335/19090 [1:08:04<7:19:51,  1.58s/it]
 12%|█▏        | 2336/19090 [1:08:06<8:39:30,  1.86s/it]
 12%|█▏        | 2337/19090 [1:08:07<7:51:35,  1.69s/it]
 12%|█▏        | 2338/19090 [1:08:09<7:17:59,  1.57s/it]
 12%|█▏        | 2339/19090 [1:08:11<8:45:10,  1.88s/it]
 12%|█▏        | 2340/19090 [1:08:13<7:56:34,  1.71s/it]
                                                        
{'loss': 0.005, 'grad_norm': 0.5381751656532288, 'learning_rate': 4.992276288511639e-05, 'recon_loss': 0.0049805700429715214, 'recon_static': 0.004007276496849954, 'recon_motion': 0.000973293639253825, 'recon_norm': 2.4161930799484255, 'recon_norm_std': 0.07560047078877688, 'target_norm': 1.620094406604767, 'recon_cos': 0.3729296840727329, 'gradnorm/compressor.query': 0.008242708574310824, 'gradnorm/compressor.layer0': 0.1462025474426042, 'gradnorm/compressor.layer1': 0.19567107663459696, 'gradnorm/compressor.layer2': 0.16639676380430649, 'gradnorm/compressor.layer3': 0.15463928665795362, 'gradnorm/compressor.layer4': 0.1461150855024511, 'gradnorm/compressor.layer5': 0.13752787128737046, 'gradnorm/compressor.layer6': 0.1322678379002154, 'gradnorm/compressor.layer7': 0.12850409690842093, 'gradnorm/decoder.layer0': 0.16244484189963873, 'gradnorm/decoder.layer1': 0.1600346219706298, 'gradnorm/decoder.layer2': 0.22810092204543914, 'gradnorm/max_param': 0.1632104516029358, 'epoch': 1.23}

 12%|█▏        | 2340/19090 [1:08:13<7:56:34,  1.71s/it]
 12%|█▏        | 2341/19090 [1:08:14<7:21:58,  1.58s/it]
 12%|█▏        | 2342/19090 [1:08:17<8:44:38,  1.88s/it]
 12%|█▏        | 2343/19090 [1:08:18<7:55:15,  1.70s/it]
 12%|█▏        | 2344/19090 [1:08:19<7:20:17,  1.58s/it]
 12%|█▏        | 2345/19090 [1:08:22<9:06:06,  1.96s/it]
 12%|█▏        | 2346/19090 [1:08:23<8:10:46,  1.76s/it]
 12%|█▏        | 2347/19090 [1:08:25<7:31:57,  1.62s/it]
 12%|█▏        | 2348/19090 [1:08:27<9:20:51,  2.01s/it]
 12%|█▏        | 2349/19090 [1:08:29<8:20:48,  1.79s/it]
 12%|█▏        | 2350/19090 [1:08:30<7:39:13,  1.65s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.5653182864189148, 'learning_rate': 4.99191306493377e-05, 'recon_loss': 0.004535833571571857, 'recon_static': 0.0035640840942505747, 'recon_motion': 0.0009717494700453244, 'recon_norm': 2.3140590727329253, 'recon_norm_std': 0.07744309343397618, 'target_norm': 1.620067024230957, 'recon_cos': 0.3879255548119545, 'gradnorm/compressor.query': 0.008784384359977858, 'gradnorm/compressor.layer0': 0.15898815872226277, 'gradnorm/compressor.layer1': 0.207087662046835, 'gradnorm/compressor.layer2': 0.17396512151558702, 'gradnorm/compressor.layer3': 0.16218342296349067, 'gradnorm/compressor.layer4': 0.15205597596062878, 'gradnorm/compressor.layer5': 0.14373360726970166, 'gradnorm/compressor.layer6': 0.13929775863555466, 'gradnorm/compressor.layer7': 0.13334143720775202, 'gradnorm/decoder.layer0': 0.17113814844773872, 'gradnorm/decoder.layer1': 0.16474940483777353, 'gradnorm/decoder.layer2': 0.2395473752127604, 'gradnorm/max_param': 0.1713476926088333, 'epoch': 1.23}

 12%|█▏        | 2350/19090 [1:08:30<7:39:13,  1.65s/it]
 12%|█▏        | 2351/19090 [1:08:33<8:51:35,  1.91s/it]
 12%|█▏        | 2352/19090 [1:08:34<8:00:09,  1.72s/it]
 12%|█▏        | 2353/19090 [1:08:35<7:24:21,  1.59s/it]
 12%|█▏        | 2354/19090 [1:08:38<8:53:32,  1.91s/it]
 12%|█▏        | 2355/19090 [1:08:39<8:01:38,  1.73s/it]
 12%|█▏        | 2356/19090 [1:08:40<7:25:12,  1.60s/it]
 12%|█▏        | 2357/19090 [1:08:43<8:52:02,  1.91s/it]
 12%|█▏        | 2358/19090 [1:08:44<8:00:06,  1.72s/it]
 12%|█▏        | 2359/19090 [1:08:46<7:24:02,  1.59s/it]
 12%|█▏        | 2360/19090 [1:08:48<8:50:34,  1.90s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.5389828085899353, 'learning_rate': 4.991541509617304e-05, 'recon_loss': 0.004599869134835899, 'recon_static': 0.0036280336556956173, 'recon_motion': 0.0009718354282085784, 'recon_norm': 2.3273626267910004, 'recon_norm_std': 0.07815677002072334, 'target_norm': 1.6213448166847229, 'recon_cos': 0.3830155819654465, 'gradnorm/compressor.query': 0.008321684925072235, 'gradnorm/compressor.layer0': 0.1476663144531617, 'gradnorm/compressor.layer1': 0.19481807115841676, 'gradnorm/compressor.layer2': 0.16900131642411545, 'gradnorm/compressor.layer3': 0.15313107055191755, 'gradnorm/compressor.layer4': 0.14582707116328575, 'gradnorm/compressor.layer5': 0.13630280573819523, 'gradnorm/compressor.layer6': 0.13239558079823266, 'gradnorm/compressor.layer7': 0.12627265200553234, 'gradnorm/decoder.layer0': 0.16374920836650433, 'gradnorm/decoder.layer1': 0.15891081680592284, 'gradnorm/decoder.layer2': 0.23082338946633887, 'gradnorm/max_param': 0.16385719180107117, 'epoch': 1.24}

 12%|█▏        | 2360/19090 [1:08:48<8:50:34,  1.90s/it]
 12%|█▏        | 2361/19090 [1:08:49<7:59:21,  1.72s/it]
 12%|█▏        | 2362/19090 [1:08:51<7:23:20,  1.59s/it]
 12%|█▏        | 2363/19090 [1:08:54<8:59:10,  1.93s/it]
 12%|█▏        | 2364/19090 [1:08:55<8:04:25,  1.74s/it]
 12%|█▏        | 2365/19090 [1:08:56<7:26:23,  1.60s/it]
 12%|█▏        | 2366/19090 [1:08:59<9:07:02,  1.96s/it]
 12%|█▏        | 2367/19090 [1:09:00<8:10:09,  1.76s/it]
 12%|█▏        | 2368/19090 [1:09:01<7:31:03,  1.62s/it]
 12%|█▏        | 2369/19090 [1:09:04<9:11:39,  1.98s/it]
 12%|█▏        | 2370/19090 [1:09:06<8:13:45,  1.77s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.39806145429611206, 'learning_rate': 4.99116162380454e-05, 'recon_loss': 0.0034529750817455353, 'recon_static': 0.002493189211236313, 'recon_motion': 0.0009597859170753509, 'recon_norm': 2.0290794670581818, 'recon_norm_std': 0.08636870514601469, 'target_norm': 1.6164136081933975, 'recon_cos': 0.44424482956528666, 'gradnorm/compressor.query': 0.006073138221004795, 'gradnorm/compressor.layer0': 0.11269836669564874, 'gradnorm/compressor.layer1': 0.1449059063732753, 'gradnorm/compressor.layer2': 0.12086971770715363, 'gradnorm/compressor.layer3': 0.11298075232977826, 'gradnorm/compressor.layer4': 0.10660594666280138, 'gradnorm/compressor.layer5': 0.10148075830741402, 'gradnorm/compressor.layer6': 0.09666686021114711, 'gradnorm/compressor.layer7': 0.09312423838553598, 'gradnorm/decoder.layer0': 0.11935432997846802, 'gradnorm/decoder.layer1': 0.11727349781596191, 'gradnorm/decoder.layer2': 0.17223521437773948, 'gradnorm/max_param': 0.12395010143518448, 'epoch': 1.24}

 12%|█▏        | 2370/19090 [1:09:06<8:13:45,  1.77s/it]
 12%|█▏        | 2371/19090 [1:09:07<7:32:54,  1.63s/it]
 12%|█▏        | 2372/19090 [1:09:10<9:29:34,  2.04s/it]
 12%|█▏        | 2373/19090 [1:09:11<8:27:39,  1.82s/it]
 12%|█▏        | 2374/19090 [1:09:12<7:42:21,  1.66s/it]
 12%|█▏        | 2375/19090 [1:09:15<9:06:30,  1.96s/it]
 12%|█▏        | 2376/19090 [1:09:16<8:09:50,  1.76s/it]
 12%|█▏        | 2377/19090 [1:09:18<7:31:00,  1.62s/it]
 12%|█▏        | 2378/19090 [1:09:21<9:09:04,  1.97s/it]
 12%|█▏        | 2379/19090 [1:09:22<8:11:39,  1.77s/it]
 12%|█▏        | 2380/19090 [1:09:23<7:32:20,  1.62s/it]
                                                        
{'loss': 0.0029, 'grad_norm': 0.3720359206199646, 'learning_rate': 4.9907734087656305e-05, 'recon_loss': 0.002912000642390922, 'recon_static': 0.0019471599167445674, 'recon_motion': 0.0009648407838540151, 'recon_norm': 1.8709800124168396, 'recon_norm_std': 0.09129069540649652, 'target_norm': 1.6203476697206498, 'recon_cos': 0.4803255632519722, 'gradnorm/compressor.query': 0.006207337116583461, 'gradnorm/compressor.layer0': 0.10394854706494924, 'gradnorm/compressor.layer1': 0.13581760973211868, 'gradnorm/compressor.layer2': 0.11442058331709734, 'gradnorm/compressor.layer3': 0.10712720161966534, 'gradnorm/compressor.layer4': 0.09927097977093399, 'gradnorm/compressor.layer5': 0.09427522024288515, 'gradnorm/compressor.layer6': 0.09012278472886735, 'gradnorm/compressor.layer7': 0.08790466408395721, 'gradnorm/decoder.layer0': 0.11113036727896887, 'gradnorm/decoder.layer1': 0.10980679225363259, 'gradnorm/decoder.layer2': 0.15985238510590605, 'gradnorm/max_param': 0.1151435598731041, 'epoch': 1.25}

 12%|█▏        | 2380/19090 [1:09:23<7:32:20,  1.62s/it]
 12%|█▏        | 2381/19090 [1:09:26<9:08:36,  1.97s/it]
 12%|█▏        | 2382/19090 [1:09:27<8:11:18,  1.76s/it]
 12%|█▏        | 2383/19090 [1:09:28<7:31:08,  1.62s/it]
 12%|█▏        | 2384/19090 [1:09:31<8:59:38,  1.94s/it]
 12%|█▏        | 2385/19090 [1:09:32<8:05:50,  1.75s/it]
 12%|█▏        | 2386/19090 [1:09:34<7:28:12,  1.61s/it]
 13%|█▎        | 2387/19090 [1:09:36<8:41:30,  1.87s/it]
 13%|█▎        | 2388/19090 [1:09:37<7:52:53,  1.70s/it]
 13%|█▎        | 2389/19090 [1:09:39<7:18:40,  1.58s/it]
 13%|█▎        | 2390/19090 [1:09:42<9:10:56,  1.98s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.42407548427581787, 'learning_rate': 4.9903768657985765e-05, 'recon_loss': 0.003172740200534463, 'recon_static': 0.0022012813948094844, 'recon_motion': 0.0009714586733025499, 'recon_norm': 1.9475127160549164, 'recon_norm_std': 0.0884545223787427, 'target_norm': 1.6209031969308854, 'recon_cos': 0.46064290776848793, 'gradnorm/compressor.query': 0.0067538245516531575, 'gradnorm/compressor.layer0': 0.12161021308139613, 'gradnorm/compressor.layer1': 0.15712762764552163, 'gradnorm/compressor.layer2': 0.13067589892958892, 'gradnorm/compressor.layer3': 0.12181182195818242, 'gradnorm/compressor.layer4': 0.1152037927488436, 'gradnorm/compressor.layer5': 0.10556781496136722, 'gradnorm/compressor.layer6': 0.10311918549329228, 'gradnorm/compressor.layer7': 0.09921688937898133, 'gradnorm/decoder.layer0': 0.1254244319108146, 'gradnorm/decoder.layer1': 0.12239254573370209, 'gradnorm/decoder.layer2': 0.18109862155631942, 'gradnorm/max_param': 0.1307309865951538, 'epoch': 1.25}

 13%|█▎        | 2390/19090 [1:09:42<9:10:56,  1.98s/it]
 13%|█▎        | 2391/19090 [1:09:43<8:13:26,  1.77s/it]
 13%|█▎        | 2392/19090 [1:09:44<7:32:58,  1.63s/it]
 13%|█▎        | 2393/19090 [1:09:47<9:07:21,  1.97s/it]
 13%|█▎        | 2394/19090 [1:09:48<8:10:56,  1.76s/it]
 13%|█▎        | 2395/19090 [1:09:50<7:31:09,  1.62s/it]
 13%|█▎        | 2396/19090 [1:09:52<8:54:36,  1.92s/it]
 13%|█▎        | 2397/19090 [1:09:54<8:01:33,  1.73s/it]
 13%|█▎        | 2398/19090 [1:09:55<7:24:52,  1.60s/it]
 13%|█▎        | 2399/19090 [1:09:58<9:25:42,  2.03s/it]
 13%|█▎        | 2400/19090 [1:09:59<8:24:01,  1.81s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.4724173843860626, 'learning_rate': 4.9899719962292235e-05, 'recon_loss': 0.003488141723209992, 'recon_static': 0.002510327805066481, 'recon_motion': 0.0009778139632544481, 'recon_norm': 2.0350608259439467, 'recon_norm_std': 0.0863979758694768, 'target_norm': 1.6233292758464812, 'recon_cos': 0.4396805465221405, 'gradnorm/compressor.query': 0.007282957440004923, 'gradnorm/compressor.layer0': 0.1339172041984983, 'gradnorm/compressor.layer1': 0.17643882356965437, 'gradnorm/compressor.layer2': 0.1453134339312823, 'gradnorm/compressor.layer3': 0.1346452493292926, 'gradnorm/compressor.layer4': 0.12444267976849652, 'gradnorm/compressor.layer5': 0.11900915646737945, 'gradnorm/compressor.layer6': 0.11481925174321521, 'gradnorm/compressor.layer7': 0.10971317208876237, 'gradnorm/decoder.layer0': 0.14109260909251842, 'gradnorm/decoder.layer1': 0.13854903838496097, 'gradnorm/decoder.layer2': 0.20209593713100982, 'gradnorm/max_param': 0.1448988914489746, 'epoch': 1.26}

 13%|█▎        | 2400/19090 [1:09:59<8:24:01,  1.81s/it]
 13%|█▎        | 2401/19090 [1:10:00<7:41:09,  1.66s/it]
 13%|█▎        | 2402/19090 [1:10:03<9:13:37,  1.99s/it]
 13%|█▎        | 2403/19090 [1:10:05<8:15:04,  1.78s/it]
 13%|█▎        | 2404/19090 [1:10:06<7:34:17,  1.63s/it]
 13%|█▎        | 2405/19090 [1:10:09<9:17:14,  2.00s/it]
 13%|█▎        | 2406/19090 [1:10:10<8:17:52,  1.79s/it]
 13%|█▎        | 2407/19090 [1:10:11<7:35:56,  1.64s/it]
 13%|█▎        | 2408/19090 [1:10:14<8:46:38,  1.89s/it]
 13%|█▎        | 2409/19090 [1:10:15<7:55:55,  1.71s/it]
 13%|█▎        | 2410/19090 [1:10:16<7:21:31,  1.59s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.4874488115310669, 'learning_rate': 4.989558801411258e-05, 'recon_loss': 0.004363903385819867, 'recon_static': 0.0033968435949645937, 'recon_motion': 0.0009670597384683788, 'recon_norm': 2.268117445707321, 'recon_norm_std': 0.08025507871061563, 'target_norm': 1.6200737327337265, 'recon_cos': 0.39407186582684517, 'gradnorm/compressor.query': 0.007466719596141338, 'gradnorm/compressor.layer0': 0.1348984588885488, 'gradnorm/compressor.layer1': 0.17903596473394612, 'gradnorm/compressor.layer2': 0.14970845161455965, 'gradnorm/compressor.layer3': 0.14219369403217122, 'gradnorm/compressor.layer4': 0.1318588095710602, 'gradnorm/compressor.layer5': 0.12307352201361618, 'gradnorm/compressor.layer6': 0.11784740453307112, 'gradnorm/compressor.layer7': 0.11601239562757935, 'gradnorm/decoder.layer0': 0.14686920479237714, 'gradnorm/decoder.layer1': 0.14188667401627958, 'gradnorm/decoder.layer2': 0.2075676177915016, 'gradnorm/max_param': 0.14963990449905396, 'epoch': 1.26}

 13%|█▎        | 2410/19090 [1:10:16<7:21:31,  1.59s/it]
 13%|█▎        | 2411/19090 [1:10:19<9:00:22,  1.94s/it]
 13%|█▎        | 2412/19090 [1:10:20<8:05:30,  1.75s/it]
 13%|█▎        | 2413/19090 [1:10:22<7:27:32,  1.61s/it]
 13%|█▎        | 2414/19090 [1:10:25<9:10:46,  1.98s/it]
 13%|█▎        | 2415/19090 [1:10:26<8:12:33,  1.77s/it]
 13%|█▎        | 2416/19090 [1:10:27<7:32:11,  1.63s/it]
 13%|█▎        | 2417/19090 [1:10:30<8:59:47,  1.94s/it]
 13%|█▎        | 2418/19090 [1:10:31<8:04:36,  1.74s/it]
 13%|█▎        | 2419/19090 [1:10:32<7:27:23,  1.61s/it]
 13%|█▎        | 2420/19090 [1:10:35<8:56:19,  1.93s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.36852124333381653, 'learning_rate': 4.989137282726202e-05, 'recon_loss': 0.0033980157051701097, 'recon_static': 0.0024302088044350968, 'recon_motion': 0.0009678068396169692, 'recon_norm': 2.0110878884792327, 'recon_norm_std': 0.08719948194921016, 'target_norm': 1.6221859604120255, 'recon_cos': 0.4484830744564533, 'gradnorm/compressor.query': 0.005595270201694406, 'gradnorm/compressor.layer0': 0.09908901127600787, 'gradnorm/compressor.layer1': 0.13163950379117728, 'gradnorm/compressor.layer2': 0.11415691893256534, 'gradnorm/compressor.layer3': 0.10692745347879483, 'gradnorm/compressor.layer4': 0.09862618376863923, 'gradnorm/compressor.layer5': 0.09471506855240593, 'gradnorm/compressor.layer6': 0.09052229767431817, 'gradnorm/compressor.layer7': 0.08712737823271793, 'gradnorm/decoder.layer0': 0.11228071039740632, 'gradnorm/decoder.layer1': 0.1088882839620678, 'gradnorm/decoder.layer2': 0.15879723174792554, 'gradnorm/max_param': 0.11519588530063629, 'epoch': 1.27}

 13%|█▎        | 2420/19090 [1:10:35<8:56:19,  1.93s/it]
 13%|█▎        | 2421/19090 [1:10:36<8:03:35,  1.74s/it]
 13%|█▎        | 2422/19090 [1:10:38<7:25:30,  1.60s/it]
 13%|█▎        | 2423/19090 [1:10:40<8:41:23,  1.88s/it]
 13%|█▎        | 2424/19090 [1:10:41<7:52:01,  1.70s/it]
 13%|█▎        | 2425/19090 [1:10:43<7:17:17,  1.57s/it]
 13%|█▎        | 2426/19090 [1:10:45<8:53:34,  1.92s/it]
 13%|█▎        | 2427/19090 [1:10:47<7:59:59,  1.73s/it]
 13%|█▎        | 2428/19090 [1:10:48<7:22:55,  1.59s/it]
 13%|█▎        | 2429/19090 [1:10:51<9:01:47,  1.95s/it]
 13%|█▎        | 2430/19090 [1:10:52<8:06:27,  1.75s/it]
                                                        
{'loss': 0.0029, 'grad_norm': 0.3592677116394043, 'learning_rate': 4.9887074415834055e-05, 'recon_loss': 0.0029153473617043344, 'recon_static': 0.0019524391333106904, 'recon_motion': 0.0009629082604078576, 'recon_norm': 1.8745641469955445, 'recon_norm_std': 0.09188990294933319, 'target_norm': 1.6198771864175796, 'recon_cos': 0.48077222481369974, 'gradnorm/compressor.query': 0.005489595234065431, 'gradnorm/compressor.layer0': 0.09754504547896506, 'gradnorm/compressor.layer1': 0.12940049787476066, 'gradnorm/compressor.layer2': 0.11051979154409784, 'gradnorm/compressor.layer3': 0.10299017990708993, 'gradnorm/compressor.layer4': 0.09755656877006533, 'gradnorm/compressor.layer5': 0.0916276447277441, 'gradnorm/compressor.layer6': 0.08849800367952668, 'gradnorm/compressor.layer7': 0.08615525519865919, 'gradnorm/decoder.layer0': 0.10819065524799652, 'gradnorm/decoder.layer1': 0.1056239037696676, 'gradnorm/decoder.layer2': 0.15470250447068568, 'gradnorm/max_param': 0.11143040657043457, 'epoch': 1.27}

 13%|█▎        | 2430/19090 [1:10:52<8:06:27,  1.75s/it]
 13%|█▎        | 2431/19090 [1:10:53<7:27:27,  1.61s/it]
 13%|█▎        | 2432/19090 [1:10:56<8:40:09,  1.87s/it]
 13%|█▎        | 2433/19090 [1:10:57<7:50:52,  1.70s/it]
 13%|█▎        | 2434/19090 [1:10:58<7:16:23,  1.57s/it]
 13%|█▎        | 2435/19090 [1:11:01<9:14:11,  2.00s/it]
 13%|█▎        | 2436/19090 [1:11:03<8:14:32,  1.78s/it]
 13%|█▎        | 2437/19090 [1:11:04<7:33:09,  1.63s/it]
 13%|█▎        | 2438/19090 [1:11:07<8:53:06,  1.92s/it]
 13%|█▎        | 2439/19090 [1:11:08<8:00:00,  1.73s/it]
 13%|█▎        | 2440/19090 [1:11:09<7:22:55,  1.60s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.40767326951026917, 'learning_rate': 4.9882692794200494e-05, 'recon_loss': 0.003172131581231952, 'recon_static': 0.0021996079129166903, 'recon_motion': 0.0009725236552185379, 'recon_norm': 1.9453109562397004, 'recon_norm_std': 0.08920769412070513, 'target_norm': 1.6223446160554886, 'recon_cos': 0.46026625111699104, 'gradnorm/compressor.query': 0.006268833165543257, 'gradnorm/compressor.layer0': 0.1130346211433499, 'gradnorm/compressor.layer1': 0.15058665689774, 'gradnorm/compressor.layer2': 0.1279506884510716, 'gradnorm/compressor.layer3': 0.11816370388638643, 'gradnorm/compressor.layer4': 0.10869190082283295, 'gradnorm/compressor.layer5': 0.10255011090292855, 'gradnorm/compressor.layer6': 0.1003653134449941, 'gradnorm/compressor.layer7': 0.09639136352478753, 'gradnorm/decoder.layer0': 0.12036175369899328, 'gradnorm/decoder.layer1': 0.1185334359404176, 'gradnorm/decoder.layer2': 0.17360561037621103, 'gradnorm/max_param': 0.1256309300661087, 'epoch': 1.28}

 13%|█▎        | 2440/19090 [1:11:09<7:22:55,  1.60s/it]
 13%|█▎        | 2441/19090 [1:11:12<9:15:10,  2.00s/it]
 13%|█▎        | 2442/19090 [1:11:13<8:15:31,  1.79s/it]
 13%|█▎        | 2443/19090 [1:11:15<7:34:17,  1.64s/it]
 13%|█▎        | 2444/19090 [1:11:17<9:14:03,  2.00s/it]
 13%|█▎        | 2445/19090 [1:11:19<8:14:21,  1.78s/it]
 13%|█▎        | 2446/19090 [1:11:20<7:33:50,  1.64s/it]
 13%|█▎        | 2447/19090 [1:11:23<8:59:54,  1.95s/it]
 13%|█▎        | 2448/19090 [1:11:24<8:04:58,  1.75s/it]
 13%|█▎        | 2449/19090 [1:11:25<7:28:05,  1.62s/it]
 13%|█▎        | 2450/19090 [1:11:28<9:03:21,  1.96s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.45795613527297974, 'learning_rate': 4.9878227977011325e-05, 'recon_loss': 0.0035731256240978835, 'recon_static': 0.0025991645292378963, 'recon_motion': 0.0009739611021359451, 'recon_norm': 2.060335543751717, 'recon_norm_std': 0.08328231889754534, 'target_norm': 1.6217389345169066, 'recon_cos': 0.4345382496714592, 'gradnorm/compressor.query': 0.007285394687048896, 'gradnorm/compressor.layer0': 0.12572656265976548, 'gradnorm/compressor.layer1': 0.1680696372963411, 'gradnorm/compressor.layer2': 0.14365488564882947, 'gradnorm/compressor.layer3': 0.13313607460844137, 'gradnorm/compressor.layer4': 0.12446481157738289, 'gradnorm/compressor.layer5': 0.11502861386760345, 'gradnorm/compressor.layer6': 0.11062016753374687, 'gradnorm/compressor.layer7': 0.10782200678084386, 'gradnorm/decoder.layer0': 0.13805255810190564, 'gradnorm/decoder.layer1': 0.13370617157323914, 'gradnorm/decoder.layer2': 0.19424960840337943, 'gradnorm/max_param': 0.1405329555273056, 'epoch': 1.28}

 13%|█▎        | 2450/19090 [1:11:28<9:03:21,  1.96s/it]
 13%|█▎        | 2451/19090 [1:11:29<8:08:39,  1.76s/it]
 13%|█▎        | 2452/19090 [1:11:31<7:30:10,  1.62s/it]
 13%|█▎        | 2453/19090 [1:11:34<9:13:22,  2.00s/it]
 13%|█▎        | 2454/19090 [1:11:35<8:13:27,  1.78s/it]
 13%|█▎        | 2455/19090 [1:11:36<7:32:39,  1.63s/it]
 13%|█▎        | 2456/19090 [1:11:39<9:01:02,  1.95s/it]
 13%|█▎        | 2457/19090 [1:11:40<8:04:59,  1.75s/it]
 13%|█▎        | 2458/19090 [1:11:41<7:25:58,  1.61s/it]
 13%|█▎        | 2459/19090 [1:11:44<8:37:27,  1.87s/it]
 13%|█▎        | 2460/19090 [1:11:45<7:48:33,  1.69s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.4759577512741089, 'learning_rate': 4.987367997919472e-05, 'recon_loss': 0.003325044072698802, 'recon_static': 0.0023551678517833353, 'recon_motion': 0.0009698762631160207, 'recon_norm': 1.9917100489139556, 'recon_norm_std': 0.08767592255026102, 'target_norm': 1.6235290557146071, 'recon_cos': 0.45152418240904807, 'gradnorm/compressor.query': 0.007322963219556802, 'gradnorm/compressor.layer0': 0.13503486579330348, 'gradnorm/compressor.layer1': 0.17645418317520242, 'gradnorm/compressor.layer2': 0.14975469692638518, 'gradnorm/compressor.layer3': 0.13735882246747938, 'gradnorm/compressor.layer4': 0.12698074920592345, 'gradnorm/compressor.layer5': 0.11996111383179822, 'gradnorm/compressor.layer6': 0.11591725546047761, 'gradnorm/compressor.layer7': 0.11088407365007436, 'gradnorm/decoder.layer0': 0.1401584488771711, 'gradnorm/decoder.layer1': 0.138451131898676, 'gradnorm/decoder.layer2': 0.2018854762096057, 'gradnorm/max_param': 0.14615680277347565, 'epoch': 1.29}

 13%|█▎        | 2460/19090 [1:11:45<7:48:33,  1.69s/it]
 13%|█▎        | 2461/19090 [1:11:46<7:14:49,  1.57s/it]
 13%|█▎        | 2462/19090 [1:11:49<8:48:29,  1.91s/it]
 13%|█▎        | 2463/19090 [1:11:50<7:56:48,  1.72s/it]
 13%|█▎        | 2464/19090 [1:11:52<7:20:19,  1.59s/it]
 13%|█▎        | 2465/19090 [1:11:54<8:52:36,  1.92s/it]
 13%|█▎        | 2466/19090 [1:11:56<7:59:32,  1.73s/it]
 13%|█▎        | 2467/19090 [1:11:57<7:22:40,  1.60s/it]
 13%|█▎        | 2468/19090 [1:11:59<8:42:37,  1.89s/it]
 13%|█▎        | 2469/19090 [1:12:01<7:52:27,  1.71s/it]
 13%|█▎        | 2470/19090 [1:12:02<7:17:29,  1.58s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.4294315278530121, 'learning_rate': 4.9869048815956946e-05, 'recon_loss': 0.0037418526655528693, 'recon_static': 0.002781267010141164, 'recon_motion': 0.0009605856947018764, 'recon_norm': 2.1119121998548507, 'recon_norm_std': 0.08329004440456629, 'target_norm': 1.6190394729375839, 'recon_cos': 0.4265995010733604, 'gradnorm/compressor.query': 0.006790792510863945, 'gradnorm/compressor.layer0': 0.11795547691731197, 'gradnorm/compressor.layer1': 0.158514499359496, 'gradnorm/compressor.layer2': 0.13345118829884875, 'gradnorm/compressor.layer3': 0.12408149666883449, 'gradnorm/compressor.layer4': 0.11528172570800248, 'gradnorm/compressor.layer5': 0.10930947060774007, 'gradnorm/compressor.layer6': 0.10419926629963641, 'gradnorm/compressor.layer7': 0.1004628474576608, 'gradnorm/decoder.layer0': 0.12890791891925016, 'gradnorm/decoder.layer1': 0.1260011560253631, 'gradnorm/decoder.layer2': 0.1828586491590989, 'gradnorm/max_param': 0.13180187344551086, 'epoch': 1.29}

 13%|█▎        | 2470/19090 [1:12:02<7:17:29,  1.58s/it]
 13%|█▎        | 2471/19090 [1:12:05<8:44:59,  1.90s/it]
 13%|█▎        | 2472/19090 [1:12:06<7:53:50,  1.71s/it]
 13%|█▎        | 2473/19090 [1:12:07<7:18:22,  1.58s/it]
 13%|█▎        | 2474/19090 [1:12:10<8:22:24,  1.81s/it]
 13%|█▎        | 2475/19090 [1:12:11<7:49:08,  1.69s/it]
 13%|█▎        | 2476/19090 [1:12:12<7:15:05,  1.57s/it]
 13%|█▎        | 2477/19090 [1:12:15<8:36:42,  1.87s/it]
 13%|█▎        | 2478/19090 [1:12:16<7:51:02,  1.70s/it]
 13%|█▎        | 2479/19090 [1:12:17<7:16:23,  1.58s/it]
 13%|█▎        | 2480/19090 [1:12:20<9:17:12,  2.01s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.5008190274238586, 'learning_rate': 4.986433450278236e-05, 'recon_loss': 0.003705383976921439, 'recon_static': 0.002739051973912865, 'recon_motion': 0.0009663320335675963, 'recon_norm': 2.104048377275467, 'recon_norm_std': 0.08262506555765867, 'target_norm': 1.6196602076292037, 'recon_cos': 0.4291616052389145, 'gradnorm/compressor.query': 0.007643388550270439, 'gradnorm/compressor.layer0': 0.1385157135685069, 'gradnorm/compressor.layer1': 0.1830906264329547, 'gradnorm/compressor.layer2': 0.15555236163921576, 'gradnorm/compressor.layer3': 0.14437771709989985, 'gradnorm/compressor.layer4': 0.13511657628684579, 'gradnorm/compressor.layer5': 0.12738760181443895, 'gradnorm/compressor.layer6': 0.12200181183485825, 'gradnorm/compressor.layer7': 0.11759754999469899, 'gradnorm/decoder.layer0': 0.14955522687805056, 'gradnorm/decoder.layer1': 0.1463207220800646, 'gradnorm/decoder.layer2': 0.21456184098253347, 'gradnorm/max_param': 0.1537623405456543, 'epoch': 1.3}

 13%|█▎        | 2480/19090 [1:12:20<9:17:12,  2.01s/it]
 13%|█▎        | 2481/19090 [1:12:22<8:16:26,  1.79s/it]
 13%|█▎        | 2482/19090 [1:12:23<7:34:13,  1.64s/it]
 13%|█▎        | 2483/19090 [1:12:26<9:11:59,  1.99s/it]
 13%|█▎        | 2484/19090 [1:12:27<8:12:38,  1.78s/it]
 13%|█▎        | 2485/19090 [1:12:28<7:31:45,  1.63s/it]
 13%|█▎        | 2486/19090 [1:12:31<9:02:28,  1.96s/it]
 13%|█▎        | 2487/19090 [1:12:32<8:06:13,  1.76s/it]
 13%|█▎        | 2488/19090 [1:12:34<7:26:40,  1.61s/it]
 13%|█▎        | 2489/19090 [1:12:37<9:35:07,  2.08s/it]
 13%|█▎        | 2490/19090 [1:12:38<8:29:42,  1.84s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.4607245624065399, 'learning_rate': 4.9859537055433326e-05, 'recon_loss': 0.0036278908723033966, 'recon_static': 0.002657888567773625, 'recon_motion': 0.0009700023991172201, 'recon_norm': 2.0797514796257017, 'recon_norm_std': 0.08239910751581192, 'target_norm': 1.6215802162885666, 'recon_cos': 0.43316895738244054, 'gradnorm/compressor.query': 0.006831117403739115, 'gradnorm/compressor.layer0': 0.12889656117180698, 'gradnorm/compressor.layer1': 0.17077184075334542, 'gradnorm/compressor.layer2': 0.1419241092673588, 'gradnorm/compressor.layer3': 0.1325278738455774, 'gradnorm/compressor.layer4': 0.12239913902147419, 'gradnorm/compressor.layer5': 0.11659910784107289, 'gradnorm/compressor.layer6': 0.1108256553609765, 'gradnorm/compressor.layer7': 0.10882914094252182, 'gradnorm/decoder.layer0': 0.13787716834999447, 'gradnorm/decoder.layer1': 0.13521762733178555, 'gradnorm/decoder.layer2': 0.19682500749459841, 'gradnorm/max_param': 0.14204265177249908, 'epoch': 1.3}

 13%|█▎        | 2490/19090 [1:12:38<8:29:42,  1.84s/it]
 13%|█▎        | 2491/19090 [1:12:39<7:43:45,  1.68s/it]
 13%|█▎        | 2492/19090 [1:12:42<9:26:10,  2.05s/it]
 13%|█▎        | 2493/19090 [1:12:44<8:22:40,  1.82s/it]
 13%|█▎        | 2494/19090 [1:12:45<7:38:33,  1.66s/it]
 13%|█▎        | 2495/19090 [1:12:48<9:03:45,  1.97s/it]
 13%|█▎        | 2496/19090 [1:12:49<8:07:05,  1.76s/it]
 13%|█▎        | 2497/19090 [1:12:50<7:27:51,  1.62s/it]
 13%|█▎        | 2498/19090 [1:12:53<9:00:25,  1.95s/it]
 13%|█▎        | 2499/19090 [1:12:54<8:04:33,  1.75s/it]
 13%|█▎        | 2500/19090 [1:12:55<7:25:32,  1.61s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.3877277672290802, 'learning_rate': 4.9854656489950144e-05, 'recon_loss': 0.003517907956847921, 'recon_static': 0.002552406012546271, 'recon_motion': 0.0009655019573983736, 'recon_norm': 2.0445575386285784, 'recon_norm_std': 0.08470449000597, 'target_norm': 1.6218648672103881, 'recon_cos': 0.4379537522792816, 'gradnorm/compressor.query': 0.005833877341656216, 'gradnorm/compressor.layer0': 0.10615109545579984, 'gradnorm/compressor.layer1': 0.1420580182696489, 'gradnorm/compressor.layer2': 0.1191145094405783, 'gradnorm/compressor.layer3': 0.11250087353579148, 'gradnorm/compressor.layer4': 0.1042086015175375, 'gradnorm/compressor.layer5': 0.09945492363746927, 'gradnorm/compressor.layer6': 0.09507174375649322, 'gradnorm/compressor.layer7': 0.09179784896009248, 'gradnorm/decoder.layer0': 0.11580921509119563, 'gradnorm/decoder.layer1': 0.11312705265996487, 'gradnorm/decoder.layer2': 0.16607173083811821, 'gradnorm/max_param': 0.11976988613605499, 'epoch': 1.31}

 13%|█▎        | 2500/19090 [1:12:55<7:25:32,  1.61s/it]
 13%|█▎        | 2501/19090 [1:12:59<10:31:25,  2.28s/it]
 13%|█▎        | 2502/19090 [1:13:01<9:08:25,  1.98s/it] 
 13%|█▎        | 2503/19090 [1:13:02<8:10:35,  1.77s/it]
 13%|█▎        | 2504/19090 [1:13:03<7:29:45,  1.63s/it]
 13%|█▎        | 2505/19090 [1:13:04<7:01:10,  1.52s/it]
 13%|█▎        | 2506/19090 [1:13:06<6:41:04,  1.45s/it]
 13%|█▎        | 2507/19090 [1:13:08<8:04:05,  1.75s/it]
 13%|█▎        | 2508/19090 [1:13:09<7:25:15,  1.61s/it]
 13%|█▎        | 2509/19090 [1:13:11<6:59:05,  1.52s/it]
 13%|█▎        | 2510/19090 [1:13:13<8:33:42,  1.86s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.4084790050983429, 'learning_rate': 4.984969282265105e-05, 'recon_loss': 0.002990705333650112, 'recon_static': 0.0020228610024787485, 'recon_motion': 0.0009678443180746399, 'recon_norm': 1.8956119507551192, 'recon_norm_std': 0.0915228247642517, 'target_norm': 1.6230654805898665, 'recon_cos': 0.4750954620540142, 'gradnorm/compressor.query': 0.006112421809628654, 'gradnorm/compressor.layer0': 0.1149255161434692, 'gradnorm/compressor.layer1': 0.15031311589192176, 'gradnorm/compressor.layer2': 0.12389610542437256, 'gradnorm/compressor.layer3': 0.11902655941380937, 'gradnorm/compressor.layer4': 0.1108397597257993, 'gradnorm/compressor.layer5': 0.10364485717038989, 'gradnorm/compressor.layer6': 0.09880094078937002, 'gradnorm/compressor.layer7': 0.09616150851278431, 'gradnorm/decoder.layer0': 0.12212369895825759, 'gradnorm/decoder.layer1': 0.11951435209646988, 'gradnorm/decoder.layer2': 0.1739655497951962, 'gradnorm/max_param': 0.12544727325439453, 'epoch': 1.31}

 13%|█▎        | 2510/19090 [1:13:13<8:33:42,  1.86s/it]
 13%|█▎        | 2511/19090 [1:13:15<7:46:41,  1.69s/it]
 13%|█▎        | 2512/19090 [1:13:16<7:13:41,  1.57s/it]
 13%|█▎        | 2513/19090 [1:13:18<8:19:17,  1.81s/it]
 13%|█▎        | 2514/19090 [1:13:20<7:35:36,  1.65s/it]
 13%|█▎        | 2515/19090 [1:13:21<7:05:31,  1.54s/it]
 13%|█▎        | 2516/19090 [1:13:24<9:05:47,  1.98s/it]
 13%|█▎        | 2517/19090 [1:13:25<8:08:16,  1.77s/it]
 13%|█▎        | 2518/19090 [1:13:27<7:28:08,  1.62s/it]
 13%|█▎        | 2519/19090 [1:13:29<8:56:59,  1.94s/it]
 13%|█▎        | 2520/19090 [1:13:30<8:02:10,  1.75s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.4375040829181671, 'learning_rate': 4.984464607013212e-05, 'recon_loss': 0.0034866852103732525, 'recon_static': 0.0025266139418818057, 'recon_motion': 0.0009600712772225961, 'recon_norm': 2.0424143373966217, 'recon_norm_std': 0.08612785022705793, 'target_norm': 1.6193658262491226, 'recon_cos': 0.44183029010891917, 'gradnorm/compressor.query': 0.006871350450685549, 'gradnorm/compressor.layer0': 0.12011216703638325, 'gradnorm/compressor.layer1': 0.16124798097935916, 'gradnorm/compressor.layer2': 0.1361455755394283, 'gradnorm/compressor.layer3': 0.1269005886826264, 'gradnorm/compressor.layer4': 0.11778359598660894, 'gradnorm/compressor.layer5': 0.11227150717146875, 'gradnorm/compressor.layer6': 0.10691143050070849, 'gradnorm/compressor.layer7': 0.1015843404614878, 'gradnorm/decoder.layer0': 0.1304595589969549, 'gradnorm/decoder.layer1': 0.1286511280795201, 'gradnorm/decoder.layer2': 0.18573444322231963, 'gradnorm/max_param': 0.1347804069519043, 'epoch': 1.32}

 13%|█▎        | 2520/19090 [1:13:31<8:02:10,  1.75s/it]
 13%|█▎        | 2521/19090 [1:13:32<7:23:51,  1.61s/it]
 13%|█▎        | 2522/19090 [1:13:34<8:49:02,  1.92s/it]
 13%|█▎        | 2523/19090 [1:13:36<8:23:58,  1.83s/it]
 13%|█▎        | 2524/19090 [1:13:37<7:39:13,  1.66s/it]
 13%|█▎        | 2525/19090 [1:13:40<8:53:31,  1.93s/it]
 13%|█▎        | 2526/19090 [1:13:41<8:06:20,  1.76s/it]
 13%|█▎        | 2527/19090 [1:13:43<7:27:08,  1.62s/it]
 13%|█▎        | 2528/19090 [1:13:45<8:54:33,  1.94s/it]
 13%|█▎        | 2529/19090 [1:13:46<8:00:31,  1.74s/it]
 13%|█▎        | 2530/19090 [1:13:48<7:22:54,  1.60s/it]
                                                        
{'loss': 0.0028, 'grad_norm': 0.362169474363327, 'learning_rate': 4.9839516249267214e-05, 'recon_loss': 0.0027993225085083395, 'recon_static': 0.001834919152315706, 'recon_motion': 0.0009644033227232285, 'recon_norm': 1.8352029889822006, 'recon_norm_std': 0.0904372887685895, 'target_norm': 1.6197641909122467, 'recon_cos': 0.48965577930212023, 'gradnorm/compressor.query': 0.005919930841765729, 'gradnorm/compressor.layer0': 0.09695967196446326, 'gradnorm/compressor.layer1': 0.13124481060350335, 'gradnorm/compressor.layer2': 0.11247864808670593, 'gradnorm/compressor.layer3': 0.10377883735857252, 'gradnorm/compressor.layer4': 0.09750087775850107, 'gradnorm/compressor.layer5': 0.09254940488836245, 'gradnorm/compressor.layer6': 0.08895520682539693, 'gradnorm/compressor.layer7': 0.08537727869844537, 'gradnorm/decoder.layer0': 0.10986991862274427, 'gradnorm/decoder.layer1': 0.10647657355335695, 'gradnorm/decoder.layer2': 0.15620830694885043, 'gradnorm/max_param': 0.11210325360298157, 'epoch': 1.33}

 13%|█▎        | 2530/19090 [1:13:48<7:22:54,  1.60s/it]
 13%|█▎        | 2531/19090 [1:13:51<9:35:22,  2.08s/it]
 13%|█▎        | 2532/19090 [1:13:52<8:28:49,  1.84s/it]
 13%|█▎        | 2533/19090 [1:13:54<7:42:22,  1.68s/it]
 13%|█▎        | 2534/19090 [1:13:56<8:59:26,  1.95s/it]
 13%|█▎        | 2535/19090 [1:13:57<8:03:42,  1.75s/it]
 13%|█▎        | 2536/19090 [1:13:59<7:24:36,  1.61s/it]
 13%|█▎        | 2537/19090 [1:14:02<9:25:26,  2.05s/it]
 13%|█▎        | 2538/19090 [1:14:03<8:22:20,  1.82s/it]
 13%|█▎        | 2539/19090 [1:14:04<7:37:48,  1.66s/it]
 13%|█▎        | 2540/19090 [1:14:07<8:52:31,  1.93s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.49322509765625, 'learning_rate': 4.9834303377207956e-05, 'recon_loss': 0.0034057922195643185, 'recon_static': 0.002436241682153195, 'recon_motion': 0.0009695505141280591, 'recon_norm': 2.016598382592201, 'recon_norm_std': 0.08634335454553366, 'target_norm': 1.6234640628099442, 'recon_cos': 0.4482577510178089, 'gradnorm/compressor.query': 0.007630129301928694, 'gradnorm/compressor.layer0': 0.1350245200402195, 'gradnorm/compressor.layer1': 0.18068909342804443, 'gradnorm/compressor.layer2': 0.1514761071410534, 'gradnorm/compressor.layer3': 0.1421079984517507, 'gradnorm/compressor.layer4': 0.13117942530456042, 'gradnorm/compressor.layer5': 0.1237736762762701, 'gradnorm/compressor.layer6': 0.11847184820508683, 'gradnorm/compressor.layer7': 0.11564692099366189, 'gradnorm/decoder.layer0': 0.14942796804508132, 'gradnorm/decoder.layer1': 0.14553590157791008, 'gradnorm/decoder.layer2': 0.21388252412335082, 'gradnorm/max_param': 0.1510952264070511, 'epoch': 1.33}

 13%|█▎        | 2540/19090 [1:14:07<8:52:31,  1.93s/it]
 13%|█▎        | 2541/19090 [1:14:08<7:58:47,  1.74s/it]
 13%|█▎        | 2542/19090 [1:14:09<7:21:16,  1.60s/it]
 13%|█▎        | 2543/19090 [1:14:12<9:01:40,  1.96s/it]
 13%|█▎        | 2544/19090 [1:14:14<8:05:20,  1.76s/it]
 13%|█▎        | 2545/19090 [1:14:15<7:26:15,  1.62s/it]
 13%|█▎        | 2546/19090 [1:14:17<8:41:36,  1.89s/it]
 13%|█▎        | 2547/19090 [1:14:19<7:51:18,  1.71s/it]
 13%|█▎        | 2548/19090 [1:14:20<7:16:27,  1.58s/it]
 13%|█▎        | 2549/19090 [1:14:23<8:48:33,  1.92s/it]
 13%|█▎        | 2550/19090 [1:14:24<7:56:34,  1.73s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 0.4856288433074951, 'learning_rate': 4.982900747138364e-05, 'recon_loss': 0.0040942945808637886, 'recon_static': 0.003129709599306807, 'recon_motion': 0.0009645849509979598, 'recon_norm': 2.203996241092682, 'recon_norm_std': 0.08030310533940792, 'target_norm': 1.6206173777580262, 'recon_cos': 0.40717838108539584, 'gradnorm/compressor.query': 0.007970597473584484, 'gradnorm/compressor.layer0': 0.13710373081642682, 'gradnorm/compressor.layer1': 0.17993161992107878, 'gradnorm/compressor.layer2': 0.15006923394026853, 'gradnorm/compressor.layer3': 0.14114010037912003, 'gradnorm/compressor.layer4': 0.12883019371060495, 'gradnorm/compressor.layer5': 0.1214373313487399, 'gradnorm/compressor.layer6': 0.11731234496583486, 'gradnorm/compressor.layer7': 0.11451396187802838, 'gradnorm/decoder.layer0': 0.1437700433855402, 'gradnorm/decoder.layer1': 0.14307103458768705, 'gradnorm/decoder.layer2': 0.2068784707835644, 'gradnorm/max_param': 0.14958523213863373, 'epoch': 1.34}

 13%|█▎        | 2550/19090 [1:14:24<7:56:34,  1.73s/it]
 13%|█▎        | 2551/19090 [1:14:25<7:19:45,  1.60s/it]
 13%|█▎        | 2552/19090 [1:14:28<8:41:32,  1.89s/it]
 13%|█▎        | 2553/19090 [1:14:29<7:51:32,  1.71s/it]
 13%|█▎        | 2554/19090 [1:14:30<7:15:58,  1.58s/it]
 13%|█▎        | 2555/19090 [1:14:33<8:37:49,  1.88s/it]
 13%|█▎        | 2556/19090 [1:14:34<7:55:27,  1.73s/it]
 13%|█▎        | 2557/19090 [1:14:36<7:19:09,  1.59s/it]
 13%|█▎        | 2558/19090 [1:14:38<8:41:13,  1.89s/it]
 13%|█▎        | 2559/19090 [1:14:40<7:59:58,  1.74s/it]
 13%|█▎        | 2560/19090 [1:14:41<7:22:15,  1.61s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.37972769141197205, 'learning_rate': 4.9823628549501185e-05, 'recon_loss': 0.003172563744010404, 'recon_static': 0.002216084615793079, 'recon_motion': 0.000956479090382345, 'recon_norm': 1.9528601527214051, 'recon_norm_std': 0.08971951026469469, 'target_norm': 1.6204005807638169, 'recon_cos': 0.46334306001663206, 'gradnorm/compressor.query': 0.00587994286061112, 'gradnorm/compressor.layer0': 0.10298057597501709, 'gradnorm/compressor.layer1': 0.1377672791950103, 'gradnorm/compressor.layer2': 0.11920526189485058, 'gradnorm/compressor.layer3': 0.11070371242009294, 'gradnorm/compressor.layer4': 0.10145537522225019, 'gradnorm/compressor.layer5': 0.09581666653333501, 'gradnorm/compressor.layer6': 0.0926128050878661, 'gradnorm/compressor.layer7': 0.0902602400184586, 'gradnorm/decoder.layer0': 0.11449388016016003, 'gradnorm/decoder.layer1': 0.11082986963238871, 'gradnorm/decoder.layer2': 0.1628445306231072, 'gradnorm/max_param': 0.11766808480024338, 'epoch': 1.34}

 13%|█▎        | 2560/19090 [1:14:41<7:22:15,  1.61s/it]
 13%|█▎        | 2561/19090 [1:14:44<9:05:29,  1.98s/it]
 13%|█▎        | 2562/19090 [1:14:45<8:07:48,  1.77s/it]
 13%|█▎        | 2563/19090 [1:14:46<7:27:37,  1.63s/it]
 13%|█▎        | 2564/19090 [1:14:49<9:09:19,  1.99s/it]
 13%|█▎        | 2565/19090 [1:14:50<8:12:17,  1.79s/it]
 13%|█▎        | 2566/19090 [1:14:52<7:31:04,  1.64s/it]
 13%|█▎        | 2567/19090 [1:14:54<8:55:47,  1.95s/it]
 13%|█▎        | 2568/19090 [1:14:56<8:01:29,  1.75s/it]
 13%|█▎        | 2569/19090 [1:14:57<7:23:29,  1.61s/it]
 13%|█▎        | 2570/19090 [1:15:00<9:13:08,  2.01s/it]
                                                        
{'loss': 0.0028, 'grad_norm': 0.38299691677093506, 'learning_rate': 4.981816662954507e-05, 'recon_loss': 0.0027675117074977607, 'recon_static': 0.001819693180732429, 'recon_motion': 0.0009478185282205231, 'recon_norm': 1.8341707587242126, 'recon_norm_std': 0.09241478405892849, 'target_norm': 1.6160206705331803, 'recon_cos': 0.493549220263958, 'gradnorm/compressor.query': 0.005400309297999524, 'gradnorm/compressor.layer0': 0.10274054849590863, 'gradnorm/compressor.layer1': 0.1379235520101539, 'gradnorm/compressor.layer2': 0.11789620992325459, 'gradnorm/compressor.layer3': 0.11037416986118716, 'gradnorm/compressor.layer4': 0.10319552773343386, 'gradnorm/compressor.layer5': 0.09727766085080798, 'gradnorm/compressor.layer6': 0.09584947509289826, 'gradnorm/compressor.layer7': 0.09192097168977388, 'gradnorm/decoder.layer0': 0.11579991831709685, 'gradnorm/decoder.layer1': 0.11304974742791173, 'gradnorm/decoder.layer2': 0.16449391319078094, 'gradnorm/max_param': 0.11823207885026932, 'epoch': 1.35}

 13%|█▎        | 2570/19090 [1:15:00<9:13:08,  2.01s/it]
 13%|█▎        | 2571/19090 [1:15:01<8:14:00,  1.79s/it]
 13%|█▎        | 2572/19090 [1:15:03<7:32:13,  1.64s/it]
 13%|█▎        | 2573/19090 [1:15:05<9:13:34,  2.01s/it]
 13%|█▎        | 2574/19090 [1:15:07<8:13:44,  1.79s/it]
 13%|█▎        | 2575/19090 [1:15:08<7:31:57,  1.64s/it]
 13%|█▎        | 2576/19090 [1:15:11<9:06:33,  1.99s/it]
 13%|█▎        | 2577/19090 [1:15:12<8:09:30,  1.78s/it]
 14%|█▎        | 2578/19090 [1:15:13<7:29:15,  1.63s/it]
 14%|█▎        | 2579/19090 [1:15:16<8:42:23,  1.90s/it]
 14%|█▎        | 2580/19090 [1:15:17<7:52:02,  1.72s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.48283877968788147, 'learning_rate': 4.9812621729777284e-05, 'recon_loss': 0.0034547292219940574, 'recon_static': 0.002493769198190421, 'recon_motion': 0.0009609599568648264, 'recon_norm': 2.0321247041225434, 'recon_norm_std': 0.08528816979378462, 'target_norm': 1.619926279783249, 'recon_cos': 0.4429264232516289, 'gradnorm/compressor.query': 0.00795560716505502, 'gradnorm/compressor.layer0': 0.1348373733484726, 'gradnorm/compressor.layer1': 0.17667121241077446, 'gradnorm/compressor.layer2': 0.1486693691958579, 'gradnorm/compressor.layer3': 0.139583966238333, 'gradnorm/compressor.layer4': 0.128745976426854, 'gradnorm/compressor.layer5': 0.12039213192595095, 'gradnorm/compressor.layer6': 0.11590897574244777, 'gradnorm/compressor.layer7': 0.11179970112580023, 'gradnorm/decoder.layer0': 0.1447267498885163, 'gradnorm/decoder.layer1': 0.140925212787588, 'gradnorm/decoder.layer2': 0.21042884320885835, 'gradnorm/max_param': 0.1497020125389099, 'epoch': 1.35}

 14%|█▎        | 2580/19090 [1:15:17<7:52:02,  1.72s/it]
 14%|█▎        | 2581/19090 [1:15:18<7:17:04,  1.59s/it]
 14%|█▎        | 2582/19090 [1:15:21<8:45:22,  1.91s/it]
 14%|█▎        | 2583/19090 [1:15:22<7:53:34,  1.72s/it]
 14%|█▎        | 2584/19090 [1:15:24<7:17:34,  1.59s/it]
 14%|█▎        | 2585/19090 [1:15:26<8:43:51,  1.90s/it]
 14%|█▎        | 2586/19090 [1:15:28<7:52:44,  1.72s/it]
 14%|█▎        | 2587/19090 [1:15:29<7:16:37,  1.59s/it]
 14%|█▎        | 2588/19090 [1:15:32<8:48:38,  1.92s/it]
 14%|█▎        | 2589/19090 [1:15:33<7:55:58,  1.73s/it]
 14%|█▎        | 2590/19090 [1:15:34<7:18:52,  1.60s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.5173030495643616, 'learning_rate': 4.980699386873727e-05, 'recon_loss': 0.003954301198245957, 'recon_static': 0.002984267013380304, 'recon_motion': 0.0009700340451672673, 'recon_norm': 2.1736544370651245, 'recon_norm_std': 0.08155311271548271, 'target_norm': 1.6221489578485488, 'recon_cos': 0.41698737666010854, 'gradnorm/compressor.query': 0.007537337433007023, 'gradnorm/compressor.layer0': 0.13943901231023437, 'gradnorm/compressor.layer1': 0.18918408629970487, 'gradnorm/compressor.layer2': 0.16157855600635815, 'gradnorm/compressor.layer3': 0.14965886276124638, 'gradnorm/compressor.layer4': 0.13753824541997092, 'gradnorm/compressor.layer5': 0.13106465252030514, 'gradnorm/compressor.layer6': 0.12579496735356063, 'gradnorm/compressor.layer7': 0.12342499355747298, 'gradnorm/decoder.layer0': 0.15594260854322808, 'gradnorm/decoder.layer1': 0.15146971276018412, 'gradnorm/decoder.layer2': 0.22224400654865437, 'gradnorm/max_param': 0.1597510129213333, 'epoch': 1.36}

 14%|█▎        | 2590/19090 [1:15:34<7:18:52,  1.60s/it]
 14%|█▎        | 2591/19090 [1:15:37<8:50:58,  1.93s/it]
 14%|█▎        | 2592/19090 [1:15:38<8:12:45,  1.79s/it]
 14%|█▎        | 2593/19090 [1:15:40<7:31:07,  1.64s/it]
 14%|█▎        | 2594/19090 [1:15:42<8:51:03,  1.93s/it]
 14%|█▎        | 2595/19090 [1:15:44<8:29:23,  1.85s/it]
 14%|█▎        | 2596/19090 [1:15:45<7:44:03,  1.69s/it]
 14%|█▎        | 2597/19090 [1:15:48<8:59:22,  1.96s/it]
 14%|█▎        | 2598/19090 [1:15:49<8:26:05,  1.84s/it]
 14%|█▎        | 2599/19090 [1:15:51<7:40:57,  1.68s/it]
 14%|█▎        | 2600/19090 [1:15:53<8:39:40,  1.89s/it]
                                                        
{'loss': 0.005, 'grad_norm': 0.5995169878005981, 'learning_rate': 4.980128306524183e-05, 'recon_loss': 0.004960044112522155, 'recon_static': 0.004001939902082086, 'recon_motion': 0.000958104187157005, 'recon_norm': 2.419656699895859, 'recon_norm_std': 0.07448479300364852, 'target_norm': 1.6178555846214295, 'recon_cos': 0.37064071372151375, 'gradnorm/compressor.query': 0.009242429333159611, 'gradnorm/compressor.layer0': 0.16401144209604088, 'gradnorm/compressor.layer1': 0.22529009621669385, 'gradnorm/compressor.layer2': 0.1866628215935188, 'gradnorm/compressor.layer3': 0.17225165710565657, 'gradnorm/compressor.layer4': 0.15824189123212115, 'gradnorm/compressor.layer5': 0.15082476857559796, 'gradnorm/compressor.layer6': 0.14518255966458296, 'gradnorm/compressor.layer7': 0.14021851149705158, 'gradnorm/decoder.layer0': 0.17795208811945634, 'gradnorm/decoder.layer1': 0.1741623077139905, 'gradnorm/decoder.layer2': 0.2581450744794054, 'gradnorm/max_param': 0.18598270416259766, 'epoch': 1.36}

 14%|█▎        | 2600/19090 [1:15:53<8:39:40,  1.89s/it]
 14%|█▎        | 2601/19090 [1:15:55<8:30:16,  1.86s/it]
 14%|█▎        | 2602/19090 [1:15:56<7:43:47,  1.69s/it]
 14%|█▎        | 2603/19090 [1:15:58<8:19:44,  1.82s/it]
 14%|█▎        | 2604/19090 [1:16:00<8:47:45,  1.92s/it]
 14%|█▎        | 2605/19090 [1:16:02<7:56:03,  1.73s/it]
 14%|█▎        | 2606/19090 [1:16:04<8:11:15,  1.79s/it]
 14%|█▎        | 2607/19090 [1:16:06<8:37:22,  1.88s/it]
 14%|█▎        | 2608/19090 [1:16:07<7:48:02,  1.70s/it]
 14%|█▎        | 2609/19090 [1:16:09<8:26:57,  1.85s/it]
 14%|█▎        | 2610/19090 [1:16:11<8:36:57,  1.88s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.46596962213516235, 'learning_rate': 4.979548933838511e-05, 'recon_loss': 0.003948708431562409, 'recon_static': 0.0029859097790904342, 'recon_motion': 0.0009627985506085679, 'recon_norm': 2.172336620092392, 'recon_norm_std': 0.07911383621394634, 'target_norm': 1.6213117480278014, 'recon_cos': 0.41881516203284264, 'gradnorm/compressor.query': 0.006941522500406259, 'gradnorm/compressor.layer0': 0.1265910413725848, 'gradnorm/compressor.layer1': 0.17005096114364263, 'gradnorm/compressor.layer2': 0.14418997940499656, 'gradnorm/compressor.layer3': 0.13351954161433374, 'gradnorm/compressor.layer4': 0.1244051945099449, 'gradnorm/compressor.layer5': 0.1178443935339287, 'gradnorm/compressor.layer6': 0.11387608822146623, 'gradnorm/compressor.layer7': 0.10985259957838309, 'gradnorm/decoder.layer0': 0.13915438054463333, 'gradnorm/decoder.layer1': 0.1372257572848669, 'gradnorm/decoder.layer2': 0.20229800307357912, 'gradnorm/max_param': 0.14563041925430298, 'epoch': 1.37}

 14%|█▎        | 2610/19090 [1:16:11<8:36:57,  1.88s/it]
 14%|█▎        | 2611/19090 [1:16:12<7:48:42,  1.71s/it]
 14%|█▎        | 2612/19090 [1:16:15<8:24:33,  1.84s/it]
 14%|█▎        | 2613/19090 [1:16:16<8:12:23,  1.79s/it]
 14%|█▎        | 2614/19090 [1:16:18<7:31:43,  1.65s/it]
 14%|█▎        | 2615/19090 [1:16:20<8:45:19,  1.91s/it]
 14%|█▎        | 2616/19090 [1:16:21<7:55:49,  1.73s/it]
 14%|█▎        | 2617/19090 [1:16:23<7:19:10,  1.60s/it]
 14%|█▎        | 2618/19090 [1:16:26<9:00:26,  1.97s/it]
 14%|█▎        | 2619/19090 [1:16:27<8:04:34,  1.77s/it]
 14%|█▎        | 2620/19090 [1:16:28<7:25:46,  1.62s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.4885280728340149, 'learning_rate': 4.9789612707538494e-05, 'recon_loss': 0.003852641553385183, 'recon_static': 0.0028819032711908223, 'recon_motion': 0.0009707382938358933, 'recon_norm': 2.139084976911545, 'recon_norm_std': 0.0818179041147232, 'target_norm': 1.6224077731370925, 'recon_cos': 0.4201086454093456, 'gradnorm/compressor.query': 0.007955008099317071, 'gradnorm/compressor.layer0': 0.13977111416481125, 'gradnorm/compressor.layer1': 0.18208142600017121, 'gradnorm/compressor.layer2': 0.15114078818164237, 'gradnorm/compressor.layer3': 0.13996645589431667, 'gradnorm/compressor.layer4': 0.12764966325564184, 'gradnorm/compressor.layer5': 0.12272150678819092, 'gradnorm/compressor.layer6': 0.1160353488082585, 'gradnorm/compressor.layer7': 0.11273601051645746, 'gradnorm/decoder.layer0': 0.14527618610749307, 'gradnorm/decoder.layer1': 0.14188996902976228, 'gradnorm/decoder.layer2': 0.21143815002206176, 'gradnorm/max_param': 0.15147560834884644, 'epoch': 1.37}

 14%|█▎        | 2620/19090 [1:16:28<7:25:46,  1.62s/it]
 14%|█▎        | 2621/19090 [1:16:31<8:58:47,  1.96s/it]
 14%|█▎        | 2622/19090 [1:16:32<8:04:02,  1.76s/it]
 14%|█▎        | 2623/19090 [1:16:33<7:25:09,  1.62s/it]
 14%|█▎        | 2624/19090 [1:16:36<8:55:58,  1.95s/it]
 14%|█▍        | 2625/19090 [1:16:38<8:10:08,  1.79s/it]
 14%|█▍        | 2626/19090 [1:16:39<7:30:11,  1.64s/it]
 14%|█▍        | 2627/19090 [1:16:42<9:14:36,  2.02s/it]
 14%|█▍        | 2628/19090 [1:16:43<8:18:52,  1.82s/it]
 14%|█▍        | 2629/19090 [1:16:44<7:36:19,  1.66s/it]
 14%|█▍        | 2630/19090 [1:16:47<8:52:33,  1.94s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.36972030997276306, 'learning_rate': 4.978365319235056e-05, 'recon_loss': 0.0033807219355367125, 'recon_static': 0.0024153641832526775, 'recon_motion': 0.0009653578163124621, 'recon_norm': 2.013632667064667, 'recon_norm_std': 0.08358337450772524, 'target_norm': 1.6191767543554305, 'recon_cos': 0.45037791877985, 'gradnorm/compressor.query': 0.005816999326889257, 'gradnorm/compressor.layer0': 0.10309725464035252, 'gradnorm/compressor.layer1': 0.13400797093983124, 'gradnorm/compressor.layer2': 0.11352418878390841, 'gradnorm/compressor.layer3': 0.106355896412473, 'gradnorm/compressor.layer4': 0.09809911445697742, 'gradnorm/compressor.layer5': 0.09439077540731232, 'gradnorm/compressor.layer6': 0.09042326984743622, 'gradnorm/compressor.layer7': 0.08628873142025223, 'gradnorm/decoder.layer0': 0.11097541366057696, 'gradnorm/decoder.layer1': 0.10994996359967865, 'gradnorm/decoder.layer2': 0.159100823085761, 'gradnorm/max_param': 0.11537452787160873, 'epoch': 1.38}

 14%|█▍        | 2630/19090 [1:16:47<8:52:33,  1.94s/it]
 14%|█▍        | 2631/19090 [1:16:48<8:06:46,  1.77s/it]
 14%|█▍        | 2632/19090 [1:16:50<7:27:46,  1.63s/it]
 14%|█▍        | 2633/19090 [1:16:52<8:29:57,  1.86s/it]
 14%|█▍        | 2634/19090 [1:16:54<8:49:33,  1.93s/it]
 14%|█▍        | 2635/19090 [1:16:56<7:57:50,  1.74s/it]
 14%|█▍        | 2636/19090 [1:16:57<8:15:21,  1.81s/it]
 14%|█▍        | 2637/19090 [1:17:00<8:35:58,  1.88s/it]
 14%|█▍        | 2638/19090 [1:17:01<7:47:07,  1.70s/it]
 14%|█▍        | 2639/19090 [1:17:03<8:28:07,  1.85s/it]
 14%|█▍        | 2640/19090 [1:17:05<8:37:28,  1.89s/it]
                                                        
{'loss': 0.0027, 'grad_norm': 0.3130733072757721, 'learning_rate': 4.977761081274701e-05, 'recon_loss': 0.0027175594004802406, 'recon_static': 0.001764227481908165, 'recon_motion': 0.0009533319884212687, 'recon_norm': 1.8140682399272918, 'recon_norm_std': 0.0942938121035695, 'target_norm': 1.6182871043682099, 'recon_cos': 0.49764613956213, 'gradnorm/compressor.query': 0.004953027630848381, 'gradnorm/compressor.layer0': 0.0872976859269953, 'gradnorm/compressor.layer1': 0.11597801957286372, 'gradnorm/compressor.layer2': 0.09627486253844734, 'gradnorm/compressor.layer3': 0.09177492264766701, 'gradnorm/compressor.layer4': 0.08340553481301678, 'gradnorm/compressor.layer5': 0.07903364291458197, 'gradnorm/compressor.layer6': 0.07649315727611873, 'gradnorm/compressor.layer7': 0.07338488982119577, 'gradnorm/decoder.layer0': 0.09275285592913289, 'gradnorm/decoder.layer1': 0.09057640836535732, 'gradnorm/decoder.layer2': 0.13409161293817837, 'gradnorm/max_param': 0.09768963605165482, 'epoch': 1.38}

 14%|█▍        | 2640/19090 [1:17:05<8:37:28,  1.89s/it]
 14%|█▍        | 2641/19090 [1:17:06<7:48:53,  1.71s/it]
 14%|█▍        | 2642/19090 [1:17:08<8:19:56,  1.82s/it]
 14%|█▍        | 2643/19090 [1:17:10<8:19:28,  1.82s/it]
 14%|█▍        | 2644/19090 [1:17:11<7:35:40,  1.66s/it]
 14%|█▍        | 2645/19090 [1:17:14<8:14:10,  1.80s/it]
 14%|█▍        | 2646/19090 [1:17:16<8:28:55,  1.86s/it]
 14%|█▍        | 2647/19090 [1:17:17<7:42:16,  1.69s/it]
 14%|█▍        | 2648/19090 [1:17:19<8:15:17,  1.81s/it]
 14%|█▍        | 2649/19090 [1:17:21<8:30:32,  1.86s/it]
 14%|█▍        | 2650/19090 [1:17:22<7:43:18,  1.69s/it]
                                                        
{'loss': 0.0021, 'grad_norm': 0.2310188263654709, 'learning_rate': 4.977148558893061e-05, 'recon_loss': 0.0020653414307162167, 'recon_static': 0.0011167645890964196, 'recon_motion': 0.0009485768416197971, 'recon_norm': 1.6008649796247483, 'recon_norm_std': 0.1029030667617917, 'target_norm': 1.6172097265720367, 'recon_cos': 0.5678575649857521, 'gradnorm/compressor.query': 0.0035105825458030785, 'gradnorm/compressor.layer0': 0.06218341838858192, 'gradnorm/compressor.layer1': 0.08438195060870438, 'gradnorm/compressor.layer2': 0.07075697760580621, 'gradnorm/compressor.layer3': 0.06633455950705544, 'gradnorm/compressor.layer4': 0.0609718929490806, 'gradnorm/compressor.layer5': 0.058923779318628494, 'gradnorm/compressor.layer6': 0.05631064615998038, 'gradnorm/compressor.layer7': 0.054202144131454875, 'gradnorm/decoder.layer0': 0.06982849348094358, 'gradnorm/decoder.layer1': 0.06892485720392579, 'gradnorm/decoder.layer2': 0.10019211382900305, 'gradnorm/max_param': 0.07240794599056244, 'epoch': 1.39}

 14%|█▍        | 2650/19090 [1:17:22<7:43:18,  1.69s/it]
 14%|█▍        | 2651/19090 [1:17:25<8:49:46,  1.93s/it]
 14%|█▍        | 2652/19090 [1:17:26<8:19:48,  1.82s/it]
 14%|█▍        | 2653/19090 [1:17:28<7:35:42,  1.66s/it]
 14%|█▍        | 2654/19090 [1:17:30<8:25:26,  1.85s/it]
 14%|█▍        | 2655/19090 [1:17:31<8:04:52,  1.77s/it]
 14%|█▍        | 2656/19090 [1:17:33<7:25:08,  1.63s/it]
 14%|█▍        | 2657/19090 [1:17:35<8:28:28,  1.86s/it]
 14%|█▍        | 2658/19090 [1:17:37<8:09:44,  1.79s/it]
 14%|█▍        | 2659/19090 [1:17:38<7:29:22,  1.64s/it]
 14%|█▍        | 2660/19090 [1:17:41<9:01:21,  1.98s/it]
                                                        
{'loss': 0.002, 'grad_norm': 0.26225244998931885, 'learning_rate': 4.9765277541381115e-05, 'recon_loss': 0.002030805425602011, 'recon_static': 0.0010638084611855447, 'recon_motion': 0.0009669970459071919, 'recon_norm': 1.57820525765419, 'recon_norm_std': 0.10089086387306452, 'target_norm': 1.6216539323329926, 'recon_cos': 0.571281698346138, 'gradnorm/compressor.query': 0.003938891518020163, 'gradnorm/compressor.layer0': 0.0734501936777384, 'gradnorm/compressor.layer1': 0.09651879206416084, 'gradnorm/compressor.layer2': 0.08165927974994201, 'gradnorm/compressor.layer3': 0.07563706782994851, 'gradnorm/compressor.layer4': 0.07101926963149074, 'gradnorm/compressor.layer5': 0.06661284440656295, 'gradnorm/compressor.layer6': 0.06363603293979127, 'gradnorm/compressor.layer7': 0.061717687971900706, 'gradnorm/decoder.layer0': 0.07793679216930716, 'gradnorm/decoder.layer1': 0.07653680523495986, 'gradnorm/decoder.layer2': 0.11130666525957844, 'gradnorm/max_param': 0.08108919858932495, 'epoch': 1.39}

 14%|█▍        | 2660/19090 [1:17:41<9:01:21,  1.98s/it]
 14%|█▍        | 2661/19090 [1:17:42<8:08:37,  1.78s/it]
 14%|█▍        | 2662/19090 [1:17:43<7:28:33,  1.64s/it]
 14%|█▍        | 2663/19090 [1:17:47<9:26:44,  2.07s/it]
 14%|█▍        | 2664/19090 [1:17:48<8:23:33,  1.84s/it]
 14%|█▍        | 2665/19090 [1:17:49<7:39:31,  1.68s/it]
 14%|█▍        | 2666/19090 [1:17:52<9:10:11,  2.01s/it]
 14%|█▍        | 2667/19090 [1:17:53<8:25:49,  1.85s/it]
 14%|█▍        | 2668/19090 [1:17:55<7:39:47,  1.68s/it]
 14%|█▍        | 2669/19090 [1:17:57<8:26:08,  1.85s/it]
 14%|█▍        | 2670/19090 [1:17:58<7:57:11,  1.74s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.5273697376251221, 'learning_rate': 4.9758986690855194e-05, 'recon_loss': 0.003017607011133805, 'recon_static': 0.002064325334504247, 'recon_motion': 0.0009532816897262819, 'recon_norm': 1.9030591666698455, 'recon_norm_std': 0.09083296731114388, 'target_norm': 1.6200082063674928, 'recon_cos': 0.48131985813379285, 'gradnorm/compressor.query': 0.008222146134512059, 'gradnorm/compressor.layer0': 0.14460260331071284, 'gradnorm/compressor.layer1': 0.19538388854947736, 'gradnorm/compressor.layer2': 0.16451877203427714, 'gradnorm/compressor.layer3': 0.15383922332475064, 'gradnorm/compressor.layer4': 0.13979358732417257, 'gradnorm/compressor.layer5': 0.1313832458155046, 'gradnorm/compressor.layer6': 0.12703317172718895, 'gradnorm/compressor.layer7': 0.12396976856638306, 'gradnorm/decoder.layer0': 0.15925479942205487, 'gradnorm/decoder.layer1': 0.1552135423985178, 'gradnorm/decoder.layer2': 0.22465627429435112, 'gradnorm/max_param': 0.16154906153678894, 'epoch': 1.4}

 14%|█▍        | 2670/19090 [1:17:58<7:57:11,  1.74s/it]
 14%|█▍        | 2671/19090 [1:18:00<7:19:40,  1.61s/it]
 14%|█▍        | 2672/19090 [1:18:02<8:16:05,  1.81s/it]
 14%|█▍        | 2673/19090 [1:18:04<8:14:23,  1.81s/it]
 14%|█▍        | 2674/19090 [1:18:05<7:31:43,  1.65s/it]
 14%|█▍        | 2675/19090 [1:18:07<7:55:50,  1.74s/it]
 14%|█▍        | 2676/19090 [1:18:10<9:02:33,  1.98s/it]
 14%|█▍        | 2677/19090 [1:18:11<8:05:53,  1.78s/it]
 14%|█▍        | 2678/19090 [1:18:13<7:56:49,  1.74s/it]
 14%|█▍        | 2679/19090 [1:18:15<9:00:21,  1.98s/it]
 14%|█▍        | 2680/19090 [1:18:16<8:06:41,  1.78s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.4916269779205322, 'learning_rate': 4.9752613058386385e-05, 'recon_loss': 0.004259147768607363, 'recon_static': 0.0032916512223891913, 'recon_motion': 0.0009674965273006819, 'recon_norm': 2.2444376826286314, 'recon_norm_std': 0.07631185371428728, 'target_norm': 1.6209616094827652, 'recon_cos': 0.3993169590830803, 'gradnorm/compressor.query': 0.007569182739539553, 'gradnorm/compressor.layer0': 0.13189444479931722, 'gradnorm/compressor.layer1': 0.177990916883649, 'gradnorm/compressor.layer2': 0.15462325380321518, 'gradnorm/compressor.layer3': 0.14186797238056173, 'gradnorm/compressor.layer4': 0.13206015699278692, 'gradnorm/compressor.layer5': 0.12546128602350348, 'gradnorm/compressor.layer6': 0.11936140319398399, 'gradnorm/compressor.layer7': 0.11588040396389397, 'gradnorm/decoder.layer0': 0.1481385797500852, 'gradnorm/decoder.layer1': 0.1456909115762953, 'gradnorm/decoder.layer2': 0.21095327918762807, 'gradnorm/max_param': 0.152659073472023, 'epoch': 1.4}

 14%|█▍        | 2680/19090 [1:18:16<8:06:41,  1.78s/it]
 14%|█▍        | 2681/19090 [1:18:18<7:48:34,  1.71s/it]
 14%|█▍        | 2682/19090 [1:18:20<8:40:09,  1.90s/it]
 14%|█▍        | 2683/19090 [1:18:22<7:50:18,  1.72s/it]
 14%|█▍        | 2684/19090 [1:18:23<7:16:08,  1.60s/it]
 14%|█▍        | 2685/19090 [1:18:26<9:31:51,  2.09s/it]
 14%|█▍        | 2686/19090 [1:18:27<8:26:14,  1.85s/it]
 14%|█▍        | 2687/19090 [1:18:29<7:39:58,  1.68s/it]
 14%|█▍        | 2688/19090 [1:18:31<8:47:33,  1.93s/it]
 14%|█▍        | 2689/19090 [1:18:33<7:55:07,  1.74s/it]
 14%|█▍        | 2690/19090 [1:18:34<7:20:16,  1.61s/it]
                                                        
{'loss': 0.004, 'grad_norm': 0.45259109139442444, 'learning_rate': 4.974615666528498e-05, 'recon_loss': 0.004023545538075268, 'recon_static': 0.003052577842026949, 'recon_motion': 0.0009709676538477652, 'recon_norm': 2.1924966394901277, 'recon_norm_std': 0.07996029239147902, 'target_norm': 1.6229683101177215, 'recon_cos': 0.4148867063224316, 'gradnorm/compressor.query': 0.007469792871558685, 'gradnorm/compressor.layer0': 0.12588117374008564, 'gradnorm/compressor.layer1': 0.16678894234047717, 'gradnorm/compressor.layer2': 0.14007883504675894, 'gradnorm/compressor.layer3': 0.13176039898185885, 'gradnorm/compressor.layer4': 0.12127361859636941, 'gradnorm/compressor.layer5': 0.11351696926589061, 'gradnorm/compressor.layer6': 0.10793045767814097, 'gradnorm/compressor.layer7': 0.10555788970950766, 'gradnorm/decoder.layer0': 0.13469784179157165, 'gradnorm/decoder.layer1': 0.13280942155972777, 'gradnorm/decoder.layer2': 0.19485556558239878, 'gradnorm/max_param': 0.14078804850578308, 'epoch': 1.41}

 14%|█▍        | 2690/19090 [1:18:34<7:20:16,  1.61s/it]
 14%|█▍        | 2691/19090 [1:18:36<8:11:44,  1.80s/it]
 14%|█▍        | 2692/19090 [1:18:37<7:29:53,  1.65s/it]
 14%|█▍        | 2693/19090 [1:18:40<8:11:25,  1.80s/it]
 14%|█▍        | 2694/19090 [1:18:42<8:42:07,  1.91s/it]
 14%|█▍        | 2695/19090 [1:18:43<7:50:30,  1.72s/it]
 14%|█▍        | 2696/19090 [1:18:45<8:03:41,  1.77s/it]
 14%|█▍        | 2697/19090 [1:18:47<8:39:34,  1.90s/it]
 14%|█▍        | 2698/19090 [1:18:48<7:48:54,  1.72s/it]
 14%|█▍        | 2699/19090 [1:18:50<7:41:53,  1.69s/it]
 14%|█▍        | 2700/19090 [1:18:52<8:21:47,  1.84s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.4714827239513397, 'learning_rate': 4.9739617533138014e-05, 'recon_loss': 0.0035558060859329998, 'recon_static': 0.0025908519804943354, 'recon_motion': 0.0009649540719692595, 'recon_norm': 2.0637941628694536, 'recon_norm_std': 0.08346284199506045, 'target_norm': 1.6211127460002899, 'recon_cos': 0.4387272834777832, 'gradnorm/compressor.query': 0.007066048655241502, 'gradnorm/compressor.layer0': 0.12360883106545512, 'gradnorm/compressor.layer1': 0.17273681948591454, 'gradnorm/compressor.layer2': 0.14709523940887323, 'gradnorm/compressor.layer3': 0.13704310497473646, 'gradnorm/compressor.layer4': 0.12454133896900511, 'gradnorm/compressor.layer5': 0.11924642805281993, 'gradnorm/compressor.layer6': 0.11775851605911962, 'gradnorm/compressor.layer7': 0.11084401127042648, 'gradnorm/decoder.layer0': 0.14385860088119806, 'gradnorm/decoder.layer1': 0.13803163672097882, 'gradnorm/decoder.layer2': 0.2026803394844283, 'gradnorm/max_param': 0.14612358808517456, 'epoch': 1.41}

 14%|█▍        | 2700/19090 [1:18:52<8:21:47,  1.84s/it]
 14%|█▍        | 2701/19090 [1:18:53<7:36:33,  1.67s/it]
 14%|█▍        | 2702/19090 [1:18:55<7:47:12,  1.71s/it]
 14%|█▍        | 2703/19090 [1:18:58<8:39:10,  1.90s/it]
 14%|█▍        | 2704/19090 [1:18:59<7:49:38,  1.72s/it]
 14%|█▍        | 2705/19090 [1:19:00<7:36:58,  1.67s/it]
 14%|█▍        | 2706/19090 [1:19:03<8:23:14,  1.84s/it]
 14%|█▍        | 2707/19090 [1:19:04<7:38:48,  1.68s/it]
 14%|█▍        | 2708/19090 [1:19:05<7:22:44,  1.62s/it]
 14%|█▍        | 2709/19090 [1:19:08<8:30:27,  1.87s/it]
 14%|█▍        | 2710/19090 [1:19:09<7:42:52,  1.70s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.383281946182251, 'learning_rate': 4.973299568380915e-05, 'recon_loss': 0.0032170079881325363, 'recon_static': 0.0022752772958483545, 'recon_motion': 0.0009417306617251597, 'recon_norm': 1.9745879173278809, 'recon_norm_std': 0.08711787983775139, 'target_norm': 1.6181893408298493, 'recon_cos': 0.46181742697954176, 'gradnorm/compressor.query': 0.005730359713418052, 'gradnorm/compressor.layer0': 0.10363251007012943, 'gradnorm/compressor.layer1': 0.14007810370322224, 'gradnorm/compressor.layer2': 0.11809201842262768, 'gradnorm/compressor.layer3': 0.11093175329179851, 'gradnorm/compressor.layer4': 0.10183785780957014, 'gradnorm/compressor.layer5': 0.0952641823900114, 'gradnorm/compressor.layer6': 0.09355530290660055, 'gradnorm/compressor.layer7': 0.09130994520360207, 'gradnorm/decoder.layer0': 0.1151831601984321, 'gradnorm/decoder.layer1': 0.11448866598309385, 'gradnorm/decoder.layer2': 0.1653720611223189, 'gradnorm/max_param': 0.11920858919620514, 'epoch': 1.42}

 14%|█▍        | 2710/19090 [1:19:09<7:42:52,  1.70s/it]
 14%|█▍        | 2711/19090 [1:19:11<7:16:02,  1.60s/it]
 14%|█▍        | 2712/19090 [1:19:13<8:59:01,  1.97s/it]
 14%|█▍        | 2713/19090 [1:19:15<8:03:45,  1.77s/it]
 14%|█▍        | 2714/19090 [1:19:16<7:38:38,  1.68s/it]
 14%|█▍        | 2715/19090 [1:19:18<8:09:12,  1.79s/it]
 14%|█▍        | 2716/19090 [1:19:20<7:27:45,  1.64s/it]
 14%|█▍        | 2717/19090 [1:19:22<8:04:36,  1.78s/it]
 14%|█▍        | 2718/19090 [1:19:24<8:19:54,  1.83s/it]
 14%|█▍        | 2719/19090 [1:19:25<7:35:09,  1.67s/it]
 14%|█▍        | 2720/19090 [1:19:27<8:19:18,  1.83s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.4817551076412201, 'learning_rate': 4.9726291139438605e-05, 'recon_loss': 0.0034326837223488836, 'recon_static': 0.0024761680338997393, 'recon_motion': 0.0009565157000906765, 'recon_norm': 2.0292681753635406, 'recon_norm_std': 0.0837893458083272, 'target_norm': 1.6179397612810136, 'recon_cos': 0.4459541693329811, 'gradnorm/compressor.query': 0.007103039805670215, 'gradnorm/compressor.layer0': 0.12585000031055985, 'gradnorm/compressor.layer1': 0.17317837999092675, 'gradnorm/compressor.layer2': 0.14779493892944978, 'gradnorm/compressor.layer3': 0.14001055098693382, 'gradnorm/compressor.layer4': 0.13097608400661898, 'gradnorm/compressor.layer5': 0.12192023779578803, 'gradnorm/compressor.layer6': 0.11839775353954358, 'gradnorm/compressor.layer7': 0.11415210545381216, 'gradnorm/decoder.layer0': 0.1442473884553547, 'gradnorm/decoder.layer1': 0.14164877137472015, 'gradnorm/decoder.layer2': 0.21171340583728657, 'gradnorm/max_param': 0.15044060349464417, 'epoch': 1.42}

 14%|█▍        | 2720/19090 [1:19:27<8:19:18,  1.83s/it]
 14%|█▍        | 2721/19090 [1:19:29<8:23:53,  1.85s/it]
 14%|█▍        | 2722/19090 [1:19:30<7:38:05,  1.68s/it]
 14%|█▍        | 2723/19090 [1:19:33<8:34:31,  1.89s/it]
 14%|█▍        | 2724/19090 [1:19:34<7:51:41,  1.73s/it]
 14%|█▍        | 2725/19090 [1:19:35<7:15:19,  1.60s/it]
 14%|█▍        | 2726/19090 [1:19:38<8:53:36,  1.96s/it]
 14%|█▍        | 2727/19090 [1:19:39<8:06:29,  1.78s/it]
 14%|█▍        | 2728/19090 [1:19:41<7:25:39,  1.63s/it]
 14%|█▍        | 2729/19090 [1:19:43<8:39:40,  1.91s/it]
 14%|█▍        | 2730/19090 [1:19:45<8:26:04,  1.86s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.3994590938091278, 'learning_rate': 4.97195039224431e-05, 'recon_loss': 0.003372471925104037, 'recon_static': 0.0024157324398402123, 'recon_motion': 0.0009567395274643787, 'recon_norm': 2.0125511944293977, 'recon_norm_std': 0.08350921515375376, 'target_norm': 1.6193092525005341, 'recon_cos': 0.44958532974123955, 'gradnorm/compressor.query': 0.005953832675369185, 'gradnorm/compressor.layer0': 0.1065164293474535, 'gradnorm/compressor.layer1': 0.14320949650568543, 'gradnorm/compressor.layer2': 0.12415311446840477, 'gradnorm/compressor.layer3': 0.11514786123598336, 'gradnorm/compressor.layer4': 0.10900150971067087, 'gradnorm/compressor.layer5': 0.10009879600062103, 'gradnorm/compressor.layer6': 0.09533286971380191, 'gradnorm/compressor.layer7': 0.09543395628967039, 'gradnorm/decoder.layer0': 0.12107305196023564, 'gradnorm/decoder.layer1': 0.11796153264356977, 'gradnorm/decoder.layer2': 0.17416564448488636, 'gradnorm/max_param': 0.12512120604515076, 'epoch': 1.43}

 14%|█▍        | 2730/19090 [1:19:45<8:26:04,  1.86s/it]
 14%|█▍        | 2731/19090 [1:19:46<7:39:27,  1.69s/it]
 14%|█▍        | 2732/19090 [1:19:49<8:20:57,  1.84s/it]
 14%|█▍        | 2733/19090 [1:19:50<8:07:05,  1.79s/it]
 14%|█▍        | 2734/19090 [1:19:51<7:26:31,  1.64s/it]
 14%|█▍        | 2735/19090 [1:19:54<8:23:26,  1.85s/it]
 14%|█▍        | 2736/19090 [1:19:55<8:02:58,  1.77s/it]
 14%|█▍        | 2737/19090 [1:19:57<7:23:00,  1.63s/it]
 14%|█▍        | 2738/19090 [1:19:59<8:36:03,  1.89s/it]
 14%|█▍        | 2739/19090 [1:20:01<8:23:05,  1.85s/it]
 14%|█▍        | 2740/19090 [1:20:02<7:37:58,  1.68s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.42973965406417847, 'learning_rate': 4.971263405551576e-05, 'recon_loss': 0.0030282717489171773, 'recon_static': 0.0020603054435923696, 'recon_motion': 0.0009679662951384671, 'recon_norm': 1.9096900343894958, 'recon_norm_std': 0.08648196328431368, 'target_norm': 1.6217268109321594, 'recon_cos': 0.47298360615968704, 'gradnorm/compressor.query': 0.006726540990217306, 'gradnorm/compressor.layer0': 0.11830764511901444, 'gradnorm/compressor.layer1': 0.1548304506570509, 'gradnorm/compressor.layer2': 0.13308257803834975, 'gradnorm/compressor.layer3': 0.12405347084992556, 'gradnorm/compressor.layer4': 0.11516124619394867, 'gradnorm/compressor.layer5': 0.10864273242725173, 'gradnorm/compressor.layer6': 0.10515346467437339, 'gradnorm/compressor.layer7': 0.10299512169451913, 'gradnorm/decoder.layer0': 0.1315489081590268, 'gradnorm/decoder.layer1': 0.12472398653127116, 'gradnorm/decoder.layer2': 0.1843087497587065, 'gradnorm/max_param': 0.13401375710964203, 'epoch': 1.44}

 14%|█▍        | 2740/19090 [1:20:02<7:37:58,  1.68s/it]
 14%|█▍        | 2741/19090 [1:20:04<8:12:45,  1.81s/it]
 14%|█▍        | 2742/19090 [1:20:06<8:17:38,  1.83s/it]
 14%|█▍        | 2743/19090 [1:20:08<7:34:27,  1.67s/it]
 14%|█▍        | 2744/19090 [1:20:09<7:58:17,  1.76s/it]
 14%|█▍        | 2745/19090 [1:20:12<8:26:59,  1.86s/it]
 14%|█▍        | 2746/19090 [1:20:13<7:40:10,  1.69s/it]
 14%|█▍        | 2747/19090 [1:20:15<7:43:16,  1.70s/it]
 14%|█▍        | 2748/19090 [1:20:17<8:32:59,  1.88s/it]
 14%|█▍        | 2749/19090 [1:20:18<7:44:21,  1.71s/it]
 14%|█▍        | 2750/19090 [1:20:20<7:55:33,  1.75s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.37387484312057495, 'learning_rate': 4.970568156162607e-05, 'recon_loss': 0.0033215027302503585, 'recon_static': 0.0023616954014869405, 'recon_motion': 0.0009598073578672483, 'recon_norm': 1.9977536648511887, 'recon_norm_std': 0.08553659673780203, 'target_norm': 1.6204240322113037, 'recon_cos': 0.45320878475904464, 'gradnorm/compressor.query': 0.005798889748949156, 'gradnorm/compressor.layer0': 0.09891575652352286, 'gradnorm/compressor.layer1': 0.13692435744212997, 'gradnorm/compressor.layer2': 0.1164505275900884, 'gradnorm/compressor.layer3': 0.10903894791115687, 'gradnorm/compressor.layer4': 0.10031061082320712, 'gradnorm/compressor.layer5': 0.09427064132495246, 'gradnorm/compressor.layer6': 0.09091698720793479, 'gradnorm/compressor.layer7': 0.08759139303315512, 'gradnorm/decoder.layer0': 0.11370365953627763, 'gradnorm/decoder.layer1': 0.10997630769082649, 'gradnorm/decoder.layer2': 0.16079593469049935, 'gradnorm/max_param': 0.11671614646911621, 'epoch': 1.44}

 14%|█▍        | 2750/19090 [1:20:20<7:55:33,  1.75s/it]
 14%|█▍        | 2751/19090 [1:20:22<8:32:32,  1.88s/it]
 14%|█▍        | 2752/19090 [1:20:24<7:45:24,  1.71s/it]
 14%|█▍        | 2753/19090 [1:20:26<8:11:46,  1.81s/it]
 14%|█▍        | 2754/19090 [1:20:27<8:04:30,  1.78s/it]
 14%|█▍        | 2755/19090 [1:20:29<7:26:43,  1.64s/it]
 14%|█▍        | 2756/19090 [1:20:31<8:12:01,  1.81s/it]
 14%|█▍        | 2757/19090 [1:20:33<8:36:32,  1.90s/it]
 14%|█▍        | 2758/19090 [1:20:34<7:47:21,  1.72s/it]
 14%|█▍        | 2759/19090 [1:20:36<8:06:16,  1.79s/it]
 14%|█▍        | 2760/19090 [1:20:38<8:39:02,  1.91s/it]
                                                        
{'loss': 0.0028, 'grad_norm': 0.4015873670578003, 'learning_rate': 4.969864646401977e-05, 'recon_loss': 0.002811048016883433, 'recon_static': 0.0018575724330730735, 'recon_motion': 0.0009534756987704896, 'recon_norm': 1.8496760487556458, 'recon_norm_std': 0.09142798818647861, 'target_norm': 1.6208028942346573, 'recon_cos': 0.4920706756412983, 'gradnorm/compressor.query': 0.006301792423574292, 'gradnorm/compressor.layer0': 0.11163950340262843, 'gradnorm/compressor.layer1': 0.14420154054527454, 'gradnorm/compressor.layer2': 0.12401895064400335, 'gradnorm/compressor.layer3': 0.1163071746803661, 'gradnorm/compressor.layer4': 0.10700129190644328, 'gradnorm/compressor.layer5': 0.10278165367571437, 'gradnorm/compressor.layer6': 0.0981169182568457, 'gradnorm/compressor.layer7': 0.0956431171530037, 'gradnorm/decoder.layer0': 0.12103520023873923, 'gradnorm/decoder.layer1': 0.11749307211276498, 'gradnorm/decoder.layer2': 0.1726995974709967, 'gradnorm/max_param': 0.12471699714660645, 'epoch': 1.45}

 14%|█▍        | 2760/19090 [1:20:38<8:39:02,  1.91s/it]
 14%|█▍        | 2761/19090 [1:20:40<7:51:30,  1.73s/it]
 14%|█▍        | 2762/19090 [1:20:42<8:33:07,  1.89s/it]
 14%|█▍        | 2763/19090 [1:20:44<8:37:33,  1.90s/it]
 14%|█▍        | 2764/19090 [1:20:45<7:48:38,  1.72s/it]
 14%|█▍        | 2765/19090 [1:20:47<8:25:07,  1.86s/it]
 14%|█▍        | 2766/19090 [1:20:49<8:21:01,  1.84s/it]
 14%|█▍        | 2767/19090 [1:20:50<7:36:46,  1.68s/it]
 14%|█▍        | 2768/19090 [1:20:53<8:28:59,  1.87s/it]
 15%|█▍        | 2769/19090 [1:20:55<8:26:37,  1.86s/it]
 15%|█▍        | 2770/19090 [1:20:56<7:40:24,  1.69s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.4393197000026703, 'learning_rate': 4.969152878621879e-05, 'recon_loss': 0.0032019792823120953, 'recon_static': 0.0022518945392221213, 'recon_motion': 0.0009500847430899739, 'recon_norm': 1.969420027732849, 'recon_norm_std': 0.08537324871867895, 'target_norm': 1.617983815073967, 'recon_cos': 0.4622343748807907, 'gradnorm/compressor.query': 0.006504514683832969, 'gradnorm/compressor.layer0': 0.1179843394923924, 'gradnorm/compressor.layer1': 0.15773069815769092, 'gradnorm/compressor.layer2': 0.13488034189471015, 'gradnorm/compressor.layer3': 0.12742700166522167, 'gradnorm/compressor.layer4': 0.11796246821370839, 'gradnorm/compressor.layer5': 0.11260323026810613, 'gradnorm/compressor.layer6': 0.10715436572044919, 'gradnorm/compressor.layer7': 0.10438338970541225, 'gradnorm/decoder.layer0': 0.13204033865486026, 'gradnorm/decoder.layer1': 0.12985800742864864, 'gradnorm/decoder.layer2': 0.19094380157610852, 'gradnorm/max_param': 0.1367582082748413, 'epoch': 1.45}

 15%|█▍        | 2770/19090 [1:20:56<7:40:24,  1.69s/it]
 15%|█▍        | 2771/19090 [1:20:57<7:23:43,  1.63s/it]
 15%|█▍        | 2772/19090 [1:21:00<8:13:54,  1.82s/it]
 15%|█▍        | 2773/19090 [1:21:01<7:31:12,  1.66s/it]
 15%|█▍        | 2774/19090 [1:21:03<7:28:48,  1.65s/it]
 15%|█▍        | 2775/19090 [1:21:04<7:44:33,  1.71s/it]
 15%|█▍        | 2776/19090 [1:21:06<7:11:08,  1.59s/it]
 15%|█▍        | 2777/19090 [1:21:08<7:49:12,  1.73s/it]
 15%|█▍        | 2778/19090 [1:21:09<7:34:07,  1.67s/it]
 15%|█▍        | 2779/19090 [1:21:11<7:02:49,  1.56s/it]
 15%|█▍        | 2780/19090 [1:21:13<8:02:52,  1.78s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.4221343696117401, 'learning_rate': 4.968432855202115e-05, 'recon_loss': 0.0036623194813728333, 'recon_static': 0.002699689124710858, 'recon_motion': 0.0009626303712138906, 'recon_norm': 2.094794863462448, 'recon_norm_std': 0.08051242735236883, 'target_norm': 1.6206682562828063, 'recon_cos': 0.4330377824604511, 'gradnorm/compressor.query': 0.006483879275522011, 'gradnorm/compressor.layer0': 0.11015316759355581, 'gradnorm/compressor.layer1': 0.1498978579965218, 'gradnorm/compressor.layer2': 0.13005908677820724, 'gradnorm/compressor.layer3': 0.1228923167508152, 'gradnorm/compressor.layer4': 0.11342765128255387, 'gradnorm/compressor.layer5': 0.10829469510615623, 'gradnorm/compressor.layer6': 0.10393170192021976, 'gradnorm/compressor.layer7': 0.10087347597645485, 'gradnorm/decoder.layer0': 0.12787602296403272, 'gradnorm/decoder.layer1': 0.12603244724788035, 'gradnorm/decoder.layer2': 0.18364948076148283, 'gradnorm/max_param': 0.13255168497562408, 'epoch': 1.46}

 15%|█▍        | 2780/19090 [1:21:13<8:02:52,  1.78s/it]
 15%|█▍        | 2781/19090 [1:21:14<7:23:13,  1.63s/it]
 15%|█▍        | 2782/19090 [1:21:15<6:55:33,  1.53s/it]
 15%|█▍        | 2783/19090 [1:21:18<8:10:51,  1.81s/it]
 15%|█▍        | 2784/19090 [1:21:19<7:28:38,  1.65s/it]
 15%|█▍        | 2785/19090 [1:21:20<6:58:53,  1.54s/it]
 15%|█▍        | 2786/19090 [1:21:23<8:25:22,  1.86s/it]
 15%|█▍        | 2787/19090 [1:21:24<7:39:22,  1.69s/it]
 15%|█▍        | 2788/19090 [1:21:26<7:06:20,  1.57s/it]
 15%|█▍        | 2789/19090 [1:21:28<8:25:07,  1.86s/it]
 15%|█▍        | 2790/19090 [1:21:29<7:39:35,  1.69s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.3770037889480591, 'learning_rate': 4.967704578550092e-05, 'recon_loss': 0.003147127421107143, 'recon_static': 0.002180825383402407, 'recon_motion': 0.0009663021031883545, 'recon_norm': 1.9482544213533401, 'recon_norm_std': 0.08636425938457251, 'target_norm': 1.622222077846527, 'recon_cos': 0.4655976496636868, 'gradnorm/compressor.query': 0.005687757756697132, 'gradnorm/compressor.layer0': 0.10081215761555794, 'gradnorm/compressor.layer1': 0.13300344821306906, 'gradnorm/compressor.layer2': 0.11626575283125022, 'gradnorm/compressor.layer3': 0.10942557136478204, 'gradnorm/compressor.layer4': 0.1019811301298887, 'gradnorm/compressor.layer5': 0.09543650330708196, 'gradnorm/compressor.layer6': 0.09354922033752928, 'gradnorm/compressor.layer7': 0.08837319192274384, 'gradnorm/decoder.layer0': 0.11492351856262283, 'gradnorm/decoder.layer1': 0.11192601674330409, 'gradnorm/decoder.layer2': 0.16416583893860956, 'gradnorm/max_param': 0.11841832846403122, 'epoch': 1.46}

 15%|█▍        | 2790/19090 [1:21:29<7:39:35,  1.69s/it]
 15%|█▍        | 2791/19090 [1:21:31<7:07:00,  1.57s/it]
 15%|█▍        | 2792/19090 [1:21:33<8:07:13,  1.79s/it]
 15%|█▍        | 2793/19090 [1:21:35<8:07:49,  1.80s/it]
 15%|█▍        | 2794/19090 [1:21:36<7:26:31,  1.64s/it]
 15%|█▍        | 2795/19090 [1:21:38<8:21:26,  1.85s/it]
 15%|█▍        | 2796/19090 [1:21:40<7:56:24,  1.75s/it]
 15%|█▍        | 2797/19090 [1:21:41<7:18:36,  1.62s/it]
 15%|█▍        | 2798/19090 [1:21:44<8:17:03,  1.83s/it]
 15%|█▍        | 2799/19090 [1:21:45<7:49:50,  1.73s/it]
 15%|█▍        | 2800/19090 [1:21:46<7:14:04,  1.60s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.4445277452468872, 'learning_rate': 4.96696805110081e-05, 'recon_loss': 0.003342047578189522, 'recon_static': 0.0023797232686774807, 'recon_motion': 0.0009623242294765078, 'recon_norm': 2.003292641043663, 'recon_norm_std': 0.08613206725567579, 'target_norm': 1.621708470582962, 'recon_cos': 0.45197202637791634, 'gradnorm/compressor.query': 0.006451222661350017, 'gradnorm/compressor.layer0': 0.1181216991558788, 'gradnorm/compressor.layer1': 0.15896578616818935, 'gradnorm/compressor.layer2': 0.13631694195850325, 'gradnorm/compressor.layer3': 0.12947348579006335, 'gradnorm/compressor.layer4': 0.11807375601837687, 'gradnorm/compressor.layer5': 0.11171939756299124, 'gradnorm/compressor.layer6': 0.10845758783165696, 'gradnorm/compressor.layer7': 0.10552339551036177, 'gradnorm/decoder.layer0': 0.13620266921271387, 'gradnorm/decoder.layer1': 0.13081706487543435, 'gradnorm/decoder.layer2': 0.1949604676595815, 'gradnorm/max_param': 0.1396206170320511, 'epoch': 1.47}

 15%|█▍        | 2800/19090 [1:21:46<7:14:04,  1.60s/it]
 15%|█▍        | 2801/19090 [1:21:49<8:28:01,  1.87s/it]
 15%|█▍        | 2802/19090 [1:21:51<8:12:24,  1.81s/it]
 15%|█▍        | 2803/19090 [1:21:52<7:29:49,  1.66s/it]
 15%|█▍        | 2804/19090 [1:21:54<8:26:09,  1.86s/it]
 15%|█▍        | 2805/19090 [1:21:56<8:05:42,  1.79s/it]
 15%|█▍        | 2806/19090 [1:21:57<7:25:08,  1.64s/it]
 15%|█▍        | 2807/19090 [1:21:59<8:18:37,  1.84s/it]
 15%|█▍        | 2808/19090 [1:22:01<8:10:22,  1.81s/it]
 15%|█▍        | 2809/19090 [1:22:03<7:28:25,  1.65s/it]
 15%|█▍        | 2810/19090 [1:22:04<7:47:12,  1.72s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.44071194529533386, 'learning_rate': 4.966223275316859e-05, 'recon_loss': 0.003703586134361103, 'recon_static': 0.002736952033592388, 'recon_motion': 0.0009666341007687152, 'recon_norm': 2.1067036509513857, 'recon_norm_std': 0.08164403736591339, 'target_norm': 1.6219843626022339, 'recon_cos': 0.431201484054327, 'gradnorm/compressor.query': 0.006551562921762908, 'gradnorm/compressor.layer0': 0.11926653337501286, 'gradnorm/compressor.layer1': 0.16054638006315025, 'gradnorm/compressor.layer2': 0.13298933094028895, 'gradnorm/compressor.layer3': 0.1273904603582521, 'gradnorm/compressor.layer4': 0.117260342052634, 'gradnorm/compressor.layer5': 0.1129123089434518, 'gradnorm/compressor.layer6': 0.10741339996891566, 'gradnorm/compressor.layer7': 0.10535632870578356, 'gradnorm/decoder.layer0': 0.1323312919387279, 'gradnorm/decoder.layer1': 0.1306579805107632, 'gradnorm/decoder.layer2': 0.1911807483535945, 'gradnorm/max_param': 0.13840538263320923, 'epoch': 1.47}

 15%|█▍        | 2810/19090 [1:22:04<7:47:12,  1.72s/it]
 15%|█▍        | 2811/19090 [1:22:07<8:41:42,  1.92s/it]
 15%|█▍        | 2812/19090 [1:22:08<7:49:51,  1.73s/it]
 15%|█▍        | 2813/19090 [1:22:10<7:46:09,  1.72s/it]
 15%|█▍        | 2814/19090 [1:22:12<9:04:29,  2.01s/it]
 15%|█▍        | 2815/19090 [1:22:14<8:06:05,  1.79s/it]
 15%|█▍        | 2816/19090 [1:22:15<7:26:44,  1.65s/it]
 15%|█▍        | 2817/19090 [1:22:17<8:31:59,  1.89s/it]
 15%|█▍        | 2818/19090 [1:22:19<7:43:10,  1.71s/it]
 15%|█▍        | 2819/19090 [1:22:21<7:57:44,  1.76s/it]
 15%|█▍        | 2820/19090 [1:22:23<8:33:57,  1.90s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.4437226355075836, 'learning_rate': 4.9654702536884024e-05, 'recon_loss': 0.003262265957891941, 'recon_static': 0.0022978824039455504, 'recon_motion': 0.000964383635437116, 'recon_norm': 1.9836974829435348, 'recon_norm_std': 0.08420932702720166, 'target_norm': 1.6206244707107544, 'recon_cos': 0.4575493671000004, 'gradnorm/compressor.query': 0.0068044081714186055, 'gradnorm/compressor.layer0': 0.11962787082768982, 'gradnorm/compressor.layer1': 0.16047716553265862, 'gradnorm/compressor.layer2': 0.13905761019007065, 'gradnorm/compressor.layer3': 0.128151439346237, 'gradnorm/compressor.layer4': 0.1186707486293194, 'gradnorm/compressor.layer5': 0.11134880275433824, 'gradnorm/compressor.layer6': 0.1083967363378707, 'gradnorm/compressor.layer7': 0.10526761107873601, 'gradnorm/decoder.layer0': 0.1331690182467279, 'gradnorm/decoder.layer1': 0.13174774610991075, 'gradnorm/decoder.layer2': 0.19135320668056724, 'gradnorm/max_param': 0.1389966756105423, 'epoch': 1.48}

 15%|█▍        | 2820/19090 [1:22:23<8:33:57,  1.90s/it]
 15%|█▍        | 2821/19090 [1:22:24<7:44:30,  1.71s/it]
 15%|█▍        | 2822/19090 [1:22:26<7:32:27,  1.67s/it]
 15%|█▍        | 2823/19090 [1:22:28<8:47:10,  1.94s/it]
 15%|█▍        | 2824/19090 [1:22:30<7:53:22,  1.75s/it]
 15%|█▍        | 2825/19090 [1:22:31<7:49:08,  1.73s/it]
 15%|█▍        | 2826/19090 [1:22:34<8:40:38,  1.92s/it]
 15%|█▍        | 2827/19090 [1:22:35<7:49:01,  1.73s/it]
 15%|█▍        | 2828/19090 [1:22:37<7:43:56,  1.71s/it]
 15%|█▍        | 2829/19090 [1:22:39<8:06:25,  1.79s/it]
 15%|█▍        | 2830/19090 [1:22:40<7:25:00,  1.64s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.40345433354377747, 'learning_rate': 4.964708988733178e-05, 'recon_loss': 0.0032293730066157876, 'recon_static': 0.002276281302329153, 'recon_motion': 0.0009530917261145078, 'recon_norm': 1.9758481174707412, 'recon_norm_std': 0.08713300917297602, 'target_norm': 1.618857192993164, 'recon_cos': 0.4600654311478138, 'gradnorm/compressor.query': 0.005929061563437606, 'gradnorm/compressor.layer0': 0.10697841140563423, 'gradnorm/compressor.layer1': 0.14327696450625096, 'gradnorm/compressor.layer2': 0.12422349518883435, 'gradnorm/compressor.layer3': 0.11479440443148671, 'gradnorm/compressor.layer4': 0.10722426531889756, 'gradnorm/compressor.layer5': 0.1026951102004121, 'gradnorm/compressor.layer6': 0.09918372050255418, 'gradnorm/compressor.layer7': 0.09648084904332317, 'gradnorm/decoder.layer0': 0.12195519932480166, 'gradnorm/decoder.layer1': 0.11965490175520244, 'gradnorm/decoder.layer2': 0.17826577279045216, 'gradnorm/max_param': 0.12603238224983215, 'epoch': 1.48}

 15%|█▍        | 2830/19090 [1:22:40<7:25:00,  1.64s/it]
 15%|█▍        | 2831/19090 [1:22:42<7:59:00,  1.77s/it]
 15%|█▍        | 2832/19090 [1:22:44<8:18:03,  1.84s/it]
 15%|█▍        | 2833/19090 [1:22:45<7:32:45,  1.67s/it]
 15%|█▍        | 2834/19090 [1:22:47<8:03:49,  1.79s/it]
 15%|█▍        | 2835/19090 [1:22:50<8:53:26,  1.97s/it]
 15%|█▍        | 2836/19090 [1:22:51<7:58:43,  1.77s/it]
 15%|█▍        | 2837/19090 [1:22:53<7:47:20,  1.73s/it]
 15%|█▍        | 2838/19090 [1:22:55<8:59:43,  1.99s/it]
 15%|█▍        | 2839/19090 [1:22:57<8:03:10,  1.78s/it]
 15%|█▍        | 2840/19090 [1:22:58<7:26:50,  1.65s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.4401213526725769, 'learning_rate': 4.963939482996482e-05, 'recon_loss': 0.003616833116393536, 'recon_static': 0.002654177520889789, 'recon_motion': 0.0009626555518480017, 'recon_norm': 2.0831920742988586, 'recon_norm_std': 0.08050900176167489, 'target_norm': 1.6205236464738846, 'recon_cos': 0.43620295226573946, 'gradnorm/compressor.query': 0.006448587485049926, 'gradnorm/compressor.layer0': 0.11441994436714518, 'gradnorm/compressor.layer1': 0.15519826496886666, 'gradnorm/compressor.layer2': 0.13696752802159892, 'gradnorm/compressor.layer3': 0.12852572137659507, 'gradnorm/compressor.layer4': 0.1184486320140653, 'gradnorm/compressor.layer5': 0.11381674674981111, 'gradnorm/compressor.layer6': 0.10954615705785622, 'gradnorm/compressor.layer7': 0.10351961955874299, 'gradnorm/decoder.layer0': 0.13443263697418825, 'gradnorm/decoder.layer1': 0.1288756926760295, 'gradnorm/decoder.layer2': 0.19190035512691556, 'gradnorm/max_param': 0.1379883885383606, 'epoch': 1.49}

 15%|█▍        | 2840/19090 [1:22:58<7:26:50,  1.65s/it]
 15%|█▍        | 2841/19090 [1:23:00<8:35:05,  1.90s/it]
 15%|█▍        | 2842/19090 [1:23:02<7:45:23,  1.72s/it]
 15%|█▍        | 2843/19090 [1:23:03<7:41:32,  1.70s/it]
 15%|█▍        | 2844/19090 [1:23:06<8:37:39,  1.91s/it]
 15%|█▍        | 2845/19090 [1:23:07<7:46:40,  1.72s/it]
 15%|█▍        | 2846/19090 [1:23:09<7:47:53,  1.73s/it]
 15%|█▍        | 2847/19090 [1:23:11<8:51:25,  1.96s/it]
 15%|█▍        | 2848/19090 [1:23:13<7:56:24,  1.76s/it]
 15%|█▍        | 2849/19090 [1:23:14<7:18:24,  1.62s/it]
 15%|█▍        | 2850/19090 [1:23:16<8:42:11,  1.93s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.4135264754295349, 'learning_rate': 4.9631617390511666e-05, 'recon_loss': 0.0032674092741217463, 'recon_static': 0.002317926974501461, 'recon_motion': 0.0009494822530541569, 'recon_norm': 1.991611933708191, 'recon_norm_std': 0.0848036352545023, 'target_norm': 1.6182214260101317, 'recon_cos': 0.45898001566529273, 'gradnorm/compressor.query': 0.00601778922013305, 'gradnorm/compressor.layer0': 0.1102576151275544, 'gradnorm/compressor.layer1': 0.14991493669231448, 'gradnorm/compressor.layer2': 0.12865744044828192, 'gradnorm/compressor.layer3': 0.120675047599482, 'gradnorm/compressor.layer4': 0.11006591758967488, 'gradnorm/compressor.layer5': 0.10503726132750618, 'gradnorm/compressor.layer6': 0.10138806622351428, 'gradnorm/compressor.layer7': 0.09589640407387579, 'gradnorm/decoder.layer0': 0.12376069224409628, 'gradnorm/decoder.layer1': 0.12187130537301735, 'gradnorm/decoder.layer2': 0.18008104019949836, 'gradnorm/max_param': 0.129397913813591, 'epoch': 1.49}

 15%|█▍        | 2850/19090 [1:23:16<8:42:11,  1.93s/it]
 15%|█▍        | 2851/19090 [1:23:18<7:50:23,  1.74s/it]
 15%|█▍        | 2852/19090 [1:23:19<7:24:28,  1.64s/it]
 15%|█▍        | 2853/19090 [1:23:21<8:09:44,  1.81s/it]
 15%|█▍        | 2854/19090 [1:23:23<7:27:48,  1.65s/it]
 15%|█▍        | 2855/19090 [1:23:24<7:40:22,  1.70s/it]
 15%|█▍        | 2856/19090 [1:23:27<8:54:41,  1.98s/it]
 15%|█▍        | 2857/19090 [1:23:28<7:58:46,  1.77s/it]
 15%|█▍        | 2858/19090 [1:23:30<8:01:09,  1.78s/it]
 15%|█▍        | 2859/19090 [1:23:32<8:39:36,  1.92s/it]
 15%|█▍        | 2860/19090 [1:23:34<7:48:39,  1.73s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.40970587730407715, 'learning_rate': 4.9623757594976264e-05, 'recon_loss': 0.0030787933908868583, 'recon_static': 0.0021218042937107386, 'recon_motion': 0.000956989123369567, 'recon_norm': 1.9290959000587464, 'recon_norm_std': 0.08866654839366675, 'target_norm': 1.620588406920433, 'recon_cos': 0.4701445013284683, 'gradnorm/compressor.query': 0.006212751280969242, 'gradnorm/compressor.layer0': 0.11029981986725633, 'gradnorm/compressor.layer1': 0.14930126195717275, 'gradnorm/compressor.layer2': 0.128502154586407, 'gradnorm/compressor.layer3': 0.11986676916399878, 'gradnorm/compressor.layer4': 0.11209493205114845, 'gradnorm/compressor.layer5': 0.10369992943082403, 'gradnorm/compressor.layer6': 0.09965569223383612, 'gradnorm/compressor.layer7': 0.0972140025834647, 'gradnorm/decoder.layer0': 0.12167409299232594, 'gradnorm/decoder.layer1': 0.11910068123786365, 'gradnorm/decoder.layer2': 0.17547707200803275, 'gradnorm/max_param': 0.12825030088424683, 'epoch': 1.5}

 15%|█▍        | 2860/19090 [1:23:34<7:48:39,  1.73s/it]
 15%|█▍        | 2861/19090 [1:23:35<7:47:48,  1.73s/it]
 15%|█▍        | 2862/19090 [1:23:38<8:14:10,  1.83s/it]
 15%|█▍        | 2863/19090 [1:23:39<7:30:36,  1.67s/it]
 15%|█▌        | 2864/19090 [1:23:41<7:53:16,  1.75s/it]
 15%|█▌        | 2865/19090 [1:23:43<8:18:44,  1.84s/it]
 15%|█▌        | 2866/19090 [1:23:44<7:33:28,  1.68s/it]
 15%|█▌        | 2867/19090 [1:23:47<8:42:49,  1.93s/it]
 15%|█▌        | 2868/19090 [1:23:48<7:54:33,  1.76s/it]
 15%|█▌        | 2869/19090 [1:23:49<7:16:33,  1.61s/it]
 15%|█▌        | 2870/19090 [1:23:52<8:47:01,  1.95s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.40522536635398865, 'learning_rate': 4.961581546963792e-05, 'recon_loss': 0.0031782588514033703, 'recon_static': 0.002221148775424808, 'recon_motion': 0.000957110158924479, 'recon_norm': 1.962192213535309, 'recon_norm_std': 0.08704861532896757, 'target_norm': 1.6198124408721923, 'recon_cos': 0.4641644813120365, 'gradnorm/compressor.query': 0.00642656264112374, 'gradnorm/compressor.layer0': 0.11000334394584937, 'gradnorm/compressor.layer1': 0.14830065995131905, 'gradnorm/compressor.layer2': 0.12600740897869517, 'gradnorm/compressor.layer3': 0.11840691713244358, 'gradnorm/compressor.layer4': 0.10791686196874133, 'gradnorm/compressor.layer5': 0.10354689359743899, 'gradnorm/compressor.layer6': 0.09728570255967818, 'gradnorm/compressor.layer7': 0.09301443920696503, 'gradnorm/decoder.layer0': 0.1215149146512118, 'gradnorm/decoder.layer1': 0.11908558058614838, 'gradnorm/decoder.layer2': 0.1753367313597691, 'gradnorm/max_param': 0.12688226997852325, 'epoch': 1.5}

 15%|█▌        | 2870/19090 [1:23:52<8:47:01,  1.95s/it]
 15%|█▌        | 2871/19090 [1:23:53<7:53:56,  1.75s/it]
 15%|█▌        | 2872/19090 [1:23:55<7:16:12,  1.61s/it]
 15%|█▌        | 2873/19090 [1:23:57<8:32:56,  1.90s/it]
 15%|█▌        | 2874/19090 [1:23:59<8:01:18,  1.78s/it]
 15%|█▌        | 2875/19090 [1:24:00<7:21:08,  1.63s/it]
 15%|█▌        | 2876/19090 [1:24:02<8:35:33,  1.91s/it]
 15%|█▌        | 2877/19090 [1:24:04<7:53:46,  1.75s/it]
 15%|█▌        | 2878/19090 [1:24:05<7:16:04,  1.61s/it]
 15%|█▌        | 2879/19090 [1:24:07<8:09:07,  1.81s/it]
 15%|█▌        | 2880/19090 [1:24:09<8:30:56,  1.89s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.46947288513183594, 'learning_rate': 4.960779104105123e-05, 'recon_loss': 0.0033239920740015805, 'recon_static': 0.0023701805796008555, 'recon_motion': 0.0009538114740280434, 'recon_norm': 2.0035510420799256, 'recon_norm_std': 0.08327549919486046, 'target_norm': 1.617647424340248, 'recon_cos': 0.45413107201457026, 'gradnorm/compressor.query': 0.007386690435517293, 'gradnorm/compressor.layer0': 0.12981147527956033, 'gradnorm/compressor.layer1': 0.1751980856821618, 'gradnorm/compressor.layer2': 0.14424142259445483, 'gradnorm/compressor.layer3': 0.13472751410267972, 'gradnorm/compressor.layer4': 0.12498911050213295, 'gradnorm/compressor.layer5': 0.11689604437158671, 'gradnorm/compressor.layer6': 0.11251917549796167, 'gradnorm/compressor.layer7': 0.11036054608397315, 'gradnorm/decoder.layer0': 0.1416550765368677, 'gradnorm/decoder.layer1': 0.13745091439963256, 'gradnorm/decoder.layer2': 0.20189569863181822, 'gradnorm/max_param': 0.14668284356594086, 'epoch': 1.51}

 15%|█▌        | 2880/19090 [1:24:10<8:30:56,  1.89s/it]
 15%|█▌        | 2881/19090 [1:24:11<7:42:07,  1.71s/it]
 15%|█▌        | 2882/19090 [1:24:13<7:55:47,  1.76s/it]
 15%|█▌        | 2883/19090 [1:24:15<8:20:36,  1.85s/it]
 15%|█▌        | 2884/19090 [1:24:16<7:34:38,  1.68s/it]
 15%|█▌        | 2885/19090 [1:24:18<7:46:30,  1.73s/it]
 15%|█▌        | 2886/19090 [1:24:20<8:33:51,  1.90s/it]
 15%|█▌        | 2887/19090 [1:24:21<7:44:20,  1.72s/it]
 15%|█▌        | 2888/19090 [1:24:23<8:06:16,  1.80s/it]
 15%|█▌        | 2889/19090 [1:24:25<8:11:44,  1.82s/it]
 15%|█▌        | 2890/19090 [1:24:27<7:28:15,  1.66s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.4506780505180359, 'learning_rate': 4.959968433604596e-05, 'recon_loss': 0.003467872377950698, 'recon_static': 0.002504808543017134, 'recon_motion': 0.0009630638174712658, 'recon_norm': 2.043884187936783, 'recon_norm_std': 0.08330262936651707, 'target_norm': 1.621859472990036, 'recon_cos': 0.44515582993626596, 'gradnorm/compressor.query': 0.006946483592744884, 'gradnorm/compressor.layer0': 0.11991607297846914, 'gradnorm/compressor.layer1': 0.16425201278771262, 'gradnorm/compressor.layer2': 0.13848984551043284, 'gradnorm/compressor.layer3': 0.12973563381091116, 'gradnorm/compressor.layer4': 0.12162504622550666, 'gradnorm/compressor.layer5': 0.11419385766640248, 'gradnorm/compressor.layer6': 0.10954637810041182, 'gradnorm/compressor.layer7': 0.10588217433466078, 'gradnorm/decoder.layer0': 0.13585962388738085, 'gradnorm/decoder.layer1': 0.13233426889914401, 'gradnorm/decoder.layer2': 0.1966686374473292, 'gradnorm/max_param': 0.14190417528152466, 'epoch': 1.51}

 15%|█▌        | 2890/19090 [1:24:27<7:28:15,  1.66s/it]
 15%|█▌        | 2891/19090 [1:24:29<8:08:22,  1.81s/it]
 15%|█▌        | 2892/19090 [1:24:31<8:18:49,  1.85s/it]
 15%|█▌        | 2893/19090 [1:24:32<7:32:58,  1.68s/it]
 15%|█▌        | 2894/19090 [1:24:34<8:19:14,  1.85s/it]
 15%|█▌        | 2895/19090 [1:24:36<8:30:45,  1.89s/it]
 15%|█▌        | 2896/19090 [1:24:38<7:41:40,  1.71s/it]
 15%|█▌        | 2897/19090 [1:24:40<8:16:35,  1.84s/it]
 15%|█▌        | 2898/19090 [1:24:42<8:34:58,  1.91s/it]
 15%|█▌        | 2899/19090 [1:24:43<7:44:19,  1.72s/it]
 15%|█▌        | 2900/19090 [1:24:45<7:57:49,  1.77s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.40149691700935364, 'learning_rate': 4.959149538172696e-05, 'recon_loss': 0.003182963718427345, 'recon_static': 0.0022186644084285943, 'recon_motion': 0.0009642992634326219, 'recon_norm': 1.9615401446819305, 'recon_norm_std': 0.08489271812140942, 'target_norm': 1.6214598000049592, 'recon_cos': 0.46379389688372613, 'gradnorm/compressor.query': 0.006006660882842237, 'gradnorm/compressor.layer0': 0.10593284383601419, 'gradnorm/compressor.layer1': 0.14262977082767728, 'gradnorm/compressor.layer2': 0.12275431685276599, 'gradnorm/compressor.layer3': 0.11607750276441653, 'gradnorm/compressor.layer4': 0.10726902230502744, 'gradnorm/compressor.layer5': 0.10132544159552986, 'gradnorm/compressor.layer6': 0.09539196461792117, 'gradnorm/compressor.layer7': 0.09454777155488742, 'gradnorm/decoder.layer0': 0.1237058816821626, 'gradnorm/decoder.layer1': 0.11891196148222222, 'gradnorm/decoder.layer2': 0.1783268140796303, 'gradnorm/max_param': 0.12757684290409088, 'epoch': 1.52}

 15%|█▌        | 2900/19090 [1:24:45<7:57:49,  1.77s/it]
 15%|█▌        | 2901/19090 [1:24:47<7:54:26,  1.76s/it]
 15%|█▌        | 2902/19090 [1:24:48<7:16:27,  1.62s/it]
 15%|█▌        | 2903/19090 [1:24:50<8:19:07,  1.85s/it]
 15%|█▌        | 2904/19090 [1:24:52<8:02:59,  1.79s/it]
 15%|█▌        | 2905/19090 [1:24:53<7:22:34,  1.64s/it]
 15%|█▌        | 2906/19090 [1:24:56<8:29:11,  1.89s/it]
 15%|█▌        | 2907/19090 [1:24:57<8:20:09,  1.85s/it]
 15%|█▌        | 2908/19090 [1:24:59<7:34:48,  1.69s/it]
 15%|█▌        | 2909/19090 [1:25:01<8:13:12,  1.83s/it]
 15%|█▌        | 2910/19090 [1:25:03<8:23:57,  1.87s/it]
                                                        
{'loss': 0.0027, 'grad_norm': 0.3185184597969055, 'learning_rate': 4.958322420547409e-05, 'recon_loss': 0.0026689725811593233, 'recon_static': 0.0017098090553190558, 'recon_motion': 0.0009591635738615878, 'recon_norm': 1.8014338046312333, 'recon_norm_std': 0.09239033330231905, 'target_norm': 1.6205068290233613, 'recon_cos': 0.5028475917875767, 'gradnorm/compressor.query': 0.004719250778165908, 'gradnorm/compressor.layer0': 0.08688000751204056, 'gradnorm/compressor.layer1': 0.11477924436202913, 'gradnorm/compressor.layer2': 0.09810540382493985, 'gradnorm/compressor.layer3': 0.09292195629111827, 'gradnorm/compressor.layer4': 0.08493014852005487, 'gradnorm/compressor.layer5': 0.07864917504345705, 'gradnorm/compressor.layer6': 0.07732743912475717, 'gradnorm/compressor.layer7': 0.07546166213438335, 'gradnorm/decoder.layer0': 0.0964695785929885, 'gradnorm/decoder.layer1': 0.09390792360417112, 'gradnorm/decoder.layer2': 0.13875253775999372, 'gradnorm/max_param': 0.10083752125501633, 'epoch': 1.52}

 15%|█▌        | 2910/19090 [1:25:03<8:23:57,  1.87s/it]
 15%|█▌        | 2911/19090 [1:25:04<7:36:59,  1.69s/it]
 15%|█▌        | 2912/19090 [1:25:07<8:30:03,  1.89s/it]
 15%|█▌        | 2913/19090 [1:25:08<8:12:17,  1.83s/it]
 15%|█▌        | 2914/19090 [1:25:10<7:29:09,  1.67s/it]
 15%|█▌        | 2915/19090 [1:25:12<8:23:58,  1.87s/it]
 15%|█▌        | 2916/19090 [1:25:13<7:49:10,  1.74s/it]
 15%|█▌        | 2917/19090 [1:25:15<7:12:57,  1.61s/it]
 15%|█▌        | 2918/19090 [1:25:17<8:51:14,  1.97s/it]
 15%|█▌        | 2919/19090 [1:25:19<7:56:16,  1.77s/it]
 15%|█▌        | 2920/19090 [1:25:20<7:17:23,  1.62s/it]
                                                        
{'loss': 0.0026, 'grad_norm': 0.35216572880744934, 'learning_rate': 4.957487083494212e-05, 'recon_loss': 0.0025822897092439233, 'recon_static': 0.0016118006693432107, 'recon_motion': 0.0009704890777356923, 'recon_norm': 1.772031980752945, 'recon_norm_std': 0.0895332757383585, 'target_norm': 1.6196691036224364, 'recon_cos': 0.5101321011781692, 'gradnorm/compressor.query': 0.005249341030548598, 'gradnorm/compressor.layer0': 0.09300502181722954, 'gradnorm/compressor.layer1': 0.12758204128844303, 'gradnorm/compressor.layer2': 0.10807525295198847, 'gradnorm/compressor.layer3': 0.10233443141329604, 'gradnorm/compressor.layer4': 0.0942635378815984, 'gradnorm/compressor.layer5': 0.08909973173589437, 'gradnorm/compressor.layer6': 0.08647565695646668, 'gradnorm/compressor.layer7': 0.08353653146068005, 'gradnorm/decoder.layer0': 0.106532905131301, 'gradnorm/decoder.layer1': 0.10409978467843894, 'gradnorm/decoder.layer2': 0.15312295104621687, 'gradnorm/max_param': 0.11107489466667175, 'epoch': 1.53}

 15%|█▌        | 2920/19090 [1:25:20<7:17:23,  1.62s/it]
 15%|█▌        | 2921/19090 [1:25:23<8:44:21,  1.95s/it]
 15%|█▌        | 2922/19090 [1:25:24<8:04:34,  1.80s/it]
 15%|█▌        | 2923/19090 [1:25:25<7:23:32,  1.65s/it]
 15%|█▌        | 2924/19090 [1:25:28<8:32:06,  1.90s/it]
 15%|█▌        | 2925/19090 [1:25:30<8:06:50,  1.81s/it]
 15%|█▌        | 2926/19090 [1:25:31<7:24:14,  1.65s/it]
 15%|█▌        | 2927/19090 [1:25:33<8:32:37,  1.90s/it]
 15%|█▌        | 2928/19090 [1:25:35<8:33:32,  1.91s/it]
 15%|█▌        | 2929/19090 [1:25:36<7:43:41,  1.72s/it]
 15%|█▌        | 2930/19090 [1:25:38<8:07:06,  1.81s/it]
                                                        
{'loss': 0.0028, 'grad_norm': 0.46028125286102295, 'learning_rate': 4.956643529806062e-05, 'recon_loss': 0.0028256165387574583, 'recon_static': 0.0018508009583456443, 'recon_motion': 0.0009748154916451313, 'recon_norm': 1.8473350822925567, 'recon_norm_std': 0.08757259380072355, 'target_norm': 1.6220145612955092, 'recon_cos': 0.4913295656442642, 'gradnorm/compressor.query': 0.007320476112189966, 'gradnorm/compressor.layer0': 0.12450562256493672, 'gradnorm/compressor.layer1': 0.16500073614281385, 'gradnorm/compressor.layer2': 0.13992673979395462, 'gradnorm/compressor.layer3': 0.13250014093679455, 'gradnorm/compressor.layer4': 0.12238966177200099, 'gradnorm/compressor.layer5': 0.11735911915674664, 'gradnorm/compressor.layer6': 0.11293408207505595, 'gradnorm/compressor.layer7': 0.11173871102106225, 'gradnorm/decoder.layer0': 0.1398259548500963, 'gradnorm/decoder.layer1': 0.13723192633948506, 'gradnorm/decoder.layer2': 0.19891356485465708, 'gradnorm/max_param': 0.143889918923378, 'epoch': 1.53}

 15%|█▌        | 2930/19090 [1:25:39<8:07:06,  1.81s/it]
 15%|█▌        | 2931/19090 [1:25:41<8:23:59,  1.87s/it]
 15%|█▌        | 2932/19090 [1:25:42<7:36:28,  1.70s/it]
 15%|█▌        | 2933/19090 [1:25:44<7:51:20,  1.75s/it]
 15%|█▌        | 2934/19090 [1:25:46<8:16:22,  1.84s/it]
 15%|█▌        | 2935/19090 [1:25:47<7:31:43,  1.68s/it]
 15%|█▌        | 2936/19090 [1:25:49<7:35:07,  1.69s/it]
 15%|█▌        | 2937/19090 [1:25:51<8:24:21,  1.87s/it]
 15%|█▌        | 2938/19090 [1:25:52<7:36:57,  1.70s/it]
 15%|█▌        | 2939/19090 [1:25:54<7:30:45,  1.67s/it]
 15%|█▌        | 2940/19090 [1:25:56<8:22:32,  1.87s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.5349568128585815, 'learning_rate': 4.955791762303393e-05, 'recon_loss': 0.004633499658666551, 'recon_static': 0.003672579413978383, 'recon_motion': 0.0009609202825231478, 'recon_norm': 2.349240654706955, 'recon_norm_std': 0.07517389254644513, 'target_norm': 1.6209571927785873, 'recon_cos': 0.38862196952104566, 'gradnorm/compressor.query': 0.008340590661163067, 'gradnorm/compressor.layer0': 0.14350004518612872, 'gradnorm/compressor.layer1': 0.19375022726660904, 'gradnorm/compressor.layer2': 0.16458176567341046, 'gradnorm/compressor.layer3': 0.15288130968162184, 'gradnorm/compressor.layer4': 0.14374022269169545, 'gradnorm/compressor.layer5': 0.13541401499817438, 'gradnorm/compressor.layer6': 0.1285531018349736, 'gradnorm/compressor.layer7': 0.1250570658718981, 'gradnorm/decoder.layer0': 0.1615061394355655, 'gradnorm/decoder.layer1': 0.15763594954143118, 'gradnorm/decoder.layer2': 0.2354171392483877, 'gradnorm/max_param': 0.16897282004356384, 'epoch': 1.54}

 15%|█▌        | 2940/19090 [1:25:56<8:22:32,  1.87s/it]
 15%|█▌        | 2941/19090 [1:25:58<7:36:03,  1.69s/it]
 15%|█▌        | 2942/19090 [1:25:59<7:31:43,  1.68s/it]
 15%|█▌        | 2943/19090 [1:26:01<8:16:27,  1.84s/it]
 15%|█▌        | 2944/19090 [1:26:03<7:31:53,  1.68s/it]
 15%|█▌        | 2945/19090 [1:26:04<7:27:12,  1.66s/it]
 15%|█▌        | 2946/19090 [1:26:07<8:37:08,  1.92s/it]
 15%|█▌        | 2947/19090 [1:26:08<7:46:05,  1.73s/it]
 15%|█▌        | 2948/19090 [1:26:10<7:39:25,  1.71s/it]
 15%|█▌        | 2949/19090 [1:26:12<8:45:52,  1.95s/it]
 15%|█▌        | 2950/19090 [1:26:14<7:53:38,  1.76s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.48239853978157043, 'learning_rate': 4.954931783834096e-05, 'recon_loss': 0.004329049325315281, 'recon_static': 0.003366509545594454, 'recon_motion': 0.0009625398626667448, 'recon_norm': 2.2721886694431306, 'recon_norm_std': 0.07618442066013813, 'target_norm': 1.6224687844514847, 'recon_cos': 0.4007429428398609, 'gradnorm/compressor.query': 0.006690028524364054, 'gradnorm/compressor.layer0': 0.12630777657514225, 'gradnorm/compressor.layer1': 0.173099501639414, 'gradnorm/compressor.layer2': 0.1491589706072077, 'gradnorm/compressor.layer3': 0.1410418002702239, 'gradnorm/compressor.layer4': 0.12788002127496206, 'gradnorm/compressor.layer5': 0.12296880788692678, 'gradnorm/compressor.layer6': 0.1175768939299152, 'gradnorm/compressor.layer7': 0.11451607232794339, 'gradnorm/decoder.layer0': 0.14571397372808936, 'gradnorm/decoder.layer1': 0.14304339904145405, 'gradnorm/decoder.layer2': 0.21094660386113762, 'gradnorm/max_param': 0.1536022573709488, 'epoch': 1.55}

 15%|█▌        | 2950/19090 [1:26:14<7:53:38,  1.76s/it]
 15%|█▌        | 2951/19090 [1:26:15<7:16:16,  1.62s/it]
 15%|█▌        | 2952/19090 [1:26:18<8:59:18,  2.01s/it]
 15%|█▌        | 2953/19090 [1:26:19<8:01:09,  1.79s/it]
 15%|█▌        | 2954/19090 [1:26:20<7:20:25,  1.64s/it]
 15%|█▌        | 2955/19090 [1:26:23<8:30:42,  1.90s/it]
 15%|█▌        | 2956/19090 [1:26:24<7:41:30,  1.72s/it]
 15%|█▌        | 2957/19090 [1:26:26<7:06:59,  1.59s/it]
 15%|█▌        | 2958/19090 [1:26:28<8:47:35,  1.96s/it]
 16%|█▌        | 2959/19090 [1:26:30<7:53:13,  1.76s/it]
 16%|█▌        | 2960/19090 [1:26:31<7:15:08,  1.62s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.46047499775886536, 'learning_rate': 4.9540635972735205e-05, 'recon_loss': 0.0033591586630791425, 'recon_static': 0.0024076652538497, 'recon_motion': 0.0009514933146419935, 'recon_norm': 2.016350209712982, 'recon_norm_std': 0.08212883695960045, 'target_norm': 1.6178526043891908, 'recon_cos': 0.4525791294872761, 'gradnorm/compressor.query': 0.007432140770638859, 'gradnorm/compressor.layer0': 0.12731619782571643, 'gradnorm/compressor.layer1': 0.1690422287279632, 'gradnorm/compressor.layer2': 0.14059878475825935, 'gradnorm/compressor.layer3': 0.1339789803756291, 'gradnorm/compressor.layer4': 0.12025924748821193, 'gradnorm/compressor.layer5': 0.11383216999959607, 'gradnorm/compressor.layer6': 0.10944865931019379, 'gradnorm/compressor.layer7': 0.10836387633178872, 'gradnorm/decoder.layer0': 0.13833583492603357, 'gradnorm/decoder.layer1': 0.1341750084451261, 'gradnorm/decoder.layer2': 0.2029713820065823, 'gradnorm/max_param': 0.14664915204048157, 'epoch': 1.55}

 16%|█▌        | 2960/19090 [1:26:31<7:15:08,  1.62s/it]
 16%|█▌        | 2961/19090 [1:26:34<9:02:32,  2.02s/it]
 16%|█▌        | 2962/19090 [1:26:35<8:03:46,  1.80s/it]
 16%|█▌        | 2963/19090 [1:26:36<7:22:21,  1.65s/it]
 16%|█▌        | 2964/19090 [1:26:39<9:03:53,  2.02s/it]
 16%|█▌        | 2965/19090 [1:26:41<8:04:59,  1.80s/it]
 16%|█▌        | 2966/19090 [1:26:42<7:23:36,  1.65s/it]
 16%|█▌        | 2967/19090 [1:26:44<8:08:52,  1.82s/it]
 16%|█▌        | 2968/19090 [1:26:45<7:25:53,  1.66s/it]
 16%|█▌        | 2969/19090 [1:26:47<6:56:00,  1.55s/it]
 16%|█▌        | 2970/19090 [1:26:49<8:26:20,  1.88s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.38935157656669617, 'learning_rate': 4.953187205524457e-05, 'recon_loss': 0.0032428127771709113, 'recon_static': 0.0022838119184598327, 'recon_motion': 0.0009590007961378433, 'recon_norm': 1.9841739058494567, 'recon_norm_std': 0.08472872152924538, 'target_norm': 1.6195376932621002, 'recon_cos': 0.4610132619738579, 'gradnorm/compressor.query': 0.006307690749279844, 'gradnorm/compressor.layer0': 0.10447037549398867, 'gradnorm/compressor.layer1': 0.1396833456198255, 'gradnorm/compressor.layer2': 0.11769725872650078, 'gradnorm/compressor.layer3': 0.11085820571116085, 'gradnorm/compressor.layer4': 0.1039895553717158, 'gradnorm/compressor.layer5': 0.09830299793323358, 'gradnorm/compressor.layer6': 0.09697210517258044, 'gradnorm/compressor.layer7': 0.09169982792141107, 'gradnorm/decoder.layer0': 0.11775927925094835, 'gradnorm/decoder.layer1': 0.11662014472918998, 'gradnorm/decoder.layer2': 0.17096760031498875, 'gradnorm/max_param': 0.12326719611883163, 'epoch': 1.56}

 16%|█▌        | 2970/19090 [1:26:49<8:26:20,  1.88s/it]
 16%|█▌        | 2971/19090 [1:26:51<7:38:04,  1.71s/it]
 16%|█▌        | 2972/19090 [1:26:52<7:04:46,  1.58s/it]
 16%|█▌        | 2973/19090 [1:26:55<8:32:41,  1.91s/it]
 16%|█▌        | 2974/19090 [1:26:56<7:43:09,  1.72s/it]
 16%|█▌        | 2975/19090 [1:26:57<7:07:48,  1.59s/it]
 16%|█▌        | 2976/19090 [1:27:00<8:22:22,  1.87s/it]
 16%|█▌        | 2977/19090 [1:27:01<7:35:50,  1.70s/it]
 16%|█▌        | 2978/19090 [1:27:02<7:02:46,  1.57s/it]
 16%|█▌        | 2979/19090 [1:27:05<8:40:14,  1.94s/it]
 16%|█▌        | 2980/19090 [1:27:06<7:48:02,  1.74s/it]
                                                        
{'loss': 0.0026, 'grad_norm': 0.3006150424480438, 'learning_rate': 4.9523026115171305e-05, 'recon_loss': 0.002626931411214173, 'recon_static': 0.0016681344190146775, 'recon_motion': 0.0009587970838765614, 'recon_norm': 1.792715209722519, 'recon_norm_std': 0.09236439317464828, 'target_norm': 1.6211600571870803, 'recon_cos': 0.5089733861386776, 'gradnorm/compressor.query': 0.004654053975644766, 'gradnorm/compressor.layer0': 0.08003661196165536, 'gradnorm/compressor.layer1': 0.1084883634764531, 'gradnorm/compressor.layer2': 0.09255442617789554, 'gradnorm/compressor.layer3': 0.08755472953722535, 'gradnorm/compressor.layer4': 0.07943715034328375, 'gradnorm/compressor.layer5': 0.07711487957690401, 'gradnorm/compressor.layer6': 0.07492507331075764, 'gradnorm/compressor.layer7': 0.07037047142115724, 'gradnorm/decoder.layer0': 0.09058770777200943, 'gradnorm/decoder.layer1': 0.08790704255945231, 'gradnorm/decoder.layer2': 0.1310858403798951, 'gradnorm/max_param': 0.09539946913719177, 'epoch': 1.56}

 16%|█▌        | 2980/19090 [1:27:06<7:48:02,  1.74s/it]
 16%|█▌        | 2981/19090 [1:27:08<7:11:44,  1.61s/it]
 16%|█▌        | 2982/19090 [1:27:10<8:23:46,  1.88s/it]
 16%|█▌        | 2983/19090 [1:27:11<7:36:16,  1.70s/it]
 16%|█▌        | 2984/19090 [1:27:13<7:02:57,  1.58s/it]
 16%|█▌        | 2985/19090 [1:27:16<8:53:32,  1.99s/it]
 16%|█▌        | 2986/19090 [1:27:17<7:57:32,  1.78s/it]
 16%|█▌        | 2987/19090 [1:27:18<7:17:42,  1.63s/it]
 16%|█▌        | 2988/19090 [1:27:21<8:46:27,  1.96s/it]
 16%|█▌        | 2989/19090 [1:27:22<7:52:29,  1.76s/it]
 16%|█▌        | 2990/19090 [1:27:24<7:14:42,  1.62s/it]
                                                        
{'loss': 0.0026, 'grad_norm': 0.3777373433113098, 'learning_rate': 4.951409818209191e-05, 'recon_loss': 0.002558963868068531, 'recon_static': 0.0015998804359696806, 'recon_motion': 0.00095908348885132, 'recon_norm': 1.771071520447731, 'recon_norm_std': 0.0921659717336297, 'target_norm': 1.6218399554491043, 'recon_cos': 0.5163895927369595, 'gradnorm/compressor.query': 0.006063437604302875, 'gradnorm/compressor.layer0': 0.10274879260119497, 'gradnorm/compressor.layer1': 0.13401667175070978, 'gradnorm/compressor.layer2': 0.11582142602600748, 'gradnorm/compressor.layer3': 0.11016630979014624, 'gradnorm/compressor.layer4': 0.10086062802128122, 'gradnorm/compressor.layer5': 0.09637491758815542, 'gradnorm/compressor.layer6': 0.09173343881192048, 'gradnorm/compressor.layer7': 0.08923589282848182, 'gradnorm/decoder.layer0': 0.11561699562461526, 'gradnorm/decoder.layer1': 0.11236205916102346, 'gradnorm/decoder.layer2': 0.16354861841997292, 'gradnorm/max_param': 0.11869270354509354, 'epoch': 1.57}

 16%|█▌        | 2990/19090 [1:27:24<7:14:42,  1.62s/it]
 16%|█▌        | 2991/19090 [1:27:27<9:02:15,  2.02s/it]
 16%|█▌        | 2992/19090 [1:27:28<8:03:21,  1.80s/it]
 16%|█▌        | 2993/19090 [1:27:29<7:22:15,  1.65s/it]
 16%|█▌        | 2994/19090 [1:27:32<9:05:59,  2.04s/it]
 16%|█▌        | 2995/19090 [1:27:33<8:06:29,  1.81s/it]
 16%|█▌        | 2996/19090 [1:27:35<7:24:20,  1.66s/it]
 16%|█▌        | 2997/19090 [1:27:37<8:13:09,  1.84s/it]
 16%|█▌        | 2998/19090 [1:27:38<7:28:30,  1.67s/it]
 16%|█▌        | 2999/19090 [1:27:40<6:57:35,  1.56s/it]
 16%|█▌        | 3000/19090 [1:27:42<8:38:02,  1.93s/it]
                                                        
{'loss': 0.0027, 'grad_norm': 0.3250802457332611, 'learning_rate': 4.950508828585704e-05, 'recon_loss': 0.0027279834728688003, 'recon_static': 0.0017722473479807378, 'recon_motion': 0.0009557361394399777, 'recon_norm': 1.8243839919567109, 'recon_norm_std': 0.08943695537745952, 'target_norm': 1.6181685745716095, 'recon_cos': 0.49848435148596765, 'gradnorm/compressor.query': 0.005308258331073985, 'gradnorm/compressor.layer0': 0.0880361623602026, 'gradnorm/compressor.layer1': 0.11770928171149937, 'gradnorm/compressor.layer2': 0.09955832631851315, 'gradnorm/compressor.layer3': 0.09335827447729399, 'gradnorm/compressor.layer4': 0.08703415053840213, 'gradnorm/compressor.layer5': 0.08171307928402419, 'gradnorm/compressor.layer6': 0.07985269986249485, 'gradnorm/compressor.layer7': 0.07689452941871275, 'gradnorm/decoder.layer0': 0.09838364070021804, 'gradnorm/decoder.layer1': 0.0958300060838339, 'gradnorm/decoder.layer2': 0.1414623694471421, 'gradnorm/max_param': 0.10271142423152924, 'epoch': 1.57}

 16%|█▌        | 3000/19090 [1:27:42<8:38:02,  1.93s/it]
 16%|█▌        | 3001/19090 [1:27:46<10:33:12,  2.36s/it]
 16%|█▌        | 3002/19090 [1:27:47<9:07:56,  2.04s/it] 
 16%|█▌        | 3003/19090 [1:27:48<8:08:04,  1.82s/it]
 16%|█▌        | 3004/19090 [1:27:50<7:25:47,  1.66s/it]
 16%|█▌        | 3005/19090 [1:27:51<6:56:30,  1.55s/it]
 16%|█▌        | 3006/19090 [1:27:52<6:35:59,  1.48s/it]
 16%|█▌        | 3007/19090 [1:27:53<6:22:36,  1.43s/it]
 16%|█▌        | 3008/19090 [1:27:55<6:12:18,  1.39s/it]
 16%|█▌        | 3009/19090 [1:27:57<6:44:20,  1.51s/it]
 16%|█▌        | 3010/19090 [1:27:58<6:27:30,  1.45s/it]
                                                        
{'loss': 0.0027, 'grad_norm': 0.41427886486053467, 'learning_rate': 4.949599645659135e-05, 'recon_loss': 0.0026781224994920196, 'recon_static': 0.0017208886303706095, 'recon_motion': 0.0009572337497957051, 'recon_norm': 1.8084152787923813, 'recon_norm_std': 0.09230690523982048, 'target_norm': 1.6211629927158355, 'recon_cos': 0.5045140638947487, 'gradnorm/compressor.query': 0.006702476288481807, 'gradnorm/compressor.layer0': 0.11280496954550755, 'gradnorm/compressor.layer1': 0.14985628072674137, 'gradnorm/compressor.layer2': 0.12828863179600497, 'gradnorm/compressor.layer3': 0.11970377155956413, 'gradnorm/compressor.layer4': 0.11112107716095354, 'gradnorm/compressor.layer5': 0.10450506576609478, 'gradnorm/compressor.layer6': 0.10058322921366045, 'gradnorm/compressor.layer7': 0.09695637730531259, 'gradnorm/decoder.layer0': 0.12511520739086276, 'gradnorm/decoder.layer1': 0.12098007096058326, 'gradnorm/decoder.layer2': 0.18037876594059354, 'gradnorm/max_param': 0.13031531870365143, 'epoch': 1.58}

 16%|█▌        | 3010/19090 [1:27:58<6:27:30,  1.45s/it]
 16%|█▌        | 3011/19090 [1:27:59<6:15:03,  1.40s/it]
 16%|█▌        | 3012/19090 [1:28:02<7:37:21,  1.71s/it]
 16%|█▌        | 3013/19090 [1:28:03<7:03:40,  1.58s/it]
 16%|█▌        | 3014/19090 [1:28:04<6:39:57,  1.49s/it]
 16%|█▌        | 3015/19090 [1:28:07<8:35:48,  1.93s/it]
 16%|█▌        | 3016/19090 [1:28:08<7:44:29,  1.73s/it]
 16%|█▌        | 3017/19090 [1:28:10<7:08:40,  1.60s/it]
 16%|█▌        | 3018/19090 [1:28:12<8:32:52,  1.91s/it]
 16%|█▌        | 3019/19090 [1:28:14<7:44:53,  1.74s/it]
 16%|█▌        | 3020/19090 [1:28:15<7:09:14,  1.60s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.5454147458076477, 'learning_rate': 4.948682272469351e-05, 'recon_loss': 0.0037659778899978846, 'recon_static': 0.0028127211262471974, 'recon_motion': 0.0009532567055430263, 'recon_norm': 2.1295858025550842, 'recon_norm_std': 0.08205827884376049, 'target_norm': 1.6211868077516556, 'recon_cos': 0.4307564660906792, 'gradnorm/compressor.query': 0.007875471376055262, 'gradnorm/compressor.layer0': 0.1434142535202617, 'gradnorm/compressor.layer1': 0.1979903985179636, 'gradnorm/compressor.layer2': 0.17020266451572416, 'gradnorm/compressor.layer3': 0.1614759259996471, 'gradnorm/compressor.layer4': 0.14873638304124207, 'gradnorm/compressor.layer5': 0.1369130524993952, 'gradnorm/compressor.layer6': 0.13158376382573528, 'gradnorm/compressor.layer7': 0.12673788163793323, 'gradnorm/decoder.layer0': 0.1634482396793008, 'gradnorm/decoder.layer1': 0.1603141796849468, 'gradnorm/decoder.layer2': 0.23652379943920765, 'gradnorm/max_param': 0.17093908786773682, 'epoch': 1.58}

 16%|█▌        | 3020/19090 [1:28:15<7:09:14,  1.60s/it]
 16%|█▌        | 3021/19090 [1:28:18<8:41:04,  1.95s/it]
 16%|█▌        | 3022/19090 [1:28:19<7:48:01,  1.75s/it]
 16%|█▌        | 3023/19090 [1:28:20<7:11:21,  1.61s/it]
 16%|█▌        | 3024/19090 [1:28:23<8:47:46,  1.97s/it]
 16%|█▌        | 3025/19090 [1:28:24<7:52:52,  1.77s/it]
 16%|█▌        | 3026/19090 [1:28:26<7:14:10,  1.62s/it]
 16%|█▌        | 3027/19090 [1:28:28<8:22:02,  1.88s/it]
 16%|█▌        | 3028/19090 [1:28:29<7:35:11,  1.70s/it]
 16%|█▌        | 3029/19090 [1:28:31<7:02:19,  1.58s/it]
 16%|█▌        | 3030/19090 [1:28:33<8:34:46,  1.92s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.4331549108028412, 'learning_rate': 4.9477567120835964e-05, 'recon_loss': 0.003880176175152883, 'recon_static': 0.0029201664787251503, 'recon_motion': 0.0009600097080692648, 'recon_norm': 2.158516389131546, 'recon_norm_std': 0.07891697976738214, 'target_norm': 1.6195845633745194, 'recon_cos': 0.42342726662755015, 'gradnorm/compressor.query': 0.006516698821334002, 'gradnorm/compressor.layer0': 0.11427925593331793, 'gradnorm/compressor.layer1': 0.1566744879556728, 'gradnorm/compressor.layer2': 0.13408454925995567, 'gradnorm/compressor.layer3': 0.12628303626726903, 'gradnorm/compressor.layer4': 0.11562262609983229, 'gradnorm/compressor.layer5': 0.11207691812709723, 'gradnorm/compressor.layer6': 0.10485554292674291, 'gradnorm/compressor.layer7': 0.10182091500818127, 'gradnorm/decoder.layer0': 0.1298360883951874, 'gradnorm/decoder.layer1': 0.1273489795968378, 'gradnorm/decoder.layer2': 0.18889367195011408, 'gradnorm/max_param': 0.13723152875900269, 'epoch': 1.59}

 16%|█▌        | 3030/19090 [1:28:33<8:34:46,  1.92s/it]
 16%|█▌        | 3031/19090 [1:28:35<7:43:44,  1.73s/it]
 16%|█▌        | 3032/19090 [1:28:36<7:07:54,  1.60s/it]
 16%|█▌        | 3033/19090 [1:28:39<8:27:08,  1.90s/it]
 16%|█▌        | 3034/19090 [1:28:40<7:37:54,  1.71s/it]
 16%|█▌        | 3035/19090 [1:28:41<7:06:42,  1.59s/it]
 16%|█▌        | 3036/19090 [1:28:44<8:31:32,  1.91s/it]
 16%|█▌        | 3037/19090 [1:28:45<7:40:59,  1.72s/it]
 16%|█▌        | 3038/19090 [1:28:47<7:12:50,  1.62s/it]
 16%|█▌        | 3039/19090 [1:28:49<8:52:09,  1.99s/it]
 16%|█▌        | 3040/19090 [1:28:51<7:55:38,  1.78s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.41312500834465027, 'learning_rate': 4.9468229675964934e-05, 'recon_loss': 0.003413979581091553, 'recon_static': 0.0024601818644441665, 'recon_motion': 0.0009537977908621542, 'recon_norm': 2.0347828030586244, 'recon_norm_std': 0.0825141565874219, 'target_norm': 1.619855108857155, 'recon_cos': 0.45035153403878214, 'gradnorm/compressor.query': 0.006133987535714729, 'gradnorm/compressor.layer0': 0.11065575356737037, 'gradnorm/compressor.layer1': 0.15036046357342037, 'gradnorm/compressor.layer2': 0.12535844015462083, 'gradnorm/compressor.layer3': 0.11847334967136128, 'gradnorm/compressor.layer4': 0.10878682793423133, 'gradnorm/compressor.layer5': 0.1047605917686604, 'gradnorm/compressor.layer6': 0.10204322458425051, 'gradnorm/compressor.layer7': 0.09621540486677527, 'gradnorm/decoder.layer0': 0.12513142620556508, 'gradnorm/decoder.layer1': 0.12184823625699429, 'gradnorm/decoder.layer2': 0.18178719585374634, 'gradnorm/max_param': 0.131147101521492, 'epoch': 1.59}

 16%|█▌        | 3040/19090 [1:28:51<7:55:38,  1.78s/it]
 16%|█▌        | 3041/19090 [1:28:52<7:16:15,  1.63s/it]
 16%|█▌        | 3042/19090 [1:28:55<8:50:18,  1.98s/it]
 16%|█▌        | 3043/19090 [1:28:56<7:54:13,  1.77s/it]
 16%|█▌        | 3044/19090 [1:28:57<7:15:03,  1.63s/it]
 16%|█▌        | 3045/19090 [1:29:00<8:39:18,  1.94s/it]
 16%|█▌        | 3046/19090 [1:29:01<7:46:33,  1.74s/it]
 16%|█▌        | 3047/19090 [1:29:03<7:09:47,  1.61s/it]
 16%|█▌        | 3048/19090 [1:29:05<8:35:22,  1.93s/it]
 16%|█▌        | 3049/19090 [1:29:07<7:44:51,  1.74s/it]
 16%|█▌        | 3050/19090 [1:29:08<7:18:50,  1.64s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.3889630138874054, 'learning_rate': 4.945881042130027e-05, 'recon_loss': 0.0030771427846048027, 'recon_static': 0.002118947863345966, 'recon_motion': 0.0009581948703271337, 'recon_norm': 1.9347954660654068, 'recon_norm_std': 0.08674854543060065, 'target_norm': 1.6210175424814224, 'recon_cos': 0.47222789749503136, 'gradnorm/compressor.query': 0.005696168559926245, 'gradnorm/compressor.layer0': 0.10215771014867055, 'gradnorm/compressor.layer1': 0.13914414698551805, 'gradnorm/compressor.layer2': 0.1171482417787734, 'gradnorm/compressor.layer3': 0.11230729556165124, 'gradnorm/compressor.layer4': 0.10426106248851787, 'gradnorm/compressor.layer5': 0.09874338146706908, 'gradnorm/compressor.layer6': 0.09435013077864052, 'gradnorm/compressor.layer7': 0.09296489835822724, 'gradnorm/decoder.layer0': 0.12012671998819054, 'gradnorm/decoder.layer1': 0.11528075626398887, 'gradnorm/decoder.layer2': 0.17099917031597167, 'gradnorm/max_param': 0.12367423623800278, 'epoch': 1.6}

 16%|█▌        | 3050/19090 [1:29:08<7:18:50,  1.64s/it]
 16%|█▌        | 3051/19090 [1:29:11<8:40:35,  1.95s/it]
 16%|█▌        | 3052/19090 [1:29:12<7:47:40,  1.75s/it]
 16%|█▌        | 3053/19090 [1:29:13<7:11:01,  1.61s/it]
 16%|█▌        | 3054/19090 [1:29:16<8:36:46,  1.93s/it]
 16%|█▌        | 3055/19090 [1:29:17<7:45:50,  1.74s/it]
 16%|█▌        | 3056/19090 [1:29:19<7:16:54,  1.63s/it]
 16%|█▌        | 3057/19090 [1:29:21<8:27:22,  1.90s/it]
 16%|█▌        | 3058/19090 [1:29:22<7:38:21,  1.72s/it]
 16%|█▌        | 3059/19090 [1:29:24<7:06:52,  1.60s/it]
 16%|█▌        | 3060/19090 [1:29:26<8:32:57,  1.92s/it]
                                                        
{'loss': 0.0028, 'grad_norm': 0.3628152310848236, 'learning_rate': 4.944930938833535e-05, 'recon_loss': 0.0027964171778876333, 'recon_static': 0.0018305081030121073, 'recon_motion': 0.0009659091199864633, 'recon_norm': 1.8415230065584183, 'recon_norm_std': 0.09068530909717083, 'target_norm': 1.624395802617073, 'recon_cos': 0.4924321912229061, 'gradnorm/compressor.query': 0.00521793958452559, 'gradnorm/compressor.layer0': 0.09477351195150313, 'gradnorm/compressor.layer1': 0.12660053146822878, 'gradnorm/compressor.layer2': 0.10966915782171746, 'gradnorm/compressor.layer3': 0.10531851435079138, 'gradnorm/compressor.layer4': 0.09778983040069501, 'gradnorm/compressor.layer5': 0.09394061558614492, 'gradnorm/compressor.layer6': 0.08770121987800295, 'gradnorm/compressor.layer7': 0.08601260425502129, 'gradnorm/decoder.layer0': 0.11095059529101184, 'gradnorm/decoder.layer1': 0.1090164395474543, 'gradnorm/decoder.layer2': 0.16064451965517507, 'gradnorm/max_param': 0.11482604593038559, 'epoch': 1.6}

 16%|█▌        | 3060/19090 [1:29:26<8:32:57,  1.92s/it]
 16%|█▌        | 3061/19090 [1:29:28<7:42:08,  1.73s/it]
 16%|█▌        | 3062/19090 [1:29:29<7:13:51,  1.62s/it]
 16%|█▌        | 3063/19090 [1:29:32<8:59:02,  2.02s/it]
 16%|█▌        | 3064/19090 [1:29:33<8:00:47,  1.80s/it]
 16%|█▌        | 3065/19090 [1:29:35<7:19:18,  1.64s/it]
 16%|█▌        | 3066/19090 [1:29:37<8:41:55,  1.95s/it]
 16%|█▌        | 3067/19090 [1:29:38<7:48:14,  1.75s/it]
 16%|█▌        | 3068/19090 [1:29:40<7:18:00,  1.64s/it]
 16%|█▌        | 3069/19090 [1:29:42<8:35:41,  1.93s/it]
 16%|█▌        | 3070/19090 [1:29:44<7:44:08,  1.74s/it]
                                                        
{'loss': 0.004, 'grad_norm': 0.5777824521064758, 'learning_rate': 4.943972660883699e-05, 'recon_loss': 0.003972099319798872, 'recon_static': 0.0030139519658405334, 'recon_motion': 0.0009581471167621203, 'recon_norm': 2.178979516029358, 'recon_norm_std': 0.07902142126113176, 'target_norm': 1.6198889404535293, 'recon_cos': 0.42211997881531715, 'gradnorm/compressor.query': 0.008949401737562715, 'gradnorm/compressor.layer0': 0.15396356798180827, 'gradnorm/compressor.layer1': 0.2018832618983249, 'gradnorm/compressor.layer2': 0.17434296819658948, 'gradnorm/compressor.layer3': 0.16472462602996407, 'gradnorm/compressor.layer4': 0.15591923154682485, 'gradnorm/compressor.layer5': 0.1472390950335068, 'gradnorm/compressor.layer6': 0.14076198422769617, 'gradnorm/compressor.layer7': 0.13781389581747613, 'gradnorm/decoder.layer0': 0.17823844735198588, 'gradnorm/decoder.layer1': 0.17525344658114, 'gradnorm/decoder.layer2': 0.2539324673541948, 'gradnorm/max_param': 0.1824609786272049, 'epoch': 1.61}

 16%|█▌        | 3070/19090 [1:29:44<7:44:08,  1.74s/it]
 16%|█▌        | 3071/19090 [1:29:45<7:10:47,  1.61s/it]
 16%|█▌        | 3072/19090 [1:29:48<8:40:36,  1.95s/it]
 16%|█▌        | 3073/19090 [1:29:49<7:47:07,  1.75s/it]
 16%|█▌        | 3074/19090 [1:29:51<7:27:38,  1.68s/it]
 16%|█▌        | 3075/19090 [1:29:53<8:50:34,  1.99s/it]
 16%|█▌        | 3076/19090 [1:29:55<7:55:06,  1.78s/it]
 16%|█▌        | 3077/19090 [1:29:56<7:15:48,  1.63s/it]
 16%|█▌        | 3078/19090 [1:29:59<8:56:03,  2.01s/it]
 16%|█▌        | 3079/19090 [1:30:00<7:58:30,  1.79s/it]
 16%|█▌        | 3080/19090 [1:30:01<7:18:45,  1.64s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.5316411852836609, 'learning_rate': 4.943006211484531e-05, 'recon_loss': 0.004632623225916177, 'recon_static': 0.003673334122868255, 'recon_motion': 0.0009592891292413697, 'recon_norm': 2.3529848635196684, 'recon_norm_std': 0.07328330986201763, 'target_norm': 1.620538556575775, 'recon_cos': 0.3886843338608742, 'gradnorm/compressor.query': 0.008227136022645864, 'gradnorm/compressor.layer0': 0.13692264339532217, 'gradnorm/compressor.layer1': 0.186318459649986, 'gradnorm/compressor.layer2': 0.15896098201460676, 'gradnorm/compressor.layer3': 0.1527842120841028, 'gradnorm/compressor.layer4': 0.14221711374362922, 'gradnorm/compressor.layer5': 0.1361782835406064, 'gradnorm/compressor.layer6': 0.13027684237307136, 'gradnorm/compressor.layer7': 0.12413986628344367, 'gradnorm/decoder.layer0': 0.1604255296583083, 'gradnorm/decoder.layer1': 0.16057787036792046, 'gradnorm/decoder.layer2': 0.24047054529294487, 'gradnorm/max_param': 0.16895288228988647, 'epoch': 1.61}

 16%|█▌        | 3080/19090 [1:30:01<7:18:45,  1.64s/it]
 16%|█▌        | 3081/19090 [1:30:04<8:50:45,  1.99s/it]
 16%|█▌        | 3082/19090 [1:30:05<7:55:05,  1.78s/it]
 16%|█▌        | 3083/19090 [1:30:07<7:19:58,  1.65s/it]
 16%|█▌        | 3084/19090 [1:30:10<8:59:34,  2.02s/it]
 16%|█▌        | 3085/19090 [1:30:11<8:00:59,  1.80s/it]
 16%|█▌        | 3086/19090 [1:30:12<7:21:31,  1.66s/it]
 16%|█▌        | 3087/19090 [1:30:15<8:43:09,  1.96s/it]
 16%|█▌        | 3088/19090 [1:30:16<7:48:59,  1.76s/it]
 16%|█▌        | 3089/19090 [1:30:18<7:10:53,  1.62s/it]
 16%|█▌        | 3090/19090 [1:30:21<9:08:48,  2.06s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.45122167468070984, 'learning_rate': 4.942031593867366e-05, 'recon_loss': 0.004180763760814443, 'recon_static': 0.003210621350444853, 'recon_motion': 0.0009701424249215051, 'recon_norm': 2.232513391971588, 'recon_norm_std': 0.07556599630042911, 'target_norm': 1.6224611908197404, 'recon_cos': 0.40799530074000356, 'gradnorm/compressor.query': 0.006288426912996311, 'gradnorm/compressor.layer0': 0.1165290173937913, 'gradnorm/compressor.layer1': 0.1581129888733571, 'gradnorm/compressor.layer2': 0.13710540863560083, 'gradnorm/compressor.layer3': 0.13184820047495138, 'gradnorm/compressor.layer4': 0.12009753477394709, 'gradnorm/compressor.layer5': 0.11513676388572217, 'gradnorm/compressor.layer6': 0.11154757587534883, 'gradnorm/compressor.layer7': 0.10633750583348897, 'gradnorm/decoder.layer0': 0.1397695682943826, 'gradnorm/decoder.layer1': 0.13489536591550738, 'gradnorm/decoder.layer2': 0.19909339588668484, 'gradnorm/max_param': 0.14428706467151642, 'epoch': 1.62}

 16%|█▌        | 3090/19090 [1:30:21<9:08:48,  2.06s/it]
 16%|█▌        | 3091/19090 [1:30:22<8:08:04,  1.83s/it]
 16%|█▌        | 3092/19090 [1:30:23<7:25:44,  1.67s/it]
 16%|█▌        | 3093/19090 [1:30:26<8:52:37,  2.00s/it]
 16%|█▌        | 3094/19090 [1:30:27<7:55:54,  1.79s/it]
 16%|█▌        | 3095/19090 [1:30:29<7:16:00,  1.64s/it]
 16%|█▌        | 3096/19090 [1:30:31<8:21:18,  1.88s/it]
 16%|█▌        | 3097/19090 [1:30:32<7:33:43,  1.70s/it]
 16%|█▌        | 3098/19090 [1:30:34<7:00:26,  1.58s/it]
 16%|█▌        | 3099/19090 [1:30:36<8:39:29,  1.95s/it]
 16%|█▌        | 3100/19090 [1:30:38<7:46:46,  1.75s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.35700735449790955, 'learning_rate': 4.9410488112908466e-05, 'recon_loss': 0.003025551250902936, 'recon_static': 0.0020658213179558516, 'recon_motion': 0.0009597298965672962, 'recon_norm': 1.9165276408195495, 'recon_norm_std': 0.08676128257066011, 'target_norm': 1.62155838906765, 'recon_cos': 0.4767653450369835, 'gradnorm/compressor.query': 0.00514220678487686, 'gradnorm/compressor.layer0': 0.09260549905130991, 'gradnorm/compressor.layer1': 0.12579777324736294, 'gradnorm/compressor.layer2': 0.10877212772705526, 'gradnorm/compressor.layer3': 0.10194033253825618, 'gradnorm/compressor.layer4': 0.09536550291018397, 'gradnorm/compressor.layer5': 0.08967501406369073, 'gradnorm/compressor.layer6': 0.08663284690795017, 'gradnorm/compressor.layer7': 0.08488588471082137, 'gradnorm/decoder.layer0': 0.10903256745415796, 'gradnorm/decoder.layer1': 0.10758342732623126, 'gradnorm/decoder.layer2': 0.15966964277503976, 'gradnorm/max_param': 0.11380741745233536, 'epoch': 1.62}

 16%|█▌        | 3100/19090 [1:30:38<7:46:46,  1.75s/it]
 16%|█▌        | 3101/19090 [1:30:39<7:10:05,  1.61s/it]
 16%|█▌        | 3102/19090 [1:30:42<8:30:28,  1.92s/it]
 16%|█▋        | 3103/19090 [1:30:43<7:40:16,  1.73s/it]
 16%|█▋        | 3104/19090 [1:30:44<7:05:08,  1.60s/it]
 16%|█▋        | 3105/19090 [1:30:47<8:40:18,  1.95s/it]
 16%|█▋        | 3106/19090 [1:30:48<7:46:41,  1.75s/it]
 16%|█▋        | 3107/19090 [1:30:50<7:09:35,  1.61s/it]
 16%|█▋        | 3108/19090 [1:30:52<8:48:22,  1.98s/it]
 16%|█▋        | 3109/19090 [1:30:54<7:52:05,  1.77s/it]
 16%|█▋        | 3110/19090 [1:30:55<7:13:03,  1.63s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.42296847701072693, 'learning_rate': 4.940057867040919e-05, 'recon_loss': 0.0029579803231172264, 'recon_static': 0.002006812210311182, 'recon_motion': 0.0009511682350421324, 'recon_norm': 1.901547548174858, 'recon_norm_std': 0.08796638138592243, 'target_norm': 1.6195296138525008, 'recon_cos': 0.4821797117590904, 'gradnorm/compressor.query': 0.0067720812753525594, 'gradnorm/compressor.layer0': 0.11520684042599107, 'gradnorm/compressor.layer1': 0.1496536042594313, 'gradnorm/compressor.layer2': 0.12864126800307835, 'gradnorm/compressor.layer3': 0.12050151978544399, 'gradnorm/compressor.layer4': 0.11349644035453173, 'gradnorm/compressor.layer5': 0.10539789380484765, 'gradnorm/compressor.layer6': 0.1031991195551169, 'gradnorm/compressor.layer7': 0.10259074770449979, 'gradnorm/decoder.layer0': 0.12923954012279149, 'gradnorm/decoder.layer1': 0.12726008644955222, 'gradnorm/decoder.layer2': 0.18454104398339713, 'gradnorm/max_param': 0.13406652212142944, 'epoch': 1.63}

 16%|█▋        | 3110/19090 [1:30:55<7:13:03,  1.63s/it]
 16%|█▋        | 3111/19090 [1:30:58<8:41:27,  1.96s/it]
 16%|█▋        | 3112/19090 [1:30:59<7:47:33,  1.76s/it]
 16%|█▋        | 3113/19090 [1:31:00<7:10:26,  1.62s/it]
 16%|█▋        | 3114/19090 [1:31:03<8:44:46,  1.97s/it]
 16%|█▋        | 3115/19090 [1:31:04<7:49:37,  1.76s/it]
 16%|█▋        | 3116/19090 [1:31:06<7:10:50,  1.62s/it]
 16%|█▋        | 3117/19090 [1:31:08<8:45:26,  1.97s/it]
 16%|█▋        | 3118/19090 [1:31:10<7:51:01,  1.77s/it]
 16%|█▋        | 3119/19090 [1:31:11<7:13:02,  1.63s/it]
 16%|█▋        | 3120/19090 [1:31:14<8:28:19,  1.91s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.38677114248275757, 'learning_rate': 4.939058764430816e-05, 'recon_loss': 0.003208864526823163, 'recon_static': 0.002256733819376677, 'recon_motion': 0.0009521306972601451, 'recon_norm': 1.9766417145729065, 'recon_norm_std': 0.0855972059071064, 'target_norm': 1.620543423295021, 'recon_cos': 0.46556226834654807, 'gradnorm/compressor.query': 0.005938229094541509, 'gradnorm/compressor.layer0': 0.10520341398600858, 'gradnorm/compressor.layer1': 0.14109480053752427, 'gradnorm/compressor.layer2': 0.11791013199188453, 'gradnorm/compressor.layer3': 0.10939592774311487, 'gradnorm/compressor.layer4': 0.10152583193918371, 'gradnorm/compressor.layer5': 0.09639865797434842, 'gradnorm/compressor.layer6': 0.09434942500523465, 'gradnorm/compressor.layer7': 0.08985450410750051, 'gradnorm/decoder.layer0': 0.11678562795906972, 'gradnorm/decoder.layer1': 0.116081365910565, 'gradnorm/decoder.layer2': 0.17035637036777046, 'gradnorm/max_param': 0.12300658226013184, 'epoch': 1.63}

 16%|█▋        | 3120/19090 [1:31:14<8:28:19,  1.91s/it]
 16%|█▋        | 3121/19090 [1:31:15<7:38:27,  1.72s/it]
 16%|█▋        | 3122/19090 [1:31:16<7:02:59,  1.59s/it]
 16%|█▋        | 3123/19090 [1:31:19<8:50:15,  1.99s/it]
 16%|█▋        | 3124/19090 [1:31:20<7:54:30,  1.78s/it]
 16%|█▋        | 3125/19090 [1:31:22<7:15:33,  1.64s/it]
 16%|█▋        | 3126/19090 [1:31:24<8:43:23,  1.97s/it]
 16%|█▋        | 3127/19090 [1:31:26<7:49:08,  1.76s/it]
 16%|█▋        | 3128/19090 [1:31:27<7:11:21,  1.62s/it]
 16%|█▋        | 3129/19090 [1:31:30<8:33:26,  1.93s/it]
 16%|█▋        | 3130/19090 [1:31:31<7:42:05,  1.74s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.5194921493530273, 'learning_rate': 4.938051506801048e-05, 'recon_loss': 0.003304876066977158, 'recon_static': 0.002345834582229145, 'recon_motion': 0.0009590415575075895, 'recon_norm': 1.9945635378360749, 'recon_norm_std': 0.08313743099570274, 'target_norm': 1.6203959196805955, 'recon_cos': 0.4592571876943111, 'gradnorm/compressor.query': 0.007959449716432788, 'gradnorm/compressor.layer0': 0.1352307199924127, 'gradnorm/compressor.layer1': 0.1871311414000055, 'gradnorm/compressor.layer2': 0.15893357798679666, 'gradnorm/compressor.layer3': 0.1510491629636681, 'gradnorm/compressor.layer4': 0.13908436367595325, 'gradnorm/compressor.layer5': 0.1333594629704208, 'gradnorm/compressor.layer6': 0.12569152018606536, 'gradnorm/compressor.layer7': 0.12079637684421457, 'gradnorm/decoder.layer0': 0.15796058807266478, 'gradnorm/decoder.layer1': 0.15299551875705963, 'gradnorm/decoder.layer2': 0.22923585901981938, 'gradnorm/max_param': 0.16476987302303314, 'epoch': 1.64}

 16%|█▋        | 3130/19090 [1:31:31<7:42:05,  1.74s/it]
 16%|█▋        | 3131/19090 [1:31:32<7:10:24,  1.62s/it]
 16%|█▋        | 3132/19090 [1:31:35<8:58:39,  2.03s/it]
 16%|█▋        | 3133/19090 [1:31:37<8:00:31,  1.81s/it]
 16%|█▋        | 3134/19090 [1:31:38<7:19:11,  1.65s/it]
 16%|█▋        | 3135/19090 [1:31:41<9:04:32,  2.05s/it]
 16%|█▋        | 3136/19090 [1:31:42<8:03:34,  1.82s/it]
 16%|█▋        | 3137/19090 [1:31:43<7:20:58,  1.66s/it]
 16%|█▋        | 3138/19090 [1:31:46<9:07:08,  2.06s/it]
 16%|█▋        | 3139/19090 [1:31:48<8:05:23,  1.83s/it]
 16%|█▋        | 3140/19090 [1:31:49<7:22:01,  1.66s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.40152809023857117, 'learning_rate': 4.9370360975193906e-05, 'recon_loss': 0.0034615172713529318, 'recon_static': 0.002492278360296041, 'recon_motion': 0.0009692389750853181, 'recon_norm': 2.041146716475487, 'recon_norm_std': 0.08126781359314919, 'target_norm': 1.623519441485405, 'recon_cos': 0.4485412947833538, 'gradnorm/compressor.query': 0.006055232040852038, 'gradnorm/compressor.layer0': 0.10607039193063798, 'gradnorm/compressor.layer1': 0.1427645052974179, 'gradnorm/compressor.layer2': 0.1231050618664425, 'gradnorm/compressor.layer3': 0.11607034978784457, 'gradnorm/compressor.layer4': 0.10818015708636843, 'gradnorm/compressor.layer5': 0.10167755890448993, 'gradnorm/compressor.layer6': 0.09829118788316075, 'gradnorm/compressor.layer7': 0.09502443691030124, 'gradnorm/decoder.layer0': 0.12181865691894385, 'gradnorm/decoder.layer1': 0.1202542678061783, 'gradnorm/decoder.layer2': 0.1757707857620053, 'gradnorm/max_param': 0.12719383835792542, 'epoch': 1.64}

 16%|█▋        | 3140/19090 [1:31:49<7:22:01,  1.66s/it]
 16%|█▋        | 3141/19090 [1:31:52<8:54:44,  2.01s/it]
 16%|█▋        | 3142/19090 [1:31:53<7:56:58,  1.79s/it]
 16%|█▋        | 3143/19090 [1:31:54<7:16:35,  1.64s/it]
 16%|█▋        | 3144/19090 [1:31:57<8:21:35,  1.89s/it]
 16%|█▋        | 3145/19090 [1:31:58<7:33:25,  1.71s/it]
 16%|█▋        | 3146/19090 [1:31:59<7:00:12,  1.58s/it]
 16%|█▋        | 3147/19090 [1:32:02<8:35:31,  1.94s/it]
 16%|█▋        | 3148/19090 [1:32:03<7:43:16,  1.74s/it]
 16%|█▋        | 3149/19090 [1:32:05<7:06:41,  1.61s/it]
 17%|█▋        | 3150/19090 [1:32:07<8:34:26,  1.94s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.35284116864204407, 'learning_rate': 4.9360125399808784e-05, 'recon_loss': 0.0029904509545303883, 'recon_static': 0.0020381621172418816, 'recon_motion': 0.0009522888678475284, 'recon_norm': 1.908245027065277, 'recon_norm_std': 0.08821396026760339, 'target_norm': 1.6200854301452636, 'recon_cos': 0.47920892983675, 'gradnorm/compressor.query': 0.005425772604723449, 'gradnorm/compressor.layer0': 0.0899559543536381, 'gradnorm/compressor.layer1': 0.1235145544129988, 'gradnorm/compressor.layer2': 0.10749793827635883, 'gradnorm/compressor.layer3': 0.10093579972904855, 'gradnorm/compressor.layer4': 0.09444622460268105, 'gradnorm/compressor.layer5': 0.08979064470717824, 'gradnorm/compressor.layer6': 0.08605207045320708, 'gradnorm/compressor.layer7': 0.0845248166049439, 'gradnorm/decoder.layer0': 0.10967402369292047, 'gradnorm/decoder.layer1': 0.10695694533436201, 'gradnorm/decoder.layer2': 0.15612706474201807, 'gradnorm/max_param': 0.1126188263297081, 'epoch': 1.65}

 17%|█▋        | 3150/19090 [1:32:07<8:34:26,  1.94s/it]
 17%|█▋        | 3151/19090 [1:32:09<7:43:05,  1.74s/it]
 17%|█▋        | 3152/19090 [1:32:10<7:06:25,  1.61s/it]
 17%|█▋        | 3153/19090 [1:32:13<8:29:31,  1.92s/it]
 17%|█▋        | 3154/19090 [1:32:14<7:38:56,  1.73s/it]
 17%|█▋        | 3155/19090 [1:32:15<7:03:42,  1.60s/it]
 17%|█▋        | 3156/19090 [1:32:19<9:27:24,  2.14s/it]
 17%|█▋        | 3157/19090 [1:32:20<8:19:43,  1.88s/it]
 17%|█▋        | 3158/19090 [1:32:21<7:31:57,  1.70s/it]
 17%|█▋        | 3159/19090 [1:32:24<9:05:10,  2.05s/it]
 17%|█▋        | 3160/19090 [1:32:25<8:04:21,  1.82s/it]
                                                        
{'loss': 0.0026, 'grad_norm': 0.39758962392807007, 'learning_rate': 4.9349808376077846e-05, 'recon_loss': 0.0026345149730332197, 'recon_static': 0.0016860646399436518, 'recon_motion': 0.0009484503651037812, 'recon_norm': 1.8040638446807862, 'recon_norm_std': 0.0918845858424902, 'target_norm': 1.6189902126789093, 'recon_cos': 0.5100634224712849, 'gradnorm/compressor.query': 0.005759044558019731, 'gradnorm/compressor.layer0': 0.10435275106230826, 'gradnorm/compressor.layer1': 0.14237077214701363, 'gradnorm/compressor.layer2': 0.12379597658714611, 'gradnorm/compressor.layer3': 0.11569446107570538, 'gradnorm/compressor.layer4': 0.10648143662801703, 'gradnorm/compressor.layer5': 0.10027294627664851, 'gradnorm/compressor.layer6': 0.0971410464844772, 'gradnorm/compressor.layer7': 0.09401708050495569, 'gradnorm/decoder.layer0': 0.12125404193251574, 'gradnorm/decoder.layer1': 0.11745654058266715, 'gradnorm/decoder.layer2': 0.173227074704716, 'gradnorm/max_param': 0.1255541443824768, 'epoch': 1.66}

 17%|█▋        | 3160/19090 [1:32:25<8:04:21,  1.82s/it]
 17%|█▋        | 3161/19090 [1:32:27<7:21:30,  1.66s/it]
 17%|█▋        | 3162/19090 [1:32:30<9:03:25,  2.05s/it]
 17%|█▋        | 3163/19090 [1:32:31<8:03:16,  1.82s/it]
 17%|█▋        | 3164/19090 [1:32:32<7:21:34,  1.66s/it]
 17%|█▋        | 3165/19090 [1:32:35<8:41:01,  1.96s/it]
 17%|█▋        | 3166/19090 [1:32:36<7:47:26,  1.76s/it]
 17%|█▋        | 3167/19090 [1:32:37<7:09:40,  1.62s/it]
 17%|█▋        | 3168/19090 [1:32:40<8:32:31,  1.93s/it]
 17%|█▋        | 3169/19090 [1:32:41<7:41:48,  1.74s/it]
 17%|█▋        | 3170/19090 [1:32:43<7:06:19,  1.61s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.5544440150260925, 'learning_rate': 4.933940993849619e-05, 'recon_loss': 0.003802337346132845, 'recon_static': 0.002855459682177752, 'recon_motion': 0.00094687782402616, 'recon_norm': 2.1441639333963396, 'recon_norm_std': 0.0792370893061161, 'target_norm': 1.6167537242174148, 'recon_cos': 0.4306220792233944, 'gradnorm/compressor.query': 0.008247999631712313, 'gradnorm/compressor.layer0': 0.14597800152288554, 'gradnorm/compressor.layer1': 0.19674244043412503, 'gradnorm/compressor.layer2': 0.17309987825811415, 'gradnorm/compressor.layer3': 0.1616405924764299, 'gradnorm/compressor.layer4': 0.14945173723540656, 'gradnorm/compressor.layer5': 0.14135963556799683, 'gradnorm/compressor.layer6': 0.1355984152830691, 'gradnorm/compressor.layer7': 0.13022469576706538, 'gradnorm/decoder.layer0': 0.1674950095668076, 'gradnorm/decoder.layer1': 0.16416027357289004, 'gradnorm/decoder.layer2': 0.24200154672031846, 'gradnorm/max_param': 0.1750912368297577, 'epoch': 1.66}

 17%|█▋        | 3170/19090 [1:32:43<7:06:19,  1.61s/it]
 17%|█▋        | 3171/19090 [1:32:46<8:52:11,  2.01s/it]
 17%|█▋        | 3172/19090 [1:32:47<7:54:23,  1.79s/it]
 17%|█▋        | 3173/19090 [1:32:48<7:14:25,  1.64s/it]
 17%|█▋        | 3174/19090 [1:32:51<8:27:34,  1.91s/it]
 17%|█▋        | 3175/19090 [1:32:52<7:37:31,  1.72s/it]
 17%|█▋        | 3176/19090 [1:32:53<7:01:58,  1.59s/it]
 17%|█▋        | 3177/19090 [1:32:56<8:25:28,  1.91s/it]
 17%|█▋        | 3178/19090 [1:32:57<7:37:32,  1.73s/it]
 17%|█▋        | 3179/19090 [1:32:58<7:02:04,  1.59s/it]
 17%|█▋        | 3180/19090 [1:33:01<8:34:42,  1.94s/it]
                                                        
{'loss': 0.004, 'grad_norm': 0.49063509702682495, 'learning_rate': 4.9328930121831104e-05, 'recon_loss': 0.003971443470800295, 'recon_static': 0.0030155997374095023, 'recon_motion': 0.0009558437770465389, 'recon_norm': 2.1856651842594146, 'recon_norm_std': 0.07555852644145489, 'target_norm': 1.6190788000822067, 'recon_cos': 0.41841228902339933, 'gradnorm/compressor.query': 0.007009054415926935, 'gradnorm/compressor.layer0': 0.13006843840127447, 'gradnorm/compressor.layer1': 0.1750472320100807, 'gradnorm/compressor.layer2': 0.15174769290485338, 'gradnorm/compressor.layer3': 0.1440500483955208, 'gradnorm/compressor.layer4': 0.13211826893461587, 'gradnorm/compressor.layer5': 0.12387660864397568, 'gradnorm/compressor.layer6': 0.11935432607696488, 'gradnorm/compressor.layer7': 0.11632667916566998, 'gradnorm/decoder.layer0': 0.14814521849742548, 'gradnorm/decoder.layer1': 0.14479675907218048, 'gradnorm/decoder.layer2': 0.2141154516120985, 'gradnorm/max_param': 0.15555039048194885, 'epoch': 1.67}

 17%|█▋        | 3180/19090 [1:33:01<8:34:42,  1.94s/it]
 17%|█▋        | 3181/19090 [1:33:03<7:42:26,  1.74s/it]
 17%|█▋        | 3182/19090 [1:33:04<7:10:01,  1.62s/it]
 17%|█▋        | 3183/19090 [1:33:06<8:19:50,  1.89s/it]
 17%|█▋        | 3184/19090 [1:33:08<7:32:03,  1.71s/it]
 17%|█▋        | 3185/19090 [1:33:09<7:30:39,  1.70s/it]
 17%|█▋        | 3186/19090 [1:33:12<8:29:35,  1.92s/it]
 17%|█▋        | 3187/19090 [1:33:13<7:39:35,  1.73s/it]
 17%|█▋        | 3188/19090 [1:33:15<7:35:35,  1.72s/it]
 17%|█▋        | 3189/19090 [1:33:17<8:19:42,  1.89s/it]
 17%|█▋        | 3190/19090 [1:33:18<7:32:14,  1.71s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.47756317257881165, 'learning_rate': 4.931836896112196e-05, 'recon_loss': 0.0037758066726382824, 'recon_static': 0.0028007233457174154, 'recon_motion': 0.0009750833472935483, 'recon_norm': 2.1296339213848112, 'recon_norm_std': 0.07768413200974464, 'target_norm': 1.6247957050800323, 'recon_cos': 0.428263983130455, 'gradnorm/compressor.query': 0.007201048193152161, 'gradnorm/compressor.layer0': 0.12506379794210298, 'gradnorm/compressor.layer1': 0.17142975095785887, 'gradnorm/compressor.layer2': 0.14839127724391607, 'gradnorm/compressor.layer3': 0.14043984542207255, 'gradnorm/compressor.layer4': 0.1259199429217181, 'gradnorm/compressor.layer5': 0.12251952867159786, 'gradnorm/compressor.layer6': 0.11591552805721304, 'gradnorm/compressor.layer7': 0.1131444795971566, 'gradnorm/decoder.layer0': 0.1445773540885653, 'gradnorm/decoder.layer1': 0.14118509544192628, 'gradnorm/decoder.layer2': 0.2080901077164857, 'gradnorm/max_param': 0.15075428783893585, 'epoch': 1.67}

 17%|█▋        | 3190/19090 [1:33:18<7:32:14,  1.71s/it]
 17%|█▋        | 3191/19090 [1:33:20<7:29:02,  1.69s/it]
 17%|█▋        | 3192/19090 [1:33:23<8:40:54,  1.97s/it]
 17%|█▋        | 3193/19090 [1:33:24<7:46:53,  1.76s/it]
 17%|█▋        | 3194/19090 [1:33:25<7:33:18,  1.71s/it]
 17%|█▋        | 3195/19090 [1:33:28<8:21:17,  1.89s/it]
 17%|█▋        | 3196/19090 [1:33:29<7:32:48,  1.71s/it]
 17%|█▋        | 3197/19090 [1:33:31<7:12:30,  1.63s/it]
 17%|█▋        | 3198/19090 [1:33:33<8:38:04,  1.96s/it]
 17%|█▋        | 3199/19090 [1:33:35<7:44:45,  1.75s/it]
 17%|█▋        | 3200/19090 [1:33:36<7:07:04,  1.61s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.38867032527923584, 'learning_rate': 4.930772649168012e-05, 'recon_loss': 0.003229837608523667, 'recon_static': 0.002265510498546064, 'recon_motion': 0.0009643271929235198, 'recon_norm': 1.978251752257347, 'recon_norm_std': 0.08485388234257699, 'target_norm': 1.6214277535676955, 'recon_cos': 0.4623539850115776, 'gradnorm/compressor.query': 0.005912289137552191, 'gradnorm/compressor.layer0': 0.10343528725023482, 'gradnorm/compressor.layer1': 0.14078765972038298, 'gradnorm/compressor.layer2': 0.11830585421624415, 'gradnorm/compressor.layer3': 0.11250174276008712, 'gradnorm/compressor.layer4': 0.10434590555188362, 'gradnorm/compressor.layer5': 0.09777532946790306, 'gradnorm/compressor.layer6': 0.09533826214060703, 'gradnorm/compressor.layer7': 0.09142998159074917, 'gradnorm/decoder.layer0': 0.11732281978224632, 'gradnorm/decoder.layer1': 0.11451420586266585, 'gradnorm/decoder.layer2': 0.1705335044949489, 'gradnorm/max_param': 0.12393835932016373, 'epoch': 1.68}

 17%|█▋        | 3200/19090 [1:33:36<7:07:04,  1.61s/it]
 17%|█▋        | 3201/19090 [1:33:39<8:51:46,  2.01s/it]
 17%|█▋        | 3202/19090 [1:33:40<7:54:27,  1.79s/it]
 17%|█▋        | 3203/19090 [1:33:41<7:13:44,  1.64s/it]
 17%|█▋        | 3204/19090 [1:33:44<8:39:34,  1.96s/it]
 17%|█▋        | 3205/19090 [1:33:45<7:45:26,  1.76s/it]
 17%|█▋        | 3206/19090 [1:33:47<7:07:50,  1.62s/it]
 17%|█▋        | 3207/19090 [1:33:49<8:37:58,  1.96s/it]
 17%|█▋        | 3208/19090 [1:33:51<7:44:20,  1.75s/it]
 17%|█▋        | 3209/19090 [1:33:52<7:13:21,  1.64s/it]
 17%|█▋        | 3210/19090 [1:33:55<9:06:46,  2.07s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.4108823239803314, 'learning_rate': 4.9297002749088804e-05, 'recon_loss': 0.003095483424840495, 'recon_static': 0.002147604519268498, 'recon_motion': 0.000947879019076936, 'recon_norm': 1.9428268849849701, 'recon_norm_std': 0.08588931262493134, 'target_norm': 1.6183816611766815, 'recon_cos': 0.47131826281547545, 'gradnorm/compressor.query': 0.006405352676455154, 'gradnorm/compressor.layer0': 0.10966788399778847, 'gradnorm/compressor.layer1': 0.14716751387132845, 'gradnorm/compressor.layer2': 0.12654370533357429, 'gradnorm/compressor.layer3': 0.11781780488999717, 'gradnorm/compressor.layer4': 0.10842452293848033, 'gradnorm/compressor.layer5': 0.10244175517238255, 'gradnorm/compressor.layer6': 0.09801182386324188, 'gradnorm/compressor.layer7': 0.09690344569747261, 'gradnorm/decoder.layer0': 0.1251772577416237, 'gradnorm/decoder.layer1': 0.12340707132264675, 'gradnorm/decoder.layer2': 0.18181902830376598, 'gradnorm/max_param': 0.130710169672966, 'epoch': 1.68}

 17%|█▋        | 3210/19090 [1:33:55<9:06:46,  2.07s/it]
 17%|█▋        | 3211/19090 [1:33:56<8:04:31,  1.83s/it]
 17%|█▋        | 3212/19090 [1:33:58<7:21:16,  1.67s/it]
 17%|█▋        | 3213/19090 [1:34:00<8:51:26,  2.01s/it]
 17%|█▋        | 3214/19090 [1:34:02<7:53:40,  1.79s/it]
 17%|█▋        | 3215/19090 [1:34:03<7:14:05,  1.64s/it]
 17%|█▋        | 3216/19090 [1:34:06<8:48:24,  2.00s/it]
 17%|█▋        | 3217/19090 [1:34:07<7:52:39,  1.79s/it]
 17%|█▋        | 3218/19090 [1:34:08<7:13:35,  1.64s/it]
 17%|█▋        | 3219/19090 [1:34:12<9:13:29,  2.09s/it]
 17%|█▋        | 3220/19090 [1:34:13<8:10:12,  1.85s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.48476094007492065, 'learning_rate': 4.928619776920296e-05, 'recon_loss': 0.0032955742732156067, 'recon_static': 0.0023395848751533777, 'recon_motion': 0.0009559894315316342, 'recon_norm': 2.0011701345443726, 'recon_norm_std': 0.08479155786335468, 'target_norm': 1.6188149988651275, 'recon_cos': 0.45936871841549876, 'gradnorm/compressor.query': 0.007048365256300365, 'gradnorm/compressor.layer0': 0.12727579778686482, 'gradnorm/compressor.layer1': 0.17285260237903624, 'gradnorm/compressor.layer2': 0.1505978135548019, 'gradnorm/compressor.layer3': 0.14283487639220566, 'gradnorm/compressor.layer4': 0.13030319731826728, 'gradnorm/compressor.layer5': 0.12330296902261784, 'gradnorm/compressor.layer6': 0.11872212591038064, 'gradnorm/compressor.layer7': 0.11353785948830315, 'gradnorm/decoder.layer0': 0.14761220360827193, 'gradnorm/decoder.layer1': 0.14182561709195604, 'gradnorm/decoder.layer2': 0.21145360116095302, 'gradnorm/max_param': 0.15308387577533722, 'epoch': 1.69}

 17%|█▋        | 3220/19090 [1:34:13<8:10:12,  1.85s/it]
 17%|█▋        | 3221/19090 [1:34:14<7:25:38,  1.68s/it]
 17%|█▋        | 3222/19090 [1:34:17<8:51:50,  2.01s/it]
 17%|█▋        | 3223/19090 [1:34:18<7:54:19,  1.79s/it]
 17%|█▋        | 3224/19090 [1:34:19<7:13:45,  1.64s/it]
 17%|█▋        | 3225/19090 [1:34:22<8:40:37,  1.97s/it]
 17%|█▋        | 3226/19090 [1:34:24<7:47:03,  1.77s/it]
 17%|█▋        | 3227/19090 [1:34:25<7:08:56,  1.62s/it]
 17%|█▋        | 3228/19090 [1:34:27<8:22:10,  1.90s/it]
 17%|█▋        | 3229/19090 [1:34:29<7:33:31,  1.72s/it]
 17%|█▋        | 3230/19090 [1:34:30<6:59:19,  1.59s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.4371190369129181, 'learning_rate': 4.927531158814914e-05, 'recon_loss': 0.003619253676151857, 'recon_static': 0.0026715914777014405, 'recon_motion': 0.0009476621824433095, 'recon_norm': 2.0963552981615066, 'recon_norm_std': 0.08305728267878294, 'target_norm': 1.6184550851583481, 'recon_cos': 0.43956447690725325, 'gradnorm/compressor.query': 0.006690614433752252, 'gradnorm/compressor.layer0': 0.11528390692722645, 'gradnorm/compressor.layer1': 0.1535316517159028, 'gradnorm/compressor.layer2': 0.13522826134114146, 'gradnorm/compressor.layer3': 0.12770803191562738, 'gradnorm/compressor.layer4': 0.11804396855495707, 'gradnorm/compressor.layer5': 0.11148923793791378, 'gradnorm/compressor.layer6': 0.10475867150917714, 'gradnorm/compressor.layer7': 0.10202156451756983, 'gradnorm/decoder.layer0': 0.13327938648723367, 'gradnorm/decoder.layer1': 0.13312306277338612, 'gradnorm/decoder.layer2': 0.19062465566072476, 'gradnorm/max_param': 0.13855236768722534, 'epoch': 1.69}

 17%|█▋        | 3230/19090 [1:34:30<6:59:19,  1.59s/it]
 17%|█▋        | 3231/19090 [1:34:33<8:26:47,  1.92s/it]
 17%|█▋        | 3232/19090 [1:34:34<7:36:53,  1.73s/it]
 17%|█▋        | 3233/19090 [1:34:35<7:01:53,  1.60s/it]
 17%|█▋        | 3234/19090 [1:34:38<8:36:03,  1.95s/it]
 17%|█▋        | 3235/19090 [1:34:39<7:43:29,  1.75s/it]
 17%|█▋        | 3236/19090 [1:34:41<7:06:23,  1.61s/it]
 17%|█▋        | 3237/19090 [1:34:43<8:00:47,  1.82s/it]
 17%|█▋        | 3238/19090 [1:34:44<7:18:40,  1.66s/it]
 17%|█▋        | 3239/19090 [1:34:45<6:49:16,  1.55s/it]
 17%|█▋        | 3240/19090 [1:34:48<7:55:06,  1.80s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.45961591601371765, 'learning_rate': 4.9264344242325436e-05, 'recon_loss': 0.0035698938474524764, 'recon_static': 0.002624535112408921, 'recon_motion': 0.000945358733588364, 'recon_norm': 2.081931209564209, 'recon_norm_std': 0.08074374198913574, 'target_norm': 1.6160863012075424, 'recon_cos': 0.4407204441726208, 'gradnorm/compressor.query': 0.00679797970757938, 'gradnorm/compressor.layer0': 0.1190537073112674, 'gradnorm/compressor.layer1': 0.16235371066203627, 'gradnorm/compressor.layer2': 0.1404653875205222, 'gradnorm/compressor.layer3': 0.13546229397702722, 'gradnorm/compressor.layer4': 0.12148314211670574, 'gradnorm/compressor.layer5': 0.11588813523164834, 'gradnorm/compressor.layer6': 0.11238700089192676, 'gradnorm/compressor.layer7': 0.1079147519129863, 'gradnorm/decoder.layer0': 0.14217240599279357, 'gradnorm/decoder.layer1': 0.1382884850737523, 'gradnorm/decoder.layer2': 0.2019558140854208, 'gradnorm/max_param': 0.14614015817642212, 'epoch': 1.7}

 17%|█▋        | 3240/19090 [1:34:48<7:55:06,  1.80s/it]
 17%|█▋        | 3241/19090 [1:34:49<7:14:48,  1.65s/it]
 17%|█▋        | 3242/19090 [1:34:50<6:46:47,  1.54s/it]
 17%|█▋        | 3243/19090 [1:34:53<7:36:08,  1.73s/it]
 17%|█▋        | 3244/19090 [1:34:54<7:03:11,  1.60s/it]
 17%|█▋        | 3245/19090 [1:34:55<6:39:18,  1.51s/it]
 17%|█▋        | 3246/19090 [1:34:58<8:04:44,  1.84s/it]
 17%|█▋        | 3247/19090 [1:34:59<7:21:40,  1.67s/it]
 17%|█▋        | 3248/19090 [1:35:00<6:51:44,  1.56s/it]
 17%|█▋        | 3249/19090 [1:35:03<8:05:46,  1.84s/it]
 17%|█▋        | 3250/19090 [1:35:04<7:23:34,  1.68s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.4316590130329132, 'learning_rate': 4.925329576840127e-05, 'recon_loss': 0.003399189148331061, 'recon_static': 0.0024467321927659214, 'recon_motion': 0.0009524569628410973, 'recon_norm': 2.0343260377645493, 'recon_norm_std': 0.08287909608334303, 'target_norm': 1.6186978340148925, 'recon_cos': 0.45203404650092127, 'gradnorm/compressor.query': 0.0065419683607957045, 'gradnorm/compressor.layer0': 0.11313165867029201, 'gradnorm/compressor.layer1': 0.15328956101232807, 'gradnorm/compressor.layer2': 0.13159825126941224, 'gradnorm/compressor.layer3': 0.12762162922724113, 'gradnorm/compressor.layer4': 0.116219299881053, 'gradnorm/compressor.layer5': 0.10830975676786564, 'gradnorm/compressor.layer6': 0.10754379809366253, 'gradnorm/compressor.layer7': 0.10065146043533255, 'gradnorm/decoder.layer0': 0.13028472010812525, 'gradnorm/decoder.layer1': 0.12942931229599214, 'gradnorm/decoder.layer2': 0.18900930389268744, 'gradnorm/max_param': 0.13702979683876038, 'epoch': 1.7}

 17%|█▋        | 3250/19090 [1:35:04<7:23:34,  1.68s/it]
 17%|█▋        | 3251/19090 [1:35:05<6:52:50,  1.56s/it]
 17%|█▋        | 3252/19090 [1:35:08<8:08:52,  1.85s/it]
 17%|█▋        | 3253/19090 [1:35:09<7:23:28,  1.68s/it]
 17%|█▋        | 3254/19090 [1:35:11<6:51:50,  1.56s/it]
 17%|█▋        | 3255/19090 [1:35:13<8:05:01,  1.84s/it]
 17%|█▋        | 3256/19090 [1:35:14<7:20:39,  1.67s/it]
 17%|█▋        | 3257/19090 [1:35:16<6:49:49,  1.55s/it]
 17%|█▋        | 3258/19090 [1:35:18<8:03:27,  1.83s/it]
 17%|█▋        | 3259/19090 [1:35:19<7:19:48,  1.67s/it]
 17%|█▋        | 3260/19090 [1:35:21<6:49:31,  1.55s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.47824186086654663, 'learning_rate': 4.9242166203317345e-05, 'recon_loss': 0.0037278077390510587, 'recon_static': 0.002772567770443857, 'recon_motion': 0.0009552400893880986, 'recon_norm': 2.1210799634456636, 'recon_norm_std': 0.07953730411827564, 'target_norm': 1.6198108047246933, 'recon_cos': 0.4313285447657108, 'gradnorm/compressor.query': 0.007547739804529168, 'gradnorm/compressor.layer0': 0.12687765225565054, 'gradnorm/compressor.layer1': 0.16878642620666645, 'gradnorm/compressor.layer2': 0.1449434806676312, 'gradnorm/compressor.layer3': 0.13788665841424755, 'gradnorm/compressor.layer4': 0.12580363654900098, 'gradnorm/compressor.layer5': 0.11989808840734073, 'gradnorm/compressor.layer6': 0.11589864635673709, 'gradnorm/compressor.layer7': 0.11398840793824681, 'gradnorm/decoder.layer0': 0.14538067505379515, 'gradnorm/decoder.layer1': 0.14299830485594658, 'gradnorm/decoder.layer2': 0.21409847869071977, 'gradnorm/max_param': 0.152414470911026, 'epoch': 1.71}

 17%|█▋        | 3260/19090 [1:35:21<6:49:31,  1.55s/it]
 17%|█▋        | 3261/19090 [1:35:23<7:27:10,  1.70s/it]
 17%|█▋        | 3262/19090 [1:35:24<6:54:40,  1.57s/it]
 17%|█▋        | 3263/19090 [1:35:25<6:31:39,  1.48s/it]
 17%|█▋        | 3264/19090 [1:35:28<7:36:36,  1.73s/it]
 17%|█▋        | 3265/19090 [1:35:29<7:00:40,  1.59s/it]
 17%|█▋        | 3266/19090 [1:35:30<6:36:04,  1.50s/it]
 17%|█▋        | 3267/19090 [1:35:33<8:11:55,  1.87s/it]
 17%|█▋        | 3268/19090 [1:35:34<7:25:31,  1.69s/it]
 17%|█▋        | 3269/19090 [1:35:36<7:12:13,  1.64s/it]
 17%|█▋        | 3270/19090 [1:35:38<8:48:15,  2.00s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.5316717028617859, 'learning_rate': 4.9230955584285475e-05, 'recon_loss': 0.004494378104573116, 'recon_static': 0.003536410559900105, 'recon_motion': 0.0009579673642292619, 'recon_norm': 2.3214051961898803, 'recon_norm_std': 0.07430609175935388, 'target_norm': 1.621467337012291, 'recon_cos': 0.39632636979222297, 'gradnorm/compressor.query': 0.0077826375779017085, 'gradnorm/compressor.layer0': 0.1402352614627118, 'gradnorm/compressor.layer1': 0.18918198915917078, 'gradnorm/compressor.layer2': 0.1651806365209901, 'gradnorm/compressor.layer3': 0.15607530245251353, 'gradnorm/compressor.layer4': 0.13977685772327028, 'gradnorm/compressor.layer5': 0.13439839542013285, 'gradnorm/compressor.layer6': 0.1293421729337994, 'gradnorm/compressor.layer7': 0.1252629673064335, 'gradnorm/decoder.layer0': 0.16305161303374777, 'gradnorm/decoder.layer1': 0.16011498082341813, 'gradnorm/decoder.layer2': 0.23080507888918292, 'gradnorm/max_param': 0.16797620058059692, 'epoch': 1.71}

 17%|█▋        | 3270/19090 [1:35:38<8:48:15,  2.00s/it]
 17%|█▋        | 3271/19090 [1:35:40<7:50:56,  1.79s/it]
 17%|█▋        | 3272/19090 [1:35:41<7:10:55,  1.63s/it]
 17%|█▋        | 3273/19090 [1:35:44<8:32:00,  1.94s/it]
 17%|█▋        | 3274/19090 [1:35:45<7:40:39,  1.75s/it]
 17%|█▋        | 3275/19090 [1:35:46<7:15:09,  1.65s/it]
 17%|█▋        | 3276/19090 [1:35:49<8:39:05,  1.97s/it]
 17%|█▋        | 3277/19090 [1:35:50<7:44:48,  1.76s/it]
 17%|█▋        | 3278/19090 [1:35:52<7:15:01,  1.65s/it]
 17%|█▋        | 3279/19090 [1:35:54<8:33:02,  1.95s/it]
 17%|█▋        | 3280/19090 [1:35:56<7:40:17,  1.75s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.4948025941848755, 'learning_rate': 4.921966394878849e-05, 'recon_loss': 0.0043435361934825775, 'recon_static': 0.0033759771671611817, 'recon_motion': 0.0009675590437836945, 'recon_norm': 2.28057182431221, 'recon_norm_std': 0.07418419234454632, 'target_norm': 1.6223142385482787, 'recon_cos': 0.4013010703027248, 'gradnorm/compressor.query': 0.007813237572662236, 'gradnorm/compressor.layer0': 0.13055005452629026, 'gradnorm/compressor.layer1': 0.1765498418405927, 'gradnorm/compressor.layer2': 0.15198732520771932, 'gradnorm/compressor.layer3': 0.14467338712548647, 'gradnorm/compressor.layer4': 0.13247928472071643, 'gradnorm/compressor.layer5': 0.1252932535883341, 'gradnorm/compressor.layer6': 0.11995821411559282, 'gradnorm/compressor.layer7': 0.11547442664877236, 'gradnorm/decoder.layer0': 0.1512251587538659, 'gradnorm/decoder.layer1': 0.14900619337071952, 'gradnorm/decoder.layer2': 0.2155910499769726, 'gradnorm/max_param': 0.15695221722126007, 'epoch': 1.72}

 17%|█▋        | 3280/19090 [1:35:56<7:40:17,  1.75s/it]
 17%|█▋        | 3281/19090 [1:35:57<7:12:34,  1.64s/it]
 17%|█▋        | 3282/19090 [1:36:00<8:25:19,  1.92s/it]
 17%|█▋        | 3283/19090 [1:36:01<7:34:52,  1.73s/it]
 17%|█▋        | 3284/19090 [1:36:03<7:28:48,  1.70s/it]
 17%|█▋        | 3285/19090 [1:36:05<8:43:55,  1.99s/it]
 17%|█▋        | 3286/19090 [1:36:07<7:48:36,  1.78s/it]
 17%|█▋        | 3287/19090 [1:36:08<7:09:42,  1.63s/it]
 17%|█▋        | 3288/19090 [1:36:10<8:19:49,  1.90s/it]
 17%|█▋        | 3289/19090 [1:36:12<7:31:49,  1.72s/it]
 17%|█▋        | 3290/19090 [1:36:13<7:12:18,  1.64s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.3979403078556061, 'learning_rate': 4.9208291334580104e-05, 'recon_loss': 0.003517965180799365, 'recon_static': 0.002560409554280341, 'recon_motion': 0.0009575556541676633, 'recon_norm': 2.066231834888458, 'recon_norm_std': 0.08078101500868798, 'target_norm': 1.6204061567783357, 'recon_cos': 0.44664219245314596, 'gradnorm/compressor.query': 0.006133862689955316, 'gradnorm/compressor.layer0': 0.10453722776332815, 'gradnorm/compressor.layer1': 0.14275281472779808, 'gradnorm/compressor.layer2': 0.12109118889592177, 'gradnorm/compressor.layer3': 0.11461631659618321, 'gradnorm/compressor.layer4': 0.10480817362460966, 'gradnorm/compressor.layer5': 0.09969587403734084, 'gradnorm/compressor.layer6': 0.09600930962150749, 'gradnorm/compressor.layer7': 0.09228416724911316, 'gradnorm/decoder.layer0': 0.12236340241909498, 'gradnorm/decoder.layer1': 0.12127307334957782, 'gradnorm/decoder.layer2': 0.17571097134667885, 'gradnorm/max_param': 0.12754333019256592, 'epoch': 1.72}

 17%|█▋        | 3290/19090 [1:36:13<7:12:18,  1.64s/it]
 17%|█▋        | 3291/19090 [1:36:16<8:42:32,  1.98s/it]
 17%|█▋        | 3292/19090 [1:36:17<7:47:34,  1.78s/it]
 17%|█▋        | 3293/19090 [1:36:19<7:14:18,  1.65s/it]
 17%|█▋        | 3294/19090 [1:36:22<9:05:19,  2.07s/it]
 17%|█▋        | 3295/19090 [1:36:23<8:04:10,  1.84s/it]
 17%|█▋        | 3296/19090 [1:36:24<7:21:59,  1.68s/it]
 17%|█▋        | 3297/19090 [1:36:27<8:55:03,  2.03s/it]
 17%|█▋        | 3298/19090 [1:36:28<7:56:44,  1.81s/it]
 17%|█▋        | 3299/19090 [1:36:30<7:15:45,  1.66s/it]
 17%|█▋        | 3300/19090 [1:36:32<8:16:40,  1.89s/it]
                                                        
{'loss': 0.0026, 'grad_norm': 0.3115735650062561, 'learning_rate': 4.9196837779684766e-05, 'recon_loss': 0.0026023173530120404, 'recon_static': 0.0016357208573026583, 'recon_motion': 0.0009665964797022752, 'recon_norm': 1.7844970881938935, 'recon_norm_std': 0.09110833629965782, 'target_norm': 1.6229503571987152, 'recon_cos': 0.5111712940037251, 'gradnorm/compressor.query': 0.004899534819368726, 'gradnorm/compressor.layer0': 0.0814624548785635, 'gradnorm/compressor.layer1': 0.11008017429098464, 'gradnorm/compressor.layer2': 0.0945932197007739, 'gradnorm/compressor.layer3': 0.09075008261101486, 'gradnorm/compressor.layer4': 0.08308588731887084, 'gradnorm/compressor.layer5': 0.07792318921733546, 'gradnorm/compressor.layer6': 0.07436751213768297, 'gradnorm/compressor.layer7': 0.07293697021993949, 'gradnorm/decoder.layer0': 0.09606793025732152, 'gradnorm/decoder.layer1': 0.09441343149456483, 'gradnorm/decoder.layer2': 0.1383773604409638, 'gradnorm/max_param': 0.09943867474794388, 'epoch': 1.73}

 17%|█▋        | 3300/19090 [1:36:32<8:16:40,  1.89s/it]
 17%|█▋        | 3301/19090 [1:36:33<7:29:35,  1.71s/it]
 17%|█▋        | 3302/19090 [1:36:35<6:56:41,  1.58s/it]
 17%|█▋        | 3303/19090 [1:36:37<8:16:30,  1.89s/it]
 17%|█▋        | 3304/19090 [1:36:39<7:29:21,  1.71s/it]
 17%|█▋        | 3305/19090 [1:36:40<6:56:11,  1.58s/it]
 17%|█▋        | 3306/19090 [1:36:42<8:20:39,  1.90s/it]
 17%|█▋        | 3307/19090 [1:36:44<7:32:14,  1.72s/it]
 17%|█▋        | 3308/19090 [1:36:45<7:08:24,  1.63s/it]
 17%|█▋        | 3309/19090 [1:36:48<8:10:50,  1.87s/it]
 17%|█▋        | 3310/19090 [1:36:49<7:25:31,  1.69s/it]
                                                        
{'loss': 0.0029, 'grad_norm': 0.42187485098838806, 'learning_rate': 4.918530332239757e-05, 'recon_loss': 0.002933613758068532, 'recon_static': 0.0019781872164458037, 'recon_motion': 0.0009554265954648145, 'recon_norm': 1.8948225498199462, 'recon_norm_std': 0.08672352321445942, 'target_norm': 1.6196963280439376, 'recon_cos': 0.4855568893253803, 'gradnorm/compressor.query': 0.00661929287954143, 'gradnorm/compressor.layer0': 0.11078571056376457, 'gradnorm/compressor.layer1': 0.15107943964137754, 'gradnorm/compressor.layer2': 0.12888990645405457, 'gradnorm/compressor.layer3': 0.12279877931205783, 'gradnorm/compressor.layer4': 0.11138932319524669, 'gradnorm/compressor.layer5': 0.10649005579471246, 'gradnorm/compressor.layer6': 0.10274405198156403, 'gradnorm/compressor.layer7': 0.10116345156948321, 'gradnorm/decoder.layer0': 0.13042983685462353, 'gradnorm/decoder.layer1': 0.12693385931698345, 'gradnorm/decoder.layer2': 0.1829800697721183, 'gradnorm/max_param': 0.13256028294563293, 'epoch': 1.73}

 17%|█▋        | 3310/19090 [1:36:49<7:25:31,  1.69s/it]
 17%|█▋        | 3311/19090 [1:36:51<7:23:44,  1.69s/it]
 17%|█▋        | 3312/19090 [1:36:53<8:22:09,  1.91s/it]
 17%|█▋        | 3313/19090 [1:36:54<7:33:20,  1.72s/it]
 17%|█▋        | 3314/19090 [1:36:56<7:24:22,  1.69s/it]
 17%|█▋        | 3315/19090 [1:36:58<8:12:50,  1.87s/it]
 17%|█▋        | 3316/19090 [1:36:59<7:26:45,  1.70s/it]
 17%|█▋        | 3317/19090 [1:37:01<7:42:59,  1.76s/it]
 17%|█▋        | 3318/19090 [1:37:04<8:32:30,  1.95s/it]
 17%|█▋        | 3319/19090 [1:37:05<7:40:30,  1.75s/it]
 17%|█▋        | 3320/19090 [1:37:06<7:08:20,  1.63s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.6643076539039612, 'learning_rate': 4.91736880012841e-05, 'recon_loss': 0.004381686460692436, 'recon_static': 0.003419359540566802, 'recon_motion': 0.0009623269288567826, 'recon_norm': 2.282413351535797, 'recon_norm_std': 0.07611105255782605, 'target_norm': 1.6238618403673173, 'recon_cos': 0.4057874567806721, 'gradnorm/compressor.query': 0.010058290485097291, 'gradnorm/compressor.layer0': 0.17556896636892252, 'gradnorm/compressor.layer1': 0.23701600777007106, 'gradnorm/compressor.layer2': 0.20496820501286178, 'gradnorm/compressor.layer3': 0.19146090330578971, 'gradnorm/compressor.layer4': 0.17286394365871774, 'gradnorm/compressor.layer5': 0.16411384278685923, 'gradnorm/compressor.layer6': 0.15850427598217026, 'gradnorm/compressor.layer7': 0.15471498360823646, 'gradnorm/decoder.layer0': 0.2005402268696146, 'gradnorm/decoder.layer1': 0.2023343599563516, 'gradnorm/decoder.layer2': 0.2974928399835216, 'gradnorm/max_param': 0.21152547001838684, 'epoch': 1.74}

 17%|█▋        | 3320/19090 [1:37:06<7:08:20,  1.63s/it]
 17%|█▋        | 3321/19090 [1:37:09<8:20:33,  1.90s/it]
 17%|█▋        | 3322/19090 [1:37:10<7:32:22,  1.72s/it]
 17%|█▋        | 3323/19090 [1:37:12<7:13:57,  1.65s/it]
 17%|█▋        | 3324/19090 [1:37:15<8:47:40,  2.01s/it]
 17%|█▋        | 3325/19090 [1:37:16<7:51:11,  1.79s/it]
 17%|█▋        | 3326/19090 [1:37:17<7:11:12,  1.64s/it]
 17%|█▋        | 3327/19090 [1:37:20<8:51:49,  2.02s/it]
 17%|█▋        | 3328/19090 [1:37:21<7:54:11,  1.81s/it]
 17%|█▋        | 3329/19090 [1:37:23<7:13:29,  1.65s/it]
 17%|█▋        | 3330/19090 [1:37:25<8:36:48,  1.97s/it]
                                                        
{'loss': 0.0068, 'grad_norm': 0.6680634021759033, 'learning_rate': 4.916199185518031e-05, 'recon_loss': 0.0068038783734664324, 'recon_static': 0.005832680535968393, 'recon_motion': 0.0009711978666018695, 'recon_norm': 2.835976642370224, 'recon_norm_std': 0.06123366924002767, 'target_norm': 1.6215037971735, 'recon_cos': 0.3236768193542957, 'gradnorm/compressor.query': 0.009562294110435849, 'gradnorm/compressor.layer0': 0.1748973824432622, 'gradnorm/compressor.layer1': 0.23599692853044213, 'gradnorm/compressor.layer2': 0.2016587892352085, 'gradnorm/compressor.layer3': 0.19456530697994587, 'gradnorm/compressor.layer4': 0.17856211276107528, 'gradnorm/compressor.layer5': 0.1661942642850236, 'gradnorm/compressor.layer6': 0.16110336728253066, 'gradnorm/compressor.layer7': 0.15816996662541583, 'gradnorm/decoder.layer0': 0.20447043345984275, 'gradnorm/decoder.layer1': 0.20449126606818338, 'gradnorm/decoder.layer2': 0.29548236815540785, 'gradnorm/max_param': 0.2137274593114853, 'epoch': 1.74}

 17%|█▋        | 3330/19090 [1:37:25<8:36:48,  1.97s/it]
 17%|█▋        | 3331/19090 [1:37:27<7:43:40,  1.77s/it]
 17%|█▋        | 3332/19090 [1:37:28<7:06:17,  1.62s/it]
 17%|█▋        | 3333/19090 [1:37:31<8:33:43,  1.96s/it]
 17%|█▋        | 3334/19090 [1:37:32<7:41:41,  1.76s/it]
 17%|█▋        | 3335/19090 [1:37:33<7:04:53,  1.62s/it]
 17%|█▋        | 3336/19090 [1:37:36<8:56:12,  2.04s/it]
 17%|█▋        | 3337/19090 [1:37:38<7:57:16,  1.82s/it]
 17%|█▋        | 3338/19090 [1:37:39<7:16:50,  1.66s/it]
 17%|█▋        | 3339/19090 [1:37:41<8:25:25,  1.93s/it]
 17%|█▋        | 3340/19090 [1:37:43<7:35:56,  1.74s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 0.3739345073699951, 'learning_rate': 4.9150214923192384e-05, 'recon_loss': 0.004133883508620784, 'recon_static': 0.0031802723242435604, 'recon_motion': 0.0009536111596389673, 'recon_norm': 2.2206039994955065, 'recon_norm_std': 0.07835333915427327, 'target_norm': 1.6193681180477142, 'recon_cos': 0.4192477449774742, 'gradnorm/compressor.query': 0.005148444869421654, 'gradnorm/compressor.layer0': 0.0963218356567541, 'gradnorm/compressor.layer1': 0.13314011781217833, 'gradnorm/compressor.layer2': 0.11304332147587594, 'gradnorm/compressor.layer3': 0.10937508940693062, 'gradnorm/compressor.layer4': 0.09819988094394726, 'gradnorm/compressor.layer5': 0.09380670163948594, 'gradnorm/compressor.layer6': 0.09142151648034351, 'gradnorm/compressor.layer7': 0.08583948115770229, 'gradnorm/decoder.layer0': 0.11310691408351413, 'gradnorm/decoder.layer1': 0.11468415719287606, 'gradnorm/decoder.layer2': 0.16715192126561504, 'gradnorm/max_param': 0.12183961272239685, 'epoch': 1.75}

 17%|█▋        | 3340/19090 [1:37:43<7:35:56,  1.74s/it]
 18%|█▊        | 3341/19090 [1:37:44<7:01:07,  1.60s/it]
 18%|█▊        | 3342/19090 [1:37:47<8:41:15,  1.99s/it]
 18%|█▊        | 3343/19090 [1:37:48<7:46:24,  1.78s/it]
 18%|█▊        | 3344/19090 [1:37:49<7:07:22,  1.63s/it]
 18%|█▊        | 3345/19090 [1:37:52<8:16:11,  1.89s/it]
 18%|█▊        | 3346/19090 [1:37:53<7:28:51,  1.71s/it]
 18%|█▊        | 3347/19090 [1:37:55<6:54:56,  1.58s/it]
 18%|█▊        | 3348/19090 [1:37:58<8:55:23,  2.04s/it]
 18%|█▊        | 3349/19090 [1:37:59<7:55:53,  1.81s/it]
 18%|█▊        | 3350/19090 [1:38:00<7:14:38,  1.66s/it]
                                                        
{'loss': 0.0024, 'grad_norm': 0.2455815076828003, 'learning_rate': 4.913835724469663e-05, 'recon_loss': 0.002398241322953254, 'recon_static': 0.0014427714253542946, 'recon_motion': 0.0009554699427098967, 'recon_norm': 1.7254314512014388, 'recon_norm_std': 0.09085036311298608, 'target_norm': 1.6200920045375824, 'recon_cos': 0.5341327011585235, 'gradnorm/compressor.query': 0.003747205201146417, 'gradnorm/compressor.layer0': 0.06304147972600072, 'gradnorm/compressor.layer1': 0.08563579353814456, 'gradnorm/compressor.layer2': 0.07404569477445562, 'gradnorm/compressor.layer3': 0.07081175094336258, 'gradnorm/compressor.layer4': 0.06493539710673996, 'gradnorm/compressor.layer5': 0.06131788383915531, 'gradnorm/compressor.layer6': 0.0599762111184949, 'gradnorm/compressor.layer7': 0.05870169590646459, 'gradnorm/decoder.layer0': 0.07577880285185551, 'gradnorm/decoder.layer1': 0.07534916422059408, 'gradnorm/decoder.layer2': 0.10978160789305447, 'gradnorm/max_param': 0.0799940973520279, 'epoch': 1.75}

 18%|█▊        | 3350/19090 [1:38:00<7:14:38,  1.66s/it]
 18%|█▊        | 3351/19090 [1:38:03<8:22:07,  1.91s/it]
 18%|█▊        | 3352/19090 [1:38:04<7:33:03,  1.73s/it]
 18%|█▊        | 3353/19090 [1:38:05<6:57:48,  1.59s/it]
 18%|█▊        | 3354/19090 [1:38:08<8:27:17,  1.93s/it]
 18%|█▊        | 3355/19090 [1:38:09<7:36:00,  1.74s/it]
 18%|█▊        | 3356/19090 [1:38:11<7:00:25,  1.60s/it]
 18%|█▊        | 3357/19090 [1:38:14<8:41:56,  1.99s/it]
 18%|█▊        | 3358/19090 [1:38:15<7:47:53,  1.78s/it]
 18%|█▊        | 3359/19090 [1:38:16<7:11:30,  1.65s/it]
 18%|█▊        | 3360/19090 [1:38:19<8:09:45,  1.87s/it]
                                                        
{'loss': 0.0023, 'grad_norm': 0.34549039602279663, 'learning_rate': 4.912641885933933e-05, 'recon_loss': 0.002300668158568442, 'recon_static': 0.0013400208845268935, 'recon_motion': 0.0009606472071027383, 'recon_norm': 1.690351766347885, 'recon_norm_std': 0.09606188889592886, 'target_norm': 1.6231804817914963, 'recon_cos': 0.5441028237342834, 'gradnorm/compressor.query': 0.0055698858384193185, 'gradnorm/compressor.layer0': 0.0917728270827588, 'gradnorm/compressor.layer1': 0.12291554981535183, 'gradnorm/compressor.layer2': 0.10516470777329076, 'gradnorm/compressor.layer3': 0.09969748545654286, 'gradnorm/compressor.layer4': 0.09073970150015688, 'gradnorm/compressor.layer5': 0.08535568567787619, 'gradnorm/compressor.layer6': 0.08299425238757387, 'gradnorm/compressor.layer7': 0.08086136375506255, 'gradnorm/decoder.layer0': 0.10545849750278184, 'gradnorm/decoder.layer1': 0.10426279093267647, 'gradnorm/decoder.layer2': 0.1544823437242065, 'gradnorm/max_param': 0.10983258485794067, 'epoch': 1.76}

 18%|█▊        | 3360/19090 [1:38:19<8:09:45,  1.87s/it]
 18%|█▊        | 3361/19090 [1:38:20<7:25:28,  1.70s/it]
 18%|█▊        | 3362/19090 [1:38:21<6:53:49,  1.58s/it]
 18%|█▊        | 3363/19090 [1:38:24<8:33:51,  1.96s/it]
 18%|█▊        | 3364/19090 [1:38:25<7:42:03,  1.76s/it]
 18%|█▊        | 3365/19090 [1:38:27<7:06:36,  1.63s/it]
 18%|█▊        | 3366/19090 [1:38:29<7:58:54,  1.83s/it]
 18%|█▊        | 3367/19090 [1:38:30<7:17:36,  1.67s/it]
 18%|█▊        | 3368/19090 [1:38:31<6:48:50,  1.56s/it]
 18%|█▊        | 3369/19090 [1:38:34<8:28:19,  1.94s/it]
 18%|█▊        | 3370/19090 [1:38:36<7:37:21,  1.75s/it]
                                                        
{'loss': 0.0025, 'grad_norm': 0.2976360321044922, 'learning_rate': 4.9114399807036596e-05, 'recon_loss': 0.0025138911849353464, 'recon_static': 0.001564935623900965, 'recon_motion': 0.0009489556032349356, 'recon_norm': 1.7655299067497254, 'recon_norm_std': 0.09203125350177288, 'target_norm': 1.61918064057827, 'recon_cos': 0.522116431593895, 'gradnorm/compressor.query': 0.0045502864844073065, 'gradnorm/compressor.layer0': 0.07894855583206663, 'gradnorm/compressor.layer1': 0.10713253936588106, 'gradnorm/compressor.layer2': 0.09217711588194066, 'gradnorm/compressor.layer3': 0.08726797484066122, 'gradnorm/compressor.layer4': 0.07853471787165986, 'gradnorm/compressor.layer5': 0.0737592091659956, 'gradnorm/compressor.layer6': 0.07208784892821463, 'gradnorm/compressor.layer7': 0.0692773423626418, 'gradnorm/decoder.layer0': 0.09031284231563226, 'gradnorm/decoder.layer1': 0.09047736770881615, 'gradnorm/decoder.layer2': 0.1296041338471947, 'gradnorm/max_param': 0.09386727958917618, 'epoch': 1.77}

 18%|█▊        | 3370/19090 [1:38:36<7:37:21,  1.75s/it]
 18%|█▊        | 3371/19090 [1:38:37<7:01:06,  1.61s/it]
 18%|█▊        | 3372/19090 [1:38:40<8:26:25,  1.93s/it]
 18%|█▊        | 3373/19090 [1:38:41<7:35:36,  1.74s/it]
 18%|█▊        | 3374/19090 [1:38:42<6:59:34,  1.60s/it]
 18%|█▊        | 3375/19090 [1:38:45<8:22:19,  1.92s/it]
 18%|█▊        | 3376/19090 [1:38:46<7:32:41,  1.73s/it]
 18%|█▊        | 3377/19090 [1:38:47<6:57:58,  1.60s/it]
 18%|█▊        | 3378/19090 [1:38:50<8:10:14,  1.87s/it]
 18%|█▊        | 3379/19090 [1:38:51<7:24:24,  1.70s/it]
 18%|█▊        | 3380/19090 [1:38:52<6:53:13,  1.58s/it]
                                                        
{'loss': 0.0025, 'grad_norm': 0.33411651849746704, 'learning_rate': 4.910230012797426e-05, 'recon_loss': 0.0024799945182166993, 'recon_static': 0.0015295805875211953, 'recon_motion': 0.0009504138579359278, 'recon_norm': 1.7552626073360442, 'recon_norm_std': 0.09123571496456861, 'target_norm': 1.618657910823822, 'recon_cos': 0.5240300446748734, 'gradnorm/compressor.query': 0.0051173609312211805, 'gradnorm/compressor.layer0': 0.08741282022012643, 'gradnorm/compressor.layer1': 0.11757989989807215, 'gradnorm/compressor.layer2': 0.10146558245480841, 'gradnorm/compressor.layer3': 0.09904925087045405, 'gradnorm/compressor.layer4': 0.08983989893911035, 'gradnorm/compressor.layer5': 0.08488774161108519, 'gradnorm/compressor.layer6': 0.0804604179966704, 'gradnorm/compressor.layer7': 0.07839111067705287, 'gradnorm/decoder.layer0': 0.10202981194976786, 'gradnorm/decoder.layer1': 0.10256085391324746, 'gradnorm/decoder.layer2': 0.14563868225065696, 'gradnorm/max_param': 0.10478895157575607, 'epoch': 1.77}

 18%|█▊        | 3380/19090 [1:38:53<6:53:13,  1.58s/it]
 18%|█▊        | 3381/19090 [1:38:55<8:35:47,  1.97s/it]
 18%|█▊        | 3382/19090 [1:38:57<7:41:59,  1.76s/it]
 18%|█▊        | 3383/19090 [1:38:58<7:04:12,  1.62s/it]
 18%|█▊        | 3384/19090 [1:39:01<8:28:01,  1.94s/it]
 18%|█▊        | 3385/19090 [1:39:02<7:36:25,  1.74s/it]
 18%|█▊        | 3386/19090 [1:39:03<7:00:14,  1.61s/it]
 18%|█▊        | 3387/19090 [1:39:06<8:41:52,  1.99s/it]
 18%|█▊        | 3388/19090 [1:39:07<7:46:23,  1.78s/it]
 18%|█▊        | 3389/19090 [1:39:09<7:07:23,  1.63s/it]
 18%|█▊        | 3390/19090 [1:39:11<7:59:37,  1.83s/it]
                                                        
{'loss': 0.0028, 'grad_norm': 0.4373237192630768, 'learning_rate': 4.9090119862607753e-05, 'recon_loss': 0.0028212342818733304, 'recon_static': 0.0018690717144636438, 'recon_motion': 0.0009521625499473885, 'recon_norm': 1.8618086189031602, 'recon_norm_std': 0.08734458778053522, 'target_norm': 1.6198467910289764, 'recon_cos': 0.4948537811636925, 'gradnorm/compressor.query': 0.006790697687359388, 'gradnorm/compressor.layer0': 0.1187568682955592, 'gradnorm/compressor.layer1': 0.15834204143703803, 'gradnorm/compressor.layer2': 0.13656688157217084, 'gradnorm/compressor.layer3': 0.12747444806055197, 'gradnorm/compressor.layer4': 0.11470095375598012, 'gradnorm/compressor.layer5': 0.10721232609835363, 'gradnorm/compressor.layer6': 0.10396833662249588, 'gradnorm/compressor.layer7': 0.1020645104988133, 'gradnorm/decoder.layer0': 0.13312748414170988, 'gradnorm/decoder.layer1': 0.13436309157323004, 'gradnorm/decoder.layer2': 0.1883522078053561, 'gradnorm/max_param': 0.13614463806152344, 'epoch': 1.78}

 18%|█▊        | 3390/19090 [1:39:11<7:59:37,  1.83s/it]
 18%|█▊        | 3391/19090 [1:39:12<7:17:54,  1.67s/it]
 18%|█▊        | 3392/19090 [1:39:14<6:47:05,  1.56s/it]
 18%|█▊        | 3393/19090 [1:39:17<8:36:23,  1.97s/it]
 18%|█▊        | 3394/19090 [1:39:18<7:42:05,  1.77s/it]
 18%|█▊        | 3395/19090 [1:39:19<7:04:03,  1.62s/it]
 18%|█▊        | 3396/19090 [1:39:22<8:31:30,  1.96s/it]
 18%|█▊        | 3397/19090 [1:39:23<7:38:58,  1.75s/it]
 18%|█▊        | 3398/19090 [1:39:24<7:01:48,  1.61s/it]
 18%|█▊        | 3399/19090 [1:39:27<8:37:53,  1.98s/it]
 18%|█▊        | 3400/19090 [1:39:28<7:43:17,  1.77s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 0.5274050235748291, 'learning_rate': 4.9077859051661896e-05, 'recon_loss': 0.004066523863002658, 'recon_static': 0.0031181600934360176, 'recon_motion': 0.0009483637666562572, 'recon_norm': 2.216858446598053, 'recon_norm_std': 0.07794585954397917, 'target_norm': 1.6188317775726317, 'recon_cos': 0.4177161201834679, 'gradnorm/compressor.query': 0.008073922338273549, 'gradnorm/compressor.layer0': 0.13950078667996543, 'gradnorm/compressor.layer1': 0.18658026982828396, 'gradnorm/compressor.layer2': 0.1619913234370294, 'gradnorm/compressor.layer3': 0.15488627888094472, 'gradnorm/compressor.layer4': 0.14008929284166166, 'gradnorm/compressor.layer5': 0.13273494223012625, 'gradnorm/compressor.layer6': 0.12878473652362216, 'gradnorm/compressor.layer7': 0.12233683672552009, 'gradnorm/decoder.layer0': 0.16166240475344376, 'gradnorm/decoder.layer1': 0.16191132063436506, 'gradnorm/decoder.layer2': 0.2290018972215188, 'gradnorm/max_param': 0.16551320254802704, 'epoch': 1.78}

 18%|█▊        | 3400/19090 [1:39:29<7:43:17,  1.77s/it]
 18%|█▊        | 3401/19090 [1:39:30<7:04:53,  1.62s/it]
 18%|█▊        | 3402/19090 [1:39:33<8:51:51,  2.03s/it]
 18%|█▊        | 3403/19090 [1:39:34<7:53:11,  1.81s/it]
 18%|█▊        | 3404/19090 [1:39:35<7:11:51,  1.65s/it]
 18%|█▊        | 3405/19090 [1:39:38<8:57:39,  2.06s/it]
 18%|█▊        | 3406/19090 [1:39:40<7:57:26,  1.83s/it]
 18%|█▊        | 3407/19090 [1:39:41<7:15:17,  1.67s/it]
 18%|█▊        | 3408/19090 [1:39:43<8:11:12,  1.88s/it]
 18%|█▊        | 3409/19090 [1:39:45<7:24:50,  1.70s/it]
 18%|█▊        | 3410/19090 [1:39:46<6:52:26,  1.58s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.510272204875946, 'learning_rate': 4.906551773613086e-05, 'recon_loss': 0.004232851217966527, 'recon_static': 0.003279074444435537, 'recon_motion': 0.0009537767051369883, 'recon_norm': 2.260012573003769, 'recon_norm_std': 0.074879683367908, 'target_norm': 1.618053635954857, 'recon_cos': 0.40810818672180177, 'gradnorm/compressor.query': 0.007753623216983054, 'gradnorm/compressor.layer0': 0.13619920925696463, 'gradnorm/compressor.layer1': 0.18245206088943047, 'gradnorm/compressor.layer2': 0.15869875964220553, 'gradnorm/compressor.layer3': 0.1506875994997238, 'gradnorm/compressor.layer4': 0.1339945711421285, 'gradnorm/compressor.layer5': 0.12644388327086922, 'gradnorm/compressor.layer6': 0.12129479301048615, 'gradnorm/compressor.layer7': 0.11639630736062229, 'gradnorm/decoder.layer0': 0.15572658130536746, 'gradnorm/decoder.layer1': 0.15528490516013377, 'gradnorm/decoder.layer2': 0.2236499446653276, 'gradnorm/max_param': 0.16102619469165802, 'epoch': 1.79}

 18%|█▊        | 3410/19090 [1:39:46<6:52:26,  1.58s/it]
 18%|█▊        | 3411/19090 [1:39:48<8:13:01,  1.89s/it]
 18%|█▊        | 3412/19090 [1:39:50<7:25:46,  1.71s/it]
 18%|█▊        | 3413/19090 [1:39:51<6:52:36,  1.58s/it]
 18%|█▊        | 3414/19090 [1:39:54<8:23:06,  1.93s/it]
 18%|█▊        | 3415/19090 [1:39:55<7:32:43,  1.73s/it]
 18%|█▊        | 3416/19090 [1:39:56<6:57:44,  1.60s/it]
 18%|█▊        | 3417/19090 [1:39:59<8:04:06,  1.85s/it]
 18%|█▊        | 3418/19090 [1:40:00<7:19:32,  1.68s/it]
 18%|█▊        | 3419/19090 [1:40:02<7:01:40,  1.61s/it]
 18%|█▊        | 3420/19090 [1:40:04<8:10:17,  1.88s/it]
                                                        
{'loss': 0.004, 'grad_norm': 0.531967282295227, 'learning_rate': 4.9053095957277964e-05, 'recon_loss': 0.004021567170275376, 'recon_static': 0.003069950669305399, 'recon_motion': 0.0009516164791421033, 'recon_norm': 2.207022178173065, 'recon_norm_std': 0.07723310180008411, 'target_norm': 1.6208915084600448, 'recon_cos': 0.4195622742176056, 'gradnorm/compressor.query': 0.008511408155635026, 'gradnorm/compressor.layer0': 0.146383505105848, 'gradnorm/compressor.layer1': 0.19371150926527508, 'gradnorm/compressor.layer2': 0.1651223271718165, 'gradnorm/compressor.layer3': 0.15515483669240449, 'gradnorm/compressor.layer4': 0.13895261580711624, 'gradnorm/compressor.layer5': 0.13077797421872814, 'gradnorm/compressor.layer6': 0.12522499875137014, 'gradnorm/compressor.layer7': 0.12145386493663092, 'gradnorm/decoder.layer0': 0.1610923600715001, 'gradnorm/decoder.layer1': 0.1614040519365274, 'gradnorm/decoder.layer2': 0.231907315020319, 'gradnorm/max_param': 0.1677820235490799, 'epoch': 1.79}

 18%|█▊        | 3420/19090 [1:40:04<8:10:17,  1.88s/it]
 18%|█▊        | 3421/19090 [1:40:05<7:23:55,  1.70s/it]
 18%|█▊        | 3422/19090 [1:40:07<7:09:32,  1.64s/it]
 18%|█▊        | 3423/19090 [1:40:10<8:55:59,  2.05s/it]
 18%|█▊        | 3424/19090 [1:40:11<7:56:10,  1.82s/it]
 18%|█▊        | 3425/19090 [1:40:12<7:14:49,  1.67s/it]
 18%|█▊        | 3426/19090 [1:40:15<8:34:29,  1.97s/it]
 18%|█▊        | 3427/19090 [1:40:16<7:40:52,  1.77s/it]
 18%|█▊        | 3428/19090 [1:40:18<7:03:09,  1.62s/it]
 18%|█▊        | 3429/19090 [1:40:20<8:11:03,  1.88s/it]
 18%|█▊        | 3430/19090 [1:40:21<7:24:51,  1.70s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.5924824476242065, 'learning_rate': 4.9040593756635565e-05, 'recon_loss': 0.004317710490431637, 'recon_static': 0.00336515624076128, 'recon_motion': 0.0009525541900075041, 'recon_norm': 2.275763577222824, 'recon_norm_std': 0.07469652947038412, 'target_norm': 1.619761660695076, 'recon_cos': 0.40292259529232977, 'gradnorm/compressor.query': 0.008669670624636194, 'gradnorm/compressor.layer0': 0.15640061986661963, 'gradnorm/compressor.layer1': 0.20816219671883926, 'gradnorm/compressor.layer2': 0.18239931379563518, 'gradnorm/compressor.layer3': 0.1691242053671592, 'gradnorm/compressor.layer4': 0.15387827793312914, 'gradnorm/compressor.layer5': 0.1461005395443769, 'gradnorm/compressor.layer6': 0.14140211067674738, 'gradnorm/compressor.layer7': 0.14186197867064515, 'gradnorm/decoder.layer0': 0.1807225931568009, 'gradnorm/decoder.layer1': 0.180130826309186, 'gradnorm/decoder.layer2': 0.26654099555440686, 'gradnorm/max_param': 0.18852855265140533, 'epoch': 1.8}

 18%|█▊        | 3430/19090 [1:40:21<7:24:51,  1.70s/it]
 18%|█▊        | 3431/19090 [1:40:23<7:12:59,  1.66s/it]
 18%|█▊        | 3432/19090 [1:40:26<8:20:34,  1.92s/it]
 18%|█▊        | 3433/19090 [1:40:27<7:31:41,  1.73s/it]
 18%|█▊        | 3434/19090 [1:40:29<7:28:00,  1.72s/it]
 18%|█▊        | 3435/19090 [1:40:30<7:48:39,  1.80s/it]
 18%|█▊        | 3436/19090 [1:40:32<7:09:04,  1.64s/it]
 18%|█▊        | 3437/19090 [1:40:34<7:35:52,  1.75s/it]
 18%|█▊        | 3438/19090 [1:40:36<8:06:10,  1.86s/it]
 18%|█▊        | 3439/19090 [1:40:37<7:20:42,  1.69s/it]
 18%|█▊        | 3440/19090 [1:40:39<7:39:12,  1.76s/it]
                                                        
{'loss': 0.0048, 'grad_norm': 0.4728197455406189, 'learning_rate': 4.902801117600492e-05, 'recon_loss': 0.0048497460025828335, 'recon_static': 0.003891832270892337, 'recon_motion': 0.0009579136458341964, 'recon_norm': 2.4064460396766663, 'recon_norm_std': 0.07206274829804897, 'target_norm': 1.6211031317710876, 'recon_cos': 0.3840305991470814, 'gradnorm/compressor.query': 0.0070366346148382905, 'gradnorm/compressor.layer0': 0.12382415470292639, 'gradnorm/compressor.layer1': 0.16894300823073524, 'gradnorm/compressor.layer2': 0.14694667979479484, 'gradnorm/compressor.layer3': 0.1408782837014542, 'gradnorm/compressor.layer4': 0.12436170329610284, 'gradnorm/compressor.layer5': 0.11683363336059427, 'gradnorm/compressor.layer6': 0.11362815635612317, 'gradnorm/compressor.layer7': 0.1095533917191362, 'gradnorm/decoder.layer0': 0.1437909005899213, 'gradnorm/decoder.layer1': 0.14413301918204782, 'gradnorm/decoder.layer2': 0.20667713398023785, 'gradnorm/max_param': 0.149999737739563, 'epoch': 1.8}

 18%|█▊        | 3440/19090 [1:40:39<7:39:12,  1.76s/it]
 18%|█▊        | 3441/19090 [1:40:41<8:08:26,  1.87s/it]
 18%|█▊        | 3442/19090 [1:40:43<7:22:39,  1.70s/it]
 18%|█▊        | 3443/19090 [1:40:45<7:49:33,  1.80s/it]
 18%|█▊        | 3444/19090 [1:40:46<7:43:45,  1.78s/it]
 18%|█▊        | 3445/19090 [1:40:48<7:05:29,  1.63s/it]
 18%|█▊        | 3446/19090 [1:40:50<8:18:04,  1.91s/it]
 18%|█▊        | 3447/19090 [1:40:51<7:29:28,  1.72s/it]
 18%|█▊        | 3448/19090 [1:40:53<7:05:20,  1.63s/it]
 18%|█▊        | 3449/19090 [1:40:56<8:35:04,  1.98s/it]
 18%|█▊        | 3450/19090 [1:40:57<7:41:30,  1.77s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.3484589755535126, 'learning_rate': 4.901534825745599e-05, 'recon_loss': 0.0032705721620004623, 'recon_static': 0.0023005143739283085, 'recon_motion': 0.0009700578812044114, 'recon_norm': 1.9901228696107864, 'recon_norm_std': 0.08231938239187002, 'target_norm': 1.6225941598415374, 'recon_cos': 0.4615844801068306, 'gradnorm/compressor.query': 0.005151233131802262, 'gradnorm/compressor.layer0': 0.09163600440868716, 'gradnorm/compressor.layer1': 0.12288371121843505, 'gradnorm/compressor.layer2': 0.10645560556044349, 'gradnorm/compressor.layer3': 0.10142367301467288, 'gradnorm/compressor.layer4': 0.09100174982293299, 'gradnorm/compressor.layer5': 0.0858335001167086, 'gradnorm/compressor.layer6': 0.08273368055738396, 'gradnorm/compressor.layer7': 0.08122978486243054, 'gradnorm/decoder.layer0': 0.1070788150752871, 'gradnorm/decoder.layer1': 0.1073148063787849, 'gradnorm/decoder.layer2': 0.15558217502137103, 'gradnorm/max_param': 0.11263679713010788, 'epoch': 1.81}

 18%|█▊        | 3450/19090 [1:40:57<7:41:30,  1.77s/it]
 18%|█▊        | 3451/19090 [1:40:58<7:17:47,  1.68s/it]
 18%|█▊        | 3452/19090 [1:41:01<8:43:17,  2.01s/it]
 18%|█▊        | 3453/19090 [1:41:02<7:47:14,  1.79s/it]
 18%|█▊        | 3454/19090 [1:41:04<7:14:27,  1.67s/it]
 18%|█▊        | 3455/19090 [1:41:07<8:39:30,  1.99s/it]
 18%|█▊        | 3456/19090 [1:41:08<7:44:39,  1.78s/it]
 18%|█▊        | 3457/19090 [1:41:09<7:05:57,  1.63s/it]
 18%|█▊        | 3458/19090 [1:41:12<8:26:39,  1.94s/it]
 18%|█▊        | 3459/19090 [1:41:13<7:36:26,  1.75s/it]
 18%|█▊        | 3460/19090 [1:41:15<7:18:01,  1.68s/it]
                                                        
{'loss': 0.0023, 'grad_norm': 0.3080367147922516, 'learning_rate': 4.900260504332742e-05, 'recon_loss': 0.002341284794965759, 'recon_static': 0.0013971472886623814, 'recon_motion': 0.000944137513579335, 'recon_norm': 1.7142094910144805, 'recon_norm_std': 0.09371698703616857, 'target_norm': 1.6179792910814286, 'recon_cos': 0.5395545780658721, 'gradnorm/compressor.query': 0.005060283453076824, 'gradnorm/compressor.layer0': 0.08369535795160885, 'gradnorm/compressor.layer1': 0.11078565592135449, 'gradnorm/compressor.layer2': 0.0936789839705982, 'gradnorm/compressor.layer3': 0.0884626792389915, 'gradnorm/compressor.layer4': 0.07959724169558274, 'gradnorm/compressor.layer5': 0.0759973115218061, 'gradnorm/compressor.layer6': 0.07327284806955311, 'gradnorm/compressor.layer7': 0.0716967422090897, 'gradnorm/decoder.layer0': 0.09539013889823109, 'gradnorm/decoder.layer1': 0.09432712379494185, 'gradnorm/decoder.layer2': 0.13551274812693417, 'gradnorm/max_param': 0.09735995531082153, 'epoch': 1.81}

 18%|█▊        | 3460/19090 [1:41:15<7:18:01,  1.68s/it]
 18%|█▊        | 3461/19090 [1:41:17<8:40:02,  2.00s/it]
 18%|█▊        | 3462/19090 [1:41:19<7:44:12,  1.78s/it]
 18%|█▊        | 3463/19090 [1:41:20<7:14:30,  1.67s/it]
 18%|█▊        | 3464/19090 [1:41:22<8:06:08,  1.87s/it]
 18%|█▊        | 3465/19090 [1:41:24<7:20:23,  1.69s/it]
 18%|█▊        | 3466/19090 [1:41:26<7:39:31,  1.76s/it]
 18%|█▊        | 3467/19090 [1:41:28<8:28:33,  1.95s/it]
 18%|█▊        | 3468/19090 [1:41:29<7:36:23,  1.75s/it]
 18%|█▊        | 3469/19090 [1:41:31<7:51:51,  1.81s/it]
 18%|█▊        | 3470/19090 [1:41:33<8:20:13,  1.92s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.5353917479515076, 'learning_rate': 4.898978157622626e-05, 'recon_loss': 0.0034595066623296588, 'recon_static': 0.0025054791185539216, 'recon_motion': 0.0009540275030303746, 'recon_norm': 2.044220983982086, 'recon_norm_std': 0.08132968172430992, 'target_norm': 1.6214809000492096, 'recon_cos': 0.4551773019134998, 'gradnorm/compressor.query': 0.007786985967276925, 'gradnorm/compressor.layer0': 0.13345501957799008, 'gradnorm/compressor.layer1': 0.18383092981055832, 'gradnorm/compressor.layer2': 0.1625681612681886, 'gradnorm/compressor.layer3': 0.15586748986403448, 'gradnorm/compressor.layer4': 0.14366495348099004, 'gradnorm/compressor.layer5': 0.1332067080824074, 'gradnorm/compressor.layer6': 0.12912430526854857, 'gradnorm/compressor.layer7': 0.12500223515419542, 'gradnorm/decoder.layer0': 0.16531625301737696, 'gradnorm/decoder.layer1': 0.16900963743899142, 'gradnorm/decoder.layer2': 0.24016779465435834, 'gradnorm/max_param': 0.16659080982208252, 'epoch': 1.82}

 18%|█▊        | 3470/19090 [1:41:33<8:20:13,  1.92s/it]
 18%|█▊        | 3471/19090 [1:41:35<7:30:34,  1.73s/it]
 18%|█▊        | 3472/19090 [1:41:37<7:40:01,  1.77s/it]
 18%|█▊        | 3473/19090 [1:41:39<7:53:42,  1.82s/it]
 18%|█▊        | 3474/19090 [1:41:40<7:12:02,  1.66s/it]
 18%|█▊        | 3475/19090 [1:41:42<7:42:30,  1.78s/it]
 18%|█▊        | 3476/19090 [1:41:44<7:53:10,  1.82s/it]
 18%|█▊        | 3477/19090 [1:41:45<7:12:30,  1.66s/it]
 18%|█▊        | 3478/19090 [1:41:47<7:42:28,  1.78s/it]
 18%|█▊        | 3479/19090 [1:41:49<7:51:14,  1.81s/it]
 18%|█▊        | 3480/19090 [1:41:50<7:10:19,  1.65s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.5263866782188416, 'learning_rate': 4.897687789902792e-05, 'recon_loss': 0.004405643464997411, 'recon_static': 0.0034515183826442806, 'recon_motion': 0.0009541251027258113, 'recon_norm': 2.3025413870811464, 'recon_norm_std': 0.07544842176139355, 'target_norm': 1.6203773111104964, 'recon_cos': 0.4005581721663475, 'gradnorm/compressor.query': 0.008850971129491491, 'gradnorm/compressor.layer0': 0.1408752030203608, 'gradnorm/compressor.layer1': 0.189116779133593, 'gradnorm/compressor.layer2': 0.15959339558532334, 'gradnorm/compressor.layer3': 0.15263826072924414, 'gradnorm/compressor.layer4': 0.1377165081403473, 'gradnorm/compressor.layer5': 0.12929531828297564, 'gradnorm/compressor.layer6': 0.12405210824415176, 'gradnorm/compressor.layer7': 0.123959107874072, 'gradnorm/decoder.layer0': 0.16038695443139153, 'gradnorm/decoder.layer1': 0.16282334001482157, 'gradnorm/decoder.layer2': 0.23227530754686096, 'gradnorm/max_param': 0.16462422907352448, 'epoch': 1.82}

 18%|█▊        | 3480/19090 [1:41:50<7:10:19,  1.65s/it]
 18%|█▊        | 3481/19090 [1:41:52<7:46:51,  1.79s/it]
 18%|█▊        | 3482/19090 [1:41:54<8:03:21,  1.86s/it]
 18%|█▊        | 3483/19090 [1:41:56<7:24:17,  1.71s/it]
 18%|█▊        | 3484/19090 [1:41:58<7:35:47,  1.75s/it]
 18%|█▊        | 3485/19090 [1:42:00<8:10:45,  1.89s/it]
 18%|█▊        | 3486/19090 [1:42:01<7:24:45,  1.71s/it]
 18%|█▊        | 3487/19090 [1:42:03<7:09:34,  1.65s/it]
 18%|█▊        | 3488/19090 [1:42:05<8:26:35,  1.95s/it]
 18%|█▊        | 3489/19090 [1:42:07<7:35:23,  1.75s/it]
 18%|█▊        | 3490/19090 [1:42:08<6:59:43,  1.61s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.41725972294807434, 'learning_rate': 4.896389405487599e-05, 'recon_loss': 0.0035157373466063293, 'recon_static': 0.0025697903824038804, 'recon_motion': 0.00094594685069751, 'recon_norm': 2.070247957110405, 'recon_norm_std': 0.08070173710584641, 'target_norm': 1.6179475486278534, 'recon_cos': 0.4469484336674213, 'gradnorm/compressor.query': 0.006505475770868621, 'gradnorm/compressor.layer0': 0.11002405482723508, 'gradnorm/compressor.layer1': 0.14724934789014502, 'gradnorm/compressor.layer2': 0.1274759603842554, 'gradnorm/compressor.layer3': 0.12199767049413157, 'gradnorm/compressor.layer4': 0.11139679346972228, 'gradnorm/compressor.layer5': 0.10484364486836878, 'gradnorm/compressor.layer6': 0.09848742581362399, 'gradnorm/compressor.layer7': 0.09884998283369198, 'gradnorm/decoder.layer0': 0.1275964356694977, 'gradnorm/decoder.layer1': 0.12844522716314505, 'gradnorm/decoder.layer2': 0.1830043360849158, 'gradnorm/max_param': 0.13161484897136688, 'epoch': 1.83}

 18%|█▊        | 3490/19090 [1:42:08<6:59:43,  1.61s/it]
 18%|█▊        | 3491/19090 [1:42:11<8:21:03,  1.93s/it]
 18%|█▊        | 3492/19090 [1:42:12<7:31:23,  1.74s/it]
 18%|█▊        | 3493/19090 [1:42:13<7:14:40,  1.67s/it]
 18%|█▊        | 3494/19090 [1:42:16<8:21:51,  1.93s/it]
 18%|█▊        | 3495/19090 [1:42:17<7:31:35,  1.74s/it]
 18%|█▊        | 3496/19090 [1:42:19<7:34:23,  1.75s/it]
 18%|█▊        | 3497/19090 [1:42:21<7:58:57,  1.84s/it]
 18%|█▊        | 3498/19090 [1:42:22<7:15:15,  1.67s/it]
 18%|█▊        | 3499/19090 [1:42:24<7:46:15,  1.79s/it]
 18%|█▊        | 3500/19090 [1:42:26<7:43:51,  1.79s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.41125214099884033, 'learning_rate': 4.89508300871821e-05, 'recon_loss': 0.0032537339895498006, 'recon_static': 0.002293868409469724, 'recon_motion': 0.0009598656004527583, 'recon_norm': 1.9941600531339645, 'recon_norm_std': 0.08276475332677365, 'target_norm': 1.6225965976715089, 'recon_cos': 0.46309767588973044, 'gradnorm/compressor.query': 0.006312173698257963, 'gradnorm/compressor.layer0': 0.11201408353801966, 'gradnorm/compressor.layer1': 0.148199549487574, 'gradnorm/compressor.layer2': 0.1283780605254064, 'gradnorm/compressor.layer3': 0.12200273550642714, 'gradnorm/compressor.layer4': 0.10716879683574913, 'gradnorm/compressor.layer5': 0.10217631982456572, 'gradnorm/compressor.layer6': 0.09784084516451334, 'gradnorm/compressor.layer7': 0.09617982107194567, 'gradnorm/decoder.layer0': 0.12500544625576954, 'gradnorm/decoder.layer1': 0.12522038760636436, 'gradnorm/decoder.layer2': 0.17651505951184282, 'gradnorm/max_param': 0.12767420709133148, 'epoch': 1.83}

 18%|█▊        | 3500/19090 [1:42:26<7:43:51,  1.79s/it]
 18%|█▊        | 3501/19090 [1:42:30<10:40:30,  2.47s/it]
 18%|█▊        | 3502/19090 [1:42:31<9:08:30,  2.11s/it] 
 18%|█▊        | 3503/19090 [1:42:33<8:03:58,  1.86s/it]
 18%|█▊        | 3504/19090 [1:42:34<7:18:56,  1.69s/it]
 18%|█▊        | 3505/19090 [1:42:35<6:47:33,  1.57s/it]
 18%|█▊        | 3506/19090 [1:42:37<6:25:36,  1.48s/it]
 18%|█▊        | 3507/19090 [1:42:38<6:09:52,  1.42s/it]
 18%|█▊        | 3508/19090 [1:42:39<5:59:03,  1.38s/it]
 18%|█▊        | 3509/19090 [1:42:42<7:21:21,  1.70s/it]
 18%|█▊        | 3510/19090 [1:42:43<6:49:22,  1.58s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.4516415596008301, 'learning_rate': 4.893768603962575e-05, 'recon_loss': 0.0032482598500791935, 'recon_static': 0.0022870441956911237, 'recon_motion': 0.0009612156834919005, 'recon_norm': 1.9904988020658494, 'recon_norm_std': 0.08235997091978789, 'target_norm': 1.6222969800233842, 'recon_cos': 0.462697796523571, 'gradnorm/compressor.query': 0.0071162037468384814, 'gradnorm/compressor.layer0': 0.12026145846288874, 'gradnorm/compressor.layer1': 0.16257532785512524, 'gradnorm/compressor.layer2': 0.14280112998617833, 'gradnorm/compressor.layer3': 0.1335244450744571, 'gradnorm/compressor.layer4': 0.1187853871811535, 'gradnorm/compressor.layer5': 0.11117301140135372, 'gradnorm/compressor.layer6': 0.10858651332618836, 'gradnorm/compressor.layer7': 0.1050442612688963, 'gradnorm/decoder.layer0': 0.1383236623367564, 'gradnorm/decoder.layer1': 0.13735366940616878, 'gradnorm/decoder.layer2': 0.193702720441655, 'gradnorm/max_param': 0.14083178341388702, 'epoch': 1.84}

 18%|█▊        | 3510/19090 [1:42:43<6:49:22,  1.58s/it]
 18%|█▊        | 3511/19090 [1:42:45<7:02:43,  1.63s/it]
 18%|█▊        | 3512/19090 [1:42:47<7:53:21,  1.82s/it]
 18%|█▊        | 3513/19090 [1:42:48<7:11:28,  1.66s/it]
 18%|█▊        | 3514/19090 [1:42:50<7:16:42,  1.68s/it]
 18%|█▊        | 3515/19090 [1:42:52<7:57:39,  1.84s/it]
 18%|█▊        | 3516/19090 [1:42:53<7:14:30,  1.67s/it]
 18%|█▊        | 3517/19090 [1:42:56<8:06:37,  1.87s/it]
 18%|█▊        | 3518/19090 [1:42:57<7:50:37,  1.81s/it]
 18%|█▊        | 3519/19090 [1:42:59<7:10:57,  1.66s/it]
 18%|█▊        | 3520/19090 [1:43:01<8:11:57,  1.90s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.5494844317436218, 'learning_rate': 4.892446195615423e-05, 'recon_loss': 0.003862047870643437, 'recon_static': 0.00290459722164087, 'recon_motion': 0.0009574506140779704, 'recon_norm': 2.158067524433136, 'recon_norm_std': 0.0768901189789176, 'target_norm': 1.6216546773910523, 'recon_cos': 0.4266915492713451, 'gradnorm/compressor.query': 0.008960319500174875, 'gradnorm/compressor.layer0': 0.14409578938965512, 'gradnorm/compressor.layer1': 0.19246113339896712, 'gradnorm/compressor.layer2': 0.16755741231466031, 'gradnorm/compressor.layer3': 0.16335788484401786, 'gradnorm/compressor.layer4': 0.1441675972449123, 'gradnorm/compressor.layer5': 0.134058708446301, 'gradnorm/compressor.layer6': 0.13160261771417936, 'gradnorm/compressor.layer7': 0.1276361358909462, 'gradnorm/decoder.layer0': 0.1715012029415877, 'gradnorm/decoder.layer1': 0.16866373842022625, 'gradnorm/decoder.layer2': 0.24302320728676774, 'gradnorm/max_param': 0.17342303693294525, 'epoch': 1.84}

 18%|█▊        | 3520/19090 [1:43:01<8:11:57,  1.90s/it]
 18%|█▊        | 3521/19090 [1:43:03<7:58:51,  1.85s/it]
 18%|█▊        | 3522/19090 [1:43:04<7:15:07,  1.68s/it]
 18%|█▊        | 3523/19090 [1:43:07<8:17:39,  1.92s/it]
 18%|█▊        | 3524/19090 [1:43:08<7:38:08,  1.77s/it]
 18%|█▊        | 3525/19090 [1:43:09<7:00:48,  1.62s/it]
 18%|█▊        | 3526/19090 [1:43:12<8:19:30,  1.93s/it]
 18%|█▊        | 3527/19090 [1:43:14<7:51:32,  1.82s/it]
 18%|█▊        | 3528/19090 [1:43:15<7:10:15,  1.66s/it]
 18%|█▊        | 3529/19090 [1:43:17<8:23:54,  1.94s/it]
 18%|█▊        | 3530/19090 [1:43:19<7:59:04,  1.85s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.5778355002403259, 'learning_rate': 4.891115788098239e-05, 'recon_loss': 0.004499591619241983, 'recon_static': 0.0035370254889130594, 'recon_motion': 0.0009625658596633003, 'recon_norm': 2.323363894224167, 'recon_norm_std': 0.07301925588399172, 'target_norm': 1.6213632762432098, 'recon_cos': 0.3957636259496212, 'gradnorm/compressor.query': 0.009705800604716205, 'gradnorm/compressor.layer0': 0.1543448289160392, 'gradnorm/compressor.layer1': 0.20622204135916333, 'gradnorm/compressor.layer2': 0.17850474151222906, 'gradnorm/compressor.layer3': 0.17084314814339968, 'gradnorm/compressor.layer4': 0.15235757765004687, 'gradnorm/compressor.layer5': 0.14448903730330215, 'gradnorm/compressor.layer6': 0.1364700507697537, 'gradnorm/compressor.layer7': 0.13455507859372157, 'gradnorm/decoder.layer0': 0.1795292447661257, 'gradnorm/decoder.layer1': 0.17696611765136974, 'gradnorm/decoder.layer2': 0.24906148432879918, 'gradnorm/max_param': 0.18032841384410858, 'epoch': 1.85}

 18%|█▊        | 3530/19090 [1:43:19<7:59:04,  1.85s/it]
 18%|█▊        | 3531/19090 [1:43:20<7:15:29,  1.68s/it]
 19%|█▊        | 3532/19090 [1:43:23<8:46:38,  2.03s/it]
 19%|█▊        | 3533/19090 [1:43:25<7:48:35,  1.81s/it]
 19%|█▊        | 3534/19090 [1:43:26<7:29:15,  1.73s/it]
 19%|█▊        | 3535/19090 [1:43:29<9:00:54,  2.09s/it]
 19%|█▊        | 3536/19090 [1:43:30<7:58:59,  1.85s/it]
 19%|█▊        | 3537/19090 [1:43:32<7:15:26,  1.68s/it]
 19%|█▊        | 3538/19090 [1:43:34<8:38:16,  2.00s/it]
 19%|█▊        | 3539/19090 [1:43:36<7:48:16,  1.81s/it]
 19%|█▊        | 3540/19090 [1:43:37<7:31:54,  1.74s/it]
                                                        
{'loss': 0.0048, 'grad_norm': 0.6082797646522522, 'learning_rate': 4.889777385859255e-05, 'recon_loss': 0.0048326797317713496, 'recon_static': 0.003875149803934619, 'recon_motion': 0.0009575298827257939, 'recon_norm': 2.407067412137985, 'recon_norm_std': 0.07217411231249571, 'target_norm': 1.6215989112854003, 'recon_cos': 0.38348616734147073, 'gradnorm/compressor.query': 0.009889253372895779, 'gradnorm/compressor.layer0': 0.16434431848538297, 'gradnorm/compressor.layer1': 0.21982956909013382, 'gradnorm/compressor.layer2': 0.1883654093715205, 'gradnorm/compressor.layer3': 0.18042272996588396, 'gradnorm/compressor.layer4': 0.15723486694771496, 'gradnorm/compressor.layer5': 0.1533045548037649, 'gradnorm/compressor.layer6': 0.14203747705252778, 'gradnorm/compressor.layer7': 0.14106189887882548, 'gradnorm/decoder.layer0': 0.18582650245491236, 'gradnorm/decoder.layer1': 0.18753186590891016, 'gradnorm/decoder.layer2': 0.26180260297109265, 'gradnorm/max_param': 0.19016927480697632, 'epoch': 1.85}

 19%|█▊        | 3540/19090 [1:43:37<7:31:54,  1.74s/it]
 19%|█▊        | 3541/19090 [1:43:40<8:28:30,  1.96s/it]
 19%|█▊        | 3542/19090 [1:43:41<7:35:21,  1.76s/it]
 19%|█▊        | 3543/19090 [1:43:42<7:08:07,  1.65s/it]
 19%|█▊        | 3544/19090 [1:43:45<8:33:05,  1.98s/it]
 19%|█▊        | 3545/19090 [1:43:46<7:40:52,  1.78s/it]
 19%|█▊        | 3546/19090 [1:43:48<7:08:24,  1.65s/it]
 19%|█▊        | 3547/19090 [1:43:50<8:21:33,  1.94s/it]
 19%|█▊        | 3548/19090 [1:43:52<7:33:36,  1.75s/it]
 19%|█▊        | 3549/19090 [1:43:53<6:59:27,  1.62s/it]
 19%|█▊        | 3550/19090 [1:43:56<8:09:19,  1.89s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.5060051679611206, 'learning_rate': 4.888430993373435e-05, 'recon_loss': 0.004451287491247058, 'recon_static': 0.003487443446647376, 'recon_motion': 0.0009638441493734718, 'recon_norm': 2.312523919343948, 'recon_norm_std': 0.07181105827912689, 'target_norm': 1.6222100019454957, 'recon_cos': 0.3986701540648937, 'gradnorm/compressor.query': 0.007871180651761237, 'gradnorm/compressor.layer0': 0.13094386177495201, 'gradnorm/compressor.layer1': 0.1756563537667251, 'gradnorm/compressor.layer2': 0.15563443642659427, 'gradnorm/compressor.layer3': 0.14901274971489453, 'gradnorm/compressor.layer4': 0.13403617702721674, 'gradnorm/compressor.layer5': 0.12592518668169425, 'gradnorm/compressor.layer6': 0.12316874493676211, 'gradnorm/compressor.layer7': 0.11932241531779036, 'gradnorm/decoder.layer0': 0.15748658211069147, 'gradnorm/decoder.layer1': 0.15795617196882594, 'gradnorm/decoder.layer2': 0.22038045437132778, 'gradnorm/max_param': 0.15849150717258453, 'epoch': 1.86}

 19%|█▊        | 3550/19090 [1:43:56<8:09:19,  1.89s/it]
 19%|█▊        | 3551/19090 [1:43:57<7:24:12,  1.72s/it]
 19%|█▊        | 3552/19090 [1:43:58<7:11:13,  1.67s/it]
 19%|█▊        | 3553/19090 [1:44:01<8:20:04,  1.93s/it]
 19%|█▊        | 3554/19090 [1:44:02<7:30:02,  1.74s/it]
 19%|█▊        | 3555/19090 [1:44:04<7:24:22,  1.72s/it]
 19%|█▊        | 3556/19090 [1:44:06<7:52:21,  1.82s/it]
 19%|█▊        | 3557/19090 [1:44:08<7:27:13,  1.73s/it]
 19%|█▊        | 3558/19090 [1:44:10<8:01:12,  1.86s/it]
 19%|█▊        | 3559/19090 [1:44:11<7:37:19,  1.77s/it]
 19%|█▊        | 3560/19090 [1:44:13<7:22:18,  1.71s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.4129175543785095, 'learning_rate': 4.887076615142456e-05, 'recon_loss': 0.0036972282745409758, 'recon_static': 0.00273876974824816, 'recon_motion': 0.0009584585262928158, 'recon_norm': 2.11528742313385, 'recon_norm_std': 0.07683103922754526, 'target_norm': 1.619247528910637, 'recon_cos': 0.4350390061736107, 'gradnorm/compressor.query': 0.00624990475989956, 'gradnorm/compressor.layer0': 0.10891104541904181, 'gradnorm/compressor.layer1': 0.1483837519762418, 'gradnorm/compressor.layer2': 0.12622667221033404, 'gradnorm/compressor.layer3': 0.12101000535746258, 'gradnorm/compressor.layer4': 0.10790240143404052, 'gradnorm/compressor.layer5': 0.10220618484732698, 'gradnorm/compressor.layer6': 0.09884264786504114, 'gradnorm/compressor.layer7': 0.096501557175458, 'gradnorm/decoder.layer0': 0.12719378768647044, 'gradnorm/decoder.layer1': 0.12797766797637541, 'gradnorm/decoder.layer2': 0.17966053594538703, 'gradnorm/max_param': 0.1287606656551361, 'epoch': 1.86}

 19%|█▊        | 3560/19090 [1:44:13<7:22:18,  1.71s/it]
 19%|█▊        | 3561/19090 [1:44:15<8:01:55,  1.86s/it]
 19%|█▊        | 3562/19090 [1:44:17<7:37:32,  1.77s/it]
 19%|█▊        | 3563/19090 [1:44:18<7:37:24,  1.77s/it]
 19%|█▊        | 3564/19090 [1:44:20<7:45:36,  1.80s/it]
 19%|█▊        | 3565/19090 [1:44:22<7:38:32,  1.77s/it]
 19%|█▊        | 3566/19090 [1:44:24<7:58:01,  1.85s/it]
 19%|█▊        | 3567/19090 [1:44:25<7:17:46,  1.69s/it]
 19%|█▊        | 3568/19090 [1:44:27<7:34:56,  1.76s/it]
 19%|█▊        | 3569/19090 [1:44:29<8:12:10,  1.90s/it]
 19%|█▊        | 3570/19090 [1:44:31<7:24:42,  1.72s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.45485496520996094, 'learning_rate': 4.885714255694698e-05, 'recon_loss': 0.0033240055607166143, 'recon_static': 0.002355857833754271, 'recon_motion': 0.000968147833191324, 'recon_norm': 2.0124078810215, 'recon_norm_std': 0.08039692156016827, 'target_norm': 1.6220686554908752, 'recon_cos': 0.4575134560465813, 'gradnorm/compressor.query': 0.006802833176707754, 'gradnorm/compressor.layer0': 0.11961115101991314, 'gradnorm/compressor.layer1': 0.16264007633610333, 'gradnorm/compressor.layer2': 0.1405519255875167, 'gradnorm/compressor.layer3': 0.13287822415476525, 'gradnorm/compressor.layer4': 0.12039729542357956, 'gradnorm/compressor.layer5': 0.11410657540618821, 'gradnorm/compressor.layer6': 0.10741114562811185, 'gradnorm/compressor.layer7': 0.10711959878415073, 'gradnorm/decoder.layer0': 0.14080958041115987, 'gradnorm/decoder.layer1': 0.14202627746136443, 'gradnorm/decoder.layer2': 0.19532100658965545, 'gradnorm/max_param': 0.1408049315214157, 'epoch': 1.87}

 19%|█▊        | 3570/19090 [1:44:31<7:24:42,  1.72s/it]
 19%|█▊        | 3571/19090 [1:44:32<7:10:08,  1.66s/it]
 19%|█▊        | 3572/19090 [1:44:35<8:09:24,  1.89s/it]
 19%|█▊        | 3573/19090 [1:44:36<7:41:17,  1.78s/it]
 19%|█▊        | 3574/19090 [1:44:38<7:25:43,  1.72s/it]
 19%|█▊        | 3575/19090 [1:44:40<8:19:33,  1.93s/it]
 19%|█▊        | 3576/19090 [1:44:42<7:30:06,  1.74s/it]
 19%|█▊        | 3577/19090 [1:44:43<7:26:21,  1.73s/it]
 19%|█▊        | 3578/19090 [1:44:46<8:13:12,  1.91s/it]
 19%|█▊        | 3579/19090 [1:44:47<7:26:02,  1.73s/it]
 19%|█▉        | 3580/19090 [1:44:48<7:15:19,  1.68s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.34508171677589417, 'learning_rate': 4.884343919585221e-05, 'recon_loss': 0.002991085307439789, 'recon_static': 0.002042196621187031, 'recon_motion': 0.0009488888012128882, 'recon_norm': 1.920443731546402, 'recon_norm_std': 0.08602794166654348, 'target_norm': 1.6200886845588685, 'recon_cos': 0.483566140383482, 'gradnorm/compressor.query': 0.00525125345970047, 'gradnorm/compressor.layer0': 0.09038308221771992, 'gradnorm/compressor.layer1': 0.12370289658677981, 'gradnorm/compressor.layer2': 0.10738666564284027, 'gradnorm/compressor.layer3': 0.1026574223740287, 'gradnorm/compressor.layer4': 0.08925171336698544, 'gradnorm/compressor.layer5': 0.08477787597303577, 'gradnorm/compressor.layer6': 0.08259445378158478, 'gradnorm/compressor.layer7': 0.08165910867504035, 'gradnorm/decoder.layer0': 0.10746689199980689, 'gradnorm/decoder.layer1': 0.10673049905368383, 'gradnorm/decoder.layer2': 0.14808145914271895, 'gradnorm/max_param': 0.10716463625431061, 'epoch': 1.88}

 19%|█▉        | 3580/19090 [1:44:48<7:15:19,  1.68s/it]
 19%|█▉        | 3581/19090 [1:44:51<8:07:58,  1.89s/it]
 19%|█▉        | 3582/19090 [1:44:52<7:22:38,  1.71s/it]
 19%|█▉        | 3583/19090 [1:44:54<7:27:21,  1.73s/it]
 19%|█▉        | 3584/19090 [1:44:56<8:31:18,  1.98s/it]
 19%|█▉        | 3585/19090 [1:44:58<7:37:53,  1.77s/it]
 19%|█▉        | 3586/19090 [1:44:59<7:00:24,  1.63s/it]
 19%|█▉        | 3587/19090 [1:45:02<8:20:45,  1.94s/it]
 19%|█▉        | 3588/19090 [1:45:03<7:31:02,  1.75s/it]
 19%|█▉        | 3589/19090 [1:45:04<6:55:09,  1.61s/it]
 19%|█▉        | 3590/19090 [1:45:07<8:14:41,  1.91s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.49854499101638794, 'learning_rate': 4.8829656113957635e-05, 'recon_loss': 0.003491675725672394, 'recon_static': 0.0025282439280999826, 'recon_motion': 0.0009634319023462013, 'recon_norm': 2.056418091058731, 'recon_norm_std': 0.07959924750030041, 'target_norm': 1.6218526631593704, 'recon_cos': 0.44869776889681817, 'gradnorm/compressor.query': 0.0077962944777338235, 'gradnorm/compressor.layer0': 0.1318751877177742, 'gradnorm/compressor.layer1': 0.17593066435149318, 'gradnorm/compressor.layer2': 0.1543890581119997, 'gradnorm/compressor.layer3': 0.14777370781587648, 'gradnorm/compressor.layer4': 0.13069853933650233, 'gradnorm/compressor.layer5': 0.12374014301009166, 'gradnorm/compressor.layer6': 0.11996839969024428, 'gradnorm/compressor.layer7': 0.11830272499138975, 'gradnorm/decoder.layer0': 0.15620639311735413, 'gradnorm/decoder.layer1': 0.15414543808481118, 'gradnorm/decoder.layer2': 0.2133720972995614, 'gradnorm/max_param': 0.15414267778396606, 'epoch': 1.88}

 19%|█▉        | 3590/19090 [1:45:07<8:14:41,  1.91s/it]
 19%|█▉        | 3591/19090 [1:45:08<7:26:03,  1.73s/it]
 19%|█▉        | 3592/19090 [1:45:10<7:09:59,  1.66s/it]
 19%|█▉        | 3593/19090 [1:45:12<8:17:02,  1.92s/it]
 19%|█▉        | 3594/19090 [1:45:14<7:39:22,  1.78s/it]
 19%|█▉        | 3595/19090 [1:45:15<7:07:37,  1.66s/it]
 19%|█▉        | 3596/19090 [1:45:18<8:14:30,  1.91s/it]
 19%|█▉        | 3597/19090 [1:45:19<7:54:53,  1.84s/it]
 19%|█▉        | 3598/19090 [1:45:21<7:13:07,  1.68s/it]
 19%|█▉        | 3599/19090 [1:45:23<8:24:05,  1.95s/it]
 19%|█▉        | 3600/19090 [1:45:24<7:35:23,  1.76s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.5204071998596191, 'learning_rate': 4.88157933573471e-05, 'recon_loss': 0.004381551535334438, 'recon_static': 0.0034215092484373598, 'recon_motion': 0.0009600423014489934, 'recon_norm': 2.296695476770401, 'recon_norm_std': 0.07376710083335639, 'target_norm': 1.620248031616211, 'recon_cos': 0.40173090174794196, 'gradnorm/compressor.query': 0.008090905005328868, 'gradnorm/compressor.layer0': 0.14222708035914083, 'gradnorm/compressor.layer1': 0.1870919297115277, 'gradnorm/compressor.layer2': 0.1606082237514871, 'gradnorm/compressor.layer3': 0.1534346136221771, 'gradnorm/compressor.layer4': 0.13674254891479065, 'gradnorm/compressor.layer5': 0.1286430272351016, 'gradnorm/compressor.layer6': 0.12152444479171405, 'gradnorm/compressor.layer7': 0.12143957072341462, 'gradnorm/decoder.layer0': 0.16095213152289328, 'gradnorm/decoder.layer1': 0.15956430832435473, 'gradnorm/decoder.layer2': 0.22373960685300653, 'gradnorm/max_param': 0.16182395815849304, 'epoch': 1.89}

 19%|█▉        | 3600/19090 [1:45:24<7:35:23,  1.76s/it]
 19%|█▉        | 3601/19090 [1:45:26<7:06:02,  1.65s/it]
 19%|█▉        | 3602/19090 [1:45:28<8:21:56,  1.94s/it]
 19%|█▉        | 3603/19090 [1:45:30<7:39:54,  1.78s/it]
 19%|█▉        | 3604/19090 [1:45:31<7:26:10,  1.73s/it]
 19%|█▉        | 3605/19090 [1:45:34<8:26:17,  1.96s/it]
 19%|█▉        | 3606/19090 [1:45:35<7:38:42,  1.78s/it]
 19%|█▉        | 3607/19090 [1:45:37<7:06:39,  1.65s/it]
 19%|█▉        | 3608/19090 [1:45:39<8:36:12,  2.00s/it]
 19%|█▉        | 3609/19090 [1:45:41<7:41:35,  1.79s/it]
 19%|█▉        | 3610/19090 [1:45:42<7:02:50,  1.64s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.42861583828926086, 'learning_rate': 4.880185097237091e-05, 'recon_loss': 0.0035860735515598208, 'recon_static': 0.002634234155993909, 'recon_motion': 0.0009518393431790173, 'recon_norm': 2.091443994641304, 'recon_norm_std': 0.07957405392080545, 'target_norm': 1.6193924248218536, 'recon_cos': 0.442910398542881, 'gradnorm/compressor.query': 0.006358726755940931, 'gradnorm/compressor.layer0': 0.113369890588913, 'gradnorm/compressor.layer1': 0.15302272027680627, 'gradnorm/compressor.layer2': 0.13420168045355524, 'gradnorm/compressor.layer3': 0.1276867649465196, 'gradnorm/compressor.layer4': 0.11214379519902981, 'gradnorm/compressor.layer5': 0.10485949088294706, 'gradnorm/compressor.layer6': 0.10245016422417977, 'gradnorm/compressor.layer7': 0.09846121430473521, 'gradnorm/decoder.layer0': 0.1322752872662371, 'gradnorm/decoder.layer1': 0.13285354372749564, 'gradnorm/decoder.layer2': 0.1848761733890306, 'gradnorm/max_param': 0.13362815976142883, 'epoch': 1.89}

 19%|█▉        | 3610/19090 [1:45:42<7:02:50,  1.64s/it]
 19%|█▉        | 3611/19090 [1:45:45<8:23:26,  1.95s/it]
 19%|█▉        | 3612/19090 [1:45:46<7:45:05,  1.80s/it]
 19%|█▉        | 3613/19090 [1:45:47<7:05:16,  1.65s/it]
 19%|█▉        | 3614/19090 [1:45:50<8:38:18,  2.01s/it]
 19%|█▉        | 3615/19090 [1:45:52<7:42:46,  1.79s/it]
 19%|█▉        | 3616/19090 [1:45:53<7:03:46,  1.64s/it]
 19%|█▉        | 3617/19090 [1:45:56<8:31:31,  1.98s/it]
 19%|█▉        | 3618/19090 [1:45:57<7:38:04,  1.78s/it]
 19%|█▉        | 3619/19090 [1:45:58<7:00:52,  1.63s/it]
 19%|█▉        | 3620/19090 [1:46:01<8:11:56,  1.91s/it]
                                                        
{'loss': 0.0027, 'grad_norm': 0.30485212802886963, 'learning_rate': 4.878782900564556e-05, 'recon_loss': 0.002739538630703464, 'recon_static': 0.0017767051322152838, 'recon_motion': 0.0009628334868466481, 'recon_norm': 1.8340937227010727, 'recon_norm_std': 0.0859028534963727, 'target_norm': 1.6200215071439743, 'recon_cos': 0.5009086340665817, 'gradnorm/compressor.query': 0.004793965318358413, 'gradnorm/compressor.layer0': 0.08202254726022361, 'gradnorm/compressor.layer1': 0.11013693741669978, 'gradnorm/compressor.layer2': 0.09374147118079268, 'gradnorm/compressor.layer3': 0.09006381036081809, 'gradnorm/compressor.layer4': 0.07905306919114177, 'gradnorm/compressor.layer5': 0.07433563991092164, 'gradnorm/compressor.layer6': 0.07251384351161723, 'gradnorm/compressor.layer7': 0.07131173152283848, 'gradnorm/decoder.layer0': 0.09386247107109988, 'gradnorm/decoder.layer1': 0.09391831150761101, 'gradnorm/decoder.layer2': 0.13192967508568645, 'gradnorm/max_param': 0.09504997730255127, 'epoch': 1.9}

 19%|█▉        | 3620/19090 [1:46:01<8:11:56,  1.91s/it]
 19%|█▉        | 3621/19090 [1:46:02<7:24:20,  1.72s/it]
 19%|█▉        | 3622/19090 [1:46:03<6:50:52,  1.59s/it]
 19%|█▉        | 3623/19090 [1:46:06<8:09:54,  1.90s/it]
 19%|█▉        | 3624/19090 [1:46:07<7:22:15,  1.72s/it]
 19%|█▉        | 3625/19090 [1:46:09<7:24:12,  1.72s/it]
 19%|█▉        | 3626/19090 [1:46:11<7:57:30,  1.85s/it]
 19%|█▉        | 3627/19090 [1:46:13<7:37:57,  1.78s/it]
 19%|█▉        | 3628/19090 [1:46:14<7:19:15,  1.70s/it]
 19%|█▉        | 3629/19090 [1:46:17<8:18:52,  1.94s/it]
 19%|█▉        | 3630/19090 [1:46:18<7:28:28,  1.74s/it]
                                                        
{'loss': 0.0026, 'grad_norm': 0.4017547070980072, 'learning_rate': 4.8773727504053665e-05, 'recon_loss': 0.0025850145902950317, 'recon_static': 0.0016386158036766573, 'recon_motion': 0.0009463987371418625, 'recon_norm': 1.7922779083251954, 'recon_norm_std': 0.09076703581959009, 'target_norm': 1.6191643089056016, 'recon_cos': 0.5167329296469688, 'gradnorm/compressor.query': 0.005908400235750047, 'gradnorm/compressor.layer0': 0.10404074826743547, 'gradnorm/compressor.layer1': 0.14356526186656682, 'gradnorm/compressor.layer2': 0.12496783915686435, 'gradnorm/compressor.layer3': 0.1199544098300654, 'gradnorm/compressor.layer4': 0.10454484468420495, 'gradnorm/compressor.layer5': 0.0982120765439125, 'gradnorm/compressor.layer6': 0.09615954728372825, 'gradnorm/compressor.layer7': 0.09568023407880352, 'gradnorm/decoder.layer0': 0.1260115035341489, 'gradnorm/decoder.layer1': 0.12576982256708719, 'gradnorm/decoder.layer2': 0.17104651983692404, 'gradnorm/max_param': 0.1231180727481842, 'epoch': 1.9}

 19%|█▉        | 3630/19090 [1:46:18<7:28:28,  1.74s/it]
 19%|█▉        | 3631/19090 [1:46:20<7:32:46,  1.76s/it]
 19%|█▉        | 3632/19090 [1:46:22<8:13:41,  1.92s/it]
 19%|█▉        | 3633/19090 [1:46:23<7:25:01,  1.73s/it]
 19%|█▉        | 3634/19090 [1:46:25<7:08:03,  1.66s/it]
 19%|█▉        | 3635/19090 [1:46:27<7:52:09,  1.83s/it]
 19%|█▉        | 3636/19090 [1:46:29<7:16:42,  1.70s/it]
 19%|█▉        | 3637/19090 [1:46:31<7:45:27,  1.81s/it]
 19%|█▉        | 3638/19090 [1:46:33<7:52:57,  1.84s/it]
 19%|█▉        | 3639/19090 [1:46:34<7:10:57,  1.67s/it]
 19%|█▉        | 3640/19090 [1:46:36<7:18:38,  1.70s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.4824238419532776, 'learning_rate': 4.875954651474375e-05, 'recon_loss': 0.0033304561802651733, 'recon_static': 0.0023773426772095265, 'recon_motion': 0.0009531135146971792, 'recon_norm': 2.017944964766502, 'recon_norm_std': 0.08168449960649013, 'target_norm': 1.619762086868286, 'recon_cos': 0.45850057527422905, 'gradnorm/compressor.query': 0.007911220912309775, 'gradnorm/compressor.layer0': 0.1293453194946268, 'gradnorm/compressor.layer1': 0.17699919599819233, 'gradnorm/compressor.layer2': 0.15122411180250517, 'gradnorm/compressor.layer3': 0.14248692195258714, 'gradnorm/compressor.layer4': 0.1256206126419601, 'gradnorm/compressor.layer5': 0.11873586485834495, 'gradnorm/compressor.layer6': 0.11325208107534096, 'gradnorm/compressor.layer7': 0.11231790547744616, 'gradnorm/decoder.layer0': 0.1493508510641498, 'gradnorm/decoder.layer1': 0.14881657833453193, 'gradnorm/decoder.layer2': 0.20412705914707954, 'gradnorm/max_param': 0.147382453083992, 'epoch': 1.91}

 19%|█▉        | 3640/19090 [1:46:36<7:18:38,  1.70s/it]
 19%|█▉        | 3641/19090 [1:46:38<8:05:29,  1.89s/it]
 19%|█▉        | 3642/19090 [1:46:39<7:32:35,  1.76s/it]
 19%|█▉        | 3643/19090 [1:46:41<7:14:42,  1.69s/it]
 19%|█▉        | 3644/19090 [1:46:43<7:50:48,  1.83s/it]
 19%|█▉        | 3645/19090 [1:46:45<8:00:03,  1.86s/it]
 19%|█▉        | 3646/19090 [1:46:46<7:15:31,  1.69s/it]
 19%|█▉        | 3647/19090 [1:46:48<7:47:38,  1.82s/it]
 19%|█▉        | 3648/19090 [1:46:51<8:23:55,  1.96s/it]
 19%|█▉        | 3649/19090 [1:46:52<7:32:09,  1.76s/it]
 19%|█▉        | 3650/19090 [1:46:54<7:41:54,  1.80s/it]
                                                        
{'loss': 0.0057, 'grad_norm': 0.7390676140785217, 'learning_rate': 4.87452860851301e-05, 'recon_loss': 0.005684305878821761, 'recon_static': 0.00472248574369587, 'recon_motion': 0.0009618201525881887, 'recon_norm': 2.589850348234177, 'recon_norm_std': 0.06788926059380174, 'target_norm': 1.6207761496305466, 'recon_cos': 0.35728998482227325, 'gradnorm/compressor.query': 0.011997158007908699, 'gradnorm/compressor.layer0': 0.1957666362566134, 'gradnorm/compressor.layer1': 0.26797412889159106, 'gradnorm/compressor.layer2': 0.2250174588476438, 'gradnorm/compressor.layer3': 0.21461981460181492, 'gradnorm/compressor.layer4': 0.19488685002735145, 'gradnorm/compressor.layer5': 0.1831194045803615, 'gradnorm/compressor.layer6': 0.1740779635017152, 'gradnorm/compressor.layer7': 0.17571946753779905, 'gradnorm/decoder.layer0': 0.2319833561567933, 'gradnorm/decoder.layer1': 0.2318585645886563, 'gradnorm/decoder.layer2': 0.3146093703014903, 'gradnorm/max_param': 0.22592779994010925, 'epoch': 1.91}

 19%|█▉        | 3650/19090 [1:46:54<7:41:54,  1.80s/it]
 19%|█▉        | 3651/19090 [1:46:56<7:59:27,  1.86s/it]
 19%|█▉        | 3652/19090 [1:46:58<7:43:21,  1.80s/it]
 19%|█▉        | 3653/19090 [1:46:59<7:29:56,  1.75s/it]
 19%|█▉        | 3654/19090 [1:47:02<8:21:36,  1.95s/it]
 19%|█▉        | 3655/19090 [1:47:03<7:30:09,  1.75s/it]
 19%|█▉        | 3656/19090 [1:47:05<7:18:42,  1.71s/it]
 19%|█▉        | 3657/19090 [1:47:07<8:30:34,  1.99s/it]
 19%|█▉        | 3658/19090 [1:47:08<7:36:51,  1.78s/it]
 19%|█▉        | 3659/19090 [1:47:10<7:25:47,  1.73s/it]
 19%|█▉        | 3660/19090 [1:47:12<8:17:48,  1.94s/it]
                                                        
{'loss': 0.0052, 'grad_norm': 0.5699127316474915, 'learning_rate': 4.873094626289263e-05, 'recon_loss': 0.0052516094874590635, 'recon_static': 0.004284968349384144, 'recon_motion': 0.0009666412122896872, 'recon_norm': 2.5040928602218626, 'recon_norm_std': 0.06748118298128247, 'target_norm': 1.6225080072879792, 'recon_cos': 0.3713330335915089, 'gradnorm/compressor.query': 0.008342976647527414, 'gradnorm/compressor.layer0': 0.14754702095380823, 'gradnorm/compressor.layer1': 0.20135366688577666, 'gradnorm/compressor.layer2': 0.1730536717873316, 'gradnorm/compressor.layer3': 0.16293361540188772, 'gradnorm/compressor.layer4': 0.14606717143303272, 'gradnorm/compressor.layer5': 0.13847813706249354, 'gradnorm/compressor.layer6': 0.1348925283265248, 'gradnorm/compressor.layer7': 0.1348700602469158, 'gradnorm/decoder.layer0': 0.18409869576465648, 'gradnorm/decoder.layer1': 0.1822932527109812, 'gradnorm/decoder.layer2': 0.24880039473779067, 'gradnorm/max_param': 0.17694979906082153, 'epoch': 1.92}

 19%|█▉        | 3660/19090 [1:47:12<8:17:48,  1.94s/it]
 19%|█▉        | 3661/19090 [1:47:14<7:27:41,  1.74s/it]
 19%|█▉        | 3662/19090 [1:47:16<7:48:35,  1.82s/it]
 19%|█▉        | 3663/19090 [1:47:18<7:47:01,  1.82s/it]
 19%|█▉        | 3664/19090 [1:47:19<7:28:02,  1.74s/it]
 19%|█▉        | 3665/19090 [1:47:21<7:45:36,  1.81s/it]
 19%|█▉        | 3666/19090 [1:47:23<8:06:23,  1.89s/it]
 19%|█▉        | 3667/19090 [1:47:24<7:19:39,  1.71s/it]
 19%|█▉        | 3668/19090 [1:47:27<8:01:12,  1.87s/it]
 19%|█▉        | 3669/19090 [1:47:29<8:06:14,  1.89s/it]
 19%|█▉        | 3670/19090 [1:47:30<7:19:14,  1.71s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.5371944308280945, 'learning_rate': 4.871652709597668e-05, 'recon_loss': 0.004300936602521688, 'recon_static': 0.0033493940951302647, 'recon_motion': 0.0009515425859717652, 'recon_norm': 2.2776224076747895, 'recon_norm_std': 0.07444693706929684, 'target_norm': 1.6219096839427949, 'recon_cos': 0.4056964419782162, 'gradnorm/compressor.query': 0.008959630474809932, 'gradnorm/compressor.layer0': 0.14656486207364075, 'gradnorm/compressor.layer1': 0.19861816119657702, 'gradnorm/compressor.layer2': 0.166884192184989, 'gradnorm/compressor.layer3': 0.1587692421980789, 'gradnorm/compressor.layer4': 0.1390120605010906, 'gradnorm/compressor.layer5': 0.13295957191833394, 'gradnorm/compressor.layer6': 0.12442949235294512, 'gradnorm/compressor.layer7': 0.12443253860541206, 'gradnorm/decoder.layer0': 0.16562051879470632, 'gradnorm/decoder.layer1': 0.16409577872315895, 'gradnorm/decoder.layer2': 0.2284355023110185, 'gradnorm/max_param': 0.16510429978370667, 'epoch': 1.92}

 19%|█▉        | 3670/19090 [1:47:30<7:19:14,  1.71s/it]
 19%|█▉        | 3671/19090 [1:47:32<7:47:20,  1.82s/it]
 19%|█▉        | 3672/19090 [1:47:34<7:54:45,  1.85s/it]
 19%|█▉        | 3673/19090 [1:47:35<7:10:55,  1.68s/it]
 19%|█▉        | 3674/19090 [1:47:37<7:20:54,  1.72s/it]
 19%|█▉        | 3675/19090 [1:47:39<7:36:56,  1.78s/it]
 19%|█▉        | 3676/19090 [1:47:40<7:15:58,  1.70s/it]
 19%|█▉        | 3677/19090 [1:47:43<7:48:41,  1.82s/it]
 19%|█▉        | 3678/19090 [1:47:44<7:39:58,  1.79s/it]
 19%|█▉        | 3679/19090 [1:47:46<7:07:45,  1.67s/it]
 19%|█▉        | 3680/19090 [1:47:48<8:07:22,  1.90s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.4286808669567108, 'learning_rate': 4.870202863259289e-05, 'recon_loss': 0.00382240453036502, 'recon_static': 0.0028552204486913977, 'recon_motion': 0.0009671840904047713, 'recon_norm': 2.146245837211609, 'recon_norm_std': 0.07742322050035, 'target_norm': 1.6227732986211776, 'recon_cos': 0.42889689803123476, 'gradnorm/compressor.query': 0.00695853534589311, 'gradnorm/compressor.layer0': 0.11125711635936744, 'gradnorm/compressor.layer1': 0.1522089288261134, 'gradnorm/compressor.layer2': 0.130170188720719, 'gradnorm/compressor.layer3': 0.12679614117986465, 'gradnorm/compressor.layer4': 0.11098266424331824, 'gradnorm/compressor.layer5': 0.10572171780070896, 'gradnorm/compressor.layer6': 0.10220126870048121, 'gradnorm/compressor.layer7': 0.10173054236368481, 'gradnorm/decoder.layer0': 0.13690265820251307, 'gradnorm/decoder.layer1': 0.13455365275778225, 'gradnorm/decoder.layer2': 0.18443472993891175, 'gradnorm/max_param': 0.13215987384319305, 'epoch': 1.93}

 19%|█▉        | 3680/19090 [1:47:48<8:07:22,  1.90s/it]
 19%|█▉        | 3681/19090 [1:47:50<7:38:48,  1.79s/it]
 19%|█▉        | 3682/19090 [1:47:51<7:17:33,  1.70s/it]
 19%|█▉        | 3683/19090 [1:47:53<7:53:34,  1.84s/it]
 19%|█▉        | 3684/19090 [1:47:55<8:00:49,  1.87s/it]
 19%|█▉        | 3685/19090 [1:47:57<7:26:16,  1.74s/it]
 19%|█▉        | 3686/19090 [1:47:59<7:57:42,  1.86s/it]
 19%|█▉        | 3687/19090 [1:48:01<7:47:10,  1.82s/it]
 19%|█▉        | 3688/19090 [1:48:02<7:05:57,  1.66s/it]
 19%|█▉        | 3689/19090 [1:48:04<7:48:42,  1.83s/it]
 19%|█▉        | 3690/19090 [1:48:06<7:49:48,  1.83s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.4138762354850769, 'learning_rate': 4.868745092121705e-05, 'recon_loss': 0.0033824745565652846, 'recon_static': 0.0024143252405337988, 'recon_motion': 0.0009681493189418689, 'recon_norm': 2.0307738155126573, 'recon_norm_std': 0.07909092418849469, 'target_norm': 1.6215676009654998, 'recon_cos': 0.45407113879919053, 'gradnorm/compressor.query': 0.006101554615181879, 'gradnorm/compressor.layer0': 0.10783177977402902, 'gradnorm/compressor.layer1': 0.14774351014085887, 'gradnorm/compressor.layer2': 0.12726037917967037, 'gradnorm/compressor.layer3': 0.12148995723962286, 'gradnorm/compressor.layer4': 0.10749628350991651, 'gradnorm/compressor.layer5': 0.10130530583503385, 'gradnorm/compressor.layer6': 0.09763001566413651, 'gradnorm/compressor.layer7': 0.09749826526779774, 'gradnorm/decoder.layer0': 0.13112954106936164, 'gradnorm/decoder.layer1': 0.128080126708238, 'gradnorm/decoder.layer2': 0.17997281601067266, 'gradnorm/max_param': 0.1271296739578247, 'epoch': 1.93}

 19%|█▉        | 3690/19090 [1:48:06<7:49:48,  1.83s/it]
 19%|█▉        | 3691/19090 [1:48:07<7:07:46,  1.67s/it]
 19%|█▉        | 3692/19090 [1:48:09<7:35:06,  1.77s/it]
 19%|█▉        | 3693/19090 [1:48:11<7:30:07,  1.75s/it]
 19%|█▉        | 3694/19090 [1:48:13<7:28:04,  1.75s/it]
 19%|█▉        | 3695/19090 [1:48:15<8:04:02,  1.89s/it]
 19%|█▉        | 3696/19090 [1:48:16<7:40:36,  1.80s/it]
 19%|█▉        | 3697/19090 [1:48:18<7:39:31,  1.79s/it]
 19%|█▉        | 3698/19090 [1:48:20<8:01:22,  1.88s/it]
 19%|█▉        | 3699/19090 [1:48:22<7:38:03,  1.79s/it]
 19%|█▉        | 3700/19090 [1:48:23<7:20:10,  1.72s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.5045895576477051, 'learning_rate': 4.867279401058987e-05, 'recon_loss': 0.003295301308389753, 'recon_static': 0.002333318954333663, 'recon_motion': 0.0009619823089451529, 'recon_norm': 2.0030495554208754, 'recon_norm_std': 0.0814401876181364, 'target_norm': 1.6234426558017732, 'recon_cos': 0.4624225653707981, 'gradnorm/compressor.query': 0.008187269956167953, 'gradnorm/compressor.layer0': 0.13478364685743527, 'gradnorm/compressor.layer1': 0.1834230542546745, 'gradnorm/compressor.layer2': 0.15651110647898997, 'gradnorm/compressor.layer3': 0.14925664200931568, 'gradnorm/compressor.layer4': 0.13004959125700224, 'gradnorm/compressor.layer5': 0.12588681861271453, 'gradnorm/compressor.layer6': 0.12020421558318284, 'gradnorm/compressor.layer7': 0.1202602658581301, 'gradnorm/decoder.layer0': 0.15621368463022337, 'gradnorm/decoder.layer1': 0.15609143675463782, 'gradnorm/decoder.layer2': 0.2133551088938174, 'gradnorm/max_param': 0.15359432995319366, 'epoch': 1.94}

 19%|█▉        | 3700/19090 [1:48:23<7:20:10,  1.72s/it]
 19%|█▉        | 3701/19090 [1:48:26<7:53:10,  1.84s/it]
 19%|█▉        | 3702/19090 [1:48:27<7:43:16,  1.81s/it]
 19%|█▉        | 3703/19090 [1:48:29<7:34:46,  1.77s/it]
 19%|█▉        | 3704/19090 [1:48:31<7:32:54,  1.77s/it]
 19%|█▉        | 3705/19090 [1:48:33<7:36:06,  1.78s/it]
 19%|█▉        | 3706/19090 [1:48:34<7:46:14,  1.82s/it]
 19%|█▉        | 3707/19090 [1:48:36<7:45:24,  1.82s/it]
 19%|█▉        | 3708/19090 [1:48:38<7:26:25,  1.74s/it]
 19%|█▉        | 3709/19090 [1:48:40<8:22:55,  1.96s/it]
 19%|█▉        | 3710/19090 [1:48:42<7:45:54,  1.82s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.5545048713684082, 'learning_rate': 4.8658057949716905e-05, 'recon_loss': 0.0046283977222628895, 'recon_static': 0.003666808019625023, 'recon_motion': 0.0009615896487957798, 'recon_norm': 2.3585257112979887, 'recon_norm_std': 0.07066307626664639, 'target_norm': 1.6213460385799408, 'recon_cos': 0.3924817368388176, 'gradnorm/compressor.query': 0.009122536462690508, 'gradnorm/compressor.layer0': 0.1510709447820361, 'gradnorm/compressor.layer1': 0.20230052589427347, 'gradnorm/compressor.layer2': 0.16956601753235126, 'gradnorm/compressor.layer3': 0.16631707143539673, 'gradnorm/compressor.layer4': 0.14285751951010472, 'gradnorm/compressor.layer5': 0.1375375479669235, 'gradnorm/compressor.layer6': 0.13027849373449002, 'gradnorm/compressor.layer7': 0.13100022237382097, 'gradnorm/decoder.layer0': 0.17375266119407828, 'gradnorm/decoder.layer1': 0.17196477907716512, 'gradnorm/decoder.layer2': 0.23240104549900356, 'gradnorm/max_param': 0.16884976625442505, 'epoch': 1.94}

 19%|█▉        | 3710/19090 [1:48:42<7:45:54,  1.82s/it]
 19%|█▉        | 3711/19090 [1:48:43<7:14:16,  1.69s/it]
 19%|█▉        | 3712/19090 [1:48:45<7:56:20,  1.86s/it]
 19%|█▉        | 3713/19090 [1:48:47<7:58:36,  1.87s/it]
 19%|█▉        | 3714/19090 [1:48:49<7:13:32,  1.69s/it]
 19%|█▉        | 3715/19090 [1:48:51<7:57:47,  1.86s/it]
 19%|█▉        | 3716/19090 [1:48:52<7:15:24,  1.70s/it]
 19%|█▉        | 3717/19090 [1:48:54<7:36:05,  1.78s/it]
 19%|█▉        | 3718/19090 [1:48:56<7:59:17,  1.87s/it]
 19%|█▉        | 3719/19090 [1:48:58<7:15:30,  1.70s/it]
 19%|█▉        | 3720/19090 [1:48:59<7:22:17,  1.73s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.5073797106742859, 'learning_rate': 4.864324278786835e-05, 'recon_loss': 0.004154638614272699, 'recon_static': 0.003195382602280006, 'recon_motion': 0.0009592560541932471, 'recon_norm': 2.2391177117824554, 'recon_norm_std': 0.07395176999270917, 'target_norm': 1.6208079010248184, 'recon_cos': 0.41194583401083945, 'gradnorm/compressor.query': 0.007903993359640978, 'gradnorm/compressor.layer0': 0.13334078764815363, 'gradnorm/compressor.layer1': 0.18130645982463944, 'gradnorm/compressor.layer2': 0.15712418984778614, 'gradnorm/compressor.layer3': 0.14986988424457925, 'gradnorm/compressor.layer4': 0.13087718756987082, 'gradnorm/compressor.layer5': 0.1257957595251698, 'gradnorm/compressor.layer6': 0.12172876641851696, 'gradnorm/compressor.layer7': 0.12040984160667575, 'gradnorm/decoder.layer0': 0.1589484026466222, 'gradnorm/decoder.layer1': 0.1606105200293145, 'gradnorm/decoder.layer2': 0.21503660042970468, 'gradnorm/max_param': 0.15438291430473328, 'epoch': 1.95}

 19%|█▉        | 3720/19090 [1:48:59<7:22:17,  1.73s/it]
 19%|█▉        | 3721/19090 [1:49:02<8:15:34,  1.93s/it]
 19%|█▉        | 3722/19090 [1:49:03<7:25:16,  1.74s/it]
 20%|█▉        | 3723/19090 [1:49:04<6:50:15,  1.60s/it]
 20%|█▉        | 3724/19090 [1:49:07<7:57:37,  1.86s/it]
 20%|█▉        | 3725/19090 [1:49:08<7:12:39,  1.69s/it]
 20%|█▉        | 3726/19090 [1:49:09<6:47:28,  1.59s/it]
 20%|█▉        | 3727/19090 [1:49:12<7:50:25,  1.84s/it]
 20%|█▉        | 3728/19090 [1:49:13<7:07:29,  1.67s/it]
 20%|█▉        | 3729/19090 [1:49:15<7:02:05,  1.65s/it]
 20%|█▉        | 3730/19090 [1:49:17<7:57:27,  1.87s/it]
                                                        
{'loss': 0.0047, 'grad_norm': 0.616204023361206, 'learning_rate': 4.8628348574578845e-05, 'recon_loss': 0.004687235073652119, 'recon_static': 0.003723097359761596, 'recon_motion': 0.0009641375843784771, 'recon_norm': 2.372179186344147, 'recon_norm_std': 0.06986184865236282, 'target_norm': 1.619863510131836, 'recon_cos': 0.3900036014616489, 'gradnorm/compressor.query': 0.009579711627195822, 'gradnorm/compressor.layer0': 0.15980326901563693, 'gradnorm/compressor.layer1': 0.22157770911717659, 'gradnorm/compressor.layer2': 0.1926557473441417, 'gradnorm/compressor.layer3': 0.18405577171281934, 'gradnorm/compressor.layer4': 0.1575187373274338, 'gradnorm/compressor.layer5': 0.15221310784892927, 'gradnorm/compressor.layer6': 0.14442820355182326, 'gradnorm/compressor.layer7': 0.14307518759588034, 'gradnorm/decoder.layer0': 0.19129017794320657, 'gradnorm/decoder.layer1': 0.1899439559020111, 'gradnorm/decoder.layer2': 0.26834213168121696, 'gradnorm/max_param': 0.1901736855506897, 'epoch': 1.95}

 20%|█▉        | 3730/19090 [1:49:17<7:57:27,  1.87s/it]
 20%|█▉        | 3731/19090 [1:49:18<7:13:24,  1.69s/it]
 20%|█▉        | 3732/19090 [1:49:20<6:50:51,  1.61s/it]
 20%|█▉        | 3733/19090 [1:49:23<8:15:56,  1.94s/it]
 20%|█▉        | 3734/19090 [1:49:24<7:25:51,  1.74s/it]
 20%|█▉        | 3735/19090 [1:49:25<7:16:50,  1.71s/it]
 20%|█▉        | 3736/19090 [1:49:28<7:52:46,  1.85s/it]
 20%|█▉        | 3737/19090 [1:49:29<7:27:23,  1.75s/it]
 20%|█▉        | 3738/19090 [1:49:31<7:52:11,  1.85s/it]
 20%|█▉        | 3739/19090 [1:49:33<8:04:17,  1.89s/it]
 20%|█▉        | 3740/19090 [1:49:35<7:41:25,  1.80s/it]
                                                        
{'loss': 0.0048, 'grad_norm': 0.5199496746063232, 'learning_rate': 4.8613375359647346e-05, 'recon_loss': 0.004774256120435893, 'recon_static': 0.0038179198163561523, 'recon_motion': 0.0009563362764311023, 'recon_norm': 2.3929082810878755, 'recon_norm_std': 0.07066057696938514, 'target_norm': 1.6186416804790498, 'recon_cos': 0.3859380818903446, 'gradnorm/compressor.query': 0.00854642122602032, 'gradnorm/compressor.layer0': 0.13916789920079298, 'gradnorm/compressor.layer1': 0.18938217604294819, 'gradnorm/compressor.layer2': 0.16080203929359424, 'gradnorm/compressor.layer3': 0.15542264283411367, 'gradnorm/compressor.layer4': 0.13356336644087544, 'gradnorm/compressor.layer5': 0.1253823898558357, 'gradnorm/compressor.layer6': 0.12120843271291525, 'gradnorm/compressor.layer7': 0.11978948287079053, 'gradnorm/decoder.layer0': 0.16294623004468256, 'gradnorm/decoder.layer1': 0.15942593116015746, 'gradnorm/decoder.layer2': 0.22463039631536338, 'gradnorm/max_param': 0.1610863357782364, 'epoch': 1.96}

 20%|█▉        | 3740/19090 [1:49:35<7:41:25,  1.80s/it]
 20%|█▉        | 3741/19090 [1:49:36<7:25:33,  1.74s/it]
 20%|█▉        | 3742/19090 [1:49:39<8:00:20,  1.88s/it]
 20%|█▉        | 3743/19090 [1:49:40<7:33:53,  1.77s/it]
 20%|█▉        | 3744/19090 [1:49:42<7:25:13,  1.74s/it]
 20%|█▉        | 3745/19090 [1:49:44<8:38:56,  2.03s/it]
 20%|█▉        | 3746/19090 [1:49:46<7:42:08,  1.81s/it]
 20%|█▉        | 3747/19090 [1:49:47<7:02:28,  1.65s/it]
 20%|█▉        | 3748/19090 [1:49:50<8:31:57,  2.00s/it]
 20%|█▉        | 3749/19090 [1:49:51<7:41:57,  1.81s/it]
 20%|█▉        | 3750/19090 [1:49:53<7:02:07,  1.65s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.4124850034713745, 'learning_rate': 4.859832319313697e-05, 'recon_loss': 0.0036492628045380116, 'recon_static': 0.002690607070690021, 'recon_motion': 0.0009586558080627583, 'recon_norm': 2.104157066345215, 'recon_norm_std': 0.0781549809500575, 'target_norm': 1.6215163052082062, 'recon_cos': 0.4393013522028923, 'gradnorm/compressor.query': 0.00688097848367508, 'gradnorm/compressor.layer0': 0.11088842008117819, 'gradnorm/compressor.layer1': 0.1486261198364176, 'gradnorm/compressor.layer2': 0.1287826465678146, 'gradnorm/compressor.layer3': 0.12230086108313565, 'gradnorm/compressor.layer4': 0.10692021534462423, 'gradnorm/compressor.layer5': 0.10025419230358071, 'gradnorm/compressor.layer6': 0.09638917991332852, 'gradnorm/compressor.layer7': 0.09648438194985307, 'gradnorm/decoder.layer0': 0.12801541666418775, 'gradnorm/decoder.layer1': 0.12907732102591987, 'gradnorm/decoder.layer2': 0.17617105073587985, 'gradnorm/max_param': 0.12694880366325378, 'epoch': 1.96}

 20%|█▉        | 3750/19090 [1:49:53<7:02:07,  1.65s/it]
 20%|█▉        | 3751/19090 [1:49:55<8:31:43,  2.00s/it]
 20%|█▉        | 3752/19090 [1:49:57<7:36:42,  1.79s/it]
 20%|█▉        | 3753/19090 [1:49:58<6:58:00,  1.64s/it]
 20%|█▉        | 3754/19090 [1:50:01<8:26:32,  1.98s/it]
 20%|█▉        | 3755/19090 [1:50:02<7:33:38,  1.77s/it]
 20%|█▉        | 3756/19090 [1:50:03<6:57:05,  1.63s/it]
 20%|█▉        | 3757/19090 [1:50:06<8:51:23,  2.08s/it]
 20%|█▉        | 3758/19090 [1:50:08<7:50:42,  1.84s/it]
 20%|█▉        | 3759/19090 [1:50:09<7:08:08,  1.68s/it]
 20%|█▉        | 3760/19090 [1:50:12<8:26:47,  1.98s/it]
                                                        
{'loss': 0.0028, 'grad_norm': 0.33873891830444336, 'learning_rate': 4.858319212537481e-05, 'recon_loss': 0.002769832272315398, 'recon_static': 0.0018173881515394896, 'recon_motion': 0.0009524441222310998, 'recon_norm': 1.8506417661905288, 'recon_norm_std': 0.08766601830720902, 'target_norm': 1.621976301074028, 'recon_cos': 0.5008156582713127, 'gradnorm/compressor.query': 0.0048658358971777945, 'gradnorm/compressor.layer0': 0.08609697056802716, 'gradnorm/compressor.layer1': 0.12013094908494663, 'gradnorm/compressor.layer2': 0.10653624019477696, 'gradnorm/compressor.layer3': 0.10193046065787455, 'gradnorm/compressor.layer4': 0.08897158495657541, 'gradnorm/compressor.layer5': 0.08298024673514819, 'gradnorm/compressor.layer6': 0.08108371479127899, 'gradnorm/compressor.layer7': 0.07992541184621275, 'gradnorm/decoder.layer0': 0.10625983623947183, 'gradnorm/decoder.layer1': 0.10630130212977418, 'gradnorm/decoder.layer2': 0.1441562918028904, 'gradnorm/max_param': 0.10384652763605118, 'epoch': 1.97}

 20%|█▉        | 3760/19090 [1:50:12<8:26:47,  1.98s/it]
 20%|█▉        | 3761/19090 [1:50:13<7:33:18,  1.77s/it]
 20%|█▉        | 3762/19090 [1:50:14<6:55:37,  1.63s/it]
 20%|█▉        | 3763/19090 [1:50:17<7:55:01,  1.86s/it]
 20%|█▉        | 3764/19090 [1:50:18<7:10:53,  1.69s/it]
 20%|█▉        | 3765/19090 [1:50:19<6:41:22,  1.57s/it]
 20%|█▉        | 3766/19090 [1:50:22<8:39:49,  2.04s/it]
 20%|█▉        | 3767/19090 [1:50:24<7:42:32,  1.81s/it]
 20%|█▉        | 3768/19090 [1:50:25<7:01:58,  1.65s/it]
 20%|█▉        | 3769/19090 [1:50:28<8:39:28,  2.03s/it]
 20%|█▉        | 3770/19090 [1:50:29<7:42:08,  1.81s/it]
                                                        
{'loss': 0.0027, 'grad_norm': 0.36244872212409973, 'learning_rate': 4.8567982206951756e-05, 'recon_loss': 0.0026668350503314285, 'recon_static': 0.0017125097278039903, 'recon_motion': 0.0009543253036099486, 'recon_norm': 1.819308865070343, 'recon_norm_std': 0.08676360063254833, 'target_norm': 1.6204681247472763, 'recon_cos': 0.5080894507467747, 'gradnorm/compressor.query': 0.005910912187713153, 'gradnorm/compressor.layer0': 0.09623201839344396, 'gradnorm/compressor.layer1': 0.13157817932896068, 'gradnorm/compressor.layer2': 0.1144667817453474, 'gradnorm/compressor.layer3': 0.10646403436767747, 'gradnorm/compressor.layer4': 0.09229486908839, 'gradnorm/compressor.layer5': 0.08820391027001324, 'gradnorm/compressor.layer6': 0.08375277578234035, 'gradnorm/compressor.layer7': 0.08397663435638786, 'gradnorm/decoder.layer0': 0.11572561380976991, 'gradnorm/decoder.layer1': 0.11302773702502612, 'gradnorm/decoder.layer2': 0.15424307355987069, 'gradnorm/max_param': 0.10960523784160614, 'epoch': 1.97}

 20%|█▉        | 3770/19090 [1:50:29<7:42:08,  1.81s/it]
 20%|█▉        | 3771/19090 [1:50:30<7:01:50,  1.65s/it]
 20%|█▉        | 3772/19090 [1:50:33<8:19:42,  1.96s/it]
 20%|█▉        | 3773/19090 [1:50:34<7:27:36,  1.75s/it]
 20%|█▉        | 3774/19090 [1:50:36<7:28:31,  1.76s/it]
 20%|█▉        | 3775/19090 [1:50:39<8:22:33,  1.97s/it]
 20%|█▉        | 3776/19090 [1:50:40<7:30:05,  1.76s/it]
 20%|█▉        | 3777/19090 [1:50:41<7:04:15,  1.66s/it]
 20%|█▉        | 3778/19090 [1:50:44<8:33:12,  2.01s/it]
 20%|█▉        | 3779/19090 [1:50:45<7:37:40,  1.79s/it]
 20%|█▉        | 3780/19090 [1:50:47<7:15:49,  1.71s/it]
                                                        
{'loss': 0.0028, 'grad_norm': 0.39742594957351685, 'learning_rate': 4.8552693488722325e-05, 'recon_loss': 0.0028453352802898734, 'recon_static': 0.0018850117456167936, 'recon_motion': 0.000960323512845207, 'recon_norm': 1.8712464421987534, 'recon_norm_std': 0.08521873988211155, 'target_norm': 1.6242163717746734, 'recon_cos': 0.4928517580032349, 'gradnorm/compressor.query': 0.006222927119451973, 'gradnorm/compressor.layer0': 0.106527506740443, 'gradnorm/compressor.layer1': 0.14558050400871134, 'gradnorm/compressor.layer2': 0.1236918100723621, 'gradnorm/compressor.layer3': 0.12037560337674012, 'gradnorm/compressor.layer4': 0.10334123911516663, 'gradnorm/compressor.layer5': 0.09887185731166231, 'gradnorm/compressor.layer6': 0.09555448452676352, 'gradnorm/compressor.layer7': 0.0922287258813799, 'gradnorm/decoder.layer0': 0.12417113528181921, 'gradnorm/decoder.layer1': 0.12177383277071702, 'gradnorm/decoder.layer2': 0.1650385376961394, 'gradnorm/max_param': 0.12003450095653534, 'epoch': 1.98}

 20%|█▉        | 3780/19090 [1:50:47<7:15:49,  1.71s/it]
 20%|█▉        | 3781/19090 [1:50:49<8:15:03,  1.94s/it]
 20%|█▉        | 3782/19090 [1:50:51<7:24:46,  1.74s/it]
 20%|█▉        | 3783/19090 [1:50:52<7:01:23,  1.65s/it]
 20%|█▉        | 3784/19090 [1:50:55<8:38:56,  2.03s/it]
 20%|█▉        | 3785/19090 [1:50:56<7:41:30,  1.81s/it]
 20%|█▉        | 3786/19090 [1:50:58<7:11:24,  1.69s/it]
 20%|█▉        | 3787/19090 [1:51:01<8:35:55,  2.02s/it]
 20%|█▉        | 3788/19090 [1:51:02<7:39:30,  1.80s/it]
 20%|█▉        | 3789/19090 [1:51:03<7:23:22,  1.74s/it]
 20%|█▉        | 3790/19090 [1:51:06<8:33:20,  2.01s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.5658965110778809, 'learning_rate': 4.853732602180452e-05, 'recon_loss': 0.0036514696723315863, 'recon_static': 0.0027025326155126093, 'recon_motion': 0.0009489371368545108, 'recon_norm': 2.1090579092502595, 'recon_norm_std': 0.07917172703891992, 'target_norm': 1.6195283442735673, 'recon_cos': 0.4416904866695404, 'gradnorm/compressor.query': 0.008930856452098072, 'gradnorm/compressor.layer0': 0.1502338124451641, 'gradnorm/compressor.layer1': 0.20478241809038555, 'gradnorm/compressor.layer2': 0.17806662431146864, 'gradnorm/compressor.layer3': 0.16864152287159634, 'gradnorm/compressor.layer4': 0.14496548605679938, 'gradnorm/compressor.layer5': 0.14149652070858115, 'gradnorm/compressor.layer6': 0.1326186504479956, 'gradnorm/compressor.layer7': 0.13565516255120152, 'gradnorm/decoder.layer0': 0.1780067076155908, 'gradnorm/decoder.layer1': 0.17593538096694966, 'gradnorm/decoder.layer2': 0.23630371440576486, 'gradnorm/max_param': 0.17006005346775055, 'epoch': 1.99}

 20%|█▉        | 3790/19090 [1:51:06<8:33:20,  2.01s/it]
 20%|█▉        | 3791/19090 [1:51:07<7:37:55,  1.80s/it]
 20%|█▉        | 3792/19090 [1:51:09<6:59:12,  1.64s/it]
 20%|█▉        | 3793/19090 [1:51:11<8:22:38,  1.97s/it]
 20%|█▉        | 3794/19090 [1:51:13<7:30:16,  1.77s/it]
 20%|█▉        | 3795/19090 [1:51:14<7:18:30,  1.72s/it]
 20%|█▉        | 3796/19090 [1:51:17<8:38:13,  2.03s/it]
 20%|█▉        | 3797/19090 [1:51:18<7:40:44,  1.81s/it]
 20%|█▉        | 3798/19090 [1:51:20<7:49:26,  1.84s/it]
 20%|█▉        | 3799/19090 [1:51:23<8:29:17,  2.00s/it]
 20%|█▉        | 3800/19090 [1:51:24<7:35:02,  1.79s/it]
                                                        
{'loss': 0.0058, 'grad_norm': 0.6682159304618835, 'learning_rate': 4.852187985757962e-05, 'recon_loss': 0.005838206456974149, 'recon_static': 0.004879134800285101, 'recon_motion': 0.0009590716785169206, 'recon_norm': 2.640266627073288, 'recon_norm_std': 0.0655246770940721, 'target_norm': 1.6210107564926148, 'recon_cos': 0.3518043585121632, 'gradnorm/compressor.query': 0.010238500358104679, 'gradnorm/compressor.layer0': 0.17826739233122074, 'gradnorm/compressor.layer1': 0.2454308406615135, 'gradnorm/compressor.layer2': 0.2100022829732097, 'gradnorm/compressor.layer3': 0.19993831607166726, 'gradnorm/compressor.layer4': 0.17477190786569388, 'gradnorm/compressor.layer5': 0.16424785656415894, 'gradnorm/compressor.layer6': 0.15659258360229453, 'gradnorm/compressor.layer7': 0.1579978812436297, 'gradnorm/decoder.layer0': 0.20651547043731, 'gradnorm/decoder.layer1': 0.2037581241924575, 'gradnorm/decoder.layer2': 0.28132309228770114, 'gradnorm/max_param': 0.2041459083557129, 'epoch': 1.99}

 20%|█▉        | 3800/19090 [1:51:24<7:35:02,  1.79s/it]
 20%|█▉        | 3801/19090 [1:51:26<7:28:08,  1.76s/it]
 20%|█▉        | 3802/19090 [1:51:28<8:12:57,  1.93s/it]
 20%|█▉        | 3803/19090 [1:51:29<7:22:55,  1.74s/it]
 20%|█▉        | 3804/19090 [1:51:31<7:13:50,  1.70s/it]
 20%|█▉        | 3805/19090 [1:51:34<8:34:59,  2.02s/it]
 20%|█▉        | 3806/19090 [1:51:35<7:38:59,  1.80s/it]
 20%|█▉        | 3807/19090 [1:51:37<7:26:50,  1.75s/it]
 20%|█▉        | 3808/19090 [1:51:40<9:09:36,  2.16s/it]
 20%|█▉        | 3809/19090 [1:51:41<8:02:53,  1.90s/it]
 20%|█▉        | 3810/19090 [1:51:42<7:16:11,  1.71s/it]
                                                        
{'loss': 0.0049, 'grad_norm': 0.5375062823295593, 'learning_rate': 4.850635504769208e-05, 'recon_loss': 0.004915478976909071, 'recon_static': 0.003965333686210215, 'recon_motion': 0.0009501453052507714, 'recon_norm': 2.4273248970508576, 'recon_norm_std': 0.06945476429536938, 'target_norm': 1.618364942073822, 'recon_cos': 0.3811496585607529, 'gradnorm/compressor.query': 0.008122724472481326, 'gradnorm/compressor.layer0': 0.1424297710579782, 'gradnorm/compressor.layer1': 0.19348000125767303, 'gradnorm/compressor.layer2': 0.1674151724267181, 'gradnorm/compressor.layer3': 0.15849695469874509, 'gradnorm/compressor.layer4': 0.13797984924255088, 'gradnorm/compressor.layer5': 0.1300823142262058, 'gradnorm/compressor.layer6': 0.12543695788130912, 'gradnorm/compressor.layer7': 0.12574450240384816, 'gradnorm/decoder.layer0': 0.17049842341070265, 'gradnorm/decoder.layer1': 0.1696553167449921, 'gradnorm/decoder.layer2': 0.22938733241842327, 'gradnorm/max_param': 0.16365943849086761, 'epoch': 2.0}

 20%|█▉        | 3810/19090 [1:51:42<7:16:11,  1.71s/it]
 20%|█▉        | 3811/19090 [1:51:45<8:14:55,  1.94s/it]
 20%|█▉        | 3812/19090 [1:51:46<7:28:12,  1.76s/it]
 20%|█▉        | 3813/19090 [1:51:47<6:51:47,  1.62s/it]
 20%|█▉        | 3814/19090 [1:51:49<6:57:11,  1.64s/it]
 20%|█▉        | 3815/19090 [1:51:50<6:28:52,  1.53s/it]
 20%|█▉        | 3816/19090 [1:51:52<6:09:01,  1.45s/it]
 20%|█▉        | 3817/19090 [1:51:53<5:55:09,  1.40s/it]
 20%|██        | 3818/19090 [1:51:55<6:17:58,  1.48s/it]
 20%|██        | 3819/19090 [1:52:02<14:33:25,  3.43s/it]
 20%|██        | 3820/19090 [1:52:04<11:50:48,  2.79s/it]
                                                         
{'loss': 0.0036, 'grad_norm': 0.3923952877521515, 'learning_rate': 4.849075164404923e-05, 'recon_loss': 0.0036495377542451026, 'recon_static': 0.002682089520385489, 'recon_motion': 0.0009674482877016999, 'recon_norm': 2.101940357685089, 'recon_norm_std': 0.0791882018558681, 'target_norm': 1.6237169295549392, 'recon_cos': 0.44024567380547525, 'gradnorm/compressor.query': 0.006456359344868844, 'gradnorm/compressor.layer0': 0.10657262188420102, 'gradnorm/compressor.layer1': 0.14315160616487602, 'gradnorm/compressor.layer2': 0.12452337547040787, 'gradnorm/compressor.layer3': 0.11569863887025793, 'gradnorm/compressor.layer4': 0.10110887679612636, 'gradnorm/compressor.layer5': 0.09455223855398853, 'gradnorm/compressor.layer6': 0.09216078193260399, 'gradnorm/compressor.layer7': 0.09108140802937088, 'gradnorm/decoder.layer0': 0.12226767797239484, 'gradnorm/decoder.layer1': 0.12166796944712194, 'gradnorm/decoder.layer2': 0.16578597284972796, 'gradnorm/max_param': 0.12013427913188934, 'epoch': 2.0}

 20%|██        | 3820/19090 [1:52:04<11:50:48,  2.79s/it]
 20%|██        | 3821/19090 [1:52:07<12:36:19,  2.97s/it]
 20%|██        | 3822/19090 [1:52:08<10:27:14,  2.46s/it]
 20%|██        | 3823/19090 [1:52:10<8:56:52,  2.11s/it] 
 20%|██        | 3824/19090 [1:52:12<8:31:36,  2.01s/it]
 20%|██        | 3825/19090 [1:52:13<7:35:44,  1.79s/it]
 20%|██        | 3826/19090 [1:52:14<6:56:38,  1.64s/it]
 20%|██        | 3827/19090 [1:52:17<8:16:57,  1.95s/it]
 20%|██        | 3828/19090 [1:52:18<7:25:20,  1.75s/it]
 20%|██        | 3829/19090 [1:52:19<6:49:33,  1.61s/it]
 20%|██        | 3830/19090 [1:52:22<7:54:14,  1.86s/it]
                                                        
{'loss': 0.0028, 'grad_norm': 0.3693449795246124, 'learning_rate': 4.8475069698821244e-05, 'recon_loss': 0.002800943562760949, 'recon_static': 0.0018534871138399466, 'recon_motion': 0.000947456550784409, 'recon_norm': 1.8616142511367797, 'recon_norm_std': 0.08538812082260847, 'target_norm': 1.6175012797117234, 'recon_cos': 0.4959526710212231, 'gradnorm/compressor.query': 0.005902891194733948, 'gradnorm/compressor.layer0': 0.09749354159226938, 'gradnorm/compressor.layer1': 0.13454769315372284, 'gradnorm/compressor.layer2': 0.1160929224888406, 'gradnorm/compressor.layer3': 0.10925708528426002, 'gradnorm/compressor.layer4': 0.09614186536099478, 'gradnorm/compressor.layer5': 0.09101144613242107, 'gradnorm/compressor.layer6': 0.08536281576978011, 'gradnorm/compressor.layer7': 0.08630339799147094, 'gradnorm/decoder.layer0': 0.11546644183989117, 'gradnorm/decoder.layer1': 0.11370805790438457, 'gradnorm/decoder.layer2': 0.1575901908869743, 'gradnorm/max_param': 0.11334579437971115, 'epoch': 2.01}

 20%|██        | 3830/19090 [1:52:22<7:54:14,  1.86s/it]
 20%|██        | 3831/19090 [1:52:23<7:09:36,  1.69s/it]
 20%|██        | 3832/19090 [1:52:24<6:38:35,  1.57s/it]
 20%|██        | 3833/19090 [1:52:27<7:29:09,  1.77s/it]
 20%|██        | 3834/19090 [1:52:28<6:52:10,  1.62s/it]
 20%|██        | 3835/19090 [1:52:29<6:26:05,  1.52s/it]
 20%|██        | 3836/19090 [1:52:31<7:11:01,  1.70s/it]
 20%|██        | 3837/19090 [1:52:33<6:39:19,  1.57s/it]
 20%|██        | 3838/19090 [1:52:34<6:17:03,  1.48s/it]
 20%|██        | 3839/19090 [1:52:36<7:35:46,  1.79s/it]
 20%|██        | 3840/19090 [1:52:38<6:56:45,  1.64s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.5503493547439575, 'learning_rate': 4.845930926444088e-05, 'recon_loss': 0.0039371500199195, 'recon_static': 0.0029829397768480702, 'recon_motion': 0.0009542103798594326, 'recon_norm': 2.178697791695595, 'recon_norm_std': 0.07659035017713904, 'target_norm': 1.6196096628904342, 'recon_cos': 0.4289625704288483, 'gradnorm/compressor.query': 0.008145556934182506, 'gradnorm/compressor.layer0': 0.14726496931582728, 'gradnorm/compressor.layer1': 0.20004995698170389, 'gradnorm/compressor.layer2': 0.17162815705039983, 'gradnorm/compressor.layer3': 0.16375306444176252, 'gradnorm/compressor.layer4': 0.1424506675924161, 'gradnorm/compressor.layer5': 0.13219317320334914, 'gradnorm/compressor.layer6': 0.12714203937095886, 'gradnorm/compressor.layer7': 0.13109846478328233, 'gradnorm/decoder.layer0': 0.17498032041999634, 'gradnorm/decoder.layer1': 0.16922039172960332, 'gradnorm/decoder.layer2': 0.23338169373023934, 'gradnorm/max_param': 0.16796354949474335, 'epoch': 2.01}

 20%|██        | 3840/19090 [1:52:38<6:56:45,  1.64s/it]
 20%|██        | 3841/19090 [1:52:39<6:29:25,  1.53s/it]
 20%|██        | 3842/19090 [1:52:41<6:55:44,  1.64s/it]
 20%|██        | 3843/19090 [1:52:42<6:28:41,  1.53s/it]
 20%|██        | 3844/19090 [1:52:43<6:09:43,  1.46s/it]
 20%|██        | 3845/19090 [1:52:46<7:41:11,  1.82s/it]
 20%|██        | 3846/19090 [1:52:47<7:00:37,  1.66s/it]
 20%|██        | 3847/19090 [1:52:49<6:32:16,  1.54s/it]
 20%|██        | 3848/19090 [1:52:51<7:30:45,  1.77s/it]
 20%|██        | 3849/19090 [1:52:52<6:53:18,  1.63s/it]
 20%|██        | 3850/19090 [1:52:53<6:28:05,  1.53s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.44570502638816833, 'learning_rate': 4.8443470393603304e-05, 'recon_loss': 0.003723707818426192, 'recon_static': 0.002766385965514928, 'recon_motion': 0.000957321806345135, 'recon_norm': 2.1255515426397324, 'recon_norm_std': 0.07745045926421881, 'target_norm': 1.62124442756176, 'recon_cos': 0.43446821719408035, 'gradnorm/compressor.query': 0.006863433384946088, 'gradnorm/compressor.layer0': 0.1167748178431226, 'gradnorm/compressor.layer1': 0.15974345158272568, 'gradnorm/compressor.layer2': 0.140336163910584, 'gradnorm/compressor.layer3': 0.1325492915425961, 'gradnorm/compressor.layer4': 0.1184302515954734, 'gradnorm/compressor.layer5': 0.11033051214051906, 'gradnorm/compressor.layer6': 0.10591582626547853, 'gradnorm/compressor.layer7': 0.10505265260411986, 'gradnorm/decoder.layer0': 0.1393538214967779, 'gradnorm/decoder.layer1': 0.1403648500021228, 'gradnorm/decoder.layer2': 0.18604685028247728, 'gradnorm/max_param': 0.13481271266937256, 'epoch': 2.02}

 20%|██        | 3850/19090 [1:52:53<6:28:05,  1.53s/it]
 20%|██        | 3851/19090 [1:52:56<7:46:33,  1.84s/it]
 20%|██        | 3852/19090 [1:52:57<7:05:04,  1.67s/it]
 20%|██        | 3853/19090 [1:52:59<6:35:38,  1.56s/it]
 20%|██        | 3854/19090 [1:53:01<8:01:44,  1.90s/it]
 20%|██        | 3855/19090 [1:53:03<7:14:57,  1.71s/it]
 20%|██        | 3856/19090 [1:53:04<6:42:34,  1.59s/it]
 20%|██        | 3857/19090 [1:53:06<7:08:02,  1.69s/it]
 20%|██        | 3858/19090 [1:53:07<6:37:27,  1.57s/it]
 20%|██        | 3859/19090 [1:53:08<6:15:49,  1.48s/it]
 20%|██        | 3860/19090 [1:53:10<6:20:00,  1.50s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.42028361558914185, 'learning_rate': 4.842755313926598e-05, 'recon_loss': 0.0032897714467253537, 'recon_static': 0.0023401371377985924, 'recon_motion': 0.0009496342594502494, 'recon_norm': 2.008978095650673, 'recon_norm_std': 0.07913691438734531, 'target_norm': 1.6178395450115204, 'recon_cos': 0.46105843409895897, 'gradnorm/compressor.query': 0.006487882191521164, 'gradnorm/compressor.layer0': 0.11266421115326603, 'gradnorm/compressor.layer1': 0.15266264648272737, 'gradnorm/compressor.layer2': 0.13194529624783918, 'gradnorm/compressor.layer3': 0.12369280018179887, 'gradnorm/compressor.layer4': 0.10880405983545585, 'gradnorm/compressor.layer5': 0.1036135327464655, 'gradnorm/compressor.layer6': 0.09943889071806544, 'gradnorm/compressor.layer7': 0.099761514788135, 'gradnorm/decoder.layer0': 0.1316953333134863, 'gradnorm/decoder.layer1': 0.130306813827247, 'gradnorm/decoder.layer2': 0.17649242331267145, 'gradnorm/max_param': 0.12639489769935608, 'epoch': 2.02}

 20%|██        | 3860/19090 [1:53:10<6:20:00,  1.50s/it]
 20%|██        | 3861/19090 [1:53:11<6:04:30,  1.44s/it]
 20%|██        | 3862/19090 [1:53:12<5:53:05,  1.39s/it]
 20%|██        | 3863/19090 [1:53:14<6:23:39,  1.51s/it]
 20%|██        | 3864/19090 [1:53:16<6:24:10,  1.51s/it]
 20%|██        | 3865/19090 [1:53:17<6:06:30,  1.44s/it]
 20%|██        | 3866/19090 [1:53:20<7:34:30,  1.79s/it]
 20%|██        | 3867/19090 [1:53:21<6:55:33,  1.64s/it]
 20%|██        | 3868/19090 [1:53:22<6:28:20,  1.53s/it]
 20%|██        | 3869/19090 [1:53:25<7:42:03,  1.82s/it]
 20%|██        | 3870/19090 [1:53:26<7:00:59,  1.66s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.5023799538612366, 'learning_rate': 4.84115575546484e-05, 'recon_loss': 0.0036338401841931045, 'recon_static': 0.002687257598154247, 'recon_motion': 0.000946582606411539, 'recon_norm': 2.1098770797252655, 'recon_norm_std': 0.08037325013428927, 'target_norm': 1.6203154772520065, 'recon_cos': 0.44249435514211655, 'gradnorm/compressor.query': 0.008198979425112251, 'gradnorm/compressor.layer0': 0.13883985499907953, 'gradnorm/compressor.layer1': 0.1843846512547602, 'gradnorm/compressor.layer2': 0.15577850729652387, 'gradnorm/compressor.layer3': 0.15010908327801953, 'gradnorm/compressor.layer4': 0.12885417030687052, 'gradnorm/compressor.layer5': 0.12063499429045407, 'gradnorm/compressor.layer6': 0.11834911563209063, 'gradnorm/compressor.layer7': 0.11905188851740121, 'gradnorm/decoder.layer0': 0.15615282613311177, 'gradnorm/decoder.layer1': 0.15555008648785976, 'gradnorm/decoder.layer2': 0.21054431687253616, 'gradnorm/max_param': 0.15170130133628845, 'epoch': 2.03}

 20%|██        | 3870/19090 [1:53:26<7:00:59,  1.66s/it]
 20%|██        | 3871/19090 [1:53:27<6:32:14,  1.55s/it]
 20%|██        | 3872/19090 [1:53:30<7:47:13,  1.84s/it]
 20%|██        | 3873/19090 [1:53:31<7:04:39,  1.67s/it]
 20%|██        | 3874/19090 [1:53:32<6:34:44,  1.56s/it]
 20%|██        | 3875/19090 [1:53:35<7:42:22,  1.82s/it]
 20%|██        | 3876/19090 [1:53:36<7:01:25,  1.66s/it]
 20%|██        | 3877/19090 [1:53:37<6:32:24,  1.55s/it]
 20%|██        | 3878/19090 [1:53:39<6:33:30,  1.55s/it]
 20%|██        | 3879/19090 [1:53:40<6:32:37,  1.55s/it]
 20%|██        | 3880/19090 [1:53:42<6:12:19,  1.47s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.5638012290000916, 'learning_rate': 4.839548369323199e-05, 'recon_loss': 0.004245578451082111, 'recon_static': 0.003293724305694923, 'recon_motion': 0.0009518541424768046, 'recon_norm': 2.262484472990036, 'recon_norm_std': 0.07330343313515186, 'target_norm': 1.6199125856161118, 'recon_cos': 0.4070489428937435, 'gradnorm/compressor.query': 0.008844456436811296, 'gradnorm/compressor.layer0': 0.14933221732057672, 'gradnorm/compressor.layer1': 0.20523836142883517, 'gradnorm/compressor.layer2': 0.17806660339070074, 'gradnorm/compressor.layer3': 0.16527621593948405, 'gradnorm/compressor.layer4': 0.14652712516039207, 'gradnorm/compressor.layer5': 0.13831787863493933, 'gradnorm/compressor.layer6': 0.13234113746955858, 'gradnorm/compressor.layer7': 0.13144230888755337, 'gradnorm/decoder.layer0': 0.1762953181966785, 'gradnorm/decoder.layer1': 0.1717640616273692, 'gradnorm/decoder.layer2': 0.24159996229550007, 'gradnorm/max_param': 0.17085300385951996, 'epoch': 2.03}

 20%|██        | 3880/19090 [1:53:42<6:12:19,  1.47s/it]
 20%|██        | 3881/19090 [1:53:44<6:54:05,  1.63s/it]
 20%|██        | 3882/19090 [1:53:46<7:14:05,  1.71s/it]
 20%|██        | 3883/19090 [1:53:47<6:42:34,  1.59s/it]
 20%|██        | 3884/19090 [1:53:49<7:01:16,  1.66s/it]
 20%|██        | 3885/19090 [1:53:50<6:41:55,  1.59s/it]
 20%|██        | 3886/19090 [1:53:52<6:19:36,  1.50s/it]
 20%|██        | 3887/19090 [1:53:54<7:51:19,  1.86s/it]
 20%|██        | 3888/19090 [1:53:56<7:08:36,  1.69s/it]
 20%|██        | 3889/19090 [1:53:57<6:38:18,  1.57s/it]
 20%|██        | 3890/19090 [1:53:58<6:32:49,  1.55s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.45162466168403625, 'learning_rate': 4.837933160875987e-05, 'recon_loss': 0.004297078913077712, 'recon_static': 0.0033425256726332007, 'recon_motion': 0.000954553195333574, 'recon_norm': 2.2755776941776276, 'recon_norm_std': 0.07217620154842734, 'target_norm': 1.6188996702432632, 'recon_cos': 0.4072936050593853, 'gradnorm/compressor.query': 0.006854199227153264, 'gradnorm/compressor.layer0': 0.11938063881573192, 'gradnorm/compressor.layer1': 0.16419354949562268, 'gradnorm/compressor.layer2': 0.14225546370599415, 'gradnorm/compressor.layer3': 0.1371940185783892, 'gradnorm/compressor.layer4': 0.11610366775075344, 'gradnorm/compressor.layer5': 0.11031298680508755, 'gradnorm/compressor.layer6': 0.10549592180273727, 'gradnorm/compressor.layer7': 0.10825612632680973, 'gradnorm/decoder.layer0': 0.1414286248013814, 'gradnorm/decoder.layer1': 0.14080348211865823, 'gradnorm/decoder.layer2': 0.18794234073647542, 'gradnorm/max_param': 0.1357150673866272, 'epoch': 2.04}

 20%|██        | 3890/19090 [1:53:58<6:32:49,  1.55s/it]
 20%|██        | 3891/19090 [1:54:01<7:51:17,  1.86s/it]
 20%|██        | 3892/19090 [1:54:02<7:08:21,  1.69s/it]
 20%|██        | 3893/19090 [1:54:04<6:56:45,  1.65s/it]
 20%|██        | 3894/19090 [1:54:06<7:31:51,  1.78s/it]
 20%|██        | 3895/19090 [1:54:07<6:54:27,  1.64s/it]
 20%|██        | 3896/19090 [1:54:08<6:32:47,  1.55s/it]
 20%|██        | 3897/19090 [1:54:11<7:50:10,  1.86s/it]
 20%|██        | 3898/19090 [1:54:12<7:07:35,  1.69s/it]
 20%|██        | 3899/19090 [1:54:14<6:37:10,  1.57s/it]
 20%|██        | 3900/19090 [1:54:16<8:06:44,  1.92s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.5217474102973938, 'learning_rate': 4.8363101355236715e-05, 'recon_loss': 0.0037003719655331222, 'recon_static': 0.0027407017769291997, 'recon_motion': 0.0009596701842383481, 'recon_norm': 2.1221142530441286, 'recon_norm_std': 0.0764232425019145, 'target_norm': 1.622223001718521, 'recon_cos': 0.4370508104562759, 'gradnorm/compressor.query': 0.00845805573604019, 'gradnorm/compressor.layer0': 0.14116426226763118, 'gradnorm/compressor.layer1': 0.19387870943417895, 'gradnorm/compressor.layer2': 0.16215256035327516, 'gradnorm/compressor.layer3': 0.15782910745879727, 'gradnorm/compressor.layer4': 0.13481456448482348, 'gradnorm/compressor.layer5': 0.12644644643930938, 'gradnorm/compressor.layer6': 0.12179064567129587, 'gradnorm/compressor.layer7': 0.12291514823738092, 'gradnorm/decoder.layer0': 0.1636926706072042, 'gradnorm/decoder.layer1': 0.16071655954444058, 'gradnorm/decoder.layer2': 0.2160999247426604, 'gradnorm/max_param': 0.1562109887599945, 'epoch': 2.04}

 20%|██        | 3900/19090 [1:54:16<8:06:44,  1.92s/it]
 20%|██        | 3901/19090 [1:54:18<7:19:05,  1.73s/it]
 20%|██        | 3902/19090 [1:54:19<6:45:47,  1.60s/it]
 20%|██        | 3903/19090 [1:54:21<7:34:52,  1.80s/it]
 20%|██        | 3904/19090 [1:54:23<6:56:40,  1.65s/it]
 20%|██        | 3905/19090 [1:54:24<6:29:54,  1.54s/it]
 20%|██        | 3906/19090 [1:54:26<7:31:24,  1.78s/it]
 20%|██        | 3907/19090 [1:54:27<6:54:07,  1.64s/it]
 20%|██        | 3908/19090 [1:54:29<6:28:23,  1.53s/it]
 20%|██        | 3909/19090 [1:54:31<6:47:32,  1.61s/it]
 20%|██        | 3910/19090 [1:54:32<6:23:44,  1.52s/it]
                                                        
{'loss': 0.0048, 'grad_norm': 0.620633065700531, 'learning_rate': 4.834679298692857e-05, 'recon_loss': 0.004781631892547012, 'recon_static': 0.003836748335743323, 'recon_motion': 0.0009448834127397277, 'recon_norm': 2.3975053429603577, 'recon_norm_std': 0.07107865484431386, 'target_norm': 1.6175579220056533, 'recon_cos': 0.38664618209004403, 'gradnorm/compressor.query': 0.010349832583675247, 'gradnorm/compressor.layer0': 0.1671997977127929, 'gradnorm/compressor.layer1': 0.22705567057277656, 'gradnorm/compressor.layer2': 0.19702302537431257, 'gradnorm/compressor.layer3': 0.18505402410899374, 'gradnorm/compressor.layer4': 0.1601265961245869, 'gradnorm/compressor.layer5': 0.15255098430627942, 'gradnorm/compressor.layer6': 0.14668395116224633, 'gradnorm/compressor.layer7': 0.14719765888273695, 'gradnorm/decoder.layer0': 0.1952567020821667, 'gradnorm/decoder.layer1': 0.19038657654100075, 'gradnorm/decoder.layer2': 0.256977216496443, 'gradnorm/max_param': 0.18589264154434204, 'epoch': 2.05}

 20%|██        | 3910/19090 [1:54:32<6:23:44,  1.52s/it]
 20%|██        | 3911/19090 [1:54:33<6:06:54,  1.45s/it]
 20%|██        | 3912/19090 [1:54:36<8:10:21,  1.94s/it]
 20%|██        | 3913/19090 [1:54:38<7:21:52,  1.75s/it]
 21%|██        | 3914/19090 [1:54:39<6:47:43,  1.61s/it]
 21%|██        | 3915/19090 [1:54:42<8:14:16,  1.95s/it]
 21%|██        | 3916/19090 [1:54:43<7:24:24,  1.76s/it]
 21%|██        | 3917/19090 [1:54:44<6:49:24,  1.62s/it]
 21%|██        | 3918/19090 [1:54:46<7:30:58,  1.78s/it]
 21%|██        | 3919/19090 [1:54:48<6:53:57,  1.64s/it]
 21%|██        | 3920/19090 [1:54:49<6:28:42,  1.54s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.4178343415260315, 'learning_rate': 4.833040655836262e-05, 'recon_loss': 0.003913798439316452, 'recon_static': 0.002951557858614251, 'recon_motion': 0.0009622405530535616, 'recon_norm': 2.1720431715250017, 'recon_norm_std': 0.07431648848578334, 'target_norm': 1.621857264637947, 'recon_cos': 0.4273072347044945, 'gradnorm/compressor.query': 0.006282811897335643, 'gradnorm/compressor.layer0': 0.11012351263175199, 'gradnorm/compressor.layer1': 0.15436225423435743, 'gradnorm/compressor.layer2': 0.13378428708206555, 'gradnorm/compressor.layer3': 0.12513825846832682, 'gradnorm/compressor.layer4': 0.10865465161390662, 'gradnorm/compressor.layer5': 0.101619314844976, 'gradnorm/compressor.layer6': 0.09511548262053225, 'gradnorm/compressor.layer7': 0.0978129417284537, 'gradnorm/decoder.layer0': 0.12974821014033405, 'gradnorm/decoder.layer1': 0.12989313041604586, 'gradnorm/decoder.layer2': 0.17487103252770692, 'gradnorm/max_param': 0.12677918374538422, 'epoch': 2.05}

 21%|██        | 3920/19090 [1:54:49<6:28:42,  1.54s/it]
 21%|██        | 3921/19090 [1:54:51<7:46:48,  1.85s/it]
 21%|██        | 3922/19090 [1:54:53<7:04:55,  1.68s/it]
 21%|██        | 3923/19090 [1:54:54<6:35:27,  1.56s/it]
 21%|██        | 3924/19090 [1:54:56<7:33:56,  1.80s/it]
 21%|██        | 3925/19090 [1:54:58<6:55:34,  1.64s/it]
 21%|██        | 3926/19090 [1:54:59<6:36:50,  1.57s/it]
 21%|██        | 3927/19090 [1:55:01<7:35:54,  1.80s/it]
 21%|██        | 3928/19090 [1:55:03<6:57:05,  1.65s/it]
 21%|██        | 3929/19090 [1:55:04<6:30:37,  1.55s/it]
 21%|██        | 3930/19090 [1:55:07<8:06:03,  1.92s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.4450814723968506, 'learning_rate': 4.831394212432708e-05, 'recon_loss': 0.003232582489727065, 'recon_static': 0.002258907840587199, 'recon_motion': 0.0009736747364513576, 'recon_norm': 1.9867353230714797, 'recon_norm_std': 0.07774406932294368, 'target_norm': 1.622943142056465, 'recon_cos': 0.4637800469994545, 'gradnorm/compressor.query': 0.00675774374285005, 'gradnorm/compressor.layer0': 0.11728619154982171, 'gradnorm/compressor.layer1': 0.16561491796109143, 'gradnorm/compressor.layer2': 0.14281800744645476, 'gradnorm/compressor.layer3': 0.13361719327323404, 'gradnorm/compressor.layer4': 0.11397551034730272, 'gradnorm/compressor.layer5': 0.11039661230600852, 'gradnorm/compressor.layer6': 0.10334103183643761, 'gradnorm/compressor.layer7': 0.10283633238518014, 'gradnorm/decoder.layer0': 0.13820332632271187, 'gradnorm/decoder.layer1': 0.13801536159345568, 'gradnorm/decoder.layer2': 0.18448865184025032, 'gradnorm/max_param': 0.13378596305847168, 'epoch': 2.06}

 21%|██        | 3930/19090 [1:55:07<8:06:03,  1.92s/it]
 21%|██        | 3931/19090 [1:55:08<7:18:21,  1.74s/it]
 21%|██        | 3932/19090 [1:55:09<6:44:42,  1.60s/it]
 21%|██        | 3933/19090 [1:55:12<7:20:47,  1.74s/it]
 21%|██        | 3934/19090 [1:55:13<6:46:32,  1.61s/it]
 21%|██        | 3935/19090 [1:55:14<6:22:18,  1.51s/it]
 21%|██        | 3936/19090 [1:55:17<7:47:41,  1.85s/it]
 21%|██        | 3937/19090 [1:55:18<7:05:49,  1.69s/it]
 21%|██        | 3938/19090 [1:55:20<6:53:09,  1.64s/it]
 21%|██        | 3939/19090 [1:55:22<7:33:55,  1.80s/it]
 21%|██        | 3940/19090 [1:55:23<6:59:28,  1.66s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.39639976620674133, 'learning_rate': 4.829739973987096e-05, 'recon_loss': 0.003434841550188139, 'recon_static': 0.002485495910514146, 'recon_motion': 0.0009493455247138627, 'recon_norm': 2.0502500623464583, 'recon_norm_std': 0.07921172678470612, 'target_norm': 1.619759091734886, 'recon_cos': 0.4529564559459686, 'gradnorm/compressor.query': 0.006367608781656911, 'gradnorm/compressor.layer0': 0.10492293977232654, 'gradnorm/compressor.layer1': 0.14279212959761983, 'gradnorm/compressor.layer2': 0.1252543499152212, 'gradnorm/compressor.layer3': 0.11928130661556069, 'gradnorm/compressor.layer4': 0.10314489049980764, 'gradnorm/compressor.layer5': 0.09822417581368521, 'gradnorm/compressor.layer6': 0.09170591128853577, 'gradnorm/compressor.layer7': 0.09434044197337957, 'gradnorm/decoder.layer0': 0.12529314952429962, 'gradnorm/decoder.layer1': 0.12263672454586097, 'gradnorm/decoder.layer2': 0.1656564192390549, 'gradnorm/max_param': 0.11911110579967499, 'epoch': 2.06}

 21%|██        | 3940/19090 [1:55:23<6:59:28,  1.66s/it]
 21%|██        | 3941/19090 [1:55:25<7:18:18,  1.74s/it]
 21%|██        | 3942/19090 [1:55:27<8:02:57,  1.91s/it]
 21%|██        | 3943/19090 [1:55:29<7:16:13,  1.73s/it]
 21%|██        | 3944/19090 [1:55:31<7:30:50,  1.79s/it]
 21%|██        | 3945/19090 [1:55:33<8:04:02,  1.92s/it]
 21%|██        | 3946/19090 [1:55:34<7:17:19,  1.73s/it]
 21%|██        | 3947/19090 [1:55:36<7:25:10,  1.76s/it]
 21%|██        | 3948/19090 [1:55:38<7:34:32,  1.80s/it]
 21%|██        | 3949/19090 [1:55:39<7:07:15,  1.69s/it]
 21%|██        | 3950/19090 [1:55:41<7:16:06,  1.73s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.5218068361282349, 'learning_rate': 4.828077946030392e-05, 'recon_loss': 0.003427691978868097, 'recon_static': 0.002469611508422531, 'recon_motion': 0.0009580804558936507, 'recon_norm': 2.041934224963188, 'recon_norm_std': 0.07869004104286433, 'target_norm': 1.6209992676973344, 'recon_cos': 0.4546757392585278, 'gradnorm/compressor.query': 0.008373909579690229, 'gradnorm/compressor.layer0': 0.14217030321655372, 'gradnorm/compressor.layer1': 0.1933907955341831, 'gradnorm/compressor.layer2': 0.16333220499019557, 'gradnorm/compressor.layer3': 0.1564830472587977, 'gradnorm/compressor.layer4': 0.13747536563195129, 'gradnorm/compressor.layer5': 0.1275675576757269, 'gradnorm/compressor.layer6': 0.11855934957170532, 'gradnorm/compressor.layer7': 0.12372310950742074, 'gradnorm/decoder.layer0': 0.16422341039779376, 'gradnorm/decoder.layer1': 0.16094125863658246, 'gradnorm/decoder.layer2': 0.21453998071418404, 'gradnorm/max_param': 0.15390731394290924, 'epoch': 2.07}

 21%|██        | 3950/19090 [1:55:41<7:16:06,  1.73s/it]
 21%|██        | 3951/19090 [1:55:43<7:16:32,  1.73s/it]
 21%|██        | 3952/19090 [1:55:44<6:43:32,  1.60s/it]
 21%|██        | 3953/19090 [1:55:46<7:28:18,  1.78s/it]
 21%|██        | 3954/19090 [1:55:48<7:46:32,  1.85s/it]
 21%|██        | 3955/19090 [1:55:50<7:05:27,  1.69s/it]
 21%|██        | 3956/19090 [1:55:51<7:11:07,  1.71s/it]
 21%|██        | 3957/19090 [1:55:53<7:32:19,  1.79s/it]
 21%|██        | 3958/19090 [1:55:55<6:54:02,  1.64s/it]
 21%|██        | 3959/19090 [1:55:56<7:00:40,  1.67s/it]
 21%|██        | 3960/19090 [1:55:58<6:39:42,  1.59s/it]
                                                        
{'loss': 0.0055, 'grad_norm': 0.6617431044578552, 'learning_rate': 4.8264081341196024e-05, 'recon_loss': 0.005502020323183388, 'recon_static': 0.004533873224863783, 'recon_motion': 0.0009681471507064998, 'recon_norm': 2.557527059316635, 'recon_norm_std': 0.06492233201861382, 'target_norm': 1.6229607701301574, 'recon_cos': 0.3626712903380394, 'gradnorm/compressor.query': 0.010513733009434003, 'gradnorm/compressor.layer0': 0.17325415306528708, 'gradnorm/compressor.layer1': 0.24035034636920313, 'gradnorm/compressor.layer2': 0.20650461981688065, 'gradnorm/compressor.layer3': 0.20074795064616444, 'gradnorm/compressor.layer4': 0.17264659496428766, 'gradnorm/compressor.layer5': 0.16431751186915267, 'gradnorm/compressor.layer6': 0.15374226426613524, 'gradnorm/compressor.layer7': 0.15397618563748808, 'gradnorm/decoder.layer0': 0.2094253006338808, 'gradnorm/decoder.layer1': 0.20693850689213583, 'gradnorm/decoder.layer2': 0.27609644052503884, 'gradnorm/max_param': 0.19785751402378082, 'epoch': 2.07}

 21%|██        | 3960/19090 [1:55:58<6:39:42,  1.59s/it]
 21%|██        | 3961/19090 [1:55:59<6:17:15,  1.50s/it]
 21%|██        | 3962/19090 [1:56:01<7:16:13,  1.73s/it]
 21%|██        | 3963/19090 [1:56:03<7:15:24,  1.73s/it]
 21%|██        | 3964/19090 [1:56:04<6:41:48,  1.59s/it]
 21%|██        | 3965/19090 [1:56:06<7:11:35,  1.71s/it]
 21%|██        | 3966/19090 [1:56:08<7:38:54,  1.82s/it]
 21%|██        | 3967/19090 [1:56:10<6:58:35,  1.66s/it]
 21%|██        | 3968/19090 [1:56:11<7:00:49,  1.67s/it]
 21%|██        | 3969/19090 [1:56:13<7:21:45,  1.75s/it]
 21%|██        | 3970/19090 [1:56:15<6:46:24,  1.61s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.5016672015190125, 'learning_rate': 4.824730543837762e-05, 'recon_loss': 0.004534653679002076, 'recon_static': 0.0035832260618917644, 'recon_motion': 0.0009514277800917626, 'recon_norm': 2.336172091960907, 'recon_norm_std': 0.07148223342373967, 'target_norm': 1.6188482165336608, 'recon_cos': 0.3977819561958313, 'gradnorm/compressor.query': 0.008124172776521391, 'gradnorm/compressor.layer0': 0.13157448451259934, 'gradnorm/compressor.layer1': 0.18197509775451287, 'gradnorm/compressor.layer2': 0.15786075633234833, 'gradnorm/compressor.layer3': 0.15294296455480919, 'gradnorm/compressor.layer4': 0.13043740547014074, 'gradnorm/compressor.layer5': 0.12387847688945927, 'gradnorm/compressor.layer6': 0.11632153513517753, 'gradnorm/compressor.layer7': 0.11609572220868877, 'gradnorm/decoder.layer0': 0.15857288336071165, 'gradnorm/decoder.layer1': 0.15586810529728418, 'gradnorm/decoder.layer2': 0.20989696503958077, 'gradnorm/max_param': 0.15223655104637146, 'epoch': 2.08}

 21%|██        | 3970/19090 [1:56:15<6:46:24,  1.61s/it]
 21%|██        | 3971/19090 [1:56:16<7:05:58,  1.69s/it]
 21%|██        | 3972/19090 [1:56:19<7:34:46,  1.80s/it]
 21%|██        | 3973/19090 [1:56:20<6:55:34,  1.65s/it]
 21%|██        | 3974/19090 [1:56:22<7:28:35,  1.78s/it]
 21%|██        | 3975/19090 [1:56:24<7:47:33,  1.86s/it]
 21%|██        | 3976/19090 [1:56:25<7:04:31,  1.69s/it]
 21%|██        | 3977/19090 [1:56:27<7:01:00,  1.67s/it]
 21%|██        | 3978/19090 [1:56:29<7:42:52,  1.84s/it]
 21%|██        | 3979/19090 [1:56:30<7:01:00,  1.67s/it]
 21%|██        | 3980/19090 [1:56:32<7:18:40,  1.74s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.44391387701034546, 'learning_rate': 4.8230451807939135e-05, 'recon_loss': 0.003613763308385387, 'recon_static': 0.0026591437635943294, 'recon_motion': 0.0009546195171424188, 'recon_norm': 2.094630327820778, 'recon_norm_std': 0.07721541225910186, 'target_norm': 1.6214557200670243, 'recon_cos': 0.440787735581398, 'gradnorm/compressor.query': 0.0071315540926534925, 'gradnorm/compressor.layer0': 0.11858241449872126, 'gradnorm/compressor.layer1': 0.1627961742540914, 'gradnorm/compressor.layer2': 0.13953942265343142, 'gradnorm/compressor.layer3': 0.13188047716328205, 'gradnorm/compressor.layer4': 0.11647267478576567, 'gradnorm/compressor.layer5': 0.11077251574084837, 'gradnorm/compressor.layer6': 0.10334575858931214, 'gradnorm/compressor.layer7': 0.10376245831754229, 'gradnorm/decoder.layer0': 0.13942581384822111, 'gradnorm/decoder.layer1': 0.1353614187023569, 'gradnorm/decoder.layer2': 0.1858356437268635, 'gradnorm/max_param': 0.13488486409187317, 'epoch': 2.08}

 21%|██        | 3980/19090 [1:56:32<7:18:40,  1.74s/it]
 21%|██        | 3981/19090 [1:56:34<7:07:17,  1.70s/it]
 21%|██        | 3982/19090 [1:56:35<6:35:54,  1.57s/it]
 21%|██        | 3983/19090 [1:56:38<7:52:16,  1.88s/it]
 21%|██        | 3984/19090 [1:56:40<8:00:43,  1.91s/it]
 21%|██        | 3985/19090 [1:56:41<7:13:09,  1.72s/it]
 21%|██        | 3986/19090 [1:56:43<7:10:48,  1.71s/it]
 21%|██        | 3987/19090 [1:56:45<8:07:45,  1.94s/it]
 21%|██        | 3988/19090 [1:56:46<7:18:03,  1.74s/it]
 21%|██        | 3989/19090 [1:56:48<7:03:15,  1.68s/it]
 21%|██        | 3990/19090 [1:56:51<8:30:06,  2.03s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.40172356367111206, 'learning_rate': 4.821352050623086e-05, 'recon_loss': 0.002985109417932108, 'recon_static': 0.002026834944263101, 'recon_motion': 0.0009582744329236448, 'recon_norm': 1.9198538422584535, 'recon_norm_std': 0.08285414539277554, 'target_norm': 1.6238867700099946, 'recon_cos': 0.4837215207517147, 'gradnorm/compressor.query': 0.006351960939402747, 'gradnorm/compressor.layer0': 0.10864789714008194, 'gradnorm/compressor.layer1': 0.14773227661918073, 'gradnorm/compressor.layer2': 0.12510243331519386, 'gradnorm/compressor.layer3': 0.12343441772234208, 'gradnorm/compressor.layer4': 0.10609530730080699, 'gradnorm/compressor.layer5': 0.09833068653493947, 'gradnorm/compressor.layer6': 0.09354761252200935, 'gradnorm/compressor.layer7': 0.09387330549243497, 'gradnorm/decoder.layer0': 0.12457833471511419, 'gradnorm/decoder.layer1': 0.12406572236841598, 'gradnorm/decoder.layer2': 0.1656050429376382, 'gradnorm/max_param': 0.11972637474536896, 'epoch': 2.09}

 21%|██        | 3990/19090 [1:56:51<8:30:06,  2.03s/it]
 21%|██        | 3991/19090 [1:56:52<7:34:01,  1.80s/it]
 21%|██        | 3992/19090 [1:56:53<6:54:28,  1.65s/it]
 21%|██        | 3993/19090 [1:56:56<7:45:37,  1.85s/it]
 21%|██        | 3994/19090 [1:56:57<7:02:37,  1.68s/it]
 21%|██        | 3995/19090 [1:56:59<6:57:27,  1.66s/it]
 21%|██        | 3996/19090 [1:57:01<7:44:06,  1.84s/it]
 21%|██        | 3997/19090 [1:57:02<7:01:44,  1.68s/it]
 21%|██        | 3998/19090 [1:57:03<6:31:56,  1.56s/it]
 21%|██        | 3999/19090 [1:57:06<7:38:31,  1.82s/it]
 21%|██        | 4000/19090 [1:57:07<6:58:34,  1.66s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.5154187083244324, 'learning_rate': 4.819651158986278e-05, 'recon_loss': 0.0034274436940904705, 'recon_static': 0.0024705811956664547, 'recon_motion': 0.0009568625537212938, 'recon_norm': 2.0417934328317644, 'recon_norm_std': 0.07987788431346417, 'target_norm': 1.6208532959222794, 'recon_cos': 0.45395840555429456, 'gradnorm/compressor.query': 0.008198414117584787, 'gradnorm/compressor.layer0': 0.13765117939022198, 'gradnorm/compressor.layer1': 0.18907231470909996, 'gradnorm/compressor.layer2': 0.16361830956941723, 'gradnorm/compressor.layer3': 0.15664704659253365, 'gradnorm/compressor.layer4': 0.13635791436701808, 'gradnorm/compressor.layer5': 0.12700117375693123, 'gradnorm/compressor.layer6': 0.11966006593305747, 'gradnorm/compressor.layer7': 0.11886525380382279, 'gradnorm/decoder.layer0': 0.16372103009872482, 'gradnorm/decoder.layer1': 0.1597945385351154, 'gradnorm/decoder.layer2': 0.20993974268193213, 'gradnorm/max_param': 0.15183189511299133, 'epoch': 2.1}

 21%|██        | 4000/19090 [1:57:07<6:58:34,  1.66s/it]
 21%|██        | 4001/19090 [1:57:11<10:10:32,  2.43s/it]
 21%|██        | 4002/19090 [1:57:13<8:44:47,  2.09s/it] 
 21%|██        | 4003/19090 [1:57:14<7:45:31,  1.85s/it]
 21%|██        | 4004/19090 [1:57:15<7:04:05,  1.69s/it]
 21%|██        | 4005/19090 [1:57:17<6:34:04,  1.57s/it]
 21%|██        | 4006/19090 [1:57:18<6:12:52,  1.48s/it]
 21%|██        | 4007/19090 [1:57:19<5:59:11,  1.43s/it]
 21%|██        | 4008/19090 [1:57:20<5:48:39,  1.39s/it]
 21%|██        | 4009/19090 [1:57:22<5:41:12,  1.36s/it]
 21%|██        | 4010/19090 [1:57:23<5:36:37,  1.34s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.6266501545906067, 'learning_rate': 4.817942511570441e-05, 'recon_loss': 0.004612270969664678, 'recon_static': 0.003650394722353667, 'recon_motion': 0.0009618762647733093, 'recon_norm': 2.351557266712189, 'recon_norm_std': 0.06989401560276746, 'target_norm': 1.6223177552223205, 'recon_cos': 0.3945058241486549, 'gradnorm/compressor.query': 0.009638817972908938, 'gradnorm/compressor.layer0': 0.16018070638987758, 'gradnorm/compressor.layer1': 0.22610443770133662, 'gradnorm/compressor.layer2': 0.19800851702494235, 'gradnorm/compressor.layer3': 0.19289651051376444, 'gradnorm/compressor.layer4': 0.16858103514162534, 'gradnorm/compressor.layer5': 0.15595105899657313, 'gradnorm/compressor.layer6': 0.14752867702768793, 'gradnorm/compressor.layer7': 0.14865844370217943, 'gradnorm/decoder.layer0': 0.1979990158339567, 'gradnorm/decoder.layer1': 0.1932094455090826, 'gradnorm/decoder.layer2': 0.25741430801129356, 'gradnorm/max_param': 0.18584561347961426, 'epoch': 2.1}

 21%|██        | 4010/19090 [1:57:23<5:36:37,  1.34s/it]
 21%|██        | 4011/19090 [1:57:24<5:33:52,  1.33s/it]
 21%|██        | 4012/19090 [1:57:26<5:31:07,  1.32s/it]
 21%|██        | 4013/19090 [1:57:27<5:28:43,  1.31s/it]
 21%|██        | 4014/19090 [1:57:28<5:27:26,  1.30s/it]
 21%|██        | 4015/19090 [1:57:29<5:26:35,  1.30s/it]
 21%|██        | 4016/19090 [1:57:31<5:25:23,  1.30s/it]
 21%|██        | 4017/19090 [1:57:32<5:25:32,  1.30s/it]
 21%|██        | 4018/19090 [1:57:33<5:25:42,  1.30s/it]
 21%|██        | 4019/19090 [1:57:35<5:25:26,  1.30s/it]
 21%|██        | 4020/19090 [1:57:36<5:50:42,  1.40s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.5427350401878357, 'learning_rate': 4.816226114088455e-05, 'recon_loss': 0.004481871030293405, 'recon_static': 0.0035207451437599957, 'recon_motion': 0.0009611259622033686, 'recon_norm': 2.321026176214218, 'recon_norm_std': 0.07297971835359932, 'target_norm': 1.6223999381065368, 'recon_cos': 0.3974997691810131, 'gradnorm/compressor.query': 0.008872169890078496, 'gradnorm/compressor.layer0': 0.14605401080031646, 'gradnorm/compressor.layer1': 0.20246230743021273, 'gradnorm/compressor.layer2': 0.1679266888052807, 'gradnorm/compressor.layer3': 0.16332748365408592, 'gradnorm/compressor.layer4': 0.14330499693539694, 'gradnorm/compressor.layer5': 0.13390138881009764, 'gradnorm/compressor.layer6': 0.12494252418326819, 'gradnorm/compressor.layer7': 0.12609833738545378, 'gradnorm/decoder.layer0': 0.17117298416621016, 'gradnorm/decoder.layer1': 0.16770064225736364, 'gradnorm/decoder.layer2': 0.2233462684284399, 'gradnorm/max_param': 0.16105815768241882, 'epoch': 2.11}

 21%|██        | 4020/19090 [1:57:36<5:50:42,  1.40s/it]
 21%|██        | 4021/19090 [1:57:38<6:07:03,  1.46s/it]
 21%|██        | 4022/19090 [1:57:39<5:58:44,  1.43s/it]
 21%|██        | 4023/19090 [1:57:41<6:10:34,  1.48s/it]
 21%|██        | 4024/19090 [1:57:42<6:19:59,  1.51s/it]
 21%|██        | 4025/19090 [1:57:44<6:14:54,  1.49s/it]
 21%|██        | 4026/19090 [1:57:45<6:17:15,  1.50s/it]
 21%|██        | 4027/19090 [1:57:47<6:45:43,  1.62s/it]
 21%|██        | 4028/19090 [1:57:49<6:24:56,  1.53s/it]
 21%|██        | 4029/19090 [1:57:50<6:16:41,  1.50s/it]
 21%|██        | 4030/19090 [1:57:52<7:03:16,  1.69s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 0.4694712162017822, 'learning_rate': 4.814501972279114e-05, 'recon_loss': 0.004140150599414483, 'recon_static': 0.0031830395164433867, 'recon_motion': 0.000957111016032286, 'recon_norm': 2.234598678350449, 'recon_norm_std': 0.07180787427350879, 'target_norm': 1.6205431699752808, 'recon_cos': 0.41325914338231085, 'gradnorm/compressor.query': 0.007941718274713273, 'gradnorm/compressor.layer0': 0.12627597093203213, 'gradnorm/compressor.layer1': 0.17556238324089146, 'gradnorm/compressor.layer2': 0.1521009564531414, 'gradnorm/compressor.layer3': 0.14216751256828422, 'gradnorm/compressor.layer4': 0.12180187082352312, 'gradnorm/compressor.layer5': 0.11438486990137785, 'gradnorm/compressor.layer6': 0.10818723344990396, 'gradnorm/compressor.layer7': 0.10926481214297605, 'gradnorm/decoder.layer0': 0.14557779792445102, 'gradnorm/decoder.layer1': 0.14460720180154213, 'gradnorm/decoder.layer2': 0.19120424666110403, 'gradnorm/max_param': 0.1392984539270401, 'epoch': 2.11}

 21%|██        | 4030/19090 [1:57:52<7:03:16,  1.69s/it]
 21%|██        | 4031/19090 [1:57:54<6:42:16,  1.60s/it]
 21%|██        | 4032/19090 [1:57:55<6:42:32,  1.60s/it]
 21%|██        | 4033/19090 [1:57:57<7:10:28,  1.72s/it]
 21%|██        | 4034/19090 [1:57:59<7:16:24,  1.74s/it]
 21%|██        | 4035/19090 [1:58:01<7:29:24,  1.79s/it]
 21%|██        | 4036/19090 [1:58:03<7:23:27,  1.77s/it]
 21%|██        | 4037/19090 [1:58:04<7:26:45,  1.78s/it]
 21%|██        | 4038/19090 [1:58:06<7:34:10,  1.81s/it]
 21%|██        | 4039/19090 [1:58:08<7:04:12,  1.69s/it]
 21%|██        | 4040/19090 [1:58:10<7:26:05,  1.78s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.4348958730697632, 'learning_rate': 4.812770091907105e-05, 'recon_loss': 0.0034411858941894025, 'recon_static': 0.0024789717455860226, 'recon_motion': 0.0009622141864383593, 'recon_norm': 2.049693164229393, 'recon_norm_std': 0.0758904641494155, 'target_norm': 1.6225221902132034, 'recon_cos': 0.4517467856407166, 'gradnorm/compressor.query': 0.006782447526921757, 'gradnorm/compressor.layer0': 0.11350640989391785, 'gradnorm/compressor.layer1': 0.16241562307893334, 'gradnorm/compressor.layer2': 0.13966476238306041, 'gradnorm/compressor.layer3': 0.1324490876495444, 'gradnorm/compressor.layer4': 0.11431127503788367, 'gradnorm/compressor.layer5': 0.10789438279117776, 'gradnorm/compressor.layer6': 0.09975928824904068, 'gradnorm/compressor.layer7': 0.10115112382764888, 'gradnorm/decoder.layer0': 0.13542860846561358, 'gradnorm/decoder.layer1': 0.13372288062439228, 'gradnorm/decoder.layer2': 0.177939391054771, 'gradnorm/max_param': 0.1288633644580841, 'epoch': 2.12}

 21%|██        | 4040/19090 [1:58:10<7:26:05,  1.78s/it]
 21%|██        | 4041/19090 [1:58:12<8:04:39,  1.93s/it]
 21%|██        | 4042/19090 [1:58:13<7:16:10,  1.74s/it]
 21%|██        | 4043/19090 [1:58:15<7:19:27,  1.75s/it]
 21%|██        | 4044/19090 [1:58:17<7:31:54,  1.80s/it]
 21%|██        | 4045/19090 [1:58:18<7:09:30,  1.71s/it]
 21%|██        | 4046/19090 [1:58:21<8:00:08,  1.91s/it]
 21%|██        | 4047/19090 [1:58:22<7:12:55,  1.73s/it]
 21%|██        | 4048/19090 [1:58:24<7:17:53,  1.75s/it]
 21%|██        | 4049/19090 [1:58:26<7:45:11,  1.86s/it]
 21%|██        | 4050/19090 [1:58:27<7:05:00,  1.70s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.42977461218833923, 'learning_rate': 4.81103047876299e-05, 'recon_loss': 0.0033033360552508382, 'recon_static': 0.0023566543706692753, 'recon_motion': 0.0009466815972700715, 'recon_norm': 2.0119797080755233, 'recon_norm_std': 0.08016893677413464, 'target_norm': 1.6183305829763412, 'recon_cos': 0.4594812482595444, 'gradnorm/compressor.query': 0.006873374737205381, 'gradnorm/compressor.layer0': 0.11506371468742559, 'gradnorm/compressor.layer1': 0.15886177894196923, 'gradnorm/compressor.layer2': 0.1353889644848955, 'gradnorm/compressor.layer3': 0.1312856208624838, 'gradnorm/compressor.layer4': 0.11432646457696236, 'gradnorm/compressor.layer5': 0.10574377370884026, 'gradnorm/compressor.layer6': 0.09906045345996277, 'gradnorm/compressor.layer7': 0.10009854479139502, 'gradnorm/decoder.layer0': 0.1334454027891035, 'gradnorm/decoder.layer1': 0.1331515962042161, 'gradnorm/decoder.layer2': 0.1760515307356519, 'gradnorm/max_param': 0.1271747648715973, 'epoch': 2.12}

 21%|██        | 4050/19090 [1:58:27<7:05:00,  1.70s/it]
 21%|██        | 4051/19090 [1:58:29<7:20:10,  1.76s/it]
 21%|██        | 4052/19090 [1:58:31<7:52:37,  1.89s/it]
 21%|██        | 4053/19090 [1:58:33<7:07:58,  1.71s/it]
 21%|██        | 4054/19090 [1:58:35<7:20:34,  1.76s/it]
 21%|██        | 4055/19090 [1:58:37<7:57:44,  1.91s/it]
 21%|██        | 4056/19090 [1:58:38<7:11:08,  1.72s/it]
 21%|██▏       | 4057/19090 [1:58:40<7:11:15,  1.72s/it]
 21%|██▏       | 4058/19090 [1:58:42<7:39:31,  1.83s/it]
 21%|██▏       | 4059/19090 [1:58:43<7:08:55,  1.71s/it]
 21%|██▏       | 4060/19090 [1:58:45<7:14:14,  1.73s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.44114166498184204, 'learning_rate': 4.8092831386631834e-05, 'recon_loss': 0.003231924015562981, 'recon_static': 0.002277207450242713, 'recon_motion': 0.0009547166191623546, 'recon_norm': 1.9943739593029022, 'recon_norm_std': 0.07907387483865022, 'target_norm': 1.6196097552776336, 'recon_cos': 0.46617005839943887, 'gradnorm/compressor.query': 0.006694815165602138, 'gradnorm/compressor.layer0': 0.11660096489451252, 'gradnorm/compressor.layer1': 0.16112089980045294, 'gradnorm/compressor.layer2': 0.14101892504436697, 'gradnorm/compressor.layer3': 0.13342099480763592, 'gradnorm/compressor.layer4': 0.11720675866505806, 'gradnorm/compressor.layer5': 0.11105337874117362, 'gradnorm/compressor.layer6': 0.09996325145456637, 'gradnorm/compressor.layer7': 0.10276634821116025, 'gradnorm/decoder.layer0': 0.13865163144923986, 'gradnorm/decoder.layer1': 0.1365954660697525, 'gradnorm/decoder.layer2': 0.18113934657931421, 'gradnorm/max_param': 0.1301344782114029, 'epoch': 2.13}

 21%|██▏       | 4060/19090 [1:58:45<7:14:14,  1.73s/it]
 21%|██▏       | 4061/19090 [1:58:47<7:29:23,  1.79s/it]
 21%|██▏       | 4062/19090 [1:58:49<7:17:03,  1.74s/it]
 21%|██▏       | 4063/19090 [1:58:51<7:18:28,  1.75s/it]
 21%|██▏       | 4064/19090 [1:58:53<7:45:53,  1.86s/it]
 21%|██▏       | 4065/19090 [1:58:54<7:06:16,  1.70s/it]
 21%|██▏       | 4066/19090 [1:58:56<7:31:00,  1.80s/it]
 21%|██▏       | 4067/19090 [1:58:58<7:43:53,  1.85s/it]
 21%|██▏       | 4068/19090 [1:58:59<7:06:17,  1.70s/it]
 21%|██▏       | 4069/19090 [1:59:01<7:34:28,  1.82s/it]
 21%|██▏       | 4070/19090 [1:59:03<7:37:14,  1.83s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.4983077347278595, 'learning_rate': 4.807528077449936e-05, 'recon_loss': 0.0035733938217163085, 'recon_static': 0.002616103965556249, 'recon_motion': 0.0009572898925398476, 'recon_norm': 2.0854872286319734, 'recon_norm_std': 0.07652743291109801, 'target_norm': 1.61892212331295, 'recon_cos': 0.4428731672465801, 'gradnorm/compressor.query': 0.008070293872574952, 'gradnorm/compressor.layer0': 0.13051203991763463, 'gradnorm/compressor.layer1': 0.18198266179033395, 'gradnorm/compressor.layer2': 0.15846788951036092, 'gradnorm/compressor.layer3': 0.1536504083426431, 'gradnorm/compressor.layer4': 0.13429789992690871, 'gradnorm/compressor.layer5': 0.12613455158356715, 'gradnorm/compressor.layer6': 0.11599368135121471, 'gradnorm/compressor.layer7': 0.11732147426325946, 'gradnorm/decoder.layer0': 0.1560446283586094, 'gradnorm/decoder.layer1': 0.15187925048264803, 'gradnorm/decoder.layer2': 0.20197013694786306, 'gradnorm/max_param': 0.14649826288223267, 'epoch': 2.13}

 21%|██▏       | 4070/19090 [1:59:03<7:37:14,  1.83s/it]
 21%|██▏       | 4071/19090 [1:59:05<7:08:07,  1.71s/it]
 21%|██▏       | 4072/19090 [1:59:07<8:00:48,  1.92s/it]
 21%|██▏       | 4073/19090 [1:59:08<7:13:25,  1.73s/it]
 21%|██▏       | 4074/19090 [1:59:10<6:55:09,  1.66s/it]
 21%|██▏       | 4075/19090 [1:59:12<8:04:53,  1.94s/it]
 21%|██▏       | 4076/19090 [1:59:14<7:25:10,  1.78s/it]
 21%|██▏       | 4077/19090 [1:59:15<7:06:17,  1.70s/it]
 21%|██▏       | 4078/19090 [1:59:18<7:50:12,  1.88s/it]
 21%|██▏       | 4079/19090 [1:59:19<7:36:56,  1.83s/it]
 21%|██▏       | 4080/19090 [1:59:21<7:04:20,  1.70s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.5172545313835144, 'learning_rate': 4.805765300991314e-05, 'recon_loss': 0.0043855101394001395, 'recon_static': 0.0034290389274246992, 'recon_motion': 0.00095647118287161, 'recon_norm': 2.300835072994232, 'recon_norm_std': 0.06999341892078519, 'target_norm': 1.6213943004608153, 'recon_cos': 0.40415624529123306, 'gradnorm/compressor.query': 0.008819993382309807, 'gradnorm/compressor.layer0': 0.135833723093519, 'gradnorm/compressor.layer1': 0.19180648181419854, 'gradnorm/compressor.layer2': 0.16554794633418787, 'gradnorm/compressor.layer3': 0.15945621179100197, 'gradnorm/compressor.layer4': 0.13913883915975478, 'gradnorm/compressor.layer5': 0.12792427837788173, 'gradnorm/compressor.layer6': 0.12142461904724591, 'gradnorm/compressor.layer7': 0.12024214300576537, 'gradnorm/decoder.layer0': 0.15933224908740476, 'gradnorm/decoder.layer1': 0.15932581928107772, 'gradnorm/decoder.layer2': 0.20905421328978202, 'gradnorm/max_param': 0.15206019580364227, 'epoch': 2.14}

 21%|██▏       | 4080/19090 [1:59:21<7:04:20,  1.70s/it]
 21%|██▏       | 4081/19090 [1:59:23<8:01:48,  1.93s/it]
 21%|██▏       | 4082/19090 [1:59:25<7:40:03,  1.84s/it]
 21%|██▏       | 4083/19090 [1:59:26<7:02:42,  1.69s/it]
 21%|██▏       | 4084/19090 [1:59:29<7:55:24,  1.90s/it]
 21%|██▏       | 4085/19090 [1:59:30<7:09:21,  1.72s/it]
 21%|██▏       | 4086/19090 [1:59:31<6:55:14,  1.66s/it]
 21%|██▏       | 4087/19090 [1:59:34<7:30:53,  1.80s/it]
 21%|██▏       | 4088/19090 [1:59:35<7:12:04,  1.73s/it]
 21%|██▏       | 4089/19090 [1:59:36<6:42:02,  1.61s/it]
 21%|██▏       | 4090/19090 [1:59:39<7:31:36,  1.81s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 0.5155460834503174, 'learning_rate': 4.80399481518118e-05, 'recon_loss': 0.004047496145358309, 'recon_static': 0.0030950928397942335, 'recon_motion': 0.0009524033375782892, 'recon_norm': 2.210234022140503, 'recon_norm_std': 0.07412108555436134, 'target_norm': 1.619340032339096, 'recon_cos': 0.41548138335347173, 'gradnorm/compressor.query': 0.007944241919866799, 'gradnorm/compressor.layer0': 0.1378607871648999, 'gradnorm/compressor.layer1': 0.18797529101386654, 'gradnorm/compressor.layer2': 0.16388337256577787, 'gradnorm/compressor.layer3': 0.15816623940622, 'gradnorm/compressor.layer4': 0.13607130015498226, 'gradnorm/compressor.layer5': 0.12726161595698718, 'gradnorm/compressor.layer6': 0.11792858952797403, 'gradnorm/compressor.layer7': 0.11946896351359389, 'gradnorm/decoder.layer0': 0.16086103465338952, 'gradnorm/decoder.layer1': 0.16076694914274073, 'gradnorm/decoder.layer2': 0.21188873807709563, 'gradnorm/max_param': 0.15252619981765747, 'epoch': 2.14}

 21%|██▏       | 4090/19090 [1:59:39<7:31:36,  1.81s/it]
 21%|██▏       | 4091/19090 [1:59:41<7:30:53,  1.80s/it]
 21%|██▏       | 4092/19090 [1:59:42<7:07:42,  1.71s/it]
 21%|██▏       | 4093/19090 [1:59:44<7:18:54,  1.76s/it]
 21%|██▏       | 4094/19090 [1:59:46<7:51:48,  1.89s/it]
 21%|██▏       | 4095/19090 [1:59:47<7:06:39,  1.71s/it]
 21%|██▏       | 4096/19090 [1:59:49<7:21:25,  1.77s/it]
 21%|██▏       | 4097/19090 [1:59:51<7:48:34,  1.88s/it]
 21%|██▏       | 4098/19090 [1:59:53<7:04:34,  1.70s/it]
 21%|██▏       | 4099/19090 [1:59:55<7:27:23,  1.79s/it]
 21%|██▏       | 4100/19090 [1:59:57<8:04:40,  1.94s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.4676938056945801, 'learning_rate': 4.802216625939174e-05, 'recon_loss': 0.003909449448110535, 'recon_static': 0.0029494733666069806, 'recon_motion': 0.0009599759694538079, 'recon_norm': 2.180555707216263, 'recon_norm_std': 0.07312352610751986, 'target_norm': 1.6212645798921586, 'recon_cos': 0.4263733811676502, 'gradnorm/compressor.query': 0.007542715012818514, 'gradnorm/compressor.layer0': 0.12265124751173664, 'gradnorm/compressor.layer1': 0.16978108394052271, 'gradnorm/compressor.layer2': 0.1479587488913423, 'gradnorm/compressor.layer3': 0.14449187979726005, 'gradnorm/compressor.layer4': 0.12495304879928809, 'gradnorm/compressor.layer5': 0.11656269563090908, 'gradnorm/compressor.layer6': 0.10778883332180238, 'gradnorm/compressor.layer7': 0.10961293413225472, 'gradnorm/decoder.layer0': 0.14733183179491346, 'gradnorm/decoder.layer1': 0.1455972192014853, 'gradnorm/decoder.layer2': 0.1905339175859221, 'gradnorm/max_param': 0.1378307193517685, 'epoch': 2.15}

 21%|██▏       | 4100/19090 [1:59:57<8:04:40,  1.94s/it]
 21%|██▏       | 4101/19090 [1:59:58<7:16:09,  1.75s/it]
 21%|██▏       | 4102/19090 [2:00:00<7:13:52,  1.74s/it]
 21%|██▏       | 4103/19090 [2:00:02<7:57:45,  1.91s/it]
 21%|██▏       | 4104/19090 [2:00:04<7:10:33,  1.72s/it]
 22%|██▏       | 4105/19090 [2:00:05<7:08:58,  1.72s/it]
 22%|██▏       | 4106/19090 [2:00:08<8:08:16,  1.96s/it]
 22%|██▏       | 4107/19090 [2:00:09<7:17:37,  1.75s/it]
 22%|██▏       | 4108/19090 [2:00:11<7:13:17,  1.74s/it]
 22%|██▏       | 4109/19090 [2:00:13<7:43:08,  1.85s/it]
 22%|██▏       | 4110/19090 [2:00:14<7:00:16,  1.68s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.424162358045578, 'learning_rate': 4.800430739210689e-05, 'recon_loss': 0.0032594576303381475, 'recon_static': 0.0022938706737477333, 'recon_motion': 0.0009655870060669258, 'recon_norm': 1.9955636829137802, 'recon_norm_std': 0.07833158951252699, 'target_norm': 1.6226550787687302, 'recon_cos': 0.46283500641584396, 'gradnorm/compressor.query': 0.006648776038513918, 'gradnorm/compressor.layer0': 0.11316212194541406, 'gradnorm/compressor.layer1': 0.1565778332947276, 'gradnorm/compressor.layer2': 0.13755041980736543, 'gradnorm/compressor.layer3': 0.13120987199246786, 'gradnorm/compressor.layer4': 0.11092525501907555, 'gradnorm/compressor.layer5': 0.10382956008694356, 'gradnorm/compressor.layer6': 0.09730295169772132, 'gradnorm/compressor.layer7': 0.10019565994916314, 'gradnorm/decoder.layer0': 0.13209381314148216, 'gradnorm/decoder.layer1': 0.13022022590368393, 'gradnorm/decoder.layer2': 0.1715021043879824, 'gradnorm/max_param': 0.12405306100845337, 'epoch': 2.15}

 22%|██▏       | 4110/19090 [2:00:14<7:00:16,  1.68s/it]
 22%|██▏       | 4111/19090 [2:00:16<7:07:46,  1.71s/it]
 22%|██▏       | 4112/19090 [2:00:18<7:36:54,  1.83s/it]
 22%|██▏       | 4113/19090 [2:00:19<6:56:03,  1.67s/it]
 22%|██▏       | 4114/19090 [2:00:21<7:18:04,  1.76s/it]
 22%|██▏       | 4115/19090 [2:00:23<7:33:46,  1.82s/it]
 22%|██▏       | 4116/19090 [2:00:25<6:53:16,  1.66s/it]
 22%|██▏       | 4117/19090 [2:00:27<7:14:13,  1.74s/it]
 22%|██▏       | 4118/19090 [2:00:29<7:35:29,  1.83s/it]
 22%|██▏       | 4119/19090 [2:00:30<6:54:33,  1.66s/it]
 22%|██▏       | 4120/19090 [2:00:32<7:35:00,  1.82s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.5104957818984985, 'learning_rate': 4.7986371609668585e-05, 'recon_loss': 0.0038111958710942417, 'recon_static': 0.0028558448189869522, 'recon_motion': 0.0009553510142723098, 'recon_norm': 2.1518293529748918, 'recon_norm_std': 0.07575287465006113, 'target_norm': 1.6213788628578185, 'recon_cos': 0.4314691752195358, 'gradnorm/compressor.query': 0.007849197791482718, 'gradnorm/compressor.layer0': 0.13486010239022328, 'gradnorm/compressor.layer1': 0.18945187637074942, 'gradnorm/compressor.layer2': 0.16282194436919667, 'gradnorm/compressor.layer3': 0.15734803500775876, 'gradnorm/compressor.layer4': 0.137488121364919, 'gradnorm/compressor.layer5': 0.12617797423463153, 'gradnorm/compressor.layer6': 0.11660273006627812, 'gradnorm/compressor.layer7': 0.11922874865593901, 'gradnorm/decoder.layer0': 0.15645172091728077, 'gradnorm/decoder.layer1': 0.15891067614997423, 'gradnorm/decoder.layer2': 0.20695504999471534, 'gradnorm/max_param': 0.14940664172172546, 'epoch': 2.16}

 22%|██▏       | 4120/19090 [2:00:32<7:35:00,  1.82s/it]
 22%|██▏       | 4121/19090 [2:00:34<7:34:55,  1.82s/it]
 22%|██▏       | 4122/19090 [2:00:35<6:54:38,  1.66s/it]
 22%|██▏       | 4123/19090 [2:00:37<7:35:38,  1.83s/it]
 22%|██▏       | 4124/19090 [2:00:39<7:51:29,  1.89s/it]
 22%|██▏       | 4125/19090 [2:00:41<7:05:48,  1.71s/it]
 22%|██▏       | 4126/19090 [2:00:43<7:40:37,  1.85s/it]
 22%|██▏       | 4127/19090 [2:00:45<7:38:49,  1.84s/it]
 22%|██▏       | 4128/19090 [2:00:46<6:57:08,  1.67s/it]
 22%|██▏       | 4129/19090 [2:00:48<7:32:18,  1.81s/it]
 22%|██▏       | 4130/19090 [2:00:50<7:46:51,  1.87s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.4120160937309265, 'learning_rate': 4.7968358972045305e-05, 'recon_loss': 0.0032590872899163513, 'recon_static': 0.002303400385426357, 'recon_motion': 0.0009556869088555686, 'recon_norm': 1.9980157285928726, 'recon_norm_std': 0.07877160292118787, 'target_norm': 1.619603732228279, 'recon_cos': 0.46368150413036346, 'gradnorm/compressor.query': 0.006444272258559975, 'gradnorm/compressor.layer0': 0.10732168814232952, 'gradnorm/compressor.layer1': 0.15248969618781422, 'gradnorm/compressor.layer2': 0.1300325176256985, 'gradnorm/compressor.layer3': 0.126346370627384, 'gradnorm/compressor.layer4': 0.11080737625713534, 'gradnorm/compressor.layer5': 0.10263043828127168, 'gradnorm/compressor.layer6': 0.09609189186780792, 'gradnorm/compressor.layer7': 0.0966735943887511, 'gradnorm/decoder.layer0': 0.12799882840387777, 'gradnorm/decoder.layer1': 0.12734893571791214, 'gradnorm/decoder.layer2': 0.16757285236520159, 'gradnorm/max_param': 0.12173933535814285, 'epoch': 2.16}

 22%|██▏       | 4130/19090 [2:00:50<7:46:51,  1.87s/it]
 22%|██▏       | 4131/19090 [2:00:51<7:04:35,  1.70s/it]
 22%|██▏       | 4132/19090 [2:00:53<7:12:33,  1.74s/it]
 22%|██▏       | 4133/19090 [2:00:56<7:59:07,  1.92s/it]
 22%|██▏       | 4134/19090 [2:00:57<7:11:13,  1.73s/it]
 22%|██▏       | 4135/19090 [2:00:58<7:05:39,  1.71s/it]
 22%|██▏       | 4136/19090 [2:01:01<8:08:11,  1.96s/it]
 22%|██▏       | 4137/19090 [2:01:02<7:17:34,  1.76s/it]
 22%|██▏       | 4138/19090 [2:01:04<7:00:26,  1.69s/it]
 22%|██▏       | 4139/19090 [2:01:06<8:05:34,  1.95s/it]
 22%|██▏       | 4140/19090 [2:01:08<7:15:42,  1.75s/it]
                                                        
{'loss': 0.0028, 'grad_norm': 0.4164798855781555, 'learning_rate': 4.7950269539462486e-05, 'recon_loss': 0.0028339310665614903, 'recon_static': 0.0018820349127054215, 'recon_motion': 0.0009518962207948789, 'recon_norm': 1.8701466500759125, 'recon_norm_std': 0.08529525063931942, 'target_norm': 1.6212386995553971, 'recon_cos': 0.49337624162435534, 'gradnorm/compressor.query': 0.006876771911728365, 'gradnorm/compressor.layer0': 0.10944825937618587, 'gradnorm/compressor.layer1': 0.1539870604286234, 'gradnorm/compressor.layer2': 0.1344674515946291, 'gradnorm/compressor.layer3': 0.12517731726185094, 'gradnorm/compressor.layer4': 0.11013854405297174, 'gradnorm/compressor.layer5': 0.10406138394447494, 'gradnorm/compressor.layer6': 0.09563630067677613, 'gradnorm/compressor.layer7': 0.09730847900703601, 'gradnorm/decoder.layer0': 0.13054404056469432, 'gradnorm/decoder.layer1': 0.12926514112229115, 'gradnorm/decoder.layer2': 0.16923564153814916, 'gradnorm/max_param': 0.12100028991699219, 'epoch': 2.17}

 22%|██▏       | 4140/19090 [2:01:08<7:15:42,  1.75s/it]
 22%|██▏       | 4141/19090 [2:01:09<6:41:21,  1.61s/it]
 22%|██▏       | 4142/19090 [2:01:12<8:15:04,  1.99s/it]
 22%|██▏       | 4143/19090 [2:01:14<8:30:39,  2.05s/it]
 22%|██▏       | 4144/19090 [2:01:15<7:33:12,  1.82s/it]
 22%|██▏       | 4145/19090 [2:01:17<7:33:04,  1.82s/it]
 22%|██▏       | 4146/19090 [2:01:18<6:52:46,  1.66s/it]
 22%|██▏       | 4147/19090 [2:01:20<6:24:27,  1.54s/it]
 22%|██▏       | 4148/19090 [2:01:22<7:51:50,  1.89s/it]
 22%|██▏       | 4149/19090 [2:01:24<7:08:06,  1.72s/it]
 22%|██▏       | 4150/19090 [2:01:25<6:38:06,  1.60s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.5016056895256042, 'learning_rate': 4.793210337240237e-05, 'recon_loss': 0.003641651972429827, 'recon_static': 0.002699609310366213, 'recon_motion': 0.0009420426824362949, 'recon_norm': 2.1121263474226, 'recon_norm_std': 0.077375696785748, 'target_norm': 1.6184318721294404, 'recon_cos': 0.4419575549662113, 'gradnorm/compressor.query': 0.007944869729533515, 'gradnorm/compressor.layer0': 0.13471076373880142, 'gradnorm/compressor.layer1': 0.18584880356230338, 'gradnorm/compressor.layer2': 0.15791632121831742, 'gradnorm/compressor.layer3': 0.15160346509903855, 'gradnorm/compressor.layer4': 0.1338569510636622, 'gradnorm/compressor.layer5': 0.12463399081909707, 'gradnorm/compressor.layer6': 0.11378536605331072, 'gradnorm/compressor.layer7': 0.1191376232331891, 'gradnorm/decoder.layer0': 0.15895550978962242, 'gradnorm/decoder.layer1': 0.1550727603090134, 'gradnorm/decoder.layer2': 0.20255996905201937, 'gradnorm/max_param': 0.14642347395420074, 'epoch': 2.17}

 22%|██▏       | 4150/19090 [2:01:25<6:38:06,  1.60s/it]
 22%|██▏       | 4151/19090 [2:01:27<7:28:16,  1.80s/it]
 22%|██▏       | 4152/19090 [2:01:29<6:52:49,  1.66s/it]
 22%|██▏       | 4153/19090 [2:01:30<6:27:26,  1.56s/it]
 22%|██▏       | 4154/19090 [2:01:32<7:40:56,  1.85s/it]
 22%|██▏       | 4155/19090 [2:01:34<7:00:32,  1.69s/it]
 22%|██▏       | 4156/19090 [2:01:35<6:31:28,  1.57s/it]
 22%|██▏       | 4157/19090 [2:01:38<8:02:06,  1.94s/it]
 22%|██▏       | 4158/19090 [2:01:39<7:14:01,  1.74s/it]
 22%|██▏       | 4159/19090 [2:01:41<6:53:02,  1.66s/it]
 22%|██▏       | 4160/19090 [2:01:43<8:14:56,  1.99s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.6171644926071167, 'learning_rate': 4.791386053160372e-05, 'recon_loss': 0.004630846163490787, 'recon_static': 0.0036857143510133027, 'recon_motion': 0.0009451318488572724, 'recon_norm': 2.359282839298248, 'recon_norm_std': 0.07090630857273936, 'target_norm': 1.6167855739593506, 'recon_cos': 0.39291359558701516, 'gradnorm/compressor.query': 0.01036670600247562, 'gradnorm/compressor.layer0': 0.16476872553831529, 'gradnorm/compressor.layer1': 0.22985726395224215, 'gradnorm/compressor.layer2': 0.19600173455499914, 'gradnorm/compressor.layer3': 0.1880791184564783, 'gradnorm/compressor.layer4': 0.16305526285022937, 'gradnorm/compressor.layer5': 0.15214630316435812, 'gradnorm/compressor.layer6': 0.1415024969900865, 'gradnorm/compressor.layer7': 0.14336096089704636, 'gradnorm/decoder.layer0': 0.193915357706272, 'gradnorm/decoder.layer1': 0.18969089296062186, 'gradnorm/decoder.layer2': 0.2512257448511644, 'gradnorm/max_param': 0.18116506934165955, 'epoch': 2.18}

 22%|██▏       | 4160/19090 [2:01:43<8:14:56,  1.99s/it]
 22%|██▏       | 4161/19090 [2:01:45<7:23:49,  1.78s/it]
 22%|██▏       | 4162/19090 [2:01:46<6:47:32,  1.64s/it]
 22%|██▏       | 4163/19090 [2:01:49<8:08:45,  1.96s/it]
 22%|██▏       | 4164/19090 [2:01:50<7:19:16,  1.77s/it]
 22%|██▏       | 4165/19090 [2:01:51<6:44:47,  1.63s/it]
 22%|██▏       | 4166/19090 [2:01:54<8:15:03,  1.99s/it]
 22%|██▏       | 4167/19090 [2:01:55<7:23:05,  1.78s/it]
 22%|██▏       | 4168/19090 [2:01:57<6:47:00,  1.64s/it]
 22%|██▏       | 4169/19090 [2:02:00<8:20:14,  2.01s/it]
 22%|██▏       | 4170/19090 [2:02:01<7:27:46,  1.80s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.5530532002449036, 'learning_rate': 4.789554107806167e-05, 'recon_loss': 0.00456407917663455, 'recon_static': 0.0036148991028312595, 'recon_motion': 0.0009491800316027366, 'recon_norm': 2.348732280731201, 'recon_norm_std': 0.07070908425375819, 'target_norm': 1.6178186595439912, 'recon_cos': 0.39569179639220237, 'gradnorm/compressor.query': 0.008663668381335619, 'gradnorm/compressor.layer0': 0.1506432170753721, 'gradnorm/compressor.layer1': 0.2039028642453712, 'gradnorm/compressor.layer2': 0.17605925404898223, 'gradnorm/compressor.layer3': 0.1706553719722735, 'gradnorm/compressor.layer4': 0.1460684593769891, 'gradnorm/compressor.layer5': 0.1362808298525668, 'gradnorm/compressor.layer6': 0.1288655317904931, 'gradnorm/compressor.layer7': 0.12680267075121815, 'gradnorm/decoder.layer0': 0.1722207871729999, 'gradnorm/decoder.layer1': 0.17056507293973305, 'gradnorm/decoder.layer2': 0.22373375426085826, 'gradnorm/max_param': 0.16197063028812408, 'epoch': 2.18}

 22%|██▏       | 4170/19090 [2:02:01<7:27:46,  1.80s/it]
 22%|██▏       | 4171/19090 [2:02:02<6:54:31,  1.67s/it]
 22%|██▏       | 4172/19090 [2:02:04<6:56:32,  1.68s/it]
 22%|██▏       | 4173/19090 [2:02:05<6:27:44,  1.56s/it]
 22%|██▏       | 4174/19090 [2:02:07<6:07:23,  1.48s/it]
 22%|██▏       | 4175/19090 [2:02:09<6:59:43,  1.69s/it]
 22%|██▏       | 4176/19090 [2:02:10<6:29:35,  1.57s/it]
 22%|██▏       | 4177/19090 [2:02:11<6:09:20,  1.49s/it]
 22%|██▏       | 4178/19090 [2:02:14<7:13:53,  1.75s/it]
 22%|██▏       | 4179/19090 [2:02:15<6:40:06,  1.61s/it]
 22%|██▏       | 4180/19090 [2:02:16<6:16:03,  1.51s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.5077250599861145, 'learning_rate': 4.787714507302753e-05, 'recon_loss': 0.004202852019807324, 'recon_static': 0.003256335807964206, 'recon_motion': 0.0009465163719141855, 'recon_norm': 2.256769818067551, 'recon_norm_std': 0.07253349963575602, 'target_norm': 1.6187669217586518, 'recon_cos': 0.41108075305819514, 'gradnorm/compressor.query': 0.008451287670920264, 'gradnorm/compressor.layer0': 0.13922018797503224, 'gradnorm/compressor.layer1': 0.1917225307101147, 'gradnorm/compressor.layer2': 0.16358517852050386, 'gradnorm/compressor.layer3': 0.15714252785103994, 'gradnorm/compressor.layer4': 0.13671394067207318, 'gradnorm/compressor.layer5': 0.12511566145696032, 'gradnorm/compressor.layer6': 0.11730685116795084, 'gradnorm/compressor.layer7': 0.11671096152117609, 'gradnorm/decoder.layer0': 0.15434396001211037, 'gradnorm/decoder.layer1': 0.15297163050767354, 'gradnorm/decoder.layer2': 0.20290470097965604, 'gradnorm/max_param': 0.14824719727039337, 'epoch': 2.19}

 22%|██▏       | 4180/19090 [2:02:16<6:16:03,  1.51s/it]
 22%|██▏       | 4181/19090 [2:02:19<7:30:47,  1.81s/it]
 22%|██▏       | 4182/19090 [2:02:20<6:51:02,  1.65s/it]
 22%|██▏       | 4183/19090 [2:02:21<6:23:29,  1.54s/it]
 22%|██▏       | 4184/19090 [2:02:24<7:31:04,  1.82s/it]
 22%|██▏       | 4185/19090 [2:02:25<6:53:07,  1.66s/it]
 22%|██▏       | 4186/19090 [2:02:26<6:25:04,  1.55s/it]
 22%|██▏       | 4187/19090 [2:02:29<7:29:08,  1.81s/it]
 22%|██▏       | 4188/19090 [2:02:30<6:50:05,  1.65s/it]
 22%|██▏       | 4189/19090 [2:02:31<6:22:34,  1.54s/it]
 22%|██▏       | 4190/19090 [2:02:34<7:50:04,  1.89s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.3798721432685852, 'learning_rate': 4.785867257800853e-05, 'recon_loss': 0.0033914732979610563, 'recon_static': 0.0024263577244710175, 'recon_motion': 0.0009651155836763792, 'recon_norm': 2.0323348551988603, 'recon_norm_std': 0.07815626282244921, 'target_norm': 1.6211986601352693, 'recon_cos': 0.45495807826519014, 'gradnorm/compressor.query': 0.006377171592818449, 'gradnorm/compressor.layer0': 0.10340183679766898, 'gradnorm/compressor.layer1': 0.13987026735716263, 'gradnorm/compressor.layer2': 0.12242359918476887, 'gradnorm/compressor.layer3': 0.1152499762173564, 'gradnorm/compressor.layer4': 0.1003836569983976, 'gradnorm/compressor.layer5': 0.0930098232530973, 'gradnorm/compressor.layer6': 0.08529255265260412, 'gradnorm/compressor.layer7': 0.08719758071184558, 'gradnorm/decoder.layer0': 0.11856423155752358, 'gradnorm/decoder.layer1': 0.11959536335188084, 'gradnorm/decoder.layer2': 0.15417516108794108, 'gradnorm/max_param': 0.11153708398342133, 'epoch': 2.19}

 22%|██▏       | 4190/19090 [2:02:34<7:50:04,  1.89s/it]
 22%|██▏       | 4191/19090 [2:02:35<7:05:09,  1.71s/it]
 22%|██▏       | 4192/19090 [2:02:37<6:34:17,  1.59s/it]
 22%|██▏       | 4193/19090 [2:02:40<8:12:04,  1.98s/it]
 22%|██▏       | 4194/19090 [2:02:41<7:20:34,  1.77s/it]
 22%|██▏       | 4195/19090 [2:02:42<6:43:46,  1.63s/it]
 22%|██▏       | 4196/19090 [2:02:46<8:53:08,  2.15s/it]
 22%|██▏       | 4197/19090 [2:02:47<7:48:53,  1.89s/it]
 22%|██▏       | 4198/19090 [2:02:48<7:04:32,  1.71s/it]
 22%|██▏       | 4199/19090 [2:02:51<8:07:11,  1.96s/it]
 22%|██▏       | 4200/19090 [2:02:52<7:17:20,  1.76s/it]
                                                        
{'loss': 0.0028, 'grad_norm': 0.45260751247406006, 'learning_rate': 4.784012365476766e-05, 'recon_loss': 0.0027921491651795804, 'recon_static': 0.0018422504334012046, 'recon_motion': 0.0009498987987171858, 'recon_norm': 1.8604270845651627, 'recon_norm_std': 0.08469996470957994, 'target_norm': 1.619741854071617, 'recon_cos': 0.49933150336146354, 'gradnorm/compressor.query': 0.007132477612842767, 'gradnorm/compressor.layer0': 0.12080794878259665, 'gradnorm/compressor.layer1': 0.17021530947947078, 'gradnorm/compressor.layer2': 0.14517768430597994, 'gradnorm/compressor.layer3': 0.1386910210407872, 'gradnorm/compressor.layer4': 0.1194196817712504, 'gradnorm/compressor.layer5': 0.11294230363928857, 'gradnorm/compressor.layer6': 0.10293823836420211, 'gradnorm/compressor.layer7': 0.10499894179225332, 'gradnorm/decoder.layer0': 0.1418278497435002, 'gradnorm/decoder.layer1': 0.13753649162353423, 'gradnorm/decoder.layer2': 0.1823557445892653, 'gradnorm/max_param': 0.1318998485803604, 'epoch': 2.2}

 22%|██▏       | 4200/19090 [2:02:52<7:17:20,  1.76s/it]
 22%|██▏       | 4201/19090 [2:02:53<6:43:16,  1.63s/it]
 22%|██▏       | 4202/19090 [2:02:56<7:58:16,  1.93s/it]
 22%|██▏       | 4203/19090 [2:02:57<7:10:49,  1.74s/it]
 22%|██▏       | 4204/19090 [2:02:58<6:37:17,  1.60s/it]
 22%|██▏       | 4205/19090 [2:03:01<7:49:17,  1.89s/it]
 22%|██▏       | 4206/19090 [2:03:02<7:04:20,  1.71s/it]
 22%|██▏       | 4207/19090 [2:03:04<6:32:23,  1.58s/it]
 22%|██▏       | 4208/19090 [2:03:06<7:59:17,  1.93s/it]
 22%|██▏       | 4209/19090 [2:03:08<7:10:50,  1.74s/it]
 22%|██▏       | 4210/19090 [2:03:09<6:37:18,  1.60s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.5261203050613403, 'learning_rate': 4.782149836532345e-05, 'recon_loss': 0.0038041210442315786, 'recon_static': 0.002850716415559873, 'recon_motion': 0.0009534046272165142, 'recon_norm': 2.1509526789188387, 'recon_norm_std': 0.07498757019639016, 'target_norm': 1.6192159324884414, 'recon_cos': 0.4309327267110348, 'gradnorm/compressor.query': 0.009233959928331991, 'gradnorm/compressor.layer0': 0.1404879153504228, 'gradnorm/compressor.layer1': 0.20011984170154676, 'gradnorm/compressor.layer2': 0.16861825497981361, 'gradnorm/compressor.layer3': 0.16041344831900226, 'gradnorm/compressor.layer4': 0.14147814930236552, 'gradnorm/compressor.layer5': 0.12930538777566308, 'gradnorm/compressor.layer6': 0.1195205665002701, 'gradnorm/compressor.layer7': 0.12273514330336344, 'gradnorm/decoder.layer0': 0.16262225517717158, 'gradnorm/decoder.layer1': 0.1610362312553987, 'gradnorm/decoder.layer2': 0.21048567206160307, 'gradnorm/max_param': 0.153067484498024, 'epoch': 2.21}

 22%|██▏       | 4210/19090 [2:03:09<6:37:18,  1.60s/it]
 22%|██▏       | 4211/19090 [2:03:12<8:00:10,  1.94s/it]
 22%|██▏       | 4212/19090 [2:03:13<7:11:35,  1.74s/it]
 22%|██▏       | 4213/19090 [2:03:14<6:37:15,  1.60s/it]
 22%|██▏       | 4214/19090 [2:03:17<8:17:05,  2.00s/it]
 22%|██▏       | 4215/19090 [2:03:18<7:23:17,  1.79s/it]
 22%|██▏       | 4216/19090 [2:03:20<6:45:23,  1.64s/it]
 22%|██▏       | 4217/19090 [2:03:22<8:06:06,  1.96s/it]
 22%|██▏       | 4218/19090 [2:03:24<7:22:59,  1.79s/it]
 22%|██▏       | 4219/19090 [2:03:25<6:46:42,  1.64s/it]
 22%|██▏       | 4220/19090 [2:03:28<8:29:56,  2.06s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.5174497961997986, 'learning_rate': 4.780279677194976e-05, 'recon_loss': 0.0035616840177681297, 'recon_static': 0.002612381026847288, 'recon_motion': 0.0009493029880104587, 'recon_norm': 2.0835554540157317, 'recon_norm_std': 0.07666172329336404, 'target_norm': 1.6180168956518173, 'recon_cos': 0.44363224059343337, 'gradnorm/compressor.query': 0.009110015825615938, 'gradnorm/compressor.layer0': 0.14572841767365616, 'gradnorm/compressor.layer1': 0.19865943895270438, 'gradnorm/compressor.layer2': 0.16913102257057333, 'gradnorm/compressor.layer3': 0.15720014172449348, 'gradnorm/compressor.layer4': 0.13421757841875645, 'gradnorm/compressor.layer5': 0.12371797566317294, 'gradnorm/compressor.layer6': 0.1153849208341355, 'gradnorm/compressor.layer7': 0.12101255663558262, 'gradnorm/decoder.layer0': 0.16127046427503622, 'gradnorm/decoder.layer1': 0.15516085711973682, 'gradnorm/decoder.layer2': 0.2056466170371947, 'gradnorm/max_param': 0.14954937994480133, 'epoch': 2.21}

 22%|██▏       | 4220/19090 [2:03:28<8:29:56,  2.06s/it]
 22%|██▏       | 4221/19090 [2:03:29<7:34:59,  1.84s/it]
 22%|██▏       | 4222/19090 [2:03:32<8:02:10,  1.95s/it]
 22%|██▏       | 4223/19090 [2:03:33<7:41:57,  1.86s/it]
 22%|██▏       | 4224/19090 [2:03:35<7:00:26,  1.70s/it]
 22%|██▏       | 4225/19090 [2:03:36<6:30:53,  1.58s/it]
 22%|██▏       | 4226/19090 [2:03:38<7:42:17,  1.87s/it]
 22%|██▏       | 4227/19090 [2:03:40<7:00:40,  1.70s/it]
 22%|██▏       | 4228/19090 [2:03:41<6:30:34,  1.58s/it]
 22%|██▏       | 4229/19090 [2:03:43<7:23:15,  1.79s/it]
 22%|██▏       | 4230/19090 [2:03:45<7:10:32,  1.74s/it]
                                                        
{'loss': 0.004, 'grad_norm': 0.49807122349739075, 'learning_rate': 4.778401893717558e-05, 'recon_loss': 0.003953169856686145, 'recon_static': 0.002994786447379738, 'recon_motion': 0.0009583833962096833, 'recon_norm': 2.1858251452445985, 'recon_norm_std': 0.0739801961928606, 'target_norm': 1.6198613971471787, 'recon_cos': 0.42076961025595666, 'gradnorm/compressor.query': 0.0085658617258631, 'gradnorm/compressor.layer0': 0.13785003194286266, 'gradnorm/compressor.layer1': 0.18655759688344895, 'gradnorm/compressor.layer2': 0.15697456091009698, 'gradnorm/compressor.layer3': 0.15162650015529944, 'gradnorm/compressor.layer4': 0.13416339547698916, 'gradnorm/compressor.layer5': 0.1249752914326874, 'gradnorm/compressor.layer6': 0.11210518407132536, 'gradnorm/compressor.layer7': 0.11524008074066931, 'gradnorm/decoder.layer0': 0.15527848169542652, 'gradnorm/decoder.layer1': 0.1537353401735352, 'gradnorm/decoder.layer2': 0.1983712250879785, 'gradnorm/max_param': 0.14422546327114105, 'epoch': 2.22}

 22%|██▏       | 4230/19090 [2:03:45<7:10:32,  1.74s/it]
 22%|██▏       | 4231/19090 [2:03:46<6:36:36,  1.60s/it]
 22%|██▏       | 4232/19090 [2:03:49<7:40:48,  1.86s/it]
 22%|██▏       | 4233/19090 [2:03:50<7:24:34,  1.80s/it]
 22%|██▏       | 4234/19090 [2:03:52<6:46:36,  1.64s/it]
 22%|██▏       | 4235/19090 [2:03:54<8:09:37,  1.98s/it]
 22%|██▏       | 4236/19090 [2:03:56<7:29:50,  1.82s/it]
 22%|██▏       | 4237/19090 [2:03:57<6:50:19,  1.66s/it]
 22%|██▏       | 4238/19090 [2:04:00<8:01:25,  1.94s/it]
 22%|██▏       | 4239/19090 [2:04:01<7:24:34,  1.80s/it]
 22%|██▏       | 4240/19090 [2:04:03<6:46:18,  1.64s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.5182536840438843, 'learning_rate': 4.776516492378479e-05, 'recon_loss': 0.0037865531223360447, 'recon_static': 0.0028349900094326586, 'recon_motion': 0.0009515631769318134, 'recon_norm': 2.147174006700516, 'recon_norm_std': 0.07582273948937654, 'target_norm': 1.617972159385681, 'recon_cos': 0.4315019600093365, 'gradnorm/compressor.query': 0.008626567394897002, 'gradnorm/compressor.layer0': 0.13793976370177288, 'gradnorm/compressor.layer1': 0.19319035627881687, 'gradnorm/compressor.layer2': 0.16626037619950582, 'gradnorm/compressor.layer3': 0.15649270638747453, 'gradnorm/compressor.layer4': 0.1369312814286863, 'gradnorm/compressor.layer5': 0.1314843044858801, 'gradnorm/compressor.layer6': 0.11982670604981391, 'gradnorm/compressor.layer7': 0.12185318519297894, 'gradnorm/decoder.layer0': 0.16184831763887736, 'gradnorm/decoder.layer1': 0.15815612895548647, 'gradnorm/decoder.layer2': 0.208750879562832, 'gradnorm/max_param': 0.14951233565807343, 'epoch': 2.22}

 22%|██▏       | 4240/19090 [2:04:03<6:46:18,  1.64s/it]
 22%|██▏       | 4241/19090 [2:04:05<7:48:49,  1.89s/it]
 22%|██▏       | 4242/19090 [2:04:07<7:20:26,  1.78s/it]
 22%|██▏       | 4243/19090 [2:04:08<6:43:37,  1.63s/it]
 22%|██▏       | 4244/19090 [2:04:10<7:39:54,  1.86s/it]
 22%|██▏       | 4245/19090 [2:04:12<7:29:54,  1.82s/it]
 22%|██▏       | 4246/19090 [2:04:13<6:50:34,  1.66s/it]
 22%|██▏       | 4247/19090 [2:04:15<7:35:00,  1.84s/it]
 22%|██▏       | 4248/19090 [2:04:17<7:19:07,  1.78s/it]
 22%|██▏       | 4249/19090 [2:04:18<6:43:47,  1.63s/it]
 22%|██▏       | 4250/19090 [2:04:21<7:58:39,  1.94s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.5305755138397217, 'learning_rate': 4.7746234794816e-05, 'recon_loss': 0.004365912918001413, 'recon_static': 0.0034168744808994234, 'recon_motion': 0.0009490383905358613, 'recon_norm': 2.295233976840973, 'recon_norm_std': 0.07478398252278566, 'target_norm': 1.618452101945877, 'recon_cos': 0.4025970324873924, 'gradnorm/compressor.query': 0.008471666684401809, 'gradnorm/compressor.layer0': 0.1399280310106736, 'gradnorm/compressor.layer1': 0.19682706105422468, 'gradnorm/compressor.layer2': 0.1692704506405148, 'gradnorm/compressor.layer3': 0.16246229843960114, 'gradnorm/compressor.layer4': 0.14381050489907668, 'gradnorm/compressor.layer5': 0.1311425589563508, 'gradnorm/compressor.layer6': 0.12132568221203048, 'gradnorm/compressor.layer7': 0.12494505106642738, 'gradnorm/decoder.layer0': 0.16601546231430162, 'gradnorm/decoder.layer1': 0.16517254549618804, 'gradnorm/decoder.layer2': 0.21217260103349403, 'gradnorm/max_param': 0.15349391102790833, 'epoch': 2.23}

 22%|██▏       | 4250/19090 [2:04:21<7:58:39,  1.94s/it]
 22%|██▏       | 4251/19090 [2:04:23<8:30:22,  2.06s/it]
 22%|██▏       | 4252/19090 [2:04:25<7:32:22,  1.83s/it]
 22%|██▏       | 4253/19090 [2:04:26<7:08:20,  1.73s/it]
 22%|██▏       | 4254/19090 [2:04:27<6:34:47,  1.60s/it]
 22%|██▏       | 4255/19090 [2:04:29<6:11:20,  1.50s/it]
 22%|██▏       | 4256/19090 [2:04:31<7:30:03,  1.82s/it]
 22%|██▏       | 4257/19090 [2:04:33<7:09:57,  1.74s/it]
 22%|██▏       | 4258/19090 [2:04:34<6:36:00,  1.60s/it]
 22%|██▏       | 4259/19090 [2:04:36<7:32:50,  1.83s/it]
 22%|██▏       | 4260/19090 [2:04:38<7:22:55,  1.79s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.5316144824028015, 'learning_rate': 4.7727228613562304e-05, 'recon_loss': 0.0035752344294451175, 'recon_static': 0.002627549332100898, 'recon_motion': 0.0009476850944338366, 'recon_norm': 2.09011689722538, 'recon_norm_std': 0.08323159012943507, 'target_norm': 1.619700887799263, 'recon_cos': 0.445305048674345, 'gradnorm/compressor.query': 0.008571730011657484, 'gradnorm/compressor.layer0': 0.13898519928523184, 'gradnorm/compressor.layer1': 0.19625454689334818, 'gradnorm/compressor.layer2': 0.17158219473694777, 'gradnorm/compressor.layer3': 0.16145007948133627, 'gradnorm/compressor.layer4': 0.14381283625456065, 'gradnorm/compressor.layer5': 0.1315888455826102, 'gradnorm/compressor.layer6': 0.12108229767573497, 'gradnorm/compressor.layer7': 0.12498835047439093, 'gradnorm/decoder.layer0': 0.16695299302070482, 'gradnorm/decoder.layer1': 0.16627686086366733, 'gradnorm/decoder.layer2': 0.21307060088605062, 'gradnorm/max_param': 0.15367497503757477, 'epoch': 2.23}

 22%|██▏       | 4260/19090 [2:04:38<7:22:55,  1.79s/it]
 22%|██▏       | 4261/19090 [2:04:39<6:45:46,  1.64s/it]
 22%|██▏       | 4262/19090 [2:04:42<7:21:51,  1.79s/it]
 22%|██▏       | 4263/19090 [2:04:44<7:42:43,  1.87s/it]
 22%|██▏       | 4264/19090 [2:04:45<6:59:14,  1.70s/it]
 22%|██▏       | 4265/19090 [2:04:48<8:23:40,  2.04s/it]
 22%|██▏       | 4266/19090 [2:04:49<7:28:02,  1.81s/it]
 22%|██▏       | 4267/19090 [2:04:50<6:49:18,  1.66s/it]
 22%|██▏       | 4268/19090 [2:04:52<7:17:44,  1.77s/it]
 22%|██▏       | 4269/19090 [2:04:54<7:36:55,  1.85s/it]
 22%|██▏       | 4270/19090 [2:04:56<6:55:20,  1.68s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.5818862915039062, 'learning_rate': 4.770814644357107e-05, 'recon_loss': 0.004258740390650928, 'recon_static': 0.003298657265258953, 'recon_motion': 0.0009600832592695952, 'recon_norm': 2.2672389686107635, 'recon_norm_std': 0.07859073337167502, 'target_norm': 1.619494053721428, 'recon_cos': 0.4076236665248871, 'gradnorm/compressor.query': 0.010294652699641752, 'gradnorm/compressor.layer0': 0.15862636113428283, 'gradnorm/compressor.layer1': 0.22030556579932067, 'gradnorm/compressor.layer2': 0.18871891667358193, 'gradnorm/compressor.layer3': 0.17887540451406594, 'gradnorm/compressor.layer4': 0.15483861896432083, 'gradnorm/compressor.layer5': 0.14407410365972148, 'gradnorm/compressor.layer6': 0.13198608748990404, 'gradnorm/compressor.layer7': 0.13450132243133822, 'gradnorm/decoder.layer0': 0.179557390376657, 'gradnorm/decoder.layer1': 0.1775014000152393, 'gradnorm/decoder.layer2': 0.23062239390867564, 'gradnorm/max_param': 0.1669156551361084, 'epoch': 2.24}

 22%|██▏       | 4270/19090 [2:04:56<6:55:20,  1.68s/it]
 22%|██▏       | 4271/19090 [2:04:58<7:13:24,  1.75s/it]
 22%|██▏       | 4272/19090 [2:05:00<7:44:53,  1.88s/it]
 22%|██▏       | 4273/19090 [2:05:01<7:01:02,  1.70s/it]
 22%|██▏       | 4274/19090 [2:05:03<7:14:58,  1.76s/it]
 22%|██▏       | 4275/19090 [2:05:05<7:37:14,  1.85s/it]
 22%|██▏       | 4276/19090 [2:05:06<6:55:21,  1.68s/it]
 22%|██▏       | 4277/19090 [2:05:08<6:58:21,  1.69s/it]
 22%|██▏       | 4278/19090 [2:05:11<8:14:01,  2.00s/it]
 22%|██▏       | 4279/19090 [2:05:12<7:21:01,  1.79s/it]
 22%|██▏       | 4280/19090 [2:05:13<6:44:10,  1.64s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.5197635293006897, 'learning_rate': 4.7688988348643746e-05, 'recon_loss': 0.004596052749548107, 'recon_static': 0.003628369205398485, 'recon_motion': 0.0009676836503786035, 'recon_norm': 2.34683558344841, 'recon_norm_std': 0.0763161949813366, 'target_norm': 1.6224576324224471, 'recon_cos': 0.39446072950959205, 'gradnorm/compressor.query': 0.008425416331326265, 'gradnorm/compressor.layer0': 0.13560078480545806, 'gradnorm/compressor.layer1': 0.19663400848799742, 'gradnorm/compressor.layer2': 0.1656455852653579, 'gradnorm/compressor.layer3': 0.1585219255833725, 'gradnorm/compressor.layer4': 0.13985181550789977, 'gradnorm/compressor.layer5': 0.1303007028691949, 'gradnorm/compressor.layer6': 0.11682376440409296, 'gradnorm/compressor.layer7': 0.12097711868453899, 'gradnorm/decoder.layer0': 0.15924006177210204, 'gradnorm/decoder.layer1': 0.16228857247308873, 'gradnorm/decoder.layer2': 0.20963764584384817, 'gradnorm/max_param': 0.15198613703250885, 'epoch': 2.24}

 22%|██▏       | 4280/19090 [2:05:13<6:44:10,  1.64s/it]
 22%|██▏       | 4281/19090 [2:05:16<7:58:21,  1.94s/it]
 22%|██▏       | 4282/19090 [2:05:17<7:10:24,  1.74s/it]
 22%|██▏       | 4283/19090 [2:05:19<6:42:15,  1.63s/it]
 22%|██▏       | 4284/19090 [2:05:21<7:57:01,  1.93s/it]
 22%|██▏       | 4285/19090 [2:05:23<7:08:46,  1.74s/it]
 22%|██▏       | 4286/19090 [2:05:24<6:35:03,  1.60s/it]
 22%|██▏       | 4287/19090 [2:05:26<7:44:29,  1.88s/it]
 22%|██▏       | 4288/19090 [2:05:28<6:59:58,  1.70s/it]
 22%|██▏       | 4289/19090 [2:05:29<6:29:04,  1.58s/it]
 22%|██▏       | 4290/19090 [2:05:32<8:15:01,  2.01s/it]
                                                        
{'loss': 0.004, 'grad_norm': 0.5311478972434998, 'learning_rate': 4.766975439283563e-05, 'recon_loss': 0.004022056300891563, 'recon_static': 0.003065653651719913, 'recon_motion': 0.000956402684096247, 'recon_norm': 2.208421128988266, 'recon_norm_std': 0.0815645543858409, 'target_norm': 1.6206117123365402, 'recon_cos': 0.419710373878479, 'gradnorm/compressor.query': 0.008945503316118427, 'gradnorm/compressor.layer0': 0.1415106843492571, 'gradnorm/compressor.layer1': 0.19383024430567422, 'gradnorm/compressor.layer2': 0.16764459269935122, 'gradnorm/compressor.layer3': 0.1619211678504687, 'gradnorm/compressor.layer4': 0.13963681952959772, 'gradnorm/compressor.layer5': 0.13069900250816738, 'gradnorm/compressor.layer6': 0.12035595028761645, 'gradnorm/compressor.layer7': 0.12417648664094914, 'gradnorm/decoder.layer0': 0.1687916624773424, 'gradnorm/decoder.layer1': 0.1675267113997821, 'gradnorm/decoder.layer2': 0.21697379102074504, 'gradnorm/max_param': 0.15294459462165833, 'epoch': 2.25}

 22%|██▏       | 4290/19090 [2:05:32<8:15:01,  2.01s/it]
 22%|██▏       | 4291/19090 [2:05:33<7:21:56,  1.79s/it]
 22%|██▏       | 4292/19090 [2:05:35<6:44:49,  1.64s/it]
 22%|██▏       | 4293/19090 [2:05:37<8:12:32,  2.00s/it]
 22%|██▏       | 4294/19090 [2:05:39<7:19:42,  1.78s/it]
 22%|██▏       | 4295/19090 [2:05:40<6:42:59,  1.63s/it]
 23%|██▎       | 4296/19090 [2:05:43<7:54:43,  1.93s/it]
 23%|██▎       | 4297/19090 [2:05:44<7:06:59,  1.73s/it]
 23%|██▎       | 4298/19090 [2:05:45<6:34:19,  1.60s/it]
 23%|██▎       | 4299/19090 [2:05:48<7:58:39,  1.94s/it]
 23%|██▎       | 4300/19090 [2:05:49<7:10:33,  1.75s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.5493249297142029, 'learning_rate': 4.765044464045567e-05, 'recon_loss': 0.0042142499587498605, 'recon_static': 0.003241206280654296, 'recon_motion': 0.0009730437348480336, 'recon_norm': 2.2536741971969603, 'recon_norm_std': 0.08005319852381945, 'target_norm': 1.6257143557071685, 'recon_cos': 0.41010838001966476, 'gradnorm/compressor.query': 0.00941327044399374, 'gradnorm/compressor.layer0': 0.15116524915935717, 'gradnorm/compressor.layer1': 0.21066170782905566, 'gradnorm/compressor.layer2': 0.1749289300202054, 'gradnorm/compressor.layer3': 0.16834888027256112, 'gradnorm/compressor.layer4': 0.14738551478331322, 'gradnorm/compressor.layer5': 0.13580625380658642, 'gradnorm/compressor.layer6': 0.12140233959002954, 'gradnorm/compressor.layer7': 0.12358852033003903, 'gradnorm/decoder.layer0': 0.16924080888078988, 'gradnorm/decoder.layer1': 0.1700585407639475, 'gradnorm/decoder.layer2': 0.2185353179339183, 'gradnorm/max_param': 0.15769119560718536, 'epoch': 2.25}

 23%|██▎       | 4300/19090 [2:05:49<7:10:33,  1.75s/it]
 23%|██▎       | 4301/19090 [2:05:50<6:36:29,  1.61s/it]
 23%|██▎       | 4302/19090 [2:05:53<7:55:57,  1.93s/it]
 23%|██▎       | 4303/19090 [2:05:54<7:08:27,  1.74s/it]
 23%|██▎       | 4304/19090 [2:05:56<6:34:43,  1.60s/it]
 23%|██▎       | 4305/19090 [2:05:59<8:02:20,  1.96s/it]
 23%|██▎       | 4306/19090 [2:06:00<7:12:47,  1.76s/it]
 23%|██▎       | 4307/19090 [2:06:01<6:38:05,  1.62s/it]
 23%|██▎       | 4308/19090 [2:06:04<8:02:03,  1.96s/it]
 23%|██▎       | 4309/19090 [2:06:05<7:12:41,  1.76s/it]
 23%|██▎       | 4310/19090 [2:06:06<6:38:45,  1.62s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.42458629608154297, 'learning_rate': 4.7631059156066234e-05, 'recon_loss': 0.0036977390584070234, 'recon_static': 0.0027430169051513077, 'recon_motion': 0.0009547220601234585, 'recon_norm': 2.1161538273096085, 'recon_norm_std': 0.08317717090249062, 'target_norm': 1.620947352051735, 'recon_cos': 0.437910021096468, 'gradnorm/compressor.query': 0.007079280097151736, 'gradnorm/compressor.layer0': 0.11063394453774547, 'gradnorm/compressor.layer1': 0.15862061902460212, 'gradnorm/compressor.layer2': 0.13526223773291035, 'gradnorm/compressor.layer3': 0.13284575522835798, 'gradnorm/compressor.layer4': 0.11413683967266677, 'gradnorm/compressor.layer5': 0.10496104304652981, 'gradnorm/compressor.layer6': 0.09701188169050579, 'gradnorm/compressor.layer7': 0.09941981548158774, 'gradnorm/decoder.layer0': 0.13431586666645026, 'gradnorm/decoder.layer1': 0.13114969587458036, 'gradnorm/decoder.layer2': 0.16822173229616183, 'gradnorm/max_param': 0.12216600030660629, 'epoch': 2.26}

 23%|██▎       | 4310/19090 [2:06:06<6:38:45,  1.62s/it]
 23%|██▎       | 4311/19090 [2:06:10<8:26:35,  2.06s/it]
 23%|██▎       | 4312/19090 [2:06:11<7:29:05,  1.82s/it]
 23%|██▎       | 4313/19090 [2:06:12<6:49:20,  1.66s/it]
 23%|██▎       | 4314/19090 [2:06:15<7:53:04,  1.92s/it]
 23%|██▎       | 4315/19090 [2:06:16<7:05:35,  1.73s/it]
 23%|██▎       | 4316/19090 [2:06:17<6:32:20,  1.59s/it]
 23%|██▎       | 4317/19090 [2:06:20<7:38:17,  1.86s/it]
 23%|██▎       | 4318/19090 [2:06:21<6:55:07,  1.69s/it]
 23%|██▎       | 4319/19090 [2:06:22<6:25:00,  1.56s/it]
 23%|██▎       | 4320/19090 [2:06:25<7:59:09,  1.95s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.44597139954566956, 'learning_rate': 4.7611598004482896e-05, 'recon_loss': 0.003058097732719034, 'recon_static': 0.0020956403895979747, 'recon_motion': 0.0009624574740882963, 'recon_norm': 1.936809554696083, 'recon_norm_std': 0.08707646392285824, 'target_norm': 1.6214958488941194, 'recon_cos': 0.4769817814230919, 'gradnorm/compressor.query': 0.007050336912679215, 'gradnorm/compressor.layer0': 0.11487406635423879, 'gradnorm/compressor.layer1': 0.16726490528391938, 'gradnorm/compressor.layer2': 0.14386853146242495, 'gradnorm/compressor.layer3': 0.13954962055212117, 'gradnorm/compressor.layer4': 0.11951933922982708, 'gradnorm/compressor.layer5': 0.10910707679714315, 'gradnorm/compressor.layer6': 0.10052600019234158, 'gradnorm/compressor.layer7': 0.10540011168129947, 'gradnorm/decoder.layer0': 0.13990537974317213, 'gradnorm/decoder.layer1': 0.13767816558561258, 'gradnorm/decoder.layer2': 0.1776576767480842, 'gradnorm/max_param': 0.1278068572282791, 'epoch': 2.26}

 23%|██▎       | 4320/19090 [2:06:25<7:59:09,  1.95s/it]
 23%|██▎       | 4321/19090 [2:06:26<7:09:52,  1.75s/it]
 23%|██▎       | 4322/19090 [2:06:28<6:35:37,  1.61s/it]
 23%|██▎       | 4323/19090 [2:06:30<7:36:50,  1.86s/it]
 23%|██▎       | 4324/19090 [2:06:31<6:54:08,  1.68s/it]
 23%|██▎       | 4325/19090 [2:06:33<6:24:22,  1.56s/it]
 23%|██▎       | 4326/19090 [2:06:35<7:44:07,  1.89s/it]
 23%|██▎       | 4327/19090 [2:06:37<6:59:42,  1.71s/it]
 23%|██▎       | 4328/19090 [2:06:38<6:31:23,  1.59s/it]
 23%|██▎       | 4329/19090 [2:06:41<8:29:07,  2.07s/it]
 23%|██▎       | 4330/19090 [2:06:42<7:31:06,  1.83s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.4837762713432312, 'learning_rate': 4.759206125077422e-05, 'recon_loss': 0.0034447487385477872, 'recon_static': 0.0024999599088914694, 'recon_motion': 0.0009447887903661467, 'recon_norm': 2.055048906803131, 'recon_norm_std': 0.08557573892176151, 'target_norm': 1.6194170832633972, 'recon_cos': 0.4524808466434479, 'gradnorm/compressor.query': 0.008080761106782353, 'gradnorm/compressor.layer0': 0.13001946022796182, 'gradnorm/compressor.layer1': 0.18286186798790155, 'gradnorm/compressor.layer2': 0.15690951634441114, 'gradnorm/compressor.layer3': 0.15199135713686232, 'gradnorm/compressor.layer4': 0.12915673627930296, 'gradnorm/compressor.layer5': 0.11965338398157842, 'gradnorm/compressor.layer6': 0.10747439661052947, 'gradnorm/compressor.layer7': 0.11164050446591234, 'gradnorm/decoder.layer0': 0.1492807816096211, 'gradnorm/decoder.layer1': 0.1475536105755441, 'gradnorm/decoder.layer2': 0.19193524615468435, 'gradnorm/max_param': 0.13745339214801788, 'epoch': 2.27}

 23%|██▎       | 4330/19090 [2:06:42<7:31:06,  1.83s/it]
 23%|██▎       | 4331/19090 [2:06:44<6:50:30,  1.67s/it]
 23%|██▎       | 4332/19090 [2:06:46<8:06:48,  1.98s/it]
 23%|██▎       | 4333/19090 [2:06:48<7:15:21,  1.77s/it]
 23%|██▎       | 4334/19090 [2:06:49<6:39:31,  1.62s/it]
 23%|██▎       | 4335/19090 [2:06:52<7:59:42,  1.95s/it]
 23%|██▎       | 4336/19090 [2:06:53<7:10:26,  1.75s/it]
 23%|██▎       | 4337/19090 [2:06:54<6:36:25,  1.61s/it]
 23%|██▎       | 4338/19090 [2:06:57<8:06:23,  1.98s/it]
 23%|██▎       | 4339/19090 [2:06:58<7:15:26,  1.77s/it]
 23%|██▎       | 4340/19090 [2:07:00<6:40:30,  1.63s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.38789597153663635, 'learning_rate': 4.7572448960261584e-05, 'recon_loss': 0.0031168177083600312, 'recon_static': 0.0021633328433381394, 'recon_motion': 0.0009534847966278903, 'recon_norm': 1.9527222275733949, 'recon_norm_std': 0.08716420140117406, 'target_norm': 1.6202117532491684, 'recon_cos': 0.47367974668741225, 'gradnorm/compressor.query': 0.006922320556465051, 'gradnorm/compressor.layer0': 0.10406348711267317, 'gradnorm/compressor.layer1': 0.14708838864676663, 'gradnorm/compressor.layer2': 0.1260960551881962, 'gradnorm/compressor.layer3': 0.11976200018023855, 'gradnorm/compressor.layer4': 0.10190719103210526, 'gradnorm/compressor.layer5': 0.09590807689537642, 'gradnorm/compressor.layer6': 0.08749343611160681, 'gradnorm/compressor.layer7': 0.09101597413417667, 'gradnorm/decoder.layer0': 0.12167101212216744, 'gradnorm/decoder.layer1': 0.11878169038445566, 'gradnorm/decoder.layer2': 0.15255370100363386, 'gradnorm/max_param': 0.11025059223175049, 'epoch': 2.27}

 23%|██▎       | 4340/19090 [2:07:00<6:40:30,  1.63s/it]
 23%|██▎       | 4341/19090 [2:07:02<8:01:22,  1.96s/it]
 23%|██▎       | 4342/19090 [2:07:04<7:12:03,  1.76s/it]
 23%|██▎       | 4343/19090 [2:07:05<6:37:05,  1.62s/it]
 23%|██▎       | 4344/19090 [2:07:08<8:03:54,  1.97s/it]
 23%|██▎       | 4345/19090 [2:07:09<7:13:47,  1.77s/it]
 23%|██▎       | 4346/19090 [2:07:10<6:38:14,  1.62s/it]
 23%|██▎       | 4347/19090 [2:07:13<8:10:55,  2.00s/it]
 23%|██▎       | 4348/19090 [2:07:14<7:18:38,  1.79s/it]
 23%|██▎       | 4349/19090 [2:07:16<6:41:51,  1.64s/it]
 23%|██▎       | 4350/19090 [2:07:18<8:02:45,  1.97s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.561663031578064, 'learning_rate': 4.755276119851887e-05, 'recon_loss': 0.0034395576978567986, 'recon_static': 0.002484279187046923, 'recon_motion': 0.0009552785035339184, 'recon_norm': 2.046202912926674, 'recon_norm_std': 0.08343325220048428, 'target_norm': 1.621120622754097, 'recon_cos': 0.4541700445115566, 'gradnorm/compressor.query': 0.010207320510527125, 'gradnorm/compressor.layer0': 0.1552130263747655, 'gradnorm/compressor.layer1': 0.21353855751069162, 'gradnorm/compressor.layer2': 0.18389906758208938, 'gradnorm/compressor.layer3': 0.1754961934373079, 'gradnorm/compressor.layer4': 0.14932236943334817, 'gradnorm/compressor.layer5': 0.13663910862215944, 'gradnorm/compressor.layer6': 0.12238806376923961, 'gradnorm/compressor.layer7': 0.12921670174042021, 'gradnorm/decoder.layer0': 0.17460502111310605, 'gradnorm/decoder.layer1': 0.16957827061009198, 'gradnorm/decoder.layer2': 0.2217356657359305, 'gradnorm/max_param': 0.1595822274684906, 'epoch': 2.28}

 23%|██▎       | 4350/19090 [2:07:18<8:02:45,  1.97s/it]
 23%|██▎       | 4351/19090 [2:07:20<7:12:56,  1.76s/it]
 23%|██▎       | 4352/19090 [2:07:21<6:37:41,  1.62s/it]
 23%|██▎       | 4353/19090 [2:07:24<7:51:25,  1.92s/it]
 23%|██▎       | 4354/19090 [2:07:25<7:04:42,  1.73s/it]
 23%|██▎       | 4355/19090 [2:07:26<6:31:45,  1.60s/it]
 23%|██▎       | 4356/19090 [2:07:29<8:00:32,  1.96s/it]
 23%|██▎       | 4357/19090 [2:07:30<7:10:57,  1.76s/it]
 23%|██▎       | 4358/19090 [2:07:32<6:35:59,  1.61s/it]
 23%|██▎       | 4359/19090 [2:07:35<8:26:53,  2.06s/it]
 23%|██▎       | 4360/19090 [2:07:36<7:29:32,  1.83s/it]
                                                        
{'loss': 0.0054, 'grad_norm': 0.6053822040557861, 'learning_rate': 4.7532998031372333e-05, 'recon_loss': 0.005380545405205339, 'recon_static': 0.004421299666864798, 'recon_motion': 0.0009592457325197756, 'recon_norm': 2.53349746465683, 'recon_norm_std': 0.06867376687005163, 'target_norm': 1.6199281573295594, 'recon_cos': 0.3642452351748943, 'gradnorm/compressor.query': 0.010675818610204488, 'gradnorm/compressor.layer0': 0.1619747648581423, 'gradnorm/compressor.layer1': 0.22731565387059788, 'gradnorm/compressor.layer2': 0.1960337482465299, 'gradnorm/compressor.layer3': 0.1859514042920781, 'gradnorm/compressor.layer4': 0.16141899588482683, 'gradnorm/compressor.layer5': 0.1504352119538848, 'gradnorm/compressor.layer6': 0.1351384107029226, 'gradnorm/compressor.layer7': 0.14278262639775552, 'gradnorm/decoder.layer0': 0.18838203115130756, 'gradnorm/decoder.layer1': 0.18688117667742424, 'gradnorm/decoder.layer2': 0.24022855984232344, 'gradnorm/max_param': 0.17449860274791718, 'epoch': 2.28}

 23%|██▎       | 4360/19090 [2:07:36<7:29:32,  1.83s/it]
 23%|██▎       | 4361/19090 [2:07:37<6:49:35,  1.67s/it]
 23%|██▎       | 4362/19090 [2:07:40<8:02:06,  1.96s/it]
 23%|██▎       | 4363/19090 [2:07:41<7:12:07,  1.76s/it]
 23%|██▎       | 4364/19090 [2:07:42<6:36:48,  1.62s/it]
 23%|██▎       | 4365/19090 [2:07:45<8:08:38,  1.99s/it]
 23%|██▎       | 4366/19090 [2:07:47<7:16:26,  1.78s/it]
 23%|██▎       | 4367/19090 [2:07:48<6:41:00,  1.63s/it]
 23%|██▎       | 4368/19090 [2:07:50<7:36:42,  1.86s/it]
 23%|██▎       | 4369/19090 [2:07:52<6:54:11,  1.69s/it]
 23%|██▎       | 4370/19090 [2:07:53<6:24:16,  1.57s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.5240724682807922, 'learning_rate': 4.751315952490034e-05, 'recon_loss': 0.004276841966202483, 'recon_static': 0.003325620183022693, 'recon_motion': 0.0009512218413874507, 'recon_norm': 2.278280657529831, 'recon_norm_std': 0.07514188420027494, 'target_norm': 1.620541763305664, 'recon_cos': 0.40977805107831955, 'gradnorm/compressor.query': 0.008343379114607322, 'gradnorm/compressor.layer0': 0.14002036870364631, 'gradnorm/compressor.layer1': 0.20082466939946186, 'gradnorm/compressor.layer2': 0.17507306068207792, 'gradnorm/compressor.layer3': 0.16432355365740545, 'gradnorm/compressor.layer4': 0.13542254294581402, 'gradnorm/compressor.layer5': 0.12837241637648955, 'gradnorm/compressor.layer6': 0.11481322495917425, 'gradnorm/compressor.layer7': 0.12181231509738275, 'gradnorm/decoder.layer0': 0.161773978123144, 'gradnorm/decoder.layer1': 0.15819677293058038, 'gradnorm/decoder.layer2': 0.2072611288279659, 'gradnorm/max_param': 0.14962619543075562, 'epoch': 2.29}

 23%|██▎       | 4370/19090 [2:07:53<6:24:16,  1.57s/it]
 23%|██▎       | 4371/19090 [2:07:56<7:44:23,  1.89s/it]
 23%|██▎       | 4372/19090 [2:07:57<6:59:30,  1.71s/it]
 23%|██▎       | 4373/19090 [2:07:58<6:27:57,  1.58s/it]
 23%|██▎       | 4374/19090 [2:08:01<7:59:45,  1.96s/it]
 23%|██▎       | 4375/19090 [2:08:02<7:10:15,  1.75s/it]
 23%|██▎       | 4376/19090 [2:08:04<6:35:39,  1.61s/it]
 23%|██▎       | 4377/19090 [2:08:06<7:38:30,  1.87s/it]
 23%|██▎       | 4378/19090 [2:08:07<6:55:43,  1.70s/it]
 23%|██▎       | 4379/19090 [2:08:09<6:25:41,  1.57s/it]
 23%|██▎       | 4380/19090 [2:08:11<8:05:20,  1.98s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.5308313965797424, 'learning_rate': 4.7493245745433145e-05, 'recon_loss': 0.00427393596037291, 'recon_static': 0.003315119043691084, 'recon_motion': 0.0009588169676135294, 'recon_norm': 2.2679755628108977, 'recon_norm_std': 0.07490423377603292, 'target_norm': 1.6209649443626404, 'recon_cos': 0.40623927116394043, 'gradnorm/compressor.query': 0.008596301157320347, 'gradnorm/compressor.layer0': 0.14473990251474603, 'gradnorm/compressor.layer1': 0.20000297797489375, 'gradnorm/compressor.layer2': 0.17480183178151387, 'gradnorm/compressor.layer3': 0.16336195539154139, 'gradnorm/compressor.layer4': 0.14105862413658363, 'gradnorm/compressor.layer5': 0.12840047506834246, 'gradnorm/compressor.layer6': 0.11755118758522722, 'gradnorm/compressor.layer7': 0.12420877365621093, 'gradnorm/decoder.layer0': 0.16272271418994544, 'gradnorm/decoder.layer1': 0.16275198088083548, 'gradnorm/decoder.layer2': 0.21193690550430241, 'gradnorm/max_param': 0.152833953499794, 'epoch': 2.29}

 23%|██▎       | 4380/19090 [2:08:11<8:05:20,  1.98s/it]
 23%|██▎       | 4381/19090 [2:08:13<7:14:54,  1.77s/it]
 23%|██▎       | 4382/19090 [2:08:14<6:40:12,  1.63s/it]
 23%|██▎       | 4383/19090 [2:08:17<7:56:52,  1.95s/it]
 23%|██▎       | 4384/19090 [2:08:18<7:09:37,  1.75s/it]
 23%|██▎       | 4385/19090 [2:08:19<6:35:19,  1.61s/it]
 23%|██▎       | 4386/19090 [2:08:22<8:15:05,  2.02s/it]
 23%|██▎       | 4387/19090 [2:08:24<7:21:18,  1.80s/it]
 23%|██▎       | 4388/19090 [2:08:25<6:43:43,  1.65s/it]
 23%|██▎       | 4389/19090 [2:08:28<7:55:18,  1.94s/it]
 23%|██▎       | 4390/19090 [2:08:29<7:07:26,  1.74s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.38600069284439087, 'learning_rate': 4.747325675955271e-05, 'recon_loss': 0.0032066222047433257, 'recon_static': 0.0022427801130106674, 'recon_motion': 0.0009638421557610855, 'recon_norm': 1.9786695033311843, 'recon_norm_std': 0.08484226018190384, 'target_norm': 1.6249956846237184, 'recon_cos': 0.46853857338428495, 'gradnorm/compressor.query': 0.0062871184545825, 'gradnorm/compressor.layer0': 0.10283986883191763, 'gradnorm/compressor.layer1': 0.14756739480957587, 'gradnorm/compressor.layer2': 0.1270820403293849, 'gradnorm/compressor.layer3': 0.12067106910756537, 'gradnorm/compressor.layer4': 0.1019223422499278, 'gradnorm/compressor.layer5': 0.09384298680147893, 'gradnorm/compressor.layer6': 0.08670131759440536, 'gradnorm/compressor.layer7': 0.0895101896201596, 'gradnorm/decoder.layer0': 0.11796483276507602, 'gradnorm/decoder.layer1': 0.11917617923970578, 'gradnorm/decoder.layer2': 0.1518253836314114, 'gradnorm/max_param': 0.10979782044887543, 'epoch': 2.3}

 23%|██▎       | 4390/19090 [2:08:29<7:07:26,  1.74s/it]
 23%|██▎       | 4391/19090 [2:08:30<6:33:48,  1.61s/it]
 23%|██▎       | 4392/19090 [2:08:33<7:46:02,  1.90s/it]
 23%|██▎       | 4393/19090 [2:08:34<7:01:04,  1.72s/it]
 23%|██▎       | 4394/19090 [2:08:35<6:29:34,  1.59s/it]
 23%|██▎       | 4395/19090 [2:08:38<7:42:37,  1.89s/it]
 23%|██▎       | 4396/19090 [2:08:39<6:59:16,  1.71s/it]
 23%|██▎       | 4397/19090 [2:08:40<6:28:05,  1.58s/it]
 23%|██▎       | 4398/19090 [2:08:43<7:47:43,  1.91s/it]
 23%|██▎       | 4399/19090 [2:08:44<7:02:17,  1.72s/it]
 23%|██▎       | 4400/19090 [2:08:46<6:30:14,  1.59s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.5106917023658752, 'learning_rate': 4.74531926340924e-05, 'recon_loss': 0.0031597491702996193, 'recon_static': 0.0021990274311974646, 'recon_motion': 0.0009607217260054312, 'recon_norm': 1.964313566684723, 'recon_norm_std': 0.0842050276696682, 'target_norm': 1.6230229914188385, 'recon_cos': 0.4713249668478966, 'gradnorm/compressor.query': 0.008697578781255653, 'gradnorm/compressor.layer0': 0.140564269621664, 'gradnorm/compressor.layer1': 0.19690980073857048, 'gradnorm/compressor.layer2': 0.17095122475766927, 'gradnorm/compressor.layer3': 0.16001272166795308, 'gradnorm/compressor.layer4': 0.13683455195094354, 'gradnorm/compressor.layer5': 0.12484021703426217, 'gradnorm/compressor.layer6': 0.11263985996751173, 'gradnorm/compressor.layer7': 0.11587654617377766, 'gradnorm/decoder.layer0': 0.15430094918775242, 'gradnorm/decoder.layer1': 0.15467624262103596, 'gradnorm/decoder.layer2': 0.1980221471453619, 'gradnorm/max_param': 0.1437065452337265, 'epoch': 2.3}

 23%|██▎       | 4400/19090 [2:08:46<6:30:14,  1.59s/it]
 23%|██▎       | 4401/19090 [2:08:49<8:07:32,  1.99s/it]
 23%|██▎       | 4402/19090 [2:08:50<7:16:57,  1.78s/it]
 23%|██▎       | 4403/19090 [2:08:51<6:40:23,  1.64s/it]
 23%|██▎       | 4404/19090 [2:08:53<7:19:23,  1.80s/it]
 23%|██▎       | 4405/19090 [2:08:55<6:45:39,  1.66s/it]
 23%|██▎       | 4406/19090 [2:08:56<6:18:25,  1.55s/it]
 23%|██▎       | 4407/19090 [2:08:59<7:35:47,  1.86s/it]
 23%|██▎       | 4408/19090 [2:09:00<6:57:22,  1.71s/it]
 23%|██▎       | 4409/19090 [2:09:01<6:35:59,  1.62s/it]
 23%|██▎       | 4410/19090 [2:09:04<8:10:45,  2.01s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.5759007334709167, 'learning_rate': 4.743305343613687e-05, 'recon_loss': 0.004324860114138573, 'recon_static': 0.003373183950316161, 'recon_motion': 0.0009516762205748819, 'recon_norm': 2.2872834503650665, 'recon_norm_std': 0.07627211455255747, 'target_norm': 1.6197209119796754, 'recon_cos': 0.40619212612509725, 'gradnorm/compressor.query': 0.009958222459810533, 'gradnorm/compressor.layer0': 0.16023831474709488, 'gradnorm/compressor.layer1': 0.22464852199244414, 'gradnorm/compressor.layer2': 0.1902004231228369, 'gradnorm/compressor.layer3': 0.17881823794463042, 'gradnorm/compressor.layer4': 0.1524734800566337, 'gradnorm/compressor.layer5': 0.14098348224809182, 'gradnorm/compressor.layer6': 0.12675405446787186, 'gradnorm/compressor.layer7': 0.13134740808172782, 'gradnorm/decoder.layer0': 0.17142975639053643, 'gradnorm/decoder.layer1': 0.17353170280905816, 'gradnorm/decoder.layer2': 0.2265445027931304, 'gradnorm/max_param': 0.16361160576343536, 'epoch': 2.31}

 23%|██▎       | 4410/19090 [2:09:04<8:10:45,  2.01s/it]
 23%|██▎       | 4411/19090 [2:09:06<7:22:15,  1.81s/it]
 23%|██▎       | 4412/19090 [2:09:07<6:44:02,  1.65s/it]
 23%|██▎       | 4413/19090 [2:09:10<8:08:25,  2.00s/it]
 23%|██▎       | 4414/19090 [2:09:11<7:21:50,  1.81s/it]
 23%|██▎       | 4415/19090 [2:09:12<6:44:19,  1.65s/it]
 23%|██▎       | 4416/19090 [2:09:15<7:46:29,  1.91s/it]
 23%|██▎       | 4417/19090 [2:09:16<7:00:52,  1.72s/it]
 23%|██▎       | 4418/19090 [2:09:17<6:29:08,  1.59s/it]
 23%|██▎       | 4419/19090 [2:09:20<8:18:26,  2.04s/it]
 23%|██▎       | 4420/19090 [2:09:22<7:23:32,  1.81s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.5669742822647095, 'learning_rate': 4.741283923302171e-05, 'recon_loss': 0.0043764161295257505, 'recon_static': 0.003430318663595244, 'recon_motion': 0.0009460975416004658, 'recon_norm': 2.295413088798523, 'recon_norm_std': 0.07907221000641584, 'target_norm': 1.6174953669309615, 'recon_cos': 0.40099346041679385, 'gradnorm/compressor.query': 0.010512558892556903, 'gradnorm/compressor.layer0': 0.1579796071571059, 'gradnorm/compressor.layer1': 0.21809115988989886, 'gradnorm/compressor.layer2': 0.18909016477722407, 'gradnorm/compressor.layer3': 0.17370220509791923, 'gradnorm/compressor.layer4': 0.148755010031629, 'gradnorm/compressor.layer5': 0.13753622753641923, 'gradnorm/compressor.layer6': 0.12591750216911665, 'gradnorm/compressor.layer7': 0.13282556329322687, 'gradnorm/decoder.layer0': 0.17382838752812482, 'gradnorm/decoder.layer1': 0.17079783612128213, 'gradnorm/decoder.layer2': 0.22129197181883412, 'gradnorm/max_param': 0.160434290766716, 'epoch': 2.32}

 23%|██▎       | 4420/19090 [2:09:22<7:23:32,  1.81s/it]
 23%|██▎       | 4421/19090 [2:09:23<6:44:44,  1.66s/it]
 23%|██▎       | 4422/19090 [2:09:26<8:03:58,  1.98s/it]
 23%|██▎       | 4423/19090 [2:09:27<7:13:07,  1.77s/it]
 23%|██▎       | 4424/19090 [2:09:28<6:37:24,  1.63s/it]
 23%|██▎       | 4425/19090 [2:09:31<8:20:01,  2.05s/it]
 23%|██▎       | 4426/19090 [2:09:33<7:24:12,  1.82s/it]
 23%|██▎       | 4427/19090 [2:09:34<6:45:03,  1.66s/it]
 23%|██▎       | 4428/19090 [2:09:36<7:29:48,  1.84s/it]
 23%|██▎       | 4429/19090 [2:09:38<6:48:48,  1.67s/it]
 23%|██▎       | 4430/19090 [2:09:39<6:20:23,  1.56s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.5351179242134094, 'learning_rate': 4.739255009233337e-05, 'recon_loss': 0.00440639064181596, 'recon_static': 0.0034571941825561224, 'recon_motion': 0.0009491966193309054, 'recon_norm': 2.3098580121994017, 'recon_norm_std': 0.08260033838450909, 'target_norm': 1.62010877430439, 'recon_cos': 0.40343557968735694, 'gradnorm/compressor.query': 0.009543603337040574, 'gradnorm/compressor.layer0': 0.14709507479171371, 'gradnorm/compressor.layer1': 0.2053546857242604, 'gradnorm/compressor.layer2': 0.1764126405411569, 'gradnorm/compressor.layer3': 0.16651337560312215, 'gradnorm/compressor.layer4': 0.14420050718498384, 'gradnorm/compressor.layer5': 0.13072036362704678, 'gradnorm/compressor.layer6': 0.11838381014599714, 'gradnorm/compressor.layer7': 0.12284046223014958, 'gradnorm/decoder.layer0': 0.1643905085510764, 'gradnorm/decoder.layer1': 0.1608660657453699, 'gradnorm/decoder.layer2': 0.20908429974426004, 'gradnorm/max_param': 0.15066389739513397, 'epoch': 2.32}

 23%|██▎       | 4430/19090 [2:09:39<6:20:23,  1.56s/it]
 23%|██▎       | 4431/19090 [2:09:42<8:00:31,  1.97s/it]
 23%|██▎       | 4432/19090 [2:09:43<7:10:22,  1.76s/it]
 23%|██▎       | 4433/19090 [2:09:44<6:35:08,  1.62s/it]
 23%|██▎       | 4434/19090 [2:09:47<7:56:13,  1.95s/it]
 23%|██▎       | 4435/19090 [2:09:48<7:07:24,  1.75s/it]
 23%|██▎       | 4436/19090 [2:09:50<6:33:27,  1.61s/it]
 23%|██▎       | 4437/19090 [2:09:52<7:55:44,  1.95s/it]
 23%|██▎       | 4438/19090 [2:09:54<7:07:29,  1.75s/it]
 23%|██▎       | 4439/19090 [2:09:55<6:33:01,  1.61s/it]
 23%|██▎       | 4440/19090 [2:09:58<8:05:55,  1.99s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.5291378498077393, 'learning_rate': 4.737218608190878e-05, 'recon_loss': 0.003742825030349195, 'recon_static': 0.0027859228139277548, 'recon_motion': 0.0009569022178766318, 'recon_norm': 2.133018708229065, 'recon_norm_std': 0.09056136049330235, 'target_norm': 1.6208148688077926, 'recon_cos': 0.4342330850660801, 'gradnorm/compressor.query': 0.008747279668178417, 'gradnorm/compressor.layer0': 0.1413200671659752, 'gradnorm/compressor.layer1': 0.2038149303992077, 'gradnorm/compressor.layer2': 0.17061579628808687, 'gradnorm/compressor.layer3': 0.1628709278044088, 'gradnorm/compressor.layer4': 0.14034414723547553, 'gradnorm/compressor.layer5': 0.13135663962576175, 'gradnorm/compressor.layer6': 0.11810727370560807, 'gradnorm/compressor.layer7': 0.12441564855358077, 'gradnorm/decoder.layer0': 0.1648810324158203, 'gradnorm/decoder.layer1': 0.16281255776958942, 'gradnorm/decoder.layer2': 0.20661179317965284, 'gradnorm/max_param': 0.14938341081142426, 'epoch': 2.33}

 23%|██▎       | 4440/19090 [2:09:58<8:05:55,  1.99s/it]
 23%|██▎       | 4441/19090 [2:09:59<7:14:28,  1.78s/it]
 23%|██▎       | 4442/19090 [2:10:00<6:38:13,  1.63s/it]
 23%|██▎       | 4443/19090 [2:10:03<7:51:16,  1.93s/it]
 23%|██▎       | 4444/19090 [2:10:04<7:04:39,  1.74s/it]
 23%|██▎       | 4445/19090 [2:10:06<6:31:46,  1.61s/it]
 23%|██▎       | 4446/19090 [2:10:08<7:51:38,  1.93s/it]
 23%|██▎       | 4447/19090 [2:10:10<7:04:36,  1.74s/it]
 23%|██▎       | 4448/19090 [2:10:11<6:31:31,  1.60s/it]
 23%|██▎       | 4449/19090 [2:10:14<8:07:21,  2.00s/it]
 23%|██▎       | 4450/19090 [2:10:15<7:15:50,  1.79s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.5903578996658325, 'learning_rate': 4.7351747269835245e-05, 'recon_loss': 0.004369744652649388, 'recon_static': 0.0034152799926232547, 'recon_motion': 0.0009544647735310718, 'recon_norm': 2.294168251752853, 'recon_norm_std': 0.08703015614300966, 'target_norm': 1.620746573805809, 'recon_cos': 0.4028044119477272, 'gradnorm/compressor.query': 0.009564482591266006, 'gradnorm/compressor.layer0': 0.16066153995714116, 'gradnorm/compressor.layer1': 0.23160569151107338, 'gradnorm/compressor.layer2': 0.1954914894111358, 'gradnorm/compressor.layer3': 0.18065584524901493, 'gradnorm/compressor.layer4': 0.15654512785852212, 'gradnorm/compressor.layer5': 0.14633322228771087, 'gradnorm/compressor.layer6': 0.12972054347795167, 'gradnorm/compressor.layer7': 0.1384413643506445, 'gradnorm/decoder.layer0': 0.1812446749661323, 'gradnorm/decoder.layer1': 0.17856805853558771, 'gradnorm/decoder.layer2': 0.22688068868793004, 'gradnorm/max_param': 0.1642633080482483, 'epoch': 2.33}

 23%|██▎       | 4450/19090 [2:10:15<7:15:50,  1.79s/it]
 23%|██▎       | 4451/19090 [2:10:16<6:39:37,  1.64s/it]
 23%|██▎       | 4452/19090 [2:10:19<7:54:58,  1.95s/it]
 23%|██▎       | 4453/19090 [2:10:20<7:06:56,  1.75s/it]
 23%|██▎       | 4454/19090 [2:10:22<6:33:45,  1.61s/it]
 23%|██▎       | 4455/19090 [2:10:24<8:02:37,  1.98s/it]
 23%|██▎       | 4456/19090 [2:10:26<7:12:27,  1.77s/it]
 23%|██▎       | 4457/19090 [2:10:27<6:37:24,  1.63s/it]
 23%|██▎       | 4458/19090 [2:10:30<7:50:17,  1.93s/it]
 23%|██▎       | 4459/19090 [2:10:31<7:04:07,  1.74s/it]
 23%|██▎       | 4460/19090 [2:10:32<6:32:27,  1.61s/it]
                                                        
{'loss': 0.0047, 'grad_norm': 0.6531025767326355, 'learning_rate': 4.7331233724450165e-05, 'recon_loss': 0.004706572426948696, 'recon_static': 0.0037471575720701366, 'recon_motion': 0.0009594147835741751, 'recon_norm': 2.3778655350208284, 'recon_norm_std': 0.08048127833753824, 'target_norm': 1.618623173236847, 'recon_cos': 0.38930867686867715, 'gradnorm/compressor.query': 0.010473129245349941, 'gradnorm/compressor.layer0': 0.17805504429058255, 'gradnorm/compressor.layer1': 0.25193100469908947, 'gradnorm/compressor.layer2': 0.21233142238759622, 'gradnorm/compressor.layer3': 0.1985582547464929, 'gradnorm/compressor.layer4': 0.17497225673484212, 'gradnorm/compressor.layer5': 0.16284393583663032, 'gradnorm/compressor.layer6': 0.14486121195868074, 'gradnorm/compressor.layer7': 0.1524161753524897, 'gradnorm/decoder.layer0': 0.20295559714959477, 'gradnorm/decoder.layer1': 0.19718509389978148, 'gradnorm/decoder.layer2': 0.2555945824384397, 'gradnorm/max_param': 0.18344978988170624, 'epoch': 2.34}

 23%|██▎       | 4460/19090 [2:10:32<6:32:27,  1.61s/it]
 23%|██▎       | 4461/19090 [2:10:35<7:24:58,  1.83s/it]
 23%|██▎       | 4462/19090 [2:10:36<6:46:08,  1.67s/it]
 23%|██▎       | 4463/19090 [2:10:37<6:18:37,  1.55s/it]
 23%|██▎       | 4464/19090 [2:10:40<7:43:09,  1.90s/it]
 23%|██▎       | 4465/19090 [2:10:41<6:58:07,  1.72s/it]
 23%|██▎       | 4466/19090 [2:10:42<6:26:48,  1.59s/it]
 23%|██▎       | 4467/19090 [2:10:45<7:48:13,  1.92s/it]
 23%|██▎       | 4468/19090 [2:10:46<7:02:08,  1.73s/it]
 23%|██▎       | 4469/19090 [2:10:48<6:31:45,  1.61s/it]
 23%|██▎       | 4470/19090 [2:10:51<8:02:57,  1.98s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.5340203642845154, 'learning_rate': 4.731064551434079e-05, 'recon_loss': 0.004463872348424047, 'recon_static': 0.0035048036312218755, 'recon_motion': 0.0009590687463060021, 'recon_norm': 2.314654511213303, 'recon_norm_std': 0.08559755440801382, 'target_norm': 1.6224582552909852, 'recon_cos': 0.3996139757335186, 'gradnorm/compressor.query': 0.009338528111456386, 'gradnorm/compressor.layer0': 0.14549398036471362, 'gradnorm/compressor.layer1': 0.2082861350159869, 'gradnorm/compressor.layer2': 0.17490017802794847, 'gradnorm/compressor.layer3': 0.1658396461367932, 'gradnorm/compressor.layer4': 0.14098453918789072, 'gradnorm/compressor.layer5': 0.13310643235639358, 'gradnorm/compressor.layer6': 0.11732447486850756, 'gradnorm/compressor.layer7': 0.1246291037341085, 'gradnorm/decoder.layer0': 0.16285077574995782, 'gradnorm/decoder.layer1': 0.16197352864632975, 'gradnorm/decoder.layer2': 0.206878884947449, 'gradnorm/max_param': 0.15017610788345337, 'epoch': 2.34}

 23%|██▎       | 4470/19090 [2:10:51<8:02:57,  1.98s/it]
 23%|██▎       | 4471/19090 [2:10:52<7:12:44,  1.78s/it]
 23%|██▎       | 4472/19090 [2:10:53<6:39:26,  1.64s/it]
 23%|██▎       | 4473/19090 [2:10:56<7:58:13,  1.96s/it]
 23%|██▎       | 4474/19090 [2:10:57<7:10:37,  1.77s/it]
 23%|██▎       | 4475/19090 [2:10:59<6:35:27,  1.62s/it]
 23%|██▎       | 4476/19090 [2:11:01<8:01:26,  1.98s/it]
 23%|██▎       | 4477/19090 [2:11:03<7:11:28,  1.77s/it]
 23%|██▎       | 4478/19090 [2:11:04<6:36:00,  1.63s/it]
 23%|██▎       | 4479/19090 [2:11:07<8:16:12,  2.04s/it]
 23%|██▎       | 4480/19090 [2:11:08<7:22:22,  1.82s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.5634416341781616, 'learning_rate': 4.7289982708344035e-05, 'recon_loss': 0.004333711706567556, 'recon_static': 0.003377563931280747, 'recon_motion': 0.0009561479455442168, 'recon_norm': 2.2886043190956116, 'recon_norm_std': 0.08557229861617088, 'target_norm': 1.6209096014499664, 'recon_cos': 0.40604673251509665, 'gradnorm/compressor.query': 0.009596412149059623, 'gradnorm/compressor.layer0': 0.1595562768784243, 'gradnorm/compressor.layer1': 0.221344126679357, 'gradnorm/compressor.layer2': 0.18611311593169982, 'gradnorm/compressor.layer3': 0.17399426353798383, 'gradnorm/compressor.layer4': 0.14926933933924844, 'gradnorm/compressor.layer5': 0.1377633041903691, 'gradnorm/compressor.layer6': 0.12401813210381164, 'gradnorm/compressor.layer7': 0.13013450357314307, 'gradnorm/decoder.layer0': 0.17140472621814182, 'gradnorm/decoder.layer1': 0.1683609342834771, 'gradnorm/decoder.layer2': 0.21612568650321412, 'gradnorm/max_param': 0.15633298456668854, 'epoch': 2.35}

 23%|██▎       | 4480/19090 [2:11:08<7:22:22,  1.82s/it]
 23%|██▎       | 4481/19090 [2:11:10<6:43:48,  1.66s/it]
 23%|██▎       | 4482/19090 [2:11:12<7:26:41,  1.83s/it]
 23%|██▎       | 4483/19090 [2:11:13<6:47:16,  1.67s/it]
 23%|██▎       | 4484/19090 [2:11:14<6:20:24,  1.56s/it]
 23%|██▎       | 4485/19090 [2:11:17<8:07:53,  2.00s/it]
 23%|██▎       | 4486/19090 [2:11:19<7:17:01,  1.80s/it]
 24%|██▎       | 4487/19090 [2:11:20<6:39:44,  1.64s/it]
 24%|██▎       | 4488/19090 [2:11:23<7:58:34,  1.97s/it]
 24%|██▎       | 4489/19090 [2:11:24<7:09:34,  1.77s/it]
 24%|██▎       | 4490/19090 [2:11:25<6:35:02,  1.62s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 0.5282389521598816, 'learning_rate': 4.7269245375546214e-05, 'recon_loss': 0.00406423524254933, 'recon_static': 0.003108146454906091, 'recon_motion': 0.0009560889389831573, 'recon_norm': 2.214228755235672, 'recon_norm_std': 0.08482074849307537, 'target_norm': 1.6217714101076126, 'recon_cos': 0.41569192707538605, 'gradnorm/compressor.query': 0.00931301052684064, 'gradnorm/compressor.layer0': 0.1438009070875987, 'gradnorm/compressor.layer1': 0.2047742954442987, 'gradnorm/compressor.layer2': 0.17131296293397222, 'gradnorm/compressor.layer3': 0.16293163195053156, 'gradnorm/compressor.layer4': 0.14285105227681003, 'gradnorm/compressor.layer5': 0.12991814383759323, 'gradnorm/compressor.layer6': 0.11538176889339555, 'gradnorm/compressor.layer7': 0.12319620093637869, 'gradnorm/decoder.layer0': 0.163225411262555, 'gradnorm/decoder.layer1': 0.16069125739467027, 'gradnorm/decoder.layer2': 0.20543611252771418, 'gradnorm/max_param': 0.1467757672071457, 'epoch': 2.35}

 24%|██▎       | 4490/19090 [2:11:25<6:35:02,  1.62s/it]
 24%|██▎       | 4491/19090 [2:11:28<7:43:14,  1.90s/it]
 24%|██▎       | 4492/19090 [2:11:29<6:58:09,  1.72s/it]
 24%|██▎       | 4493/19090 [2:11:30<6:26:14,  1.59s/it]
 24%|██▎       | 4494/19090 [2:11:33<7:58:10,  1.97s/it]
 24%|██▎       | 4495/19090 [2:11:35<7:08:46,  1.76s/it]
 24%|██▎       | 4496/19090 [2:11:36<6:33:48,  1.62s/it]
 24%|██▎       | 4497/19090 [2:11:39<8:14:08,  2.03s/it]
 24%|██▎       | 4498/19090 [2:11:40<7:19:45,  1.81s/it]
 24%|██▎       | 4499/19090 [2:11:41<6:41:40,  1.65s/it]
 24%|██▎       | 4500/19090 [2:11:44<7:36:40,  1.88s/it]
                                                        
{'loss': 0.005, 'grad_norm': 0.6725409030914307, 'learning_rate': 4.724843358528282e-05, 'recon_loss': 0.005014591111103073, 'recon_static': 0.004060315247625112, 'recon_motion': 0.0009542757616145536, 'recon_norm': 2.4459417939186094, 'recon_norm_std': 0.07578790187835693, 'target_norm': 1.6178069591522217, 'recon_cos': 0.3803360804915428, 'gradnorm/compressor.query': 0.01267490787502309, 'gradnorm/compressor.layer0': 0.18581623803467748, 'gradnorm/compressor.layer1': 0.2616451074416157, 'gradnorm/compressor.layer2': 0.21715514333094124, 'gradnorm/compressor.layer3': 0.20709904874127744, 'gradnorm/compressor.layer4': 0.17642637127917296, 'gradnorm/compressor.layer5': 0.1661642923584432, 'gradnorm/compressor.layer6': 0.14699130399399432, 'gradnorm/compressor.layer7': 0.1538557737135675, 'gradnorm/decoder.layer0': 0.1995937520462115, 'gradnorm/decoder.layer1': 0.20515021918875717, 'gradnorm/decoder.layer2': 0.27051663986841507, 'gradnorm/max_param': 0.1899847388267517, 'epoch': 2.36}

 24%|██▎       | 4500/19090 [2:11:44<7:36:40,  1.88s/it]
 24%|██▎       | 4501/19090 [2:11:48<10:18:38,  2.54s/it]
 24%|██▎       | 4502/19090 [2:11:49<8:46:43,  2.17s/it] 
 24%|██▎       | 4503/19090 [2:11:50<7:42:11,  1.90s/it]
 24%|██▎       | 4504/19090 [2:11:52<6:57:17,  1.72s/it]
 24%|██▎       | 4505/19090 [2:11:53<6:25:49,  1.59s/it]
 24%|██▎       | 4506/19090 [2:11:54<6:03:40,  1.50s/it]
 24%|██▎       | 4507/19090 [2:11:56<5:48:34,  1.43s/it]
 24%|██▎       | 4508/19090 [2:11:57<5:37:40,  1.39s/it]
 24%|██▎       | 4509/19090 [2:11:58<5:30:16,  1.36s/it]
 24%|██▎       | 4510/19090 [2:11:59<5:24:47,  1.34s/it]
                                                        
{'loss': 0.005, 'grad_norm': 0.5686471462249756, 'learning_rate': 4.7227547407138316e-05, 'recon_loss': 0.004988781141582877, 'recon_static': 0.004030917200725526, 'recon_motion': 0.0009578640529070981, 'recon_norm': 2.4438862979412077, 'recon_norm_std': 0.07259221784770489, 'target_norm': 1.6207954704761505, 'recon_cos': 0.3781305395066738, 'gradnorm/compressor.query': 0.009296138604284972, 'gradnorm/compressor.layer0': 0.15455457438067716, 'gradnorm/compressor.layer1': 0.21846503190663705, 'gradnorm/compressor.layer2': 0.18464201084587584, 'gradnorm/compressor.layer3': 0.1765505170553669, 'gradnorm/compressor.layer4': 0.15231218989234635, 'gradnorm/compressor.layer5': 0.14054907630057434, 'gradnorm/compressor.layer6': 0.12616876238122135, 'gradnorm/compressor.layer7': 0.1313594401579353, 'gradnorm/decoder.layer0': 0.1782070936959336, 'gradnorm/decoder.layer1': 0.17546256635339477, 'gradnorm/decoder.layer2': 0.21845945579469145, 'gradnorm/max_param': 0.157553568482399, 'epoch': 2.36}

 24%|██▎       | 4510/19090 [2:11:59<5:24:47,  1.34s/it]
 24%|██▎       | 4511/19090 [2:12:01<5:21:02,  1.32s/it]
 24%|██▎       | 4512/19090 [2:12:02<5:18:15,  1.31s/it]
 24%|██▎       | 4513/19090 [2:12:03<5:26:09,  1.34s/it]
 24%|██▎       | 4514/19090 [2:12:05<5:38:47,  1.39s/it]
 24%|██▎       | 4515/19090 [2:12:07<6:13:18,  1.54s/it]
 24%|██▎       | 4516/19090 [2:12:09<6:38:02,  1.64s/it]
 24%|██▎       | 4517/19090 [2:12:11<7:06:32,  1.76s/it]
 24%|██▎       | 4518/19090 [2:12:12<6:52:40,  1.70s/it]
 24%|██▎       | 4519/19090 [2:12:14<7:10:18,  1.77s/it]
 24%|██▎       | 4520/19090 [2:12:16<6:55:06,  1.71s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.5277787446975708, 'learning_rate': 4.720658691094586e-05, 'recon_loss': 0.003952149476390332, 'recon_static': 0.0029934023215901105, 'recon_motion': 0.0009587471591657959, 'recon_norm': 2.1856578290462494, 'recon_norm_std': 0.07955378219485283, 'target_norm': 1.6220402628183366, 'recon_cos': 0.4223577082157135, 'gradnorm/compressor.query': 0.00883517322335842, 'gradnorm/compressor.layer0': 0.14229549195000862, 'gradnorm/compressor.layer1': 0.20538397189479876, 'gradnorm/compressor.layer2': 0.17847119076769272, 'gradnorm/compressor.layer3': 0.1645677144651157, 'gradnorm/compressor.layer4': 0.13882239326146908, 'gradnorm/compressor.layer5': 0.1299882976716419, 'gradnorm/compressor.layer6': 0.11298949390179361, 'gradnorm/compressor.layer7': 0.12291139001013417, 'gradnorm/decoder.layer0': 0.16307695444002554, 'gradnorm/decoder.layer1': 0.16033408131254773, 'gradnorm/decoder.layer2': 0.20190020998747354, 'gradnorm/max_param': 0.14469116926193237, 'epoch': 2.37}

 24%|██▎       | 4520/19090 [2:12:16<6:55:06,  1.71s/it]
 24%|██▎       | 4521/19090 [2:12:18<7:13:38,  1.79s/it]
 24%|██▎       | 4522/19090 [2:12:20<7:11:15,  1.78s/it]
 24%|██▎       | 4523/19090 [2:12:21<7:17:14,  1.80s/it]
 24%|██▎       | 4524/19090 [2:12:23<7:12:21,  1.78s/it]
 24%|██▎       | 4525/19090 [2:12:25<7:15:04,  1.79s/it]
 24%|██▎       | 4526/19090 [2:12:27<7:19:13,  1.81s/it]
 24%|██▎       | 4527/19090 [2:12:29<7:18:38,  1.81s/it]
 24%|██▎       | 4528/19090 [2:12:30<7:01:58,  1.74s/it]
 24%|██▎       | 4529/19090 [2:12:32<7:23:42,  1.83s/it]
 24%|██▎       | 4530/19090 [2:12:34<7:09:18,  1.77s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.5651185512542725, 'learning_rate': 4.718555216678709e-05, 'recon_loss': 0.003639492386719212, 'recon_static': 0.002688285952899605, 'recon_motion': 0.0009512063668807968, 'recon_norm': 2.1029908299446105, 'recon_norm_std': 0.08150062542408705, 'target_norm': 1.618692362308502, 'recon_cos': 0.43889460191130636, 'gradnorm/compressor.query': 0.009139433509552732, 'gradnorm/compressor.layer0': 0.14912980909179407, 'gradnorm/compressor.layer1': 0.21462079530539704, 'gradnorm/compressor.layer2': 0.1836755752839778, 'gradnorm/compressor.layer3': 0.17337075078806674, 'gradnorm/compressor.layer4': 0.14880682775348508, 'gradnorm/compressor.layer5': 0.14095266897019726, 'gradnorm/compressor.layer6': 0.12463887771246326, 'gradnorm/compressor.layer7': 0.13284541171068157, 'gradnorm/decoder.layer0': 0.17474978666367558, 'gradnorm/decoder.layer1': 0.17791339745261356, 'gradnorm/decoder.layer2': 0.22314793690559634, 'gradnorm/max_param': 0.1542903631925583, 'epoch': 2.37}

 24%|██▎       | 4530/19090 [2:12:34<7:09:18,  1.77s/it]
 24%|██▎       | 4531/19090 [2:12:36<7:00:44,  1.73s/it]
 24%|██▎       | 4532/19090 [2:12:38<7:28:50,  1.85s/it]
 24%|██▎       | 4533/19090 [2:12:39<7:08:12,  1.76s/it]
 24%|██▍       | 4534/19090 [2:12:41<7:05:27,  1.75s/it]
 24%|██▍       | 4535/19090 [2:12:43<7:49:09,  1.93s/it]
 24%|██▍       | 4536/19090 [2:12:45<7:32:57,  1.87s/it]
 24%|██▍       | 4537/19090 [2:12:47<7:08:21,  1.77s/it]
 24%|██▍       | 4538/19090 [2:12:49<7:42:47,  1.91s/it]
 24%|██▍       | 4539/19090 [2:12:50<7:16:48,  1.80s/it]
 24%|██▍       | 4540/19090 [2:12:52<6:49:13,  1.69s/it]
                                                        
{'loss': 0.0055, 'grad_norm': 0.6680550575256348, 'learning_rate': 4.7164443244991916e-05, 'recon_loss': 0.00546627645380795, 'recon_static': 0.004505266703199595, 'recon_motion': 0.0009610097404220141, 'recon_norm': 2.552603894472122, 'recon_norm_std': 0.06894983984529972, 'target_norm': 1.620591014623642, 'recon_cos': 0.360985042899847, 'gradnorm/compressor.query': 0.011303260198961962, 'gradnorm/compressor.layer0': 0.1897241384356474, 'gradnorm/compressor.layer1': 0.26009445602099335, 'gradnorm/compressor.layer2': 0.21517281906888142, 'gradnorm/compressor.layer3': 0.20468133050779022, 'gradnorm/compressor.layer4': 0.17521415468785687, 'gradnorm/compressor.layer5': 0.1659694551331889, 'gradnorm/compressor.layer6': 0.14823320409320975, 'gradnorm/compressor.layer7': 0.156568072347403, 'gradnorm/decoder.layer0': 0.20693303423630816, 'gradnorm/decoder.layer1': 0.20228093177396747, 'gradnorm/decoder.layer2': 0.25663243052785056, 'gradnorm/max_param': 0.1829230636358261, 'epoch': 2.38}

 24%|██▍       | 4540/19090 [2:12:52<6:49:13,  1.69s/it]
 24%|██▍       | 4541/19090 [2:12:54<7:30:59,  1.86s/it]
 24%|██▍       | 4542/19090 [2:12:56<7:25:22,  1.84s/it]
 24%|██▍       | 4543/19090 [2:12:57<6:49:35,  1.69s/it]
 24%|██▍       | 4544/19090 [2:12:59<6:57:39,  1.72s/it]
 24%|██▍       | 4545/19090 [2:13:01<7:22:30,  1.83s/it]
 24%|██▍       | 4546/19090 [2:13:02<6:43:08,  1.66s/it]
 24%|██▍       | 4547/19090 [2:13:04<7:00:19,  1.73s/it]
 24%|██▍       | 4548/19090 [2:13:06<7:28:54,  1.85s/it]
 24%|██▍       | 4549/19090 [2:13:08<6:47:32,  1.68s/it]
 24%|██▍       | 4550/19090 [2:13:09<6:39:21,  1.65s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.6111655831336975, 'learning_rate': 4.7143260216138255e-05, 'recon_loss': 0.004525777383241803, 'recon_static': 0.0035738300648517907, 'recon_motion': 0.0009519472951069474, 'recon_norm': 2.338759160041809, 'recon_norm_std': 0.07458526212722064, 'target_norm': 1.6197348475456237, 'recon_cos': 0.3982994414865971, 'gradnorm/compressor.query': 0.01081620429141227, 'gradnorm/compressor.layer0': 0.17061310517940698, 'gradnorm/compressor.layer1': 0.2364176746035866, 'gradnorm/compressor.layer2': 0.20173010178824322, 'gradnorm/compressor.layer3': 0.18835093209937548, 'gradnorm/compressor.layer4': 0.15865096530599598, 'gradnorm/compressor.layer5': 0.14909759366292674, 'gradnorm/compressor.layer6': 0.13395422465572632, 'gradnorm/compressor.layer7': 0.14353823405324523, 'gradnorm/decoder.layer0': 0.18528810827861442, 'gradnorm/decoder.layer1': 0.1884429881971535, 'gradnorm/decoder.layer2': 0.23757830100788394, 'gradnorm/max_param': 0.16645340621471405, 'epoch': 2.38}

 24%|██▍       | 4550/19090 [2:13:09<6:39:21,  1.65s/it]
 24%|██▍       | 4551/19090 [2:13:12<7:33:56,  1.87s/it]
 24%|██▍       | 4552/19090 [2:13:13<6:50:59,  1.70s/it]
 24%|██▍       | 4553/19090 [2:13:14<6:38:30,  1.64s/it]
 24%|██▍       | 4554/19090 [2:13:17<7:22:02,  1.82s/it]
 24%|██▍       | 4555/19090 [2:13:18<6:42:57,  1.66s/it]
 24%|██▍       | 4556/19090 [2:13:20<6:41:57,  1.66s/it]
 24%|██▍       | 4557/19090 [2:13:22<7:36:27,  1.88s/it]
 24%|██▍       | 4558/19090 [2:13:23<6:52:35,  1.70s/it]
 24%|██▍       | 4559/19090 [2:13:25<6:53:07,  1.71s/it]
 24%|██▍       | 4560/19090 [2:13:27<7:32:59,  1.87s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.5002221465110779, 'learning_rate': 4.7122003151051774e-05, 'recon_loss': 0.004237195861060172, 'recon_static': 0.003283920686226338, 'recon_motion': 0.0009532751675578766, 'recon_norm': 2.2588979244232177, 'recon_norm_std': 0.07444931576028466, 'target_norm': 1.6187788605690003, 'recon_cos': 0.40883345901966095, 'gradnorm/compressor.query': 0.008200522507376963, 'gradnorm/compressor.layer0': 0.13409289088937454, 'gradnorm/compressor.layer1': 0.1948431383473601, 'gradnorm/compressor.layer2': 0.1690987237739172, 'gradnorm/compressor.layer3': 0.1570013637001347, 'gradnorm/compressor.layer4': 0.13154213982245158, 'gradnorm/compressor.layer5': 0.11999928883376589, 'gradnorm/compressor.layer6': 0.10797096282607013, 'gradnorm/compressor.layer7': 0.11700345926005509, 'gradnorm/decoder.layer0': 0.15313372830407598, 'gradnorm/decoder.layer1': 0.15410543589202313, 'gradnorm/decoder.layer2': 0.19161742102956794, 'gradnorm/max_param': 0.13643941283226013, 'epoch': 2.39}

 24%|██▍       | 4560/19090 [2:13:27<7:32:59,  1.87s/it]
 24%|██▍       | 4561/19090 [2:13:28<6:50:24,  1.69s/it]
 24%|██▍       | 4562/19090 [2:13:30<6:46:48,  1.68s/it]
 24%|██▍       | 4563/19090 [2:13:33<8:21:05,  2.07s/it]
 24%|██▍       | 4564/19090 [2:13:34<7:24:05,  1.83s/it]
 24%|██▍       | 4565/19090 [2:13:36<6:44:00,  1.67s/it]
 24%|██▍       | 4566/19090 [2:13:39<8:20:18,  2.07s/it]
 24%|██▍       | 4567/19090 [2:13:40<7:23:25,  1.83s/it]
 24%|██▍       | 4568/19090 [2:13:41<6:43:31,  1.67s/it]
 24%|██▍       | 4569/19090 [2:13:44<8:06:46,  2.01s/it]
 24%|██▍       | 4570/19090 [2:13:45<7:14:14,  1.79s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.4173370897769928, 'learning_rate': 4.7100672120805714e-05, 'recon_loss': 0.0031250529282260686, 'recon_static': 0.0021755614463472737, 'recon_motion': 0.0009494915619143284, 'recon_norm': 1.9582064419984817, 'recon_norm_std': 0.08419404476881027, 'target_norm': 1.618926328420639, 'recon_cos': 0.47198629602789877, 'gradnorm/compressor.query': 0.006537905077412736, 'gradnorm/compressor.layer0': 0.11445728235863924, 'gradnorm/compressor.layer1': 0.16498188856816132, 'gradnorm/compressor.layer2': 0.13942133169764115, 'gradnorm/compressor.layer3': 0.1297340042543587, 'gradnorm/compressor.layer4': 0.10936260153115744, 'gradnorm/compressor.layer5': 0.09948942966181049, 'gradnorm/compressor.layer6': 0.09054829253238283, 'gradnorm/compressor.layer7': 0.09589743837041975, 'gradnorm/decoder.layer0': 0.12860017621379038, 'gradnorm/decoder.layer1': 0.12958623252959717, 'gradnorm/decoder.layer2': 0.15797009200555226, 'gradnorm/max_param': 0.11320837587118149, 'epoch': 2.39}

 24%|██▍       | 4570/19090 [2:13:45<7:14:14,  1.79s/it]
 24%|██▍       | 4571/19090 [2:13:47<6:37:13,  1.64s/it]
 24%|██▍       | 4572/19090 [2:13:49<7:57:59,  1.98s/it]
 24%|██▍       | 4573/19090 [2:13:51<7:07:24,  1.77s/it]
 24%|██▍       | 4574/19090 [2:13:52<6:32:21,  1.62s/it]
 24%|██▍       | 4575/19090 [2:13:55<8:05:12,  2.01s/it]
 24%|██▍       | 4576/19090 [2:13:56<7:12:42,  1.79s/it]
 24%|██▍       | 4577/19090 [2:13:57<6:35:57,  1.64s/it]
 24%|██▍       | 4578/19090 [2:14:00<8:04:09,  2.00s/it]
 24%|██▍       | 4579/19090 [2:14:02<7:12:39,  1.79s/it]
 24%|██▍       | 4580/19090 [2:14:03<6:35:57,  1.64s/it]
                                                        
{'loss': 0.0028, 'grad_norm': 0.4003365933895111, 'learning_rate': 4.707926719672062e-05, 'recon_loss': 0.0027653821744024754, 'recon_static': 0.0018027886108029634, 'recon_motion': 0.0009625935010262765, 'recon_norm': 1.8433805882930756, 'recon_norm_std': 0.08562031295150518, 'target_norm': 1.621517613530159, 'recon_cos': 0.49801270589232444, 'gradnorm/compressor.query': 0.006328936573276452, 'gradnorm/compressor.layer0': 0.1067621127692153, 'gradnorm/compressor.layer1': 0.15948907939303164, 'gradnorm/compressor.layer2': 0.1404151144941817, 'gradnorm/compressor.layer3': 0.1276597312373197, 'gradnorm/compressor.layer4': 0.1048940569073112, 'gradnorm/compressor.layer5': 0.09745980053775531, 'gradnorm/compressor.layer6': 0.08505319650477917, 'gradnorm/compressor.layer7': 0.0910417513124973, 'gradnorm/decoder.layer0': 0.12076451552063162, 'gradnorm/decoder.layer1': 0.12195269449562464, 'gradnorm/decoder.layer2': 0.14798866979782854, 'gradnorm/max_param': 0.10666610300540924, 'epoch': 2.4}

 24%|██▍       | 4580/19090 [2:14:03<6:35:57,  1.64s/it]
 24%|██▍       | 4581/19090 [2:14:05<7:38:43,  1.90s/it]
 24%|██▍       | 4582/19090 [2:14:07<6:53:52,  1.71s/it]
 24%|██▍       | 4583/19090 [2:14:08<6:22:36,  1.58s/it]
 24%|██▍       | 4584/19090 [2:14:10<7:29:13,  1.86s/it]
 24%|██▍       | 4585/19090 [2:14:12<6:48:47,  1.69s/it]
 24%|██▍       | 4586/19090 [2:14:13<6:19:28,  1.57s/it]
 24%|██▍       | 4587/19090 [2:14:15<7:27:28,  1.85s/it]
 24%|██▍       | 4588/19090 [2:14:17<6:47:01,  1.68s/it]
 24%|██▍       | 4589/19090 [2:14:18<6:18:01,  1.56s/it]
 24%|██▍       | 4590/19090 [2:14:21<7:33:38,  1.88s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.6286261081695557, 'learning_rate': 4.705778845036407e-05, 'recon_loss': 0.0033326156146358697, 'recon_static': 0.0023737791721941902, 'recon_motion': 0.0009588363929651678, 'recon_norm': 2.0065126955509185, 'recon_norm_std': 0.08038252461701631, 'target_norm': 1.6209967136383057, 'recon_cos': 0.4633399210870266, 'gradnorm/compressor.query': 0.010053375731484925, 'gradnorm/compressor.layer0': 0.16845867322038602, 'gradnorm/compressor.layer1': 0.24789236167301779, 'gradnorm/compressor.layer2': 0.21187568352175484, 'gradnorm/compressor.layer3': 0.19646091434877433, 'gradnorm/compressor.layer4': 0.16832162694977731, 'gradnorm/compressor.layer5': 0.1534301522296126, 'gradnorm/compressor.layer6': 0.1359209420544559, 'gradnorm/compressor.layer7': 0.14647372524966948, 'gradnorm/decoder.layer0': 0.19231613087938035, 'gradnorm/decoder.layer1': 0.19205989943635868, 'gradnorm/decoder.layer2': 0.2368081396174155, 'gradnorm/max_param': 0.1696324646472931, 'epoch': 2.4}

 24%|██▍       | 4590/19090 [2:14:21<7:33:38,  1.88s/it]
 24%|██▍       | 4591/19090 [2:14:22<6:50:51,  1.70s/it]
 24%|██▍       | 4592/19090 [2:14:23<6:21:13,  1.58s/it]
 24%|██▍       | 4593/19090 [2:14:26<7:58:59,  1.98s/it]
 24%|██▍       | 4594/19090 [2:14:27<7:08:38,  1.77s/it]
 24%|██▍       | 4595/19090 [2:14:29<6:33:26,  1.63s/it]
 24%|██▍       | 4596/19090 [2:14:32<8:18:41,  2.06s/it]
 24%|██▍       | 4597/19090 [2:14:33<7:22:16,  1.83s/it]
 24%|██▍       | 4598/19090 [2:14:34<6:42:45,  1.67s/it]
 24%|██▍       | 4599/19090 [2:14:37<7:56:37,  1.97s/it]
 24%|██▍       | 4600/19090 [2:14:38<7:06:58,  1.77s/it]
                                                        
{'loss': 0.0048, 'grad_norm': 0.6472981572151184, 'learning_rate': 4.70362359535505e-05, 'recon_loss': 0.004837263550143689, 'recon_static': 0.003882867528591305, 'recon_motion': 0.0009543960200971924, 'recon_norm': 2.4086138904094696, 'recon_norm_std': 0.06957631399855017, 'target_norm': 1.6197010576725006, 'recon_cos': 0.38229767307639123, 'gradnorm/compressor.query': 0.011275900550295882, 'gradnorm/compressor.layer0': 0.18174214756136314, 'gradnorm/compressor.layer1': 0.25861408868195035, 'gradnorm/compressor.layer2': 0.2177022124681819, 'gradnorm/compressor.layer3': 0.19911169466788234, 'gradnorm/compressor.layer4': 0.1679465423363632, 'gradnorm/compressor.layer5': 0.15764783610905952, 'gradnorm/compressor.layer6': 0.1415711666237402, 'gradnorm/compressor.layer7': 0.15211455515116515, 'gradnorm/decoder.layer0': 0.19570029860109717, 'gradnorm/decoder.layer1': 0.19580642235447063, 'gradnorm/decoder.layer2': 0.24297675613263628, 'gradnorm/max_param': 0.17442363500595093, 'epoch': 2.41}

 24%|██▍       | 4600/19090 [2:14:38<7:06:58,  1.77s/it]
 24%|██▍       | 4601/19090 [2:14:40<6:32:20,  1.62s/it]
 24%|██▍       | 4602/19090 [2:14:42<7:45:01,  1.93s/it]
 24%|██▍       | 4603/19090 [2:14:44<6:58:51,  1.73s/it]
 24%|██▍       | 4604/19090 [2:14:45<6:26:48,  1.60s/it]
 24%|██▍       | 4605/19090 [2:14:47<7:09:31,  1.78s/it]
 24%|██▍       | 4606/19090 [2:14:48<6:33:42,  1.63s/it]
 24%|██▍       | 4607/19090 [2:14:50<6:09:15,  1.53s/it]
 24%|██▍       | 4608/19090 [2:14:52<7:31:40,  1.87s/it]
 24%|██▍       | 4609/19090 [2:14:54<6:49:14,  1.70s/it]
 24%|██▍       | 4610/19090 [2:14:55<6:19:44,  1.57s/it]
                                                        
{'loss': 0.004, 'grad_norm': 0.5785634517669678, 'learning_rate': 4.701460977834092e-05, 'recon_loss': 0.004005160828819498, 'recon_static': 0.0030493481492158024, 'recon_motion': 0.00095581265486544, 'recon_norm': 2.2011483728885652, 'recon_norm_std': 0.07365673985332251, 'target_norm': 1.619805520772934, 'recon_cos': 0.4214831404387951, 'gradnorm/compressor.query': 0.009580779828855244, 'gradnorm/compressor.layer0': 0.16038719831356096, 'gradnorm/compressor.layer1': 0.2310959879835811, 'gradnorm/compressor.layer2': 0.19211814799837904, 'gradnorm/compressor.layer3': 0.1796155453783961, 'gradnorm/compressor.layer4': 0.152773946626147, 'gradnorm/compressor.layer5': 0.13989217202225837, 'gradnorm/compressor.layer6': 0.12484541289734326, 'gradnorm/compressor.layer7': 0.1345373445497207, 'gradnorm/decoder.layer0': 0.17349175794432053, 'gradnorm/decoder.layer1': 0.17814329287643624, 'gradnorm/decoder.layer2': 0.21888943894208124, 'gradnorm/max_param': 0.15663282573223114, 'epoch': 2.41}

 24%|██▍       | 4610/19090 [2:14:55<6:19:44,  1.57s/it]
 24%|██▍       | 4611/19090 [2:14:58<8:02:13,  2.00s/it]
 24%|██▍       | 4612/19090 [2:14:59<7:10:32,  1.78s/it]
 24%|██▍       | 4613/19090 [2:15:00<6:34:17,  1.63s/it]
 24%|██▍       | 4614/19090 [2:15:03<8:05:11,  2.01s/it]
 24%|██▍       | 4615/19090 [2:15:05<7:12:26,  1.79s/it]
 24%|██▍       | 4616/19090 [2:15:06<6:35:50,  1.64s/it]
 24%|██▍       | 4617/19090 [2:15:09<8:05:43,  2.01s/it]
 24%|██▍       | 4618/19090 [2:15:10<7:13:02,  1.80s/it]
 24%|██▍       | 4619/19090 [2:15:11<6:35:55,  1.64s/it]
 24%|██▍       | 4620/19090 [2:15:14<7:58:35,  1.98s/it]
                                                        
{'loss': 0.0047, 'grad_norm': 0.6594323515892029, 'learning_rate': 4.699290999704267e-05, 'recon_loss': 0.004729724023491144, 'recon_static': 0.003775634180055931, 'recon_motion': 0.000954089815786574, 'recon_norm': 2.385104250907898, 'recon_norm_std': 0.06731166671961546, 'target_norm': 1.617600393295288, 'recon_cos': 0.388305851072073, 'gradnorm/compressor.query': 0.010801730753369224, 'gradnorm/compressor.layer0': 0.18359934006966624, 'gradnorm/compressor.layer1': 0.2603923007374385, 'gradnorm/compressor.layer2': 0.21636349086905937, 'gradnorm/compressor.layer3': 0.20563480571219803, 'gradnorm/compressor.layer4': 0.17226815773276308, 'gradnorm/compressor.layer5': 0.15942467518416395, 'gradnorm/compressor.layer6': 0.14036923567706022, 'gradnorm/compressor.layer7': 0.1570158369651485, 'gradnorm/decoder.layer0': 0.20311054765198588, 'gradnorm/decoder.layer1': 0.20256799658580227, 'gradnorm/decoder.layer2': 0.24976977536091452, 'gradnorm/max_param': 0.17625652253627777, 'epoch': 2.42}

 24%|██▍       | 4620/19090 [2:15:14<7:58:35,  1.98s/it]
 24%|██▍       | 4621/19090 [2:15:15<7:07:46,  1.77s/it]
 24%|██▍       | 4622/19090 [2:15:17<6:32:22,  1.63s/it]
 24%|██▍       | 4623/19090 [2:15:19<7:46:49,  1.94s/it]
 24%|██▍       | 4624/19090 [2:15:21<6:59:41,  1.74s/it]
 24%|██▍       | 4625/19090 [2:15:22<6:26:44,  1.60s/it]
 24%|██▍       | 4626/19090 [2:15:25<7:44:04,  1.93s/it]
 24%|██▍       | 4627/19090 [2:15:26<6:58:00,  1.73s/it]
 24%|██▍       | 4628/19090 [2:15:27<6:26:28,  1.60s/it]
 24%|██▍       | 4629/19090 [2:15:30<7:37:26,  1.90s/it]
 24%|██▍       | 4630/19090 [2:15:31<6:53:35,  1.72s/it]
                                                        
{'loss': 0.0047, 'grad_norm': 0.5839895009994507, 'learning_rate': 4.697113668220922e-05, 'recon_loss': 0.004655115946661681, 'recon_static': 0.0037010762956924737, 'recon_motion': 0.0009540396597003565, 'recon_norm': 2.361668574810028, 'recon_norm_std': 0.07094991588965058, 'target_norm': 1.6184275835752486, 'recon_cos': 0.3892657786607742, 'gradnorm/compressor.query': 0.009728707254570531, 'gradnorm/compressor.layer0': 0.16347538823443195, 'gradnorm/compressor.layer1': 0.23605282567041347, 'gradnorm/compressor.layer2': 0.19609930820870722, 'gradnorm/compressor.layer3': 0.18145871025693033, 'gradnorm/compressor.layer4': 0.15191430917505516, 'gradnorm/compressor.layer5': 0.14218893233824928, 'gradnorm/compressor.layer6': 0.12365459428039785, 'gradnorm/compressor.layer7': 0.13240122226139392, 'gradnorm/decoder.layer0': 0.1781746995757712, 'gradnorm/decoder.layer1': 0.17858640573908216, 'gradnorm/decoder.layer2': 0.21763775965354312, 'gradnorm/max_param': 0.15456992387771606, 'epoch': 2.43}

 24%|██▍       | 4630/19090 [2:15:31<6:53:35,  1.72s/it]
 24%|██▍       | 4631/19090 [2:15:32<6:23:29,  1.59s/it]
 24%|██▍       | 4632/19090 [2:15:35<7:35:11,  1.89s/it]
 24%|██▍       | 4633/19090 [2:15:36<6:51:23,  1.71s/it]
 24%|██▍       | 4634/19090 [2:15:38<6:21:05,  1.58s/it]
 24%|██▍       | 4635/19090 [2:15:40<7:58:56,  1.99s/it]
 24%|██▍       | 4636/19090 [2:15:42<7:08:23,  1.78s/it]
 24%|██▍       | 4637/19090 [2:15:43<6:33:10,  1.63s/it]
 24%|██▍       | 4638/19090 [2:15:46<7:59:51,  1.99s/it]
 24%|██▍       | 4639/19090 [2:15:47<7:08:49,  1.78s/it]
 24%|██▍       | 4640/19090 [2:15:48<6:33:45,  1.63s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.5971921682357788, 'learning_rate': 4.694928990663988e-05, 'recon_loss': 0.004647848044987768, 'recon_static': 0.0036839489766862243, 'recon_motion': 0.0009638989737140946, 'recon_norm': 2.3586941003799438, 'recon_norm_std': 0.08166504800319671, 'target_norm': 1.621716719865799, 'recon_cos': 0.3894231341779232, 'gradnorm/compressor.query': 0.010282091023327762, 'gradnorm/compressor.layer0': 0.163864578344513, 'gradnorm/compressor.layer1': 0.23670334624226555, 'gradnorm/compressor.layer2': 0.20017810840651581, 'gradnorm/compressor.layer3': 0.18675770452991358, 'gradnorm/compressor.layer4': 0.15808300473693168, 'gradnorm/compressor.layer5': 0.14601542749896707, 'gradnorm/compressor.layer6': 0.12852644599415744, 'gradnorm/compressor.layer7': 0.13893156182533295, 'gradnorm/decoder.layer0': 0.18329055586895476, 'gradnorm/decoder.layer1': 0.1804389995146633, 'gradnorm/decoder.layer2': 0.2245385514428025, 'gradnorm/max_param': 0.15965867042541504, 'epoch': 2.43}

 24%|██▍       | 4640/19090 [2:15:48<6:33:45,  1.63s/it]
 24%|██▍       | 4641/19090 [2:15:51<7:52:21,  1.96s/it]
 24%|██▍       | 4642/19090 [2:15:53<7:03:54,  1.76s/it]
 24%|██▍       | 4643/19090 [2:15:54<6:29:18,  1.62s/it]
 24%|██▍       | 4644/19090 [2:15:57<8:01:28,  2.00s/it]
 24%|██▍       | 4645/19090 [2:15:58<7:09:54,  1.79s/it]
 24%|██▍       | 4646/19090 [2:15:59<6:34:48,  1.64s/it]
 24%|██▍       | 4647/19090 [2:16:02<7:48:45,  1.95s/it]
 24%|██▍       | 4648/19090 [2:16:03<7:00:42,  1.75s/it]
 24%|██▍       | 4649/19090 [2:16:04<6:27:11,  1.61s/it]
 24%|██▍       | 4650/19090 [2:16:07<8:04:43,  2.01s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.5639619827270508, 'learning_rate': 4.692736974337959e-05, 'recon_loss': 0.004455485264770687, 'recon_static': 0.003500153654022142, 'recon_motion': 0.0009553316311212256, 'recon_norm': 2.3163511037826536, 'recon_norm_std': 0.09340393133461475, 'target_norm': 1.6209274351596832, 'recon_cos': 0.39965967610478403, 'gradnorm/compressor.query': 0.00930505694682632, 'gradnorm/compressor.layer0': 0.15619832614931078, 'gradnorm/compressor.layer1': 0.22883781314917404, 'gradnorm/compressor.layer2': 0.19216434060095344, 'gradnorm/compressor.layer3': 0.17858760517793054, 'gradnorm/compressor.layer4': 0.14717285487335496, 'gradnorm/compressor.layer5': 0.13266878907888735, 'gradnorm/compressor.layer6': 0.12064785154719469, 'gradnorm/compressor.layer7': 0.1276728257211673, 'gradnorm/decoder.layer0': 0.17052340638295913, 'gradnorm/decoder.layer1': 0.16973970228458093, 'gradnorm/decoder.layer2': 0.21065204343456578, 'gradnorm/max_param': 0.15005595982074738, 'epoch': 2.44}

 24%|██▍       | 4650/19090 [2:16:07<8:04:43,  2.01s/it]
 24%|██▍       | 4651/19090 [2:16:09<7:12:36,  1.80s/it]
 24%|██▍       | 4652/19090 [2:16:10<6:35:17,  1.64s/it]
 24%|██▍       | 4653/19090 [2:16:13<8:21:32,  2.08s/it]
 24%|██▍       | 4654/19090 [2:16:14<7:23:50,  1.84s/it]
 24%|██▍       | 4655/19090 [2:16:16<6:43:24,  1.68s/it]
 24%|██▍       | 4656/19090 [2:16:18<7:49:15,  1.95s/it]
 24%|██▍       | 4657/19090 [2:16:20<7:01:05,  1.75s/it]
 24%|██▍       | 4658/19090 [2:16:21<6:27:47,  1.61s/it]
 24%|██▍       | 4659/19090 [2:16:24<8:04:05,  2.01s/it]
 24%|██▍       | 4660/19090 [2:16:25<7:12:07,  1.80s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.4156206548213959, 'learning_rate': 4.690537626571864e-05, 'recon_loss': 0.0032292471965774893, 'recon_static': 0.0022806330292951317, 'recon_motion': 0.0009486141992965713, 'recon_norm': 1.9857161104679109, 'recon_norm_std': 0.110809494741261, 'target_norm': 1.6193865090608597, 'recon_cos': 0.46545354574918746, 'gradnorm/compressor.query': 0.0074760611804766, 'gradnorm/compressor.layer0': 0.1150244481724339, 'gradnorm/compressor.layer1': 0.17059375305650956, 'gradnorm/compressor.layer2': 0.14232057014186958, 'gradnorm/compressor.layer3': 0.13180328953887802, 'gradnorm/compressor.layer4': 0.1071526143869304, 'gradnorm/compressor.layer5': 0.09849409697817871, 'gradnorm/compressor.layer6': 0.08686679722890148, 'gradnorm/compressor.layer7': 0.09476412195647474, 'gradnorm/decoder.layer0': 0.12397049727649741, 'gradnorm/decoder.layer1': 0.12498259795769158, 'gradnorm/decoder.layer2': 0.1550043100961429, 'gradnorm/max_param': 0.10989288985729218, 'epoch': 2.44}

 24%|██▍       | 4660/19090 [2:16:25<7:12:07,  1.80s/it]
 24%|██▍       | 4661/19090 [2:16:26<6:35:33,  1.64s/it]
 24%|██▍       | 4662/19090 [2:16:29<7:54:12,  1.97s/it]
 24%|██▍       | 4663/19090 [2:16:30<7:04:45,  1.77s/it]
 24%|██▍       | 4664/19090 [2:16:32<6:30:12,  1.62s/it]
 24%|██▍       | 4665/19090 [2:16:34<7:34:47,  1.89s/it]
 24%|██▍       | 4666/19090 [2:16:36<6:51:18,  1.71s/it]
 24%|██▍       | 4667/19090 [2:16:37<6:20:32,  1.58s/it]
 24%|██▍       | 4668/19090 [2:16:40<7:56:12,  1.98s/it]
 24%|██▍       | 4669/19090 [2:16:41<7:05:52,  1.77s/it]
 24%|██▍       | 4670/19090 [2:16:42<6:30:59,  1.63s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.43737146258354187, 'learning_rate': 4.6883309547192476e-05, 'recon_loss': 0.003070097375893965, 'recon_static': 0.002105895147542469, 'recon_motion': 0.0009642023738706485, 'recon_norm': 1.9344928413629532, 'recon_norm_std': 0.12190461698919534, 'target_norm': 1.6243282109498978, 'recon_cos': 0.4770346872508526, 'gradnorm/compressor.query': 0.007393362828893581, 'gradnorm/compressor.layer0': 0.11748011930672163, 'gradnorm/compressor.layer1': 0.17846561749821388, 'gradnorm/compressor.layer2': 0.15373403770657137, 'gradnorm/compressor.layer3': 0.14026730796235212, 'gradnorm/compressor.layer4': 0.1148830205571101, 'gradnorm/compressor.layer5': 0.10474997211498036, 'gradnorm/compressor.layer6': 0.09168748221888041, 'gradnorm/compressor.layer7': 0.09981827242192846, 'gradnorm/decoder.layer0': 0.1296519328373449, 'gradnorm/decoder.layer1': 0.13218751581566177, 'gradnorm/decoder.layer2': 0.15954432818713402, 'gradnorm/max_param': 0.11397039145231247, 'epoch': 2.45}

 24%|██▍       | 4670/19090 [2:16:42<6:30:59,  1.63s/it]
 24%|██▍       | 4671/19090 [2:16:45<7:51:58,  1.96s/it]
 24%|██▍       | 4672/19090 [2:16:46<7:02:58,  1.76s/it]
 24%|██▍       | 4673/19090 [2:16:48<6:28:40,  1.62s/it]
 24%|██▍       | 4674/19090 [2:16:50<7:46:40,  1.94s/it]
 24%|██▍       | 4675/19090 [2:16:52<6:59:50,  1.75s/it]
 24%|██▍       | 4676/19090 [2:16:53<6:26:51,  1.61s/it]
 24%|██▍       | 4677/19090 [2:16:55<7:38:21,  1.91s/it]
 25%|██▍       | 4678/19090 [2:16:57<6:53:22,  1.72s/it]
 25%|██▍       | 4679/19090 [2:16:58<6:21:45,  1.59s/it]
 25%|██▍       | 4680/19090 [2:17:01<7:59:59,  2.00s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.6743322014808655, 'learning_rate': 4.686116966158142e-05, 'recon_loss': 0.003568691055988893, 'recon_static': 0.002609263511840254, 'recon_motion': 0.0009594276052666828, 'recon_norm': 2.0672622501850126, 'recon_norm_std': 0.12089770194143057, 'target_norm': 1.6201472133398056, 'recon_cos': 0.4488539665937424, 'gradnorm/compressor.query': 0.012337448108709145, 'gradnorm/compressor.layer0': 0.18925194037335913, 'gradnorm/compressor.layer1': 0.27102278844273964, 'gradnorm/compressor.layer2': 0.22456940004963286, 'gradnorm/compressor.layer3': 0.21140931486007192, 'gradnorm/compressor.layer4': 0.17692782197973453, 'gradnorm/compressor.layer5': 0.163592710446149, 'gradnorm/compressor.layer6': 0.14299850675330095, 'gradnorm/compressor.layer7': 0.15791427474823175, 'gradnorm/decoder.layer0': 0.1999934225192909, 'gradnorm/decoder.layer1': 0.20439924771425677, 'gradnorm/decoder.layer2': 0.25485043667195856, 'gradnorm/max_param': 0.17774245142936707, 'epoch': 2.45}

 25%|██▍       | 4680/19090 [2:17:01<7:59:59,  2.00s/it]
 25%|██▍       | 4681/19090 [2:17:02<7:08:24,  1.78s/it]
 25%|██▍       | 4682/19090 [2:17:04<6:32:58,  1.64s/it]
 25%|██▍       | 4683/19090 [2:17:06<7:59:32,  2.00s/it]
 25%|██▍       | 4684/19090 [2:17:08<7:08:34,  1.79s/it]
 25%|██▍       | 4685/19090 [2:17:09<6:32:18,  1.63s/it]
 25%|██▍       | 4686/19090 [2:17:11<7:25:04,  1.85s/it]
 25%|██▍       | 4687/19090 [2:17:13<6:44:05,  1.68s/it]
 25%|██▍       | 4688/19090 [2:17:14<6:15:45,  1.57s/it]
 25%|██▍       | 4689/19090 [2:17:17<7:42:33,  1.93s/it]
 25%|██▍       | 4690/19090 [2:17:18<6:56:32,  1.74s/it]
                                                        
{'loss': 0.0061, 'grad_norm': 0.6567772030830383, 'learning_rate': 4.68389566829104e-05, 'recon_loss': 0.006061903748195619, 'recon_static': 0.005098297307267785, 'recon_motion': 0.000963606548612006, 'recon_norm': 2.6812046706676482, 'recon_norm_std': 0.08879823442548514, 'target_norm': 1.622402411699295, 'recon_cos': 0.34528368785977365, 'gradnorm/compressor.query': 0.01121297217120818, 'gradnorm/compressor.layer0': 0.1837750209588828, 'gradnorm/compressor.layer1': 0.2669139471155713, 'gradnorm/compressor.layer2': 0.21781351379311845, 'gradnorm/compressor.layer3': 0.20495602742343197, 'gradnorm/compressor.layer4': 0.17148171207732635, 'gradnorm/compressor.layer5': 0.15685773320369092, 'gradnorm/compressor.layer6': 0.1405180484648003, 'gradnorm/compressor.layer7': 0.14937751298051513, 'gradnorm/decoder.layer0': 0.19852855275473158, 'gradnorm/decoder.layer1': 0.19899533403739555, 'gradnorm/decoder.layer2': 0.24835971568259135, 'gradnorm/max_param': 0.17679525911808014, 'epoch': 2.46}

 25%|██▍       | 4690/19090 [2:17:18<6:56:32,  1.74s/it]
 25%|██▍       | 4691/19090 [2:17:19<6:24:12,  1.60s/it]
 25%|██▍       | 4692/19090 [2:17:22<8:07:38,  2.03s/it]
 25%|██▍       | 4693/19090 [2:17:24<7:13:54,  1.81s/it]
 25%|██▍       | 4694/19090 [2:17:25<6:36:26,  1.65s/it]
 25%|██▍       | 4695/19090 [2:17:28<8:00:56,  2.00s/it]
 25%|██▍       | 4696/19090 [2:17:29<7:09:08,  1.79s/it]
 25%|██▍       | 4697/19090 [2:17:30<6:33:16,  1.64s/it]
 25%|██▍       | 4698/19090 [2:17:33<8:07:35,  2.03s/it]
 25%|██▍       | 4699/19090 [2:17:35<7:14:04,  1.81s/it]
 25%|██▍       | 4700/19090 [2:17:36<6:36:42,  1.65s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.500624418258667, 'learning_rate': 4.6816670685448775e-05, 'recon_loss': 0.003754335374105722, 'recon_static': 0.002814493386540562, 'recon_motion': 0.0009398420195793733, 'recon_norm': 2.1336436688899996, 'recon_norm_std': 0.10306935235857964, 'target_norm': 1.6183603078126907, 'recon_cos': 0.4352046735584736, 'gradnorm/compressor.query': 0.00912673675124813, 'gradnorm/compressor.layer0': 0.13863868047441927, 'gradnorm/compressor.layer1': 0.2024395546504138, 'gradnorm/compressor.layer2': 0.17289880333306246, 'gradnorm/compressor.layer3': 0.15596204091567303, 'gradnorm/compressor.layer4': 0.12839446197148582, 'gradnorm/compressor.layer5': 0.1209159014202708, 'gradnorm/compressor.layer6': 0.10668159211482874, 'gradnorm/compressor.layer7': 0.1142137760496149, 'gradnorm/decoder.layer0': 0.1517852240492413, 'gradnorm/decoder.layer1': 0.15410868721315077, 'gradnorm/decoder.layer2': 0.1838508388625162, 'gradnorm/max_param': 0.1306295245885849, 'epoch': 2.46}

 25%|██▍       | 4700/19090 [2:17:36<6:36:42,  1.65s/it]
 25%|██▍       | 4701/19090 [2:17:39<8:11:19,  2.05s/it]
 25%|██▍       | 4702/19090 [2:17:40<7:16:13,  1.82s/it]
 25%|██▍       | 4703/19090 [2:17:41<6:37:43,  1.66s/it]
 25%|██▍       | 4704/19090 [2:17:44<8:13:29,  2.06s/it]
 25%|██▍       | 4705/19090 [2:17:46<7:17:41,  1.83s/it]
 25%|██▍       | 4706/19090 [2:17:47<6:38:55,  1.66s/it]
 25%|██▍       | 4707/19090 [2:17:49<7:24:56,  1.86s/it]
 25%|██▍       | 4708/19090 [2:17:51<6:44:53,  1.69s/it]
 25%|██▍       | 4709/19090 [2:17:52<6:15:52,  1.57s/it]
 25%|██▍       | 4710/19090 [2:17:55<7:57:22,  1.99s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.4909747838973999, 'learning_rate': 4.679431174371001e-05, 'recon_loss': 0.003428854967933148, 'recon_static': 0.002467371488455683, 'recon_motion': 0.000961483565333765, 'recon_norm': 2.0390974044799806, 'recon_norm_std': 0.10892887394875288, 'target_norm': 1.620641642808914, 'recon_cos': 0.4504993610084057, 'gradnorm/compressor.query': 0.008324989263160507, 'gradnorm/compressor.layer0': 0.13452340209480174, 'gradnorm/compressor.layer1': 0.19674067001687634, 'gradnorm/compressor.layer2': 0.16602569998295036, 'gradnorm/compressor.layer3': 0.1538985156037717, 'gradnorm/compressor.layer4': 0.12860259501929883, 'gradnorm/compressor.layer5': 0.11916920834973002, 'gradnorm/compressor.layer6': 0.10480201546457925, 'gradnorm/compressor.layer7': 0.1142727605708764, 'gradnorm/decoder.layer0': 0.15023062604609633, 'gradnorm/decoder.layer1': 0.14849518608709955, 'gradnorm/decoder.layer2': 0.1831187942742312, 'gradnorm/max_param': 0.1285954713821411, 'epoch': 2.47}

 25%|██▍       | 4710/19090 [2:17:55<7:57:22,  1.99s/it]
 25%|██▍       | 4711/19090 [2:17:56<7:06:38,  1.78s/it]
 25%|██▍       | 4712/19090 [2:17:57<6:31:17,  1.63s/it]
 25%|██▍       | 4713/19090 [2:18:00<7:56:52,  1.99s/it]
 25%|██▍       | 4714/19090 [2:18:01<7:05:52,  1.78s/it]
 25%|██▍       | 4715/19090 [2:18:03<6:30:33,  1.63s/it]
 25%|██▍       | 4716/19090 [2:18:05<7:43:14,  1.93s/it]
 25%|██▍       | 4717/19090 [2:18:07<6:56:21,  1.74s/it]
 25%|██▍       | 4718/19090 [2:18:08<6:23:45,  1.60s/it]
 25%|██▍       | 4719/19090 [2:18:11<8:08:16,  2.04s/it]
 25%|██▍       | 4720/19090 [2:18:12<7:14:14,  1.81s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.5324130654335022, 'learning_rate': 4.6771879932451464e-05, 'recon_loss': 0.0036679462355095893, 'recon_static': 0.0027331317134667186, 'recon_motion': 0.0009348146864795126, 'recon_norm': 2.110608074069023, 'recon_norm_std': 0.10548655986785889, 'target_norm': 1.6180791288614274, 'recon_cos': 0.4388491176068783, 'gradnorm/compressor.query': 0.009224451885311628, 'gradnorm/compressor.layer0': 0.14778389207289538, 'gradnorm/compressor.layer1': 0.22036085342163375, 'gradnorm/compressor.layer2': 0.18371893287632718, 'gradnorm/compressor.layer3': 0.1682253197672046, 'gradnorm/compressor.layer4': 0.1398192541416276, 'gradnorm/compressor.layer5': 0.12836135950951663, 'gradnorm/compressor.layer6': 0.11228589025003063, 'gradnorm/compressor.layer7': 0.12257582706706896, 'gradnorm/decoder.layer0': 0.15633948267218362, 'gradnorm/decoder.layer1': 0.15971042640177685, 'gradnorm/decoder.layer2': 0.19318049277077867, 'gradnorm/max_param': 0.13792826235294342, 'epoch': 2.47}

 25%|██▍       | 4720/19090 [2:18:12<7:14:14,  1.81s/it]
 25%|██▍       | 4721/19090 [2:18:14<6:36:32,  1.66s/it]
 25%|██▍       | 4722/19090 [2:18:16<7:53:10,  1.98s/it]
 25%|██▍       | 4723/19090 [2:18:18<7:03:36,  1.77s/it]
 25%|██▍       | 4724/19090 [2:18:19<6:29:00,  1.62s/it]
 25%|██▍       | 4725/19090 [2:18:21<7:37:32,  1.91s/it]
 25%|██▍       | 4726/19090 [2:18:23<6:52:29,  1.72s/it]
 25%|██▍       | 4727/19090 [2:18:24<6:20:52,  1.59s/it]
 25%|██▍       | 4728/19090 [2:18:27<7:50:37,  1.97s/it]
 25%|██▍       | 4729/19090 [2:18:28<7:01:31,  1.76s/it]
 25%|██▍       | 4730/19090 [2:18:29<6:27:18,  1.62s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.5145432353019714, 'learning_rate': 4.6749375326674155e-05, 'recon_loss': 0.003573722264263779, 'recon_static': 0.002623199427034706, 'recon_motion': 0.0009505228052148595, 'recon_norm': 2.0832361400127413, 'recon_norm_std': 0.09920882638543845, 'target_norm': 1.6185571372509002, 'recon_cos': 0.44202988743782046, 'gradnorm/compressor.query': 0.008844591762919747, 'gradnorm/compressor.layer0': 0.14704397758585935, 'gradnorm/compressor.layer1': 0.21304059652009602, 'gradnorm/compressor.layer2': 0.1762279556570426, 'gradnorm/compressor.layer3': 0.15929447318189185, 'gradnorm/compressor.layer4': 0.13255828481930185, 'gradnorm/compressor.layer5': 0.12151696099288417, 'gradnorm/compressor.layer6': 0.10651928842183866, 'gradnorm/compressor.layer7': 0.11722762295180707, 'gradnorm/decoder.layer0': 0.154716440344984, 'gradnorm/decoder.layer1': 0.1563893650326991, 'gradnorm/decoder.layer2': 0.18819052975932288, 'gradnorm/max_param': 0.13386520743370056, 'epoch': 2.48}

 25%|██▍       | 4730/19090 [2:18:29<6:27:18,  1.62s/it]
 25%|██▍       | 4731/19090 [2:18:33<8:09:08,  2.04s/it]
 25%|██▍       | 4732/19090 [2:18:34<7:14:24,  1.82s/it]
 25%|██▍       | 4733/19090 [2:18:35<6:36:18,  1.66s/it]
 25%|██▍       | 4734/19090 [2:18:38<7:33:13,  1.89s/it]
 25%|██▍       | 4735/19090 [2:18:39<6:49:34,  1.71s/it]
 25%|██▍       | 4736/19090 [2:18:40<6:18:52,  1.58s/it]
 25%|██▍       | 4737/19090 [2:18:43<7:41:27,  1.93s/it]
 25%|██▍       | 4738/19090 [2:18:44<6:55:15,  1.74s/it]
 25%|██▍       | 4739/19090 [2:18:45<6:23:28,  1.60s/it]
 25%|██▍       | 4740/19090 [2:18:48<7:35:16,  1.90s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.5482240915298462, 'learning_rate': 4.672679800162248e-05, 'recon_loss': 0.0036904749053064735, 'recon_static': 0.0027375391335226597, 'recon_motion': 0.0009529357092105783, 'recon_norm': 2.1132011145353315, 'recon_norm_std': 0.08582032509148121, 'target_norm': 1.6168148905038833, 'recon_cos': 0.43425996154546737, 'gradnorm/compressor.query': 0.008378047487154636, 'gradnorm/compressor.layer0': 0.1493844894363734, 'gradnorm/compressor.layer1': 0.22590973145249066, 'gradnorm/compressor.layer2': 0.18940133438881476, 'gradnorm/compressor.layer3': 0.16901593578909918, 'gradnorm/compressor.layer4': 0.1448082331801383, 'gradnorm/compressor.layer5': 0.1305817249130978, 'gradnorm/compressor.layer6': 0.11280260003654638, 'gradnorm/compressor.layer7': 0.12380223936283206, 'gradnorm/decoder.layer0': 0.1674277108510838, 'gradnorm/decoder.layer1': 0.16883217325259983, 'gradnorm/decoder.layer2': 0.20020128566855974, 'gradnorm/max_param': 0.1406964808702469, 'epoch': 2.48}

 25%|██▍       | 4740/19090 [2:18:48<7:35:16,  1.90s/it]
 25%|██▍       | 4741/19090 [2:18:49<6:51:08,  1.72s/it]
 25%|██▍       | 4742/19090 [2:18:51<6:20:56,  1.59s/it]
 25%|██▍       | 4743/19090 [2:18:53<7:49:45,  1.96s/it]
 25%|██▍       | 4744/19090 [2:18:55<7:01:32,  1.76s/it]
 25%|██▍       | 4745/19090 [2:18:56<6:28:34,  1.63s/it]
 25%|██▍       | 4746/19090 [2:18:59<7:36:57,  1.91s/it]
 25%|██▍       | 4747/19090 [2:19:00<6:52:06,  1.72s/it]
 25%|██▍       | 4748/19090 [2:19:01<6:21:59,  1.60s/it]
 25%|██▍       | 4749/19090 [2:19:04<8:04:17,  2.03s/it]
 25%|██▍       | 4750/19090 [2:19:06<7:11:16,  1.80s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 0.6334925293922424, 'learning_rate': 4.670414803278397e-05, 'recon_loss': 0.004102414811495691, 'recon_static': 0.003148035041522235, 'recon_motion': 0.0009543797496007755, 'recon_norm': 2.223742699623108, 'recon_norm_std': 0.07956387232989073, 'target_norm': 1.618587499856949, 'recon_cos': 0.4142663426697254, 'gradnorm/compressor.query': 0.011018855566730209, 'gradnorm/compressor.layer0': 0.18087499522240139, 'gradnorm/compressor.layer1': 0.2577787579506042, 'gradnorm/compressor.layer2': 0.21072025945101036, 'gradnorm/compressor.layer3': 0.1943119264270885, 'gradnorm/compressor.layer4': 0.16567837736729077, 'gradnorm/compressor.layer5': 0.15422271201659135, 'gradnorm/compressor.layer6': 0.13373386330277684, 'gradnorm/compressor.layer7': 0.1462348911538376, 'gradnorm/decoder.layer0': 0.19437460778187618, 'gradnorm/decoder.layer1': 0.18957520513786294, 'gradnorm/decoder.layer2': 0.23588554460793398, 'gradnorm/max_param': 0.16638317704200745, 'epoch': 2.49}

 25%|██▍       | 4750/19090 [2:19:06<7:11:16,  1.80s/it]
 25%|██▍       | 4751/19090 [2:19:07<6:33:55,  1.65s/it]
 25%|██▍       | 4752/19090 [2:19:09<7:29:42,  1.88s/it]
 25%|██▍       | 4753/19090 [2:19:11<6:46:39,  1.70s/it]
 25%|██▍       | 4754/19090 [2:19:12<6:16:43,  1.58s/it]
 25%|██▍       | 4755/19090 [2:19:14<7:27:07,  1.87s/it]
 25%|██▍       | 4756/19090 [2:19:16<6:44:48,  1.69s/it]
 25%|██▍       | 4757/19090 [2:19:17<6:15:09,  1.57s/it]
 25%|██▍       | 4758/19090 [2:19:19<7:18:30,  1.84s/it]
 25%|██▍       | 4759/19090 [2:19:21<6:38:35,  1.67s/it]
 25%|██▍       | 4760/19090 [2:19:22<6:11:12,  1.55s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.6379106044769287, 'learning_rate': 4.6681425495889034e-05, 'recon_loss': 0.004457480646669865, 'recon_static': 0.0035111613164190204, 'recon_motion': 0.0009463193055125885, 'recon_norm': 2.3200827717781065, 'recon_norm_std': 0.07412985675036907, 'target_norm': 1.618619579076767, 'recon_cos': 0.39981279596686364, 'gradnorm/compressor.query': 0.011267842813172684, 'gradnorm/compressor.layer0': 0.18184046877942306, 'gradnorm/compressor.layer1': 0.2661074016516468, 'gradnorm/compressor.layer2': 0.21834862022324286, 'gradnorm/compressor.layer3': 0.2009509734832076, 'gradnorm/compressor.layer4': 0.16520434911641133, 'gradnorm/compressor.layer5': 0.15211255920047564, 'gradnorm/compressor.layer6': 0.13188209432123124, 'gradnorm/compressor.layer7': 0.14511506629718168, 'gradnorm/decoder.layer0': 0.1894338637323616, 'gradnorm/decoder.layer1': 0.18958276069121457, 'gradnorm/decoder.layer2': 0.23271798256995604, 'gradnorm/max_param': 0.1656738817691803, 'epoch': 2.49}

 25%|██▍       | 4760/19090 [2:19:22<6:11:12,  1.55s/it]
 25%|██▍       | 4761/19090 [2:19:25<8:03:46,  2.03s/it]
 25%|██▍       | 4762/19090 [2:19:26<7:10:29,  1.80s/it]
 25%|██▍       | 4763/19090 [2:19:28<6:32:57,  1.65s/it]
 25%|██▍       | 4764/19090 [2:19:30<7:39:30,  1.92s/it]
 25%|██▍       | 4765/19090 [2:19:31<6:53:19,  1.73s/it]
 25%|██▍       | 4766/19090 [2:19:33<6:21:07,  1.60s/it]
 25%|██▍       | 4767/19090 [2:19:36<7:47:18,  1.96s/it]
 25%|██▍       | 4768/19090 [2:19:37<6:58:52,  1.75s/it]
 25%|██▍       | 4769/19090 [2:19:38<6:24:57,  1.61s/it]
 25%|██▍       | 4770/19090 [2:19:41<7:30:16,  1.89s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.6264873743057251, 'learning_rate': 4.665863046691073e-05, 'recon_loss': 0.004462011787109077, 'recon_static': 0.0035199865233153105, 'recon_motion': 0.0009420253292773851, 'recon_norm': 2.319826918840408, 'recon_norm_std': 0.07261080695316195, 'target_norm': 1.6181469857692719, 'recon_cos': 0.3987536758184433, 'gradnorm/compressor.query': 0.0098447400123401, 'gradnorm/compressor.layer0': 0.17345187905134446, 'gradnorm/compressor.layer1': 0.25665730983896634, 'gradnorm/compressor.layer2': 0.21336088824634722, 'gradnorm/compressor.layer3': 0.1974309190973835, 'gradnorm/compressor.layer4': 0.16646995645790896, 'gradnorm/compressor.layer5': 0.14915739715489892, 'gradnorm/compressor.layer6': 0.13289445564315186, 'gradnorm/compressor.layer7': 0.1422791416255812, 'gradnorm/decoder.layer0': 0.18853270727195393, 'gradnorm/decoder.layer1': 0.1891348909078467, 'gradnorm/decoder.layer2': 0.2296693807708236, 'gradnorm/max_param': 0.16206927597522736, 'epoch': 2.5}

 25%|██▍       | 4770/19090 [2:19:41<7:30:16,  1.89s/it]
 25%|██▍       | 4771/19090 [2:19:42<6:47:09,  1.71s/it]
 25%|██▍       | 4772/19090 [2:19:43<6:16:37,  1.58s/it]
 25%|██▌       | 4773/19090 [2:19:46<7:49:17,  1.97s/it]
 25%|██▌       | 4774/19090 [2:19:47<7:00:10,  1.76s/it]
 25%|██▌       | 4775/19090 [2:19:49<6:25:53,  1.62s/it]
 25%|██▌       | 4776/19090 [2:19:51<7:48:47,  1.97s/it]
 25%|██▌       | 4777/19090 [2:19:53<7:00:23,  1.76s/it]
 25%|██▌       | 4778/19090 [2:19:54<6:26:46,  1.62s/it]
 25%|██▌       | 4779/19090 [2:19:57<7:47:50,  1.96s/it]
 25%|██▌       | 4780/19090 [2:19:58<6:59:54,  1.76s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.554523766040802, 'learning_rate': 4.66357630220645e-05, 'recon_loss': 0.00422329040011391, 'recon_static': 0.003265661740442738, 'recon_motion': 0.0009576286291121505, 'recon_norm': 2.251953476667404, 'recon_norm_std': 0.07361970115453005, 'target_norm': 1.6204699516296386, 'recon_cos': 0.4072064869105816, 'gradnorm/compressor.query': 0.008466277354739567, 'gradnorm/compressor.layer0': 0.15472341082107066, 'gradnorm/compressor.layer1': 0.23374958836422338, 'gradnorm/compressor.layer2': 0.18530551873126844, 'gradnorm/compressor.layer3': 0.173967503807534, 'gradnorm/compressor.layer4': 0.14109481373889707, 'gradnorm/compressor.layer5': 0.1312719289875403, 'gradnorm/compressor.layer6': 0.11656858802242355, 'gradnorm/compressor.layer7': 0.12517713870108432, 'gradnorm/decoder.layer0': 0.1674035565798892, 'gradnorm/decoder.layer1': 0.16675023067714342, 'gradnorm/decoder.layer2': 0.2049796640090448, 'gradnorm/max_param': 0.1449037790298462, 'epoch': 2.5}

 25%|██▌       | 4780/19090 [2:19:58<6:59:54,  1.76s/it]
 25%|██▌       | 4781/19090 [2:19:59<6:26:07,  1.62s/it]
 25%|██▌       | 4782/19090 [2:20:02<7:31:01,  1.89s/it]
 25%|██▌       | 4783/19090 [2:20:03<6:47:17,  1.71s/it]
 25%|██▌       | 4784/19090 [2:20:04<6:17:05,  1.58s/it]
 25%|██▌       | 4785/19090 [2:20:07<7:45:41,  1.95s/it]
 25%|██▌       | 4786/19090 [2:20:09<6:57:44,  1.75s/it]
 25%|██▌       | 4787/19090 [2:20:10<6:24:00,  1.61s/it]
 25%|██▌       | 4788/19090 [2:20:13<7:46:55,  1.96s/it]
 25%|██▌       | 4789/19090 [2:20:14<6:58:46,  1.76s/it]
 25%|██▌       | 4790/19090 [2:20:15<6:24:57,  1.62s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.4983430802822113, 'learning_rate': 4.661282323780787e-05, 'recon_loss': 0.003169293125392869, 'recon_static': 0.00221996150794439, 'recon_motion': 0.0009493316611042246, 'recon_norm': 1.9667511343955995, 'recon_norm_std': 0.08081372920423746, 'target_norm': 1.6191579401493073, 'recon_cos': 0.46744219958782196, 'gradnorm/compressor.query': 0.007919075857636034, 'gradnorm/compressor.layer0': 0.1386461838577601, 'gradnorm/compressor.layer1': 0.2077566878582737, 'gradnorm/compressor.layer2': 0.17635830363629806, 'gradnorm/compressor.layer3': 0.16091938333483397, 'gradnorm/compressor.layer4': 0.1289730188094471, 'gradnorm/compressor.layer5': 0.11779221041913938, 'gradnorm/compressor.layer6': 0.10055605898293704, 'gradnorm/compressor.layer7': 0.11098769909154188, 'gradnorm/decoder.layer0': 0.14698444205138936, 'gradnorm/decoder.layer1': 0.14853980278457052, 'gradnorm/decoder.layer2': 0.1797021880573604, 'gradnorm/max_param': 0.12707559764385223, 'epoch': 2.51}

 25%|██▌       | 4790/19090 [2:20:15<6:24:57,  1.62s/it]
 25%|██▌       | 4791/19090 [2:20:18<7:38:11,  1.92s/it]
 25%|██▌       | 4792/19090 [2:20:19<6:52:32,  1.73s/it]
 25%|██▌       | 4793/19090 [2:20:20<6:20:34,  1.60s/it]
 25%|██▌       | 4794/19090 [2:20:23<7:32:37,  1.90s/it]
 25%|██▌       | 4795/19090 [2:20:24<6:48:39,  1.72s/it]
 25%|██▌       | 4796/19090 [2:20:26<6:17:41,  1.59s/it]
 25%|██▌       | 4797/19090 [2:20:28<7:55:12,  1.99s/it]
 25%|██▌       | 4798/19090 [2:20:30<7:04:30,  1.78s/it]
 25%|██▌       | 4799/19090 [2:20:31<6:28:58,  1.63s/it]
 25%|██▌       | 4800/19090 [2:20:34<8:00:01,  2.02s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.5772759914398193, 'learning_rate': 4.658981119084028e-05, 'recon_loss': 0.0034988651110325008, 'recon_static': 0.002545802417444065, 'recon_motion': 0.0009530627969070338, 'recon_norm': 2.063894733786583, 'recon_norm_std': 0.07424878459423781, 'target_norm': 1.6200262129306793, 'recon_cos': 0.44686908051371577, 'gradnorm/compressor.query': 0.008836428593131688, 'gradnorm/compressor.layer0': 0.15778580123425404, 'gradnorm/compressor.layer1': 0.24142238291439544, 'gradnorm/compressor.layer2': 0.201922700682052, 'gradnorm/compressor.layer3': 0.181776641864022, 'gradnorm/compressor.layer4': 0.1528148519592428, 'gradnorm/compressor.layer5': 0.1380411431526396, 'gradnorm/compressor.layer6': 0.12101178702492237, 'gradnorm/compressor.layer7': 0.13263230862600145, 'gradnorm/decoder.layer0': 0.17186586940961707, 'gradnorm/decoder.layer1': 0.1702785202128894, 'gradnorm/decoder.layer2': 0.20735778788382542, 'gradnorm/max_param': 0.14702363312244415, 'epoch': 2.51}

 25%|██▌       | 4800/19090 [2:20:34<8:00:01,  2.02s/it]
 25%|██▌       | 4801/19090 [2:20:35<7:07:49,  1.80s/it]
 25%|██▌       | 4802/19090 [2:20:37<6:31:34,  1.64s/it]
 25%|██▌       | 4803/19090 [2:20:39<7:45:39,  1.96s/it]
 25%|██▌       | 4804/19090 [2:20:41<6:57:47,  1.75s/it]
 25%|██▌       | 4805/19090 [2:20:42<6:24:15,  1.61s/it]
 25%|██▌       | 4806/19090 [2:20:44<7:26:51,  1.88s/it]
 25%|██▌       | 4807/19090 [2:20:46<6:44:49,  1.70s/it]
 25%|██▌       | 4808/19090 [2:20:47<6:15:13,  1.58s/it]
 25%|██▌       | 4809/19090 [2:20:49<7:13:01,  1.82s/it]
 25%|██▌       | 4810/19090 [2:20:51<6:34:54,  1.66s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.45712393522262573, 'learning_rate': 4.656672695810276e-05, 'recon_loss': 0.0033372902835253626, 'recon_static': 0.0023904127825517207, 'recon_motion': 0.0009468776173889637, 'recon_norm': 2.016960659623146, 'recon_norm_std': 0.07609328152611852, 'target_norm': 1.6207034200429917, 'recon_cos': 0.45974729433655737, 'gradnorm/compressor.query': 0.006635108029566098, 'gradnorm/compressor.layer0': 0.12396444960405753, 'gradnorm/compressor.layer1': 0.1959105871051802, 'gradnorm/compressor.layer2': 0.16034039517730983, 'gradnorm/compressor.layer3': 0.14601027378381815, 'gradnorm/compressor.layer4': 0.1174691075058003, 'gradnorm/compressor.layer5': 0.1064617249321211, 'gradnorm/compressor.layer6': 0.09252020255214292, 'gradnorm/compressor.layer7': 0.10014560760456472, 'gradnorm/decoder.layer0': 0.13676920368572745, 'gradnorm/decoder.layer1': 0.1403600063592172, 'gradnorm/decoder.layer2': 0.16128575551436927, 'gradnorm/max_param': 0.11468765884637833, 'epoch': 2.52}

 25%|██▌       | 4810/19090 [2:20:51<6:34:54,  1.66s/it]
 25%|██▌       | 4811/19090 [2:20:52<6:08:09,  1.55s/it]
 25%|██▌       | 4812/19090 [2:20:54<7:04:08,  1.78s/it]
 25%|██▌       | 4813/19090 [2:20:55<6:28:39,  1.63s/it]
 25%|██▌       | 4814/19090 [2:20:57<6:03:48,  1.53s/it]
 25%|██▌       | 4815/19090 [2:20:59<6:55:33,  1.75s/it]
 25%|██▌       | 4816/19090 [2:21:00<6:22:40,  1.61s/it]
 25%|██▌       | 4817/19090 [2:21:02<5:59:48,  1.51s/it]
 25%|██▌       | 4818/19090 [2:21:03<6:00:22,  1.52s/it]
 25%|██▌       | 4819/19090 [2:21:04<5:43:58,  1.45s/it]
 25%|██▌       | 4820/19090 [2:21:06<5:33:43,  1.40s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.5072214007377625, 'learning_rate': 4.6543570616777677e-05, 'recon_loss': 0.003090219380101189, 'recon_static': 0.0021392837923485787, 'recon_motion': 0.0009509355426416732, 'recon_norm': 1.9433802932500839, 'recon_norm_std': 0.07507694195955991, 'target_norm': 1.6203561872243881, 'recon_cos': 0.47244684472680093, 'gradnorm/compressor.query': 0.007629811275544252, 'gradnorm/compressor.layer0': 0.14010476867967145, 'gradnorm/compressor.layer1': 0.22101186577198004, 'gradnorm/compressor.layer2': 0.1808534505874835, 'gradnorm/compressor.layer3': 0.1585976250179793, 'gradnorm/compressor.layer4': 0.1307967662494113, 'gradnorm/compressor.layer5': 0.1195320087296584, 'gradnorm/compressor.layer6': 0.10093320232873366, 'gradnorm/compressor.layer7': 0.11313071607932955, 'gradnorm/decoder.layer0': 0.14903693506752388, 'gradnorm/decoder.layer1': 0.14920523018030452, 'gradnorm/decoder.layer2': 0.1789486247690655, 'gradnorm/max_param': 0.1271415650844574, 'epoch': 2.52}

 25%|██▌       | 4820/19090 [2:21:06<5:33:43,  1.40s/it]
 25%|██▌       | 4821/19090 [2:21:08<6:22:01,  1.61s/it]
 25%|██▌       | 4822/19090 [2:21:09<5:59:46,  1.51s/it]
 25%|██▌       | 4823/19090 [2:21:10<5:44:16,  1.45s/it]
 25%|██▌       | 4824/19090 [2:21:13<6:39:28,  1.68s/it]
 25%|██▌       | 4825/19090 [2:21:14<6:11:36,  1.56s/it]
 25%|██▌       | 4826/19090 [2:21:15<5:51:58,  1.48s/it]
 25%|██▌       | 4827/19090 [2:21:18<6:59:48,  1.77s/it]
 25%|██▌       | 4828/19090 [2:21:19<6:26:05,  1.62s/it]
 25%|██▌       | 4829/19090 [2:21:20<6:02:08,  1.52s/it]
 25%|██▌       | 4830/19090 [2:21:22<6:49:35,  1.72s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.623028576374054, 'learning_rate': 4.652034224428853e-05, 'recon_loss': 0.00423111038398929, 'recon_static': 0.0032680776668712495, 'recon_motion': 0.0009630327112972736, 'recon_norm': 2.2464152067899703, 'recon_norm_std': 0.06702874936163425, 'target_norm': 1.6206677317619325, 'recon_cos': 0.4096638783812523, 'gradnorm/compressor.query': 0.009662511834650153, 'gradnorm/compressor.layer0': 0.1772820197232766, 'gradnorm/compressor.layer1': 0.2613869456877611, 'gradnorm/compressor.layer2': 0.21473488216788536, 'gradnorm/compressor.layer3': 0.19214812356078295, 'gradnorm/compressor.layer4': 0.15882797820069922, 'gradnorm/compressor.layer5': 0.1508336848170135, 'gradnorm/compressor.layer6': 0.13007303520073613, 'gradnorm/compressor.layer7': 0.1438086786442156, 'gradnorm/decoder.layer0': 0.18411015869895855, 'gradnorm/decoder.layer1': 0.18579638920756497, 'gradnorm/decoder.layer2': 0.22648040336397796, 'gradnorm/max_param': 0.15927337110042572, 'epoch': 2.53}

 25%|██▌       | 4830/19090 [2:21:22<6:49:35,  1.72s/it]
 25%|██▌       | 4831/19090 [2:21:24<6:19:24,  1.60s/it]
 25%|██▌       | 4832/19090 [2:21:25<5:57:21,  1.50s/it]
 25%|██▌       | 4833/19090 [2:21:27<6:57:42,  1.76s/it]
 25%|██▌       | 4834/19090 [2:21:29<6:23:53,  1.62s/it]
 25%|██▌       | 4835/19090 [2:21:30<6:00:36,  1.52s/it]
 25%|██▌       | 4836/19090 [2:21:32<7:03:56,  1.78s/it]
 25%|██▌       | 4837/19090 [2:21:34<6:28:16,  1.63s/it]
 25%|██▌       | 4838/19090 [2:21:35<6:03:23,  1.53s/it]
 25%|██▌       | 4839/19090 [2:21:37<6:56:58,  1.76s/it]
 25%|██▌       | 4840/19090 [2:21:38<6:24:29,  1.62s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.6061400175094604, 'learning_rate': 4.649704191829962e-05, 'recon_loss': 0.004160442808642984, 'recon_static': 0.003196206991560757, 'recon_motion': 0.0009642358578275889, 'recon_norm': 2.237585175037384, 'recon_norm_std': 0.06673889774829149, 'target_norm': 1.621567714214325, 'recon_cos': 0.4105090342462063, 'gradnorm/compressor.query': 0.009745070686496023, 'gradnorm/compressor.layer0': 0.1704637230132773, 'gradnorm/compressor.layer1': 0.2520576969751336, 'gradnorm/compressor.layer2': 0.21403703955830436, 'gradnorm/compressor.layer3': 0.19528322954192462, 'gradnorm/compressor.layer4': 0.15874832306374162, 'gradnorm/compressor.layer5': 0.1432888527977785, 'gradnorm/compressor.layer6': 0.12451726863684343, 'gradnorm/compressor.layer7': 0.13675892784494112, 'gradnorm/decoder.layer0': 0.1770911472129961, 'gradnorm/decoder.layer1': 0.1774526291249387, 'gradnorm/decoder.layer2': 0.21904663001243452, 'gradnorm/max_param': 0.1549905687570572, 'epoch': 2.54}

 25%|██▌       | 4840/19090 [2:21:38<6:24:29,  1.62s/it]
 25%|██▌       | 4841/19090 [2:21:40<6:00:46,  1.52s/it]
 25%|██▌       | 4842/19090 [2:21:42<6:54:17,  1.74s/it]
 25%|██▌       | 4843/19090 [2:21:43<6:21:55,  1.61s/it]
 25%|██▌       | 4844/19090 [2:21:45<5:58:35,  1.51s/it]
 25%|██▌       | 4845/19090 [2:21:47<7:04:26,  1.79s/it]
 25%|██▌       | 4846/19090 [2:21:48<6:29:24,  1.64s/it]
 25%|██▌       | 4847/19090 [2:21:50<6:05:00,  1.54s/it]
 25%|██▌       | 4848/19090 [2:21:52<6:56:11,  1.75s/it]
 25%|██▌       | 4849/19090 [2:21:53<6:22:55,  1.61s/it]
 25%|██▌       | 4850/19090 [2:21:54<5:59:49,  1.52s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.6938854455947876, 'learning_rate': 4.647366971671586e-05, 'recon_loss': 0.004611982154892757, 'recon_static': 0.003648276999592781, 'recon_motion': 0.0009637051276513376, 'recon_norm': 2.34583665728569, 'recon_norm_std': 0.06322462931275367, 'target_norm': 1.6210949957370757, 'recon_cos': 0.3906924448907375, 'gradnorm/compressor.query': 0.010464111290010068, 'gradnorm/compressor.layer0': 0.19499675978817335, 'gradnorm/compressor.layer1': 0.28778409500861524, 'gradnorm/compressor.layer2': 0.23739637950339487, 'gradnorm/compressor.layer3': 0.21754744054820394, 'gradnorm/compressor.layer4': 0.18236689831685812, 'gradnorm/compressor.layer5': 0.1668171774235935, 'gradnorm/compressor.layer6': 0.14399892489252156, 'gradnorm/compressor.layer7': 0.16037943454916223, 'gradnorm/decoder.layer0': 0.20957927169207388, 'gradnorm/decoder.layer1': 0.2059034542832421, 'gradnorm/decoder.layer2': 0.25095337638179727, 'gradnorm/max_param': 0.1766217201948166, 'epoch': 2.54}

 25%|██▌       | 4850/19090 [2:21:54<5:59:49,  1.52s/it]
 25%|██▌       | 4851/19090 [2:21:57<7:20:23,  1.86s/it]
 25%|██▌       | 4852/19090 [2:21:58<6:40:02,  1.69s/it]
 25%|██▌       | 4853/19090 [2:22:00<6:11:29,  1.57s/it]
 25%|██▌       | 4854/19090 [2:22:02<7:19:02,  1.85s/it]
 25%|██▌       | 4855/19090 [2:22:03<6:38:58,  1.68s/it]
 25%|██▌       | 4856/19090 [2:22:05<6:11:00,  1.56s/it]
 25%|██▌       | 4857/19090 [2:22:07<7:31:52,  1.90s/it]
 25%|██▌       | 4858/19090 [2:22:09<6:47:40,  1.72s/it]
 25%|██▌       | 4859/19090 [2:22:10<6:16:23,  1.59s/it]
 25%|██▌       | 4860/19090 [2:22:13<7:37:27,  1.93s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 0.5874965786933899, 'learning_rate': 4.645022571768245e-05, 'recon_loss': 0.0040576380037236955, 'recon_static': 0.003102025354746729, 'recon_motion': 0.0009556125805829652, 'recon_norm': 2.211099123954773, 'recon_norm_std': 0.066477582603693, 'target_norm': 1.6204570442438126, 'recon_cos': 0.4161852538585663, 'gradnorm/compressor.query': 0.009184818675771131, 'gradnorm/compressor.layer0': 0.16342179897666567, 'gradnorm/compressor.layer1': 0.2475555163912642, 'gradnorm/compressor.layer2': 0.2051836741262446, 'gradnorm/compressor.layer3': 0.18458613575113286, 'gradnorm/compressor.layer4': 0.1531775775969762, 'gradnorm/compressor.layer5': 0.13758457411338068, 'gradnorm/compressor.layer6': 0.12188631304784119, 'gradnorm/compressor.layer7': 0.13352141792404298, 'gradnorm/decoder.layer0': 0.1747756912647481, 'gradnorm/decoder.layer1': 0.17426685622758625, 'gradnorm/decoder.layer2': 0.21192564690916313, 'gradnorm/max_param': 0.14989785850048065, 'epoch': 2.55}

 25%|██▌       | 4860/19090 [2:22:13<7:37:27,  1.93s/it]
 25%|██▌       | 4861/19090 [2:22:14<6:51:36,  1.74s/it]
 25%|██▌       | 4862/19090 [2:22:15<6:19:16,  1.60s/it]
 25%|██▌       | 4863/19090 [2:22:18<7:29:50,  1.90s/it]
 25%|██▌       | 4864/19090 [2:22:19<6:46:21,  1.71s/it]
 25%|██▌       | 4865/19090 [2:22:20<6:15:43,  1.58s/it]
 25%|██▌       | 4866/19090 [2:22:23<7:04:51,  1.79s/it]
 25%|██▌       | 4867/19090 [2:22:24<6:28:51,  1.64s/it]
 26%|██▌       | 4868/19090 [2:22:25<6:03:37,  1.53s/it]
 26%|██▌       | 4869/19090 [2:22:28<7:29:26,  1.90s/it]
 26%|██▌       | 4870/19090 [2:22:29<6:46:14,  1.71s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.5653154253959656, 'learning_rate': 4.6426709999584664e-05, 'recon_loss': 0.003823116404237226, 'recon_static': 0.002880201587686315, 'recon_motion': 0.0009429148311028257, 'recon_norm': 2.152935838699341, 'recon_norm_std': 0.06829805122688412, 'target_norm': 1.6179765105247497, 'recon_cos': 0.4278628401458263, 'gradnorm/compressor.query': 0.008075443816493041, 'gradnorm/compressor.layer0': 0.1595964242308892, 'gradnorm/compressor.layer1': 0.24101395227726635, 'gradnorm/compressor.layer2': 0.19718553786956503, 'gradnorm/compressor.layer3': 0.17797495729842694, 'gradnorm/compressor.layer4': 0.14821286520876356, 'gradnorm/compressor.layer5': 0.13582301306377978, 'gradnorm/compressor.layer6': 0.11222351291813776, 'gradnorm/compressor.layer7': 0.1261125835667028, 'gradnorm/decoder.layer0': 0.16829382135760398, 'gradnorm/decoder.layer1': 0.1630832077985731, 'gradnorm/decoder.layer2': 0.20388216336808368, 'gradnorm/max_param': 0.14474862813949585, 'epoch': 2.55}

 26%|██▌       | 4870/19090 [2:22:29<6:46:14,  1.71s/it]
 26%|██▌       | 4871/19090 [2:22:31<6:15:34,  1.58s/it]
 26%|██▌       | 4872/19090 [2:22:33<6:56:29,  1.76s/it]
 26%|██▌       | 4873/19090 [2:22:34<6:22:40,  1.62s/it]
 26%|██▌       | 4874/19090 [2:22:35<5:59:07,  1.52s/it]
 26%|██▌       | 4875/19090 [2:22:38<7:05:24,  1.80s/it]
 26%|██▌       | 4876/19090 [2:22:39<6:28:49,  1.64s/it]
 26%|██▌       | 4877/19090 [2:22:40<6:03:27,  1.53s/it]
 26%|██▌       | 4878/19090 [2:22:43<6:56:38,  1.76s/it]
 26%|██▌       | 4879/19090 [2:22:44<6:22:37,  1.62s/it]
 26%|██▌       | 4880/19090 [2:22:45<5:59:11,  1.52s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.4719265103340149, 'learning_rate': 4.640312264104756e-05, 'recon_loss': 0.003150773042580113, 'recon_static': 0.0021922887215623634, 'recon_motion': 0.0009584843937773257, 'recon_norm': 1.9578895092010498, 'recon_norm_std': 0.07219840548932552, 'target_norm': 1.6209359169006348, 'recon_cos': 0.46857959777116776, 'gradnorm/compressor.query': 0.006599323734964281, 'gradnorm/compressor.layer0': 0.12931920144050854, 'gradnorm/compressor.layer1': 0.20305251698950016, 'gradnorm/compressor.layer2': 0.1715150879487231, 'gradnorm/compressor.layer3': 0.1504238513061057, 'gradnorm/compressor.layer4': 0.11923585665875377, 'gradnorm/compressor.layer5': 0.10918710099984226, 'gradnorm/compressor.layer6': 0.09310546487688585, 'gradnorm/compressor.layer7': 0.10396213767331523, 'gradnorm/decoder.layer0': 0.14288032851291008, 'gradnorm/decoder.layer1': 0.13964372907050301, 'gradnorm/decoder.layer2': 0.16474707579984268, 'gradnorm/max_param': 0.11700813472270966, 'epoch': 2.56}

 26%|██▌       | 4880/19090 [2:22:45<5:59:11,  1.52s/it]
 26%|██▌       | 4881/19090 [2:22:48<7:13:04,  1.83s/it]
 26%|██▌       | 4882/19090 [2:22:49<6:34:09,  1.66s/it]
 26%|██▌       | 4883/19090 [2:22:50<6:07:33,  1.55s/it]
 26%|██▌       | 4884/19090 [2:22:53<6:58:29,  1.77s/it]
 26%|██▌       | 4885/19090 [2:22:54<6:23:50,  1.62s/it]
 26%|██▌       | 4886/19090 [2:22:55<5:59:49,  1.52s/it]
 26%|██▌       | 4887/19090 [2:22:57<6:58:11,  1.77s/it]
 26%|██▌       | 4888/19090 [2:22:59<6:23:33,  1.62s/it]
 26%|██▌       | 4889/19090 [2:23:00<5:59:44,  1.52s/it]
 26%|██▌       | 4890/19090 [2:23:02<6:58:13,  1.77s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.5017119646072388, 'learning_rate': 4.6379463720935726e-05, 'recon_loss': 0.003000779333524406, 'recon_static': 0.0020509548048721626, 'recon_motion': 0.0009498245330178179, 'recon_norm': 1.917264637351036, 'recon_norm_std': 0.07326493989676237, 'target_norm': 1.6188567757606507, 'recon_cos': 0.47908162996172904, 'gradnorm/compressor.query': 0.007687549676222184, 'gradnorm/compressor.layer0': 0.13613563572981302, 'gradnorm/compressor.layer1': 0.21631698075475544, 'gradnorm/compressor.layer2': 0.18197313248671412, 'gradnorm/compressor.layer3': 0.16328286941309328, 'gradnorm/compressor.layer4': 0.12536057976813375, 'gradnorm/compressor.layer5': 0.11661521330117082, 'gradnorm/compressor.layer6': 0.10029380003111975, 'gradnorm/compressor.layer7': 0.11100472357806418, 'gradnorm/decoder.layer0': 0.1489293270001663, 'gradnorm/decoder.layer1': 0.14765671497857083, 'gradnorm/decoder.layer2': 0.17571744820582205, 'gradnorm/max_param': 0.12432632595300674, 'epoch': 2.56}

 26%|██▌       | 4890/19090 [2:23:02<6:58:13,  1.77s/it]
 26%|██▌       | 4891/19090 [2:23:04<6:23:52,  1.62s/it]
 26%|██▌       | 4892/19090 [2:23:05<5:59:44,  1.52s/it]
 26%|██▌       | 4893/19090 [2:23:07<6:42:33,  1.70s/it]
 26%|██▌       | 4894/19090 [2:23:08<6:12:28,  1.57s/it]
 26%|██▌       | 4895/19090 [2:23:10<5:51:32,  1.49s/it]
 26%|██▌       | 4896/19090 [2:23:13<7:37:00,  1.93s/it]
 26%|██▌       | 4897/19090 [2:23:14<6:50:54,  1.74s/it]
 26%|██▌       | 4898/19090 [2:23:15<6:19:14,  1.60s/it]
 26%|██▌       | 4899/19090 [2:23:18<7:43:17,  1.96s/it]
 26%|██▌       | 4900/19090 [2:23:19<6:55:15,  1.76s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.6801241040229797, 'learning_rate': 4.635573331835302e-05, 'recon_loss': 0.003688898927066475, 'recon_static': 0.0027388479735236614, 'recon_motion': 0.0009500510466750711, 'recon_norm': 2.1093966722488404, 'recon_norm_std': 0.0667529758065939, 'target_norm': 1.6180746108293533, 'recon_cos': 0.4371460862457752, 'gradnorm/compressor.query': 0.010658708561757605, 'gradnorm/compressor.layer0': 0.1923705837797536, 'gradnorm/compressor.layer1': 0.29069839699233985, 'gradnorm/compressor.layer2': 0.2355368110695655, 'gradnorm/compressor.layer3': 0.2209521466975307, 'gradnorm/compressor.layer4': 0.1764913679411811, 'gradnorm/compressor.layer5': 0.16272926731856976, 'gradnorm/compressor.layer6': 0.13931911157302104, 'gradnorm/compressor.layer7': 0.15406954593486125, 'gradnorm/decoder.layer0': 0.19935059392064175, 'gradnorm/decoder.layer1': 0.1978758447777264, 'gradnorm/decoder.layer2': 0.23869001784723445, 'gradnorm/max_param': 0.17554010450839996, 'epoch': 2.57}

 26%|██▌       | 4900/19090 [2:23:19<6:55:15,  1.76s/it]
 26%|██▌       | 4901/19090 [2:23:21<6:21:51,  1.61s/it]
 26%|██▌       | 4902/19090 [2:23:23<7:47:28,  1.98s/it]
 26%|██▌       | 4903/19090 [2:23:25<6:57:58,  1.77s/it]
 26%|██▌       | 4904/19090 [2:23:26<6:23:29,  1.62s/it]
 26%|██▌       | 4905/19090 [2:23:29<7:48:10,  1.98s/it]
 26%|██▌       | 4906/19090 [2:23:30<6:58:43,  1.77s/it]
 26%|██▌       | 4907/19090 [2:23:31<6:24:44,  1.63s/it]
 26%|██▌       | 4908/19090 [2:23:34<7:46:49,  1.98s/it]
 26%|██▌       | 4909/19090 [2:23:35<6:58:08,  1.77s/it]
 26%|██▌       | 4910/19090 [2:23:37<6:24:04,  1.63s/it]
                                                        
{'loss': 0.005, 'grad_norm': 0.7159276604652405, 'learning_rate': 4.633193151264231e-05, 'recon_loss': 0.005000222765374929, 'recon_static': 0.004045409988611937, 'recon_motion': 0.000954812859708909, 'recon_norm': 2.443163686990738, 'recon_norm_std': 0.05953235765919089, 'target_norm': 1.619187143445015, 'recon_cos': 0.3748579822480679, 'gradnorm/compressor.query': 0.010767178495675388, 'gradnorm/compressor.layer0': 0.2053284343650095, 'gradnorm/compressor.layer1': 0.3080127050462338, 'gradnorm/compressor.layer2': 0.2515670568633963, 'gradnorm/compressor.layer3': 0.22247378083649574, 'gradnorm/compressor.layer4': 0.18845546189227094, 'gradnorm/compressor.layer5': 0.17000430398815056, 'gradnorm/compressor.layer6': 0.1468219174173684, 'gradnorm/compressor.layer7': 0.16093792544929045, 'gradnorm/decoder.layer0': 0.21308224479909213, 'gradnorm/decoder.layer1': 0.2046035998688449, 'gradnorm/decoder.layer2': 0.25195828513842117, 'gradnorm/max_param': 0.18440017104148865, 'epoch': 2.57}

 26%|██▌       | 4910/19090 [2:23:37<6:24:04,  1.63s/it]
 26%|██▌       | 4911/19090 [2:23:40<8:07:58,  2.06s/it]
 26%|██▌       | 4912/19090 [2:23:41<7:13:01,  1.83s/it]
 26%|██▌       | 4913/19090 [2:23:42<6:34:24,  1.67s/it]
 26%|██▌       | 4914/19090 [2:23:45<7:34:48,  1.92s/it]
 26%|██▌       | 4915/19090 [2:23:46<6:49:56,  1.74s/it]
 26%|██▌       | 4916/19090 [2:23:47<6:18:09,  1.60s/it]
 26%|██▌       | 4917/19090 [2:23:50<7:31:45,  1.91s/it]
 26%|██▌       | 4918/19090 [2:23:51<6:47:34,  1.73s/it]
 26%|██▌       | 4919/19090 [2:23:53<6:16:14,  1.59s/it]
 26%|██▌       | 4920/19090 [2:23:56<7:49:34,  1.99s/it]
                                                        
{'loss': 0.0055, 'grad_norm': 0.6863983869552612, 'learning_rate': 4.630805838338519e-05, 'recon_loss': 0.005526293162256479, 'recon_static': 0.004577780922409147, 'recon_motion': 0.0009485124246566556, 'recon_norm': 2.5655751883983613, 'recon_norm_std': 0.05882205590605736, 'target_norm': 1.6193776905536652, 'recon_cos': 0.3589874267578125, 'gradnorm/compressor.query': 0.010943641120007782, 'gradnorm/compressor.layer0': 0.19241028797670148, 'gradnorm/compressor.layer1': 0.29319980411087143, 'gradnorm/compressor.layer2': 0.2424717895822706, 'gradnorm/compressor.layer3': 0.22020464279332752, 'gradnorm/compressor.layer4': 0.1765686150141763, 'gradnorm/compressor.layer5': 0.16260808046393255, 'gradnorm/compressor.layer6': 0.13923938570455419, 'gradnorm/compressor.layer7': 0.15331120676590787, 'gradnorm/decoder.layer0': 0.20116639361274768, 'gradnorm/decoder.layer1': 0.20033038165162798, 'gradnorm/decoder.layer2': 0.244218195989643, 'gradnorm/max_param': 0.1734716147184372, 'epoch': 2.58}

 26%|██▌       | 4920/19090 [2:23:56<7:49:34,  1.99s/it]
 26%|██▌       | 4921/19090 [2:23:57<6:59:58,  1.78s/it]
 26%|██▌       | 4922/19090 [2:23:58<6:25:16,  1.63s/it]
 26%|██▌       | 4923/19090 [2:24:01<7:22:23,  1.87s/it]
 26%|██▌       | 4924/19090 [2:24:02<6:40:39,  1.70s/it]
 26%|██▌       | 4925/19090 [2:24:03<6:11:29,  1.57s/it]
 26%|██▌       | 4926/19090 [2:24:06<7:23:50,  1.88s/it]
 26%|██▌       | 4927/19090 [2:24:07<6:41:35,  1.70s/it]
 26%|██▌       | 4928/19090 [2:24:08<6:12:05,  1.58s/it]
 26%|██▌       | 4929/19090 [2:24:11<7:38:21,  1.94s/it]
 26%|██▌       | 4930/19090 [2:24:12<6:52:15,  1.75s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.5052264928817749, 'learning_rate': 4.628411401040173e-05, 'recon_loss': 0.003847013699123636, 'recon_static': 0.002891974290832877, 'recon_motion': 0.0009550394577672705, 'recon_norm': 2.1504815191030504, 'recon_norm_std': 0.06610191678628326, 'target_norm': 1.6205011367797852, 'recon_cos': 0.43047637268900873, 'gradnorm/compressor.query': 0.007525730393491311, 'gradnorm/compressor.layer0': 0.1428990410401013, 'gradnorm/compressor.layer1': 0.22333616880021315, 'gradnorm/compressor.layer2': 0.1817628285322348, 'gradnorm/compressor.layer3': 0.16497907168967746, 'gradnorm/compressor.layer4': 0.12672882071810568, 'gradnorm/compressor.layer5': 0.11239785600547475, 'gradnorm/compressor.layer6': 0.09984555947241633, 'gradnorm/compressor.layer7': 0.10926043948144402, 'gradnorm/decoder.layer0': 0.14905701155325035, 'gradnorm/decoder.layer1': 0.1466160436786764, 'gradnorm/decoder.layer2': 0.17414790646136355, 'gradnorm/max_param': 0.12892818450927734, 'epoch': 2.58}

 26%|██▌       | 4930/19090 [2:24:12<6:52:15,  1.75s/it]
 26%|██▌       | 4931/19090 [2:24:14<6:19:21,  1.61s/it]
 26%|██▌       | 4932/19090 [2:24:16<7:41:08,  1.95s/it]
 26%|██▌       | 4933/19090 [2:24:18<6:53:31,  1.75s/it]
 26%|██▌       | 4934/19090 [2:24:19<6:20:11,  1.61s/it]
 26%|██▌       | 4935/19090 [2:24:22<7:44:20,  1.97s/it]
 26%|██▌       | 4936/19090 [2:24:23<6:55:59,  1.76s/it]
 26%|██▌       | 4937/19090 [2:24:24<6:22:36,  1.62s/it]
 26%|██▌       | 4938/19090 [2:24:27<7:48:37,  1.99s/it]
 26%|██▌       | 4939/19090 [2:24:29<6:59:03,  1.78s/it]
 26%|██▌       | 4940/19090 [2:24:30<6:24:30,  1.63s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.5667673945426941, 'learning_rate': 4.62600984737502e-05, 'recon_loss': 0.0030738353845663367, 'recon_static': 0.002121896084281616, 'recon_motion': 0.0009519391998765059, 'recon_norm': 1.9367226392030716, 'recon_norm_std': 0.07126304684206844, 'target_norm': 1.6198981136083603, 'recon_cos': 0.4737855553627014, 'gradnorm/compressor.query': 0.008218139559108887, 'gradnorm/compressor.layer0': 0.16129148935537202, 'gradnorm/compressor.layer1': 0.2557890263880116, 'gradnorm/compressor.layer2': 0.20184960137612076, 'gradnorm/compressor.layer3': 0.1846961445245687, 'gradnorm/compressor.layer4': 0.14330581579246157, 'gradnorm/compressor.layer5': 0.1290468258611103, 'gradnorm/compressor.layer6': 0.11132739719353742, 'gradnorm/compressor.layer7': 0.12049592406182753, 'gradnorm/decoder.layer0': 0.16316345222948986, 'gradnorm/decoder.layer1': 0.16529195922918305, 'gradnorm/decoder.layer2': 0.19186556454300657, 'gradnorm/max_param': 0.14564168453216553, 'epoch': 2.59}

 26%|██▌       | 4940/19090 [2:24:30<6:24:30,  1.63s/it]
 26%|██▌       | 4941/19090 [2:24:32<7:20:37,  1.87s/it]
 26%|██▌       | 4942/19090 [2:24:34<6:39:44,  1.70s/it]
 26%|██▌       | 4943/19090 [2:24:35<6:10:55,  1.57s/it]
 26%|██▌       | 4944/19090 [2:24:37<7:12:27,  1.83s/it]
 26%|██▌       | 4945/19090 [2:24:39<6:34:01,  1.67s/it]
 26%|██▌       | 4946/19090 [2:24:40<6:07:12,  1.56s/it]
 26%|██▌       | 4947/19090 [2:24:42<7:19:54,  1.87s/it]
 26%|██▌       | 4948/19090 [2:24:44<6:39:03,  1.69s/it]
 26%|██▌       | 4949/19090 [2:24:45<6:10:45,  1.57s/it]
 26%|██▌       | 4950/19090 [2:24:48<7:33:52,  1.93s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.5050322413444519, 'learning_rate': 4.6236011853726804e-05, 'recon_loss': 0.0037699174543377014, 'recon_static': 0.002804362197639421, 'recon_motion': 0.0009655553076299839, 'recon_norm': 2.1318692952394485, 'recon_norm_std': 0.06560478322207927, 'target_norm': 1.622620490193367, 'recon_cos': 0.432260861992836, 'gradnorm/compressor.query': 0.0070853036780995785, 'gradnorm/compressor.layer0': 0.13549267182557512, 'gradnorm/compressor.layer1': 0.23500622148851202, 'gradnorm/compressor.layer2': 0.18601076461805877, 'gradnorm/compressor.layer3': 0.16478058364092463, 'gradnorm/compressor.layer4': 0.12534899718791123, 'gradnorm/compressor.layer5': 0.11345790785209349, 'gradnorm/compressor.layer6': 0.09806131279410892, 'gradnorm/compressor.layer7': 0.10744298814150963, 'gradnorm/decoder.layer0': 0.14274766065698294, 'gradnorm/decoder.layer1': 0.1438624009955292, 'gradnorm/decoder.layer2': 0.16985340492790166, 'gradnorm/max_param': 0.12198124080896378, 'epoch': 2.59}

 26%|██▌       | 4950/19090 [2:24:48<7:33:52,  1.93s/it]
 26%|██▌       | 4951/19090 [2:24:49<6:48:43,  1.73s/it]
 26%|██▌       | 4952/19090 [2:24:50<6:17:09,  1.60s/it]
 26%|██▌       | 4953/19090 [2:24:53<7:35:46,  1.93s/it]
 26%|██▌       | 4954/19090 [2:24:54<6:49:30,  1.74s/it]
 26%|██▌       | 4955/19090 [2:24:56<6:17:41,  1.60s/it]
 26%|██▌       | 4956/19090 [2:24:58<7:31:53,  1.92s/it]
 26%|██▌       | 4957/19090 [2:25:00<6:47:11,  1.73s/it]
 26%|██▌       | 4958/19090 [2:25:01<6:15:50,  1.60s/it]
 26%|██▌       | 4959/19090 [2:25:04<7:33:53,  1.93s/it]
 26%|██▌       | 4960/19090 [2:25:05<6:48:39,  1.74s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.5802269577980042, 'learning_rate': 4.6211854230865435e-05, 'recon_loss': 0.003235851530916989, 'recon_static': 0.002277931763092056, 'recon_motion': 0.0009579197838320396, 'recon_norm': 1.982075148820877, 'recon_norm_std': 0.07085107946768403, 'target_norm': 1.6229396224021913, 'recon_cos': 0.46202142611145974, 'gradnorm/compressor.query': 0.008734451444008556, 'gradnorm/compressor.layer0': 0.16648359538396548, 'gradnorm/compressor.layer1': 0.25827147348204643, 'gradnorm/compressor.layer2': 0.20725160244332566, 'gradnorm/compressor.layer3': 0.18646142338122718, 'gradnorm/compressor.layer4': 0.14665813948102793, 'gradnorm/compressor.layer5': 0.13382231862866498, 'gradnorm/compressor.layer6': 0.1146190020646591, 'gradnorm/compressor.layer7': 0.1286050717131281, 'gradnorm/decoder.layer0': 0.16944117900128638, 'gradnorm/decoder.layer1': 0.16416367181557096, 'gradnorm/decoder.layer2': 0.19922647274232666, 'gradnorm/max_param': 0.14932182431221008, 'epoch': 2.6}

 26%|██▌       | 4960/19090 [2:25:05<6:48:39,  1.74s/it]
 26%|██▌       | 4961/19090 [2:25:06<6:16:57,  1.60s/it]
 26%|██▌       | 4962/19090 [2:25:09<7:18:55,  1.86s/it]
 26%|██▌       | 4963/19090 [2:25:10<6:38:08,  1.69s/it]
 26%|██▌       | 4964/19090 [2:25:11<6:09:34,  1.57s/it]
 26%|██▌       | 4965/19090 [2:25:13<7:02:06,  1.79s/it]
 26%|██▌       | 4966/19090 [2:25:15<6:25:59,  1.64s/it]
 26%|██▌       | 4967/19090 [2:25:16<6:00:59,  1.53s/it]
 26%|██▌       | 4968/19090 [2:25:19<7:22:44,  1.88s/it]
 26%|██▌       | 4969/19090 [2:25:20<6:40:37,  1.70s/it]
 26%|██▌       | 4970/19090 [2:25:21<6:11:01,  1.58s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.605969250202179, 'learning_rate': 4.618762568593736e-05, 'recon_loss': 0.003783459367696196, 'recon_static': 0.0028267695917747913, 'recon_motion': 0.000956689850136172, 'recon_norm': 2.135248625278473, 'recon_norm_std': 0.06477730236947536, 'target_norm': 1.6193490356206894, 'recon_cos': 0.42769239768385886, 'gradnorm/compressor.query': 0.009829585531417698, 'gradnorm/compressor.layer0': 0.17458041951627898, 'gradnorm/compressor.layer1': 0.2772772265963192, 'gradnorm/compressor.layer2': 0.21725131853634141, 'gradnorm/compressor.layer3': 0.19312079960535944, 'gradnorm/compressor.layer4': 0.14935401258311246, 'gradnorm/compressor.layer5': 0.13797541461845303, 'gradnorm/compressor.layer6': 0.11933946042394013, 'gradnorm/compressor.layer7': 0.13122390391136757, 'gradnorm/decoder.layer0': 0.1752194752610982, 'gradnorm/decoder.layer1': 0.1716514564190689, 'gradnorm/decoder.layer2': 0.20561065560023326, 'gradnorm/max_param': 0.15665066242218018, 'epoch': 2.6}

 26%|██▌       | 4970/19090 [2:25:21<6:11:01,  1.58s/it]
 26%|██▌       | 4971/19090 [2:25:24<7:43:54,  1.97s/it]
 26%|██▌       | 4972/19090 [2:25:25<6:55:10,  1.76s/it]
 26%|██▌       | 4973/19090 [2:25:27<6:21:05,  1.62s/it]
 26%|██▌       | 4974/19090 [2:25:29<7:34:58,  1.93s/it]
 26%|██▌       | 4975/19090 [2:25:31<6:49:14,  1.74s/it]
 26%|██▌       | 4976/19090 [2:25:32<6:16:58,  1.60s/it]
 26%|██▌       | 4977/19090 [2:25:35<7:53:15,  2.01s/it]
 26%|██▌       | 4978/19090 [2:25:36<7:01:32,  1.79s/it]
 26%|██▌       | 4979/19090 [2:25:38<6:25:33,  1.64s/it]
 26%|██▌       | 4980/19090 [2:25:40<7:31:28,  1.92s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.6720647811889648, 'learning_rate': 4.616332629995098e-05, 'recon_loss': 0.0037261601944919677, 'recon_static': 0.0027794839697889984, 'recon_motion': 0.000946676179592032, 'recon_norm': 2.123691463470459, 'recon_norm_std': 0.06554490020498634, 'target_norm': 1.6195668190717698, 'recon_cos': 0.4336418904364109, 'gradnorm/compressor.query': 0.010746139983193343, 'gradnorm/compressor.layer0': 0.1994802124271587, 'gradnorm/compressor.layer1': 0.3026145328687444, 'gradnorm/compressor.layer2': 0.23746874911095817, 'gradnorm/compressor.layer3': 0.21310511117296319, 'gradnorm/compressor.layer4': 0.1693217158257872, 'gradnorm/compressor.layer5': 0.15532486741621965, 'gradnorm/compressor.layer6': 0.1327269222329418, 'gradnorm/compressor.layer7': 0.14603112990066028, 'gradnorm/decoder.layer0': 0.19271607748572112, 'gradnorm/decoder.layer1': 0.18829467377354545, 'gradnorm/decoder.layer2': 0.23248847039409534, 'gradnorm/max_param': 0.1796015501022339, 'epoch': 2.61}

 26%|██▌       | 4980/19090 [2:25:40<7:31:28,  1.92s/it]
 26%|██▌       | 4981/19090 [2:25:41<6:46:33,  1.73s/it]
 26%|██▌       | 4982/19090 [2:25:43<6:14:54,  1.59s/it]
 26%|██▌       | 4983/19090 [2:25:45<7:21:46,  1.88s/it]
 26%|██▌       | 4984/19090 [2:25:47<6:40:16,  1.70s/it]
 26%|██▌       | 4985/19090 [2:25:48<6:11:02,  1.58s/it]
 26%|██▌       | 4986/19090 [2:25:51<7:56:03,  2.03s/it]
 26%|██▌       | 4987/19090 [2:25:52<7:03:55,  1.80s/it]
 26%|██▌       | 4988/19090 [2:25:53<6:27:44,  1.65s/it]
 26%|██▌       | 4989/19090 [2:25:56<7:36:56,  1.94s/it]
 26%|██▌       | 4990/19090 [2:25:57<6:52:12,  1.75s/it]
                                                        
{'loss': 0.004, 'grad_norm': 0.6334631443023682, 'learning_rate': 4.6138956154151566e-05, 'recon_loss': 0.003989743528654799, 'recon_static': 0.0030429159523919225, 'recon_motion': 0.0009468275500694289, 'recon_norm': 2.1968684017658235, 'recon_norm_std': 0.06462371665984393, 'target_norm': 1.6189087897539138, 'recon_cos': 0.4202683314681053, 'gradnorm/compressor.query': 0.009760296008358791, 'gradnorm/compressor.layer0': 0.19009873412918313, 'gradnorm/compressor.layer1': 0.29974992233240366, 'gradnorm/compressor.layer2': 0.22388553914918546, 'gradnorm/compressor.layer3': 0.19903354824989491, 'gradnorm/compressor.layer4': 0.15630074511481287, 'gradnorm/compressor.layer5': 0.14305279377307875, 'gradnorm/compressor.layer6': 0.12247308278431568, 'gradnorm/compressor.layer7': 0.13417892317019772, 'gradnorm/decoder.layer0': 0.17772630526489547, 'gradnorm/decoder.layer1': 0.1743683457130365, 'gradnorm/decoder.layer2': 0.21345971539220168, 'gradnorm/max_param': 0.17173397541046143, 'epoch': 2.61}

 26%|██▌       | 4990/19090 [2:25:57<6:52:12,  1.75s/it]
 26%|██▌       | 4991/19090 [2:25:59<6:20:56,  1.62s/it]
 26%|██▌       | 4992/19090 [2:26:01<7:36:12,  1.94s/it]
 26%|██▌       | 4993/19090 [2:26:03<6:51:14,  1.75s/it]
 26%|██▌       | 4994/19090 [2:26:04<6:19:53,  1.62s/it]
 26%|██▌       | 4995/19090 [2:26:07<7:43:06,  1.97s/it]
 26%|██▌       | 4996/19090 [2:26:08<6:55:39,  1.77s/it]
 26%|██▌       | 4997/19090 [2:26:09<6:22:57,  1.63s/it]
 26%|██▌       | 4998/19090 [2:26:12<7:52:47,  2.01s/it]
 26%|██▌       | 4999/19090 [2:26:14<7:01:57,  1.80s/it]
 26%|██▌       | 5000/19090 [2:26:15<6:27:38,  1.65s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.5984669327735901, 'learning_rate': 4.6114515330020965e-05, 'recon_loss': 0.003827333147637546, 'recon_static': 0.0028671944863162937, 'recon_motion': 0.0009601387413567863, 'recon_norm': 2.1488991022109984, 'recon_norm_std': 0.06536982869729399, 'target_norm': 1.622508442401886, 'recon_cos': 0.42705378606915473, 'gradnorm/compressor.query': 0.00977019300017983, 'gradnorm/compressor.layer0': 0.17693923365148884, 'gradnorm/compressor.layer1': 0.27069268639985683, 'gradnorm/compressor.layer2': 0.2175052856482688, 'gradnorm/compressor.layer3': 0.1921286186486787, 'gradnorm/compressor.layer4': 0.14756238365734253, 'gradnorm/compressor.layer5': 0.13594559996139205, 'gradnorm/compressor.layer6': 0.11447440509407257, 'gradnorm/compressor.layer7': 0.12924772607686433, 'gradnorm/decoder.layer0': 0.17325615810415107, 'gradnorm/decoder.layer1': 0.16634885778796032, 'gradnorm/decoder.layer2': 0.20373606467422195, 'gradnorm/max_param': 0.1595616191625595, 'epoch': 2.62}

 26%|██▌       | 5000/19090 [2:26:15<6:27:38,  1.65s/it]
 26%|██▌       | 5001/19090 [2:26:18<8:40:55,  2.22s/it]
 26%|██▌       | 5002/19090 [2:26:20<7:35:46,  1.94s/it]
 26%|██▌       | 5003/19090 [2:26:21<6:50:09,  1.75s/it]
 26%|██▌       | 5004/19090 [2:26:23<6:47:29,  1.74s/it]
 26%|██▌       | 5005/19090 [2:26:24<6:16:03,  1.60s/it]
 26%|██▌       | 5006/19090 [2:26:25<5:54:29,  1.51s/it]
 26%|██▌       | 5007/19090 [2:26:28<7:14:40,  1.85s/it]
 26%|██▌       | 5008/19090 [2:26:29<6:34:59,  1.68s/it]
 26%|██▌       | 5009/19090 [2:26:31<6:06:59,  1.56s/it]
 26%|██▌       | 5010/19090 [2:26:33<7:20:22,  1.88s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.7705490589141846, 'learning_rate': 4.6090003909277335e-05, 'recon_loss': 0.00432699789525941, 'recon_static': 0.003359288035426289, 'recon_motion': 0.0009677099267719313, 'recon_norm': 2.271625280380249, 'recon_norm_std': 0.061389331985265014, 'target_norm': 1.6257084637880326, 'recon_cos': 0.4030908279120922, 'gradnorm/compressor.query': 0.01156920579202099, 'gradnorm/compressor.layer0': 0.23695422222916007, 'gradnorm/compressor.layer1': 0.3477139478660253, 'gradnorm/compressor.layer2': 0.2789496283897773, 'gradnorm/compressor.layer3': 0.24459285051593, 'gradnorm/compressor.layer4': 0.1957862544200297, 'gradnorm/compressor.layer5': 0.1708624011601578, 'gradnorm/compressor.layer6': 0.15111440655438588, 'gradnorm/compressor.layer7': 0.16977682719200232, 'gradnorm/decoder.layer0': 0.2173405265980438, 'gradnorm/decoder.layer1': 0.21350570507713124, 'gradnorm/decoder.layer2': 0.2586766268640202, 'gradnorm/max_param': 0.2145872563123703, 'epoch': 2.62}

 26%|██▌       | 5010/19090 [2:26:33<7:20:22,  1.88s/it]
 26%|██▌       | 5011/19090 [2:26:34<6:39:10,  1.70s/it]
 26%|██▋       | 5012/19090 [2:26:36<6:10:06,  1.58s/it]
 26%|██▋       | 5013/19090 [2:26:39<7:37:07,  1.95s/it]
 26%|██▋       | 5014/19090 [2:26:40<6:50:39,  1.75s/it]
 26%|██▋       | 5015/19090 [2:26:41<6:17:58,  1.61s/it]
 26%|██▋       | 5016/19090 [2:26:44<7:43:27,  1.98s/it]
 26%|██▋       | 5017/19090 [2:26:45<6:54:59,  1.77s/it]
 26%|██▋       | 5018/19090 [2:26:47<6:20:46,  1.62s/it]
 26%|██▋       | 5019/19090 [2:26:49<7:34:43,  1.94s/it]
 26%|██▋       | 5020/19090 [2:26:51<6:52:57,  1.76s/it]
                                                        
{'loss': 0.0064, 'grad_norm': 0.745945394039154, 'learning_rate': 4.606542197387489e-05, 'recon_loss': 0.006376624992117286, 'recon_static': 0.005421734496485442, 'recon_motion': 0.0009548904141411185, 'recon_norm': 2.744931024312973, 'recon_norm_std': 0.054335523303598164, 'target_norm': 1.6210369527339936, 'recon_cos': 0.333069834113121, 'gradnorm/compressor.query': 0.011732293538105596, 'gradnorm/compressor.layer0': 0.21914186460310991, 'gradnorm/compressor.layer1': 0.3228351184974992, 'gradnorm/compressor.layer2': 0.26391048231508585, 'gradnorm/compressor.layer3': 0.23370167653478552, 'gradnorm/compressor.layer4': 0.19094882530363996, 'gradnorm/compressor.layer5': 0.17726101562909316, 'gradnorm/compressor.layer6': 0.14780664656180742, 'gradnorm/compressor.layer7': 0.16426791093154303, 'gradnorm/decoder.layer0': 0.21995372473831507, 'gradnorm/decoder.layer1': 0.2133424234518044, 'gradnorm/decoder.layer2': 0.26301654333764357, 'gradnorm/max_param': 0.19771383702754974, 'epoch': 2.63}

 26%|██▋       | 5020/19090 [2:26:51<6:52:57,  1.76s/it]
 26%|██▋       | 5021/19090 [2:26:52<6:19:26,  1.62s/it]
 26%|██▋       | 5022/19090 [2:26:55<7:44:04,  1.98s/it]
 26%|██▋       | 5023/19090 [2:26:56<6:54:45,  1.77s/it]
 26%|██▋       | 5024/19090 [2:26:57<6:20:36,  1.62s/it]
 26%|██▋       | 5025/19090 [2:27:00<7:27:54,  1.91s/it]
 26%|██▋       | 5026/19090 [2:27:01<6:55:00,  1.77s/it]
 26%|██▋       | 5027/19090 [2:27:03<6:22:16,  1.63s/it]
 26%|██▋       | 5028/19090 [2:27:05<7:33:03,  1.93s/it]
 26%|██▋       | 5029/19090 [2:27:07<7:23:41,  1.89s/it]
 26%|██▋       | 5030/19090 [2:27:08<6:40:59,  1.71s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.6195242404937744, 'learning_rate': 4.6040769606003586e-05, 'recon_loss': 0.004348212818149478, 'recon_static': 0.003401995840249583, 'recon_motion': 0.000946216974989511, 'recon_norm': 2.2889050483703612, 'recon_norm_std': 0.06085589285939932, 'target_norm': 1.6187396019697189, 'recon_cos': 0.40378988459706305, 'gradnorm/compressor.query': 0.010237039163996596, 'gradnorm/compressor.layer0': 0.18025869307143585, 'gradnorm/compressor.layer1': 0.29062181747908983, 'gradnorm/compressor.layer2': 0.22409978362187463, 'gradnorm/compressor.layer3': 0.193126953000931, 'gradnorm/compressor.layer4': 0.15358716375291326, 'gradnorm/compressor.layer5': 0.14060075868840044, 'gradnorm/compressor.layer6': 0.11793504542953669, 'gradnorm/compressor.layer7': 0.13246284768097713, 'gradnorm/decoder.layer0': 0.1783842007857789, 'gradnorm/decoder.layer1': 0.1731886190621253, 'gradnorm/decoder.layer2': 0.2064088697216688, 'gradnorm/max_param': 0.16194531321525574, 'epoch': 2.63}

 26%|██▋       | 5030/19090 [2:27:08<6:40:59,  1.71s/it]
 26%|██▋       | 5031/19090 [2:27:11<7:19:38,  1.88s/it]
 26%|██▋       | 5032/19090 [2:27:12<7:23:00,  1.89s/it]
 26%|██▋       | 5033/19090 [2:27:14<6:40:48,  1.71s/it]
 26%|██▋       | 5034/19090 [2:27:16<7:22:42,  1.89s/it]
 26%|██▋       | 5035/19090 [2:27:18<7:23:01,  1.89s/it]
 26%|██▋       | 5036/19090 [2:27:19<6:40:31,  1.71s/it]
 26%|██▋       | 5037/19090 [2:27:21<7:14:27,  1.85s/it]
 26%|██▋       | 5038/19090 [2:27:23<7:20:56,  1.88s/it]
 26%|██▋       | 5039/19090 [2:27:25<6:39:03,  1.70s/it]
 26%|██▋       | 5040/19090 [2:27:27<6:54:35,  1.77s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.5614473223686218, 'learning_rate': 4.601604688808889e-05, 'recon_loss': 0.0035921125439926983, 'recon_static': 0.0026331657718401402, 'recon_motion': 0.0009589467939804308, 'recon_norm': 2.084196150302887, 'recon_norm_std': 0.06603044858202338, 'target_norm': 1.6215969800949097, 'recon_cos': 0.4412085026502609, 'gradnorm/compressor.query': 0.007622928740580274, 'gradnorm/compressor.layer0': 0.15841652758646738, 'gradnorm/compressor.layer1': 0.2670839856497366, 'gradnorm/compressor.layer2': 0.20509884139284262, 'gradnorm/compressor.layer3': 0.17635270584148835, 'gradnorm/compressor.layer4': 0.13572714742864833, 'gradnorm/compressor.layer5': 0.12557395652468079, 'gradnorm/compressor.layer6': 0.10865586017470692, 'gradnorm/compressor.layer7': 0.1198139588892036, 'gradnorm/decoder.layer0': 0.1593979409947735, 'gradnorm/decoder.layer1': 0.15978340036052024, 'gradnorm/decoder.layer2': 0.1851457476152302, 'gradnorm/max_param': 0.14348913729190826, 'epoch': 2.64}

 26%|██▋       | 5040/19090 [2:27:27<6:54:35,  1.77s/it]
 26%|██▋       | 5041/19090 [2:27:28<7:02:44,  1.81s/it]
 26%|██▋       | 5042/19090 [2:27:30<6:26:33,  1.65s/it]
 26%|██▋       | 5043/19090 [2:27:32<6:55:44,  1.78s/it]
 26%|██▋       | 5044/19090 [2:27:34<7:16:17,  1.86s/it]
 26%|██▋       | 5045/19090 [2:27:35<6:36:14,  1.69s/it]
 26%|██▋       | 5046/19090 [2:27:37<6:48:08,  1.74s/it]
 26%|██▋       | 5047/19090 [2:27:39<7:09:50,  1.84s/it]
 26%|██▋       | 5048/19090 [2:27:40<6:31:07,  1.67s/it]
 26%|██▋       | 5049/19090 [2:27:43<7:05:50,  1.82s/it]
 26%|██▋       | 5050/19090 [2:27:45<7:20:35,  1.88s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.546842634677887, 'learning_rate': 4.5991253902791486e-05, 'recon_loss': 0.0029803086537867783, 'recon_static': 0.0020334807195467874, 'recon_motion': 0.0009468280361033977, 'recon_norm': 1.9082088947296143, 'recon_norm_std': 0.06856981785967946, 'target_norm': 1.6166402310132981, 'recon_cos': 0.47910369709134104, 'gradnorm/compressor.query': 0.007342706879904424, 'gradnorm/compressor.layer0': 0.14782633572519616, 'gradnorm/compressor.layer1': 0.27094430514526446, 'gradnorm/compressor.layer2': 0.20902276800565625, 'gradnorm/compressor.layer3': 0.17861775538126454, 'gradnorm/compressor.layer4': 0.1329681101961802, 'gradnorm/compressor.layer5': 0.11753287680371567, 'gradnorm/compressor.layer6': 0.09774868405290474, 'gradnorm/compressor.layer7': 0.11213252097786734, 'gradnorm/decoder.layer0': 0.15220317712831286, 'gradnorm/decoder.layer1': 0.14835187048309803, 'gradnorm/decoder.layer2': 0.17125300546605296, 'gradnorm/max_param': 0.1342276930809021, 'epoch': 2.65}

 26%|██▋       | 5050/19090 [2:27:45<7:20:35,  1.88s/it]
 26%|██▋       | 5051/19090 [2:27:46<6:38:39,  1.70s/it]
 26%|██▋       | 5052/19090 [2:27:48<7:02:41,  1.81s/it]
 26%|██▋       | 5053/19090 [2:27:50<6:59:23,  1.79s/it]
 26%|██▋       | 5054/19090 [2:27:51<6:24:13,  1.64s/it]
 26%|██▋       | 5055/19090 [2:27:53<6:52:10,  1.76s/it]
 26%|██▋       | 5056/19090 [2:27:55<7:26:34,  1.91s/it]
 26%|██▋       | 5057/19090 [2:27:57<6:42:31,  1.72s/it]
 26%|██▋       | 5058/19090 [2:27:58<6:26:57,  1.65s/it]
 27%|██▋       | 5059/19090 [2:28:01<7:36:01,  1.95s/it]
 27%|██▋       | 5060/19090 [2:28:02<6:50:21,  1.75s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.5842386484146118, 'learning_rate': 4.596639073300699e-05, 'recon_loss': 0.003275915316771716, 'recon_static': 0.002307760325493291, 'recon_motion': 0.000968155030568596, 'recon_norm': 1.9888795167207718, 'recon_norm_std': 0.06660619881004096, 'target_norm': 1.6233798623085023, 'recon_cos': 0.457206217944622, 'gradnorm/compressor.query': 0.008455938850686124, 'gradnorm/compressor.layer0': 0.16302787866592625, 'gradnorm/compressor.layer1': 0.28130679087003857, 'gradnorm/compressor.layer2': 0.2141694062997193, 'gradnorm/compressor.layer3': 0.18505979150067434, 'gradnorm/compressor.layer4': 0.14146895941298962, 'gradnorm/compressor.layer5': 0.12939258789906374, 'gradnorm/compressor.layer6': 0.11164599764091036, 'gradnorm/compressor.layer7': 0.12458664378846714, 'gradnorm/decoder.layer0': 0.16610946247212274, 'gradnorm/decoder.layer1': 0.160964288148711, 'gradnorm/decoder.layer2': 0.19285127562095317, 'gradnorm/max_param': 0.1481320708990097, 'epoch': 2.65}

 27%|██▋       | 5060/19090 [2:28:02<6:50:21,  1.75s/it]
 27%|██▋       | 5061/19090 [2:28:03<6:26:18,  1.65s/it]
 27%|██▋       | 5062/19090 [2:28:06<7:52:03,  2.02s/it]
 27%|██▋       | 5063/19090 [2:28:08<7:01:09,  1.80s/it]
 27%|██▋       | 5064/19090 [2:28:09<6:26:00,  1.65s/it]
 27%|██▋       | 5065/19090 [2:28:11<7:20:47,  1.89s/it]
 27%|██▋       | 5066/19090 [2:28:13<6:39:43,  1.71s/it]
 27%|██▋       | 5067/19090 [2:28:14<6:18:16,  1.62s/it]
 27%|██▋       | 5068/19090 [2:28:16<7:14:50,  1.86s/it]
 27%|██▋       | 5069/19090 [2:28:18<6:34:25,  1.69s/it]
 27%|██▋       | 5070/19090 [2:28:19<6:34:05,  1.69s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.5685221552848816, 'learning_rate': 4.5941457461865676e-05, 'recon_loss': 0.00302788193221204, 'recon_static': 0.0020789885602425784, 'recon_motion': 0.000948893443273846, 'recon_norm': 1.925094097852707, 'recon_norm_std': 0.06796734668314457, 'target_norm': 1.6194689691066741, 'recon_cos': 0.4769654363393784, 'gradnorm/compressor.query': 0.008350770939931368, 'gradnorm/compressor.layer0': 0.16618039425424386, 'gradnorm/compressor.layer1': 0.284413061634006, 'gradnorm/compressor.layer2': 0.2086480455778034, 'gradnorm/compressor.layer3': 0.18159467874872742, 'gradnorm/compressor.layer4': 0.13518310236027908, 'gradnorm/compressor.layer5': 0.12455334808949255, 'gradnorm/compressor.layer6': 0.1037289562484978, 'gradnorm/compressor.layer7': 0.11506603358726038, 'gradnorm/decoder.layer0': 0.15611508028007257, 'gradnorm/decoder.layer1': 0.15269599442074777, 'gradnorm/decoder.layer2': 0.18070552453255082, 'gradnorm/max_param': 0.1511928290128708, 'epoch': 2.66}

 27%|██▋       | 5070/19090 [2:28:19<6:34:05,  1.69s/it]
 27%|██▋       | 5071/19090 [2:28:22<7:22:49,  1.90s/it]
 27%|██▋       | 5072/19090 [2:28:23<6:40:28,  1.71s/it]
 27%|██▋       | 5073/19090 [2:28:25<6:56:53,  1.78s/it]
 27%|██▋       | 5074/19090 [2:28:27<7:20:13,  1.88s/it]
 27%|██▋       | 5075/19090 [2:28:28<6:38:03,  1.70s/it]
 27%|██▋       | 5076/19090 [2:28:30<6:51:48,  1.76s/it]
 27%|██▋       | 5077/19090 [2:28:32<7:17:01,  1.87s/it]
 27%|██▋       | 5078/19090 [2:28:34<6:35:46,  1.69s/it]
 27%|██▋       | 5079/19090 [2:28:36<7:26:22,  1.91s/it]
 27%|██▋       | 5080/19090 [2:28:38<7:06:40,  1.83s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.6751470565795898, 'learning_rate': 4.591645417273222e-05, 'recon_loss': 0.003746424417477101, 'recon_static': 0.002779343817383051, 'recon_motion': 0.0009670806510257534, 'recon_norm': 2.1219374001026154, 'recon_norm_std': 0.06286362065002322, 'target_norm': 1.6235293090343474, 'recon_cos': 0.4298584841191769, 'gradnorm/compressor.query': 0.010193666267215575, 'gradnorm/compressor.layer0': 0.20038149480116138, 'gradnorm/compressor.layer1': 0.3253257421167443, 'gradnorm/compressor.layer2': 0.24221002566271735, 'gradnorm/compressor.layer3': 0.20955154960662833, 'gradnorm/compressor.layer4': 0.16816605025123493, 'gradnorm/compressor.layer5': 0.1493415594277544, 'gradnorm/compressor.layer6': 0.12798596374658242, 'gradnorm/compressor.layer7': 0.14403679398055194, 'gradnorm/decoder.layer0': 0.1888469456394866, 'gradnorm/decoder.layer1': 0.18461291512780942, 'gradnorm/decoder.layer2': 0.22292976144107965, 'gradnorm/max_param': 0.18177126348018646, 'epoch': 2.66}

 27%|██▋       | 5080/19090 [2:28:38<7:06:40,  1.83s/it]
 27%|██▋       | 5081/19090 [2:28:39<6:29:04,  1.67s/it]
 27%|██▋       | 5082/19090 [2:28:42<7:25:31,  1.91s/it]
 27%|██▋       | 5083/19090 [2:28:43<6:58:51,  1.79s/it]
 27%|██▋       | 5084/19090 [2:28:44<6:23:10,  1.64s/it]
 27%|██▋       | 5085/19090 [2:28:47<7:40:40,  1.97s/it]
 27%|██▋       | 5086/19090 [2:28:48<6:52:27,  1.77s/it]
 27%|██▋       | 5087/19090 [2:28:50<6:18:42,  1.62s/it]
 27%|██▋       | 5088/19090 [2:28:52<7:25:42,  1.91s/it]
 27%|██▋       | 5089/19090 [2:28:54<6:51:57,  1.77s/it]
 27%|██▋       | 5090/19090 [2:28:55<6:18:36,  1.62s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.7307231426239014, 'learning_rate': 4.589138094920537e-05, 'recon_loss': 0.004197310842573643, 'recon_static': 0.0032540912739932536, 'recon_motion': 0.0009432194288820029, 'recon_norm': 2.2468713760375976, 'recon_norm_std': 0.062189249135553834, 'target_norm': 1.6202998399734496, 'recon_cos': 0.4095199279487133, 'gradnorm/compressor.query': 0.011951232236127478, 'gradnorm/compressor.layer0': 0.22400945904197492, 'gradnorm/compressor.layer1': 0.3539535591891726, 'gradnorm/compressor.layer2': 0.2641873019614559, 'gradnorm/compressor.layer3': 0.22538372571358797, 'gradnorm/compressor.layer4': 0.17879738306374884, 'gradnorm/compressor.layer5': 0.16404691122563148, 'gradnorm/compressor.layer6': 0.13622821272756053, 'gradnorm/compressor.layer7': 0.15178532835767072, 'gradnorm/decoder.layer0': 0.2046113833648939, 'gradnorm/decoder.layer1': 0.19542722193801812, 'gradnorm/decoder.layer2': 0.2392131374455688, 'gradnorm/max_param': 0.20286867022514343, 'epoch': 2.67}

 27%|██▋       | 5090/19090 [2:28:55<6:18:36,  1.62s/it]
 27%|██▋       | 5091/19090 [2:28:57<7:19:56,  1.89s/it]
 27%|██▋       | 5092/19090 [2:28:59<6:58:34,  1.79s/it]
 27%|██▋       | 5093/19090 [2:29:00<6:22:56,  1.64s/it]
 27%|██▋       | 5094/19090 [2:29:03<7:10:41,  1.85s/it]
 27%|██▋       | 5095/19090 [2:29:05<7:13:50,  1.86s/it]
 27%|██▋       | 5096/19090 [2:29:06<6:33:43,  1.69s/it]
 27%|██▋       | 5097/19090 [2:29:08<7:19:14,  1.88s/it]
 27%|██▋       | 5098/19090 [2:29:10<6:47:17,  1.75s/it]
 27%|██▋       | 5099/19090 [2:29:11<6:14:52,  1.61s/it]
 27%|██▋       | 5100/19090 [2:29:14<7:40:23,  1.97s/it]
                                                        
{'loss': 0.004, 'grad_norm': 0.6720208525657654, 'learning_rate': 4.5866237875117754e-05, 'recon_loss': 0.004034136637346819, 'recon_static': 0.003079967718804255, 'recon_motion': 0.00095416885742452, 'recon_norm': 2.1999746382236482, 'recon_norm_std': 0.06023328760638833, 'target_norm': 1.6190928846597672, 'recon_cos': 0.4146853871643543, 'gradnorm/compressor.query': 0.009584819918492454, 'gradnorm/compressor.layer0': 0.19175524910577046, 'gradnorm/compressor.layer1': 0.3201935942853764, 'gradnorm/compressor.layer2': 0.2379974410431398, 'gradnorm/compressor.layer3': 0.20319637548434158, 'gradnorm/compressor.layer4': 0.1700843166123221, 'gradnorm/compressor.layer5': 0.15294945565951748, 'gradnorm/compressor.layer6': 0.12937925716435597, 'gradnorm/compressor.layer7': 0.14792024057955505, 'gradnorm/decoder.layer0': 0.1929282477959044, 'gradnorm/decoder.layer1': 0.18889283377916152, 'gradnorm/decoder.layer2': 0.22461340527443074, 'gradnorm/max_param': 0.17431801557540894, 'epoch': 2.67}

 27%|██▋       | 5100/19090 [2:29:14<7:40:23,  1.97s/it]
 27%|██▋       | 5101/19090 [2:29:15<6:52:11,  1.77s/it]
 27%|██▋       | 5102/19090 [2:29:16<6:18:19,  1.62s/it]
 27%|██▋       | 5103/19090 [2:29:19<7:54:45,  2.04s/it]
 27%|██▋       | 5104/19090 [2:29:21<7:02:11,  1.81s/it]
 27%|██▋       | 5105/19090 [2:29:22<6:25:26,  1.65s/it]
 27%|██▋       | 5106/19090 [2:29:25<7:44:29,  1.99s/it]
 27%|██▋       | 5107/19090 [2:29:26<6:54:51,  1.78s/it]
 27%|██▋       | 5108/19090 [2:29:27<6:20:18,  1.63s/it]
 27%|██▋       | 5109/19090 [2:29:30<7:37:46,  1.96s/it]
 27%|██▋       | 5110/19090 [2:29:31<6:53:45,  1.78s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.66436368227005, 'learning_rate': 4.58410250345355e-05, 'recon_loss': 0.00372181887505576, 'recon_static': 0.002775773796020076, 'recon_motion': 0.0009460450397455134, 'recon_norm': 2.1224377751350403, 'recon_norm_std': 0.06498458376154304, 'target_norm': 1.618914970755577, 'recon_cos': 0.4328094065189362, 'gradnorm/compressor.query': 0.009550769028602184, 'gradnorm/compressor.layer0': 0.19113241771389816, 'gradnorm/compressor.layer1': 0.3160019988339318, 'gradnorm/compressor.layer2': 0.24641960762165527, 'gradnorm/compressor.layer3': 0.2085940914651312, 'gradnorm/compressor.layer4': 0.16729636680865734, 'gradnorm/compressor.layer5': 0.1495931399926442, 'gradnorm/compressor.layer6': 0.12708744132979619, 'gradnorm/compressor.layer7': 0.1401742094179545, 'gradnorm/decoder.layer0': 0.18678994636975896, 'gradnorm/decoder.layer1': 0.17937995672853949, 'gradnorm/decoder.layer2': 0.217786788531174, 'gradnorm/max_param': 0.17376941442489624, 'epoch': 2.68}

 27%|██▋       | 5110/19090 [2:29:31<6:53:45,  1.78s/it]
 27%|██▋       | 5111/19090 [2:29:33<6:19:57,  1.63s/it]
 27%|██▋       | 5112/19090 [2:29:35<7:32:04,  1.94s/it]
 27%|██▋       | 5113/19090 [2:29:37<6:51:15,  1.77s/it]
 27%|██▋       | 5114/19090 [2:29:38<6:17:59,  1.62s/it]
 27%|██▋       | 5115/19090 [2:29:40<7:18:51,  1.88s/it]
 27%|██▋       | 5116/19090 [2:29:42<6:39:16,  1.71s/it]
 27%|██▋       | 5117/19090 [2:29:43<6:09:18,  1.59s/it]
 27%|██▋       | 5118/19090 [2:29:46<7:46:36,  2.00s/it]
 27%|██▋       | 5119/19090 [2:29:47<6:57:02,  1.79s/it]
 27%|██▋       | 5120/19090 [2:29:49<6:22:07,  1.64s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.5887855887413025, 'learning_rate': 4.5815742511758005e-05, 'recon_loss': 0.0033598467824049296, 'recon_static': 0.002405779907712713, 'recon_motion': 0.0009540669052512385, 'recon_norm': 2.0176339238882064, 'recon_norm_std': 0.06587304286658764, 'target_norm': 1.620276728272438, 'recon_cos': 0.4549629673361778, 'gradnorm/compressor.query': 0.008547368295461604, 'gradnorm/compressor.layer0': 0.16742875104197913, 'gradnorm/compressor.layer1': 0.2855342815247007, 'gradnorm/compressor.layer2': 0.22816534175556633, 'gradnorm/compressor.layer3': 0.1876744214041508, 'gradnorm/compressor.layer4': 0.1424117750896231, 'gradnorm/compressor.layer5': 0.1278243759812258, 'gradnorm/compressor.layer6': 0.1052518442050689, 'gradnorm/compressor.layer7': 0.12104972300484985, 'gradnorm/decoder.layer0': 0.16526341284596513, 'gradnorm/decoder.layer1': 0.15890609304219674, 'gradnorm/decoder.layer2': 0.186648811026684, 'gradnorm/max_param': 0.15176653861999512, 'epoch': 2.68}

 27%|██▋       | 5120/19090 [2:29:49<6:22:07,  1.64s/it]
 27%|██▋       | 5121/19090 [2:29:52<8:08:09,  2.10s/it]
 27%|██▋       | 5122/19090 [2:29:53<7:12:09,  1.86s/it]
 27%|██▋       | 5123/19090 [2:29:54<6:33:29,  1.69s/it]
 27%|██▋       | 5124/19090 [2:29:57<7:54:28,  2.04s/it]
 27%|██▋       | 5125/19090 [2:29:58<7:02:08,  1.81s/it]
 27%|██▋       | 5126/19090 [2:30:00<6:25:29,  1.66s/it]
 27%|██▋       | 5127/19090 [2:30:02<7:09:17,  1.84s/it]
 27%|██▋       | 5128/19090 [2:30:03<6:29:58,  1.68s/it]
 27%|██▋       | 5129/19090 [2:30:05<6:02:36,  1.56s/it]
 27%|██▋       | 5130/19090 [2:30:07<7:17:05,  1.88s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.7561405301094055, 'learning_rate': 4.5790390391317675e-05, 'recon_loss': 0.0038971997797489165, 'recon_static': 0.002947237022453919, 'recon_motion': 0.0009499626379692927, 'recon_norm': 2.1628354370594023, 'recon_norm_std': 0.06395598445087672, 'target_norm': 1.6207349091768264, 'recon_cos': 0.4230883188545704, 'gradnorm/compressor.query': 0.011463047223744693, 'gradnorm/compressor.layer0': 0.22797328626787774, 'gradnorm/compressor.layer1': 0.37271838679791314, 'gradnorm/compressor.layer2': 0.2753328303300944, 'gradnorm/compressor.layer3': 0.23550086607996515, 'gradnorm/compressor.layer4': 0.1873988832568723, 'gradnorm/compressor.layer5': 0.16353401714909127, 'gradnorm/compressor.layer6': 0.13884587855427843, 'gradnorm/compressor.layer7': 0.15989529470317798, 'gradnorm/decoder.layer0': 0.20653661984184152, 'gradnorm/decoder.layer1': 0.20020900772515704, 'gradnorm/decoder.layer2': 0.24496943363425702, 'gradnorm/max_param': 0.2076028436422348, 'epoch': 2.69}

 27%|██▋       | 5130/19090 [2:30:07<7:17:05,  1.88s/it]
 27%|██▋       | 5131/19090 [2:30:09<6:35:37,  1.70s/it]
 27%|██▋       | 5132/19090 [2:30:10<6:06:36,  1.58s/it]
 27%|██▋       | 5133/19090 [2:30:13<7:25:01,  1.91s/it]
 27%|██▋       | 5134/19090 [2:30:14<6:41:14,  1.73s/it]
 27%|██▋       | 5135/19090 [2:30:15<6:10:42,  1.59s/it]
 27%|██▋       | 5136/19090 [2:30:18<7:41:46,  1.99s/it]
 27%|██▋       | 5137/19090 [2:30:19<6:53:24,  1.78s/it]
 27%|██▋       | 5138/19090 [2:30:21<6:19:38,  1.63s/it]
 27%|██▋       | 5139/19090 [2:30:24<7:53:15,  2.04s/it]
 27%|██▋       | 5140/19090 [2:30:25<7:01:30,  1.81s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.6872150897979736, 'learning_rate': 4.5764968757979595e-05, 'recon_loss': 0.004608519794419408, 'recon_static': 0.003660892846528441, 'recon_motion': 0.0009476268271100708, 'recon_norm': 2.3469665884971618, 'recon_norm_std': 0.05889917826279998, 'target_norm': 1.619290292263031, 'recon_cos': 0.38965821713209153, 'gradnorm/compressor.query': 0.010457617373627059, 'gradnorm/compressor.layer0': 0.20007760844374367, 'gradnorm/compressor.layer1': 0.33919568588306925, 'gradnorm/compressor.layer2': 0.24720699314220515, 'gradnorm/compressor.layer3': 0.21249203596898975, 'gradnorm/compressor.layer4': 0.16540911342766432, 'gradnorm/compressor.layer5': 0.1515124270221738, 'gradnorm/compressor.layer6': 0.12924980130678396, 'gradnorm/compressor.layer7': 0.1444302412137917, 'gradnorm/decoder.layer0': 0.1935815979143155, 'gradnorm/decoder.layer1': 0.18618795172606775, 'gradnorm/decoder.layer2': 0.22535473258436542, 'gradnorm/max_param': 0.1816040724515915, 'epoch': 2.69}

 27%|██▋       | 5140/19090 [2:30:25<7:01:30,  1.81s/it]
 27%|██▋       | 5141/19090 [2:30:26<6:24:33,  1.65s/it]
 27%|██▋       | 5142/19090 [2:30:29<7:35:31,  1.96s/it]
 27%|██▋       | 5143/19090 [2:30:30<6:49:47,  1.76s/it]
 27%|██▋       | 5144/19090 [2:30:31<6:17:16,  1.62s/it]
 27%|██▋       | 5145/19090 [2:30:34<7:30:56,  1.94s/it]
 27%|██▋       | 5146/19090 [2:30:35<6:45:59,  1.75s/it]
 27%|██▋       | 5147/19090 [2:30:37<6:14:33,  1.61s/it]
 27%|██▋       | 5148/19090 [2:30:39<7:38:42,  1.97s/it]
 27%|██▋       | 5149/19090 [2:30:41<6:50:30,  1.77s/it]
 27%|██▋       | 5150/19090 [2:30:42<6:17:01,  1.62s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.6989783048629761, 'learning_rate': 4.573947769674129e-05, 'recon_loss': 0.004627060634084046, 'recon_static': 0.003674566629342735, 'recon_motion': 0.0009524941866402514, 'recon_norm': 2.3503151714801787, 'recon_norm_std': 0.05866296784952283, 'target_norm': 1.621120497584343, 'recon_cos': 0.3899416461586952, 'gradnorm/compressor.query': 0.009899892025012038, 'gradnorm/compressor.layer0': 0.20226988162103207, 'gradnorm/compressor.layer1': 0.33900693930427156, 'gradnorm/compressor.layer2': 0.24820966710158182, 'gradnorm/compressor.layer3': 0.21594905896861719, 'gradnorm/compressor.layer4': 0.17213303530186108, 'gradnorm/compressor.layer5': 0.1552653640471549, 'gradnorm/compressor.layer6': 0.13026227946158894, 'gradnorm/compressor.layer7': 0.14736932472896083, 'gradnorm/decoder.layer0': 0.200661205518693, 'gradnorm/decoder.layer1': 0.19129561126573233, 'gradnorm/decoder.layer2': 0.2347546681213178, 'gradnorm/max_param': 0.18405760824680328, 'epoch': 2.7}

 27%|██▋       | 5150/19090 [2:30:42<6:17:01,  1.62s/it]
 27%|██▋       | 5151/19090 [2:30:45<7:37:58,  1.97s/it]
 27%|██▋       | 5152/19090 [2:30:46<6:50:13,  1.77s/it]
 27%|██▋       | 5153/19090 [2:30:47<6:17:06,  1.62s/it]
 27%|██▋       | 5154/19090 [2:30:50<7:38:22,  1.97s/it]
 27%|██▋       | 5155/19090 [2:30:52<6:50:40,  1.77s/it]
 27%|██▋       | 5156/19090 [2:30:53<6:17:22,  1.62s/it]
 27%|██▋       | 5157/19090 [2:30:55<7:23:41,  1.91s/it]
 27%|██▋       | 5158/19090 [2:30:57<6:40:05,  1.72s/it]
 27%|██▋       | 5159/19090 [2:30:58<6:09:52,  1.59s/it]
 27%|██▋       | 5160/19090 [2:31:01<8:05:11,  2.09s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.6585870385169983, 'learning_rate': 4.5713917292832384e-05, 'recon_loss': 0.004169344081310556, 'recon_static': 0.003210673644207418, 'recon_motion': 0.0009586704356479459, 'recon_norm': 2.2374320030212402, 'recon_norm_std': 0.059809421002864835, 'target_norm': 1.6218705624341965, 'recon_cos': 0.40977186933159826, 'gradnorm/compressor.query': 0.009777847851396352, 'gradnorm/compressor.layer0': 0.19043634843180782, 'gradnorm/compressor.layer1': 0.33567331705084164, 'gradnorm/compressor.layer2': 0.24176976879323245, 'gradnorm/compressor.layer3': 0.2059163175869644, 'gradnorm/compressor.layer4': 0.15683550211084724, 'gradnorm/compressor.layer5': 0.14163253708486015, 'gradnorm/compressor.layer6': 0.11847609313912119, 'gradnorm/compressor.layer7': 0.1343786516358764, 'gradnorm/decoder.layer0': 0.18174143014154864, 'gradnorm/decoder.layer1': 0.17721388132668672, 'gradnorm/decoder.layer2': 0.2065393524214061, 'gradnorm/max_param': 0.17317542433738708, 'epoch': 2.7}

 27%|██▋       | 5160/19090 [2:31:01<8:05:11,  2.09s/it]
 27%|██▋       | 5161/19090 [2:31:02<7:10:21,  1.85s/it]
 27%|██▋       | 5162/19090 [2:31:04<6:31:35,  1.69s/it]
 27%|██▋       | 5163/19090 [2:31:06<6:48:51,  1.76s/it]
 27%|██▋       | 5164/19090 [2:31:07<6:16:06,  1.62s/it]
 27%|██▋       | 5165/19090 [2:31:08<5:53:23,  1.52s/it]
 27%|██▋       | 5166/19090 [2:31:11<7:27:17,  1.93s/it]
 27%|██▋       | 5167/19090 [2:31:12<6:43:20,  1.74s/it]
 27%|██▋       | 5168/19090 [2:31:14<6:12:30,  1.61s/it]
 27%|██▋       | 5169/19090 [2:31:16<7:15:51,  1.88s/it]
 27%|██▋       | 5170/19090 [2:31:18<6:35:00,  1.70s/it]
                                                        
{'loss': 0.0049, 'grad_norm': 0.7876556515693665, 'learning_rate': 4.568828763171437e-05, 'recon_loss': 0.004908119200263172, 'recon_static': 0.003954669047379867, 'recon_motion': 0.0009534501150483266, 'recon_norm': 2.4158167600631715, 'recon_norm_std': 0.056998575758188966, 'target_norm': 1.6203563272953034, 'recon_cos': 0.38245116248726846, 'gradnorm/compressor.query': 0.011747412101580293, 'gradnorm/compressor.layer0': 0.23759198164366474, 'gradnorm/compressor.layer1': 0.38023828580094116, 'gradnorm/compressor.layer2': 0.28033105078630527, 'gradnorm/compressor.layer3': 0.2480289047219727, 'gradnorm/compressor.layer4': 0.19808187720926135, 'gradnorm/compressor.layer5': 0.17587115849486393, 'gradnorm/compressor.layer6': 0.14734152196532532, 'gradnorm/compressor.layer7': 0.16568646617460386, 'gradnorm/decoder.layer0': 0.2196090977659209, 'gradnorm/decoder.layer1': 0.21351892167975445, 'gradnorm/decoder.layer2': 0.25686838213431257, 'gradnorm/max_param': 0.2164241373538971, 'epoch': 2.71}

 27%|██▋       | 5170/19090 [2:31:18<6:35:00,  1.70s/it]
 27%|██▋       | 5171/19090 [2:31:19<6:06:26,  1.58s/it]
 27%|██▋       | 5172/19090 [2:31:22<7:44:53,  2.00s/it]
 27%|██▋       | 5173/19090 [2:31:23<6:56:49,  1.80s/it]
 27%|██▋       | 5174/19090 [2:31:24<6:21:47,  1.65s/it]
 27%|██▋       | 5175/19090 [2:31:27<7:48:10,  2.02s/it]
 27%|██▋       | 5176/19090 [2:31:29<6:57:59,  1.80s/it]
 27%|██▋       | 5177/19090 [2:31:30<6:22:51,  1.65s/it]
 27%|██▋       | 5178/19090 [2:31:33<7:40:17,  1.99s/it]
 27%|██▋       | 5179/19090 [2:31:34<6:52:08,  1.78s/it]
 27%|██▋       | 5180/19090 [2:31:35<6:19:02,  1.63s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.7051214575767517, 'learning_rate': 4.566258879908031e-05, 'recon_loss': 0.004528938711155206, 'recon_static': 0.003578116453718394, 'recon_motion': 0.0009508221497526393, 'recon_norm': 2.325359123945236, 'recon_norm_std': 0.05809692787006497, 'target_norm': 1.6210076928138732, 'recon_cos': 0.39372079595923426, 'gradnorm/compressor.query': 0.011089669410229415, 'gradnorm/compressor.layer0': 0.20458122180624685, 'gradnorm/compressor.layer1': 0.35471389605068004, 'gradnorm/compressor.layer2': 0.2610773225886464, 'gradnorm/compressor.layer3': 0.21954058087702405, 'gradnorm/compressor.layer4': 0.1703617442534053, 'gradnorm/compressor.layer5': 0.153535825065279, 'gradnorm/compressor.layer6': 0.1289103897470371, 'gradnorm/compressor.layer7': 0.14692682838102297, 'gradnorm/decoder.layer0': 0.1949097964074407, 'gradnorm/decoder.layer1': 0.185093868788849, 'gradnorm/decoder.layer2': 0.2235779840814289, 'gradnorm/max_param': 0.18775200843811035, 'epoch': 2.71}

 27%|██▋       | 5180/19090 [2:31:35<6:19:02,  1.63s/it]
 27%|██▋       | 5181/19090 [2:31:38<7:44:24,  2.00s/it]
 27%|██▋       | 5182/19090 [2:31:39<6:55:51,  1.79s/it]
 27%|██▋       | 5183/19090 [2:31:41<6:21:48,  1.65s/it]
 27%|██▋       | 5184/19090 [2:31:44<7:38:38,  1.98s/it]
 27%|██▋       | 5185/19090 [2:31:45<6:50:47,  1.77s/it]
 27%|██▋       | 5186/19090 [2:31:46<6:17:29,  1.63s/it]
 27%|██▋       | 5187/19090 [2:31:49<7:53:35,  2.04s/it]
 27%|██▋       | 5188/19090 [2:31:50<7:01:27,  1.82s/it]
 27%|██▋       | 5189/19090 [2:31:52<6:25:11,  1.66s/it]
 27%|██▋       | 5190/19090 [2:31:54<7:12:40,  1.87s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.6961624622344971, 'learning_rate': 4.563682088085452e-05, 'recon_loss': 0.003787963039940223, 'recon_static': 0.002824192662956193, 'recon_motion': 0.0009637704206397757, 'recon_norm': 2.1293293297290803, 'recon_norm_std': 0.06056939642876387, 'target_norm': 1.6218004196882247, 'recon_cos': 0.4259026810526848, 'gradnorm/compressor.query': 0.010213659054941282, 'gradnorm/compressor.layer0': 0.20200302127429398, 'gradnorm/compressor.layer1': 0.3402642794232692, 'gradnorm/compressor.layer2': 0.2604374765982456, 'gradnorm/compressor.layer3': 0.22001449369241685, 'gradnorm/compressor.layer4': 0.16997245057831542, 'gradnorm/compressor.layer5': 0.15184163830936068, 'gradnorm/compressor.layer6': 0.12689845301759475, 'gradnorm/compressor.layer7': 0.14585171453120765, 'gradnorm/decoder.layer0': 0.19920142416318073, 'gradnorm/decoder.layer1': 0.1847762820535486, 'gradnorm/decoder.layer2': 0.2203925994925233, 'gradnorm/max_param': 0.18394005298614502, 'epoch': 2.72}

 27%|██▋       | 5190/19090 [2:31:54<7:12:40,  1.87s/it]
 27%|██▋       | 5191/19090 [2:31:55<6:32:24,  1.69s/it]
 27%|██▋       | 5192/19090 [2:31:57<6:04:15,  1.57s/it]
 27%|██▋       | 5193/19090 [2:31:59<7:31:47,  1.95s/it]
 27%|██▋       | 5194/19090 [2:32:01<6:46:17,  1.75s/it]
 27%|██▋       | 5195/19090 [2:32:02<6:14:05,  1.62s/it]
 27%|██▋       | 5196/19090 [2:32:05<7:24:54,  1.92s/it]
 27%|██▋       | 5197/19090 [2:32:06<6:40:40,  1.73s/it]
 27%|██▋       | 5198/19090 [2:32:07<6:10:02,  1.60s/it]
 27%|██▋       | 5199/19090 [2:32:10<7:56:30,  2.06s/it]
 27%|██▋       | 5200/19090 [2:32:12<7:03:06,  1.83s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.756432294845581, 'learning_rate': 4.561098396319232e-05, 'recon_loss': 0.0046265011536888775, 'recon_static': 0.003665087459376082, 'recon_motion': 0.0009614136070013046, 'recon_norm': 2.3512897074222563, 'recon_norm_std': 0.05761832091957331, 'target_norm': 1.6222485065460206, 'recon_cos': 0.38923249766230583, 'gradnorm/compressor.query': 0.011512221670912008, 'gradnorm/compressor.layer0': 0.22122002725923576, 'gradnorm/compressor.layer1': 0.3728062317889626, 'gradnorm/compressor.layer2': 0.27810729056136085, 'gradnorm/compressor.layer3': 0.23333131152083156, 'gradnorm/compressor.layer4': 0.18367777585732836, 'gradnorm/compressor.layer5': 0.16730917019464311, 'gradnorm/compressor.layer6': 0.1393031138907303, 'gradnorm/compressor.layer7': 0.160653030031559, 'gradnorm/decoder.layer0': 0.20983815706231645, 'gradnorm/decoder.layer1': 0.20071971410466052, 'gradnorm/decoder.layer2': 0.24714827532557182, 'gradnorm/max_param': 0.20152772963047028, 'epoch': 2.72}

 27%|██▋       | 5200/19090 [2:32:12<7:03:06,  1.83s/it]
 27%|██▋       | 5201/19090 [2:32:13<6:25:56,  1.67s/it]
 27%|██▋       | 5202/19090 [2:32:16<7:56:49,  2.06s/it]
 27%|██▋       | 5203/19090 [2:32:17<7:02:57,  1.83s/it]
 27%|██▋       | 5204/19090 [2:32:19<6:25:39,  1.67s/it]
 27%|██▋       | 5205/19090 [2:32:21<7:30:25,  1.95s/it]
 27%|██▋       | 5206/19090 [2:32:22<6:44:39,  1.75s/it]
 27%|██▋       | 5207/19090 [2:32:24<6:12:19,  1.61s/it]
 27%|██▋       | 5208/19090 [2:32:26<7:14:08,  1.88s/it]
 27%|██▋       | 5209/19090 [2:32:28<6:33:11,  1.70s/it]
 27%|██▋       | 5210/19090 [2:32:29<6:04:29,  1.58s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.6801363825798035, 'learning_rate': 4.558507813247974e-05, 'recon_loss': 0.004462022078223526, 'recon_static': 0.003513314720476046, 'recon_motion': 0.0009487074465141631, 'recon_norm': 2.309445345401764, 'recon_norm_std': 0.05941619984805584, 'target_norm': 1.619240438938141, 'recon_cos': 0.3948442257940769, 'gradnorm/compressor.query': 0.010267830445969142, 'gradnorm/compressor.layer0': 0.19531322479113564, 'gradnorm/compressor.layer1': 0.3389786748501368, 'gradnorm/compressor.layer2': 0.24953535233041435, 'gradnorm/compressor.layer3': 0.21458099944194195, 'gradnorm/compressor.layer4': 0.16547014153915723, 'gradnorm/compressor.layer5': 0.14563288209933042, 'gradnorm/compressor.layer6': 0.12446000390280021, 'gradnorm/compressor.layer7': 0.13994451627183782, 'gradnorm/decoder.layer0': 0.190919178522282, 'gradnorm/decoder.layer1': 0.18442788254712722, 'gradnorm/decoder.layer2': 0.21676156039906833, 'gradnorm/max_param': 0.1771073043346405, 'epoch': 2.73}

 27%|██▋       | 5210/19090 [2:32:29<6:04:29,  1.58s/it]
 27%|██▋       | 5211/19090 [2:32:31<7:14:51,  1.88s/it]
 27%|██▋       | 5212/19090 [2:32:33<6:34:00,  1.70s/it]
 27%|██▋       | 5213/19090 [2:32:34<6:05:42,  1.58s/it]
 27%|██▋       | 5214/19090 [2:32:37<7:26:57,  1.93s/it]
 27%|██▋       | 5215/19090 [2:32:38<6:42:43,  1.74s/it]
 27%|██▋       | 5216/19090 [2:32:39<6:12:07,  1.61s/it]
 27%|██▋       | 5217/19090 [2:32:42<7:36:42,  1.98s/it]
 27%|██▋       | 5218/19090 [2:32:43<6:49:48,  1.77s/it]
 27%|██▋       | 5219/19090 [2:32:45<6:17:19,  1.63s/it]
 27%|██▋       | 5220/19090 [2:32:48<7:33:53,  1.96s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.620083749294281, 'learning_rate': 4.555910347533319e-05, 'recon_loss': 0.003038492996711284, 'recon_static': 0.0020867185870883987, 'recon_motion': 0.0009517744139884599, 'recon_norm': 1.9227963238954544, 'recon_norm_std': 0.0684882958419621, 'target_norm': 1.621422353386879, 'recon_cos': 0.4753511607646942, 'gradnorm/compressor.query': 0.008551686878744367, 'gradnorm/compressor.layer0': 0.16966308420728934, 'gradnorm/compressor.layer1': 0.33701179303212925, 'gradnorm/compressor.layer2': 0.23329798879185729, 'gradnorm/compressor.layer3': 0.19458871245522005, 'gradnorm/compressor.layer4': 0.14151370512822628, 'gradnorm/compressor.layer5': 0.12835407480637906, 'gradnorm/compressor.layer6': 0.10638037284299845, 'gradnorm/compressor.layer7': 0.12063378607791447, 'gradnorm/decoder.layer0': 0.16650275400607656, 'gradnorm/decoder.layer1': 0.15874966652339542, 'gradnorm/decoder.layer2': 0.18568980082919637, 'gradnorm/max_param': 0.1768527626991272, 'epoch': 2.73}

 27%|██▋       | 5220/19090 [2:32:48<7:33:53,  1.96s/it]
 27%|██▋       | 5221/19090 [2:32:49<6:47:50,  1.76s/it]
 27%|██▋       | 5222/19090 [2:32:50<6:15:48,  1.63s/it]
 27%|██▋       | 5223/19090 [2:32:53<7:36:22,  1.97s/it]
 27%|██▋       | 5224/19090 [2:32:54<6:48:50,  1.77s/it]
 27%|██▋       | 5225/19090 [2:32:55<6:16:09,  1.63s/it]
 27%|██▋       | 5226/19090 [2:32:58<7:13:24,  1.88s/it]
 27%|██▋       | 5227/19090 [2:32:59<6:33:09,  1.70s/it]
 27%|██▋       | 5228/19090 [2:33:01<6:05:27,  1.58s/it]
 27%|██▋       | 5229/19090 [2:33:03<7:13:00,  1.87s/it]
 27%|██▋       | 5230/19090 [2:33:04<6:32:12,  1.70s/it]
                                                        
{'loss': 0.0038, 'grad_norm': 0.7581437826156616, 'learning_rate': 4.5533060078599226e-05, 'recon_loss': 0.0038159671938046815, 'recon_static': 0.0028710224840324374, 'recon_motion': 0.0009449447868973948, 'recon_norm': 2.142537075281143, 'recon_norm_std': 0.06294537549838423, 'target_norm': 1.6187051594257356, 'recon_cos': 0.4269678257405758, 'gradnorm/compressor.query': 0.01155515942591532, 'gradnorm/compressor.layer0': 0.22476274810053037, 'gradnorm/compressor.layer1': 0.39846246304569743, 'gradnorm/compressor.layer2': 0.28161663424125644, 'gradnorm/compressor.layer3': 0.2368266545860578, 'gradnorm/compressor.layer4': 0.17705247881370628, 'gradnorm/compressor.layer5': 0.16057062043057604, 'gradnorm/compressor.layer6': 0.13634111839830665, 'gradnorm/compressor.layer7': 0.15166320773147973, 'gradnorm/decoder.layer0': 0.2024054252863319, 'gradnorm/decoder.layer1': 0.1954878210923108, 'gradnorm/decoder.layer2': 0.2281632192158867, 'gradnorm/max_param': 0.20586153864860535, 'epoch': 2.74}

 27%|██▋       | 5230/19090 [2:33:04<6:32:12,  1.70s/it]
 27%|██▋       | 5231/19090 [2:33:06<6:03:55,  1.58s/it]
 27%|██▋       | 5232/19090 [2:33:09<7:34:35,  1.97s/it]
 27%|██▋       | 5233/19090 [2:33:10<6:46:58,  1.76s/it]
 27%|██▋       | 5234/19090 [2:33:11<6:14:00,  1.62s/it]
 27%|██▋       | 5235/19090 [2:33:14<7:31:05,  1.95s/it]
 27%|██▋       | 5236/19090 [2:33:15<6:45:24,  1.76s/it]
 27%|██▋       | 5237/19090 [2:33:16<6:12:53,  1.62s/it]
 27%|██▋       | 5238/19090 [2:33:19<7:09:58,  1.86s/it]
 27%|██▋       | 5239/19090 [2:33:20<6:29:57,  1.69s/it]
 27%|██▋       | 5240/19090 [2:33:21<6:02:06,  1.57s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.7323849201202393, 'learning_rate': 4.550694802935423e-05, 'recon_loss': 0.004219849035143853, 'recon_static': 0.0032660848170053213, 'recon_motion': 0.0009537643723888323, 'recon_norm': 2.25004985332489, 'recon_norm_std': 0.05881251152604818, 'target_norm': 1.619559755921364, 'recon_cos': 0.40653659477829934, 'gradnorm/compressor.query': 0.0117733392967259, 'gradnorm/compressor.layer0': 0.2119281781676831, 'gradnorm/compressor.layer1': 0.37928253838171866, 'gradnorm/compressor.layer2': 0.27665034100659575, 'gradnorm/compressor.layer3': 0.22853538232147796, 'gradnorm/compressor.layer4': 0.17362338771064614, 'gradnorm/compressor.layer5': 0.1571844174782436, 'gradnorm/compressor.layer6': 0.12941964823386765, 'gradnorm/compressor.layer7': 0.15099689351685314, 'gradnorm/decoder.layer0': 0.19657164961741908, 'gradnorm/decoder.layer1': 0.19120427588605649, 'gradnorm/decoder.layer2': 0.22161988783575765, 'gradnorm/max_param': 0.19339600205421448, 'epoch': 2.74}

 27%|██▋       | 5240/19090 [2:33:21<6:02:06,  1.57s/it]
 27%|██▋       | 5241/19090 [2:33:24<7:32:15,  1.96s/it]
 27%|██▋       | 5242/19090 [2:33:26<6:46:57,  1.76s/it]
 27%|██▋       | 5243/19090 [2:33:27<6:14:37,  1.62s/it]
 27%|██▋       | 5244/19090 [2:33:30<7:33:57,  1.97s/it]
 27%|██▋       | 5245/19090 [2:33:31<6:47:27,  1.77s/it]
 27%|██▋       | 5246/19090 [2:33:32<6:16:06,  1.63s/it]
 27%|██▋       | 5247/19090 [2:33:35<7:19:48,  1.91s/it]
 27%|██▋       | 5248/19090 [2:33:36<6:38:12,  1.73s/it]
 27%|██▋       | 5249/19090 [2:33:37<6:08:05,  1.60s/it]
 28%|██▊       | 5250/19090 [2:33:40<7:34:47,  1.97s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.7970021963119507, 'learning_rate': 4.548076741490414e-05, 'recon_loss': 0.004581045208033175, 'recon_static': 0.003621254349127412, 'recon_motion': 0.0009597908239811659, 'recon_norm': 2.3361356973648073, 'recon_norm_std': 0.058498440310359, 'target_norm': 1.6229792028665542, 'recon_cos': 0.3897288404405117, 'gradnorm/compressor.query': 0.012570826520613773, 'gradnorm/compressor.layer0': 0.2358092847024045, 'gradnorm/compressor.layer1': 0.4023114163637642, 'gradnorm/compressor.layer2': 0.29072141176161775, 'gradnorm/compressor.layer3': 0.2463886371285367, 'gradnorm/compressor.layer4': 0.19517976064909293, 'gradnorm/compressor.layer5': 0.17248354814346606, 'gradnorm/compressor.layer6': 0.1465719152183698, 'gradnorm/compressor.layer7': 0.16593367292908365, 'gradnorm/decoder.layer0': 0.22205401002601863, 'gradnorm/decoder.layer1': 0.20477370419768146, 'gradnorm/decoder.layer2': 0.2535042882186053, 'gradnorm/max_param': 0.21558913588523865, 'epoch': 2.75}

 28%|██▊       | 5250/19090 [2:33:40<7:34:47,  1.97s/it]
 28%|██▊       | 5251/19090 [2:33:42<6:47:17,  1.77s/it]
 28%|██▊       | 5252/19090 [2:33:43<6:14:13,  1.62s/it]
 28%|██▊       | 5253/19090 [2:33:46<7:36:52,  1.98s/it]
 28%|██▊       | 5254/19090 [2:33:47<6:49:40,  1.78s/it]
 28%|██▊       | 5255/19090 [2:33:48<6:16:23,  1.63s/it]
 28%|██▊       | 5256/19090 [2:33:51<7:03:02,  1.83s/it]
 28%|██▊       | 5257/19090 [2:33:52<6:24:50,  1.67s/it]
 28%|██▊       | 5258/19090 [2:33:53<5:58:18,  1.55s/it]
 28%|██▊       | 5259/19090 [2:33:56<7:21:41,  1.92s/it]
 28%|██▊       | 5260/19090 [2:33:57<6:37:51,  1.73s/it]
                                                        
{'loss': 0.004, 'grad_norm': 0.7353459000587463, 'learning_rate': 4.54545183227841e-05, 'recon_loss': 0.004021349124377594, 'recon_static': 0.0030691366002429275, 'recon_motion': 0.0009522124819341116, 'recon_norm': 2.202315467596054, 'recon_norm_std': 0.06084668710827827, 'target_norm': 1.6209849387407302, 'recon_cos': 0.41799160838127136, 'gradnorm/compressor.query': 0.011576586180346857, 'gradnorm/compressor.layer0': 0.21948337236480786, 'gradnorm/compressor.layer1': 0.39238544499305533, 'gradnorm/compressor.layer2': 0.2771644012391468, 'gradnorm/compressor.layer3': 0.2351809314435993, 'gradnorm/compressor.layer4': 0.17108541886412912, 'gradnorm/compressor.layer5': 0.1521520692541227, 'gradnorm/compressor.layer6': 0.12640046355179096, 'gradnorm/compressor.layer7': 0.14348747989776695, 'gradnorm/decoder.layer0': 0.1933746331870429, 'gradnorm/decoder.layer1': 0.18306639190774704, 'gradnorm/decoder.layer2': 0.21549222354952924, 'gradnorm/max_param': 0.20056456327438354, 'epoch': 2.76}

 28%|██▊       | 5260/19090 [2:33:57<6:37:51,  1.73s/it]
 28%|██▊       | 5261/19090 [2:33:58<6:07:33,  1.59s/it]
 28%|██▊       | 5262/19090 [2:34:01<7:22:42,  1.92s/it]
 28%|██▊       | 5263/19090 [2:34:02<6:38:27,  1.73s/it]
 28%|██▊       | 5264/19090 [2:34:04<6:07:51,  1.60s/it]
 28%|██▊       | 5265/19090 [2:34:06<7:07:06,  1.85s/it]
 28%|██▊       | 5266/19090 [2:34:08<6:36:31,  1.72s/it]
 28%|██▊       | 5267/19090 [2:34:09<6:06:22,  1.59s/it]
 28%|██▊       | 5268/19090 [2:34:11<6:55:52,  1.81s/it]
 28%|██▊       | 5269/19090 [2:34:13<7:04:30,  1.84s/it]
 28%|██▊       | 5270/19090 [2:34:14<6:26:51,  1.68s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.7742187976837158, 'learning_rate': 4.542820084075825e-05, 'recon_loss': 0.004399317468050867, 'recon_static': 0.0034384229918941857, 'recon_motion': 0.0009608942840714007, 'recon_norm': 2.2868717312812805, 'recon_norm_std': 0.057922568172216415, 'target_norm': 1.6216996282339096, 'recon_cos': 0.3962414778769016, 'gradnorm/compressor.query': 0.01120359052417611, 'gradnorm/compressor.layer0': 0.22721418018123826, 'gradnorm/compressor.layer1': 0.39522536753675336, 'gradnorm/compressor.layer2': 0.28615737913733164, 'gradnorm/compressor.layer3': 0.24460686224563946, 'gradnorm/compressor.layer4': 0.1884271329586952, 'gradnorm/compressor.layer5': 0.16744941445763978, 'gradnorm/compressor.layer6': 0.14064000632433205, 'gradnorm/compressor.layer7': 0.1570691036983292, 'gradnorm/decoder.layer0': 0.21090416291617695, 'gradnorm/decoder.layer1': 0.19994163257540978, 'gradnorm/decoder.layer2': 0.23935261619234585, 'gradnorm/max_param': 0.20759043097496033, 'epoch': 2.76}

 28%|██▊       | 5270/19090 [2:34:14<6:26:51,  1.68s/it]
 28%|██▊       | 5271/19090 [2:34:17<7:06:37,  1.85s/it]
 28%|██▊       | 5272/19090 [2:34:18<6:36:23,  1.72s/it]
 28%|██▊       | 5273/19090 [2:34:19<6:06:13,  1.59s/it]
 28%|██▊       | 5274/19090 [2:34:22<7:07:32,  1.86s/it]
 28%|██▊       | 5275/19090 [2:34:24<6:55:34,  1.80s/it]
 28%|██▊       | 5276/19090 [2:34:25<6:19:39,  1.65s/it]
 28%|██▊       | 5277/19090 [2:34:27<7:04:08,  1.84s/it]
 28%|██▊       | 5278/19090 [2:34:28<6:27:53,  1.69s/it]
 28%|██▊       | 5279/19090 [2:34:30<6:00:09,  1.56s/it]
 28%|██▊       | 5280/19090 [2:34:32<7:13:06,  1.88s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.8226209282875061, 'learning_rate': 4.540181505681939e-05, 'recon_loss': 0.004295859707053751, 'recon_static': 0.0033345628704410045, 'recon_motion': 0.0009612969734007493, 'recon_norm': 2.2672992587089538, 'recon_norm_std': 0.05707613015547395, 'target_norm': 1.6228493690490722, 'recon_cos': 0.40275986343622205, 'gradnorm/compressor.query': 0.013304715822711059, 'gradnorm/compressor.layer0': 0.2570280219805847, 'gradnorm/compressor.layer1': 0.4331636271516269, 'gradnorm/compressor.layer2': 0.306107062926057, 'gradnorm/compressor.layer3': 0.25982369942531836, 'gradnorm/compressor.layer4': 0.19375453412995622, 'gradnorm/compressor.layer5': 0.17317009256828053, 'gradnorm/compressor.layer6': 0.14392079490030202, 'gradnorm/compressor.layer7': 0.15844400340808573, 'gradnorm/decoder.layer0': 0.21791020191558336, 'gradnorm/decoder.layer1': 0.20397934608294438, 'gradnorm/decoder.layer2': 0.2432802604429765, 'gradnorm/max_param': 0.2344650775194168, 'epoch': 2.77}

 28%|██▊       | 5280/19090 [2:34:32<7:13:06,  1.88s/it]
 28%|██▊       | 5281/19090 [2:34:34<6:41:49,  1.75s/it]
 28%|██▊       | 5282/19090 [2:34:35<6:10:12,  1.61s/it]
 28%|██▊       | 5283/19090 [2:34:37<7:03:48,  1.84s/it]
 28%|██▊       | 5284/19090 [2:34:39<6:25:20,  1.67s/it]
 28%|██▊       | 5285/19090 [2:34:40<5:58:36,  1.56s/it]
 28%|██▊       | 5286/19090 [2:34:42<6:35:59,  1.72s/it]
 28%|██▊       | 5287/19090 [2:34:44<6:31:02,  1.70s/it]
 28%|██▊       | 5288/19090 [2:34:45<6:02:21,  1.58s/it]
 28%|██▊       | 5289/19090 [2:34:47<6:43:11,  1.75s/it]
 28%|██▊       | 5290/19090 [2:34:49<6:13:26,  1.62s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.794340193271637, 'learning_rate': 4.5375361059188656e-05, 'recon_loss': 0.004631569015327841, 'recon_static': 0.003675531910266727, 'recon_motion': 0.0009560371487168595, 'recon_norm': 2.3505747616291046, 'recon_norm_std': 0.05749081689864397, 'target_norm': 1.6210298925638198, 'recon_cos': 0.3882544659078121, 'gradnorm/compressor.query': 0.01198055158205723, 'gradnorm/compressor.layer0': 0.2335117495978739, 'gradnorm/compressor.layer1': 0.42927961930921, 'gradnorm/compressor.layer2': 0.29063419973517574, 'gradnorm/compressor.layer3': 0.2509218000343819, 'gradnorm/compressor.layer4': 0.18594310014770232, 'gradnorm/compressor.layer5': 0.16923262580627632, 'gradnorm/compressor.layer6': 0.13612248642889066, 'gradnorm/compressor.layer7': 0.15477756878761126, 'gradnorm/decoder.layer0': 0.20932890220874237, 'gradnorm/decoder.layer1': 0.1992252573184691, 'gradnorm/decoder.layer2': 0.2353399360919072, 'gradnorm/max_param': 0.21597357094287872, 'epoch': 2.77}

 28%|██▊       | 5290/19090 [2:34:49<6:13:26,  1.62s/it]
 28%|██▊       | 5291/19090 [2:34:50<5:50:35,  1.52s/it]
 28%|██▊       | 5292/19090 [2:34:52<7:05:38,  1.85s/it]
 28%|██▊       | 5293/19090 [2:34:54<6:26:55,  1.68s/it]
 28%|██▊       | 5294/19090 [2:34:55<5:59:41,  1.56s/it]
 28%|██▊       | 5295/19090 [2:34:58<7:32:09,  1.97s/it]
 28%|██▊       | 5296/19090 [2:34:59<6:45:47,  1.77s/it]
 28%|██▊       | 5297/19090 [2:35:01<6:13:06,  1.62s/it]
 28%|██▊       | 5298/19090 [2:35:04<7:54:30,  2.06s/it]
 28%|██▊       | 5299/19090 [2:35:05<7:01:00,  1.83s/it]
 28%|██▊       | 5300/19090 [2:35:06<6:23:56,  1.67s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.6796645522117615, 'learning_rate': 4.534883893631529e-05, 'recon_loss': 0.003937926929211244, 'recon_static': 0.0029910662153270094, 'recon_motion': 0.0009468606949667446, 'recon_norm': 2.1764726519584654, 'recon_norm_std': 0.05802848692983389, 'target_norm': 1.6149815320968628, 'recon_cos': 0.41964696943759916, 'gradnorm/compressor.query': 0.009404938641008673, 'gradnorm/compressor.layer0': 0.18523881334950887, 'gradnorm/compressor.layer1': 0.3817241924708309, 'gradnorm/compressor.layer2': 0.24904093961358, 'gradnorm/compressor.layer3': 0.22055827088056484, 'gradnorm/compressor.layer4': 0.15403282532609044, 'gradnorm/compressor.layer5': 0.1347843516510542, 'gradnorm/compressor.layer6': 0.11361452520698401, 'gradnorm/compressor.layer7': 0.12882762736495623, 'gradnorm/decoder.layer0': 0.17411203451760576, 'gradnorm/decoder.layer1': 0.17108374766719864, 'gradnorm/decoder.layer2': 0.2004467571382078, 'gradnorm/max_param': 0.20940037071704865, 'epoch': 2.78}

 28%|██▊       | 5300/19090 [2:35:06<6:23:56,  1.67s/it]
 28%|██▊       | 5301/19090 [2:35:09<7:44:39,  2.02s/it]
 28%|██▊       | 5302/19090 [2:35:10<6:54:08,  1.80s/it]
 28%|██▊       | 5303/19090 [2:35:12<6:18:23,  1.65s/it]
 28%|██▊       | 5304/19090 [2:35:14<7:36:28,  1.99s/it]
 28%|██▊       | 5305/19090 [2:35:16<6:47:57,  1.78s/it]
 28%|██▊       | 5306/19090 [2:35:17<6:14:02,  1.63s/it]
 28%|██▊       | 5307/19090 [2:35:20<7:32:37,  1.97s/it]
 28%|██▊       | 5308/19090 [2:35:21<6:45:28,  1.77s/it]
 28%|██▊       | 5309/19090 [2:35:22<6:12:21,  1.62s/it]
 28%|██▊       | 5310/19090 [2:35:25<7:20:42,  1.92s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.6680678129196167, 'learning_rate': 4.5322248776876294e-05, 'recon_loss': 0.0033237801573704926, 'recon_static': 0.0023825280426535756, 'recon_motion': 0.0009412521350895986, 'recon_norm': 2.0075830936431887, 'recon_norm_std': 0.06444519851356745, 'target_norm': 1.6189775317907333, 'recon_cos': 0.4553848497569561, 'gradnorm/compressor.query': 0.008863205903314306, 'gradnorm/compressor.layer0': 0.17714360324105352, 'gradnorm/compressor.layer1': 0.38494038364886773, 'gradnorm/compressor.layer2': 0.2509222305801773, 'gradnorm/compressor.layer3': 0.2154676222408621, 'gradnorm/compressor.layer4': 0.15068274774023405, 'gradnorm/compressor.layer5': 0.13104381658319778, 'gradnorm/compressor.layer6': 0.10769517983645299, 'gradnorm/compressor.layer7': 0.12541288493190605, 'gradnorm/decoder.layer0': 0.1690447630365569, 'gradnorm/decoder.layer1': 0.16466612126966973, 'gradnorm/decoder.layer2': 0.18545424508288375, 'gradnorm/max_param': 0.21674208343029022, 'epoch': 2.78}

 28%|██▊       | 5310/19090 [2:35:25<7:20:42,  1.92s/it]
 28%|██▊       | 5311/19090 [2:35:26<6:38:01,  1.73s/it]
 28%|██▊       | 5312/19090 [2:35:28<6:07:19,  1.60s/it]
 28%|██▊       | 5313/19090 [2:35:30<7:33:47,  1.98s/it]
 28%|██▊       | 5314/19090 [2:35:32<6:46:52,  1.77s/it]
 28%|██▊       | 5315/19090 [2:35:33<6:13:45,  1.63s/it]
 28%|██▊       | 5316/19090 [2:35:35<7:13:34,  1.89s/it]
 28%|██▊       | 5317/19090 [2:35:37<6:32:14,  1.71s/it]
 28%|██▊       | 5318/19090 [2:35:38<6:03:53,  1.59s/it]
 28%|██▊       | 5319/19090 [2:35:41<7:20:25,  1.92s/it]
 28%|██▊       | 5320/19090 [2:35:42<6:37:06,  1.73s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.7869019508361816, 'learning_rate': 4.529559066977617e-05, 'recon_loss': 0.003251060418551788, 'recon_static': 0.0022950372949708253, 'recon_motion': 0.0009560230901115574, 'recon_norm': 1.9795573502779007, 'recon_norm_std': 0.06311855725944042, 'target_norm': 1.61952702999115, 'recon_cos': 0.45963781550526617, 'gradnorm/compressor.query': 0.011653308884626172, 'gradnorm/compressor.layer0': 0.23304560659372225, 'gradnorm/compressor.layer1': 0.43049959255713166, 'gradnorm/compressor.layer2': 0.28906639199600886, 'gradnorm/compressor.layer3': 0.24358718704181603, 'gradnorm/compressor.layer4': 0.1848213466633575, 'gradnorm/compressor.layer5': 0.1639537450818124, 'gradnorm/compressor.layer6': 0.13342501541714422, 'gradnorm/compressor.layer7': 0.1536061968567272, 'gradnorm/decoder.layer0': 0.20595376313729066, 'gradnorm/decoder.layer1': 0.20015125255018876, 'gradnorm/decoder.layer2': 0.2272230089869502, 'gradnorm/max_param': 0.2191746085882187, 'epoch': 2.79}

 28%|██▊       | 5320/19090 [2:35:42<6:37:06,  1.73s/it]
 28%|██▊       | 5321/19090 [2:35:43<6:07:16,  1.60s/it]
 28%|██▊       | 5322/19090 [2:35:46<7:06:29,  1.86s/it]
 28%|██▊       | 5323/19090 [2:35:47<6:26:55,  1.69s/it]
 28%|██▊       | 5324/19090 [2:35:48<5:59:27,  1.57s/it]
 28%|██▊       | 5325/19090 [2:35:51<7:24:48,  1.94s/it]
 28%|██▊       | 5326/19090 [2:35:52<6:40:33,  1.75s/it]
 28%|██▊       | 5327/19090 [2:35:54<6:09:38,  1.61s/it]
 28%|██▊       | 5328/19090 [2:35:57<7:33:29,  1.98s/it]
 28%|██▊       | 5329/19090 [2:35:58<6:46:25,  1.77s/it]
 28%|██▊       | 5330/19090 [2:35:59<6:13:30,  1.63s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.8228665590286255, 'learning_rate': 4.5268864704146584e-05, 'recon_loss': 0.004451621533371508, 'recon_static': 0.0035012099018786103, 'recon_motion': 0.0009504115558229386, 'recon_norm': 2.304031628370285, 'recon_norm_std': 0.05822173822671175, 'target_norm': 1.6193371742963791, 'recon_cos': 0.39347571432590484, 'gradnorm/compressor.query': 0.011397934336591907, 'gradnorm/compressor.layer0': 0.24537166744481814, 'gradnorm/compressor.layer1': 0.4453884193889056, 'gradnorm/compressor.layer2': 0.2889190395158942, 'gradnorm/compressor.layer3': 0.2687094918484014, 'gradnorm/compressor.layer4': 0.19452668418275315, 'gradnorm/compressor.layer5': 0.17118949083909443, 'gradnorm/compressor.layer6': 0.142874970451946, 'gradnorm/compressor.layer7': 0.16268029847767942, 'gradnorm/decoder.layer0': 0.21654575857046302, 'gradnorm/decoder.layer1': 0.20617658616234436, 'gradnorm/decoder.layer2': 0.2435095450774665, 'gradnorm/max_param': 0.22437684237957, 'epoch': 2.79}

 28%|██▊       | 5330/19090 [2:35:59<6:13:30,  1.63s/it]
 28%|██▊       | 5331/19090 [2:36:02<7:47:49,  2.04s/it]
 28%|██▊       | 5332/19090 [2:36:03<6:56:04,  1.81s/it]
 28%|██▊       | 5333/19090 [2:36:05<6:19:25,  1.65s/it]
 28%|██▊       | 5334/19090 [2:36:07<7:10:19,  1.88s/it]
 28%|██▊       | 5335/19090 [2:36:08<6:29:30,  1.70s/it]
 28%|██▊       | 5336/19090 [2:36:10<6:01:28,  1.58s/it]
 28%|██▊       | 5337/19090 [2:36:13<7:52:50,  2.06s/it]
 28%|██▊       | 5338/19090 [2:36:14<6:59:03,  1.83s/it]
 28%|██▊       | 5339/19090 [2:36:15<6:22:10,  1.67s/it]
 28%|██▊       | 5340/19090 [2:36:18<7:35:04,  1.99s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.8396710753440857, 'learning_rate': 4.524207096934609e-05, 'recon_loss': 0.004573763126973063, 'recon_static': 0.00361859689000994, 'recon_motion': 0.0009551662602461875, 'recon_norm': 2.3355052351951597, 'recon_norm_std': 0.056205952726304534, 'target_norm': 1.620494505763054, 'recon_cos': 0.3900181017816067, 'gradnorm/compressor.query': 0.012559176303758423, 'gradnorm/compressor.layer0': 0.23863124158138932, 'gradnorm/compressor.layer1': 0.470196431434278, 'gradnorm/compressor.layer2': 0.30807777921182544, 'gradnorm/compressor.layer3': 0.27170102647886524, 'gradnorm/compressor.layer4': 0.19023055399791736, 'gradnorm/compressor.layer5': 0.16912764153634424, 'gradnorm/compressor.layer6': 0.14227491955924956, 'gradnorm/compressor.layer7': 0.1607655298520271, 'gradnorm/decoder.layer0': 0.21483040681498783, 'gradnorm/decoder.layer1': 0.2102948571920435, 'gradnorm/decoder.layer2': 0.23847716516517015, 'gradnorm/max_param': 0.24886669218540192, 'epoch': 2.8}

 28%|██▊       | 5340/19090 [2:36:18<7:35:04,  1.99s/it]
 28%|██▊       | 5341/19090 [2:36:19<6:46:57,  1.78s/it]
 28%|██▊       | 5342/19090 [2:36:21<6:14:13,  1.63s/it]
 28%|██▊       | 5343/19090 [2:36:24<7:43:07,  2.02s/it]
 28%|██▊       | 5344/19090 [2:36:25<6:52:37,  1.80s/it]
 28%|██▊       | 5345/19090 [2:36:26<6:17:36,  1.65s/it]
 28%|██▊       | 5346/19090 [2:36:29<7:11:53,  1.89s/it]
 28%|██▊       | 5347/19090 [2:36:30<6:48:40,  1.78s/it]
 28%|██▊       | 5348/19090 [2:36:32<6:15:20,  1.64s/it]
 28%|██▊       | 5349/19090 [2:36:34<7:10:54,  1.88s/it]
 28%|██▊       | 5350/19090 [2:36:36<7:01:34,  1.84s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.8597242832183838, 'learning_rate': 4.521520955495983e-05, 'recon_loss': 0.00438921062159352, 'recon_static': 0.0034377570380456747, 'recon_motion': 0.0009514535748166963, 'recon_norm': 2.2922159910202025, 'recon_norm_std': 0.05771930841729045, 'target_norm': 1.620195859670639, 'recon_cos': 0.39880060106515886, 'gradnorm/compressor.query': 0.013506800156945093, 'gradnorm/compressor.layer0': 0.25865286355995093, 'gradnorm/compressor.layer1': 0.4732072944073252, 'gradnorm/compressor.layer2': 0.31470828632622233, 'gradnorm/compressor.layer3': 0.2661021658999559, 'gradnorm/compressor.layer4': 0.20155937872547333, 'gradnorm/compressor.layer5': 0.17397057129496032, 'gradnorm/compressor.layer6': 0.15092762527782752, 'gradnorm/compressor.layer7': 0.1687080889772932, 'gradnorm/decoder.layer0': 0.22321638123346074, 'gradnorm/decoder.layer1': 0.2108710767220466, 'gradnorm/decoder.layer2': 0.2486758180689776, 'gradnorm/max_param': 0.24057894945144653, 'epoch': 2.8}

 28%|██▊       | 5350/19090 [2:36:36<7:01:34,  1.84s/it]
 28%|██▊       | 5351/19090 [2:36:37<6:23:28,  1.67s/it]
 28%|██▊       | 5352/19090 [2:36:40<7:38:16,  2.00s/it]
 28%|██▊       | 5353/19090 [2:36:41<7:08:44,  1.87s/it]
 28%|██▊       | 5354/19090 [2:36:43<6:28:21,  1.70s/it]
 28%|██▊       | 5355/19090 [2:36:45<7:15:42,  1.90s/it]
 28%|██▊       | 5356/19090 [2:36:47<7:13:34,  1.89s/it]
 28%|██▊       | 5357/19090 [2:36:48<6:31:46,  1.71s/it]
 28%|██▊       | 5358/19090 [2:36:51<7:13:47,  1.90s/it]
 28%|██▊       | 5359/19090 [2:36:52<6:51:27,  1.80s/it]
 28%|██▊       | 5360/19090 [2:36:53<6:16:11,  1.64s/it]
                                                        
{'loss': 0.0055, 'grad_norm': 0.8709811568260193, 'learning_rate': 4.518828055079925e-05, 'recon_loss': 0.005507167696487158, 'recon_static': 0.004557415802264586, 'recon_motion': 0.0009497519204160199, 'recon_norm': 2.554895794391632, 'recon_norm_std': 0.05233275340870023, 'target_norm': 1.619602656364441, 'recon_cos': 0.3574801504611969, 'gradnorm/compressor.query': 0.01356546575166482, 'gradnorm/compressor.layer0': 0.2666063273356898, 'gradnorm/compressor.layer1': 0.4657201599322956, 'gradnorm/compressor.layer2': 0.31796794066455014, 'gradnorm/compressor.layer3': 0.27041504574119923, 'gradnorm/compressor.layer4': 0.20483856741651013, 'gradnorm/compressor.layer5': 0.18118619988031925, 'gradnorm/compressor.layer6': 0.15217102480544223, 'gradnorm/compressor.layer7': 0.17291697649085389, 'gradnorm/decoder.layer0': 0.22869544005234863, 'gradnorm/decoder.layer1': 0.21810626781607953, 'gradnorm/decoder.layer2': 0.26173759506473815, 'gradnorm/max_param': 0.2445889264345169, 'epoch': 2.81}

 28%|██▊       | 5360/19090 [2:36:53<6:16:11,  1.64s/it]
 28%|██▊       | 5361/19090 [2:36:56<7:12:07,  1.89s/it]
 28%|██▊       | 5362/19090 [2:36:58<7:03:07,  1.85s/it]
 28%|██▊       | 5363/19090 [2:36:59<6:24:50,  1.68s/it]
 28%|██▊       | 5364/19090 [2:37:01<7:05:00,  1.86s/it]
 28%|██▊       | 5365/19090 [2:37:03<6:49:54,  1.79s/it]
 28%|██▊       | 5366/19090 [2:37:04<6:15:05,  1.64s/it]
 28%|██▊       | 5367/19090 [2:37:06<6:57:07,  1.82s/it]
 28%|██▊       | 5368/19090 [2:37:08<6:37:01,  1.74s/it]
 28%|██▊       | 5369/19090 [2:37:09<6:06:18,  1.60s/it]
 28%|██▊       | 5370/19090 [2:37:12<7:38:46,  2.01s/it]
                                                        
{'loss': 0.0052, 'grad_norm': 0.9626774191856384, 'learning_rate': 4.5161284046901766e-05, 'recon_loss': 0.0051724050776101645, 'recon_static': 0.004216986458050087, 'recon_motion': 0.0009554184696753509, 'recon_norm': 2.476079285144806, 'recon_norm_std': 0.052238731924444434, 'target_norm': 1.619472822546959, 'recon_cos': 0.3694939509034157, 'gradnorm/compressor.query': 0.014332055218585058, 'gradnorm/compressor.layer0': 0.289731270179803, 'gradnorm/compressor.layer1': 0.5393137968045332, 'gradnorm/compressor.layer2': 0.35716371645638334, 'gradnorm/compressor.layer3': 0.28775723341940346, 'gradnorm/compressor.layer4': 0.22173888302582526, 'gradnorm/compressor.layer5': 0.2016410171995668, 'gradnorm/compressor.layer6': 0.16430686735276426, 'gradnorm/compressor.layer7': 0.1858757417358755, 'gradnorm/decoder.layer0': 0.24479130058396473, 'gradnorm/decoder.layer1': 0.2340982631232518, 'gradnorm/decoder.layer2': 0.2743538488390614, 'gradnorm/max_param': 0.27678412199020386, 'epoch': 2.81}

 28%|██▊       | 5370/19090 [2:37:12<7:38:46,  2.01s/it]
 28%|██▊       | 5371/19090 [2:37:13<6:49:48,  1.79s/it]
 28%|██▊       | 5372/19090 [2:37:15<6:15:16,  1.64s/it]
 28%|██▊       | 5373/19090 [2:37:18<7:37:37,  2.00s/it]
 28%|██▊       | 5374/19090 [2:37:19<6:48:49,  1.79s/it]
 28%|██▊       | 5375/19090 [2:37:20<6:14:48,  1.64s/it]
 28%|██▊       | 5376/19090 [2:37:23<7:29:34,  1.97s/it]
 28%|██▊       | 5377/19090 [2:37:24<6:47:20,  1.78s/it]
 28%|██▊       | 5378/19090 [2:37:26<6:13:41,  1.64s/it]
 28%|██▊       | 5379/19090 [2:37:28<7:37:40,  2.00s/it]
 28%|██▊       | 5380/19090 [2:37:30<6:48:47,  1.79s/it]
                                                        
{'loss': 0.004, 'grad_norm': 0.7429243922233582, 'learning_rate': 4.513422013353047e-05, 'recon_loss': 0.004008866025833413, 'recon_static': 0.0030503678892273454, 'recon_motion': 0.0009584982457454316, 'recon_norm': 2.184913581609726, 'recon_norm_std': 0.059409153461456296, 'target_norm': 1.6225149244070054, 'recon_cos': 0.4174698181450367, 'gradnorm/compressor.query': 0.010552397993854153, 'gradnorm/compressor.layer0': 0.20378009905853525, 'gradnorm/compressor.layer1': 0.4344450523643083, 'gradnorm/compressor.layer2': 0.27669630908493087, 'gradnorm/compressor.layer3': 0.24279105444015303, 'gradnorm/compressor.layer4': 0.16161152214911353, 'gradnorm/compressor.layer5': 0.14323126112244633, 'gradnorm/compressor.layer6': 0.1170309730075772, 'gradnorm/compressor.layer7': 0.1355918559544125, 'gradnorm/decoder.layer0': 0.18576193944159675, 'gradnorm/decoder.layer1': 0.17822275031100365, 'gradnorm/decoder.layer2': 0.2026386950293077, 'gradnorm/max_param': 0.2453431487083435, 'epoch': 2.82}

 28%|██▊       | 5380/19090 [2:37:30<6:48:47,  1.79s/it]
 28%|██▊       | 5381/19090 [2:37:31<6:15:18,  1.64s/it]
 28%|██▊       | 5382/19090 [2:37:34<7:36:41,  2.00s/it]
 28%|██▊       | 5383/19090 [2:37:35<6:48:03,  1.79s/it]
 28%|██▊       | 5384/19090 [2:37:36<6:13:52,  1.64s/it]
 28%|██▊       | 5385/19090 [2:37:39<7:16:20,  1.91s/it]
 28%|██▊       | 5386/19090 [2:37:40<6:33:37,  1.72s/it]
 28%|██▊       | 5387/19090 [2:37:42<6:03:59,  1.59s/it]
 28%|██▊       | 5388/19090 [2:37:45<7:42:13,  2.02s/it]
 28%|██▊       | 5389/19090 [2:37:46<6:52:51,  1.81s/it]
 28%|██▊       | 5390/19090 [2:37:47<6:19:09,  1.66s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.7064776420593262, 'learning_rate': 4.510708890117388e-05, 'recon_loss': 0.003122359811095521, 'recon_static': 0.0021735565562266855, 'recon_motion': 0.0009488032883382402, 'recon_norm': 1.9456832617521287, 'recon_norm_std': 0.06439275871962309, 'target_norm': 1.6194950670003891, 'recon_cos': 0.4678597591817379, 'gradnorm/compressor.query': 0.00936742314362632, 'gradnorm/compressor.layer0': 0.18336435962170636, 'gradnorm/compressor.layer1': 0.42399384734912415, 'gradnorm/compressor.layer2': 0.2768557513200236, 'gradnorm/compressor.layer3': 0.22322081215717804, 'gradnorm/compressor.layer4': 0.15162971864183886, 'gradnorm/compressor.layer5': 0.13124444870335222, 'gradnorm/compressor.layer6': 0.10704288819655983, 'gradnorm/compressor.layer7': 0.12599544235815652, 'gradnorm/decoder.layer0': 0.17487465401254637, 'gradnorm/decoder.layer1': 0.1672524103381582, 'gradnorm/decoder.layer2': 0.18179332303461207, 'gradnorm/max_param': 0.24849872291088104, 'epoch': 2.82}

 28%|██▊       | 5390/19090 [2:37:47<6:19:09,  1.66s/it]
 28%|██▊       | 5391/19090 [2:37:49<7:01:11,  1.84s/it]
 28%|██▊       | 5392/19090 [2:37:51<6:24:56,  1.69s/it]
 28%|██▊       | 5393/19090 [2:37:52<5:59:46,  1.58s/it]
 28%|██▊       | 5394/19090 [2:37:55<7:13:08,  1.90s/it]
 28%|██▊       | 5395/19090 [2:37:56<6:31:38,  1.72s/it]
 28%|██▊       | 5396/19090 [2:37:57<6:02:58,  1.59s/it]
 28%|██▊       | 5397/19090 [2:38:00<7:18:39,  1.92s/it]
 28%|██▊       | 5398/19090 [2:38:01<6:35:29,  1.73s/it]
 28%|██▊       | 5399/19090 [2:38:03<6:05:26,  1.60s/it]
 28%|██▊       | 5400/19090 [2:38:05<7:30:25,  1.97s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 0.7968848347663879, 'learning_rate': 4.507989044054554e-05, 'recon_loss': 0.003389611089369282, 'recon_static': 0.0024367342703044415, 'recon_motion': 0.0009528768743621186, 'recon_norm': 2.0212870210409166, 'recon_norm_std': 0.06185726281255484, 'target_norm': 1.6209265291690826, 'recon_cos': 0.44981858804821967, 'gradnorm/compressor.query': 0.011551215118473224, 'gradnorm/compressor.layer0': 0.22546770044030948, 'gradnorm/compressor.layer1': 0.4893570302536015, 'gradnorm/compressor.layer2': 0.3020584189346283, 'gradnorm/compressor.layer3': 0.24369249796578882, 'gradnorm/compressor.layer4': 0.16699424541610774, 'gradnorm/compressor.layer5': 0.14444133178121396, 'gradnorm/compressor.layer6': 0.12191388979957324, 'gradnorm/compressor.layer7': 0.13927943822826236, 'gradnorm/decoder.layer0': 0.18871385332435786, 'gradnorm/decoder.layer1': 0.18551370424612706, 'gradnorm/decoder.layer2': 0.20228779177354436, 'gradnorm/max_param': 0.2833508253097534, 'epoch': 2.83}

 28%|██▊       | 5400/19090 [2:38:05<7:30:25,  1.97s/it]
 28%|██▊       | 5401/19090 [2:38:07<6:44:17,  1.77s/it]
 28%|██▊       | 5402/19090 [2:38:08<6:11:52,  1.63s/it]
 28%|██▊       | 5403/19090 [2:38:11<7:23:06,  1.94s/it]
 28%|██▊       | 5404/19090 [2:38:12<6:39:03,  1.75s/it]
 28%|██▊       | 5405/19090 [2:38:13<6:08:07,  1.61s/it]
 28%|██▊       | 5406/19090 [2:38:16<7:30:46,  1.98s/it]
 28%|██▊       | 5407/19090 [2:38:17<6:44:16,  1.77s/it]
 28%|██▊       | 5408/19090 [2:38:19<6:11:27,  1.63s/it]
 28%|██▊       | 5409/19090 [2:38:21<7:30:24,  1.98s/it]
 28%|██▊       | 5410/19090 [2:38:23<6:44:07,  1.77s/it]
                                                        
{'loss': 0.0032, 'grad_norm': 0.7356327176094055, 'learning_rate': 4.505262484258381e-05, 'recon_loss': 0.00323573867790401, 'recon_static': 0.0022758009436074644, 'recon_motion': 0.0009599377895938233, 'recon_norm': 1.9748256474733352, 'recon_norm_std': 0.0635081484913826, 'target_norm': 1.6235730588436126, 'recon_cos': 0.4610752061009407, 'gradnorm/compressor.query': 0.0101445503199798, 'gradnorm/compressor.layer0': 0.19707795460063862, 'gradnorm/compressor.layer1': 0.45881958296047376, 'gradnorm/compressor.layer2': 0.2637234955166429, 'gradnorm/compressor.layer3': 0.2309244709572344, 'gradnorm/compressor.layer4': 0.1549870230641485, 'gradnorm/compressor.layer5': 0.13516420357515282, 'gradnorm/compressor.layer6': 0.11488693602813359, 'gradnorm/compressor.layer7': 0.12862235654935744, 'gradnorm/decoder.layer0': 0.17722484366659855, 'gradnorm/decoder.layer1': 0.1688669497055649, 'gradnorm/decoder.layer2': 0.1919986937187879, 'gradnorm/max_param': 0.27179133892059326, 'epoch': 2.83}

 28%|██▊       | 5410/19090 [2:38:23<6:44:07,  1.77s/it]
 28%|██▊       | 5411/19090 [2:38:24<6:11:40,  1.63s/it]
 28%|██▊       | 5412/19090 [2:38:27<7:43:58,  2.04s/it]
 28%|██▊       | 5413/19090 [2:38:28<6:52:50,  1.81s/it]
 28%|██▊       | 5414/19090 [2:38:30<6:17:33,  1.66s/it]
 28%|██▊       | 5415/19090 [2:38:33<7:43:57,  2.04s/it]
 28%|██▊       | 5416/19090 [2:38:34<6:53:06,  1.81s/it]
 28%|██▊       | 5417/19090 [2:38:35<6:17:44,  1.66s/it]
 28%|██▊       | 5418/19090 [2:38:38<7:06:52,  1.87s/it]
 28%|██▊       | 5419/19090 [2:38:39<6:27:07,  1.70s/it]
 28%|██▊       | 5420/19090 [2:38:40<5:59:52,  1.58s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.6957176923751831, 'learning_rate': 4.502529219845153e-05, 'recon_loss': 0.0030669228115584703, 'recon_static': 0.002108689848682843, 'recon_motion': 0.0009582329541444779, 'recon_norm': 1.9230684459209442, 'recon_norm_std': 0.063297622743994, 'target_norm': 1.6204890936613083, 'recon_cos': 0.4695032998919487, 'gradnorm/compressor.query': 0.009457939337993097, 'gradnorm/compressor.layer0': 0.18075679770293335, 'gradnorm/compressor.layer1': 0.42096470711622747, 'gradnorm/compressor.layer2': 0.2652880129842396, 'gradnorm/compressor.layer3': 0.22150431851004812, 'gradnorm/compressor.layer4': 0.14643795549498512, 'gradnorm/compressor.layer5': 0.129306583385802, 'gradnorm/compressor.layer6': 0.10875070313385311, 'gradnorm/compressor.layer7': 0.12493949781936843, 'gradnorm/decoder.layer0': 0.17298087439390314, 'gradnorm/decoder.layer1': 0.16481655967791511, 'gradnorm/decoder.layer2': 0.17866078699732352, 'gradnorm/max_param': 0.2458890974521637, 'epoch': 2.84}

 28%|██▊       | 5420/19090 [2:38:40<5:59:52,  1.58s/it]
 28%|██▊       | 5421/19090 [2:38:43<7:21:47,  1.94s/it]
 28%|██▊       | 5422/19090 [2:38:44<6:39:13,  1.75s/it]
 28%|██▊       | 5423/19090 [2:38:46<6:07:57,  1.62s/it]
 28%|██▊       | 5424/19090 [2:38:48<7:27:09,  1.96s/it]
 28%|██▊       | 5425/19090 [2:38:50<6:42:01,  1.77s/it]
 28%|██▊       | 5426/19090 [2:38:51<6:10:12,  1.63s/it]
 28%|██▊       | 5427/19090 [2:38:53<7:06:39,  1.87s/it]
 28%|██▊       | 5428/19090 [2:38:55<6:27:14,  1.70s/it]
 28%|██▊       | 5429/19090 [2:38:56<5:59:17,  1.58s/it]
 28%|██▊       | 5430/19090 [2:38:59<7:20:52,  1.94s/it]
                                                        
{'loss': 0.0031, 'grad_norm': 0.777490496635437, 'learning_rate': 4.4997892599535676e-05, 'recon_loss': 0.0030532678589224815, 'recon_static': 0.002089920750586316, 'recon_motion': 0.0009633470690459944, 'recon_norm': 1.919979304075241, 'recon_norm_std': 0.06312368651852011, 'target_norm': 1.6216654241085053, 'recon_cos': 0.4718701906502247, 'gradnorm/compressor.query': 0.01103864142410991, 'gradnorm/compressor.layer0': 0.21271744079657703, 'gradnorm/compressor.layer1': 0.4828829396044692, 'gradnorm/compressor.layer2': 0.27927834776136246, 'gradnorm/compressor.layer3': 0.24659207012023446, 'gradnorm/compressor.layer4': 0.1664470060791, 'gradnorm/compressor.layer5': 0.14629952529548607, 'gradnorm/compressor.layer6': 0.11956384797116576, 'gradnorm/compressor.layer7': 0.1381058153901506, 'gradnorm/decoder.layer0': 0.18490915631035706, 'gradnorm/decoder.layer1': 0.17667778057964428, 'gradnorm/decoder.layer2': 0.1981354223933586, 'gradnorm/max_param': 0.2823900580406189, 'epoch': 2.84}

 28%|██▊       | 5430/19090 [2:38:59<7:20:52,  1.94s/it]
 28%|██▊       | 5431/19090 [2:39:00<6:37:47,  1.75s/it]
 28%|██▊       | 5432/19090 [2:39:01<6:07:04,  1.61s/it]
 28%|██▊       | 5433/19090 [2:39:04<7:23:59,  1.95s/it]
 28%|██▊       | 5434/19090 [2:39:05<6:39:04,  1.75s/it]
 28%|██▊       | 5435/19090 [2:39:07<6:07:15,  1.61s/it]
 28%|██▊       | 5436/19090 [2:39:09<7:31:10,  1.98s/it]
 28%|██▊       | 5437/19090 [2:39:11<6:44:05,  1.78s/it]
 28%|██▊       | 5438/19090 [2:39:12<6:11:12,  1.63s/it]
 28%|██▊       | 5439/19090 [2:39:15<7:40:12,  2.02s/it]
 28%|██▊       | 5440/19090 [2:39:16<6:50:03,  1.80s/it]
                                                        
{'loss': 0.0042, 'grad_norm': 0.99717116355896, 'learning_rate': 4.497042613744711e-05, 'recon_loss': 0.00418992256745696, 'recon_static': 0.003236859827302396, 'recon_motion': 0.000953062839107588, 'recon_norm': 2.230798068642616, 'recon_norm_std': 0.057263716403394935, 'target_norm': 1.6183839023113251, 'recon_cos': 0.4104311473667622, 'gradnorm/compressor.query': 0.015976153448051745, 'gradnorm/compressor.layer0': 0.3080111448389509, 'gradnorm/compressor.layer1': 0.5558781538235422, 'gradnorm/compressor.layer2': 0.3678453112854206, 'gradnorm/compressor.layer3': 0.3076855659176117, 'gradnorm/compressor.layer4': 0.22830985659633607, 'gradnorm/compressor.layer5': 0.20561626309962078, 'gradnorm/compressor.layer6': 0.17112797717011774, 'gradnorm/compressor.layer7': 0.18936313379521474, 'gradnorm/decoder.layer0': 0.25608146288886213, 'gradnorm/decoder.layer1': 0.23524753756708977, 'gradnorm/decoder.layer2': 0.2824490735898295, 'gradnorm/max_param': 0.28324106335639954, 'epoch': 2.85}

 28%|██▊       | 5440/19090 [2:39:16<6:50:03,  1.80s/it]
 29%|██▊       | 5441/19090 [2:39:18<6:15:00,  1.65s/it]
 29%|██▊       | 5442/19090 [2:39:21<7:45:48,  2.05s/it]
 29%|██▊       | 5443/19090 [2:39:22<6:53:41,  1.82s/it]
 29%|██▊       | 5444/19090 [2:39:23<6:17:22,  1.66s/it]
 29%|██▊       | 5445/19090 [2:39:26<7:17:49,  1.93s/it]
 29%|██▊       | 5446/19090 [2:39:27<6:34:18,  1.73s/it]
 29%|██▊       | 5447/19090 [2:39:28<6:03:32,  1.60s/it]
 29%|██▊       | 5448/19090 [2:39:31<7:20:19,  1.94s/it]
 29%|██▊       | 5449/19090 [2:39:32<6:36:12,  1.74s/it]
 29%|██▊       | 5450/19090 [2:39:34<6:05:30,  1.61s/it]
                                                        
{'loss': 0.0052, 'grad_norm': 0.9258949160575867, 'learning_rate': 4.4942892904020265e-05, 'recon_loss': 0.005245941586326808, 'recon_static': 0.004290860024048016, 'recon_motion': 0.0009550815389957279, 'recon_norm': 2.493437623977661, 'recon_norm_std': 0.052169555053114894, 'target_norm': 1.6209543168544769, 'recon_cos': 0.36500167548656465, 'gradnorm/compressor.query': 0.013202257213381438, 'gradnorm/compressor.layer0': 0.27197658734588165, 'gradnorm/compressor.layer1': 0.5381794483050771, 'gradnorm/compressor.layer2': 0.34288189468168817, 'gradnorm/compressor.layer3': 0.2750378371348077, 'gradnorm/compressor.layer4': 0.20227555410827622, 'gradnorm/compressor.layer5': 0.18569817648373887, 'gradnorm/compressor.layer6': 0.15578833564859237, 'gradnorm/compressor.layer7': 0.1760662206101602, 'gradnorm/decoder.layer0': 0.23228374351508724, 'gradnorm/decoder.layer1': 0.21903658724505798, 'gradnorm/decoder.layer2': 0.2601607336625937, 'gradnorm/max_param': 0.29217758774757385, 'epoch': 2.85}

 29%|██▊       | 5450/19090 [2:39:34<6:05:30,  1.61s/it]
 29%|██▊       | 5451/19090 [2:39:37<7:47:11,  2.06s/it]
 29%|██▊       | 5452/19090 [2:39:38<6:54:39,  1.82s/it]
 29%|██▊       | 5453/19090 [2:39:39<6:18:11,  1.66s/it]
 29%|██▊       | 5454/19090 [2:39:42<7:12:59,  1.91s/it]
 29%|██▊       | 5455/19090 [2:39:43<6:31:50,  1.72s/it]
 29%|██▊       | 5456/19090 [2:39:44<6:02:38,  1.60s/it]
 29%|██▊       | 5457/19090 [2:39:47<7:06:48,  1.88s/it]
 29%|██▊       | 5458/19090 [2:39:48<6:26:40,  1.70s/it]
 29%|██▊       | 5459/19090 [2:39:49<5:58:28,  1.58s/it]
 29%|██▊       | 5460/19090 [2:39:52<7:19:15,  1.93s/it]
                                                        
{'loss': 0.0045, 'grad_norm': 0.9108772873878479, 'learning_rate': 4.4915292991312784e-05, 'recon_loss': 0.004516544367652387, 'recon_static': 0.0035552855872083454, 'recon_motion': 0.0009612588735762983, 'recon_norm': 2.3163177192211153, 'recon_norm_std': 0.05490497360005975, 'target_norm': 1.6206055939197541, 'recon_cos': 0.3905004791915417, 'gradnorm/compressor.query': 0.014163739269035445, 'gradnorm/compressor.layer0': 0.2668201542496055, 'gradnorm/compressor.layer1': 0.5180555337093964, 'gradnorm/compressor.layer2': 0.32141406517560106, 'gradnorm/compressor.layer3': 0.2865623660903668, 'gradnorm/compressor.layer4': 0.20799884866375198, 'gradnorm/compressor.layer5': 0.18392749644202103, 'gradnorm/compressor.layer6': 0.15445789554582098, 'gradnorm/compressor.layer7': 0.1767581180970306, 'gradnorm/decoder.layer0': 0.2337190190097053, 'gradnorm/decoder.layer1': 0.22364240734056676, 'gradnorm/decoder.layer2': 0.26012163934365284, 'gradnorm/max_param': 0.27394264936447144, 'epoch': 2.86}

 29%|██▊       | 5460/19090 [2:39:52<7:19:15,  1.93s/it]
 29%|██▊       | 5461/19090 [2:39:53<6:35:27,  1.74s/it]
 29%|██▊       | 5462/19090 [2:39:55<6:04:54,  1.61s/it]
 29%|██▊       | 5463/19090 [2:39:58<7:23:24,  1.95s/it]
 29%|██▊       | 5464/19090 [2:39:59<6:38:34,  1.76s/it]
 29%|██▊       | 5465/19090 [2:40:00<6:07:17,  1.62s/it]
 29%|██▊       | 5466/19090 [2:40:03<7:18:25,  1.93s/it]
 29%|██▊       | 5467/19090 [2:40:04<6:45:51,  1.79s/it]
 29%|██▊       | 5468/19090 [2:40:06<6:11:59,  1.64s/it]
 29%|██▊       | 5469/19090 [2:40:08<7:06:34,  1.88s/it]
 29%|██▊       | 5470/19090 [2:40:10<6:53:52,  1.82s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 0.8035340905189514, 'learning_rate': 4.488762649160529e-05, 'recon_loss': 0.004083112970693037, 'recon_static': 0.0031234778580255806, 'recon_motion': 0.0009596352130756714, 'recon_norm': 2.205144685506821, 'recon_norm_std': 0.057262111268937585, 'target_norm': 1.620815673470497, 'recon_cos': 0.4123041622340679, 'gradnorm/compressor.query': 0.011591961804448297, 'gradnorm/compressor.layer0': 0.21713794479093979, 'gradnorm/compressor.layer1': 0.5029688551382125, 'gradnorm/compressor.layer2': 0.3046051916797416, 'gradnorm/compressor.layer3': 0.23766178374208868, 'gradnorm/compressor.layer4': 0.17175296222222838, 'gradnorm/compressor.layer5': 0.14817048828005477, 'gradnorm/compressor.layer6': 0.12295036088141813, 'gradnorm/compressor.layer7': 0.13876232367420943, 'gradnorm/decoder.layer0': 0.19110320878032258, 'gradnorm/decoder.layer1': 0.17923981417228746, 'gradnorm/decoder.layer2': 0.20412242362994734, 'gradnorm/max_param': 0.29804494976997375, 'epoch': 2.87}

 29%|██▊       | 5470/19090 [2:40:10<6:53:52,  1.82s/it]
 29%|██▊       | 5471/19090 [2:40:11<6:17:39,  1.66s/it]
 29%|██▊       | 5472/19090 [2:40:13<6:56:37,  1.84s/it]
 29%|██▊       | 5473/19090 [2:40:15<6:39:23,  1.76s/it]
 29%|██▊       | 5474/19090 [2:40:16<6:07:35,  1.62s/it]
 29%|██▊       | 5475/19090 [2:40:19<7:35:15,  2.01s/it]
 29%|██▊       | 5476/19090 [2:40:20<6:45:46,  1.79s/it]
 29%|██▊       | 5477/19090 [2:40:22<6:11:26,  1.64s/it]
 29%|██▊       | 5478/19090 [2:40:24<7:27:12,  1.97s/it]
 29%|██▊       | 5479/19090 [2:40:26<6:40:10,  1.76s/it]
 29%|██▊       | 5480/19090 [2:40:27<6:08:18,  1.62s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.8713739514350891, 'learning_rate': 4.4859893497401025e-05, 'recon_loss': 0.0042939776787534354, 'recon_static': 0.003345074725802988, 'recon_motion': 0.0009489029267570004, 'recon_norm': 2.2664020895957946, 'recon_norm_std': 0.057856842409819365, 'target_norm': 1.6205789417028427, 'recon_cos': 0.40462245121598245, 'gradnorm/compressor.query': 0.012441246667599022, 'gradnorm/compressor.layer0': 0.24518907868156403, 'gradnorm/compressor.layer1': 0.518070490584601, 'gradnorm/compressor.layer2': 0.3120629353641504, 'gradnorm/compressor.layer3': 0.2736912639739779, 'gradnorm/compressor.layer4': 0.18888591133097363, 'gradnorm/compressor.layer5': 0.1686620487497731, 'gradnorm/compressor.layer6': 0.14244139660472097, 'gradnorm/compressor.layer7': 0.16069563310858154, 'gradnorm/decoder.layer0': 0.22232227206959262, 'gradnorm/decoder.layer1': 0.2129451874273803, 'gradnorm/decoder.layer2': 0.2309169290902023, 'gradnorm/max_param': 0.2952005863189697, 'epoch': 2.87}

 29%|██▊       | 5480/19090 [2:40:27<6:08:18,  1.62s/it]
 29%|██▊       | 5481/19090 [2:40:29<7:17:15,  1.93s/it]
 29%|██▊       | 5482/19090 [2:40:31<6:49:43,  1.81s/it]
 29%|██▊       | 5483/19090 [2:40:32<6:14:47,  1.65s/it]
 29%|██▊       | 5484/19090 [2:40:35<7:21:55,  1.95s/it]
 29%|██▊       | 5485/19090 [2:40:37<7:02:01,  1.86s/it]
 29%|██▊       | 5486/19090 [2:40:38<6:23:17,  1.69s/it]
 29%|██▊       | 5487/19090 [2:40:40<7:06:26,  1.88s/it]
 29%|██▊       | 5488/19090 [2:40:42<6:48:50,  1.80s/it]
 29%|██▉       | 5489/19090 [2:40:43<6:14:19,  1.65s/it]
 29%|██▉       | 5490/19090 [2:40:46<7:16:26,  1.93s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 0.8915979862213135, 'learning_rate': 4.483209410142556e-05, 'recon_loss': 0.004131739481817931, 'recon_static': 0.0031769110006280243, 'recon_motion': 0.0009548285015625879, 'recon_norm': 2.218080300092697, 'recon_norm_std': 0.05763939209282398, 'target_norm': 1.619776576757431, 'recon_cos': 0.40599223002791407, 'gradnorm/compressor.query': 0.013079591758646144, 'gradnorm/compressor.layer0': 0.2612554662990444, 'gradnorm/compressor.layer1': 0.5312707560355331, 'gradnorm/compressor.layer2': 0.337214372004269, 'gradnorm/compressor.layer3': 0.27168169329387537, 'gradnorm/compressor.layer4': 0.19889278484654888, 'gradnorm/compressor.layer5': 0.1692087071680077, 'gradnorm/compressor.layer6': 0.14182601765849678, 'gradnorm/compressor.layer7': 0.16274925702301932, 'gradnorm/decoder.layer0': 0.21363325694071592, 'gradnorm/decoder.layer1': 0.20377551049273956, 'gradnorm/decoder.layer2': 0.23407865704080716, 'gradnorm/max_param': 0.2963849902153015, 'epoch': 2.88}

 29%|██▉       | 5490/19090 [2:40:46<7:16:26,  1.93s/it]
 29%|██▉       | 5491/19090 [2:40:47<6:57:23,  1.84s/it]
 29%|██▉       | 5492/19090 [2:40:49<6:19:47,  1.68s/it]
 29%|██▉       | 5493/19090 [2:40:51<7:37:59,  2.02s/it]
 29%|██▉       | 5494/19090 [2:40:53<6:48:13,  1.80s/it]
 29%|██▉       | 5495/19090 [2:40:54<6:14:01,  1.65s/it]
 29%|██▉       | 5496/19090 [2:40:57<7:36:34,  2.02s/it]
 29%|██▉       | 5497/19090 [2:40:58<6:56:47,  1.84s/it]
 29%|██▉       | 5498/19090 [2:41:00<6:19:31,  1.68s/it]
 29%|██▉       | 5499/19090 [2:41:02<7:20:11,  1.94s/it]
 29%|██▉       | 5500/19090 [2:41:04<6:36:23,  1.75s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.9890919327735901, 'learning_rate': 4.480422839662648e-05, 'recon_loss': 0.004286747385049239, 'recon_static': 0.0033234940841794015, 'recon_motion': 0.0009632532222894951, 'recon_norm': 2.256129467487335, 'recon_norm_std': 0.05472076777368784, 'target_norm': 1.6207666128873826, 'recon_cos': 0.40080008581280707, 'gradnorm/compressor.query': 0.015950793410748863, 'gradnorm/compressor.layer0': 0.30486259564482643, 'gradnorm/compressor.layer1': 0.5698473091317675, 'gradnorm/compressor.layer2': 0.35475523045711693, 'gradnorm/compressor.layer3': 0.30665745051199617, 'gradnorm/compressor.layer4': 0.2203078739546326, 'gradnorm/compressor.layer5': 0.20147968557461432, 'gradnorm/compressor.layer6': 0.16716667451122408, 'gradnorm/compressor.layer7': 0.186312561264847, 'gradnorm/decoder.layer0': 0.25004393966297267, 'gradnorm/decoder.layer1': 0.22956135295983282, 'gradnorm/decoder.layer2': 0.2710795368569209, 'gradnorm/max_param': 0.29791730642318726, 'epoch': 2.88}

 29%|██▉       | 5500/19090 [2:41:04<6:36:23,  1.75s/it]
 29%|██▉       | 5501/19090 [2:41:08<9:14:48,  2.45s/it]
 29%|██▉       | 5502/19090 [2:41:09<7:55:55,  2.10s/it]
 29%|██▉       | 5503/19090 [2:41:10<7:00:15,  1.86s/it]
 29%|██▉       | 5504/19090 [2:41:11<6:21:44,  1.69s/it]
 29%|██▉       | 5505/19090 [2:41:13<5:55:11,  1.57s/it]
 29%|██▉       | 5506/19090 [2:41:14<5:35:53,  1.48s/it]
 29%|██▉       | 5507/19090 [2:41:15<5:22:14,  1.42s/it]
 29%|██▉       | 5508/19090 [2:41:17<5:13:16,  1.38s/it]
 29%|██▉       | 5509/19090 [2:41:18<5:06:30,  1.35s/it]
 29%|██▉       | 5510/19090 [2:41:19<5:02:06,  1.33s/it]
                                                        
{'loss': 0.0046, 'grad_norm': 0.8938696980476379, 'learning_rate': 4.477629647617307e-05, 'recon_loss': 0.004562894918490201, 'recon_static': 0.0036092941998504104, 'recon_motion': 0.0009536006415146403, 'recon_norm': 2.3303393363952636, 'recon_norm_std': 0.053972738515585664, 'target_norm': 1.6195434659719468, 'recon_cos': 0.3894917756319046, 'gradnorm/compressor.query': 0.012641641985849616, 'gradnorm/compressor.layer0': 0.2477792211943391, 'gradnorm/compressor.layer1': 0.5362637052585287, 'gradnorm/compressor.layer2': 0.32285069617255, 'gradnorm/compressor.layer3': 0.26706205847677117, 'gradnorm/compressor.layer4': 0.19944810754163161, 'gradnorm/compressor.layer5': 0.17632129152279621, 'gradnorm/compressor.layer6': 0.14742034707139293, 'gradnorm/compressor.layer7': 0.1656077591901541, 'gradnorm/decoder.layer0': 0.2337446876549504, 'gradnorm/decoder.layer1': 0.20517134591292135, 'gradnorm/decoder.layer2': 0.2399081286657053, 'gradnorm/max_param': 0.30502182245254517, 'epoch': 2.89}

 29%|██▉       | 5510/19090 [2:41:19<5:02:06,  1.33s/it]
 29%|██▉       | 5511/19090 [2:41:20<4:59:00,  1.32s/it]
 29%|██▉       | 5512/19090 [2:41:22<4:56:28,  1.31s/it]
 29%|██▉       | 5513/19090 [2:41:23<4:55:13,  1.30s/it]
 29%|██▉       | 5514/19090 [2:41:25<5:25:18,  1.44s/it]
 29%|██▉       | 5515/19090 [2:41:27<6:24:02,  1.70s/it]
 29%|██▉       | 5516/19090 [2:41:29<6:09:59,  1.64s/it]
 29%|██▉       | 5517/19090 [2:41:30<6:22:03,  1.69s/it]
 29%|██▉       | 5518/19090 [2:41:33<6:59:25,  1.85s/it]
 29%|██▉       | 5519/19090 [2:41:34<6:27:12,  1.71s/it]
 29%|██▉       | 5520/19090 [2:41:36<6:19:54,  1.68s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.8751580119132996, 'learning_rate': 4.474829843345602e-05, 'recon_loss': 0.0038653643045108766, 'recon_static': 0.0029075435653794558, 'recon_motion': 0.000957820778421592, 'recon_norm': 2.15042382478714, 'recon_norm_std': 0.057834033947438, 'target_norm': 1.6207007765769958, 'recon_cos': 0.42197628766298295, 'gradnorm/compressor.query': 0.012256368829769597, 'gradnorm/compressor.layer0': 0.247309934192581, 'gradnorm/compressor.layer1': 0.537407365845151, 'gradnorm/compressor.layer2': 0.31052663725161905, 'gradnorm/compressor.layer3': 0.2808494099214021, 'gradnorm/compressor.layer4': 0.18765429267847167, 'gradnorm/compressor.layer5': 0.163686975360669, 'gradnorm/compressor.layer6': 0.13774423203707004, 'gradnorm/compressor.layer7': 0.15899686904852034, 'gradnorm/decoder.layer0': 0.2090177509359154, 'gradnorm/decoder.layer1': 0.20148328178082556, 'gradnorm/decoder.layer2': 0.22345965148920627, 'gradnorm/max_param': 0.3096868693828583, 'epoch': 2.89}

 29%|██▉       | 5520/19090 [2:41:36<6:19:54,  1.68s/it]
 29%|██▉       | 5521/19090 [2:41:38<7:20:29,  1.95s/it]
 29%|██▉       | 5522/19090 [2:41:40<6:38:36,  1.76s/it]
 29%|██▉       | 5523/19090 [2:41:41<6:20:00,  1.68s/it]
 29%|██▉       | 5524/19090 [2:41:44<7:22:52,  1.96s/it]
 29%|██▉       | 5525/19090 [2:41:45<6:39:27,  1.77s/it]
 29%|██▉       | 5526/19090 [2:41:46<6:16:33,  1.67s/it]
 29%|██▉       | 5527/19090 [2:41:49<7:01:39,  1.87s/it]
 29%|██▉       | 5528/19090 [2:41:50<6:28:27,  1.72s/it]
 29%|██▉       | 5529/19090 [2:41:52<6:37:39,  1.76s/it]
 29%|██▉       | 5530/19090 [2:41:54<6:49:37,  1.81s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 0.8316292762756348, 'learning_rate': 4.472023436208709e-05, 'recon_loss': 0.0037476519239135085, 'recon_static': 0.0027840416063554586, 'recon_motion': 0.0009636103350203484, 'recon_norm': 2.11135296523571, 'recon_norm_std': 0.058281924575567245, 'target_norm': 1.6217808187007905, 'recon_cos': 0.42676766589283943, 'gradnorm/compressor.query': 0.010751108575219713, 'gradnorm/compressor.layer0': 0.20821063581020208, 'gradnorm/compressor.layer1': 0.5539577411774216, 'gradnorm/compressor.layer2': 0.2986728788534103, 'gradnorm/compressor.layer3': 0.2666060199295054, 'gradnorm/compressor.layer4': 0.16557782733959236, 'gradnorm/compressor.layer5': 0.13797847229690083, 'gradnorm/compressor.layer6': 0.11531115253302157, 'gradnorm/compressor.layer7': 0.13346717565907984, 'gradnorm/decoder.layer0': 0.19372841268687319, 'gradnorm/decoder.layer1': 0.1719945253572419, 'gradnorm/decoder.layer2': 0.190550868284604, 'gradnorm/max_param': 0.3423934280872345, 'epoch': 2.9}

 29%|██▉       | 5530/19090 [2:41:54<6:49:37,  1.81s/it]
 29%|██▉       | 5531/19090 [2:41:56<6:40:35,  1.77s/it]
 29%|██▉       | 5532/19090 [2:41:57<6:37:44,  1.76s/it]
 29%|██▉       | 5533/19090 [2:41:59<6:50:14,  1.82s/it]
 29%|██▉       | 5534/19090 [2:42:01<6:44:02,  1.79s/it]
 29%|██▉       | 5535/19090 [2:42:03<6:29:52,  1.73s/it]
 29%|██▉       | 5536/19090 [2:42:05<7:05:47,  1.88s/it]
 29%|██▉       | 5537/19090 [2:42:06<6:27:18,  1.71s/it]
 29%|██▉       | 5538/19090 [2:42:08<6:32:37,  1.74s/it]
 29%|██▉       | 5539/19090 [2:42:10<7:03:24,  1.87s/it]
 29%|██▉       | 5540/19090 [2:42:12<6:32:15,  1.74s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 0.9196954369544983, 'learning_rate': 4.469210435589879e-05, 'recon_loss': 0.0032961566699668764, 'recon_static': 0.002339087409200147, 'recon_motion': 0.0009570692927809432, 'recon_norm': 1.994850516319275, 'recon_norm_std': 0.0613321197219193, 'target_norm': 1.6212552040815353, 'recon_cos': 0.45601586624979973, 'gradnorm/compressor.query': 0.012843349810718281, 'gradnorm/compressor.layer0': 0.2509226165861244, 'gradnorm/compressor.layer1': 0.5960166731332796, 'gradnorm/compressor.layer2': 0.3493014316692494, 'gradnorm/compressor.layer3': 0.2713138950701324, 'gradnorm/compressor.layer4': 0.18447057865198682, 'gradnorm/compressor.layer5': 0.15938006117208217, 'gradnorm/compressor.layer6': 0.13453381407461898, 'gradnorm/compressor.layer7': 0.15283238468799612, 'gradnorm/decoder.layer0': 0.21418078175803879, 'gradnorm/decoder.layer1': 0.19496665852949352, 'gradnorm/decoder.layer2': 0.2169796714297641, 'gradnorm/max_param': 0.35920923948287964, 'epoch': 2.9}

 29%|██▉       | 5540/19090 [2:42:12<6:32:15,  1.74s/it]
 29%|██▉       | 5541/19090 [2:42:13<6:35:11,  1.75s/it]
 29%|██▉       | 5542/19090 [2:42:16<7:29:52,  1.99s/it]
 29%|██▉       | 5543/19090 [2:42:17<6:41:59,  1.78s/it]
 29%|██▉       | 5544/19090 [2:42:19<6:21:12,  1.69s/it]
 29%|██▉       | 5545/19090 [2:42:21<7:21:15,  1.95s/it]
 29%|██▉       | 5546/19090 [2:42:22<6:37:03,  1.76s/it]
 29%|██▉       | 5547/19090 [2:42:24<6:25:28,  1.71s/it]
 29%|██▉       | 5548/19090 [2:42:27<7:20:29,  1.95s/it]
 29%|██▉       | 5549/19090 [2:42:28<6:36:40,  1.76s/it]
 29%|██▉       | 5550/19090 [2:42:30<6:33:12,  1.74s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.9900282025337219, 'learning_rate': 4.466390850894414e-05, 'recon_loss': 0.00444194859592244, 'recon_static': 0.0034869207709562035, 'recon_motion': 0.0009550277609378099, 'recon_norm': 2.2974149763584135, 'recon_norm_std': 0.05453208219259977, 'target_norm': 1.6207966446876525, 'recon_cos': 0.3925709009170532, 'gradnorm/compressor.query': 0.015680403941737888, 'gradnorm/compressor.layer0': 0.2917875255225673, 'gradnorm/compressor.layer1': 0.6175688760391815, 'gradnorm/compressor.layer2': 0.35337405241551445, 'gradnorm/compressor.layer3': 0.3050833600287455, 'gradnorm/compressor.layer4': 0.2130803916062596, 'gradnorm/compressor.layer5': 0.18037243580796505, 'gradnorm/compressor.layer6': 0.15024627222849388, 'gradnorm/compressor.layer7': 0.1732286176990138, 'gradnorm/decoder.layer0': 0.23146586588446455, 'gradnorm/decoder.layer1': 0.22178384459763584, 'gradnorm/decoder.layer2': 0.24928985257062247, 'gradnorm/max_param': 0.3565080761909485, 'epoch': 2.91}

 29%|██▉       | 5550/19090 [2:42:30<6:33:12,  1.74s/it]
 29%|██▉       | 5551/19090 [2:42:32<7:17:03,  1.94s/it]
 29%|██▉       | 5552/19090 [2:42:33<6:32:57,  1.74s/it]
 29%|██▉       | 5553/19090 [2:42:35<6:33:36,  1.74s/it]
 29%|██▉       | 5554/19090 [2:42:37<6:57:57,  1.85s/it]
 29%|██▉       | 5555/19090 [2:42:38<6:20:15,  1.69s/it]
 29%|██▉       | 5556/19090 [2:42:40<6:38:37,  1.77s/it]
 29%|██▉       | 5557/19090 [2:42:43<7:08:19,  1.90s/it]
 29%|██▉       | 5558/19090 [2:42:44<6:45:26,  1.80s/it]
 29%|██▉       | 5559/19090 [2:42:46<6:42:13,  1.78s/it]
 29%|██▉       | 5560/19090 [2:42:48<7:11:19,  1.91s/it]
                                                        
{'loss': 0.0044, 'grad_norm': 0.970079779624939, 'learning_rate': 4.463564691549623e-05, 'recon_loss': 0.004452810517977923, 'recon_static': 0.003489692381117493, 'recon_motion': 0.0009631181848817505, 'recon_norm': 2.3004514038562776, 'recon_norm_std': 0.054116493463516234, 'target_norm': 1.6231303691864014, 'recon_cos': 0.3931502282619476, 'gradnorm/compressor.query': 0.014748950667291117, 'gradnorm/compressor.layer0': 0.28290694962127116, 'gradnorm/compressor.layer1': 0.6098242839625901, 'gradnorm/compressor.layer2': 0.36084211501050667, 'gradnorm/compressor.layer3': 0.2950084761234306, 'gradnorm/compressor.layer4': 0.2027386236756239, 'gradnorm/compressor.layer5': 0.17769399121276064, 'gradnorm/compressor.layer6': 0.14524346219054848, 'gradnorm/compressor.layer7': 0.16552325330873355, 'gradnorm/decoder.layer0': 0.22295494289460427, 'gradnorm/decoder.layer1': 0.20933038819864894, 'gradnorm/decoder.layer2': 0.23870167616424273, 'gradnorm/max_param': 0.3537043631076813, 'epoch': 2.91}

 29%|██▉       | 5560/19090 [2:42:48<7:11:19,  1.91s/it]
 29%|██▉       | 5561/19090 [2:42:49<6:29:52,  1.73s/it]
 29%|██▉       | 5562/19090 [2:42:51<6:29:11,  1.73s/it]
 29%|██▉       | 5563/19090 [2:42:54<7:22:14,  1.96s/it]
 29%|██▉       | 5564/19090 [2:42:55<6:36:24,  1.76s/it]
 29%|██▉       | 5565/19090 [2:42:57<6:22:31,  1.70s/it]
 29%|██▉       | 5566/19090 [2:42:59<7:00:24,  1.87s/it]
 29%|██▉       | 5567/19090 [2:43:00<6:23:44,  1.70s/it]
 29%|██▉       | 5568/19090 [2:43:02<6:47:40,  1.81s/it]
 29%|██▉       | 5569/19090 [2:43:04<6:59:44,  1.86s/it]
 29%|██▉       | 5570/19090 [2:43:05<6:21:07,  1.69s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.998092770576477, 'learning_rate': 4.460731967004802e-05, 'recon_loss': 0.004256363800959661, 'recon_static': 0.003309506591176614, 'recon_motion': 0.0009468570468015968, 'recon_norm': 2.253069579601288, 'recon_norm_std': 0.056389189697802065, 'target_norm': 1.6214533150196075, 'recon_cos': 0.4022008813917637, 'gradnorm/compressor.query': 0.01568326758672626, 'gradnorm/compressor.layer0': 0.29260755525887744, 'gradnorm/compressor.layer1': 0.6283458084374004, 'gradnorm/compressor.layer2': 0.37209984769764226, 'gradnorm/compressor.layer3': 0.29768406895677135, 'gradnorm/compressor.layer4': 0.20926756700258217, 'gradnorm/compressor.layer5': 0.178006686687781, 'gradnorm/compressor.layer6': 0.15112741615006275, 'gradnorm/compressor.layer7': 0.17065351647019844, 'gradnorm/decoder.layer0': 0.22365418377836352, 'gradnorm/decoder.layer1': 0.22409509578668316, 'gradnorm/decoder.layer2': 0.24647576336696497, 'gradnorm/max_param': 0.36531978845596313, 'epoch': 2.92}

 29%|██▉       | 5570/19090 [2:43:05<6:21:07,  1.69s/it]
 29%|██▉       | 5571/19090 [2:43:07<6:24:49,  1.71s/it]
 29%|██▉       | 5572/19090 [2:43:09<7:01:11,  1.87s/it]
 29%|██▉       | 5573/19090 [2:43:11<6:45:41,  1.80s/it]
 29%|██▉       | 5574/19090 [2:43:13<6:50:58,  1.82s/it]
 29%|██▉       | 5575/19090 [2:43:15<6:33:41,  1.75s/it]
 29%|██▉       | 5576/19090 [2:43:16<6:42:03,  1.79s/it]
 29%|██▉       | 5577/19090 [2:43:18<6:24:29,  1.71s/it]
 29%|██▉       | 5578/19090 [2:43:20<6:52:17,  1.83s/it]
 29%|██▉       | 5579/19090 [2:43:21<6:27:51,  1.72s/it]
 29%|██▉       | 5580/19090 [2:43:23<6:18:10,  1.68s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 0.9912829399108887, 'learning_rate': 4.4578926867311965e-05, 'recon_loss': 0.004047858703415841, 'recon_static': 0.0030959087947849185, 'recon_motion': 0.000951950013404712, 'recon_norm': 2.198461836576462, 'recon_norm_std': 0.05772940246388316, 'target_norm': 1.6202964305877685, 'recon_cos': 0.41338675990700724, 'gradnorm/compressor.query': 0.014167049185449605, 'gradnorm/compressor.layer0': 0.26327091389184654, 'gradnorm/compressor.layer1': 0.6718975994833685, 'gradnorm/compressor.layer2': 0.35290952750524773, 'gradnorm/compressor.layer3': 0.2959688436511857, 'gradnorm/compressor.layer4': 0.19289159545098103, 'gradnorm/compressor.layer5': 0.1641110337086482, 'gradnorm/compressor.layer6': 0.1380350170105109, 'gradnorm/compressor.layer7': 0.15956878497957927, 'gradnorm/decoder.layer0': 0.22201950646201699, 'gradnorm/decoder.layer1': 0.20889905275202672, 'gradnorm/decoder.layer2': 0.21898297267577768, 'gradnorm/max_param': 0.41737496852874756, 'epoch': 2.92}

 29%|██▉       | 5580/19090 [2:43:23<6:18:10,  1.68s/it]
 29%|██▉       | 5581/19090 [2:43:26<7:19:54,  1.95s/it]
 29%|██▉       | 5582/19090 [2:43:27<6:39:00,  1.77s/it]
 29%|██▉       | 5583/19090 [2:43:28<6:06:09,  1.63s/it]
 29%|██▉       | 5584/19090 [2:43:31<7:21:11,  1.96s/it]
 29%|██▉       | 5585/19090 [2:43:33<6:49:47,  1.82s/it]
 29%|██▉       | 5586/19090 [2:43:34<6:13:59,  1.66s/it]
 29%|██▉       | 5587/19090 [2:43:36<7:16:43,  1.94s/it]
 29%|██▉       | 5588/19090 [2:43:38<7:10:14,  1.91s/it]
 29%|██▉       | 5589/19090 [2:43:40<6:28:08,  1.72s/it]
 29%|██▉       | 5590/19090 [2:43:42<6:50:57,  1.83s/it]
                                                        
{'loss': 0.0043, 'grad_norm': 0.964535117149353, 'learning_rate': 4.4550468602219716e-05, 'recon_loss': 0.0043309700326062735, 'recon_static': 0.003384519537212327, 'recon_motion': 0.000946450512856245, 'recon_norm': 2.2732268631458283, 'recon_norm_std': 0.054944674205034974, 'target_norm': 1.617579048871994, 'recon_cos': 0.39900661185383796, 'gradnorm/compressor.query': 0.014232821579623878, 'gradnorm/compressor.layer0': 0.27093961659872967, 'gradnorm/compressor.layer1': 0.6372919369543508, 'gradnorm/compressor.layer2': 0.3493287862536072, 'gradnorm/compressor.layer3': 0.28675111489216415, 'gradnorm/compressor.layer4': 0.19365197016708344, 'gradnorm/compressor.layer5': 0.16685200546190845, 'gradnorm/compressor.layer6': 0.13721733461838925, 'gradnorm/compressor.layer7': 0.15810284556426987, 'gradnorm/decoder.layer0': 0.21710425572298475, 'gradnorm/decoder.layer1': 0.19600505114634803, 'gradnorm/decoder.layer2': 0.22715905181934826, 'gradnorm/max_param': 0.3822256624698639, 'epoch': 2.93}

 29%|██▉       | 5590/19090 [2:43:42<6:50:57,  1.83s/it]
 29%|██▉       | 5591/19090 [2:43:44<6:57:53,  1.86s/it]
 29%|██▉       | 5592/19090 [2:43:45<6:19:35,  1.69s/it]
 29%|██▉       | 5593/19090 [2:43:47<6:45:15,  1.80s/it]
 29%|██▉       | 5594/19090 [2:43:49<6:58:50,  1.86s/it]
 29%|██▉       | 5595/19090 [2:43:50<6:20:27,  1.69s/it]
 29%|██▉       | 5596/19090 [2:43:52<6:32:22,  1.74s/it]
 29%|██▉       | 5597/19090 [2:43:54<6:50:55,  1.83s/it]
 29%|██▉       | 5598/19090 [2:43:55<6:14:24,  1.67s/it]
 29%|██▉       | 5599/19090 [2:43:58<6:51:25,  1.83s/it]
 29%|██▉       | 5600/19090 [2:43:59<6:35:41,  1.76s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 0.8884871006011963, 'learning_rate': 4.452194496992179e-05, 'recon_loss': 0.003919934865552932, 'recon_static': 0.0029727460583671927, 'recon_motion': 0.0009471890371059999, 'recon_norm': 2.163668930530548, 'recon_norm_std': 0.056737899873405696, 'target_norm': 1.6181231796741486, 'recon_cos': 0.4193216092884541, 'gradnorm/compressor.query': 0.011502617276424404, 'gradnorm/compressor.layer0': 0.22134028092008204, 'gradnorm/compressor.layer1': 0.5988470195220571, 'gradnorm/compressor.layer2': 0.33088822807599966, 'gradnorm/compressor.layer3': 0.2666959770058698, 'gradnorm/compressor.layer4': 0.16826345406299995, 'gradnorm/compressor.layer5': 0.14757477869104138, 'gradnorm/compressor.layer6': 0.1220798832002528, 'gradnorm/compressor.layer7': 0.14126975547997406, 'gradnorm/decoder.layer0': 0.2061033507822946, 'gradnorm/decoder.layer1': 0.1842234647469463, 'gradnorm/decoder.layer2': 0.19921227051912657, 'gradnorm/max_param': 0.37554457783699036, 'epoch': 2.93}

 29%|██▉       | 5600/19090 [2:43:59<6:35:41,  1.76s/it]
 29%|██▉       | 5601/19090 [2:44:00<6:04:11,  1.62s/it]
 29%|██▉       | 5602/19090 [2:44:03<6:35:53,  1.76s/it]
 29%|██▉       | 5603/19090 [2:44:05<7:08:47,  1.91s/it]
 29%|██▉       | 5604/19090 [2:44:06<6:27:54,  1.73s/it]
 29%|██▉       | 5605/19090 [2:44:08<6:56:03,  1.85s/it]
 29%|██▉       | 5606/19090 [2:44:10<7:00:22,  1.87s/it]
 29%|██▉       | 5607/19090 [2:44:11<6:21:15,  1.70s/it]
 29%|██▉       | 5608/19090 [2:44:14<6:47:24,  1.81s/it]
 29%|██▉       | 5609/19090 [2:44:15<6:34:14,  1.75s/it]
 29%|██▉       | 5610/19090 [2:44:16<6:02:57,  1.62s/it]
                                                        
{'loss': 0.0041, 'grad_norm': 1.0461432933807373, 'learning_rate': 4.449335606578725e-05, 'recon_loss': 0.004086258850293234, 'recon_static': 0.003128864843165502, 'recon_motion': 0.0009573941104463301, 'recon_norm': 2.2071312725543977, 'recon_norm_std': 0.05640745935961604, 'target_norm': 1.62266047000885, 'recon_cos': 0.41104049012064936, 'gradnorm/compressor.query': 0.015009767332089339, 'gradnorm/compressor.layer0': 0.2905485898752217, 'gradnorm/compressor.layer1': 0.6812728939234238, 'gradnorm/compressor.layer2': 0.3984382105802296, 'gradnorm/compressor.layer3': 0.3036317507431918, 'gradnorm/compressor.layer4': 0.21408230488065974, 'gradnorm/compressor.layer5': 0.1810642758395273, 'gradnorm/compressor.layer6': 0.15262830885589448, 'gradnorm/compressor.layer7': 0.168196628930591, 'gradnorm/decoder.layer0': 0.237980097276603, 'gradnorm/decoder.layer1': 0.21861852373244345, 'gradnorm/decoder.layer2': 0.24508070199260332, 'gradnorm/max_param': 0.4070318937301636, 'epoch': 2.94}

 29%|██▉       | 5610/19090 [2:44:16<6:02:57,  1.62s/it]
 29%|██▉       | 5611/19090 [2:44:19<7:16:29,  1.94s/it]
 29%|██▉       | 5612/19090 [2:44:20<6:32:16,  1.75s/it]
 29%|██▉       | 5613/19090 [2:44:22<6:01:25,  1.61s/it]
 29%|██▉       | 5614/19090 [2:44:24<7:17:03,  1.95s/it]
 29%|██▉       | 5615/19090 [2:44:26<6:32:13,  1.75s/it]
 29%|██▉       | 5616/19090 [2:44:27<6:01:16,  1.61s/it]
 29%|██▉       | 5617/19090 [2:44:30<7:01:01,  1.87s/it]
 29%|██▉       | 5618/19090 [2:44:31<6:37:50,  1.77s/it]
 29%|██▉       | 5619/19090 [2:44:32<6:05:24,  1.63s/it]
 29%|██▉       | 5620/19090 [2:44:35<7:07:58,  1.91s/it]
                                                        
{'loss': 0.0039, 'grad_norm': 1.0455033779144287, 'learning_rate': 4.4464701985403434e-05, 'recon_loss': 0.003898843372007832, 'recon_static': 0.002945333399111405, 'recon_motion': 0.0009535100049106404, 'recon_norm': 2.1582369089126585, 'recon_norm_std': 0.05761419683694839, 'target_norm': 1.62097307741642, 'recon_cos': 0.41915545910596846, 'gradnorm/compressor.query': 0.014511276869893982, 'gradnorm/compressor.layer0': 0.2833719605290007, 'gradnorm/compressor.layer1': 0.7107711429678201, 'gradnorm/compressor.layer2': 0.38080700245252147, 'gradnorm/compressor.layer3': 0.28819782218077167, 'gradnorm/compressor.layer4': 0.20495647273608905, 'gradnorm/compressor.layer5': 0.17811066753883256, 'gradnorm/compressor.layer6': 0.1461319498619431, 'gradnorm/compressor.layer7': 0.1697753625386445, 'gradnorm/decoder.layer0': 0.2264814642975417, 'gradnorm/decoder.layer1': 0.21925219723177913, 'gradnorm/decoder.layer2': 0.2366966416638638, 'gradnorm/max_param': 0.4366135597229004, 'epoch': 2.94}

 29%|██▉       | 5620/19090 [2:44:35<7:07:58,  1.91s/it]
 29%|██▉       | 5621/19090 [2:44:37<6:59:16,  1.87s/it]
 29%|██▉       | 5622/19090 [2:44:38<6:19:46,  1.69s/it]
 29%|██▉       | 5623/19090 [2:44:40<6:57:34,  1.86s/it]
 29%|██▉       | 5624/19090 [2:44:42<6:55:38,  1.85s/it]
 29%|██▉       | 5625/19090 [2:44:43<6:17:54,  1.68s/it]
 29%|██▉       | 5626/19090 [2:44:45<6:42:46,  1.79s/it]
 29%|██▉       | 5627/19090 [2:44:48<7:21:46,  1.97s/it]
 29%|██▉       | 5628/19090 [2:44:49<6:36:22,  1.77s/it]
 29%|██▉       | 5629/19090 [2:44:50<6:06:33,  1.63s/it]
 29%|██▉       | 5630/19090 [2:44:53<7:29:23,  2.00s/it]
                                                        
{'loss': 0.004, 'grad_norm': 1.037230134010315, 'learning_rate': 4.443598282457556e-05, 'recon_loss': 0.003936365467961877, 'recon_static': 0.0029974760313052685, 'recon_motion': 0.0009388894046423957, 'recon_norm': 2.172002041339874, 'recon_norm_std': 0.05545246647670865, 'target_norm': 1.6161724597215652, 'recon_cos': 0.41814384087920187, 'gradnorm/compressor.query': 0.014755254447644094, 'gradnorm/compressor.layer0': 0.27455810079039566, 'gradnorm/compressor.layer1': 0.7163158746621062, 'gradnorm/compressor.layer2': 0.38107022048990064, 'gradnorm/compressor.layer3': 0.31622345671631896, 'gradnorm/compressor.layer4': 0.18985409924276306, 'gradnorm/compressor.layer5': 0.16202759116004523, 'gradnorm/compressor.layer6': 0.13665668577460144, 'gradnorm/compressor.layer7': 0.155747128942366, 'gradnorm/decoder.layer0': 0.21693697689678018, 'gradnorm/decoder.layer1': 0.20358601851871405, 'gradnorm/decoder.layer2': 0.21951657751879877, 'gradnorm/max_param': 0.4499337077140808, 'epoch': 2.95}

 29%|██▉       | 5630/19090 [2:44:53<7:29:23,  2.00s/it]
 29%|██▉       | 5631/19090 [2:44:55<6:42:35,  1.79s/it]
 30%|██▉       | 5632/19090 [2:44:56<6:08:25,  1.64s/it]
 30%|██▉       | 5633/19090 [2:44:59<7:15:42,  1.94s/it]
 30%|██▉       | 5634/19090 [2:45:00<6:32:09,  1.75s/it]
 30%|██▉       | 5635/19090 [2:45:01<6:10:23,  1.65s/it]
 30%|██▉       | 5636/19090 [2:45:04<7:19:15,  1.96s/it]
 30%|██▉       | 5637/19090 [2:45:05<6:33:52,  1.76s/it]
 30%|██▉       | 5638/19090 [2:45:06<6:02:15,  1.62s/it]
 30%|██▉       | 5639/19090 [2:45:09<7:12:03,  1.93s/it]
 30%|██▉       | 5640/19090 [2:45:10<6:28:47,  1.73s/it]
                                                        
{'loss': 0.0036, 'grad_norm': 0.9326608180999756, 'learning_rate': 4.440719867932645e-05, 'recon_loss': 0.003603855223627761, 'recon_static': 0.0026413870975375174, 'recon_motion': 0.0009624681828427128, 'recon_norm': 2.07341183423996, 'recon_norm_std': 0.05929103847593069, 'target_norm': 1.6218720883131028, 'recon_cos': 0.43872622326016425, 'gradnorm/compressor.query': 0.010477665867172817, 'gradnorm/compressor.layer0': 0.20840023079326686, 'gradnorm/compressor.layer1': 0.6752856647926699, 'gradnorm/compressor.layer2': 0.3505336754398616, 'gradnorm/compressor.layer3': 0.265135271549666, 'gradnorm/compressor.layer4': 0.16062190235616927, 'gradnorm/compressor.layer5': 0.13849459316875484, 'gradnorm/compressor.layer6': 0.11109634582153359, 'gradnorm/compressor.layer7': 0.13187943200261695, 'gradnorm/decoder.layer0': 0.19130596141552267, 'gradnorm/decoder.layer1': 0.17967856426411227, 'gradnorm/decoder.layer2': 0.1831321289998822, 'gradnorm/max_param': 0.4397820234298706, 'epoch': 2.95}

 30%|██▉       | 5640/19090 [2:45:10<6:28:47,  1.73s/it]
 30%|██▉       | 5641/19090 [2:45:12<6:32:36,  1.75s/it]
 30%|██▉       | 5642/19090 [2:45:14<6:58:23,  1.87s/it]
 30%|██▉       | 5643/19090 [2:45:16<6:19:52,  1.69s/it]
 30%|██▉       | 5644/19090 [2:45:18<6:47:37,  1.82s/it]
 30%|██▉       | 5645/19090 [2:45:19<6:35:25,  1.76s/it]
 30%|██▉       | 5646/19090 [2:45:21<6:04:10,  1.63s/it]
 30%|██▉       | 5647/19090 [2:45:23<7:08:43,  1.91s/it]
 30%|██▉       | 5648/19090 [2:45:25<6:26:08,  1.72s/it]
 30%|██▉       | 5649/19090 [2:45:26<5:57:08,  1.59s/it]
 30%|██▉       | 5650/19090 [2:45:29<7:12:54,  1.93s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 1.0481863021850586, 'learning_rate': 4.437834964589623e-05, 'recon_loss': 0.0036854912352282552, 'recon_static': 0.00272828625747934, 'recon_motion': 0.0009572049675625749, 'recon_norm': 2.097451573610306, 'recon_norm_std': 0.0580436572432518, 'target_norm': 1.620019683241844, 'recon_cos': 0.43226738795638087, 'gradnorm/compressor.query': 0.0142395756143626, 'gradnorm/compressor.layer0': 0.28323083939268223, 'gradnorm/compressor.layer1': 0.6689161976513665, 'gradnorm/compressor.layer2': 0.3947182580375369, 'gradnorm/compressor.layer3': 0.32137886346092004, 'gradnorm/compressor.layer4': 0.21680971056083148, 'gradnorm/compressor.layer5': 0.18725297470556568, 'gradnorm/compressor.layer6': 0.15656129702374288, 'gradnorm/compressor.layer7': 0.17523325165904277, 'gradnorm/decoder.layer0': 0.24633820063984535, 'gradnorm/decoder.layer1': 0.22505374571181774, 'gradnorm/decoder.layer2': 0.2512294964229188, 'gradnorm/max_param': 0.39994722604751587, 'epoch': 2.96}

 30%|██▉       | 5650/19090 [2:45:29<7:12:54,  1.93s/it]
 30%|██▉       | 5651/19090 [2:45:30<6:29:11,  1.74s/it]
 30%|██▉       | 5652/19090 [2:45:31<5:58:44,  1.60s/it]
 30%|██▉       | 5653/19090 [2:45:34<7:47:43,  2.09s/it]
 30%|██▉       | 5654/19090 [2:45:36<6:54:40,  1.85s/it]
 30%|██▉       | 5655/19090 [2:45:37<6:16:57,  1.68s/it]
 30%|██▉       | 5656/19090 [2:45:40<7:51:33,  2.11s/it]
 30%|██▉       | 5657/19090 [2:45:41<6:56:26,  1.86s/it]
 30%|██▉       | 5658/19090 [2:45:43<6:17:48,  1.69s/it]
 30%|██▉       | 5659/19090 [2:45:46<7:57:57,  2.14s/it]
 30%|██▉       | 5660/19090 [2:45:47<7:01:03,  1.88s/it]
                                                        
{'loss': 0.0037, 'grad_norm': 1.014497995376587, 'learning_rate': 4.4349435820741934e-05, 'recon_loss': 0.003679309005383402, 'recon_static': 0.002725508564617485, 'recon_motion': 0.0009538004858768545, 'recon_norm': 2.0931448519229887, 'recon_norm_std': 0.0571710810996592, 'target_norm': 1.6202434033155442, 'recon_cos': 0.43244988694787023, 'gradnorm/compressor.query': 0.014225965116156794, 'gradnorm/compressor.layer0': 0.2698819049439525, 'gradnorm/compressor.layer1': 0.6866499674334345, 'gradnorm/compressor.layer2': 0.3660879513537332, 'gradnorm/compressor.layer3': 0.2918414489119465, 'gradnorm/compressor.layer4': 0.19937213901631506, 'gradnorm/compressor.layer5': 0.17380622125702397, 'gradnorm/compressor.layer6': 0.144348015492807, 'gradnorm/compressor.layer7': 0.16415988211804614, 'gradnorm/decoder.layer0': 0.22624451421177963, 'gradnorm/decoder.layer1': 0.21051565126316832, 'gradnorm/decoder.layer2': 0.2285642735138896, 'gradnorm/max_param': 0.42433667182922363, 'epoch': 2.96}

 30%|██▉       | 5660/19090 [2:45:47<7:01:03,  1.88s/it]
 30%|██▉       | 5661/19090 [2:45:48<6:21:21,  1.70s/it]
 30%|██▉       | 5662/19090 [2:45:51<7:04:12,  1.90s/it]
 30%|██▉       | 5663/19090 [2:45:52<6:23:11,  1.71s/it]
 30%|██▉       | 5664/19090 [2:45:53<5:54:41,  1.59s/it]
 30%|██▉       | 5665/19090 [2:45:56<7:24:30,  1.99s/it]
 30%|██▉       | 5666/19090 [2:45:57<6:37:38,  1.78s/it]
 30%|██▉       | 5667/19090 [2:45:59<6:05:44,  1.63s/it]
 30%|██▉       | 5668/19090 [2:46:01<7:15:34,  1.95s/it]
 30%|██▉       | 5669/19090 [2:46:03<6:31:04,  1.75s/it]
 30%|██▉       | 5670/19090 [2:46:04<6:00:07,  1.61s/it]
                                                        
{'loss': 0.0035, 'grad_norm': 0.9759389162063599, 'learning_rate': 4.432045730053728e-05, 'recon_loss': 0.0035065408155787736, 'recon_static': 0.0025580672081559896, 'recon_motion': 0.0009484736787271686, 'recon_norm': 2.050006178021431, 'recon_norm_std': 0.057945546507835385, 'target_norm': 1.619752436876297, 'recon_cos': 0.44128682613372805, 'gradnorm/compressor.query': 0.012852579565008312, 'gradnorm/compressor.layer0': 0.25280626150197827, 'gradnorm/compressor.layer1': 0.6620542115044042, 'gradnorm/compressor.layer2': 0.3656154826955172, 'gradnorm/compressor.layer3': 0.30438545122944155, 'gradnorm/compressor.layer4': 0.1837173816703274, 'gradnorm/compressor.layer5': 0.15534531830882525, 'gradnorm/compressor.layer6': 0.13100652818939182, 'gradnorm/compressor.layer7': 0.1482825662757488, 'gradnorm/decoder.layer0': 0.21138634435033152, 'gradnorm/decoder.layer1': 0.19578196370937748, 'gradnorm/decoder.layer2': 0.20902199272921124, 'gradnorm/max_param': 0.41337850689888, 'epoch': 2.97}

 30%|██▉       | 5670/19090 [2:46:04<6:00:07,  1.61s/it]
 30%|██▉       | 5671/19090 [2:46:07<7:25:05,  1.99s/it]
 30%|██▉       | 5672/19090 [2:46:08<6:38:36,  1.78s/it]
 30%|██▉       | 5673/19090 [2:46:10<6:05:10,  1.63s/it]
 30%|██▉       | 5674/19090 [2:46:12<7:05:53,  1.90s/it]
 30%|██▉       | 5675/19090 [2:46:13<6:24:13,  1.72s/it]
 30%|██▉       | 5676/19090 [2:46:15<5:55:09,  1.59s/it]
 30%|██▉       | 5677/19090 [2:46:17<6:50:22,  1.84s/it]
 30%|██▉       | 5678/19090 [2:46:18<6:24:44,  1.72s/it]
 30%|██▉       | 5679/19090 [2:46:20<5:55:30,  1.59s/it]
 30%|██▉       | 5680/19090 [2:46:22<7:07:29,  1.91s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 1.1140532493591309, 'learning_rate': 4.429141418217223e-05, 'recon_loss': 0.00334664573892951, 'recon_static': 0.0023861481226049364, 'recon_motion': 0.0009604975843103602, 'recon_norm': 2.0029425263404845, 'recon_norm_std': 0.059911100380122664, 'target_norm': 1.622479584813118, 'recon_cos': 0.4542450487613678, 'gradnorm/compressor.query': 0.014417203075499848, 'gradnorm/compressor.layer0': 0.2814872985696489, 'gradnorm/compressor.layer1': 0.7977614240961068, 'gradnorm/compressor.layer2': 0.4142555470564289, 'gradnorm/compressor.layer3': 0.3215835056914057, 'gradnorm/compressor.layer4': 0.19899057897594805, 'gradnorm/compressor.layer5': 0.16644171842729194, 'gradnorm/compressor.layer6': 0.13716855986113502, 'gradnorm/compressor.layer7': 0.16000754733464342, 'gradnorm/decoder.layer0': 0.2242313521030386, 'gradnorm/decoder.layer1': 0.204544663277213, 'gradnorm/decoder.layer2': 0.21538066531051303, 'gradnorm/max_param': 0.5132741332054138, 'epoch': 2.98}

 30%|██▉       | 5680/19090 [2:46:22<7:07:29,  1.91s/it]
 30%|██▉       | 5681/19090 [2:46:24<6:32:42,  1.76s/it]
 30%|██▉       | 5682/19090 [2:46:25<6:01:15,  1.62s/it]
 30%|██▉       | 5683/19090 [2:46:28<7:08:08,  1.92s/it]
 30%|██▉       | 5684/19090 [2:46:29<6:42:30,  1.80s/it]
 30%|██▉       | 5685/19090 [2:46:31<6:07:43,  1.65s/it]
 30%|██▉       | 5686/19090 [2:46:33<6:58:20,  1.87s/it]
 30%|██▉       | 5687/19090 [2:46:35<6:49:39,  1.83s/it]
 30%|██▉       | 5688/19090 [2:46:36<6:13:32,  1.67s/it]
 30%|██▉       | 5689/19090 [2:46:38<6:58:00,  1.87s/it]
 30%|██▉       | 5690/19090 [2:46:40<6:28:00,  1.74s/it]
                                                        
{'loss': 0.0034, 'grad_norm': 1.0695929527282715, 'learning_rate': 4.42623065627528e-05, 'recon_loss': 0.0033839923969935628, 'recon_static': 0.002422664174810052, 'recon_motion': 0.0009613281989004463, 'recon_norm': 2.0094745427370073, 'recon_norm_std': 0.057504831533879044, 'target_norm': 1.6209254413843155, 'recon_cos': 0.44824324250221254, 'gradnorm/compressor.query': 0.013082034726468477, 'gradnorm/compressor.layer0': 0.2619560788309746, 'gradnorm/compressor.layer1': 0.7614775838301556, 'gradnorm/compressor.layer2': 0.4036354330220164, 'gradnorm/compressor.layer3': 0.29800318626486144, 'gradnorm/compressor.layer4': 0.1897429284868871, 'gradnorm/compressor.layer5': 0.16544472199346463, 'gradnorm/compressor.layer6': 0.13361510920295339, 'gradnorm/compressor.layer7': 0.15472599908716045, 'gradnorm/decoder.layer0': 0.22623768081373277, 'gradnorm/decoder.layer1': 0.20502759220062378, 'gradnorm/decoder.layer2': 0.21277736135653139, 'gradnorm/max_param': 0.4877532124519348, 'epoch': 2.98}

 30%|██▉       | 5690/19090 [2:46:40<6:28:00,  1.74s/it]
 30%|██▉       | 5691/19090 [2:46:41<5:58:04,  1.60s/it]
 30%|██▉       | 5692/19090 [2:46:44<7:16:08,  1.95s/it]
 30%|██▉       | 5693/19090 [2:46:45<6:55:30,  1.86s/it]
 30%|██▉       | 5694/19090 [2:46:47<6:17:07,  1.69s/it]
 30%|██▉       | 5695/19090 [2:46:49<7:07:06,  1.91s/it]
 30%|██▉       | 5696/19090 [2:46:51<6:46:38,  1.82s/it]
 30%|██▉       | 5697/19090 [2:46:52<6:10:46,  1.66s/it]
 30%|██▉       | 5698/19090 [2:46:55<7:12:13,  1.94s/it]
 30%|██▉       | 5699/19090 [2:46:56<6:44:35,  1.81s/it]
 30%|██▉       | 5700/19090 [2:46:57<6:11:34,  1.67s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 1.0752356052398682, 'learning_rate': 4.4233134539600626e-05, 'recon_loss': 0.003341915592318401, 'recon_static': 0.0023852356447605417, 'recon_motion': 0.0009566800290485844, 'recon_norm': 1.9974240213632584, 'recon_norm_std': 0.05831954451277852, 'target_norm': 1.6207983076572419, 'recon_cos': 0.4502194426953793, 'gradnorm/compressor.query': 0.013616013505854903, 'gradnorm/compressor.layer0': 0.26444916771501537, 'gradnorm/compressor.layer1': 0.7804411324778283, 'gradnorm/compressor.layer2': 0.3971290796326713, 'gradnorm/compressor.layer3': 0.31148388889601075, 'gradnorm/compressor.layer4': 0.18257658808024177, 'gradnorm/compressor.layer5': 0.15911785633281667, 'gradnorm/compressor.layer6': 0.12894142277908255, 'gradnorm/compressor.layer7': 0.15066856240081913, 'gradnorm/decoder.layer0': 0.20856691716517578, 'gradnorm/decoder.layer1': 0.19734426385083204, 'gradnorm/decoder.layer2': 0.20436775152377298, 'gradnorm/max_param': 0.5027769207954407, 'epoch': 2.99}

 30%|██▉       | 5700/19090 [2:46:57<6:11:34,  1.67s/it]
 30%|██▉       | 5701/19090 [2:47:00<7:10:10,  1.93s/it]
 30%|██▉       | 5702/19090 [2:47:02<6:40:57,  1.80s/it]
 30%|██▉       | 5703/19090 [2:47:03<6:07:28,  1.65s/it]
 30%|██▉       | 5704/19090 [2:47:06<7:28:52,  2.01s/it]
 30%|██▉       | 5705/19090 [2:47:07<6:51:59,  1.85s/it]
 30%|██▉       | 5706/19090 [2:47:08<6:14:31,  1.68s/it]
 30%|██▉       | 5707/19090 [2:47:11<7:08:30,  1.92s/it]
 30%|██▉       | 5708/19090 [2:47:13<6:54:39,  1.86s/it]
 30%|██▉       | 5709/19090 [2:47:14<6:17:00,  1.69s/it]
 30%|██▉       | 5710/19090 [2:47:16<6:50:57,  1.84s/it]
                                                        
{'loss': 0.0033, 'grad_norm': 1.1476410627365112, 'learning_rate': 4.420389821025268e-05, 'recon_loss': 0.00334453386021778, 'recon_static': 0.0023852831858675928, 'recon_motion': 0.0009592505768523552, 'recon_norm': 1.9992127478122712, 'recon_norm_std': 0.05724372901022434, 'target_norm': 1.619749480485916, 'recon_cos': 0.45026651695370673, 'gradnorm/compressor.query': 0.014941803005192865, 'gradnorm/compressor.layer0': 0.2895948945740449, 'gradnorm/compressor.layer1': 0.8237786948802331, 'gradnorm/compressor.layer2': 0.4290156747726477, 'gradnorm/compressor.layer3': 0.32466477427471246, 'gradnorm/compressor.layer4': 0.1990276991338309, 'gradnorm/compressor.layer5': 0.16846485949032453, 'gradnorm/compressor.layer6': 0.14273065090280052, 'gradnorm/compressor.layer7': 0.1626461575315015, 'gradnorm/decoder.layer0': 0.23859454492712592, 'gradnorm/decoder.layer1': 0.21394142186344967, 'gradnorm/decoder.layer2': 0.21772017918228936, 'gradnorm/max_param': 0.5254333019256592, 'epoch': 2.99}

 30%|██▉       | 5710/19090 [2:47:16<6:50:57,  1.84s/it]
 30%|██▉       | 5711/19090 [2:47:18<7:05:28,  1.91s/it]
 30%|██▉       | 5712/19090 [2:47:19<6:23:59,  1.72s/it]
 30%|██▉       | 5713/19090 [2:47:21<6:22:23,  1.72s/it]
 30%|██▉       | 5714/19090 [2:47:23<6:51:08,  1.84s/it]
 30%|██▉       | 5715/19090 [2:47:25<6:14:56,  1.68s/it]
 30%|██▉       | 5716/19090 [2:47:27<6:48:41,  1.83s/it]
 30%|██▉       | 5717/19090 [2:47:29<6:40:48,  1.80s/it]
 30%|██▉       | 5718/19090 [2:47:30<6:07:33,  1.65s/it]
 30%|██▉       | 5719/19090 [2:47:32<7:00:35,  1.89s/it]
 30%|██▉       | 5720/19090 [2:47:34<6:37:46,  1.79s/it]
                                                        
{'loss': 0.003, 'grad_norm': 0.9924050569534302, 'learning_rate': 4.417459767246096e-05, 'recon_loss': 0.0030110681254882365, 'recon_static': 0.0020625080389436333, 'recon_motion': 0.0009485600763582625, 'recon_norm': 1.902650049328804, 'recon_norm_std': 0.060755613073706624, 'target_norm': 1.6190776109695435, 'recon_cos': 0.47341316044330595, 'gradnorm/compressor.query': 0.011830745914191131, 'gradnorm/compressor.layer0': 0.2322170332613011, 'gradnorm/compressor.layer1': 0.7305638868423341, 'gradnorm/compressor.layer2': 0.3452526433340778, 'gradnorm/compressor.layer3': 0.28822585765169373, 'gradnorm/compressor.layer4': 0.1727961596908686, 'gradnorm/compressor.layer5': 0.14673693818404723, 'gradnorm/compressor.layer6': 0.11749767738694708, 'gradnorm/compressor.layer7': 0.1396836389844358, 'gradnorm/decoder.layer0': 0.2036125128203238, 'gradnorm/decoder.layer1': 0.18611490737920422, 'gradnorm/decoder.layer2': 0.18439534891305043, 'gradnorm/max_param': 0.4762941002845764, 'epoch': 3.0}

 30%|██▉       | 5720/19090 [2:47:34<6:37:46,  1.79s/it]
 30%|██▉       | 5721/19090 [2:47:35<6:08:43,  1.65s/it]
 30%|██▉       | 5722/19090 [2:47:38<6:56:51,  1.87s/it]
 30%|██▉       | 5723/19090 [2:47:39<6:18:19,  1.70s/it]
 30%|██▉       | 5724/19090 [2:47:40<5:50:09,  1.57s/it]
 30%|██▉       | 5725/19090 [2:47:41<5:29:43,  1.48s/it]
 30%|██▉       | 5726/19090 [2:47:43<5:15:47,  1.42s/it]
 30%|███       | 5727/19090 [2:47:44<5:35:22,  1.51s/it]
 30%|███       | 5728/19090 [2:47:52<12:56:01,  3.48s/it]
 30%|███       | 5729/19090 [2:47:54<10:29:26,  2.83s/it]
 30%|███       | 5730/19090 [2:47:57<11:26:58,  3.09s/it]
                                                         
{'loss': 0.0032, 'grad_norm': 1.1562058925628662, 'learning_rate': 4.4145233024192124e-05, 'recon_loss': 0.0031716806290205566, 'recon_static': 0.002210867451503873, 'recon_motion': 0.0009608132037101314, 'recon_norm': 1.951593604683876, 'recon_norm_std': 0.059226933680474755, 'target_norm': 1.6232890456914901, 'recon_cos': 0.4639748767018318, 'gradnorm/compressor.query': 0.013628401994123516, 'gradnorm/compressor.layer0': 0.2610254784704886, 'gradnorm/compressor.layer1': 0.8578509949209402, 'gradnorm/compressor.layer2': 0.4560248877777452, 'gradnorm/compressor.layer3': 0.3226324709980921, 'gradnorm/compressor.layer4': 0.1836514078059816, 'gradnorm/compressor.layer5': 0.16095759372823518, 'gradnorm/compressor.layer6': 0.12815175184231262, 'gradnorm/compressor.layer7': 0.14570723059957552, 'gradnorm/decoder.layer0': 0.208417605204136, 'gradnorm/decoder.layer1': 0.20110840397981872, 'gradnorm/decoder.layer2': 0.19839935461749772, 'gradnorm/max_param': 0.5641019344329834, 'epoch': 3.0}

 30%|███       | 5730/19090 [2:47:57<11:26:58,  3.09s/it]
 30%|███       | 5731/19090 [2:47:59<9:27:29,  2.55s/it] 
 30%|███       | 5732/19090 [2:48:00<8:03:41,  2.17s/it]
 30%|███       | 5733/19090 [2:48:03<8:30:32,  2.29s/it]
 30%|███       | 5734/19090 [2:48:04<7:23:30,  1.99s/it]
 30%|███       | 5735/19090 [2:48:05<6:36:51,  1.78s/it]
 30%|███       | 5736/19090 [2:48:07<7:02:44,  1.90s/it]
 30%|███       | 5737/19090 [2:48:09<6:21:57,  1.72s/it]
 30%|███       | 5738/19090 [2:48:10<5:53:23,  1.59s/it]
 30%|███       | 5739/19090 [2:48:12<6:21:36,  1.71s/it]
 30%|███       | 5740/19090 [2:48:13<5:53:16,  1.59s/it]
                                                        
{'loss': 0.0029, 'grad_norm': 1.0663594007492065, 'learning_rate': 4.41158043636272e-05, 'recon_loss': 0.002859105682000518, 'recon_static': 0.001917247666278854, 'recon_motion': 0.000941858124861028, 'recon_norm': 1.8577600091695785, 'recon_norm_std': 0.06231480669230223, 'target_norm': 1.6176128536462784, 'recon_cos': 0.4864471472799778, 'gradnorm/compressor.query': 0.010987216529423858, 'gradnorm/compressor.layer0': 0.22597130560798126, 'gradnorm/compressor.layer1': 0.8065791071970806, 'gradnorm/compressor.layer2': 0.40931317102735765, 'gradnorm/compressor.layer3': 0.3111791711479114, 'gradnorm/compressor.layer4': 0.1639068182761757, 'gradnorm/compressor.layer5': 0.13523660129056694, 'gradnorm/compressor.layer6': 0.10795369284095778, 'gradnorm/compressor.layer7': 0.1274366926001153, 'gradnorm/decoder.layer0': 0.1992923656657798, 'gradnorm/decoder.layer1': 0.1721313255812683, 'gradnorm/decoder.layer2': 0.1691268540870722, 'gradnorm/max_param': 0.5376380085945129, 'epoch': 3.01}

 30%|███       | 5740/19090 [2:48:13<5:53:16,  1.59s/it]
 30%|███       | 5741/19090 [2:48:15<5:33:49,  1.50s/it]
 30%|███       | 5742/19090 [2:48:17<6:38:36,  1.79s/it]
 30%|███       | 5743/19090 [2:48:18<6:04:48,  1.64s/it]
 30%|███       | 5744/19090 [2:48:20<5:41:08,  1.53s/it]
 30%|███       | 5745/19090 [2:48:22<7:04:51,  1.91s/it]
 30%|███       | 5746/19090 [2:48:24<6:23:19,  1.72s/it]
 30%|███       | 5747/19090 [2:48:25<5:54:03,  1.59s/it]
 30%|███       | 5748/19090 [2:48:27<6:31:27,  1.76s/it]
 30%|███       | 5749/19090 [2:48:28<5:59:46,  1.62s/it]
 30%|███       | 5750/19090 [2:48:30<5:38:04,  1.52s/it]
                                                        
{'loss': 0.0027, 'grad_norm': 1.1511605978012085, 'learning_rate': 4.4086311789161225e-05, 'recon_loss': 0.002682916144840419, 'recon_static': 0.0017284404719248414, 'recon_motion': 0.0009544757369440049, 'recon_norm': 1.7942497700452804, 'recon_norm_std': 0.0634085494093597, 'target_norm': 1.6194980829954146, 'recon_cos': 0.5006105192005634, 'gradnorm/compressor.query': 0.012734983888941849, 'gradnorm/compressor.layer0': 0.2687326153965346, 'gradnorm/compressor.layer1': 0.838702220039072, 'gradnorm/compressor.layer2': 0.4340393594680073, 'gradnorm/compressor.layer3': 0.32868300402537287, 'gradnorm/compressor.layer4': 0.19363155859409809, 'gradnorm/compressor.layer5': 0.16835601099640748, 'gradnorm/compressor.layer6': 0.13952525915204853, 'gradnorm/compressor.layer7': 0.1566988220932538, 'gradnorm/decoder.layer0': 0.2306206412795125, 'gradnorm/decoder.layer1': 0.20818371565994429, 'gradnorm/decoder.layer2': 0.21578246427742664, 'gradnorm/max_param': 0.5438373684883118, 'epoch': 3.01}

 30%|███       | 5750/19090 [2:48:30<5:38:04,  1.52s/it]
 30%|███       | 5751/19090 [2:48:32<6:35:16,  1.78s/it]
 30%|███       | 5752/19090 [2:48:33<6:02:36,  1.63s/it]
 30%|███       | 5753/19090 [2:48:35<5:39:55,  1.53s/it]
 30%|███       | 5754/19090 [2:48:37<6:01:58,  1.63s/it]
 30%|███       | 5755/19090 [2:48:38<5:39:24,  1.53s/it]
 30%|███       | 5756/19090 [2:48:39<5:23:51,  1.46s/it]
 30%|███       | 5757/19090 [2:48:41<6:10:40,  1.67s/it]
 30%|███       | 5758/19090 [2:48:43<5:52:42,  1.59s/it]
 30%|███       | 5759/19090 [2:48:44<5:33:03,  1.50s/it]
 30%|███       | 5760/19090 [2:48:46<6:11:38,  1.67s/it]
                                                        
{'loss': 0.0027, 'grad_norm': 1.0269595384597778, 'learning_rate': 4.405675539940297e-05, 'recon_loss': 0.0027013828803319484, 'recon_static': 0.0017494281171821057, 'recon_motion': 0.0009519547878880985, 'recon_norm': 1.8059966295957566, 'recon_norm_std': 0.06241558101028204, 'target_norm': 1.6202091515064239, 'recon_cos': 0.49882944747805597, 'gradnorm/compressor.query': 0.011712019572096445, 'gradnorm/compressor.layer0': 0.21387052315675592, 'gradnorm/compressor.layer1': 0.7792715580013968, 'gradnorm/compressor.layer2': 0.3871099246855528, 'gradnorm/compressor.layer3': 0.3023652071571457, 'gradnorm/compressor.layer4': 0.15746486564284617, 'gradnorm/compressor.layer5': 0.13431362702263847, 'gradnorm/compressor.layer6': 0.10578906021980208, 'gradnorm/compressor.layer7': 0.1217896171570408, 'gradnorm/decoder.layer0': 0.18592172203118584, 'gradnorm/decoder.layer1': 0.17188290014317184, 'gradnorm/decoder.layer2': 0.16509139342993634, 'gradnorm/max_param': 0.5184545516967773, 'epoch': 3.02}

 30%|███       | 5760/19090 [2:48:46<6:11:38,  1.67s/it]
 30%|███       | 5761/19090 [2:48:47<5:45:44,  1.56s/it]
 30%|███       | 5762/19090 [2:48:49<5:27:34,  1.47s/it]
 30%|███       | 5763/19090 [2:48:51<6:16:35,  1.70s/it]
 30%|███       | 5764/19090 [2:48:53<6:28:25,  1.75s/it]
 30%|███       | 5765/19090 [2:48:54<5:57:18,  1.61s/it]
 30%|███       | 5766/19090 [2:48:56<6:31:53,  1.76s/it]
 30%|███       | 5767/19090 [2:48:57<5:59:39,  1.62s/it]
 30%|███       | 5768/19090 [2:48:59<5:37:00,  1.52s/it]
 30%|███       | 5769/19090 [2:49:01<6:10:25,  1.67s/it]
 30%|███       | 5770/19090 [2:49:02<5:45:44,  1.56s/it]
                                                        
{'loss': 0.0029, 'grad_norm': 1.0555850267410278, 'learning_rate': 4.402713529317453e-05, 'recon_loss': 0.0028657075599767268, 'recon_static': 0.001907769509125501, 'recon_motion': 0.0009579380086506717, 'recon_norm': 1.855296215415001, 'recon_norm_std': 0.06073578642681241, 'target_norm': 1.6207985997200012, 'recon_cos': 0.48549428433179853, 'gradnorm/compressor.query': 0.010553362915490931, 'gradnorm/compressor.layer0': 0.217844998358429, 'gradnorm/compressor.layer1': 0.8076705453721557, 'gradnorm/compressor.layer2': 0.3957203235259045, 'gradnorm/compressor.layer3': 0.30204511132084777, 'gradnorm/compressor.layer4': 0.15653910712454577, 'gradnorm/compressor.layer5': 0.13347515815485064, 'gradnorm/compressor.layer6': 0.1078914608878116, 'gradnorm/compressor.layer7': 0.12511528927175625, 'gradnorm/decoder.layer0': 0.19378006571097317, 'gradnorm/decoder.layer1': 0.17696441252288575, 'gradnorm/decoder.layer2': 0.16748916558910049, 'gradnorm/max_param': 0.5402776598930359, 'epoch': 3.02}

 30%|███       | 5770/19090 [2:49:02<5:45:44,  1.56s/it]
 30%|███       | 5771/19090 [2:49:03<5:28:02,  1.48s/it]
 30%|███       | 5772/19090 [2:49:06<6:34:43,  1.78s/it]
 30%|███       | 5773/19090 [2:49:07<6:11:29,  1.67s/it]
 30%|███       | 5774/19090 [2:49:08<5:45:42,  1.56s/it]
 30%|███       | 5775/19090 [2:49:11<6:48:23,  1.84s/it]
 30%|███       | 5776/19090 [2:49:12<6:11:01,  1.67s/it]
 30%|███       | 5777/19090 [2:49:14<5:45:14,  1.56s/it]
 30%|███       | 5778/19090 [2:49:16<6:54:46,  1.87s/it]
 30%|███       | 5779/19090 [2:49:17<6:15:50,  1.69s/it]
 30%|███       | 5780/19090 [2:49:19<5:48:41,  1.57s/it]
                                                        
{'loss': 0.0028, 'grad_norm': 1.110464334487915, 'learning_rate': 4.399745156951105e-05, 'recon_loss': 0.0027669193979818372, 'recon_static': 0.0018175913486629725, 'recon_motion': 0.0009493280027527362, 'recon_norm': 1.8297984808683396, 'recon_norm_std': 0.062616918515414, 'target_norm': 1.6209366172552109, 'recon_cos': 0.4958308398723602, 'gradnorm/compressor.query': 0.011203321007120565, 'gradnorm/compressor.layer0': 0.21611496501944125, 'gradnorm/compressor.layer1': 0.869170524854378, 'gradnorm/compressor.layer2': 0.4317212957196004, 'gradnorm/compressor.layer3': 0.30799444165958073, 'gradnorm/compressor.layer4': 0.15720126143938554, 'gradnorm/compressor.layer5': 0.12924904471640192, 'gradnorm/compressor.layer6': 0.09798601739072985, 'gradnorm/compressor.layer7': 0.11893496194287359, 'gradnorm/decoder.layer0': 0.18038200836947446, 'gradnorm/decoder.layer1': 0.16813429171294944, 'gradnorm/decoder.layer2': 0.15371268173255734, 'gradnorm/max_param': 0.5872563719749451, 'epoch': 3.03}

 30%|███       | 5780/19090 [2:49:19<5:48:41,  1.57s/it]
 30%|███       | 5781/19090 [2:49:21<6:34:28,  1.78s/it]
 30%|███       | 5782/19090 [2:49:22<6:01:19,  1.63s/it]
 30%|███       | 5783/19090 [2:49:24<5:38:27,  1.53s/it]
 30%|███       | 5784/19090 [2:49:26<6:16:30,  1.70s/it]
 30%|███       | 5785/19090 [2:49:27<6:04:16,  1.64s/it]
 30%|███       | 5786/19090 [2:49:28<5:40:08,  1.53s/it]
 30%|███       | 5787/19090 [2:49:31<6:39:18,  1.80s/it]
 30%|███       | 5788/19090 [2:49:32<6:17:54,  1.70s/it]
 30%|███       | 5789/19090 [2:49:34<5:51:23,  1.59s/it]
 30%|███       | 5790/19090 [2:49:36<6:49:48,  1.85s/it]
                                                        
{'loss': 0.0024, 'grad_norm': 1.0833309888839722, 'learning_rate': 4.396770432766039e-05, 'recon_loss': 0.0024261929385829715, 'recon_static': 0.0014661815686849878, 'recon_motion': 0.0009600113291526214, 'recon_norm': 1.7103818625211715, 'recon_norm_std': 0.06553501598536968, 'target_norm': 1.6228107541799546, 'recon_cos': 0.5244201712310315, 'gradnorm/compressor.query': 0.010106922882542876, 'gradnorm/compressor.layer0': 0.20692530206303605, 'gradnorm/compressor.layer1': 0.8271701273951122, 'gradnorm/compressor.layer2': 0.4259433612714074, 'gradnorm/compressor.layer3': 0.29663796749938953, 'gradnorm/compressor.layer4': 0.16423973092836483, 'gradnorm/compressor.layer5': 0.1383852548697601, 'gradnorm/compressor.layer6': 0.10727737417978701, 'gradnorm/compressor.layer7': 0.13076791129003495, 'gradnorm/decoder.layer0': 0.19831121586511, 'gradnorm/decoder.layer1': 0.18513943962048007, 'gradnorm/decoder.layer2': 0.16945685960726622, 'gradnorm/max_param': 0.5556060075759888, 'epoch': 3.03}

 30%|███       | 5790/19090 [2:49:36<6:49:48,  1.85s/it]
 30%|███       | 5791/19090 [2:49:37<6:13:34,  1.69s/it]
 30%|███       | 5792/19090 [2:49:39<5:47:49,  1.57s/it]
 30%|███       | 5793/19090 [2:49:41<6:49:26,  1.85s/it]
 30%|███       | 5794/19090 [2:49:42<6:12:04,  1.68s/it]
 30%|███       | 5795/19090 [2:49:44<5:47:04,  1.57s/it]
 30%|███       | 5796/19090 [2:49:46<6:39:50,  1.80s/it]
 30%|███       | 5797/19090 [2:49:47<6:05:17,  1.65s/it]
 30%|███       | 5798/19090 [2:49:49<5:41:39,  1.54s/it]
 30%|███       | 5799/19090 [2:49:51<7:00:22,  1.90s/it]
 30%|███       | 5800/19090 [2:49:53<6:19:58,  1.72s/it]
                                                        
{'loss': 0.0026, 'grad_norm': 1.099693775177002, 'learning_rate': 4.393789366708279e-05, 'recon_loss': 0.00260485180770047, 'recon_static': 0.0016517692944034933, 'recon_motion': 0.0009530824521789327, 'recon_norm': 1.7730442404747009, 'recon_norm_std': 0.06262310007587076, 'target_norm': 1.6204594552516938, 'recon_cos': 0.5072408944368363, 'gradnorm/compressor.query': 0.01167374651974331, 'gradnorm/compressor.layer0': 0.22151120543068534, 'gradnorm/compressor.layer1': 0.8560075848323584, 'gradnorm/compressor.layer2': 0.4106915205516475, 'gradnorm/compressor.layer3': 0.297900574200606, 'gradnorm/compressor.layer4': 0.16564024953994824, 'gradnorm/compressor.layer5': 0.13660000685388818, 'gradnorm/compressor.layer6': 0.10794827491371106, 'gradnorm/compressor.layer7': 0.12808998635498664, 'gradnorm/decoder.layer0': 0.1822738377617111, 'gradnorm/decoder.layer1': 0.18258489231402808, 'gradnorm/decoder.layer2': 0.17101077610947124, 'gradnorm/max_param': 0.5753939747810364, 'epoch': 3.04}

 30%|███       | 5800/19090 [2:49:53<6:19:58,  1.72s/it]
 30%|███       | 5801/19090 [2:49:54<5:51:53,  1.59s/it]
 30%|███       | 5802/19090 [2:49:56<6:39:37,  1.80s/it]
 30%|███       | 5803/19090 [2:49:58<6:05:03,  1.65s/it]
 30%|███       | 5804/19090 [2:49:59<5:41:13,  1.54s/it]
 30%|███       | 5805/19090 [2:50:01<5:59:16,  1.62s/it]
 30%|███       | 5806/19090 [2:50:02<5:47:16,  1.57s/it]
 30%|███       | 5807/19090 [2:50:03<5:28:35,  1.48s/it]
 30%|███       | 5808/19090 [2:50:06<6:17:54,  1.71s/it]
 30%|███       | 5809/19090 [2:50:07<6:00:38,  1.63s/it]
 30%|███       | 5810/19090 [2:50:08<5:38:02,  1.53s/it]
                                                        
{'loss': 0.0024, 'grad_norm': 1.0984357595443726, 'learning_rate': 4.3908019687450496e-05, 'recon_loss': 0.002441105229081586, 'recon_static': 0.0014850121515337378, 'recon_motion': 0.000956093096465338, 'recon_norm': 1.7181951224803924, 'recon_norm_std': 0.06392530873417854, 'target_norm': 1.619944441318512, 'recon_cos': 0.5247419320046902, 'gradnorm/compressor.query': 0.008493635696236923, 'gradnorm/compressor.layer0': 0.17105820408508182, 'gradnorm/compressor.layer1': 0.8935624252567979, 'gradnorm/compressor.layer2': 0.4270936356054002, 'gradnorm/compressor.layer3': 0.29046454512254594, 'gradnorm/compressor.layer4': 0.1366661788138932, 'gradnorm/compressor.layer5': 0.10471308167887895, 'gradnorm/compressor.layer6': 0.07731946566359753, 'gradnorm/compressor.layer7': 0.10358266638817033, 'gradnorm/decoder.layer0': 0.16080290805128808, 'gradnorm/decoder.layer1': 0.1581731520622602, 'gradnorm/decoder.layer2': 0.12152907355692502, 'gradnorm/max_param': 0.6148701906204224, 'epoch': 3.04}

 30%|███       | 5810/19090 [2:50:08<5:38:02,  1.53s/it]
 30%|███       | 5811/19090 [2:50:10<6:14:03,  1.69s/it]
 30%|███       | 5812/19090 [2:50:12<6:20:49,  1.72s/it]
 30%|███       | 5813/19090 [2:50:14<5:52:17,  1.59s/it]
 30%|███       | 5814/19090 [2:50:16<6:21:22,  1.72s/it]
 30%|███       | 5815/19090 [2:50:18<6:54:21,  1.87s/it]
 30%|███       | 5816/19090 [2:50:19<6:15:48,  1.70s/it]
 30%|███       | 5817/19090 [2:50:20<5:48:39,  1.58s/it]
 30%|███       | 5818/19090 [2:50:23<6:54:30,  1.87s/it]
 30%|███       | 5819/19090 [2:50:24<6:15:15,  1.70s/it]
 30%|███       | 5820/19090 [2:50:26<5:48:28,  1.58s/it]
                                                        
{'loss': 0.0024, 'grad_norm': 1.1924031972885132, 'learning_rate': 4.387808248864751e-05, 'recon_loss': 0.002437314245617017, 'recon_static': 0.0014839114301139489, 'recon_motion': 0.0009534027369227261, 'recon_norm': 1.7153572410345077, 'recon_norm_std': 0.06432431945577263, 'target_norm': 1.620260152220726, 'recon_cos': 0.5234477639198303, 'gradnorm/compressor.query': 0.011588252920980757, 'gradnorm/compressor.layer0': 0.23216701607176957, 'gradnorm/compressor.layer1': 0.9194822131412892, 'gradnorm/compressor.layer2': 0.47895224756381927, 'gradnorm/compressor.layer3': 0.33683321438882746, 'gradnorm/compressor.layer4': 0.18157752797214333, 'gradnorm/compressor.layer5': 0.1391841131636749, 'gradnorm/compressor.layer6': 0.10859697220502038, 'gradnorm/compressor.layer7': 0.12880443395646474, 'gradnorm/decoder.layer0': 0.1864351293219379, 'gradnorm/decoder.layer1': 0.18984966465091802, 'gradnorm/decoder.layer2': 0.1672202220874051, 'gradnorm/max_param': 0.6197419166564941, 'epoch': 3.05}

 30%|███       | 5820/19090 [2:50:26<5:48:28,  1.58s/it]
 30%|███       | 5821/19090 [2:50:28<6:55:22,  1.88s/it]
 30%|███       | 5822/19090 [2:50:29<6:16:31,  1.70s/it]
 31%|███       | 5823/19090 [2:50:31<5:49:37,  1.58s/it]
 31%|███       | 5824/19090 [2:50:33<6:16:26,  1.70s/it]
 31%|███       | 5825/19090 [2:50:34<5:49:25,  1.58s/it]
 31%|███       | 5826/19090 [2:50:35<5:30:23,  1.49s/it]
 31%|███       | 5827/19090 [2:50:38<6:56:43,  1.89s/it]
 31%|███       | 5828/19090 [2:50:39<6:17:32,  1.71s/it]
 31%|███       | 5829/19090 [2:50:41<5:49:27,  1.58s/it]
 31%|███       | 5830/19090 [2:50:43<6:35:19,  1.79s/it]
                                                        
{'loss': 0.0024, 'grad_norm': 1.031747817993164, 'learning_rate': 4.384808217076917e-05, 'recon_loss': 0.0024422709830105305, 'recon_static': 0.0014857638365356252, 'recon_motion': 0.0009565071057295426, 'recon_norm': 1.7177977442741394, 'recon_norm_std': 0.06340912096202374, 'target_norm': 1.6206650972366332, 'recon_cos': 0.5226032361388206, 'gradnorm/compressor.query': 0.009478695639561021, 'gradnorm/compressor.layer0': 0.18730125225206573, 'gradnorm/compressor.layer1': 0.7955413204397419, 'gradnorm/compressor.layer2': 0.42471956299033237, 'gradnorm/compressor.layer3': 0.2899503218917257, 'gradnorm/compressor.layer4': 0.1484947596084528, 'gradnorm/compressor.layer5': 0.12438071596007885, 'gradnorm/compressor.layer6': 0.09008232864536572, 'gradnorm/compressor.layer7': 0.11302751043111964, 'gradnorm/decoder.layer0': 0.16607001479542513, 'gradnorm/decoder.layer1': 0.1586748613389243, 'gradnorm/decoder.layer2': 0.14429147265141518, 'gradnorm/max_param': 0.5391502380371094, 'epoch': 3.05}

 31%|███       | 5830/19090 [2:50:43<6:35:19,  1.79s/it]
 31%|███       | 5831/19090 [2:50:44<6:01:48,  1.64s/it]
 31%|███       | 5832/19090 [2:50:46<5:38:29,  1.53s/it]
 31%|███       | 5833/19090 [2:50:48<6:32:27,  1.78s/it]
 31%|███       | 5834/19090 [2:50:49<6:00:39,  1.63s/it]
 31%|███       | 5835/19090 [2:50:50<5:40:14,  1.54s/it]
 31%|███       | 5836/19090 [2:50:53<6:49:20,  1.85s/it]
 31%|███       | 5837/19090 [2:50:54<6:12:25,  1.69s/it]
 31%|███       | 5838/19090 [2:50:56<5:47:06,  1.57s/it]
 31%|███       | 5839/19090 [2:50:58<6:59:07,  1.90s/it]
 31%|███       | 5840/19090 [2:51:00<6:20:09,  1.72s/it]
                                                        
{'loss': 0.0024, 'grad_norm': 1.1789172887802124, 'learning_rate': 4.381801883412187e-05, 'recon_loss': 0.002348939870717004, 'recon_static': 0.001404540127259679, 'recon_motion': 0.0009443998336791992, 'recon_norm': 1.6885802030563355, 'recon_norm_std': 0.06603503618389368, 'target_norm': 1.618884229660034, 'recon_cos': 0.5332528851926327, 'gradnorm/compressor.query': 0.010120679777199913, 'gradnorm/compressor.layer0': 0.19153146151610184, 'gradnorm/compressor.layer1': 0.9544040352752031, 'gradnorm/compressor.layer2': 0.4562736002161464, 'gradnorm/compressor.layer3': 0.3090644578322064, 'gradnorm/compressor.layer4': 0.15110224640753953, 'gradnorm/compressor.layer5': 0.12284079202765177, 'gradnorm/compressor.layer6': 0.08830892574244353, 'gradnorm/compressor.layer7': 0.11126735768832614, 'gradnorm/decoder.layer0': 0.1817730964057048, 'gradnorm/decoder.layer1': 0.168872624675287, 'gradnorm/decoder.layer2': 0.13713323570277006, 'gradnorm/max_param': 0.6558526754379272, 'epoch': 3.06}

 31%|███       | 5840/19090 [2:51:00<6:20:09,  1.72s/it]
 31%|███       | 5841/19090 [2:51:01<5:52:22,  1.60s/it]
 31%|███       | 5842/19090 [2:51:03<6:43:43,  1.83s/it]
 31%|███       | 5843/19090 [2:51:05<6:10:12,  1.68s/it]
 31%|███       | 5844/19090 [2:51:06<5:48:37,  1.58s/it]
 31%|███       | 5845/19090 [2:51:09<6:59:55,  1.90s/it]
 31%|███       | 5846/19090 [2:51:10<6:21:15,  1.73s/it]
 31%|███       | 5847/19090 [2:51:11<5:52:56,  1.60s/it]
 31%|███       | 5848/19090 [2:51:14<6:59:52,  1.90s/it]
 31%|███       | 5849/19090 [2:51:15<6:19:40,  1.72s/it]
 31%|███       | 5850/19090 [2:51:16<5:51:50,  1.59s/it]
                                                        
{'loss': 0.0021, 'grad_norm': 1.1554944515228271, 'learning_rate': 4.378789257922271e-05, 'recon_loss': 0.0021166971913771705, 'recon_static': 0.0011595759075134992, 'recon_motion': 0.0009571212634909898, 'recon_norm': 1.601555860042572, 'recon_norm_std': 0.06606896249577403, 'target_norm': 1.6203271627426148, 'recon_cos': 0.5574383333325386, 'gradnorm/compressor.query': 0.010589662821028635, 'gradnorm/compressor.layer0': 0.19844161652628034, 'gradnorm/compressor.layer1': 0.9229477928085218, 'gradnorm/compressor.layer2': 0.4387922616042417, 'gradnorm/compressor.layer3': 0.32100051166591687, 'gradnorm/compressor.layer4': 0.15436756349714054, 'gradnorm/compressor.layer5': 0.12462819952765153, 'gradnorm/compressor.layer6': 0.09557300109842227, 'gradnorm/compressor.layer7': 0.11524857013362573, 'gradnorm/decoder.layer0': 0.18232644748398233, 'gradnorm/decoder.layer1': 0.17674476617539672, 'gradnorm/decoder.layer2': 0.14833785778371214, 'gradnorm/max_param': 0.629843533039093, 'epoch': 3.06}

 31%|███       | 5850/19090 [2:51:16<5:51:50,  1.59s/it]
 31%|███       | 5851/19090 [2:51:19<6:57:17,  1.89s/it]
 31%|███       | 5852/19090 [2:51:20<6:18:42,  1.72s/it]
 31%|███       | 5853/19090 [2:51:22<5:51:14,  1.59s/it]
 31%|███       | 5854/19090 [2:51:24<6:39:05,  1.81s/it]
 31%|███       | 5855/19090 [2:51:25<6:05:07,  1.66s/it]
 31%|███       | 5856/19090 [2:51:27<5:41:51,  1.55s/it]
 31%|███       | 5857/19090 [2:51:30<7:32:58,  2.05s/it]
 31%|███       | 5858/19090 [2:51:31<6:43:05,  1.83s/it]
 31%|███       | 5859/19090 [2:51:32<6:08:18,  1.67s/it]
 31%|███       | 5860/19090 [2:51:35<6:54:41,  1.88s/it]
                                                        
{'loss': 0.0021, 'grad_norm': 1.2403267621994019, 'learning_rate': 4.375770350679914e-05, 'recon_loss': 0.0020952688850229607, 'recon_static': 0.001147753909754101, 'recon_motion': 0.0009475149840000085, 'recon_norm': 1.6029131591320038, 'recon_norm_std': 0.06650286233052612, 'target_norm': 1.6206929236650467, 'recon_cos': 0.562709367275238, 'gradnorm/compressor.query': 0.011079042021791899, 'gradnorm/compressor.layer0': 0.22323905228409882, 'gradnorm/compressor.layer1': 1.0001870814802098, 'gradnorm/compressor.layer2': 0.45696546619098616, 'gradnorm/compressor.layer3': 0.3398237441740543, 'gradnorm/compressor.layer4': 0.16527027659953764, 'gradnorm/compressor.layer5': 0.13775690202850763, 'gradnorm/compressor.layer6': 0.10028261913174467, 'gradnorm/compressor.layer7': 0.11874924537142853, 'gradnorm/decoder.layer0': 0.18805620031686102, 'gradnorm/decoder.layer1': 0.18632080895477993, 'gradnorm/decoder.layer2': 0.15248049809682923, 'gradnorm/max_param': 0.6817191243171692, 'epoch': 3.07}

 31%|███       | 5860/19090 [2:51:35<6:54:41,  1.88s/it]
 31%|███       | 5861/19090 [2:51:36<6:16:15,  1.71s/it]
 31%|███       | 5862/19090 [2:51:37<5:49:53,  1.59s/it]
 31%|███       | 5863/19090 [2:51:40<6:53:39,  1.88s/it]
 31%|███       | 5864/19090 [2:51:41<6:15:32,  1.70s/it]
 31%|███       | 5865/19090 [2:51:43<5:48:46,  1.58s/it]
 31%|███       | 5866/19090 [2:51:45<7:19:13,  1.99s/it]
 31%|███       | 5867/19090 [2:51:47<6:34:28,  1.79s/it]
 31%|███       | 5868/19090 [2:51:48<6:01:22,  1.64s/it]
 31%|███       | 5869/19090 [2:51:50<6:25:07,  1.75s/it]
 31%|███       | 5870/19090 [2:51:51<5:54:52,  1.61s/it]
                                                        
{'loss': 0.002, 'grad_norm': 1.1866991519927979, 'learning_rate': 4.372745171778867e-05, 'recon_loss': 0.0020106908050365746, 'recon_static': 0.0010617356514558197, 'recon_motion': 0.0009489551681326702, 'recon_norm': 1.5704708814620971, 'recon_norm_std': 0.06816288717091083, 'target_norm': 1.6204338371753693, 'recon_cos': 0.5729558169841766, 'gradnorm/compressor.query': 0.010136824264601966, 'gradnorm/compressor.layer0': 0.18109545367067356, 'gradnorm/compressor.layer1': 0.9669018312761776, 'gradnorm/compressor.layer2': 0.44422250494508797, 'gradnorm/compressor.layer3': 0.3283524065823495, 'gradnorm/compressor.layer4': 0.15284381004500786, 'gradnorm/compressor.layer5': 0.11644215382128027, 'gradnorm/compressor.layer6': 0.08090325652928151, 'gradnorm/compressor.layer7': 0.10809967177095965, 'gradnorm/decoder.layer0': 0.18879946783599977, 'gradnorm/decoder.layer1': 0.17109409031768816, 'gradnorm/decoder.layer2': 0.12344736440090455, 'gradnorm/max_param': 0.6670695543289185, 'epoch': 3.07}

 31%|███       | 5870/19090 [2:51:51<5:54:52,  1.61s/it]
 31%|███       | 5871/19090 [2:51:53<5:34:19,  1.52s/it]
 31%|███       | 5872/19090 [2:51:55<6:31:51,  1.78s/it]
 31%|███       | 5873/19090 [2:51:56<5:59:33,  1.63s/it]
 31%|███       | 5874/19090 [2:51:58<5:36:43,  1.53s/it]
 31%|███       | 5875/19090 [2:52:00<6:43:04,  1.83s/it]
 31%|███       | 5876/19090 [2:52:01<6:07:12,  1.67s/it]
 31%|███       | 5877/19090 [2:52:03<5:42:04,  1.55s/it]
 31%|███       | 5878/19090 [2:52:05<6:55:00,  1.88s/it]
 31%|███       | 5879/19090 [2:52:07<6:15:34,  1.71s/it]
 31%|███       | 5880/19090 [2:52:08<5:48:21,  1.58s/it]
                                                        
{'loss': 0.002, 'grad_norm': 1.2315304279327393, 'learning_rate': 4.369713731333847e-05, 'recon_loss': 0.002003709282143973, 'recon_static': 0.0010447306733112783, 'recon_motion': 0.0009589785360731185, 'recon_norm': 1.5591514885425568, 'recon_norm_std': 0.06652538273483515, 'target_norm': 1.6216659098863602, 'recon_cos': 0.5720352128148078, 'gradnorm/compressor.query': 0.009855962315199605, 'gradnorm/compressor.layer0': 0.19629301000772562, 'gradnorm/compressor.layer1': 1.0021849841760424, 'gradnorm/compressor.layer2': 0.47019478348102234, 'gradnorm/compressor.layer3': 0.31287768905130275, 'gradnorm/compressor.layer4': 0.16515263481297648, 'gradnorm/compressor.layer5': 0.1283523261258657, 'gradnorm/compressor.layer6': 0.09210694984466024, 'gradnorm/compressor.layer7': 0.11875492339845024, 'gradnorm/decoder.layer0': 0.19083524702152355, 'gradnorm/decoder.layer1': 0.17737697466543487, 'gradnorm/decoder.layer2': 0.14308048608649548, 'gradnorm/max_param': 0.6885062456130981, 'epoch': 3.08}

 31%|███       | 5880/19090 [2:52:08<5:48:21,  1.58s/it]
 31%|███       | 5881/19090 [2:52:10<6:31:13,  1.78s/it]
 31%|███       | 5882/19090 [2:52:12<5:58:52,  1.63s/it]
 31%|███       | 5883/19090 [2:52:13<5:35:53,  1.53s/it]
 31%|███       | 5884/19090 [2:52:15<6:32:04,  1.78s/it]
 31%|███       | 5885/19090 [2:52:16<5:59:27,  1.63s/it]
 31%|███       | 5886/19090 [2:52:18<5:36:36,  1.53s/it]
 31%|███       | 5887/19090 [2:52:21<7:04:29,  1.93s/it]
 31%|███       | 5888/19090 [2:52:22<6:22:24,  1.74s/it]
 31%|███       | 5889/19090 [2:52:23<5:52:54,  1.60s/it]
 31%|███       | 5890/19090 [2:52:26<6:43:26,  1.83s/it]
                                                        
{'loss': 0.0019, 'grad_norm': 1.2876625061035156, 'learning_rate': 4.36667603948051e-05, 'recon_loss': 0.0018557251954916866, 'recon_static': 0.0009092122258152813, 'recon_motion': 0.0009465129624004476, 'recon_norm': 1.5092299699783325, 'recon_norm_std': 0.0698631071485579, 'target_norm': 1.6180013686418533, 'recon_cos': 0.593437311053276, 'gradnorm/compressor.query': 0.008828788210608077, 'gradnorm/compressor.layer0': 0.16646742770731654, 'gradnorm/compressor.layer1': 1.0734045098578364, 'gradnorm/compressor.layer2': 0.4928887382419213, 'gradnorm/compressor.layer3': 0.323729110876029, 'gradnorm/compressor.layer4': 0.1405749563076495, 'gradnorm/compressor.layer5': 0.11845092002910691, 'gradnorm/compressor.layer6': 0.07559056871602832, 'gradnorm/compressor.layer7': 0.09804795386188166, 'gradnorm/decoder.layer0': 0.19384195814237737, 'gradnorm/decoder.layer1': 0.17206096883598976, 'gradnorm/decoder.layer2': 0.11853094523089225, 'gradnorm/max_param': 0.7454608082771301, 'epoch': 3.09}

 31%|███       | 5890/19090 [2:52:26<6:43:26,  1.83s/it]
 31%|███       | 5891/19090 [2:52:27<6:07:35,  1.67s/it]
 31%|███       | 5892/19090 [2:52:28<5:42:29,  1.56s/it]
 31%|███       | 5893/19090 [2:52:31<7:29:13,  2.04s/it]
 31%|███       | 5894/19090 [2:52:33<6:39:37,  1.82s/it]
 31%|███       | 5895/19090 [2:52:34<6:04:22,  1.66s/it]
 31%|███       | 5896/19090 [2:52:36<6:57:31,  1.90s/it]
 31%|███       | 5897/19090 [2:52:38<6:17:23,  1.72s/it]
 31%|███       | 5898/19090 [2:52:39<5:48:56,  1.59s/it]
 31%|███       | 5899/19090 [2:52:42<7:09:16,  1.95s/it]
 31%|███       | 5900/19090 [2:52:43<6:25:28,  1.75s/it]
                                                        
{'loss': 0.0019, 'grad_norm': 1.3168538808822632, 'learning_rate': 4.363632106375412e-05, 'recon_loss': 0.0018652087368536741, 'recon_static': 0.0009116196044487879, 'recon_motion': 0.0009535891309496947, 'recon_norm': 1.5130655258893966, 'recon_norm_std': 0.0698783453553915, 'target_norm': 1.6197980880737304, 'recon_cos': 0.59302918612957, 'gradnorm/compressor.query': 0.009572671732054871, 'gradnorm/compressor.layer0': 0.18444970645992345, 'gradnorm/compressor.layer1': 1.094224772269914, 'gradnorm/compressor.layer2': 0.5058300053970406, 'gradnorm/compressor.layer3': 0.34138299133895317, 'gradnorm/compressor.layer4': 0.15247817100180366, 'gradnorm/compressor.layer5': 0.11753481418922816, 'gradnorm/compressor.layer6': 0.08205212032403161, 'gradnorm/compressor.layer7': 0.10950656624298694, 'gradnorm/decoder.layer0': 0.17759584960219946, 'gradnorm/decoder.layer1': 0.1616690642283142, 'gradnorm/decoder.layer2': 0.12958042538876868, 'gradnorm/max_param': 0.7577736973762512, 'epoch': 3.09}

 31%|███       | 5900/19090 [2:52:43<6:25:28,  1.75s/it]
 31%|███       | 5901/19090 [2:52:44<5:54:50,  1.61s/it]
 31%|███       | 5902/19090 [2:52:47<6:56:55,  1.90s/it]
 31%|███       | 5903/19090 [2:52:48<6:17:50,  1.72s/it]
 31%|███       | 5904/19090 [2:52:49<5:49:26,  1.59s/it]
 31%|███       | 5905/19090 [2:52:52<6:55:51,  1.89s/it]
 31%|███       | 5906/19090 [2:52:53<6:16:10,  1.71s/it]
 31%|███       | 5907/19090 [2:52:55<5:48:13,  1.58s/it]
 31%|███       | 5908/19090 [2:52:58<7:27:18,  2.04s/it]
 31%|███       | 5909/19090 [2:52:59<6:37:50,  1.81s/it]
 31%|███       | 5910/19090 [2:53:00<6:03:36,  1.66s/it]
                                                        
{'loss': 0.0017, 'grad_norm': 1.240746021270752, 'learning_rate': 4.360581942195975e-05, 'recon_loss': 0.0017451108025852591, 'recon_static': 0.0007917766852187924, 'recon_motion': 0.0009533341930364258, 'recon_norm': 1.4642984211444854, 'recon_norm_std': 0.07139294724911452, 'target_norm': 1.6203853100538255, 'recon_cos': 0.6111443147063256, 'gradnorm/compressor.query': 0.007899979237618959, 'gradnorm/compressor.layer0': 0.154758848247021, 'gradnorm/compressor.layer1': 1.033505314440967, 'gradnorm/compressor.layer2': 0.4825583971340986, 'gradnorm/compressor.layer3': 0.3095534647981055, 'gradnorm/compressor.layer4': 0.13898540701223, 'gradnorm/compressor.layer5': 0.10919767293038908, 'gradnorm/compressor.layer6': 0.07114260469958478, 'gradnorm/compressor.layer7': 0.09761353512236018, 'gradnorm/decoder.layer0': 0.17972801622459056, 'gradnorm/decoder.layer1': 0.1679759633932833, 'gradnorm/decoder.layer2': 0.11077024988998076, 'gradnorm/max_param': 0.7187587022781372, 'epoch': 3.1}

 31%|███       | 5910/19090 [2:53:00<6:03:36,  1.66s/it]
 31%|███       | 5911/19090 [2:53:03<7:16:03,  1.99s/it]
 31%|███       | 5912/19090 [2:53:04<6:30:26,  1.78s/it]
 31%|███       | 5913/19090 [2:53:06<5:58:05,  1.63s/it]
 31%|███       | 5914/19090 [2:53:08<6:56:07,  1.89s/it]
 31%|███       | 5915/19090 [2:53:09<6:16:19,  1.71s/it]
 31%|███       | 5916/19090 [2:53:11<5:48:24,  1.59s/it]
 31%|███       | 5917/19090 [2:53:13<7:00:43,  1.92s/it]
 31%|███       | 5918/19090 [2:53:15<6:19:25,  1.73s/it]
 31%|███       | 5919/19090 [2:53:16<5:50:07,  1.59s/it]
 31%|███       | 5920/19090 [2:53:19<7:28:24,  2.04s/it]
                                                        
{'loss': 0.0017, 'grad_norm': 1.30172860622406, 'learning_rate': 4.357525557140457e-05, 'recon_loss': 0.0016613539890386164, 'recon_static': 0.0007132759536034427, 'recon_motion': 0.000948078055807855, 'recon_norm': 1.4339959442615509, 'recon_norm_std': 0.07199267856776714, 'target_norm': 1.6180730998516082, 'recon_cos': 0.6252312272787094, 'gradnorm/compressor.query': 0.007282035518530934, 'gradnorm/compressor.layer0': 0.14162657286661967, 'gradnorm/compressor.layer1': 1.0970211703292854, 'gradnorm/compressor.layer2': 0.5051776607942303, 'gradnorm/compressor.layer3': 0.328949618708292, 'gradnorm/compressor.layer4': 0.13490307060058038, 'gradnorm/compressor.layer5': 0.10473735070785316, 'gradnorm/compressor.layer6': 0.06180950483332809, 'gradnorm/compressor.layer7': 0.0919970905916803, 'gradnorm/decoder.layer0': 0.16439994101536348, 'gradnorm/decoder.layer1': 0.16966564217710461, 'gradnorm/decoder.layer2': 0.10125240339264158, 'gradnorm/max_param': 0.7657578587532043, 'epoch': 3.1}

 31%|███       | 5920/19090 [2:53:19<7:28:24,  2.04s/it]
 31%|███       | 5921/19090 [2:53:20<6:38:49,  1.82s/it]
 31%|███       | 5922/19090 [2:53:22<6:04:22,  1.66s/it]
 31%|███       | 5923/19090 [2:53:25<7:28:11,  2.04s/it]
 31%|███       | 5924/19090 [2:53:26<6:38:31,  1.82s/it]
 31%|███       | 5925/19090 [2:53:27<6:04:03,  1.66s/it]
 31%|███       | 5926/19090 [2:53:30<7:34:03,  2.07s/it]
 31%|███       | 5927/19090 [2:53:32<6:42:38,  1.84s/it]
 31%|███       | 5928/19090 [2:53:33<6:06:20,  1.67s/it]
 31%|███       | 5929/19090 [2:53:35<6:51:06,  1.87s/it]
 31%|███       | 5930/19090 [2:53:36<6:12:44,  1.70s/it]
                                                        
{'loss': 0.0016, 'grad_norm': 1.2816951274871826, 'learning_rate': 4.354462961427915e-05, 'recon_loss': 0.0016323956078849733, 'recon_static': 0.0006762784498278052, 'recon_motion': 0.0009561172715621069, 'recon_norm': 1.4150078237056731, 'recon_norm_std': 0.07242365600541234, 'target_norm': 1.62131667137146, 'recon_cos': 0.6301268711686134, 'gradnorm/compressor.query': 0.006295075218037048, 'gradnorm/compressor.layer0': 0.13172228824257162, 'gradnorm/compressor.layer1': 1.0868611421409073, 'gradnorm/compressor.layer2': 0.4748428147392439, 'gradnorm/compressor.layer3': 0.33131826929834457, 'gradnorm/compressor.layer4': 0.14583453038836963, 'gradnorm/compressor.layer5': 0.1050335239923541, 'gradnorm/compressor.layer6': 0.058388052929272405, 'gradnorm/compressor.layer7': 0.091054926110556, 'gradnorm/decoder.layer0': 0.1695306648297033, 'gradnorm/decoder.layer1': 0.163653892357703, 'gradnorm/decoder.layer2': 0.096510541703599, 'gradnorm/max_param': 0.7594544887542725, 'epoch': 3.11}

 31%|███       | 5930/19090 [2:53:36<6:12:44,  1.70s/it]
 31%|███       | 5931/19090 [2:53:38<5:45:25,  1.57s/it]
 31%|███       | 5932/19090 [2:53:40<6:53:23,  1.89s/it]
 31%|███       | 5933/19090 [2:53:42<6:14:15,  1.71s/it]
 31%|███       | 5934/19090 [2:53:43<5:46:31,  1.58s/it]
 31%|███       | 5935/19090 [2:53:46<7:10:16,  1.96s/it]
 31%|███       | 5936/19090 [2:53:47<6:26:21,  1.76s/it]
 31%|███       | 5937/19090 [2:53:48<5:55:44,  1.62s/it]
 31%|███       | 5938/19090 [2:53:51<7:30:18,  2.05s/it]
 31%|███       | 5939/19090 [2:53:53<6:41:38,  1.83s/it]
 31%|███       | 5940/19090 [2:53:54<6:08:05,  1.68s/it]
                                                        
{'loss': 0.0016, 'grad_norm': 1.4535185098648071, 'learning_rate': 4.351394165298172e-05, 'recon_loss': 0.0016420988831669092, 'recon_static': 0.0006993650444201194, 'recon_motion': 0.0009427338402019814, 'recon_norm': 1.4253805577754974, 'recon_norm_std': 0.07520841360092163, 'target_norm': 1.6205429911613465, 'recon_cos': 0.6291863918304443, 'gradnorm/compressor.query': 0.009909124979201982, 'gradnorm/compressor.layer0': 0.185232900696498, 'gradnorm/compressor.layer1': 1.2201210037789736, 'gradnorm/compressor.layer2': 0.5637899020129354, 'gradnorm/compressor.layer3': 0.3692468964530879, 'gradnorm/compressor.layer4': 0.16079116784212222, 'gradnorm/compressor.layer5': 0.12023614404974425, 'gradnorm/compressor.layer6': 0.0750078179834533, 'gradnorm/compressor.layer7': 0.10498286399770211, 'gradnorm/decoder.layer0': 0.18715974611541508, 'gradnorm/decoder.layer1': 0.1718359653622899, 'gradnorm/decoder.layer2': 0.11787415637167477, 'gradnorm/max_param': 0.8489505648612976, 'epoch': 3.11}

 31%|███       | 5940/19090 [2:53:54<6:08:05,  1.68s/it]
 31%|███       | 5941/19090 [2:53:57<7:17:53,  2.00s/it]
 31%|███       | 5942/19090 [2:53:58<6:31:07,  1.78s/it]
 31%|███       | 5943/19090 [2:53:59<5:58:38,  1.64s/it]
 31%|███       | 5944/19090 [2:54:02<7:09:23,  1.96s/it]
 31%|███       | 5945/19090 [2:54:03<6:25:45,  1.76s/it]
 31%|███       | 5946/19090 [2:54:05<5:55:16,  1.62s/it]
 31%|███       | 5947/19090 [2:54:07<7:02:03,  1.93s/it]
 31%|███       | 5948/19090 [2:54:09<6:20:06,  1.74s/it]
 31%|███       | 5949/19090 [2:54:10<5:50:59,  1.60s/it]
 31%|███       | 5950/19090 [2:54:13<7:00:59,  1.92s/it]
                                                        
{'loss': 0.0016, 'grad_norm': 1.4825425148010254, 'learning_rate': 4.348319179011781e-05, 'recon_loss': 0.0015960412536514923, 'recon_static': 0.0006432462498196401, 'recon_motion': 0.0009527949645416811, 'recon_norm': 1.4016435533761977, 'recon_norm_std': 0.07038974389433861, 'target_norm': 1.6192995488643647, 'recon_cos': 0.6358880996704102, 'gradnorm/compressor.query': 0.00855036374921608, 'gradnorm/compressor.layer0': 0.1613875196636242, 'gradnorm/compressor.layer1': 1.2647395863327686, 'gradnorm/compressor.layer2': 0.5597587388035004, 'gradnorm/compressor.layer3': 0.36690345897478877, 'gradnorm/compressor.layer4': 0.15022126483300124, 'gradnorm/compressor.layer5': 0.11678241413757084, 'gradnorm/compressor.layer6': 0.07038803149789698, 'gradnorm/compressor.layer7': 0.1047093950334602, 'gradnorm/decoder.layer0': 0.18161632006067976, 'gradnorm/decoder.layer1': 0.1635336469751243, 'gradnorm/decoder.layer2': 0.11157255777867504, 'gradnorm/max_param': 0.8830541968345642, 'epoch': 3.12}

 31%|███       | 5950/19090 [2:54:13<7:00:59,  1.92s/it]
 31%|███       | 5951/19090 [2:54:14<6:19:24,  1.73s/it]
 31%|███       | 5952/19090 [2:54:15<5:50:11,  1.60s/it]
 31%|███       | 5953/19090 [2:54:18<7:06:31,  1.95s/it]
 31%|███       | 5954/19090 [2:54:19<6:23:26,  1.75s/it]
 31%|███       | 5955/19090 [2:54:20<5:52:50,  1.61s/it]
 31%|███       | 5956/19090 [2:54:23<6:34:07,  1.80s/it]
 31%|███       | 5957/19090 [2:54:24<6:00:19,  1.65s/it]
 31%|███       | 5958/19090 [2:54:25<5:37:01,  1.54s/it]
 31%|███       | 5959/19090 [2:54:28<6:36:57,  1.81s/it]
 31%|███       | 5960/19090 [2:54:29<6:02:46,  1.66s/it]
                                                        
{'loss': 0.0016, 'grad_norm': 1.3354690074920654, 'learning_rate': 4.345238012849991e-05, 'recon_loss': 0.0015676952782087028, 'recon_static': 0.0006143828883068636, 'recon_motion': 0.0009533124262816273, 'recon_norm': 1.3923010379076004, 'recon_norm_std': 0.07222850210964679, 'target_norm': 1.6196686655282975, 'recon_cos': 0.6421912014484406, 'gradnorm/compressor.query': 0.007093196207563805, 'gradnorm/compressor.layer0': 0.14252450666385724, 'gradnorm/compressor.layer1': 1.1405693458670667, 'gradnorm/compressor.layer2': 0.4891567620791559, 'gradnorm/compressor.layer3': 0.32401620903942985, 'gradnorm/compressor.layer4': 0.14926740517199139, 'gradnorm/compressor.layer5': 0.1098678418378503, 'gradnorm/compressor.layer6': 0.06414468250613414, 'gradnorm/compressor.layer7': 0.1011758238342132, 'gradnorm/decoder.layer0': 0.1723450226879423, 'gradnorm/decoder.layer1': 0.1709466702653269, 'gradnorm/decoder.layer2': 0.10161783013353604, 'gradnorm/max_param': 0.7963917255401611, 'epoch': 3.12}

 31%|███       | 5960/19090 [2:54:29<6:02:46,  1.66s/it]
 31%|███       | 5961/19090 [2:54:30<5:39:27,  1.55s/it]
 31%|███       | 5962/19090 [2:54:33<6:50:54,  1.88s/it]
 31%|███       | 5963/19090 [2:54:34<6:12:32,  1.70s/it]
 31%|███       | 5964/19090 [2:54:36<5:45:28,  1.58s/it]
 31%|███       | 5965/19090 [2:54:38<7:04:15,  1.94s/it]
 31%|███▏      | 5966/19090 [2:54:40<6:22:08,  1.75s/it]
 31%|███▏      | 5967/19090 [2:54:41<5:52:32,  1.61s/it]
 31%|███▏      | 5968/19090 [2:54:43<5:55:10,  1.62s/it]
 31%|███▏      | 5969/19090 [2:54:44<5:33:27,  1.52s/it]
 31%|███▏      | 5970/19090 [2:54:45<5:18:12,  1.46s/it]
                                                        
{'loss': 0.0016, 'grad_norm': 1.4269464015960693, 'learning_rate': 4.342150677114716e-05, 'recon_loss': 0.0015557713224552572, 'recon_static': 0.0006144190090708434, 'recon_motion': 0.0009413522915565409, 'recon_norm': 1.3907948553562164, 'recon_norm_std': 0.07568352818489074, 'target_norm': 1.6193569749593735, 'recon_cos': 0.6444857105612755, 'gradnorm/compressor.query': 0.0074318279782081146, 'gradnorm/compressor.layer0': 0.14690187720651907, 'gradnorm/compressor.layer1': 1.2236955022317897, 'gradnorm/compressor.layer2': 0.5435072060838135, 'gradnorm/compressor.layer3': 0.3215824283582342, 'gradnorm/compressor.layer4': 0.1486315901376435, 'gradnorm/compressor.layer5': 0.11450192465710642, 'gradnorm/compressor.layer6': 0.06553090507960468, 'gradnorm/compressor.layer7': 0.09769363641349744, 'gradnorm/decoder.layer0': 0.17633438502343166, 'gradnorm/decoder.layer1': 0.16666527029783693, 'gradnorm/decoder.layer2': 0.10189792830668958, 'gradnorm/max_param': 0.8556270599365234, 'epoch': 3.13}

 31%|███▏      | 5970/19090 [2:54:45<5:18:12,  1.46s/it]
 31%|███▏      | 5971/19090 [2:54:47<6:10:35,  1.69s/it]
 31%|███▏      | 5972/19090 [2:54:49<5:44:30,  1.58s/it]
 31%|███▏      | 5973/19090 [2:54:50<5:26:06,  1.49s/it]
 31%|███▏      | 5974/19090 [2:54:52<5:48:37,  1.59s/it]
 31%|███▏      | 5975/19090 [2:54:53<5:28:37,  1.50s/it]
 31%|███▏      | 5976/19090 [2:54:54<5:14:50,  1.44s/it]
 31%|███▏      | 5977/19090 [2:54:57<6:00:51,  1.65s/it]
 31%|███▏      | 5978/19090 [2:54:58<5:36:56,  1.54s/it]
 31%|███▏      | 5979/19090 [2:54:59<5:20:20,  1.47s/it]
 31%|███▏      | 5980/19090 [2:55:01<5:42:42,  1.57s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 1.4051436185836792, 'learning_rate': 4.339057182128497e-05, 'recon_loss': 0.0015358740405645222, 'recon_static': 0.0005746046866988763, 'recon_motion': 0.0009612693727831356, 'recon_norm': 1.3757910013198853, 'recon_norm_std': 0.07263874541968107, 'target_norm': 1.6213655531406403, 'recon_cos': 0.6483097866177558, 'gradnorm/compressor.query': 0.007243067468997146, 'gradnorm/compressor.layer0': 0.14009279632754312, 'gradnorm/compressor.layer1': 1.1886263073609027, 'gradnorm/compressor.layer2': 0.5471626342753054, 'gradnorm/compressor.layer3': 0.35803946830118716, 'gradnorm/compressor.layer4': 0.1513791725582184, 'gradnorm/compressor.layer5': 0.10707904121116052, 'gradnorm/compressor.layer6': 0.05923099889997878, 'gradnorm/compressor.layer7': 0.09362285654978349, 'gradnorm/decoder.layer0': 0.17205238400432807, 'gradnorm/decoder.layer1': 0.168190881316289, 'gradnorm/decoder.layer2': 0.09896115614813518, 'gradnorm/max_param': 0.8314539194107056, 'epoch': 3.13}

 31%|███▏      | 5980/19090 [2:55:01<5:42:42,  1.57s/it]
 31%|███▏      | 5981/19090 [2:55:02<5:24:31,  1.49s/it]
 31%|███▏      | 5982/19090 [2:55:04<5:11:45,  1.43s/it]
 31%|███▏      | 5983/19090 [2:55:06<6:03:19,  1.66s/it]
 31%|███▏      | 5984/19090 [2:55:07<5:38:52,  1.55s/it]
 31%|███▏      | 5985/19090 [2:55:08<5:21:32,  1.47s/it]
 31%|███▏      | 5986/19090 [2:55:11<6:33:43,  1.80s/it]
 31%|███▏      | 5987/19090 [2:55:12<6:00:24,  1.65s/it]
 31%|███▏      | 5988/19090 [2:55:14<5:36:34,  1.54s/it]
 31%|███▏      | 5989/19090 [2:55:16<6:35:02,  1.81s/it]
 31%|███▏      | 5990/19090 [2:55:17<6:01:11,  1.65s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 1.3942791223526, 'learning_rate': 4.335957538234468e-05, 'recon_loss': 0.0015530229749856517, 'recon_static': 0.0005909242347115651, 'recon_motion': 0.0009620987562811934, 'recon_norm': 1.379502508044243, 'recon_norm_std': 0.07172908000648022, 'target_norm': 1.6204320728778838, 'recon_cos': 0.644552794098854, 'gradnorm/compressor.query': 0.006702598684547871, 'gradnorm/compressor.layer0': 0.1331142265925901, 'gradnorm/compressor.layer1': 1.19017760914821, 'gradnorm/compressor.layer2': 0.5253079147184136, 'gradnorm/compressor.layer3': 0.35346474465782113, 'gradnorm/compressor.layer4': 0.13778891662219647, 'gradnorm/compressor.layer5': 0.10596391769743085, 'gradnorm/compressor.layer6': 0.05773151499022619, 'gradnorm/compressor.layer7': 0.0984812885105198, 'gradnorm/decoder.layer0': 0.16831758228246957, 'gradnorm/decoder.layer1': 0.16809559069411265, 'gradnorm/decoder.layer2': 0.09491655134648833, 'gradnorm/max_param': 0.8328393697738647, 'epoch': 3.14}

 31%|███▏      | 5990/19090 [2:55:17<6:01:11,  1.65s/it]
 31%|███▏      | 5991/19090 [2:55:19<5:37:39,  1.55s/it]
 31%|███▏      | 5992/19090 [2:55:21<6:30:48,  1.79s/it]
 31%|███▏      | 5993/19090 [2:55:22<5:57:49,  1.64s/it]
 31%|███▏      | 5994/19090 [2:55:23<5:35:04,  1.54s/it]
 31%|███▏      | 5995/19090 [2:55:26<6:22:56,  1.75s/it]
 31%|███▏      | 5996/19090 [2:55:27<5:51:53,  1.61s/it]
 31%|███▏      | 5997/19090 [2:55:28<5:31:01,  1.52s/it]
 31%|███▏      | 5998/19090 [2:55:31<6:32:34,  1.80s/it]
 31%|███▏      | 5999/19090 [2:55:32<5:58:52,  1.64s/it]
 31%|███▏      | 6000/19090 [2:55:33<5:35:31,  1.54s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 1.4756797552108765, 'learning_rate': 4.3328517557963225e-05, 'recon_loss': 0.0015125914360396564, 'recon_static': 0.0005530763206479606, 'recon_motion': 0.0009595151379471645, 'recon_norm': 1.3619378119707108, 'recon_norm_std': 0.07243864219635725, 'target_norm': 1.6197101384401322, 'recon_cos': 0.6515352934598923, 'gradnorm/compressor.query': 0.007196423105683475, 'gradnorm/compressor.layer0': 0.1430232467908479, 'gradnorm/compressor.layer1': 1.2686152043638623, 'gradnorm/compressor.layer2': 0.5432574924631695, 'gradnorm/compressor.layer3': 0.36621396890021385, 'gradnorm/compressor.layer4': 0.15509538818668667, 'gradnorm/compressor.layer5': 0.11021768838367392, 'gradnorm/compressor.layer6': 0.05935277656269023, 'gradnorm/compressor.layer7': 0.09811705114429718, 'gradnorm/decoder.layer0': 0.1801166902732563, 'gradnorm/decoder.layer1': 0.16540084776640912, 'gradnorm/decoder.layer2': 0.09667408088842015, 'gradnorm/max_param': 0.8881795406341553, 'epoch': 3.14}

 31%|███▏      | 6000/19090 [2:55:33<5:35:31,  1.54s/it]
 31%|███▏      | 6001/19090 [2:55:38<8:37:05,  2.37s/it]
 31%|███▏      | 6002/19090 [2:55:39<7:26:26,  2.05s/it]
 31%|███▏      | 6003/19090 [2:55:40<6:36:30,  1.82s/it]
 31%|███▏      | 6004/19090 [2:55:42<6:01:38,  1.66s/it]
 31%|███▏      | 6005/19090 [2:55:43<5:37:25,  1.55s/it]
 31%|███▏      | 6006/19090 [2:55:44<5:21:29,  1.47s/it]
 31%|███▏      | 6007/19090 [2:55:45<5:09:14,  1.42s/it]
 31%|███▏      | 6008/19090 [2:55:47<5:00:39,  1.38s/it]
 31%|███▏      | 6009/19090 [2:55:48<4:54:30,  1.35s/it]
 31%|███▏      | 6010/19090 [2:55:49<4:50:24,  1.33s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 1.4802032709121704, 'learning_rate': 4.329739845198277e-05, 'recon_loss': 0.0014696323487441987, 'recon_static': 0.0005183406894502696, 'recon_motion': 0.0009512916862149723, 'recon_norm': 1.3503054976463318, 'recon_norm_std': 0.07411549240350723, 'target_norm': 1.618976154923439, 'recon_cos': 0.6614076480269432, 'gradnorm/compressor.query': 0.006956402746169481, 'gradnorm/compressor.layer0': 0.1282194583142487, 'gradnorm/compressor.layer1': 1.2679314495624319, 'gradnorm/compressor.layer2': 0.5744144339821298, 'gradnorm/compressor.layer3': 0.3631073730445079, 'gradnorm/compressor.layer4': 0.14835160681548962, 'gradnorm/compressor.layer5': 0.10897986101899698, 'gradnorm/compressor.layer6': 0.04956415473321143, 'gradnorm/compressor.layer7': 0.09105432264826865, 'gradnorm/decoder.layer0': 0.1658344402068625, 'gradnorm/decoder.layer1': 0.1596097869326242, 'gradnorm/decoder.layer2': 0.0863437532717621, 'gradnorm/max_param': 0.889026403427124, 'epoch': 3.15}

 31%|███▏      | 6010/19090 [2:55:49<4:50:24,  1.33s/it]
 31%|███▏      | 6011/19090 [2:55:51<4:47:34,  1.32s/it]
 31%|███▏      | 6012/19090 [2:55:52<4:45:04,  1.31s/it]
 31%|███▏      | 6013/19090 [2:55:54<5:29:51,  1.51s/it]
 32%|███▏      | 6014/19090 [2:55:55<5:15:02,  1.45s/it]
 32%|███▏      | 6015/19090 [2:55:56<5:05:15,  1.40s/it]
 32%|███▏      | 6016/19090 [2:55:59<6:43:30,  1.85s/it]
 32%|███▏      | 6017/19090 [2:56:01<6:07:36,  1.69s/it]
 32%|███▏      | 6018/19090 [2:56:02<5:42:11,  1.57s/it]
 32%|███▏      | 6019/19090 [2:56:05<7:25:42,  2.05s/it]
 32%|███▏      | 6020/19090 [2:56:06<6:36:27,  1.82s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 1.5750081539154053, 'learning_rate': 4.3266218168450396e-05, 'recon_loss': 0.0015017733530839905, 'recon_static': 0.0005458960658870637, 'recon_motion': 0.0009558772755553946, 'recon_norm': 1.3612279564142227, 'recon_norm_std': 0.07280547078698874, 'target_norm': 1.622167345881462, 'recon_cos': 0.6555197104811669, 'gradnorm/compressor.query': 0.007508067301671692, 'gradnorm/compressor.layer0': 0.15406091973614322, 'gradnorm/compressor.layer1': 1.3439080788535944, 'gradnorm/compressor.layer2': 0.6093646562138553, 'gradnorm/compressor.layer3': 0.3918439348127448, 'gradnorm/compressor.layer4': 0.15801561098222228, 'gradnorm/compressor.layer5': 0.1128855156319747, 'gradnorm/compressor.layer6': 0.06323844925126881, 'gradnorm/compressor.layer7': 0.09706328614132108, 'gradnorm/decoder.layer0': 0.1932311451665828, 'gradnorm/decoder.layer1': 0.16277193346603464, 'gradnorm/decoder.layer2': 0.10458643397481755, 'gradnorm/max_param': 0.9405731558799744, 'epoch': 3.15}

 32%|███▏      | 6020/19090 [2:56:06<6:36:27,  1.82s/it]
 32%|███▏      | 6021/19090 [2:56:08<6:01:27,  1.66s/it]
 32%|███▏      | 6022/19090 [2:56:10<7:07:28,  1.96s/it]
 32%|███▏      | 6023/19090 [2:56:12<6:22:51,  1.76s/it]
 32%|███▏      | 6024/19090 [2:56:13<5:51:50,  1.62s/it]
 32%|███▏      | 6025/19090 [2:56:16<7:21:07,  2.03s/it]
 32%|███▏      | 6026/19090 [2:56:17<6:32:47,  1.80s/it]
 32%|███▏      | 6027/19090 [2:56:18<5:58:38,  1.65s/it]
 32%|███▏      | 6028/19090 [2:56:21<7:24:22,  2.04s/it]
 32%|███▏      | 6029/19090 [2:56:23<6:34:50,  1.81s/it]
 32%|███▏      | 6030/19090 [2:56:24<6:00:51,  1.66s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 1.6586905717849731, 'learning_rate': 4.3234976811617723e-05, 'recon_loss': 0.0014836285234196111, 'recon_static': 0.0005305641105223912, 'recon_motion': 0.0009530644500046037, 'recon_norm': 1.3540623158216476, 'recon_norm_std': 0.07289846893399954, 'target_norm': 1.6186607301235199, 'recon_cos': 0.6578411027789116, 'gradnorm/compressor.query': 0.008174762288614652, 'gradnorm/compressor.layer0': 0.15374350002939507, 'gradnorm/compressor.layer1': 1.449702508597484, 'gradnorm/compressor.layer2': 0.5933725261423709, 'gradnorm/compressor.layer3': 0.39981797368159605, 'gradnorm/compressor.layer4': 0.15223069152831392, 'gradnorm/compressor.layer5': 0.11014474206698559, 'gradnorm/compressor.layer6': 0.05751452553733768, 'gradnorm/compressor.layer7': 0.0997503815956752, 'gradnorm/decoder.layer0': 0.172968049319329, 'gradnorm/decoder.layer1': 0.16194623183519893, 'gradnorm/decoder.layer2': 0.09622394185803813, 'gradnorm/max_param': 1.016474962234497, 'epoch': 3.16}

 32%|███▏      | 6030/19090 [2:56:24<6:00:51,  1.66s/it]
 32%|███▏      | 6031/19090 [2:56:27<7:06:37,  1.96s/it]
 32%|███▏      | 6032/19090 [2:56:28<6:22:42,  1.76s/it]
 32%|███▏      | 6033/19090 [2:56:29<5:52:02,  1.62s/it]
 32%|███▏      | 6034/19090 [2:56:32<7:10:23,  1.98s/it]
 32%|███▏      | 6035/19090 [2:56:33<6:26:26,  1.78s/it]
 32%|███▏      | 6036/19090 [2:56:35<5:54:39,  1.63s/it]
 32%|███▏      | 6037/19090 [2:56:37<7:02:17,  1.94s/it]
 32%|███▏      | 6038/19090 [2:56:39<6:19:27,  1.74s/it]
 32%|███▏      | 6039/19090 [2:56:40<5:49:56,  1.61s/it]
 32%|███▏      | 6040/19090 [2:56:42<6:45:20,  1.86s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 1.4959161281585693, 'learning_rate': 4.320367448594057e-05, 'recon_loss': 0.0014419783372431993, 'recon_static': 0.00048260225157719104, 'recon_motion': 0.0009593761089490727, 'recon_norm': 1.3333636224269867, 'recon_norm_std': 0.07473372556269169, 'target_norm': 1.6203978151082992, 'recon_cos': 0.6673557907342911, 'gradnorm/compressor.query': 0.0070521300450980895, 'gradnorm/compressor.layer0': 0.13922283032835261, 'gradnorm/compressor.layer1': 1.277437585772921, 'gradnorm/compressor.layer2': 0.5723821632018663, 'gradnorm/compressor.layer3': 0.37334530063306176, 'gradnorm/compressor.layer4': 0.15615379829039433, 'gradnorm/compressor.layer5': 0.10882229032163088, 'gradnorm/compressor.layer6': 0.05459925224147827, 'gradnorm/compressor.layer7': 0.10319083471270266, 'gradnorm/decoder.layer0': 0.18440347023352238, 'gradnorm/decoder.layer1': 0.16619752567567642, 'gradnorm/decoder.layer2': 0.08941729548497741, 'gradnorm/max_param': 0.8955687880516052, 'epoch': 3.16}

 32%|███▏      | 6040/19090 [2:56:42<6:45:20,  1.86s/it]
 32%|███▏      | 6041/19090 [2:56:44<6:07:53,  1.69s/it]
 32%|███▏      | 6042/19090 [2:56:45<5:41:33,  1.57s/it]
 32%|███▏      | 6043/19090 [2:56:48<7:12:14,  1.99s/it]
 32%|███▏      | 6044/19090 [2:56:49<6:26:36,  1.78s/it]
 32%|███▏      | 6045/19090 [2:56:50<5:55:03,  1.63s/it]
 32%|███▏      | 6046/19090 [2:56:53<7:24:11,  2.04s/it]
 32%|███▏      | 6047/19090 [2:56:55<6:34:51,  1.82s/it]
 32%|███▏      | 6048/19090 [2:56:56<6:00:20,  1.66s/it]
 32%|███▏      | 6049/19090 [2:56:59<7:05:53,  1.96s/it]
 32%|███▏      | 6050/19090 [2:57:00<6:22:18,  1.76s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 1.585653305053711, 'learning_rate': 4.3172311296078595e-05, 'recon_loss': 0.001464803473209031, 'recon_static': 0.0005124605042510666, 'recon_motion': 0.000952342982054688, 'recon_norm': 1.3451059281826019, 'recon_norm_std': 0.07374178580939769, 'target_norm': 1.6194838106632232, 'recon_cos': 0.6621927544474602, 'gradnorm/compressor.query': 0.006913115326998382, 'gradnorm/compressor.layer0': 0.12373908930421722, 'gradnorm/compressor.layer1': 1.384039038543054, 'gradnorm/compressor.layer2': 0.5832995734209463, 'gradnorm/compressor.layer3': 0.3663880797153015, 'gradnorm/compressor.layer4': 0.15021097922496918, 'gradnorm/compressor.layer5': 0.10381763863835902, 'gradnorm/compressor.layer6': 0.04733223733418723, 'gradnorm/compressor.layer7': 0.09140314724538569, 'gradnorm/decoder.layer0': 0.1755489828155512, 'gradnorm/decoder.layer1': 0.16556369753628056, 'gradnorm/decoder.layer2': 0.08203148841823263, 'gradnorm/max_param': 0.971884548664093, 'epoch': 3.17}

 32%|███▏      | 6050/19090 [2:57:00<6:22:18,  1.76s/it]
 32%|███▏      | 6051/19090 [2:57:01<5:51:36,  1.62s/it]
 32%|███▏      | 6052/19090 [2:57:04<6:47:38,  1.88s/it]
 32%|███▏      | 6053/19090 [2:57:05<6:09:01,  1.70s/it]
 32%|███▏      | 6054/19090 [2:57:06<5:42:11,  1.57s/it]
 32%|███▏      | 6055/19090 [2:57:09<6:51:36,  1.89s/it]
 32%|███▏      | 6056/19090 [2:57:10<6:11:49,  1.71s/it]
 32%|███▏      | 6057/19090 [2:57:12<5:44:04,  1.58s/it]
 32%|███▏      | 6058/19090 [2:57:14<7:11:25,  1.99s/it]
 32%|███▏      | 6059/19090 [2:57:16<6:25:56,  1.78s/it]
 32%|███▏      | 6060/19090 [2:57:17<5:53:59,  1.63s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 1.6255131959915161, 'learning_rate': 4.3140887346894974e-05, 'recon_loss': 0.0014415861282031984, 'recon_static': 0.0004881596330960747, 'recon_motion': 0.0009534265016554854, 'recon_norm': 1.335596150159836, 'recon_norm_std': 0.07414803095161915, 'target_norm': 1.6223563969135284, 'recon_cos': 0.668103052675724, 'gradnorm/compressor.query': 0.006833473574699048, 'gradnorm/compressor.layer0': 0.1374940077061156, 'gradnorm/compressor.layer1': 1.4186896391140371, 'gradnorm/compressor.layer2': 0.6016939688358891, 'gradnorm/compressor.layer3': 0.36626350534165125, 'gradnorm/compressor.layer4': 0.1487388125087362, 'gradnorm/compressor.layer5': 0.112061652217558, 'gradnorm/compressor.layer6': 0.05513986963301674, 'gradnorm/compressor.layer7': 0.09731891064883244, 'gradnorm/decoder.layer0': 0.18077961087617145, 'gradnorm/decoder.layer1': 0.15960962938763648, 'gradnorm/decoder.layer2': 0.09554113087734016, 'gradnorm/max_param': 0.9952378869056702, 'epoch': 3.17}

 32%|███▏      | 6060/19090 [2:57:17<5:53:59,  1.63s/it]
 32%|███▏      | 6061/19090 [2:57:20<7:11:42,  1.99s/it]
 32%|███▏      | 6062/19090 [2:57:21<6:25:39,  1.78s/it]
 32%|███▏      | 6063/19090 [2:57:22<5:53:40,  1.63s/it]
 32%|███▏      | 6064/19090 [2:57:25<7:12:07,  1.99s/it]
 32%|███▏      | 6065/19090 [2:57:27<6:26:16,  1.78s/it]
 32%|███▏      | 6066/19090 [2:57:28<5:54:11,  1.63s/it]
 32%|███▏      | 6067/19090 [2:57:31<7:22:52,  2.04s/it]
 32%|███▏      | 6068/19090 [2:57:32<6:33:34,  1.81s/it]
 32%|███▏      | 6069/19090 [2:57:33<5:59:04,  1.65s/it]
 32%|███▏      | 6070/19090 [2:57:36<7:22:05,  2.04s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 1.7017662525177002, 'learning_rate': 4.310940274345603e-05, 'recon_loss': 0.001402778972988017, 'recon_static': 0.000446641820599325, 'recon_motion': 0.0009561371378367766, 'recon_norm': 1.3149354755878448, 'recon_norm_std': 0.07579184789210558, 'target_norm': 1.6207516938447952, 'recon_cos': 0.675671610236168, 'gradnorm/compressor.query': 0.006510707108655021, 'gradnorm/compressor.layer0': 0.1172107713163322, 'gradnorm/compressor.layer1': 1.487432122485004, 'gradnorm/compressor.layer2': 0.637640350450171, 'gradnorm/compressor.layer3': 0.3872920708998769, 'gradnorm/compressor.layer4': 0.1475346425257671, 'gradnorm/compressor.layer5': 0.10807408098566462, 'gradnorm/compressor.layer6': 0.046611357400574475, 'gradnorm/compressor.layer7': 0.09909731481047385, 'gradnorm/decoder.layer0': 0.17972687621525574, 'gradnorm/decoder.layer1': 0.16904439942070199, 'gradnorm/decoder.layer2': 0.08292142355644869, 'gradnorm/max_param': 1.045224666595459, 'epoch': 3.18}

 32%|███▏      | 6070/19090 [2:57:36<7:22:05,  2.04s/it]
 32%|███▏      | 6071/19090 [2:57:38<6:33:07,  1.81s/it]
 32%|███▏      | 6072/19090 [2:57:39<5:58:52,  1.65s/it]
 32%|███▏      | 6073/19090 [2:57:42<7:06:21,  1.97s/it]
 32%|███▏      | 6074/19090 [2:57:43<6:21:57,  1.76s/it]
 32%|███▏      | 6075/19090 [2:57:44<5:51:17,  1.62s/it]
 32%|███▏      | 6076/19090 [2:57:47<7:12:57,  2.00s/it]
 32%|███▏      | 6077/19090 [2:57:48<6:26:25,  1.78s/it]
 32%|███▏      | 6078/19090 [2:57:50<5:54:21,  1.63s/it]
 32%|███▏      | 6079/19090 [2:57:52<7:12:03,  1.99s/it]
 32%|███▏      | 6080/19090 [2:57:54<6:26:13,  1.78s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 1.685541033744812, 'learning_rate': 4.307785759103087e-05, 'recon_loss': 0.001383169327164069, 'recon_static': 0.00041944281401811167, 'recon_motion': 0.0009637265407945961, 'recon_norm': 1.3029275447130204, 'recon_norm_std': 0.0745932662859559, 'target_norm': 1.6207855612039566, 'recon_cos': 0.6798884198069572, 'gradnorm/compressor.query': 0.005941560598925304, 'gradnorm/compressor.layer0': 0.12447249972939509, 'gradnorm/compressor.layer1': 1.4658238908392365, 'gradnorm/compressor.layer2': 0.6294826685570116, 'gradnorm/compressor.layer3': 0.39895256260747947, 'gradnorm/compressor.layer4': 0.1591868837754267, 'gradnorm/compressor.layer5': 0.1115144709307836, 'gradnorm/compressor.layer6': 0.047621553719975875, 'gradnorm/compressor.layer7': 0.10004628421645358, 'gradnorm/decoder.layer0': 0.18542277552867456, 'gradnorm/decoder.layer1': 0.17207543650173704, 'gradnorm/decoder.layer2': 0.08616414851690127, 'gradnorm/max_param': 1.0297318696975708, 'epoch': 3.18}

 32%|███▏      | 6080/19090 [2:57:54<6:26:13,  1.78s/it]
 32%|███▏      | 6081/19090 [2:57:55<5:54:36,  1.64s/it]
 32%|███▏      | 6082/19090 [2:57:58<7:11:07,  1.99s/it]
 32%|███▏      | 6083/19090 [2:57:59<6:25:23,  1.78s/it]
 32%|███▏      | 6084/19090 [2:58:00<5:53:24,  1.63s/it]
 32%|███▏      | 6085/19090 [2:58:03<6:55:08,  1.92s/it]
 32%|███▏      | 6086/19090 [2:58:04<6:14:11,  1.73s/it]
 32%|███▏      | 6087/19090 [2:58:06<5:51:08,  1.62s/it]
 32%|███▏      | 6088/19090 [2:58:08<6:52:55,  1.91s/it]
 32%|███▏      | 6089/19090 [2:58:09<6:12:15,  1.72s/it]
 32%|███▏      | 6090/19090 [2:58:11<6:12:52,  1.72s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 1.8092470169067383, 'learning_rate': 4.304625199509107e-05, 'recon_loss': 0.0013826735346810892, 'recon_static': 0.0004289267024432775, 'recon_motion': 0.0009537468155031093, 'recon_norm': 1.3083735436201096, 'recon_norm_std': 0.07500946931540967, 'target_norm': 1.6199655741453172, 'recon_cos': 0.6802330613136292, 'gradnorm/compressor.query': 0.006569342316460947, 'gradnorm/compressor.layer0': 0.12908048125802654, 'gradnorm/compressor.layer1': 1.5797050799590528, 'gradnorm/compressor.layer2': 0.6708047904187587, 'gradnorm/compressor.layer3': 0.43826122136174717, 'gradnorm/compressor.layer4': 0.15488185930379678, 'gradnorm/compressor.layer5': 0.11948105764915817, 'gradnorm/compressor.layer6': 0.048562732917670166, 'gradnorm/compressor.layer7': 0.10352517929609738, 'gradnorm/decoder.layer0': 0.17932988945349798, 'gradnorm/decoder.layer1': 0.16796827316216495, 'gradnorm/decoder.layer2': 0.08623866121163744, 'gradnorm/max_param': 1.1098361015319824, 'epoch': 3.19}

 32%|███▏      | 6090/19090 [2:58:11<6:12:52,  1.72s/it]
 32%|███▏      | 6091/19090 [2:58:14<7:05:50,  1.97s/it]
 32%|███▏      | 6092/19090 [2:58:15<6:21:20,  1.76s/it]
 32%|███▏      | 6093/19090 [2:58:16<6:01:31,  1.67s/it]
 32%|███▏      | 6094/19090 [2:58:19<6:44:00,  1.87s/it]
 32%|███▏      | 6095/19090 [2:58:20<6:06:08,  1.69s/it]
 32%|███▏      | 6096/19090 [2:58:22<6:21:23,  1.76s/it]
 32%|███▏      | 6097/19090 [2:58:24<6:44:16,  1.87s/it]
 32%|███▏      | 6098/19090 [2:58:25<6:06:48,  1.69s/it]
 32%|███▏      | 6099/19090 [2:58:27<6:26:17,  1.78s/it]
 32%|███▏      | 6100/19090 [2:58:29<6:38:45,  1.84s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 1.7455790042877197, 'learning_rate': 4.301458606131026e-05, 'recon_loss': 0.001379374897805974, 'recon_static': 0.00042161385135841556, 'recon_motion': 0.0009577610471751541, 'recon_norm': 1.3033467382192612, 'recon_norm_std': 0.0758617864921689, 'target_norm': 1.620815846323967, 'recon_cos': 0.6810414731502533, 'gradnorm/compressor.query': 0.006355385247190752, 'gradnorm/compressor.layer0': 0.11931645988304543, 'gradnorm/compressor.layer1': 1.538515874930158, 'gradnorm/compressor.layer2': 0.6145592754849317, 'gradnorm/compressor.layer3': 0.40982310980111514, 'gradnorm/compressor.layer4': 0.15705253020376458, 'gradnorm/compressor.layer5': 0.10815664336197199, 'gradnorm/compressor.layer6': 0.04552707949839475, 'gradnorm/compressor.layer7': 0.10243455011136429, 'gradnorm/decoder.layer0': 0.1882505692616502, 'gradnorm/decoder.layer1': 0.17082466167929783, 'gradnorm/decoder.layer2': 0.08094320903990827, 'gradnorm/max_param': 1.0814135074615479, 'epoch': 3.2}

 32%|███▏      | 6100/19090 [2:58:29<6:38:45,  1.84s/it]
 32%|███▏      | 6101/19090 [2:58:31<6:02:45,  1.68s/it]
 32%|███▏      | 6102/19090 [2:58:33<6:52:19,  1.90s/it]
 32%|███▏      | 6103/19090 [2:58:35<6:27:43,  1.79s/it]
 32%|███▏      | 6104/19090 [2:58:36<5:54:55,  1.64s/it]
 32%|███▏      | 6105/19090 [2:58:38<6:45:41,  1.87s/it]
 32%|███▏      | 6106/19090 [2:58:40<6:16:54,  1.74s/it]
 32%|███▏      | 6107/19090 [2:58:41<5:50:24,  1.62s/it]
 32%|███▏      | 6108/19090 [2:58:44<6:51:17,  1.90s/it]
 32%|███▏      | 6109/19090 [2:58:46<6:48:46,  1.89s/it]
 32%|███▏      | 6110/19090 [2:58:47<6:09:45,  1.71s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 1.727946400642395, 'learning_rate': 4.298285989556386e-05, 'recon_loss': 0.0013642622012412175, 'recon_static': 0.00040564592854934743, 'recon_motion': 0.0009586163039784878, 'recon_norm': 1.2970224797725678, 'recon_norm_std': 0.0738068986684084, 'target_norm': 1.622213861346245, 'recon_cos': 0.6854019552469254, 'gradnorm/compressor.query': 0.00547762683463602, 'gradnorm/compressor.layer0': 0.1139566005023526, 'gradnorm/compressor.layer1': 1.5079613646710908, 'gradnorm/compressor.layer2': 0.6297220424141796, 'gradnorm/compressor.layer3': 0.4231714531939673, 'gradnorm/compressor.layer4': 0.16284951187456803, 'gradnorm/compressor.layer5': 0.1198216656426092, 'gradnorm/compressor.layer6': 0.04248639043553432, 'gradnorm/compressor.layer7': 0.10120116660370151, 'gradnorm/decoder.layer0': 0.17700381573979562, 'gradnorm/decoder.layer1': 0.1801051489853054, 'gradnorm/decoder.layer2': 0.08073607121699942, 'gradnorm/max_param': 1.0600730180740356, 'epoch': 3.2}

 32%|███▏      | 6110/19090 [2:58:47<6:09:45,  1.71s/it]
 32%|███▏      | 6111/19090 [2:58:49<6:35:09,  1.83s/it]
 32%|███▏      | 6112/19090 [2:58:51<6:38:31,  1.84s/it]
 32%|███▏      | 6113/19090 [2:58:52<6:02:52,  1.68s/it]
 32%|███▏      | 6114/19090 [2:58:54<6:43:40,  1.87s/it]
 32%|███▏      | 6115/19090 [2:58:56<6:37:30,  1.84s/it]
 32%|███▏      | 6116/19090 [2:58:57<6:01:31,  1.67s/it]
 32%|███▏      | 6117/19090 [2:59:00<6:39:20,  1.85s/it]
 32%|███▏      | 6118/19090 [2:59:02<6:48:05,  1.89s/it]
 32%|███▏      | 6119/19090 [2:59:03<6:09:07,  1.71s/it]
 32%|███▏      | 6120/19090 [2:59:05<6:21:39,  1.77s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 1.9035097360610962, 'learning_rate': 4.2951073603928646e-05, 'recon_loss': 0.0013529103161999956, 'recon_static': 0.0004100178106455132, 'recon_motion': 0.0009428925113752484, 'recon_norm': 1.3006668359041214, 'recon_norm_std': 0.07535932827740907, 'target_norm': 1.6180043399333954, 'recon_cos': 0.686334790289402, 'gradnorm/compressor.query': 0.00639573985373243, 'gradnorm/compressor.layer0': 0.1263892857540395, 'gradnorm/compressor.layer1': 1.6861576109871717, 'gradnorm/compressor.layer2': 0.6825008702534587, 'gradnorm/compressor.layer3': 0.41752868730814835, 'gradnorm/compressor.layer4': 0.16594756919318102, 'gradnorm/compressor.layer5': 0.11293414392456382, 'gradnorm/compressor.layer6': 0.04610544557416502, 'gradnorm/compressor.layer7': 0.10040694115297263, 'gradnorm/decoder.layer0': 0.1827943931480887, 'gradnorm/decoder.layer1': 0.17647765281475872, 'gradnorm/decoder.layer2': 0.08373428412559035, 'gradnorm/max_param': 1.1853326559066772, 'epoch': 3.21}

 32%|███▏      | 6120/19090 [2:59:05<6:21:39,  1.77s/it]
 32%|███▏      | 6121/19090 [2:59:07<6:32:11,  1.81s/it]
 32%|███▏      | 6122/19090 [2:59:08<5:57:26,  1.65s/it]
 32%|███▏      | 6123/19090 [2:59:11<6:48:10,  1.89s/it]
 32%|███▏      | 6124/19090 [2:59:12<6:22:34,  1.77s/it]
 32%|███▏      | 6125/19090 [2:59:13<5:50:42,  1.62s/it]
 32%|███▏      | 6126/19090 [2:59:16<6:36:55,  1.84s/it]
 32%|███▏      | 6127/19090 [2:59:17<6:03:27,  1.68s/it]
 32%|███▏      | 6128/19090 [2:59:18<5:37:21,  1.56s/it]
 32%|███▏      | 6129/19090 [2:59:21<6:57:39,  1.93s/it]
 32%|███▏      | 6130/19090 [2:59:22<6:24:56,  1.78s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 1.8782683610916138, 'learning_rate': 4.291922729268243e-05, 'recon_loss': 0.001324161549564451, 'recon_static': 0.00037639591027982533, 'recon_motion': 0.0009477655781665817, 'recon_norm': 1.2841884404420854, 'recon_norm_std': 0.07492780834436416, 'target_norm': 1.6186036795377732, 'recon_cos': 0.6933953955769538, 'gradnorm/compressor.query': 0.006312635332333761, 'gradnorm/compressor.layer0': 0.1194502450060657, 'gradnorm/compressor.layer1': 1.644538136374945, 'gradnorm/compressor.layer2': 0.7092479289511723, 'gradnorm/compressor.layer3': 0.42488542863707285, 'gradnorm/compressor.layer4': 0.1488574825077275, 'gradnorm/compressor.layer5': 0.12145700499038331, 'gradnorm/compressor.layer6': 0.042184758262816965, 'gradnorm/compressor.layer7': 0.10280905101811691, 'gradnorm/decoder.layer0': 0.1964423401995144, 'gradnorm/decoder.layer1': 0.17586820359101948, 'gradnorm/decoder.layer2': 0.082089695659198, 'gradnorm/max_param': 1.1566884517669678, 'epoch': 3.21}

 32%|███▏      | 6130/19090 [2:59:22<6:24:56,  1.78s/it]
 32%|███▏      | 6131/19090 [2:59:24<5:52:51,  1.63s/it]
 32%|███▏      | 6132/19090 [2:59:26<7:00:11,  1.95s/it]
 32%|███▏      | 6133/19090 [2:59:28<6:20:03,  1.76s/it]
 32%|███▏      | 6134/19090 [2:59:29<5:54:45,  1.64s/it]
 32%|███▏      | 6135/19090 [2:59:32<7:04:38,  1.97s/it]
 32%|███▏      | 6136/19090 [2:59:33<6:20:30,  1.76s/it]
 32%|███▏      | 6137/19090 [2:59:34<5:49:18,  1.62s/it]
 32%|███▏      | 6138/19090 [2:59:37<7:01:16,  1.95s/it]
 32%|███▏      | 6139/19090 [2:59:38<6:18:12,  1.75s/it]
 32%|███▏      | 6140/19090 [2:59:40<5:48:16,  1.61s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 1.8339331150054932, 'learning_rate': 4.2887321068303696e-05, 'recon_loss': 0.001321478094905615, 'recon_static': 0.0003814190538832918, 'recon_motion': 0.000940059068670962, 'recon_norm': 1.2843123614788055, 'recon_norm_std': 0.07447463124990464, 'target_norm': 1.6156243801116943, 'recon_cos': 0.6928260371088981, 'gradnorm/compressor.query': 0.005688496784554913, 'gradnorm/compressor.layer0': 0.1137656878366547, 'gradnorm/compressor.layer1': 1.6239777797719757, 'gradnorm/compressor.layer2': 0.6436854274548601, 'gradnorm/compressor.layer3': 0.4157958188778383, 'gradnorm/compressor.layer4': 0.15566605287856486, 'gradnorm/compressor.layer5': 0.11552780180297746, 'gradnorm/compressor.layer6': 0.04374819628369272, 'gradnorm/compressor.layer7': 0.10398192016604342, 'gradnorm/decoder.layer0': 0.19290942040635517, 'gradnorm/decoder.layer1': 0.17749892348648078, 'gradnorm/decoder.layer2': 0.08239944173056782, 'gradnorm/max_param': 1.1422905921936035, 'epoch': 3.22}

 32%|███▏      | 6140/19090 [2:59:40<5:48:16,  1.61s/it]
 32%|███▏      | 6141/19090 [2:59:42<6:54:44,  1.92s/it]
 32%|███▏      | 6142/19090 [2:59:44<6:34:37,  1.83s/it]
 32%|███▏      | 6143/19090 [2:59:45<6:00:27,  1.67s/it]
 32%|███▏      | 6144/19090 [2:59:48<7:10:54,  2.00s/it]
 32%|███▏      | 6145/19090 [2:59:49<6:24:39,  1.78s/it]
 32%|███▏      | 6146/19090 [2:59:51<5:52:22,  1.63s/it]
 32%|███▏      | 6147/19090 [2:59:53<7:12:37,  2.01s/it]
 32%|███▏      | 6148/19090 [2:59:55<6:26:13,  1.79s/it]
 32%|███▏      | 6149/19090 [2:59:56<5:53:22,  1.64s/it]
 32%|███▏      | 6150/19090 [2:59:59<7:14:07,  2.01s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 1.8823421001434326, 'learning_rate': 4.285535503747127e-05, 'recon_loss': 0.0013313692266819998, 'recon_static': 0.0003816135285887867, 'recon_motion': 0.0009497556995484047, 'recon_norm': 1.2840002447366714, 'recon_norm_std': 0.07492467239499093, 'target_norm': 1.6205023735761643, 'recon_cos': 0.6924637630581856, 'gradnorm/compressor.query': 0.006518294955643978, 'gradnorm/compressor.layer0': 0.1313204825842348, 'gradnorm/compressor.layer1': 1.6624590531662997, 'gradnorm/compressor.layer2': 0.6709678869380022, 'gradnorm/compressor.layer3': 0.42410271183366594, 'gradnorm/compressor.layer4': 0.17167936917428459, 'gradnorm/compressor.layer5': 0.1243727862578015, 'gradnorm/compressor.layer6': 0.04796147617134521, 'gradnorm/compressor.layer7': 0.10580754172401759, 'gradnorm/decoder.layer0': 0.1793062581617686, 'gradnorm/decoder.layer1': 0.18390275436553558, 'gradnorm/decoder.layer2': 0.08766177370819542, 'gradnorm/max_param': 1.1684068441390991, 'epoch': 3.22}

 32%|███▏      | 6150/19090 [2:59:59<7:14:07,  2.01s/it]
 32%|███▏      | 6151/19090 [3:00:00<6:27:20,  1.80s/it]
 32%|███▏      | 6152/19090 [3:00:02<5:54:55,  1.65s/it]
 32%|███▏      | 6153/19090 [3:00:04<7:05:24,  1.97s/it]
 32%|███▏      | 6154/19090 [3:00:06<6:23:01,  1.78s/it]
 32%|███▏      | 6155/19090 [3:00:07<5:51:32,  1.63s/it]
 32%|███▏      | 6156/19090 [3:00:10<7:18:44,  2.04s/it]
 32%|███▏      | 6157/19090 [3:00:11<6:31:45,  1.82s/it]
 32%|███▏      | 6158/19090 [3:00:12<5:58:17,  1.66s/it]
 32%|███▏      | 6159/19090 [3:00:15<7:25:39,  2.07s/it]
 32%|███▏      | 6160/19090 [3:00:17<6:36:36,  1.84s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.0462121963500977, 'learning_rate': 4.282332930706391e-05, 'recon_loss': 0.0013431769504677505, 'recon_static': 0.000381332232063869, 'recon_motion': 0.0009618446871172637, 'recon_norm': 1.2868184477090836, 'recon_norm_std': 0.07323545385152101, 'target_norm': 1.6205480068922042, 'recon_cos': 0.6898215278983116, 'gradnorm/compressor.query': 0.006768365500848685, 'gradnorm/compressor.layer0': 0.11768520285850728, 'gradnorm/compressor.layer1': 1.8096503514366957, 'gradnorm/compressor.layer2': 0.7609370597822416, 'gradnorm/compressor.layer3': 0.43049036792121154, 'gradnorm/compressor.layer4': 0.17246868811673816, 'gradnorm/compressor.layer5': 0.11992373822283822, 'gradnorm/compressor.layer6': 0.04037478613473877, 'gradnorm/compressor.layer7': 0.10731380401701059, 'gradnorm/decoder.layer0': 0.1901468768979509, 'gradnorm/decoder.layer1': 0.18433023410388222, 'gradnorm/decoder.layer2': 0.08061928025553407, 'gradnorm/max_param': 1.2736150026321411, 'epoch': 3.23}

 32%|███▏      | 6160/19090 [3:00:17<6:36:36,  1.84s/it]
 32%|███▏      | 6161/19090 [3:00:18<6:02:39,  1.68s/it]
 32%|███▏      | 6162/19090 [3:00:21<7:03:37,  1.97s/it]
 32%|███▏      | 6163/19090 [3:00:22<6:20:12,  1.76s/it]
 32%|███▏      | 6164/19090 [3:00:23<5:49:25,  1.62s/it]
 32%|███▏      | 6165/19090 [3:00:26<6:45:32,  1.88s/it]
 32%|███▏      | 6166/19090 [3:00:27<6:08:05,  1.71s/it]
 32%|███▏      | 6167/19090 [3:00:28<5:41:05,  1.58s/it]
 32%|███▏      | 6168/19090 [3:00:31<6:55:52,  1.93s/it]
 32%|███▏      | 6169/19090 [3:00:32<6:14:24,  1.74s/it]
 32%|███▏      | 6170/19090 [3:00:34<5:45:47,  1.61s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 1.7871078252792358, 'learning_rate': 4.279124398416e-05, 'recon_loss': 0.0013653867063112557, 'recon_static': 0.000396853408892639, 'recon_motion': 0.0009685332959634251, 'recon_norm': 1.288452473282814, 'recon_norm_std': 0.0730592267587781, 'target_norm': 1.6221557945013045, 'recon_cos': 0.6842857703566552, 'gradnorm/compressor.query': 0.006657633394062051, 'gradnorm/compressor.layer0': 0.13686076690100166, 'gradnorm/compressor.layer1': 1.5699044831492641, 'gradnorm/compressor.layer2': 0.6331565769225402, 'gradnorm/compressor.layer3': 0.4191727454400087, 'gradnorm/compressor.layer4': 0.1670281778557986, 'gradnorm/compressor.layer5': 0.11676193291932042, 'gradnorm/compressor.layer6': 0.051434637068536196, 'gradnorm/compressor.layer7': 0.1033284726733351, 'gradnorm/decoder.layer0': 0.1885940600584653, 'gradnorm/decoder.layer1': 0.1859734600620021, 'gradnorm/decoder.layer2': 0.09369296584144399, 'gradnorm/max_param': 1.1024162769317627, 'epoch': 3.23}

 32%|███▏      | 6170/19090 [3:00:34<5:45:47,  1.61s/it]
 32%|███▏      | 6171/19090 [3:00:36<6:44:30,  1.88s/it]
 32%|███▏      | 6172/19090 [3:00:38<6:09:46,  1.72s/it]
 32%|███▏      | 6173/19090 [3:00:39<5:50:43,  1.63s/it]
 32%|███▏      | 6174/19090 [3:00:42<7:03:58,  1.97s/it]
 32%|███▏      | 6175/19090 [3:00:43<6:19:33,  1.76s/it]
 32%|███▏      | 6176/19090 [3:00:44<5:48:26,  1.62s/it]
 32%|███▏      | 6177/19090 [3:00:47<7:27:10,  2.08s/it]
 32%|███▏      | 6178/19090 [3:00:49<6:36:06,  1.84s/it]
 32%|███▏      | 6179/19090 [3:00:50<6:00:21,  1.67s/it]
 32%|███▏      | 6180/19090 [3:00:53<7:10:33,  2.00s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 2.0217976570129395, 'learning_rate': 4.2759099176037176e-05, 'recon_loss': 0.001361324347089976, 'recon_static': 0.0004045080524520017, 'recon_motion': 0.0009568162786308676, 'recon_norm': 1.2943151652812959, 'recon_norm_std': 0.07362730074673891, 'target_norm': 1.6210808098316192, 'recon_cos': 0.6853960126638412, 'gradnorm/compressor.query': 0.005870536514828489, 'gradnorm/compressor.layer0': 0.11821310349783624, 'gradnorm/compressor.layer1': 1.7908563407862592, 'gradnorm/compressor.layer2': 0.7346188972145659, 'gradnorm/compressor.layer3': 0.42955967128804134, 'gradnorm/compressor.layer4': 0.16639100998952064, 'gradnorm/compressor.layer5': 0.11973528511296848, 'gradnorm/compressor.layer6': 0.04354119990085701, 'gradnorm/compressor.layer7': 0.10882381366999541, 'gradnorm/decoder.layer0': 0.21286195159441382, 'gradnorm/decoder.layer1': 0.18334703913485917, 'gradnorm/decoder.layer2': 0.08727521548254139, 'gradnorm/max_param': 1.2599616050720215, 'epoch': 3.24}

 32%|███▏      | 6180/19090 [3:00:53<7:10:33,  2.00s/it]
 32%|███▏      | 6181/19090 [3:00:54<6:24:32,  1.79s/it]
 32%|███▏      | 6182/19090 [3:00:55<5:52:24,  1.64s/it]
 32%|███▏      | 6183/19090 [3:00:58<7:05:08,  1.98s/it]
 32%|███▏      | 6184/19090 [3:00:59<6:20:44,  1.77s/it]
 32%|███▏      | 6185/19090 [3:01:01<5:49:33,  1.63s/it]
 32%|███▏      | 6186/19090 [3:01:04<7:06:46,  1.98s/it]
 32%|███▏      | 6187/19090 [3:01:05<6:22:08,  1.78s/it]
 32%|███▏      | 6188/19090 [3:01:06<5:50:43,  1.63s/it]
 32%|███▏      | 6189/19090 [3:01:09<7:02:41,  1.97s/it]
 32%|███▏      | 6190/19090 [3:01:10<6:23:15,  1.78s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.0476880073547363, 'learning_rate': 4.272689499017194e-05, 'recon_loss': 0.0013411673426162452, 'recon_static': 0.00037319551265682096, 'recon_motion': 0.0009679718481493182, 'recon_norm': 1.2779284745454789, 'recon_norm_std': 0.0721500551328063, 'target_norm': 1.6233384370803834, 'recon_cos': 0.6907424747943878, 'gradnorm/compressor.query': 0.006383596971191464, 'gradnorm/compressor.layer0': 0.1316167068259822, 'gradnorm/compressor.layer1': 1.8004081157448013, 'gradnorm/compressor.layer2': 0.7594626639043456, 'gradnorm/compressor.layer3': 0.45614057953253834, 'gradnorm/compressor.layer4': 0.17411183660506663, 'gradnorm/compressor.layer5': 0.12397490700817101, 'gradnorm/compressor.layer6': 0.04715878543348101, 'gradnorm/compressor.layer7': 0.11348241997733532, 'gradnorm/decoder.layer0': 0.2134663819366126, 'gradnorm/decoder.layer1': 0.1865927382745053, 'gradnorm/decoder.layer2': 0.09041289241537634, 'gradnorm/max_param': 1.2663960456848145, 'epoch': 3.24}

 32%|███▏      | 6190/19090 [3:01:10<6:23:15,  1.78s/it]
 32%|███▏      | 6191/19090 [3:01:12<5:51:48,  1.64s/it]
 32%|███▏      | 6192/19090 [3:01:14<6:46:41,  1.89s/it]
 32%|███▏      | 6193/19090 [3:01:15<6:13:29,  1.74s/it]
 32%|███▏      | 6194/19090 [3:01:17<5:44:25,  1.60s/it]
 32%|███▏      | 6195/19090 [3:01:19<6:35:58,  1.84s/it]
 32%|███▏      | 6196/19090 [3:01:21<6:30:00,  1.81s/it]
 32%|███▏      | 6197/19090 [3:01:22<5:55:59,  1.66s/it]
 32%|███▏      | 6198/19090 [3:01:24<6:17:09,  1.76s/it]
 32%|███▏      | 6199/19090 [3:01:26<6:52:42,  1.92s/it]
 32%|███▏      | 6200/19090 [3:01:28<6:12:17,  1.73s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.144881010055542, 'learning_rate': 4.269463153423936e-05, 'recon_loss': 0.001315380708547309, 'recon_static': 0.0003669606034236494, 'recon_motion': 0.0009484201233135537, 'recon_norm': 1.2771610915660858, 'recon_norm_std': 0.0732956706546247, 'target_norm': 1.6176974922418594, 'recon_cos': 0.6949698626995087, 'gradnorm/compressor.query': 0.006171766097486448, 'gradnorm/compressor.layer0': 0.1258918048296277, 'gradnorm/compressor.layer1': 1.9161074555115365, 'gradnorm/compressor.layer2': 0.7404019208937777, 'gradnorm/compressor.layer3': 0.4650671926783801, 'gradnorm/compressor.layer4': 0.17991091480737523, 'gradnorm/compressor.layer5': 0.12371726804916694, 'gradnorm/compressor.layer6': 0.04549821140450334, 'gradnorm/compressor.layer7': 0.10786253970334769, 'gradnorm/decoder.layer0': 0.20253362216098825, 'gradnorm/decoder.layer1': 0.19461897744799722, 'gradnorm/decoder.layer2': 0.09166546302875775, 'gradnorm/max_param': 1.3482836484909058, 'epoch': 3.25}

 32%|███▏      | 6200/19090 [3:01:28<6:12:17,  1.73s/it]
 32%|███▏      | 6201/19090 [3:01:30<6:19:21,  1.77s/it]
 32%|███▏      | 6202/19090 [3:01:32<6:43:35,  1.88s/it]
 32%|███▏      | 6203/19090 [3:01:33<6:05:20,  1.70s/it]
 32%|███▏      | 6204/19090 [3:01:35<6:34:04,  1.83s/it]
 33%|███▎      | 6205/19090 [3:01:37<6:32:50,  1.83s/it]
 33%|███▎      | 6206/19090 [3:01:38<5:57:50,  1.67s/it]
 33%|███▎      | 6207/19090 [3:01:40<6:08:41,  1.72s/it]
 33%|███▎      | 6208/19090 [3:01:42<6:37:49,  1.85s/it]
 33%|███▎      | 6209/19090 [3:01:44<6:01:28,  1.68s/it]
 33%|███▎      | 6210/19090 [3:01:45<6:08:27,  1.72s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 1.9954321384429932, 'learning_rate': 4.2662308916112646e-05, 'recon_loss': 0.0013308950932696463, 'recon_static': 0.0003697335683682468, 'recon_motion': 0.0009611615314497612, 'recon_norm': 1.275250318646431, 'recon_norm_std': 0.07514263223856688, 'target_norm': 1.6222470611333848, 'recon_cos': 0.6930271178483963, 'gradnorm/compressor.query': 0.006340910176371528, 'gradnorm/compressor.layer0': 0.12049755488656272, 'gradnorm/compressor.layer1': 1.768672844380042, 'gradnorm/compressor.layer2': 0.7173795904081475, 'gradnorm/compressor.layer3': 0.42493929429425165, 'gradnorm/compressor.layer4': 0.17470059945393154, 'gradnorm/compressor.layer5': 0.12648306166760478, 'gradnorm/compressor.layer6': 0.04413273897028149, 'gradnorm/compressor.layer7': 0.10718647567103876, 'gradnorm/decoder.layer0': 0.203544888799524, 'gradnorm/decoder.layer1': 0.18651741577334388, 'gradnorm/decoder.layer2': 0.08720419709319485, 'gradnorm/max_param': 1.2443625926971436, 'epoch': 3.25}

 33%|███▎      | 6210/19090 [3:01:45<6:08:27,  1.72s/it]
 33%|███▎      | 6211/19090 [3:01:48<6:53:29,  1.93s/it]
 33%|███▎      | 6212/19090 [3:01:49<6:12:11,  1.73s/it]
 33%|███▎      | 6213/19090 [3:01:51<6:18:07,  1.76s/it]
 33%|███▎      | 6214/19090 [3:01:53<6:52:58,  1.92s/it]
 33%|███▎      | 6215/19090 [3:01:54<6:12:06,  1.73s/it]
 33%|███▎      | 6216/19090 [3:01:56<6:11:03,  1.73s/it]
 33%|███▎      | 6217/19090 [3:01:59<6:57:39,  1.95s/it]
 33%|███▎      | 6218/19090 [3:02:00<6:15:16,  1.75s/it]
 33%|███▎      | 6219/19090 [3:02:01<5:56:12,  1.66s/it]
 33%|███▎      | 6220/19090 [3:02:04<6:55:04,  1.94s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.061964511871338, 'learning_rate': 4.262992724386282e-05, 'recon_loss': 0.001349418074823916, 'recon_static': 0.00039478260005125775, 'recon_motion': 0.0009546355009661056, 'recon_norm': 1.2888179123401642, 'recon_norm_std': 0.07339058946818114, 'target_norm': 1.6221059709787369, 'recon_cos': 0.6888905331492424, 'gradnorm/compressor.query': 0.007346676628686922, 'gradnorm/compressor.layer0': 0.14023583259993852, 'gradnorm/compressor.layer1': 1.82246736348133, 'gradnorm/compressor.layer2': 0.748872585101079, 'gradnorm/compressor.layer3': 0.44566978376689864, 'gradnorm/compressor.layer4': 0.17751635289754378, 'gradnorm/compressor.layer5': 0.1393168186623916, 'gradnorm/compressor.layer6': 0.04654050981811136, 'gradnorm/compressor.layer7': 0.11070165548378964, 'gradnorm/decoder.layer0': 0.19944260679692405, 'gradnorm/decoder.layer1': 0.1963732429786379, 'gradnorm/decoder.layer2': 0.08647364029430188, 'gradnorm/max_param': 1.2816903591156006, 'epoch': 3.26}

 33%|███▎      | 6220/19090 [3:02:04<6:55:04,  1.94s/it]
 33%|███▎      | 6221/19090 [3:02:05<6:13:51,  1.74s/it]
 33%|███▎      | 6222/19090 [3:02:07<5:57:56,  1.67s/it]
 33%|███▎      | 6223/19090 [3:02:09<6:46:47,  1.90s/it]
 33%|███▎      | 6224/19090 [3:02:10<6:07:23,  1.71s/it]
 33%|███▎      | 6225/19090 [3:02:12<5:46:38,  1.62s/it]
 33%|███▎      | 6226/19090 [3:02:15<6:59:34,  1.96s/it]
 33%|███▎      | 6227/19090 [3:02:16<6:16:10,  1.75s/it]
 33%|███▎      | 6228/19090 [3:02:17<5:55:17,  1.66s/it]
 33%|███▎      | 6229/19090 [3:02:20<7:17:59,  2.04s/it]
 33%|███▎      | 6230/19090 [3:02:22<6:29:32,  1.82s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.096470832824707, 'learning_rate': 4.259748662575835e-05, 'recon_loss': 0.001347073510987684, 'recon_static': 0.00039674076178926044, 'recon_motion': 0.0009503327717538923, 'recon_norm': 1.2875984907150269, 'recon_norm_std': 0.07449942361563444, 'target_norm': 1.6200547069311142, 'recon_cos': 0.6882095992565155, 'gradnorm/compressor.query': 0.00798139738456243, 'gradnorm/compressor.layer0': 0.15024991078620079, 'gradnorm/compressor.layer1': 1.8546069261873879, 'gradnorm/compressor.layer2': 0.7423630356255678, 'gradnorm/compressor.layer3': 0.4590167419708797, 'gradnorm/compressor.layer4': 0.19072146437312093, 'gradnorm/compressor.layer5': 0.1289782250923565, 'gradnorm/compressor.layer6': 0.0493617722554668, 'gradnorm/compressor.layer7': 0.12291029131312264, 'gradnorm/decoder.layer0': 0.22251848509703395, 'gradnorm/decoder.layer1': 0.20693443841838324, 'gradnorm/decoder.layer2': 0.09108106037304173, 'gradnorm/max_param': 1.3044027090072632, 'epoch': 3.26}

 33%|███▎      | 6230/19090 [3:02:22<6:29:32,  1.82s/it]
 33%|███▎      | 6231/19090 [3:02:23<5:55:22,  1.66s/it]
 33%|███▎      | 6232/19090 [3:02:26<7:11:33,  2.01s/it]
 33%|███▎      | 6233/19090 [3:02:27<6:24:42,  1.80s/it]
 33%|███▎      | 6234/19090 [3:02:28<5:51:44,  1.64s/it]
 33%|███▎      | 6235/19090 [3:02:31<7:12:19,  2.02s/it]
 33%|███▎      | 6236/19090 [3:02:32<6:25:17,  1.80s/it]
 33%|███▎      | 6237/19090 [3:02:34<5:52:18,  1.64s/it]
 33%|███▎      | 6238/19090 [3:02:36<6:55:09,  1.94s/it]
 33%|███▎      | 6239/19090 [3:02:38<6:13:14,  1.74s/it]
 33%|███▎      | 6240/19090 [3:02:39<5:44:27,  1.61s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 2.099830150604248, 'learning_rate': 4.2564987170264825e-05, 'recon_loss': 0.001360350110917352, 'recon_static': 0.00039936821412993594, 'recon_motion': 0.0009609818836906925, 'recon_norm': 1.2890424281358719, 'recon_norm_std': 0.07201022021472454, 'target_norm': 1.6196520328521729, 'recon_cos': 0.6848338529467582, 'gradnorm/compressor.query': 0.006962550399561636, 'gradnorm/compressor.layer0': 0.13803523291441505, 'gradnorm/compressor.layer1': 1.875506078451421, 'gradnorm/compressor.layer2': 0.7240663946357335, 'gradnorm/compressor.layer3': 0.44748465826066275, 'gradnorm/compressor.layer4': 0.1737301689775785, 'gradnorm/compressor.layer5': 0.13041995415210503, 'gradnorm/compressor.layer6': 0.04786504827930108, 'gradnorm/compressor.layer7': 0.11397949784834523, 'gradnorm/decoder.layer0': 0.20851090532709227, 'gradnorm/decoder.layer1': 0.18426880606971427, 'gradnorm/decoder.layer2': 0.09077649443837398, 'gradnorm/max_param': 1.319146990776062, 'epoch': 3.27}

 33%|███▎      | 6240/19090 [3:02:39<5:44:27,  1.61s/it]
 33%|███▎      | 6241/19090 [3:02:42<6:47:31,  1.90s/it]
 33%|███▎      | 6242/19090 [3:02:43<6:08:27,  1.72s/it]
 33%|███▎      | 6243/19090 [3:02:44<5:40:33,  1.59s/it]
 33%|███▎      | 6244/19090 [3:02:47<6:47:44,  1.90s/it]
 33%|███▎      | 6245/19090 [3:02:48<6:08:06,  1.72s/it]
 33%|███▎      | 6246/19090 [3:02:49<5:40:11,  1.59s/it]
 33%|███▎      | 6247/19090 [3:02:52<6:56:20,  1.95s/it]
 33%|███▎      | 6248/19090 [3:02:53<6:14:15,  1.75s/it]
 33%|███▎      | 6249/19090 [3:02:55<5:44:26,  1.61s/it]
 33%|███▎      | 6250/19090 [3:02:57<6:55:01,  1.94s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 2.07572603225708, 'learning_rate': 4.25324289860445e-05, 'recon_loss': 0.0013666053622728215, 'recon_static': 0.0004135436094657052, 'recon_motion': 0.0009530617855489254, 'recon_norm': 1.2953415781259536, 'recon_norm_std': 0.07255111690610647, 'target_norm': 1.6202425152063369, 'recon_cos': 0.683660514652729, 'gradnorm/compressor.query': 0.007161492425302797, 'gradnorm/compressor.layer0': 0.13891023647307857, 'gradnorm/compressor.layer1': 1.8307605883604554, 'gradnorm/compressor.layer2': 0.7377432826016546, 'gradnorm/compressor.layer3': 0.46982841418463966, 'gradnorm/compressor.layer4': 0.18606138670219538, 'gradnorm/compressor.layer5': 0.13526993529159093, 'gradnorm/compressor.layer6': 0.05039543759112238, 'gradnorm/compressor.layer7': 0.11490446483956868, 'gradnorm/decoder.layer0': 0.2321361500496877, 'gradnorm/decoder.layer1': 0.20180571800702196, 'gradnorm/decoder.layer2': 0.0965964990161108, 'gradnorm/max_param': 1.2874445915222168, 'epoch': 3.27}

 33%|███▎      | 6250/19090 [3:02:57<6:55:01,  1.94s/it]
 33%|███▎      | 6251/19090 [3:02:59<6:17:11,  1.76s/it]
 33%|███▎      | 6252/19090 [3:03:00<5:47:06,  1.62s/it]
 33%|███▎      | 6253/19090 [3:03:03<7:00:27,  1.97s/it]
 33%|███▎      | 6254/19090 [3:03:04<6:16:51,  1.76s/it]
 33%|███▎      | 6255/19090 [3:03:05<5:46:28,  1.62s/it]
 33%|███▎      | 6256/19090 [3:03:08<6:51:00,  1.92s/it]
 33%|███▎      | 6257/19090 [3:03:09<6:10:08,  1.73s/it]
 33%|███▎      | 6258/19090 [3:03:11<5:41:21,  1.60s/it]
 33%|███▎      | 6259/19090 [3:03:14<7:13:41,  2.03s/it]
 33%|███▎      | 6260/19090 [3:03:15<6:26:14,  1.81s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 2.0904414653778076, 'learning_rate': 4.249981218195604e-05, 'recon_loss': 0.0013658317038789392, 'recon_static': 0.0004132963658776134, 'recon_motion': 0.000952535327814985, 'recon_norm': 1.2938743561506272, 'recon_norm_std': 0.07188571821898222, 'target_norm': 1.618183746933937, 'recon_cos': 0.6829128384590148, 'gradnorm/compressor.query': 0.006748584641439467, 'gradnorm/compressor.layer0': 0.1356486883549457, 'gradnorm/compressor.layer1': 1.8623082081401894, 'gradnorm/compressor.layer2': 0.7237591137748708, 'gradnorm/compressor.layer3': 0.43459268576646065, 'gradnorm/compressor.layer4': 0.1926991819236417, 'gradnorm/compressor.layer5': 0.14736040743281195, 'gradnorm/compressor.layer6': 0.05014705555487464, 'gradnorm/compressor.layer7': 0.11668825696593417, 'gradnorm/decoder.layer0': 0.22133750385323844, 'gradnorm/decoder.layer1': 0.19345329392745222, 'gradnorm/decoder.layer2': 0.09600583683836658, 'gradnorm/max_param': 1.3097528219223022, 'epoch': 3.28}

 33%|███▎      | 6260/19090 [3:03:15<6:26:14,  1.81s/it]
 33%|███▎      | 6261/19090 [3:03:16<5:53:26,  1.65s/it]
 33%|███▎      | 6262/19090 [3:03:19<6:40:41,  1.87s/it]
 33%|███▎      | 6263/19090 [3:03:20<6:02:56,  1.70s/it]
 33%|███▎      | 6264/19090 [3:03:21<5:39:18,  1.59s/it]
 33%|███▎      | 6265/19090 [3:03:24<6:38:21,  1.86s/it]
 33%|███▎      | 6266/19090 [3:03:25<6:01:03,  1.69s/it]
 33%|███▎      | 6267/19090 [3:03:27<6:15:50,  1.76s/it]
 33%|███▎      | 6268/19090 [3:03:29<7:02:55,  1.98s/it]
 33%|███▎      | 6269/19090 [3:03:31<6:18:12,  1.77s/it]
 33%|███▎      | 6270/19090 [3:03:32<5:53:42,  1.66s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 2.1397337913513184, 'learning_rate': 4.246713686705408e-05, 'recon_loss': 0.001364330534124747, 'recon_static': 0.00040471160973538646, 'recon_motion': 0.0009596188683644868, 'recon_norm': 1.289051678776741, 'recon_norm_std': 0.07204351555556059, 'target_norm': 1.6216127157211304, 'recon_cos': 0.6846352770924569, 'gradnorm/compressor.query': 0.006635327891145681, 'gradnorm/compressor.layer0': 0.1351709145655259, 'gradnorm/compressor.layer1': 1.9085433797166182, 'gradnorm/compressor.layer2': 0.7433835080309218, 'gradnorm/compressor.layer3': 0.45880990466199784, 'gradnorm/compressor.layer4': 0.19148328748913238, 'gradnorm/compressor.layer5': 0.12821456988273183, 'gradnorm/compressor.layer6': 0.04776452625113418, 'gradnorm/compressor.layer7': 0.1121542254395677, 'gradnorm/decoder.layer0': 0.205303776438123, 'gradnorm/decoder.layer1': 0.19062673693006302, 'gradnorm/decoder.layer2': 0.08960659405251865, 'gradnorm/max_param': 1.3426244258880615, 'epoch': 3.28}

 33%|███▎      | 6270/19090 [3:03:32<5:53:42,  1.66s/it]
 33%|███▎      | 6271/19090 [3:03:35<7:03:12,  1.98s/it]
 33%|███▎      | 6272/19090 [3:03:36<6:18:36,  1.77s/it]
 33%|███▎      | 6273/19090 [3:03:37<5:47:55,  1.63s/it]
 33%|███▎      | 6274/19090 [3:03:40<6:52:41,  1.93s/it]
 33%|███▎      | 6275/19090 [3:03:41<6:11:06,  1.74s/it]
 33%|███▎      | 6276/19090 [3:03:43<5:42:09,  1.60s/it]
 33%|███▎      | 6277/19090 [3:03:46<7:33:15,  2.12s/it]
 33%|███▎      | 6278/19090 [3:03:47<6:39:48,  1.87s/it]
 33%|███▎      | 6279/19090 [3:03:48<6:02:13,  1.70s/it]
 33%|███▎      | 6280/19090 [3:03:51<6:40:01,  1.87s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.345428228378296, 'learning_rate': 4.2434403150588895e-05, 'recon_loss': 0.0013359520089579746, 'recon_static': 0.0003815099174971692, 'recon_motion': 0.0009544421001919546, 'recon_norm': 1.2791601508855819, 'recon_norm_std': 0.0719780147075653, 'target_norm': 1.619482097029686, 'recon_cos': 0.6908237710595131, 'gradnorm/compressor.query': 0.006791276518690007, 'gradnorm/compressor.layer0': 0.13760865643338469, 'gradnorm/compressor.layer1': 2.1043963417124387, 'gradnorm/compressor.layer2': 0.7991094863199638, 'gradnorm/compressor.layer3': 0.4965001941864925, 'gradnorm/compressor.layer4': 0.1944703255659591, 'gradnorm/compressor.layer5': 0.1377773311184856, 'gradnorm/compressor.layer6': 0.048766636279387675, 'gradnorm/compressor.layer7': 0.11478644939717804, 'gradnorm/decoder.layer0': 0.21488595894689283, 'gradnorm/decoder.layer1': 0.20391855750813617, 'gradnorm/decoder.layer2': 0.09145962367291682, 'gradnorm/max_param': 1.4809155464172363, 'epoch': 3.29}

 33%|███▎      | 6280/19090 [3:03:51<6:40:01,  1.87s/it]
 33%|███▎      | 6281/19090 [3:03:52<6:03:45,  1.70s/it]
 33%|███▎      | 6282/19090 [3:03:53<5:40:55,  1.60s/it]
 33%|███▎      | 6283/19090 [3:03:56<6:36:08,  1.86s/it]
 33%|███▎      | 6284/19090 [3:03:57<5:59:09,  1.68s/it]
 33%|███▎      | 6285/19090 [3:03:59<6:17:41,  1.77s/it]
 33%|███▎      | 6286/19090 [3:04:01<6:41:37,  1.88s/it]
 33%|███▎      | 6287/19090 [3:04:03<6:03:11,  1.70s/it]
 33%|███▎      | 6288/19090 [3:04:05<6:23:08,  1.80s/it]
 33%|███▎      | 6289/19090 [3:04:06<6:14:37,  1.76s/it]
 33%|███▎      | 6290/19090 [3:04:08<5:44:36,  1.62s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.2397804260253906, 'learning_rate': 4.2401611142006016e-05, 'recon_loss': 0.0013049022003542633, 'recon_static': 0.00035811345078400335, 'recon_motion': 0.0009467887270147912, 'recon_norm': 1.2702561259269713, 'recon_norm_std': 0.07163425143808126, 'target_norm': 1.6146544516086578, 'recon_cos': 0.6963855311274528, 'gradnorm/compressor.query': 0.006151176570879761, 'gradnorm/compressor.layer0': 0.12360812659387233, 'gradnorm/compressor.layer1': 2.000772327074012, 'gradnorm/compressor.layer2': 0.7785976016481208, 'gradnorm/compressor.layer3': 0.4768628909072248, 'gradnorm/compressor.layer4': 0.1927155265668427, 'gradnorm/compressor.layer5': 0.1355791553679297, 'gradnorm/compressor.layer6': 0.042335879416548226, 'gradnorm/compressor.layer7': 0.11956981445104091, 'gradnorm/decoder.layer0': 0.2051615681507338, 'gradnorm/decoder.layer1': 0.20773577923136918, 'gradnorm/decoder.layer2': 0.087981389121902, 'gradnorm/max_param': 1.4083547592163086, 'epoch': 3.29}

 33%|███▎      | 6290/19090 [3:04:08<5:44:36,  1.62s/it]
 33%|███▎      | 6291/19090 [3:04:10<6:15:53,  1.76s/it]
 33%|███▎      | 6292/19090 [3:04:12<6:31:24,  1.84s/it]
 33%|███▎      | 6293/19090 [3:04:13<5:56:01,  1.67s/it]
 33%|███▎      | 6294/19090 [3:04:15<6:14:56,  1.76s/it]
 33%|███▎      | 6295/19090 [3:04:17<6:53:11,  1.94s/it]
 33%|███▎      | 6296/19090 [3:04:19<6:11:54,  1.74s/it]
 33%|███▎      | 6297/19090 [3:04:20<6:08:43,  1.73s/it]
 33%|███▎      | 6298/19090 [3:04:23<7:11:19,  2.02s/it]
 33%|███▎      | 6299/19090 [3:04:24<6:23:59,  1.80s/it]
 33%|███▎      | 6300/19090 [3:04:26<6:14:56,  1.76s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.0896263122558594, 'learning_rate': 4.2368760950945896e-05, 'recon_loss': 0.001308350870385766, 'recon_static': 0.0003575022332370281, 'recon_motion': 0.0009508486444246955, 'recon_norm': 1.268956834077835, 'recon_norm_std': 0.0720552546903491, 'target_norm': 1.618471196293831, 'recon_cos': 0.6969878107309342, 'gradnorm/compressor.query': 0.006165558371488847, 'gradnorm/compressor.layer0': 0.12187287576364733, 'gradnorm/compressor.layer1': 1.8600151178299353, 'gradnorm/compressor.layer2': 0.7227024785282362, 'gradnorm/compressor.layer3': 0.4549495040466996, 'gradnorm/compressor.layer4': 0.18462113785691228, 'gradnorm/compressor.layer5': 0.14322766534423048, 'gradnorm/compressor.layer6': 0.04690702407393259, 'gradnorm/compressor.layer7': 0.10629329412911773, 'gradnorm/decoder.layer0': 0.21724689446798745, 'gradnorm/decoder.layer1': 0.19760906425229038, 'gradnorm/decoder.layer2': 0.09148170767820335, 'gradnorm/max_param': 1.3087397813796997, 'epoch': 3.3}

 33%|███▎      | 6300/19090 [3:04:26<6:14:56,  1.76s/it]
 33%|███▎      | 6301/19090 [3:04:28<6:49:17,  1.92s/it]
 33%|███▎      | 6302/19090 [3:04:29<6:08:40,  1.73s/it]
 33%|███▎      | 6303/19090 [3:04:31<5:47:14,  1.63s/it]
 33%|███▎      | 6304/19090 [3:04:33<6:34:03,  1.85s/it]
 33%|███▎      | 6305/19090 [3:04:35<5:57:45,  1.68s/it]
 33%|███▎      | 6306/19090 [3:04:36<5:53:56,  1.66s/it]
 33%|███▎      | 6307/19090 [3:04:38<6:38:03,  1.87s/it]
 33%|███▎      | 6308/19090 [3:04:40<6:00:40,  1.69s/it]
 33%|███▎      | 6309/19090 [3:04:41<5:39:13,  1.59s/it]
 33%|███▎      | 6310/19090 [3:04:44<7:05:35,  2.00s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.422088623046875, 'learning_rate': 4.233585268724349e-05, 'recon_loss': 0.0013320999249117448, 'recon_static': 0.00036982050660299136, 'recon_motion': 0.0009622793964808807, 'recon_norm': 1.2737123280763627, 'recon_norm_std': 0.07078576963394881, 'target_norm': 1.6215679049491882, 'recon_cos': 0.6924017488956451, 'gradnorm/compressor.query': 0.007520406458441295, 'gradnorm/compressor.layer0': 0.1460623447301494, 'gradnorm/compressor.layer1': 2.1657602419328668, 'gradnorm/compressor.layer2': 0.847026327203668, 'gradnorm/compressor.layer3': 0.5106238671223693, 'gradnorm/compressor.layer4': 0.19141627304926276, 'gradnorm/compressor.layer5': 0.13763709224013115, 'gradnorm/compressor.layer6': 0.044877945629453644, 'gradnorm/compressor.layer7': 0.12317415118676021, 'gradnorm/decoder.layer0': 0.22530705276073656, 'gradnorm/decoder.layer1': 0.21181825150723266, 'gradnorm/decoder.layer2': 0.0895165206283747, 'gradnorm/max_param': 1.5247267484664917, 'epoch': 3.31}

 33%|███▎      | 6310/19090 [3:04:44<7:05:35,  2.00s/it]
 33%|███▎      | 6311/19090 [3:04:45<6:21:31,  1.79s/it]
 33%|███▎      | 6312/19090 [3:04:47<5:49:51,  1.64s/it]
 33%|███▎      | 6313/19090 [3:04:49<6:52:43,  1.94s/it]
 33%|███▎      | 6314/19090 [3:04:51<6:11:26,  1.74s/it]
 33%|███▎      | 6315/19090 [3:04:52<5:42:33,  1.61s/it]
 33%|███▎      | 6316/19090 [3:04:55<6:54:41,  1.95s/it]
 33%|███▎      | 6317/19090 [3:04:56<6:12:22,  1.75s/it]
 33%|███▎      | 6318/19090 [3:04:57<5:45:53,  1.62s/it]
 33%|███▎      | 6319/19090 [3:05:00<6:34:39,  1.85s/it]
 33%|███▎      | 6320/19090 [3:05:01<5:58:27,  1.68s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.529942274093628, 'learning_rate': 4.230288646092795e-05, 'recon_loss': 0.0013148183119483293, 'recon_static': 0.00036374770716065543, 'recon_motion': 0.0009510706222499721, 'recon_norm': 1.2690214961767197, 'recon_norm_std': 0.07018326940014959, 'target_norm': 1.6165894895792008, 'recon_cos': 0.6945421069860458, 'gradnorm/compressor.query': 0.006854639483762028, 'gradnorm/compressor.layer0': 0.14054339080656417, 'gradnorm/compressor.layer1': 2.2953470558593185, 'gradnorm/compressor.layer2': 0.8079802171154935, 'gradnorm/compressor.layer3': 0.511556701627295, 'gradnorm/compressor.layer4': 0.2020639528652387, 'gradnorm/compressor.layer5': 0.14917854371572573, 'gradnorm/compressor.layer6': 0.0471707984912675, 'gradnorm/compressor.layer7': 0.1223245757400929, 'gradnorm/decoder.layer0': 0.2543449719675371, 'gradnorm/decoder.layer1': 0.21106295916614962, 'gradnorm/decoder.layer2': 0.0903133115187487, 'gradnorm/max_param': 1.6157352924346924, 'epoch': 3.31}

 33%|███▎      | 6320/19090 [3:05:01<5:58:27,  1.68s/it]
 33%|███▎      | 6321/19090 [3:05:03<5:54:22,  1.67s/it]
 33%|███▎      | 6322/19090 [3:05:05<6:58:02,  1.96s/it]
 33%|███▎      | 6323/19090 [3:05:06<6:14:32,  1.76s/it]
 33%|███▎      | 6324/19090 [3:05:08<5:44:12,  1.62s/it]
 33%|███▎      | 6325/19090 [3:05:11<7:07:12,  2.01s/it]
 33%|███▎      | 6326/19090 [3:05:12<6:21:11,  1.79s/it]
 33%|███▎      | 6327/19090 [3:05:13<5:48:53,  1.64s/it]
 33%|███▎      | 6328/19090 [3:05:16<7:00:15,  1.98s/it]
 33%|███▎      | 6329/19090 [3:05:17<6:15:59,  1.77s/it]
 33%|███▎      | 6330/19090 [3:05:19<5:49:01,  1.64s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.409842014312744, 'learning_rate': 4.2269862382222224e-05, 'recon_loss': 0.0013145005737897008, 'recon_static': 0.00035670395809574986, 'recon_motion': 0.0009577966018696315, 'recon_norm': 1.2669584929943085, 'recon_norm_std': 0.0707629220560193, 'target_norm': 1.6194132268428802, 'recon_cos': 0.6957761585712433, 'gradnorm/compressor.query': 0.0073924608277189064, 'gradnorm/compressor.layer0': 0.13493806062681732, 'gradnorm/compressor.layer1': 2.1739761836792484, 'gradnorm/compressor.layer2': 0.791355766350674, 'gradnorm/compressor.layer3': 0.5011059926615802, 'gradnorm/compressor.layer4': 0.20839728128958016, 'gradnorm/compressor.layer5': 0.15213022184225508, 'gradnorm/compressor.layer6': 0.049045939999248264, 'gradnorm/compressor.layer7': 0.12172950089320249, 'gradnorm/decoder.layer0': 0.2253276039357169, 'gradnorm/decoder.layer1': 0.20484610556929217, 'gradnorm/decoder.layer2': 0.09497285560973273, 'gradnorm/max_param': 1.530320644378662, 'epoch': 3.32}

 33%|███▎      | 6330/19090 [3:05:19<5:49:01,  1.64s/it]
 33%|███▎      | 6331/19090 [3:05:21<7:02:08,  1.99s/it]
 33%|███▎      | 6332/19090 [3:05:23<6:17:09,  1.77s/it]
 33%|███▎      | 6333/19090 [3:05:24<5:46:15,  1.63s/it]
 33%|███▎      | 6334/19090 [3:05:28<8:03:14,  2.27s/it]
 33%|███▎      | 6335/19090 [3:05:29<7:00:09,  1.98s/it]
 33%|███▎      | 6336/19090 [3:05:30<6:15:43,  1.77s/it]
 33%|███▎      | 6337/19090 [3:05:32<6:13:02,  1.76s/it]
 33%|███▎      | 6338/19090 [3:05:33<5:42:55,  1.61s/it]
 33%|███▎      | 6339/19090 [3:05:35<5:33:51,  1.57s/it]
 33%|███▎      | 6340/19090 [3:05:38<6:46:57,  1.92s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.416905403137207, 'learning_rate': 4.2236780561542655e-05, 'recon_loss': 0.0013102880242513494, 'recon_static': 0.000361110099038342, 'recon_motion': 0.0009491779448580929, 'recon_norm': 1.268933293223381, 'recon_norm_std': 0.07304382640868426, 'target_norm': 1.6202019065618516, 'recon_cos': 0.6974748864769935, 'gradnorm/compressor.query': 0.006706986640461479, 'gradnorm/compressor.layer0': 0.1324954244997739, 'gradnorm/compressor.layer1': 2.1888689389855354, 'gradnorm/compressor.layer2': 0.7725650606257234, 'gradnorm/compressor.layer3': 0.5018391164300536, 'gradnorm/compressor.layer4': 0.20229512112676967, 'gradnorm/compressor.layer5': 0.15515365418820243, 'gradnorm/compressor.layer6': 0.04841301418343371, 'gradnorm/compressor.layer7': 0.12385250687857251, 'gradnorm/decoder.layer0': 0.22221678312834275, 'gradnorm/decoder.layer1': 0.20874830977675016, 'gradnorm/decoder.layer2': 0.09137493925901448, 'gradnorm/max_param': 1.5409587621688843, 'epoch': 3.32}

 33%|███▎      | 6340/19090 [3:05:38<6:46:57,  1.92s/it]
 33%|███▎      | 6341/19090 [3:05:39<6:06:30,  1.72s/it]
 33%|███▎      | 6342/19090 [3:05:40<5:38:00,  1.59s/it]
 33%|███▎      | 6343/19090 [3:05:43<6:58:14,  1.97s/it]
 33%|███▎      | 6344/19090 [3:05:44<6:14:29,  1.76s/it]
 33%|███▎      | 6345/19090 [3:05:46<5:44:26,  1.62s/it]
 33%|███▎      | 6346/19090 [3:05:48<6:47:38,  1.92s/it]
 33%|███▎      | 6347/19090 [3:05:49<6:07:33,  1.73s/it]
 33%|███▎      | 6348/19090 [3:05:51<5:39:27,  1.60s/it]
 33%|███▎      | 6349/19090 [3:05:54<7:17:48,  2.06s/it]
 33%|███▎      | 6350/19090 [3:05:55<6:28:18,  1.83s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.4975807666778564, 'learning_rate': 4.2203641109498696e-05, 'recon_loss': 0.0013039520941674708, 'recon_static': 0.0003489056056423578, 'recon_motion': 0.0009550464907079004, 'recon_norm': 1.2623411208391189, 'recon_norm_std': 0.07259506024420262, 'target_norm': 1.6206191301345825, 'recon_cos': 0.699110995233059, 'gradnorm/compressor.query': 0.006625248988347597, 'gradnorm/compressor.layer0': 0.1332253882185766, 'gradnorm/compressor.layer1': 2.269078326047285, 'gradnorm/compressor.layer2': 0.7856571630359787, 'gradnorm/compressor.layer3': 0.5097645142057162, 'gradnorm/compressor.layer4': 0.20992489882500015, 'gradnorm/compressor.layer5': 0.15463273398656052, 'gradnorm/compressor.layer6': 0.045956308830961044, 'gradnorm/compressor.layer7': 0.12791870885434475, 'gradnorm/decoder.layer0': 0.23595330962208033, 'gradnorm/decoder.layer1': 0.20969724730564085, 'gradnorm/decoder.layer2': 0.09045318005873651, 'gradnorm/max_param': 1.5976883172988892, 'epoch': 3.33}

 33%|███▎      | 6350/19090 [3:05:55<6:28:18,  1.83s/it]
 33%|███▎      | 6351/19090 [3:05:56<5:53:46,  1.67s/it]
 33%|███▎      | 6352/19090 [3:05:59<6:59:42,  1.98s/it]
 33%|███▎      | 6353/19090 [3:06:00<6:15:41,  1.77s/it]
 33%|███▎      | 6354/19090 [3:06:02<5:45:03,  1.63s/it]
 33%|███▎      | 6355/19090 [3:06:05<7:03:51,  2.00s/it]
 33%|███▎      | 6356/19090 [3:06:06<6:18:18,  1.78s/it]
 33%|███▎      | 6357/19090 [3:06:07<5:46:51,  1.63s/it]
 33%|███▎      | 6358/19090 [3:06:10<6:45:30,  1.91s/it]
 33%|███▎      | 6359/19090 [3:06:11<6:05:44,  1.72s/it]
 33%|███▎      | 6360/19090 [3:06:12<5:38:37,  1.60s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.680495023727417, 'learning_rate': 4.217044413689246e-05, 'recon_loss': 0.0012906986841699109, 'recon_static': 0.0003345588076626882, 'recon_motion': 0.0009561398852383718, 'recon_norm': 1.2555949956178665, 'recon_norm_std': 0.07088336329907179, 'target_norm': 1.6197516471147537, 'recon_cos': 0.7018708184361457, 'gradnorm/compressor.query': 0.006978607845004008, 'gradnorm/compressor.layer0': 0.1370656245276006, 'gradnorm/compressor.layer1': 2.440932961936974, 'gradnorm/compressor.layer2': 0.8520147540064178, 'gradnorm/compressor.layer3': 0.5302953557495378, 'gradnorm/compressor.layer4': 0.21121775588439257, 'gradnorm/compressor.layer5': 0.15654352750815312, 'gradnorm/compressor.layer6': 0.04516010245716015, 'gradnorm/compressor.layer7': 0.12297193576413497, 'gradnorm/decoder.layer0': 0.24416485475468933, 'gradnorm/decoder.layer1': 0.21526032722324223, 'gradnorm/decoder.layer2': 0.09320839261072435, 'gradnorm/max_param': 1.719224214553833, 'epoch': 3.33}

 33%|███▎      | 6360/19090 [3:06:12<5:38:37,  1.60s/it]
 33%|███▎      | 6361/19090 [3:06:15<7:08:28,  2.02s/it]
 33%|███▎      | 6362/19090 [3:06:17<6:21:33,  1.80s/it]
 33%|███▎      | 6363/19090 [3:06:18<5:49:07,  1.65s/it]
 33%|███▎      | 6364/19090 [3:06:20<6:43:03,  1.90s/it]
 33%|███▎      | 6365/19090 [3:06:22<6:04:03,  1.72s/it]
 33%|███▎      | 6366/19090 [3:06:23<5:38:15,  1.60s/it]
 33%|███▎      | 6367/19090 [3:06:26<6:43:45,  1.90s/it]
 33%|███▎      | 6368/19090 [3:06:27<6:04:02,  1.72s/it]
 33%|███▎      | 6369/19090 [3:06:28<5:36:34,  1.59s/it]
 33%|███▎      | 6370/19090 [3:06:31<6:53:20,  1.95s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.5724244117736816, 'learning_rate': 4.2137189754718374e-05, 'recon_loss': 0.0012815879483241588, 'recon_static': 0.00033117196362582036, 'recon_motion': 0.0009504160057986155, 'recon_norm': 1.2527193605899811, 'recon_norm_std': 0.07353286873549222, 'target_norm': 1.6188844054937364, 'recon_cos': 0.703849746286869, 'gradnorm/compressor.query': 0.00613240764913735, 'gradnorm/compressor.layer0': 0.13095649986529087, 'gradnorm/compressor.layer1': 2.314577561264555, 'gradnorm/compressor.layer2': 0.8597673040890969, 'gradnorm/compressor.layer3': 0.5401900013251467, 'gradnorm/compressor.layer4': 0.21941841558532588, 'gradnorm/compressor.layer5': 0.15351706229937237, 'gradnorm/compressor.layer6': 0.044647443644324755, 'gradnorm/compressor.layer7': 0.1191224686280113, 'gradnorm/decoder.layer0': 0.25381080594186706, 'gradnorm/decoder.layer1': 0.22359397084447163, 'gradnorm/decoder.layer2': 0.09795853560037525, 'gradnorm/max_param': 1.6298829317092896, 'epoch': 3.34}

 33%|███▎      | 6370/19090 [3:06:31<6:53:20,  1.95s/it]
 33%|███▎      | 6371/19090 [3:06:32<6:11:29,  1.75s/it]
 33%|███▎      | 6372/19090 [3:06:34<5:41:57,  1.61s/it]
 33%|███▎      | 6373/19090 [3:06:36<7:03:47,  2.00s/it]
 33%|███▎      | 6374/19090 [3:06:38<6:18:16,  1.78s/it]
 33%|███▎      | 6375/19090 [3:06:39<5:46:46,  1.64s/it]
 33%|███▎      | 6376/19090 [3:06:42<6:48:21,  1.93s/it]
 33%|███▎      | 6377/19090 [3:06:43<6:07:34,  1.73s/it]
 33%|███▎      | 6378/19090 [3:06:44<5:39:16,  1.60s/it]
 33%|███▎      | 6379/19090 [3:06:47<6:45:02,  1.91s/it]
 33%|███▎      | 6380/19090 [3:06:48<6:05:45,  1.73s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.601036310195923, 'learning_rate': 4.210387807416285e-05, 'recon_loss': 0.001278461652691476, 'recon_static': 0.00032941234967438505, 'recon_motion': 0.000949049333576113, 'recon_norm': 1.2538520842790604, 'recon_norm_std': 0.0711309409700334, 'target_norm': 1.6189518928527833, 'recon_cos': 0.7048500210046769, 'gradnorm/compressor.query': 0.006076468501664578, 'gradnorm/compressor.layer0': 0.12857998393039985, 'gradnorm/compressor.layer1': 2.348635959424867, 'gradnorm/compressor.layer2': 0.8557341351232322, 'gradnorm/compressor.layer3': 0.5392979646660955, 'gradnorm/compressor.layer4': 0.2161784501724977, 'gradnorm/compressor.layer5': 0.15298783644971078, 'gradnorm/compressor.layer6': 0.04798252324239446, 'gradnorm/compressor.layer7': 0.11958648162930334, 'gradnorm/decoder.layer0': 0.24738493025328184, 'gradnorm/decoder.layer1': 0.22821541979983612, 'gradnorm/decoder.layer2': 0.0978902730684279, 'gradnorm/max_param': 1.6539454460144043, 'epoch': 3.34}

 33%|███▎      | 6380/19090 [3:06:48<6:05:45,  1.73s/it]
 33%|███▎      | 6381/19090 [3:06:49<5:37:47,  1.59s/it]
 33%|███▎      | 6382/19090 [3:06:52<7:08:04,  2.02s/it]
 33%|███▎      | 6383/19090 [3:06:54<6:21:43,  1.80s/it]
 33%|███▎      | 6384/19090 [3:06:55<5:48:49,  1.65s/it]
 33%|███▎      | 6385/19090 [3:06:58<6:56:26,  1.97s/it]
 33%|███▎      | 6386/19090 [3:06:59<6:13:12,  1.76s/it]
 33%|███▎      | 6387/19090 [3:07:01<6:40:56,  1.89s/it]
 33%|███▎      | 6388/19090 [3:07:03<7:03:59,  2.00s/it]
 33%|███▎      | 6389/19090 [3:07:05<6:18:44,  1.79s/it]
 33%|███▎      | 6390/19090 [3:07:06<5:46:41,  1.64s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.541766405105591, 'learning_rate': 4.2070509206603846e-05, 'recon_loss': 0.0012751272151945159, 'recon_static': 0.00032284534245263785, 'recon_motion': 0.0009522819396806881, 'recon_norm': 1.248786461353302, 'recon_norm_std': 0.07245303336530924, 'target_norm': 1.6182674586772918, 'recon_cos': 0.7052688524127007, 'gradnorm/compressor.query': 0.006961299928882907, 'gradnorm/compressor.layer0': 0.1213474385056813, 'gradnorm/compressor.layer1': 2.301355468388646, 'gradnorm/compressor.layer2': 0.8253953318861111, 'gradnorm/compressor.layer3': 0.5236012216989414, 'gradnorm/compressor.layer4': 0.217759127638538, 'gradnorm/compressor.layer5': 0.15200514944743204, 'gradnorm/compressor.layer6': 0.04518812514765681, 'gradnorm/compressor.layer7': 0.117535190569212, 'gradnorm/decoder.layer0': 0.23452565596311575, 'gradnorm/decoder.layer1': 0.2094861984321526, 'gradnorm/decoder.layer2': 0.09414245002212208, 'gradnorm/max_param': 1.6208906173706055, 'epoch': 3.35}

 33%|███▎      | 6390/19090 [3:07:06<5:46:41,  1.64s/it]
 33%|███▎      | 6391/19090 [3:07:09<7:19:36,  2.08s/it]
 33%|███▎      | 6392/19090 [3:07:10<6:29:53,  1.84s/it]
 33%|███▎      | 6393/19090 [3:07:12<5:54:59,  1.68s/it]
 33%|███▎      | 6394/19090 [3:07:14<6:57:03,  1.97s/it]
 33%|███▎      | 6395/19090 [3:07:16<6:14:16,  1.77s/it]
 34%|███▎      | 6396/19090 [3:07:17<5:44:01,  1.63s/it]
 34%|███▎      | 6397/19090 [3:07:19<6:38:40,  1.88s/it]
 34%|███▎      | 6398/19090 [3:07:21<6:00:46,  1.71s/it]
 34%|███▎      | 6399/19090 [3:07:22<5:33:55,  1.58s/it]
 34%|███▎      | 6400/19090 [3:07:25<6:51:19,  1.94s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.7550501823425293, 'learning_rate': 4.2037083263610536e-05, 'recon_loss': 0.0012813169742003083, 'recon_static': 0.0003263949743995909, 'recon_motion': 0.0009549219801556319, 'recon_norm': 1.2493068099021911, 'recon_norm_std': 0.0703191900625825, 'target_norm': 1.6206061512231826, 'recon_cos': 0.704708245396614, 'gradnorm/compressor.query': 0.0069675155772711645, 'gradnorm/compressor.layer0': 0.13851840297755508, 'gradnorm/compressor.layer1': 2.4944196409065147, 'gradnorm/compressor.layer2': 0.9197092640745224, 'gradnorm/compressor.layer3': 0.5421389269439528, 'gradnorm/compressor.layer4': 0.21813058861019338, 'gradnorm/compressor.layer5': 0.15275437085122842, 'gradnorm/compressor.layer6': 0.04600700767280824, 'gradnorm/compressor.layer7': 0.13154678424152869, 'gradnorm/decoder.layer0': 0.24066700785425374, 'gradnorm/decoder.layer1': 0.22689750173600565, 'gradnorm/decoder.layer2': 0.09704139267160175, 'gradnorm/max_param': 1.7570157051086426, 'epoch': 3.35}

 34%|███▎      | 6400/19090 [3:07:25<6:51:19,  1.94s/it]
 34%|███▎      | 6401/19090 [3:07:26<6:09:47,  1.75s/it]
 34%|███▎      | 6402/19090 [3:07:27<5:40:28,  1.61s/it]
 34%|███▎      | 6403/19090 [3:07:30<6:56:49,  1.97s/it]
 34%|███▎      | 6404/19090 [3:07:31<6:13:24,  1.77s/it]
 34%|███▎      | 6405/19090 [3:07:34<6:50:57,  1.94s/it]
 34%|███▎      | 6406/19090 [3:07:35<6:28:20,  1.84s/it]
 34%|███▎      | 6407/19090 [3:07:37<5:53:23,  1.67s/it]
 34%|███▎      | 6408/19090 [3:07:38<5:28:58,  1.56s/it]
 34%|███▎      | 6409/19090 [3:07:41<7:07:21,  2.02s/it]
 34%|███▎      | 6410/19090 [3:07:42<6:21:12,  1.80s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.862740993499756, 'learning_rate': 4.200360035694293e-05, 'recon_loss': 0.0012749714107485488, 'recon_static': 0.00032187197430175727, 'recon_motion': 0.0009530994619126431, 'recon_norm': 1.249460804462433, 'recon_norm_std': 0.07073191329836845, 'target_norm': 1.6206401467323304, 'recon_cos': 0.7066152349114418, 'gradnorm/compressor.query': 0.005882992928012681, 'gradnorm/compressor.layer0': 0.12608999128541926, 'gradnorm/compressor.layer1': 2.6247111570298385, 'gradnorm/compressor.layer2': 0.8760744038080843, 'gradnorm/compressor.layer3': 0.5567224542803441, 'gradnorm/compressor.layer4': 0.215652020461437, 'gradnorm/compressor.layer5': 0.15694713047517897, 'gradnorm/compressor.layer6': 0.045845770040360945, 'gradnorm/compressor.layer7': 0.12258037814275327, 'gradnorm/decoder.layer0': 0.2572574854680163, 'gradnorm/decoder.layer1': 0.22021115589648674, 'gradnorm/decoder.layer2': 0.09928564215962665, 'gradnorm/max_param': 1.8490036725997925, 'epoch': 3.36}

 34%|███▎      | 6410/19090 [3:07:42<6:21:12,  1.80s/it]
 34%|███▎      | 6411/19090 [3:07:44<5:48:25,  1.65s/it]
 34%|███▎      | 6412/19090 [3:07:46<6:40:00,  1.89s/it]
 34%|███▎      | 6413/19090 [3:07:47<6:01:33,  1.71s/it]
 34%|███▎      | 6414/19090 [3:07:49<5:34:33,  1.58s/it]
 34%|███▎      | 6415/19090 [3:07:52<6:59:12,  1.98s/it]
 34%|███▎      | 6416/19090 [3:07:53<6:14:59,  1.78s/it]
 34%|███▎      | 6417/19090 [3:07:54<5:43:50,  1.63s/it]
 34%|███▎      | 6418/19090 [3:07:57<7:02:04,  2.00s/it]
 34%|███▎      | 6419/19090 [3:07:58<6:17:52,  1.79s/it]
 34%|███▎      | 6420/19090 [3:08:00<5:46:21,  1.64s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.719219923019409, 'learning_rate': 4.1970060598551486e-05, 'recon_loss': 0.0012782346224412323, 'recon_static': 0.0003142341047350783, 'recon_motion': 0.0009640005111577921, 'recon_norm': 1.2418090939521789, 'recon_norm_std': 0.07029222343116999, 'target_norm': 1.621676042675972, 'recon_cos': 0.7057486370205879, 'gradnorm/compressor.query': 0.00674226455676213, 'gradnorm/compressor.layer0': 0.12833581769796804, 'gradnorm/compressor.layer1': 2.4765954241083374, 'gradnorm/compressor.layer2': 0.8586341438822861, 'gradnorm/compressor.layer3': 0.5377621100324397, 'gradnorm/compressor.layer4': 0.21880724702798776, 'gradnorm/compressor.layer5': 0.16211601611402598, 'gradnorm/compressor.layer6': 0.04720137145931168, 'gradnorm/compressor.layer7': 0.1335406119646665, 'gradnorm/decoder.layer0': 0.2517110711648539, 'gradnorm/decoder.layer1': 0.22688861100691557, 'gradnorm/decoder.layer2': 0.09283953370112226, 'gradnorm/max_param': 1.74458646774292, 'epoch': 3.36}

 34%|███▎      | 6420/19090 [3:08:00<5:46:21,  1.64s/it]
 34%|███▎      | 6421/19090 [3:08:02<6:44:24,  1.92s/it]
 34%|███▎      | 6422/19090 [3:08:04<6:04:34,  1.73s/it]
 34%|███▎      | 6423/19090 [3:08:05<5:36:31,  1.59s/it]
 34%|███▎      | 6424/19090 [3:08:07<6:39:48,  1.89s/it]
 34%|███▎      | 6425/19090 [3:08:09<6:01:28,  1.71s/it]
 34%|███▎      | 6426/19090 [3:08:10<5:34:32,  1.58s/it]
 34%|███▎      | 6427/19090 [3:08:13<6:55:33,  1.97s/it]
 34%|███▎      | 6428/19090 [3:08:14<6:12:32,  1.77s/it]
 34%|███▎      | 6429/19090 [3:08:15<5:42:26,  1.62s/it]
 34%|███▎      | 6430/19090 [3:08:18<6:37:31,  1.88s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.723076343536377, 'learning_rate': 4.1936464100576755e-05, 'recon_loss': 0.0012681252817856147, 'recon_static': 0.0003127557487459853, 'recon_motion': 0.000955369501025416, 'recon_norm': 1.2426450431346894, 'recon_norm_std': 0.07099890718236565, 'target_norm': 1.619557362794876, 'recon_cos': 0.7076569125056267, 'gradnorm/compressor.query': 0.0061607266358105275, 'gradnorm/compressor.layer0': 0.12317594313614663, 'gradnorm/compressor.layer1': 2.472318171188207, 'gradnorm/compressor.layer2': 0.882123796660846, 'gradnorm/compressor.layer3': 0.5410248518064031, 'gradnorm/compressor.layer4': 0.22041470747319145, 'gradnorm/compressor.layer5': 0.154405857179872, 'gradnorm/compressor.layer6': 0.04597210552162188, 'gradnorm/compressor.layer7': 0.12820955777577955, 'gradnorm/decoder.layer0': 0.25932179445319004, 'gradnorm/decoder.layer1': 0.22195117139913215, 'gradnorm/decoder.layer2': 0.09614392866171236, 'gradnorm/max_param': 1.7417097091674805, 'epoch': 3.37}

 34%|███▎      | 6430/19090 [3:08:18<6:37:31,  1.88s/it]
 34%|███▎      | 6431/19090 [3:08:19<5:59:42,  1.70s/it]
 34%|███▎      | 6432/19090 [3:08:21<5:33:20,  1.58s/it]
 34%|███▎      | 6433/19090 [3:08:23<6:42:32,  1.91s/it]
 34%|███▎      | 6434/19090 [3:08:24<6:03:22,  1.72s/it]
 34%|███▎      | 6435/19090 [3:08:26<5:36:02,  1.59s/it]
 34%|███▎      | 6436/19090 [3:08:28<6:47:30,  1.93s/it]
 34%|███▎      | 6437/19090 [3:08:30<6:07:00,  1.74s/it]
 34%|███▎      | 6438/19090 [3:08:31<5:57:50,  1.70s/it]
 34%|███▎      | 6439/19090 [3:08:34<7:13:13,  2.05s/it]
 34%|███▎      | 6440/19090 [3:08:36<6:24:49,  1.83s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.9707486629486084, 'learning_rate': 4.190281097534901e-05, 'recon_loss': 0.0012814967631129548, 'recon_static': 0.00032289593655150385, 'recon_motion': 0.0009586008483893238, 'recon_norm': 1.245678648352623, 'recon_norm_std': 0.07070237621665001, 'target_norm': 1.6221042633056642, 'recon_cos': 0.7053287193179131, 'gradnorm/compressor.query': 0.007510678774648194, 'gradnorm/compressor.layer0': 0.1424848368905269, 'gradnorm/compressor.layer1': 2.7312387845924264, 'gradnorm/compressor.layer2': 0.8827583498931235, 'gradnorm/compressor.layer3': 0.5927601393287367, 'gradnorm/compressor.layer4': 0.23564793299319448, 'gradnorm/compressor.layer5': 0.1598849324436035, 'gradnorm/compressor.layer6': 0.05025318803878165, 'gradnorm/compressor.layer7': 0.12561128577060124, 'gradnorm/decoder.layer0': 0.23199224433185262, 'gradnorm/decoder.layer1': 0.22581669948230024, 'gradnorm/decoder.layer2': 0.10239802606630517, 'gradnorm/max_param': 1.9237922430038452, 'epoch': 3.37}

 34%|███▎      | 6440/19090 [3:08:36<6:24:49,  1.83s/it]
 34%|███▎      | 6441/19090 [3:08:37<5:50:33,  1.66s/it]
 34%|███▎      | 6442/19090 [3:08:40<6:54:19,  1.97s/it]
 34%|███▍      | 6443/19090 [3:08:41<6:11:42,  1.76s/it]
 34%|███▍      | 6444/19090 [3:08:42<5:41:33,  1.62s/it]
 34%|███▍      | 6445/19090 [3:08:45<6:54:14,  1.97s/it]
 34%|███▍      | 6446/19090 [3:08:46<6:11:18,  1.76s/it]
 34%|███▍      | 6447/19090 [3:08:47<5:41:23,  1.62s/it]
 34%|███▍      | 6448/19090 [3:08:50<6:56:38,  1.98s/it]
 34%|███▍      | 6449/19090 [3:08:52<6:13:05,  1.77s/it]
 34%|███▍      | 6450/19090 [3:08:53<5:42:47,  1.63s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.7981362342834473, 'learning_rate': 4.1869101335387816e-05, 'recon_loss': 0.001292390952585265, 'recon_static': 0.0003267233965743799, 'recon_motion': 0.00096566756546963, 'recon_norm': 1.2485584974288941, 'recon_norm_std': 0.0688831578940153, 'target_norm': 1.6220680862665176, 'recon_cos': 0.7025812819600106, 'gradnorm/compressor.query': 0.006790400350835896, 'gradnorm/compressor.layer0': 0.13985138930867325, 'gradnorm/compressor.layer1': 2.552333762409436, 'gradnorm/compressor.layer2': 0.8938218681410537, 'gradnorm/compressor.layer3': 0.522910892282446, 'gradnorm/compressor.layer4': 0.22598656258839397, 'gradnorm/compressor.layer5': 0.16542074546381488, 'gradnorm/compressor.layer6': 0.05141968084604782, 'gradnorm/compressor.layer7': 0.13380942920562644, 'gradnorm/decoder.layer0': 0.24691168023692098, 'gradnorm/decoder.layer1': 0.2297453518756361, 'gradnorm/decoder.layer2': 0.1023211525218997, 'gradnorm/max_param': 1.7974622249603271, 'epoch': 3.38}

 34%|███▍      | 6450/19090 [3:08:53<5:42:47,  1.63s/it]
 34%|███▍      | 6451/19090 [3:08:56<6:56:21,  1.98s/it]
 34%|███▍      | 6452/19090 [3:08:57<6:12:58,  1.77s/it]
 34%|███▍      | 6453/19090 [3:08:59<6:03:46,  1.73s/it]
 34%|███▍      | 6454/19090 [3:09:01<6:43:14,  1.91s/it]
 34%|███▍      | 6455/19090 [3:09:02<6:03:22,  1.73s/it]
 34%|███▍      | 6456/19090 [3:09:04<6:01:30,  1.72s/it]
 34%|███▍      | 6457/19090 [3:09:07<7:01:07,  2.00s/it]
 34%|███▍      | 6458/19090 [3:09:08<6:16:12,  1.79s/it]
 34%|███▍      | 6459/19090 [3:09:09<5:46:17,  1.64s/it]
 34%|███▍      | 6460/19090 [3:09:12<6:36:49,  1.89s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.9422900676727295, 'learning_rate': 4.183533529340174e-05, 'recon_loss': 0.0012716800963971763, 'recon_static': 0.00032163559808395803, 'recon_motion': 0.0009500445012236014, 'recon_norm': 1.2468949645757674, 'recon_norm_std': 0.07032080823555589, 'target_norm': 1.6183492720127106, 'recon_cos': 0.7061189174652099, 'gradnorm/compressor.query': 0.007969025391344172, 'gradnorm/compressor.layer0': 0.15020008526028514, 'gradnorm/compressor.layer1': 2.6793235776365645, 'gradnorm/compressor.layer2': 0.9181891967994571, 'gradnorm/compressor.layer3': 0.6018515053636689, 'gradnorm/compressor.layer4': 0.23572392902604328, 'gradnorm/compressor.layer5': 0.16906086602717046, 'gradnorm/compressor.layer6': 0.05161951590185861, 'gradnorm/compressor.layer7': 0.13325745014553583, 'gradnorm/decoder.layer0': 0.2723846055166974, 'gradnorm/decoder.layer1': 0.24481865382403348, 'gradnorm/decoder.layer2': 0.10806052920191919, 'gradnorm/max_param': 1.8870164155960083, 'epoch': 3.38}

 34%|███▍      | 6460/19090 [3:09:12<6:36:49,  1.89s/it]
 34%|███▍      | 6461/19090 [3:09:13<5:59:47,  1.71s/it]
 34%|███▍      | 6462/19090 [3:09:14<5:33:49,  1.59s/it]
 34%|███▍      | 6463/19090 [3:09:17<7:03:28,  2.01s/it]
 34%|███▍      | 6464/19090 [3:09:18<6:17:44,  1.80s/it]
 34%|███▍      | 6465/19090 [3:09:20<5:45:56,  1.64s/it]
 34%|███▍      | 6466/19090 [3:09:22<6:49:27,  1.95s/it]
 34%|███▍      | 6467/19090 [3:09:24<6:07:38,  1.75s/it]
 34%|███▍      | 6468/19090 [3:09:25<5:38:39,  1.61s/it]
 34%|███▍      | 6469/19090 [3:09:28<7:04:45,  2.02s/it]
 34%|███▍      | 6470/19090 [3:09:29<6:19:01,  1.80s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.89821457862854, 'learning_rate': 4.180151296228792e-05, 'recon_loss': 0.0012800900207366795, 'recon_static': 0.00032630298082949596, 'recon_motion': 0.0009537870486383326, 'recon_norm': 1.2466696053743362, 'recon_norm_std': 0.06961012659594416, 'target_norm': 1.6216327846050262, 'recon_cos': 0.7054598420858383, 'gradnorm/compressor.query': 0.006305599096536466, 'gradnorm/compressor.layer0': 0.13110813296123783, 'gradnorm/compressor.layer1': 2.6600692219439415, 'gradnorm/compressor.layer2': 0.8897401195062195, 'gradnorm/compressor.layer3': 0.5473678275607777, 'gradnorm/compressor.layer4': 0.2304011827768201, 'gradnorm/compressor.layer5': 0.1596228742935906, 'gradnorm/compressor.layer6': 0.049774745280614194, 'gradnorm/compressor.layer7': 0.12713550525299117, 'gradnorm/decoder.layer0': 0.24073767470116716, 'gradnorm/decoder.layer1': 0.2223783230793257, 'gradnorm/decoder.layer2': 0.10290277100186798, 'gradnorm/max_param': 1.8738512992858887, 'epoch': 3.39}

 34%|███▍      | 6470/19090 [3:09:29<6:19:01,  1.80s/it]
 34%|███▍      | 6471/19090 [3:09:31<5:47:37,  1.65s/it]
 34%|███▍      | 6472/19090 [3:09:33<6:59:00,  1.99s/it]
 34%|███▍      | 6473/19090 [3:09:35<6:14:35,  1.78s/it]
 34%|███▍      | 6474/19090 [3:09:36<5:43:13,  1.63s/it]
 34%|███▍      | 6475/19090 [3:09:39<7:12:44,  2.06s/it]
 34%|███▍      | 6476/19090 [3:09:40<6:24:12,  1.83s/it]
 34%|███▍      | 6477/19090 [3:09:42<5:50:10,  1.67s/it]
 34%|███▍      | 6478/19090 [3:09:44<6:50:16,  1.95s/it]
 34%|███▍      | 6479/19090 [3:09:45<6:08:20,  1.75s/it]
 34%|███▍      | 6480/19090 [3:09:47<5:38:58,  1.61s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.9888381958007812, 'learning_rate': 4.176763445513167e-05, 'recon_loss': 0.0012776376534020528, 'recon_static': 0.00031349565470009113, 'recon_motion': 0.0009641419877880253, 'recon_norm': 1.2402324050664901, 'recon_norm_std': 0.07037347164005041, 'target_norm': 1.6208247691392899, 'recon_cos': 0.70591149777174, 'gradnorm/compressor.query': 0.008325426683678713, 'gradnorm/compressor.layer0': 0.14694731357608715, 'gradnorm/compressor.layer1': 2.7364521965090667, 'gradnorm/compressor.layer2': 0.9231173339504777, 'gradnorm/compressor.layer3': 0.5785732078356691, 'gradnorm/compressor.layer4': 0.23099614269887933, 'gradnorm/compressor.layer5': 0.16914315849503547, 'gradnorm/compressor.layer6': 0.05300312305768346, 'gradnorm/compressor.layer7': 0.1309036208549196, 'gradnorm/decoder.layer0': 0.2677360820242765, 'gradnorm/decoder.layer1': 0.22968036971410505, 'gradnorm/decoder.layer2': 0.100389678000648, 'gradnorm/max_param': 1.927824854850769, 'epoch': 3.39}

 34%|███▍      | 6480/19090 [3:09:47<5:38:58,  1.61s/it]
 34%|███▍      | 6481/19090 [3:09:49<6:38:31,  1.90s/it]
 34%|███▍      | 6482/19090 [3:09:51<5:59:53,  1.71s/it]
 34%|███▍      | 6483/19090 [3:09:52<5:32:56,  1.58s/it]
 34%|███▍      | 6484/19090 [3:09:55<6:45:55,  1.93s/it]
 34%|███▍      | 6485/19090 [3:09:56<6:05:17,  1.74s/it]
 34%|███▍      | 6486/19090 [3:09:57<5:36:32,  1.60s/it]
 34%|███▍      | 6487/19090 [3:10:00<7:15:18,  2.07s/it]
 34%|███▍      | 6488/19090 [3:10:02<6:26:03,  1.84s/it]
 34%|███▍      | 6489/19090 [3:10:03<5:51:49,  1.68s/it]
 34%|███▍      | 6490/19090 [3:10:05<5:49:46,  1.67s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 2.9842870235443115, 'learning_rate': 4.173369988520617e-05, 'recon_loss': 0.0012592940591275692, 'recon_static': 0.00031090731281437913, 'recon_motion': 0.0009483867310336791, 'recon_norm': 1.2384721100330354, 'recon_norm_std': 0.07034865068271756, 'target_norm': 1.6190800428390504, 'recon_cos': 0.7096333146095276, 'gradnorm/compressor.query': 0.00563377805322516, 'gradnorm/compressor.layer0': 0.11956993907407805, 'gradnorm/compressor.layer1': 2.7333440962633815, 'gradnorm/compressor.layer2': 0.9218318735153768, 'gradnorm/compressor.layer3': 0.5949891861966728, 'gradnorm/compressor.layer4': 0.2223749475178863, 'gradnorm/compressor.layer5': 0.16099703298746196, 'gradnorm/compressor.layer6': 0.05085750045339588, 'gradnorm/compressor.layer7': 0.1322047831236699, 'gradnorm/decoder.layer0': 0.24281003373905236, 'gradnorm/decoder.layer1': 0.22779808527340217, 'gradnorm/decoder.layer2': 0.10197494737337826, 'gradnorm/max_param': 1.9259934425354004, 'epoch': 3.4}

 34%|███▍      | 6490/19090 [3:10:05<5:49:46,  1.67s/it]
 34%|███▍      | 6491/19090 [3:10:06<5:26:35,  1.56s/it]
 34%|███▍      | 6492/19090 [3:10:07<5:10:01,  1.48s/it]
 34%|███▍      | 6493/19090 [3:10:09<5:50:12,  1.67s/it]
 34%|███▍      | 6494/19090 [3:10:11<5:27:47,  1.56s/it]
 34%|███▍      | 6495/19090 [3:10:12<5:10:54,  1.48s/it]
 34%|███▍      | 6496/19090 [3:10:14<5:33:59,  1.59s/it]
 34%|███▍      | 6497/19090 [3:10:15<5:16:17,  1.51s/it]
 34%|███▍      | 6498/19090 [3:10:16<5:02:22,  1.44s/it]
 34%|███▍      | 6499/19090 [3:10:18<5:27:28,  1.56s/it]
 34%|███▍      | 6500/19090 [3:10:19<5:10:29,  1.48s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.148757219314575, 'learning_rate': 4.169970936597201e-05, 'recon_loss': 0.0012716358323814347, 'recon_static': 0.0003083054492890369, 'recon_motion': 0.0009633303983719088, 'recon_norm': 1.2377372443675996, 'recon_norm_std': 0.07040919633582235, 'target_norm': 1.6242297381162643, 'recon_cos': 0.7089389488101006, 'gradnorm/compressor.query': 0.006763124484601272, 'gradnorm/compressor.layer0': 0.12842003563207538, 'gradnorm/compressor.layer1': 2.8925331210100493, 'gradnorm/compressor.layer2': 0.9335105571630017, 'gradnorm/compressor.layer3': 0.638613193627575, 'gradnorm/compressor.layer4': 0.2442607050275572, 'gradnorm/compressor.layer5': 0.17329877984868788, 'gradnorm/compressor.layer6': 0.05371346467748479, 'gradnorm/compressor.layer7': 0.13858882673823614, 'gradnorm/decoder.layer0': 0.25911877339615863, 'gradnorm/decoder.layer1': 0.2472929723962786, 'gradnorm/decoder.layer2': 0.10970198340719141, 'gradnorm/max_param': 2.038257122039795, 'epoch': 3.4}

 34%|███▍      | 6500/19090 [3:10:19<5:10:29,  1.48s/it]
 34%|███▍      | 6501/19090 [3:10:24<7:56:22,  2.27s/it]
 34%|███▍      | 6502/19090 [3:10:25<6:55:57,  1.98s/it]
 34%|███▍      | 6503/19090 [3:10:26<6:13:25,  1.78s/it]
 34%|███▍      | 6504/19090 [3:10:27<5:42:23,  1.63s/it]
 34%|███▍      | 6505/19090 [3:10:29<5:21:19,  1.53s/it]
 34%|███▍      | 6506/19090 [3:10:30<5:06:20,  1.46s/it]
 34%|███▍      | 6507/19090 [3:10:31<4:55:34,  1.41s/it]
 34%|███▍      | 6508/19090 [3:10:33<4:47:54,  1.37s/it]
 34%|███▍      | 6509/19090 [3:10:34<4:42:46,  1.35s/it]
 34%|███▍      | 6510/19090 [3:10:35<4:39:21,  1.33s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.051161766052246, 'learning_rate': 4.166566301107687e-05, 'recon_loss': 0.00126119417254813, 'recon_static': 0.000310946830722969, 'recon_motion': 0.0009502473578322679, 'recon_norm': 1.2400522470474242, 'recon_norm_std': 0.07030261447653174, 'target_norm': 1.6185872495174407, 'recon_cos': 0.7093646228313446, 'gradnorm/compressor.query': 0.008367786921161365, 'gradnorm/compressor.layer0': 0.14744735175194879, 'gradnorm/compressor.layer1': 2.7885942680481093, 'gradnorm/compressor.layer2': 0.956809349025467, 'gradnorm/compressor.layer3': 0.6030500523542428, 'gradnorm/compressor.layer4': 0.2435658134972157, 'gradnorm/compressor.layer5': 0.16928059598695339, 'gradnorm/compressor.layer6': 0.05095868336070817, 'gradnorm/compressor.layer7': 0.13241047880484166, 'gradnorm/decoder.layer0': 0.2538002673284477, 'gradnorm/decoder.layer1': 0.22361800298837634, 'gradnorm/decoder.layer2': 0.099180011433226, 'gradnorm/max_param': 1.9647752046585083, 'epoch': 3.41}

 34%|███▍      | 6510/19090 [3:10:35<4:39:21,  1.33s/it]
 34%|███▍      | 6511/19090 [3:10:37<4:37:10,  1.32s/it]
 34%|███▍      | 6512/19090 [3:10:38<4:35:07,  1.31s/it]
 34%|███▍      | 6513/19090 [3:10:39<4:33:55,  1.31s/it]
 34%|███▍      | 6514/19090 [3:10:40<4:32:52,  1.30s/it]
 34%|███▍      | 6515/19090 [3:10:42<4:32:29,  1.30s/it]
 34%|███▍      | 6516/19090 [3:10:43<4:31:50,  1.30s/it]
 34%|███▍      | 6517/19090 [3:10:44<4:31:07,  1.29s/it]
 34%|███▍      | 6518/19090 [3:10:46<4:30:41,  1.29s/it]
 34%|███▍      | 6519/19090 [3:10:47<4:30:22,  1.29s/it]
 34%|███▍      | 6520/19090 [3:10:48<4:49:55,  1.38s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.276854991912842, 'learning_rate': 4.163156093435511e-05, 'recon_loss': 0.0012663927831454203, 'recon_static': 0.00030940798023948446, 'recon_motion': 0.0009569848290993832, 'recon_norm': 1.2378281623125076, 'recon_norm_std': 0.0702098316513002, 'target_norm': 1.6204562455415725, 'recon_cos': 0.7085150003433227, 'gradnorm/compressor.query': 0.007210165850981136, 'gradnorm/compressor.layer0': 0.14215207785284625, 'gradnorm/compressor.layer1': 3.0022606279558457, 'gradnorm/compressor.layer2': 1.022636391233427, 'gradnorm/compressor.layer3': 0.6408918801769307, 'gradnorm/compressor.layer4': 0.24683284278857803, 'gradnorm/compressor.layer5': 0.16433518885828918, 'gradnorm/compressor.layer6': 0.05223413134377421, 'gradnorm/compressor.layer7': 0.14282624328812646, 'gradnorm/decoder.layer0': 0.2670861615305477, 'gradnorm/decoder.layer1': 0.23259189541882125, 'gradnorm/decoder.layer2': 0.1045380607518783, 'gradnorm/max_param': 2.1154682636260986, 'epoch': 3.42}

 34%|███▍      | 6520/19090 [3:10:48<4:49:55,  1.38s/it]
 34%|███▍      | 6521/19090 [3:10:50<5:03:16,  1.45s/it]
 34%|███▍      | 6522/19090 [3:10:51<4:53:22,  1.40s/it]
 34%|███▍      | 6523/19090 [3:10:53<5:14:42,  1.50s/it]
 34%|███▍      | 6524/19090 [3:10:55<5:30:54,  1.58s/it]
 34%|███▍      | 6525/19090 [3:10:56<5:13:04,  1.49s/it]
 34%|███▍      | 6526/19090 [3:10:58<5:12:21,  1.49s/it]
 34%|███▍      | 6527/19090 [3:10:59<5:26:25,  1.56s/it]
 34%|███▍      | 6528/19090 [3:11:01<5:09:24,  1.48s/it]
 34%|███▍      | 6529/19090 [3:11:03<5:34:06,  1.60s/it]
 34%|███▍      | 6530/19090 [3:11:04<5:15:23,  1.51s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.2442662715911865, 'learning_rate': 4.1597403249827405e-05, 'recon_loss': 0.001252034047502093, 'recon_static': 0.0003009781357832253, 'recon_motion': 0.000951055895711761, 'recon_norm': 1.2343197911977768, 'recon_norm_std': 0.07011976428329944, 'target_norm': 1.6204445958137512, 'recon_cos': 0.7123251080513, 'gradnorm/compressor.query': 0.008372789078565207, 'gradnorm/compressor.layer0': 0.15245122664085725, 'gradnorm/compressor.layer1': 2.9855667051709616, 'gradnorm/compressor.layer2': 0.9803139613793972, 'gradnorm/compressor.layer3': 0.6148930647387746, 'gradnorm/compressor.layer4': 0.24970559577395024, 'gradnorm/compressor.layer5': 0.17150804512704257, 'gradnorm/compressor.layer6': 0.0520870515370362, 'gradnorm/compressor.layer7': 0.136656931116294, 'gradnorm/decoder.layer0': 0.26186270099974335, 'gradnorm/decoder.layer1': 0.2367631282061748, 'gradnorm/decoder.layer2': 0.10700517810230968, 'gradnorm/max_param': 2.10368275642395, 'epoch': 3.42}

 34%|███▍      | 6530/19090 [3:11:04<5:15:23,  1.51s/it]
 34%|███▍      | 6531/19090 [3:11:05<5:02:27,  1.44s/it]
 34%|███▍      | 6532/19090 [3:11:07<5:38:08,  1.62s/it]
 34%|███▍      | 6533/19090 [3:11:09<5:36:00,  1.61s/it]
 34%|███▍      | 6534/19090 [3:11:10<5:16:12,  1.51s/it]
 34%|███▍      | 6535/19090 [3:11:12<5:26:50,  1.56s/it]
 34%|███▍      | 6536/19090 [3:11:13<5:31:58,  1.59s/it]
 34%|███▍      | 6537/19090 [3:11:15<5:14:02,  1.50s/it]
 34%|███▍      | 6538/19090 [3:11:16<5:24:55,  1.55s/it]
 34%|███▍      | 6539/19090 [3:11:18<5:24:48,  1.55s/it]
 34%|███▍      | 6540/19090 [3:11:19<5:09:24,  1.48s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.105447769165039, 'learning_rate': 4.156319007170032e-05, 'recon_loss': 0.0012663748813793064, 'recon_static': 0.0003049681967240758, 'recon_motion': 0.0009614067443180829, 'recon_norm': 1.2347960948944092, 'recon_norm_std': 0.06861332906410098, 'target_norm': 1.6194952577352524, 'recon_cos': 0.7080676972866058, 'gradnorm/compressor.query': 0.008777629598637775, 'gradnorm/compressor.layer0': 0.14947164585275965, 'gradnorm/compressor.layer1': 2.8459015137606904, 'gradnorm/compressor.layer2': 0.9684298817032428, 'gradnorm/compressor.layer3': 0.5817140828738727, 'gradnorm/compressor.layer4': 0.23291858521890002, 'gradnorm/compressor.layer5': 0.1752716519031237, 'gradnorm/compressor.layer6': 0.05447937539611045, 'gradnorm/compressor.layer7': 0.14738383393144766, 'gradnorm/decoder.layer0': 0.26175192722296037, 'gradnorm/decoder.layer1': 0.23702587813122275, 'gradnorm/decoder.layer2': 0.10783422396350766, 'gradnorm/max_param': 2.004925489425659, 'epoch': 3.43}

 34%|███▍      | 6540/19090 [3:11:19<5:09:24,  1.48s/it]
 34%|███▍      | 6541/19090 [3:11:21<5:12:16,  1.49s/it]
 34%|███▍      | 6542/19090 [3:11:23<5:33:07,  1.59s/it]
 34%|███▍      | 6543/19090 [3:11:24<5:14:29,  1.50s/it]
 34%|███▍      | 6544/19090 [3:11:25<5:13:37,  1.50s/it]
 34%|███▍      | 6545/19090 [3:11:27<5:30:09,  1.58s/it]
 34%|███▍      | 6546/19090 [3:11:28<5:12:38,  1.50s/it]
 34%|███▍      | 6547/19090 [3:11:30<5:23:07,  1.55s/it]
 34%|███▍      | 6548/19090 [3:11:32<5:43:02,  1.64s/it]
 34%|███▍      | 6549/19090 [3:11:33<5:21:18,  1.54s/it]
 34%|███▍      | 6550/19090 [3:11:34<5:06:25,  1.47s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.275651454925537, 'learning_rate': 4.152892151436602e-05, 'recon_loss': 0.0012713397853076458, 'recon_static': 0.0003081347014813218, 'recon_motion': 0.0009632050918298773, 'recon_norm': 1.2357277572154999, 'recon_norm_std': 0.0697520355693996, 'target_norm': 1.622642347216606, 'recon_cos': 0.7080738961696624, 'gradnorm/compressor.query': 0.00667504497927838, 'gradnorm/compressor.layer0': 0.1352426820359998, 'gradnorm/compressor.layer1': 2.9999014520353544, 'gradnorm/compressor.layer2': 1.0159487868415127, 'gradnorm/compressor.layer3': 0.6469001930970101, 'gradnorm/compressor.layer4': 0.24537140934668872, 'gradnorm/compressor.layer5': 0.17001128858714287, 'gradnorm/compressor.layer6': 0.05659291913859199, 'gradnorm/compressor.layer7': 0.144984777377031, 'gradnorm/decoder.layer0': 0.2798763080062049, 'gradnorm/decoder.layer1': 0.24332709764275837, 'gradnorm/decoder.layer2': 0.10315395096953175, 'gradnorm/max_param': 2.113792657852173, 'epoch': 3.43}

 34%|███▍      | 6550/19090 [3:11:34<5:06:25,  1.47s/it]
 34%|███▍      | 6551/19090 [3:11:36<5:31:40,  1.59s/it]
 34%|███▍      | 6552/19090 [3:11:38<5:13:09,  1.50s/it]
 34%|███▍      | 6553/19090 [3:11:39<5:05:30,  1.46s/it]
 34%|███▍      | 6554/19090 [3:11:41<5:30:47,  1.58s/it]
 34%|███▍      | 6555/19090 [3:11:42<5:13:00,  1.50s/it]
 34%|███▍      | 6556/19090 [3:11:44<5:21:06,  1.54s/it]
 34%|███▍      | 6557/19090 [3:11:46<5:31:38,  1.59s/it]
 34%|███▍      | 6558/19090 [3:11:47<5:13:32,  1.50s/it]
 34%|███▍      | 6559/19090 [3:11:48<5:09:02,  1.48s/it]
 34%|███▍      | 6560/19090 [3:11:50<5:37:49,  1.62s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.262646198272705, 'learning_rate': 4.1494597692401796e-05, 'recon_loss': 0.0012591696257004514, 'recon_static': 0.0002951603019027971, 'recon_motion': 0.0009640093456255272, 'recon_norm': 1.228601649403572, 'recon_norm_std': 0.06884317426010966, 'target_norm': 1.6217506557703019, 'recon_cos': 0.7112148523330688, 'gradnorm/compressor.query': 0.005725459352243485, 'gradnorm/compressor.layer0': 0.11853282701784354, 'gradnorm/compressor.layer1': 2.9663037310273928, 'gradnorm/compressor.layer2': 1.0840816419796786, 'gradnorm/compressor.layer3': 0.62099515984134, 'gradnorm/compressor.layer4': 0.25788153099891065, 'gradnorm/compressor.layer5': 0.18052286352375968, 'gradnorm/compressor.layer6': 0.05223172204046948, 'gradnorm/compressor.layer7': 0.14191941727632446, 'gradnorm/decoder.layer0': 0.27596551666508184, 'gradnorm/decoder.layer1': 0.24718525440628372, 'gradnorm/decoder.layer2': 0.10832062927210781, 'gradnorm/max_param': 2.0904366970062256, 'epoch': 3.44}

 34%|███▍      | 6560/19090 [3:11:50<5:37:49,  1.62s/it]
 34%|███▍      | 6561/19090 [3:11:51<5:17:40,  1.52s/it]
 34%|███▍      | 6562/19090 [3:11:53<5:08:43,  1.48s/it]
 34%|███▍      | 6563/19090 [3:11:55<5:30:14,  1.58s/it]
 34%|███▍      | 6564/19090 [3:11:56<5:12:14,  1.50s/it]
 34%|███▍      | 6565/19090 [3:11:57<5:13:28,  1.50s/it]
 34%|███▍      | 6566/19090 [3:11:59<5:30:22,  1.58s/it]
 34%|███▍      | 6567/19090 [3:12:01<5:12:10,  1.50s/it]
 34%|███▍      | 6568/19090 [3:12:02<5:31:25,  1.59s/it]
 34%|███▍      | 6569/19090 [3:12:04<5:27:08,  1.57s/it]
 34%|███▍      | 6570/19090 [3:12:06<5:31:49,  1.59s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.458907127380371, 'learning_rate': 4.14602187205697e-05, 'recon_loss': 0.0012570342369144782, 'recon_static': 0.00029823995428159835, 'recon_motion': 0.0009587943000951782, 'recon_norm': 1.230912059545517, 'recon_norm_std': 0.07012857738882303, 'target_norm': 1.6228488683700562, 'recon_cos': 0.7122087985277176, 'gradnorm/compressor.query': 0.006306897952623764, 'gradnorm/compressor.layer0': 0.12377481245536365, 'gradnorm/compressor.layer1': 3.1841761892922267, 'gradnorm/compressor.layer2': 1.0540764628078438, 'gradnorm/compressor.layer3': 0.6552900605004869, 'gradnorm/compressor.layer4': 0.2536087934694056, 'gradnorm/compressor.layer5': 0.17682988715349263, 'gradnorm/compressor.layer6': 0.05113147358274263, 'gradnorm/compressor.layer7': 0.13869560065599723, 'gradnorm/decoder.layer0': 0.281927868078354, 'gradnorm/decoder.layer1': 0.24604994520240256, 'gradnorm/decoder.layer2': 0.10842103550744411, 'gradnorm/max_param': 2.244169235229492, 'epoch': 3.44}

 34%|███▍      | 6570/19090 [3:12:06<5:31:49,  1.59s/it]
 34%|███▍      | 6571/19090 [3:12:07<5:19:41,  1.53s/it]
 34%|███▍      | 6572/19090 [3:12:09<5:27:44,  1.57s/it]
 34%|███▍      | 6573/19090 [3:12:10<5:10:11,  1.49s/it]
 34%|███▍      | 6574/19090 [3:12:11<5:12:29,  1.50s/it]
 34%|███▍      | 6575/19090 [3:12:13<5:20:23,  1.54s/it]
 34%|███▍      | 6576/19090 [3:12:14<5:10:05,  1.49s/it]
 34%|███▍      | 6577/19090 [3:12:16<5:15:30,  1.51s/it]
 34%|███▍      | 6578/19090 [3:12:18<5:47:45,  1.67s/it]
 34%|███▍      | 6579/19090 [3:12:19<5:24:40,  1.56s/it]
 34%|███▍      | 6580/19090 [3:12:21<5:08:51,  1.48s/it]
                                                        
{'loss': 0.0012, 'grad_norm': 3.481058359146118, 'learning_rate': 4.142578471381622e-05, 'recon_loss': 0.0012379918596707284, 'recon_static': 0.00028599850047612564, 'recon_motion': 0.0009519933635601773, 'recon_norm': 1.2264352023601532, 'recon_norm_std': 0.06863828310742974, 'target_norm': 1.6179727077484132, 'recon_cos': 0.7149741664528847, 'gradnorm/compressor.query': 0.005841127744568955, 'gradnorm/compressor.layer0': 0.1202332742044525, 'gradnorm/compressor.layer1': 3.1991200787019705, 'gradnorm/compressor.layer2': 1.0926659252469906, 'gradnorm/compressor.layer3': 0.6401080162195382, 'gradnorm/compressor.layer4': 0.24650050411259866, 'gradnorm/compressor.layer5': 0.18749154588395714, 'gradnorm/compressor.layer6': 0.052755047137208466, 'gradnorm/compressor.layer7': 0.14172783187441584, 'gradnorm/decoder.layer0': 0.26649678445472147, 'gradnorm/decoder.layer1': 0.2543468027865533, 'gradnorm/decoder.layer2': 0.10369166663922703, 'gradnorm/max_param': 2.2547643184661865, 'epoch': 3.45}

 34%|███▍      | 6580/19090 [3:12:21<5:08:51,  1.48s/it]
 34%|███▍      | 6581/19090 [3:12:23<5:50:25,  1.68s/it]
 34%|███▍      | 6582/19090 [3:12:24<5:26:31,  1.57s/it]
 34%|███▍      | 6583/19090 [3:12:25<5:09:44,  1.49s/it]
 34%|███▍      | 6584/19090 [3:12:27<5:41:10,  1.64s/it]
 34%|███▍      | 6585/19090 [3:12:29<5:19:48,  1.53s/it]
 34%|███▍      | 6586/19090 [3:12:30<5:05:05,  1.46s/it]
 35%|███▍      | 6587/19090 [3:12:32<5:34:30,  1.61s/it]
 35%|███▍      | 6588/19090 [3:12:33<5:15:18,  1.51s/it]
 35%|███▍      | 6589/19090 [3:12:34<5:01:53,  1.45s/it]
 35%|███▍      | 6590/19090 [3:12:37<6:06:16,  1.76s/it]
                                                        
{'loss': 0.0012, 'grad_norm': 3.5285422801971436, 'learning_rate': 4.1391295787271836e-05, 'recon_loss': 0.0012374497717246413, 'recon_static': 0.00028392958047334104, 'recon_motion': 0.0009535201679682359, 'recon_norm': 1.2237626910209656, 'recon_norm_std': 0.06914239414036275, 'target_norm': 1.6184884637594223, 'recon_cos': 0.7157543271780014, 'gradnorm/compressor.query': 0.006835495240516876, 'gradnorm/compressor.layer0': 0.1462263887248143, 'gradnorm/compressor.layer1': 3.251440609247142, 'gradnorm/compressor.layer2': 1.0746111898569848, 'gradnorm/compressor.layer3': 0.6545274470572364, 'gradnorm/compressor.layer4': 0.2591356080437592, 'gradnorm/compressor.layer5': 0.1799871392755174, 'gradnorm/compressor.layer6': 0.05005471264740449, 'gradnorm/compressor.layer7': 0.14646079191470368, 'gradnorm/decoder.layer0': 0.28285685449608916, 'gradnorm/decoder.layer1': 0.24186049966201312, 'gradnorm/decoder.layer2': 0.11005619489869997, 'gradnorm/max_param': 2.2915046215057373, 'epoch': 3.45}

 35%|███▍      | 6590/19090 [3:12:37<6:06:16,  1.76s/it]
 35%|███▍      | 6591/19090 [3:12:38<5:36:42,  1.62s/it]
 35%|███▍      | 6592/19090 [3:12:40<5:16:09,  1.52s/it]
 35%|███▍      | 6593/19090 [3:12:42<6:34:13,  1.89s/it]
 35%|███▍      | 6594/19090 [3:12:44<5:56:22,  1.71s/it]
 35%|███▍      | 6595/19090 [3:12:45<5:30:16,  1.59s/it]
 35%|███▍      | 6596/19090 [3:12:47<6:27:02,  1.86s/it]
 35%|███▍      | 6597/19090 [3:12:49<5:58:35,  1.72s/it]
 35%|███▍      | 6598/19090 [3:12:50<5:50:14,  1.68s/it]
 35%|███▍      | 6599/19090 [3:12:53<6:35:53,  1.90s/it]
 35%|███▍      | 6600/19090 [3:12:54<6:01:36,  1.74s/it]
                                                        
{'loss': 0.0012, 'grad_norm': 3.702876329421997, 'learning_rate': 4.135675205625065e-05, 'recon_loss': 0.0012479140277719125, 'recon_static': 0.0002910396644438151, 'recon_motion': 0.0009568743640556932, 'recon_norm': 1.2277122616767884, 'recon_norm_std': 0.0683579158037901, 'target_norm': 1.6205145806074142, 'recon_cos': 0.7137444525957107, 'gradnorm/compressor.query': 0.00747975904211189, 'gradnorm/compressor.layer0': 0.14513206611026988, 'gradnorm/compressor.layer1': 3.437787962903015, 'gradnorm/compressor.layer2': 1.0994372055212054, 'gradnorm/compressor.layer3': 0.6327887401123387, 'gradnorm/compressor.layer4': 0.2542646519509826, 'gradnorm/compressor.layer5': 0.18694454610217057, 'gradnorm/compressor.layer6': 0.0513381408110602, 'gradnorm/compressor.layer7': 0.14287045309844867, 'gradnorm/decoder.layer0': 0.25881173376427324, 'gradnorm/decoder.layer1': 0.24826807386847247, 'gradnorm/decoder.layer2': 0.10647653419303364, 'gradnorm/max_param': 2.4228529930114746, 'epoch': 3.46}

 35%|███▍      | 6600/19090 [3:12:54<6:01:36,  1.74s/it]
 35%|███▍      | 6601/19090 [3:12:56<6:17:47,  1.82s/it]
 35%|███▍      | 6602/19090 [3:12:58<6:31:34,  1.88s/it]
 35%|███▍      | 6603/19090 [3:13:00<6:00:12,  1.73s/it]
 35%|███▍      | 6604/19090 [3:13:02<6:24:37,  1.85s/it]
 35%|███▍      | 6605/19090 [3:13:03<6:21:30,  1.83s/it]
 35%|███▍      | 6606/19090 [3:13:05<5:47:36,  1.67s/it]
 35%|███▍      | 6607/19090 [3:13:07<6:10:56,  1.78s/it]
 35%|███▍      | 6608/19090 [3:13:09<6:27:12,  1.86s/it]
 35%|███▍      | 6609/19090 [3:13:10<5:51:23,  1.69s/it]
 35%|███▍      | 6610/19090 [3:13:12<6:22:08,  1.84s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.47804594039917, 'learning_rate': 4.132215363624999e-05, 'recon_loss': 0.001267480693059042, 'recon_static': 0.00030883907384122723, 'recon_motion': 0.000958641606848687, 'recon_norm': 1.2348232924938203, 'recon_norm_std': 0.06869107540696859, 'target_norm': 1.6219045579433442, 'recon_cos': 0.7091147914528847, 'gradnorm/compressor.query': 0.007180163197991151, 'gradnorm/compressor.layer0': 0.15034700355545697, 'gradnorm/compressor.layer1': 3.198558250127011, 'gradnorm/compressor.layer2': 1.0726677812705245, 'gradnorm/compressor.layer3': 0.632769901113391, 'gradnorm/compressor.layer4': 0.24703664135584164, 'gradnorm/compressor.layer5': 0.2069977787239569, 'gradnorm/compressor.layer6': 0.05617761457934639, 'gradnorm/compressor.layer7': 0.14840901879575122, 'gradnorm/decoder.layer0': 0.2929867548178377, 'gradnorm/decoder.layer1': 0.25500745393106045, 'gradnorm/decoder.layer2': 0.1114983553831945, 'gradnorm/max_param': 2.253770351409912, 'epoch': 3.46}

 35%|███▍      | 6610/19090 [3:13:12<6:22:08,  1.84s/it]
 35%|███▍      | 6611/19090 [3:13:14<6:30:58,  1.88s/it]
 35%|███▍      | 6612/19090 [3:13:16<5:54:54,  1.71s/it]
 35%|███▍      | 6613/19090 [3:13:18<6:17:25,  1.82s/it]
 35%|███▍      | 6614/19090 [3:13:19<6:15:04,  1.80s/it]
 35%|███▍      | 6615/19090 [3:13:21<5:43:44,  1.65s/it]
 35%|███▍      | 6616/19090 [3:13:23<6:00:56,  1.74s/it]
 35%|███▍      | 6617/19090 [3:13:25<6:28:49,  1.87s/it]
 35%|███▍      | 6618/19090 [3:13:26<5:52:15,  1.69s/it]
 35%|███▍      | 6619/19090 [3:13:28<5:57:50,  1.72s/it]
 35%|███▍      | 6620/19090 [3:13:30<6:39:33,  1.92s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.5675957202911377, 'learning_rate': 4.128750064295007e-05, 'recon_loss': 0.0012745111103868111, 'recon_static': 0.00031980326530174353, 'recon_motion': 0.0009547078690957278, 'recon_norm': 1.2391005158424377, 'recon_norm_std': 0.0697833695448935, 'target_norm': 1.622002550959587, 'recon_cos': 0.7070525676012039, 'gradnorm/compressor.query': 0.008121473904543942, 'gradnorm/compressor.layer0': 0.16173185488912104, 'gradnorm/compressor.layer1': 3.2784067096524256, 'gradnorm/compressor.layer2': 1.1090436628670801, 'gradnorm/compressor.layer3': 0.6582058869708604, 'gradnorm/compressor.layer4': 0.2539964222346287, 'gradnorm/compressor.layer5': 0.1829086466582856, 'gradnorm/compressor.layer6': 0.059980681132366334, 'gradnorm/compressor.layer7': 0.1473834042369167, 'gradnorm/decoder.layer0': 0.2906957442821144, 'gradnorm/decoder.layer1': 0.2613216490577504, 'gradnorm/decoder.layer2': 0.11945649393020136, 'gradnorm/max_param': 2.3097100257873535, 'epoch': 3.47}

 35%|███▍      | 6620/19090 [3:13:30<6:39:33,  1.92s/it]
 35%|███▍      | 6621/19090 [3:13:32<5:59:58,  1.73s/it]
 35%|███▍      | 6622/19090 [3:13:34<6:11:56,  1.79s/it]
 35%|███▍      | 6623/19090 [3:13:36<6:34:37,  1.90s/it]
 35%|███▍      | 6624/19090 [3:13:37<5:56:33,  1.72s/it]
 35%|███▍      | 6625/19090 [3:13:39<6:11:45,  1.79s/it]
 35%|███▍      | 6626/19090 [3:13:41<6:38:42,  1.92s/it]
 35%|███▍      | 6627/19090 [3:13:42<5:59:34,  1.73s/it]
 35%|███▍      | 6628/19090 [3:13:44<5:57:26,  1.72s/it]
 35%|███▍      | 6629/19090 [3:13:46<6:16:53,  1.81s/it]
 35%|███▍      | 6630/19090 [3:13:47<5:44:10,  1.66s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.7957332134246826, 'learning_rate': 4.125279319221354e-05, 'recon_loss': 0.0012697539641521871, 'recon_static': 0.0003138418556773104, 'recon_motion': 0.0009559121375787072, 'recon_norm': 1.235445249080658, 'recon_norm_std': 0.06817871183156968, 'target_norm': 1.6209950178861618, 'recon_cos': 0.7079477742314338, 'gradnorm/compressor.query': 0.009777754463050046, 'gradnorm/compressor.layer0': 0.17755977732186598, 'gradnorm/compressor.layer1': 3.50438864628921, 'gradnorm/compressor.layer2': 1.15608301761212, 'gradnorm/compressor.layer3': 0.6828708859409989, 'gradnorm/compressor.layer4': 0.26760482654858925, 'gradnorm/compressor.layer5': 0.1968606720357919, 'gradnorm/compressor.layer6': 0.0575815106513226, 'gradnorm/compressor.layer7': 0.14633470518446529, 'gradnorm/decoder.layer0': 0.28328703590876514, 'gradnorm/decoder.layer1': 0.25207249083465216, 'gradnorm/decoder.layer2': 0.11618720541583756, 'gradnorm/max_param': 2.4691665172576904, 'epoch': 3.47}

 35%|███▍      | 6630/19090 [3:13:47<5:44:10,  1.66s/it]
 35%|███▍      | 6631/19090 [3:13:50<6:14:06,  1.80s/it]
 35%|███▍      | 6632/19090 [3:13:52<6:31:32,  1.89s/it]
 35%|███▍      | 6633/19090 [3:13:53<5:54:15,  1.71s/it]
 35%|███▍      | 6634/19090 [3:13:55<6:07:34,  1.77s/it]
 35%|███▍      | 6635/19090 [3:13:57<6:19:29,  1.83s/it]
 35%|███▍      | 6636/19090 [3:13:58<5:48:06,  1.68s/it]
 35%|███▍      | 6637/19090 [3:14:00<6:04:18,  1.76s/it]
 35%|███▍      | 6638/19090 [3:14:02<6:14:03,  1.80s/it]
 35%|███▍      | 6639/19090 [3:14:03<5:42:10,  1.65s/it]
 35%|███▍      | 6640/19090 [3:14:05<6:08:38,  1.78s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.426344394683838, 'learning_rate': 4.1218031400085146e-05, 'recon_loss': 0.0012790603388566524, 'recon_static': 0.00031878931040409955, 'recon_motion': 0.0009602710779290646, 'recon_norm': 1.2388051062822343, 'recon_norm_std': 0.06774986339733005, 'target_norm': 1.6208130210638045, 'recon_cos': 0.7055266693234443, 'gradnorm/compressor.query': 0.007106193317515539, 'gradnorm/compressor.layer0': 0.15505330057619932, 'gradnorm/compressor.layer1': 3.1374165862040355, 'gradnorm/compressor.layer2': 1.0761499419980254, 'gradnorm/compressor.layer3': 0.6385503530819011, 'gradnorm/compressor.layer4': 0.29392543991690595, 'gradnorm/compressor.layer5': 0.18758308040264887, 'gradnorm/compressor.layer6': 0.058064031250849975, 'gradnorm/compressor.layer7': 0.15537708949443338, 'gradnorm/decoder.layer0': 0.28492987805343123, 'gradnorm/decoder.layer1': 0.25360067025530025, 'gradnorm/decoder.layer2': 0.10920000552591809, 'gradnorm/max_param': 2.210218906402588, 'epoch': 3.48}

 35%|███▍      | 6640/19090 [3:14:05<6:08:38,  1.78s/it]
 35%|███▍      | 6641/19090 [3:14:07<6:19:11,  1.83s/it]
 35%|███▍      | 6642/19090 [3:14:09<5:45:51,  1.67s/it]
 35%|███▍      | 6643/19090 [3:14:11<6:10:41,  1.79s/it]
 35%|███▍      | 6644/19090 [3:14:13<6:31:33,  1.89s/it]
 35%|███▍      | 6645/19090 [3:14:14<5:54:19,  1.71s/it]
 35%|███▍      | 6646/19090 [3:14:16<6:02:59,  1.75s/it]
 35%|███▍      | 6647/19090 [3:14:18<6:42:29,  1.94s/it]
 35%|███▍      | 6648/19090 [3:14:20<6:01:51,  1.75s/it]
 35%|███▍      | 6649/19090 [3:14:22<6:12:05,  1.79s/it]
 35%|███▍      | 6650/19090 [3:14:24<6:38:45,  1.92s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.361677646636963, 'learning_rate': 4.118321538279132e-05, 'recon_loss': 0.0012785698607331141, 'recon_static': 0.00032209746495936997, 'recon_motion': 0.0009564723950461484, 'recon_norm': 1.240754246711731, 'recon_norm_std': 0.06743024252355098, 'target_norm': 1.6222312301397324, 'recon_cos': 0.7060782819986343, 'gradnorm/compressor.query': 0.0074686017559944555, 'gradnorm/compressor.layer0': 0.16110219375540694, 'gradnorm/compressor.layer1': 3.0777576827652955, 'gradnorm/compressor.layer2': 1.0492180751136546, 'gradnorm/compressor.layer3': 0.6328598758247311, 'gradnorm/compressor.layer4': 0.2696844163322742, 'gradnorm/compressor.layer5': 0.1999977718914613, 'gradnorm/compressor.layer6': 0.06136448702464405, 'gradnorm/compressor.layer7': 0.15679355490476177, 'gradnorm/decoder.layer0': 0.27860113332039355, 'gradnorm/decoder.layer1': 0.2614184977414914, 'gradnorm/decoder.layer2': 0.11775821157188016, 'gradnorm/max_param': 2.167943000793457, 'epoch': 3.48}

 35%|███▍      | 6650/19090 [3:14:24<6:38:45,  1.92s/it]
 35%|███▍      | 6651/19090 [3:14:25<5:59:44,  1.74s/it]
 35%|███▍      | 6652/19090 [3:14:27<6:02:05,  1.75s/it]
 35%|███▍      | 6653/19090 [3:14:29<6:37:01,  1.92s/it]
 35%|███▍      | 6654/19090 [3:14:30<5:58:53,  1.73s/it]
 35%|███▍      | 6655/19090 [3:14:32<5:52:21,  1.70s/it]
 35%|███▍      | 6656/19090 [3:14:34<6:25:00,  1.86s/it]
 35%|███▍      | 6657/19090 [3:14:36<5:49:58,  1.69s/it]
 35%|███▍      | 6658/19090 [3:14:37<5:59:59,  1.74s/it]
 35%|███▍      | 6659/19090 [3:14:40<6:33:10,  1.90s/it]
 35%|███▍      | 6660/19090 [3:14:41<5:55:37,  1.72s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.552145481109619, 'learning_rate': 4.11483452567398e-05, 'recon_loss': 0.001280909136403352, 'recon_static': 0.00032824102963786573, 'recon_motion': 0.0009526680747512728, 'recon_norm': 1.2403501093387603, 'recon_norm_std': 0.06749357599765063, 'target_norm': 1.6189114272594451, 'recon_cos': 0.7038500934839249, 'gradnorm/compressor.query': 0.008690599155590137, 'gradnorm/compressor.layer0': 0.1634698335454519, 'gradnorm/compressor.layer1': 3.2355295967654834, 'gradnorm/compressor.layer2': 1.1544803789800107, 'gradnorm/compressor.layer3': 0.6873166966994834, 'gradnorm/compressor.layer4': 0.27604547384023836, 'gradnorm/compressor.layer5': 0.19568111920317502, 'gradnorm/compressor.layer6': 0.05893087953292164, 'gradnorm/compressor.layer7': 0.16622121646892155, 'gradnorm/decoder.layer0': 0.28772613557095444, 'gradnorm/decoder.layer1': 0.2708412759178667, 'gradnorm/decoder.layer2': 0.12377665966467417, 'gradnorm/max_param': 2.279405117034912, 'epoch': 3.49}

 35%|███▍      | 6660/19090 [3:14:41<5:55:37,  1.72s/it]
 35%|███▍      | 6661/19090 [3:14:43<6:01:41,  1.75s/it]
 35%|███▍      | 6662/19090 [3:14:45<6:20:40,  1.84s/it]
 35%|███▍      | 6663/19090 [3:14:46<5:48:18,  1.68s/it]
 35%|███▍      | 6664/19090 [3:14:48<6:10:18,  1.79s/it]
 35%|███▍      | 6665/19090 [3:14:50<6:16:28,  1.82s/it]
 35%|███▍      | 6666/19090 [3:14:51<5:43:53,  1.66s/it]
 35%|███▍      | 6667/19090 [3:14:54<6:20:31,  1.84s/it]
 35%|███▍      | 6668/19090 [3:14:55<6:07:46,  1.78s/it]
 35%|███▍      | 6669/19090 [3:14:57<5:37:44,  1.63s/it]
 35%|███▍      | 6670/19090 [3:14:59<6:22:13,  1.85s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.242388725280762, 'learning_rate': 4.111342113851924e-05, 'recon_loss': 0.0012876674940343947, 'recon_static': 0.0003206109198799822, 'recon_motion': 0.0009670565923443064, 'recon_norm': 1.2372588038444519, 'recon_norm_std': 0.06735172979533673, 'target_norm': 1.6225150376558304, 'recon_cos': 0.7038654550909996, 'gradnorm/compressor.query': 0.008567466542124945, 'gradnorm/compressor.layer0': 0.1748531476836926, 'gradnorm/compressor.layer1': 3.92086791220216, 'gradnorm/compressor.layer2': 1.2846900534969814, 'gradnorm/compressor.layer3': 0.7713005522941749, 'gradnorm/compressor.layer4': 0.3058669541200626, 'gradnorm/compressor.layer5': 0.2093967753801481, 'gradnorm/compressor.layer6': 0.06162586087922595, 'gradnorm/compressor.layer7': 0.16008022876658456, 'gradnorm/decoder.layer0': 0.3108107327591841, 'gradnorm/decoder.layer1': 0.2644913126027211, 'gradnorm/decoder.layer2': 0.1236560177532133, 'gradnorm/max_param': 2.76286244392395, 'epoch': 3.49}

 35%|███▍      | 6670/19090 [3:14:59<6:22:13,  1.85s/it]
 35%|███▍      | 6671/19090 [3:15:01<6:14:21,  1.81s/it]
 35%|███▍      | 6672/19090 [3:15:02<5:41:59,  1.65s/it]
 35%|███▍      | 6673/19090 [3:15:04<6:26:10,  1.87s/it]
 35%|███▍      | 6674/19090 [3:15:06<6:03:49,  1.76s/it]
 35%|███▍      | 6675/19090 [3:15:07<5:34:18,  1.62s/it]
 35%|███▍      | 6676/19090 [3:15:10<6:51:27,  1.99s/it]
 35%|███▍      | 6677/19090 [3:15:11<6:08:18,  1.78s/it]
 35%|███▍      | 6678/19090 [3:15:13<5:37:51,  1.63s/it]
 35%|███▍      | 6679/19090 [3:15:15<6:48:15,  1.97s/it]
 35%|███▍      | 6680/19090 [3:15:17<6:06:12,  1.77s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.8721022605895996, 'learning_rate': 4.107844314489878e-05, 'recon_loss': 0.0012673548306338489, 'recon_static': 0.00031406787456944585, 'recon_motion': 0.0009532869764370844, 'recon_norm': 1.2345487207174302, 'recon_norm_std': 0.06771789006888866, 'target_norm': 1.6191787540912628, 'recon_cos': 0.7076524287462235, 'gradnorm/compressor.query': 0.008301389545703782, 'gradnorm/compressor.layer0': 0.17080308706516711, 'gradnorm/compressor.layer1': 3.5334896310989077, 'gradnorm/compressor.layer2': 1.2760992024224518, 'gradnorm/compressor.layer3': 0.7102831669401045, 'gradnorm/compressor.layer4': 0.30524089881441585, 'gradnorm/compressor.layer5': 0.20212301380040312, 'gradnorm/compressor.layer6': 0.06198827420498525, 'gradnorm/compressor.layer7': 0.16818703285203443, 'gradnorm/decoder.layer0': 0.3026729890667867, 'gradnorm/decoder.layer1': 0.269568744760445, 'gradnorm/decoder.layer2': 0.12255591887332497, 'gradnorm/max_param': 2.4896373748779297, 'epoch': 3.5}

 35%|███▍      | 6680/19090 [3:15:17<6:06:12,  1.77s/it]
 35%|███▍      | 6681/19090 [3:15:18<5:36:28,  1.63s/it]
 35%|███▌      | 6682/19090 [3:15:21<6:45:21,  1.96s/it]
 35%|███▌      | 6683/19090 [3:15:22<6:04:23,  1.76s/it]
 35%|███▌      | 6684/19090 [3:15:23<5:35:22,  1.62s/it]
 35%|███▌      | 6685/19090 [3:15:26<6:35:23,  1.91s/it]
 35%|███▌      | 6686/19090 [3:15:27<5:56:28,  1.72s/it]
 35%|███▌      | 6687/19090 [3:15:28<5:29:42,  1.59s/it]
 35%|███▌      | 6688/19090 [3:15:31<6:57:55,  2.02s/it]
 35%|███▌      | 6689/19090 [3:15:33<6:12:44,  1.80s/it]
 35%|███▌      | 6690/19090 [3:15:34<5:41:41,  1.65s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.8632559776306152, 'learning_rate': 4.104341139282776e-05, 'recon_loss': 0.0012646469695027918, 'recon_static': 0.0003053660133446101, 'recon_motion': 0.0009592809888999909, 'recon_norm': 1.2307619363069535, 'recon_norm_std': 0.06608775267377495, 'target_norm': 1.6204070955514909, 'recon_cos': 0.7090490460395813, 'gradnorm/compressor.query': 0.007618412935967267, 'gradnorm/compressor.layer0': 0.14714178701693148, 'gradnorm/compressor.layer1': 3.536378090818444, 'gradnorm/compressor.layer2': 1.2528480984889063, 'gradnorm/compressor.layer3': 0.7021907533672972, 'gradnorm/compressor.layer4': 0.2867948015471106, 'gradnorm/compressor.layer5': 0.19939693261530622, 'gradnorm/compressor.layer6': 0.053937481906430336, 'gradnorm/compressor.layer7': 0.16391611948654888, 'gradnorm/decoder.layer0': 0.3177803015363248, 'gradnorm/decoder.layer1': 0.2584312844826321, 'gradnorm/decoder.layer2': 0.12169149748677814, 'gradnorm/max_param': 2.4921412467956543, 'epoch': 3.5}

 35%|███▌      | 6690/19090 [3:15:34<5:41:41,  1.65s/it]
 35%|███▌      | 6691/19090 [3:15:37<6:58:32,  2.03s/it]
 35%|███▌      | 6692/19090 [3:15:38<6:13:17,  1.81s/it]
 35%|███▌      | 6693/19090 [3:15:39<5:40:46,  1.65s/it]
 35%|███▌      | 6694/19090 [3:15:42<6:56:59,  2.02s/it]
 35%|███▌      | 6695/19090 [3:15:44<6:11:41,  1.80s/it]
 35%|███▌      | 6696/19090 [3:15:45<5:39:53,  1.65s/it]
 35%|███▌      | 6697/19090 [3:15:47<6:33:24,  1.90s/it]
 35%|███▌      | 6698/19090 [3:15:49<5:55:36,  1.72s/it]
 35%|███▌      | 6699/19090 [3:15:50<5:29:47,  1.60s/it]
 35%|███▌      | 6700/19090 [3:15:53<6:27:03,  1.87s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.7665841579437256, 'learning_rate': 4.1008325999435215e-05, 'recon_loss': 0.0012646724964724854, 'recon_static': 0.00030684883313369937, 'recon_motion': 0.0009578236829838715, 'recon_norm': 1.2320026844739913, 'recon_norm_std': 0.0671735399402678, 'target_norm': 1.6208776265382767, 'recon_cos': 0.7094656109809876, 'gradnorm/compressor.query': 0.008153285621280728, 'gradnorm/compressor.layer0': 0.16387406949674577, 'gradnorm/compressor.layer1': 3.423462103403147, 'gradnorm/compressor.layer2': 1.2561238961007575, 'gradnorm/compressor.layer3': 0.7376601821093245, 'gradnorm/compressor.layer4': 0.2577543193018631, 'gradnorm/compressor.layer5': 0.2099464677659572, 'gradnorm/compressor.layer6': 0.0623091133333489, 'gradnorm/compressor.layer7': 0.16181655658169325, 'gradnorm/decoder.layer0': 0.30279985765613804, 'gradnorm/decoder.layer1': 0.26541762110530054, 'gradnorm/decoder.layer2': 0.1242443617622274, 'gradnorm/max_param': 2.4125401973724365, 'epoch': 3.51}

 35%|███▌      | 6700/19090 [3:15:53<6:27:03,  1.87s/it]
 35%|███▌      | 6701/19090 [3:15:54<5:50:47,  1.70s/it]
 35%|███▌      | 6702/19090 [3:15:55<5:25:32,  1.58s/it]
 35%|███▌      | 6703/19090 [3:15:58<6:21:18,  1.85s/it]
 35%|███▌      | 6704/19090 [3:15:59<5:58:59,  1.74s/it]
 35%|███▌      | 6705/19090 [3:16:00<5:30:55,  1.60s/it]
 35%|███▌      | 6706/19090 [3:16:03<6:17:49,  1.83s/it]
 35%|███▌      | 6707/19090 [3:16:05<6:17:37,  1.83s/it]
 35%|███▌      | 6708/19090 [3:16:06<5:43:52,  1.67s/it]
 35%|███▌      | 6709/19090 [3:16:08<6:22:33,  1.85s/it]
 35%|███▌      | 6710/19090 [3:16:10<6:25:47,  1.87s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.254398822784424, 'learning_rate': 4.097318708202954e-05, 'recon_loss': 0.0012703566841082647, 'recon_static': 0.0003186676032782998, 'recon_motion': 0.0009516890902887098, 'recon_norm': 1.2342774212360381, 'recon_norm_std': 0.06905663488432764, 'target_norm': 1.6219746679067613, 'recon_cos': 0.7081574246287345, 'gradnorm/compressor.query': 0.009557270464835201, 'gradnorm/compressor.layer0': 0.1811162393378359, 'gradnorm/compressor.layer1': 3.9322599632888307, 'gradnorm/compressor.layer2': 1.2836785444599816, 'gradnorm/compressor.layer3': 0.7636890067986262, 'gradnorm/compressor.layer4': 0.3025593528568122, 'gradnorm/compressor.layer5': 0.21946921643002937, 'gradnorm/compressor.layer6': 0.06114644585650047, 'gradnorm/compressor.layer7': 0.1762743761659549, 'gradnorm/decoder.layer0': 0.3171950034019795, 'gradnorm/decoder.layer1': 0.28508032480375384, 'gradnorm/decoder.layer2': 0.1299787181618855, 'gradnorm/max_param': 2.770850419998169, 'epoch': 3.51}

 35%|███▌      | 6710/19090 [3:16:10<6:25:47,  1.87s/it]
 35%|███▌      | 6711/19090 [3:16:11<5:49:36,  1.69s/it]
 35%|███▌      | 6712/19090 [3:16:14<6:32:14,  1.90s/it]
 35%|███▌      | 6713/19090 [3:16:15<6:22:22,  1.85s/it]
 35%|███▌      | 6714/19090 [3:16:17<5:47:27,  1.68s/it]
 35%|███▌      | 6715/19090 [3:16:19<6:28:43,  1.88s/it]
 35%|███▌      | 6716/19090 [3:16:21<6:18:16,  1.83s/it]
 35%|███▌      | 6717/19090 [3:16:22<5:44:45,  1.67s/it]
 35%|███▌      | 6718/19090 [3:16:24<6:23:27,  1.86s/it]
 35%|███▌      | 6719/19090 [3:16:26<6:11:52,  1.80s/it]
 35%|███▌      | 6720/19090 [3:16:27<5:40:39,  1.65s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.1845316886901855, 'learning_rate': 4.093799475809808e-05, 'recon_loss': 0.001277108848444186, 'recon_static': 0.00031338306798716075, 'recon_motion': 0.0009637257782742382, 'recon_norm': 1.234385234117508, 'recon_norm_std': 0.06668431591242552, 'target_norm': 1.622243481874466, 'recon_cos': 0.7066182851791382, 'gradnorm/compressor.query': 0.008332958422937709, 'gradnorm/compressor.layer0': 0.17023275153133066, 'gradnorm/compressor.layer1': 3.870322449925076, 'gradnorm/compressor.layer2': 1.275914130033059, 'gradnorm/compressor.layer3': 0.72663095623458, 'gradnorm/compressor.layer4': 0.3025286190626649, 'gradnorm/compressor.layer5': 0.20948593168718102, 'gradnorm/compressor.layer6': 0.05875283936033484, 'gradnorm/compressor.layer7': 0.15876691342350896, 'gradnorm/decoder.layer0': 0.30644295024751517, 'gradnorm/decoder.layer1': 0.2659931016481033, 'gradnorm/decoder.layer2': 0.13109803143899382, 'gradnorm/max_param': 2.7274668216705322, 'epoch': 3.52}

 35%|███▌      | 6720/19090 [3:16:27<5:40:39,  1.65s/it]
 35%|███▌      | 6721/19090 [3:16:30<6:28:49,  1.89s/it]
 35%|███▌      | 6722/19090 [3:16:32<6:33:12,  1.91s/it]
 35%|███▌      | 6723/19090 [3:16:33<5:55:19,  1.72s/it]
 35%|███▌      | 6724/19090 [3:16:35<5:45:23,  1.68s/it]
 35%|███▌      | 6725/19090 [3:16:37<6:17:24,  1.83s/it]
 35%|███▌      | 6726/19090 [3:16:38<5:43:48,  1.67s/it]
 35%|███▌      | 6727/19090 [3:16:40<5:54:07,  1.72s/it]
 35%|███▌      | 6728/19090 [3:16:42<6:18:56,  1.84s/it]
 35%|███▌      | 6729/19090 [3:16:43<5:44:58,  1.67s/it]
 35%|███▌      | 6730/19090 [3:16:46<6:30:43,  1.90s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.174973011016846, 'learning_rate': 4.090274914530677e-05, 'recon_loss': 0.0012694429053226486, 'recon_static': 0.0003069406273425557, 'recon_motion': 0.0009625022983527742, 'recon_norm': 1.228860503435135, 'recon_norm_std': 0.06851515863090754, 'target_norm': 1.6220045506954193, 'recon_cos': 0.7083747893571853, 'gradnorm/compressor.query': 0.008284075138143271, 'gradnorm/compressor.layer0': 0.16841917303628673, 'gradnorm/compressor.layer1': 3.846367541758898, 'gradnorm/compressor.layer2': 1.2995827225250214, 'gradnorm/compressor.layer3': 0.7302227319956642, 'gradnorm/compressor.layer4': 0.31873204938046756, 'gradnorm/compressor.layer5': 0.23019828119957833, 'gradnorm/compressor.layer6': 0.06180039766521848, 'gradnorm/compressor.layer7': 0.16409612492717202, 'gradnorm/decoder.layer0': 0.3200713156789839, 'gradnorm/decoder.layer1': 0.28381842496106163, 'gradnorm/decoder.layer2': 0.13020260241930176, 'gradnorm/max_param': 2.7103681564331055, 'epoch': 3.53}

 35%|███▌      | 6730/19090 [3:16:46<6:30:43,  1.90s/it]
 35%|███▌      | 6731/19090 [3:16:47<6:15:04,  1.82s/it]
 35%|███▌      | 6732/19090 [3:16:49<5:41:54,  1.66s/it]
 35%|███▌      | 6733/19090 [3:16:51<6:28:26,  1.89s/it]
 35%|███▌      | 6734/19090 [3:16:53<6:11:54,  1.81s/it]
 35%|███▌      | 6735/19090 [3:16:54<5:39:46,  1.65s/it]
 35%|███▌      | 6736/19090 [3:16:57<6:35:40,  1.92s/it]
 35%|███▌      | 6737/19090 [3:16:58<6:16:05,  1.83s/it]
 35%|███▌      | 6738/19090 [3:17:00<5:43:08,  1.67s/it]
 35%|███▌      | 6739/19090 [3:17:02<6:08:53,  1.79s/it]
 35%|███▌      | 6740/19090 [3:17:04<6:19:33,  1.84s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.048499584197998, 'learning_rate': 4.08674503614997e-05, 'recon_loss': 0.0012737397715682163, 'recon_static': 0.00030980346564319917, 'recon_motion': 0.0009639363270252943, 'recon_norm': 1.228909793496132, 'recon_norm_std': 0.06811501430347562, 'target_norm': 1.6244001686573029, 'recon_cos': 0.7084287986159324, 'gradnorm/compressor.query': 0.008962867662258456, 'gradnorm/compressor.layer0': 0.17162393526134578, 'gradnorm/compressor.layer1': 3.7147596732812764, 'gradnorm/compressor.layer2': 1.313334699152383, 'gradnorm/compressor.layer3': 0.7176435080505252, 'gradnorm/compressor.layer4': 0.28420139505123154, 'gradnorm/compressor.layer5': 0.2026156586882979, 'gradnorm/compressor.layer6': 0.06027026788964899, 'gradnorm/compressor.layer7': 0.16214651808295347, 'gradnorm/decoder.layer0': 0.2799404572077091, 'gradnorm/decoder.layer1': 0.2730274668332439, 'gradnorm/decoder.layer2': 0.12983174144562062, 'gradnorm/max_param': 2.617804527282715, 'epoch': 3.53}

 35%|███▌      | 6740/19090 [3:17:04<6:19:33,  1.84s/it]
 35%|███▌      | 6741/19090 [3:17:05<5:45:28,  1.68s/it]
 35%|███▌      | 6742/19090 [3:17:07<5:55:40,  1.73s/it]
 35%|███▌      | 6743/19090 [3:17:09<6:27:38,  1.88s/it]
 35%|███▌      | 6744/19090 [3:17:10<5:51:01,  1.71s/it]
 35%|███▌      | 6745/19090 [3:17:12<6:01:09,  1.76s/it]
 35%|███▌      | 6746/19090 [3:17:14<6:24:08,  1.87s/it]
 35%|███▌      | 6747/19090 [3:17:16<5:48:18,  1.69s/it]
 35%|███▌      | 6748/19090 [3:17:17<5:44:53,  1.68s/it]
 35%|███▌      | 6749/19090 [3:17:19<6:22:24,  1.86s/it]
 35%|███▌      | 6750/19090 [3:17:21<5:47:29,  1.69s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 3.891484498977661, 'learning_rate': 4.083209852469874e-05, 'recon_loss': 0.0012746218853862956, 'recon_static': 0.000316768614720786, 'recon_motion': 0.0009578532641171478, 'recon_norm': 1.2341340094804765, 'recon_norm_std': 0.068163519538939, 'target_norm': 1.6220310837030412, 'recon_cos': 0.7071537107229233, 'gradnorm/compressor.query': 0.008749650798954412, 'gradnorm/compressor.layer0': 0.1752349258002623, 'gradnorm/compressor.layer1': 3.5228097699033873, 'gradnorm/compressor.layer2': 1.3215195144009202, 'gradnorm/compressor.layer3': 0.7813851430353915, 'gradnorm/compressor.layer4': 0.30126693452430087, 'gradnorm/compressor.layer5': 0.20914194815569453, 'gradnorm/compressor.layer6': 0.06380348747356165, 'gradnorm/compressor.layer7': 0.17063286991973367, 'gradnorm/decoder.layer0': 0.29474360402199334, 'gradnorm/decoder.layer1': 0.27151166655219056, 'gradnorm/decoder.layer2': 0.13199175351114717, 'gradnorm/max_param': 2.4821324348449707, 'epoch': 3.54}

 35%|███▌      | 6750/19090 [3:17:21<5:47:29,  1.69s/it]
 35%|███▌      | 6751/19090 [3:17:22<5:51:40,  1.71s/it]
 35%|███▌      | 6752/19090 [3:17:24<5:56:00,  1.73s/it]
 35%|███▌      | 6753/19090 [3:17:26<5:28:53,  1.60s/it]
 35%|███▌      | 6754/19090 [3:17:27<5:25:57,  1.59s/it]
 35%|███▌      | 6755/19090 [3:17:29<5:32:21,  1.62s/it]
 35%|███▌      | 6756/19090 [3:17:30<5:12:21,  1.52s/it]
 35%|███▌      | 6757/19090 [3:17:32<5:13:04,  1.52s/it]
 35%|███▌      | 6758/19090 [3:17:34<5:40:11,  1.66s/it]
 35%|███▌      | 6759/19090 [3:17:35<5:18:36,  1.55s/it]
 35%|███▌      | 6760/19090 [3:17:36<5:12:43,  1.52s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.228830337524414, 'learning_rate': 4.0796693753103135e-05, 'recon_loss': 0.0012666949536651374, 'recon_static': 0.00032218158958130516, 'recon_motion': 0.0009445133953704498, 'recon_norm': 1.2365567326545714, 'recon_norm_std': 0.06617365125566721, 'target_norm': 1.6189418971538543, 'recon_cos': 0.7079068452119828, 'gradnorm/compressor.query': 0.00887435392491342, 'gradnorm/compressor.layer0': 0.18294132250241696, 'gradnorm/compressor.layer1': 3.8729155071407972, 'gradnorm/compressor.layer2': 1.374393806346198, 'gradnorm/compressor.layer3': 0.7704024829112454, 'gradnorm/compressor.layer4': 0.3148516031615788, 'gradnorm/compressor.layer5': 0.22983963009002878, 'gradnorm/compressor.layer6': 0.06256093839312257, 'gradnorm/compressor.layer7': 0.17687339586534737, 'gradnorm/decoder.layer0': 0.2958814660168777, 'gradnorm/decoder.layer1': 0.27265566358530546, 'gradnorm/decoder.layer2': 0.13588606118789803, 'gradnorm/max_param': 2.729060649871826, 'epoch': 3.54}

 35%|███▌      | 6760/19090 [3:17:36<5:12:43,  1.52s/it]
 35%|███▌      | 6761/19090 [3:17:38<5:26:35,  1.59s/it]
 35%|███▌      | 6762/19090 [3:17:39<5:08:24,  1.50s/it]
 35%|███▌      | 6763/19090 [3:17:41<5:15:16,  1.53s/it]
 35%|███▌      | 6764/19090 [3:17:43<5:39:25,  1.65s/it]
 35%|███▌      | 6765/19090 [3:17:44<5:16:51,  1.54s/it]
 35%|███▌      | 6766/19090 [3:17:46<5:00:48,  1.46s/it]
 35%|███▌      | 6767/19090 [3:17:48<5:39:20,  1.65s/it]
 35%|███▌      | 6768/19090 [3:17:49<5:16:43,  1.54s/it]
 35%|███▌      | 6769/19090 [3:17:50<5:13:21,  1.53s/it]
 35%|███▌      | 6770/19090 [3:17:52<5:46:35,  1.69s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.395832061767578, 'learning_rate': 4.0761236165089124e-05, 'recon_loss': 0.0012662752880714834, 'recon_static': 0.0003086445729422849, 'recon_motion': 0.0009576307347742841, 'recon_norm': 1.22976753115654, 'recon_norm_std': 0.06737398449331522, 'target_norm': 1.620496666431427, 'recon_cos': 0.7086526364088058, 'gradnorm/compressor.query': 0.008876177573711092, 'gradnorm/compressor.layer0': 0.17974871131875386, 'gradnorm/compressor.layer1': 4.0604611489116955, 'gradnorm/compressor.layer2': 1.3547839542884859, 'gradnorm/compressor.layer3': 0.7664744666612651, 'gradnorm/compressor.layer4': 0.3132644720174739, 'gradnorm/compressor.layer5': 0.229702905540499, 'gradnorm/compressor.layer6': 0.066115165610344, 'gradnorm/compressor.layer7': 0.17309088222747154, 'gradnorm/decoder.layer0': 0.3168086918744798, 'gradnorm/decoder.layer1': 0.2749128729258908, 'gradnorm/decoder.layer2': 0.1396899661757451, 'gradnorm/max_param': 2.8616323471069336, 'epoch': 3.55}

 35%|███▌      | 6770/19090 [3:17:52<5:46:35,  1.69s/it]
 35%|███▌      | 6771/19090 [3:17:54<5:21:36,  1.57s/it]
 35%|███▌      | 6772/19090 [3:17:55<5:25:32,  1.59s/it]
 35%|███▌      | 6773/19090 [3:17:57<5:54:36,  1.73s/it]
 35%|███▌      | 6774/19090 [3:17:59<5:27:16,  1.59s/it]
 35%|███▌      | 6775/19090 [3:18:01<5:43:23,  1.67s/it]
 35%|███▌      | 6776/19090 [3:18:02<5:48:09,  1.70s/it]
 36%|███▌      | 6777/19090 [3:18:04<5:22:38,  1.57s/it]
 36%|███▌      | 6778/19090 [3:18:05<5:39:35,  1.65s/it]
 36%|███▌      | 6779/19090 [3:18:07<5:33:49,  1.63s/it]
 36%|███▌      | 6780/19090 [3:18:08<5:12:50,  1.52s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.59299373626709, 'learning_rate': 4.072572587920955e-05, 'recon_loss': 0.0012649695883737877, 'recon_static': 0.00030978813956608067, 'recon_motion': 0.000955181433528196, 'recon_norm': 1.2302493929862977, 'recon_norm_std': 0.06618432635441422, 'target_norm': 1.6195453584194184, 'recon_cos': 0.708708643913269, 'gradnorm/compressor.query': 0.008689772359898502, 'gradnorm/compressor.layer0': 0.18003393032235382, 'gradnorm/compressor.layer1': 4.2419348091752855, 'gradnorm/compressor.layer2': 1.4146765961425898, 'gradnorm/compressor.layer3': 0.8011092839930781, 'gradnorm/compressor.layer4': 0.3132047336201168, 'gradnorm/compressor.layer5': 0.2318250221311669, 'gradnorm/compressor.layer6': 0.07037696267587573, 'gradnorm/compressor.layer7': 0.17970497626858642, 'gradnorm/decoder.layer0': 0.35478165002923134, 'gradnorm/decoder.layer1': 0.3024604912052165, 'gradnorm/decoder.layer2': 0.13987133936741, 'gradnorm/max_param': 2.989198684692383, 'epoch': 3.55}

 36%|███▌      | 6780/19090 [3:18:08<5:12:50,  1.52s/it]
 36%|███▌      | 6781/19090 [3:18:10<5:32:34,  1.62s/it]
 36%|███▌      | 6782/19090 [3:18:12<5:27:25,  1.60s/it]
 36%|███▌      | 6783/19090 [3:18:13<5:08:32,  1.50s/it]
 36%|███▌      | 6784/19090 [3:18:15<5:18:49,  1.55s/it]
 36%|███▌      | 6785/19090 [3:18:16<5:28:19,  1.60s/it]
 36%|███▌      | 6786/19090 [3:18:18<5:08:56,  1.51s/it]
 36%|███▌      | 6787/19090 [3:18:19<5:15:10,  1.54s/it]
 36%|███▌      | 6788/19090 [3:18:21<5:23:33,  1.58s/it]
 36%|███▌      | 6789/19090 [3:18:22<5:05:45,  1.49s/it]
 36%|███▌      | 6790/19090 [3:18:24<5:08:13,  1.50s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.501053333282471, 'learning_rate': 4.069016301419345e-05, 'recon_loss': 0.0012616662861546502, 'recon_static': 0.0003028122300747782, 'recon_motion': 0.00095885409245966, 'recon_norm': 1.225072506070137, 'recon_norm_std': 0.06627809489145875, 'target_norm': 1.6203514486551285, 'recon_cos': 0.709872841835022, 'gradnorm/compressor.query': 0.008457413112369926, 'gradnorm/compressor.layer0': 0.1629015914280897, 'gradnorm/compressor.layer1': 4.149244556838757, 'gradnorm/compressor.layer2': 1.3821300395809182, 'gradnorm/compressor.layer3': 0.8370997112818238, 'gradnorm/compressor.layer4': 0.3205236227632989, 'gradnorm/compressor.layer5': 0.23241862927903337, 'gradnorm/compressor.layer6': 0.06189481160674274, 'gradnorm/compressor.layer7': 0.1794085098992059, 'gradnorm/decoder.layer0': 0.32963396695395303, 'gradnorm/decoder.layer1': 0.2944868302177398, 'gradnorm/decoder.layer2': 0.13216176917755204, 'gradnorm/max_param': 2.9245760440826416, 'epoch': 3.56}

 36%|███▌      | 6790/19090 [3:18:24<5:08:13,  1.50s/it]
 36%|███▌      | 6791/19090 [3:18:25<5:25:17,  1.59s/it]
 36%|███▌      | 6792/19090 [3:18:27<5:07:00,  1.50s/it]
 36%|███▌      | 6793/19090 [3:18:28<5:14:51,  1.54s/it]
 36%|███▌      | 6794/19090 [3:18:30<5:15:16,  1.54s/it]
 36%|███▌      | 6795/19090 [3:18:31<5:00:06,  1.46s/it]
 36%|███▌      | 6796/19090 [3:18:33<5:06:39,  1.50s/it]
 36%|███▌      | 6797/19090 [3:18:35<5:18:42,  1.56s/it]
 36%|███▌      | 6798/19090 [3:18:36<5:02:33,  1.48s/it]
 36%|███▌      | 6799/19090 [3:18:37<5:01:27,  1.47s/it]
 36%|███▌      | 6800/19090 [3:18:39<5:12:11,  1.52s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.340192794799805, 'learning_rate': 4.0654547688945636e-05, 'recon_loss': 0.0012556783825857566, 'recon_static': 0.00029799258336424825, 'recon_motion': 0.0009576858530635946, 'recon_norm': 1.2236096650362014, 'recon_norm_std': 0.06636869087815285, 'target_norm': 1.619611892104149, 'recon_cos': 0.711157776415348, 'gradnorm/compressor.query': 0.008029053791167983, 'gradnorm/compressor.layer0': 0.16244519162213675, 'gradnorm/compressor.layer1': 3.9914318347468574, 'gradnorm/compressor.layer2': 1.3777438443288132, 'gradnorm/compressor.layer3': 0.7711624826024129, 'gradnorm/compressor.layer4': 0.3235065612223405, 'gradnorm/compressor.layer5': 0.2324751384194222, 'gradnorm/compressor.layer6': 0.06542768404156532, 'gradnorm/compressor.layer7': 0.17624036383834144, 'gradnorm/decoder.layer0': 0.31139709637209384, 'gradnorm/decoder.layer1': 0.2792962481084871, 'gradnorm/decoder.layer2': 0.13192989392178872, 'gradnorm/max_param': 2.8130788803100586, 'epoch': 3.56}

 36%|███▌      | 6800/19090 [3:18:39<5:12:11,  1.52s/it]
 36%|███▌      | 6801/19090 [3:18:40<4:58:35,  1.46s/it]
 36%|███▌      | 6802/19090 [3:18:42<5:09:35,  1.51s/it]
 36%|███▌      | 6803/19090 [3:18:44<5:30:20,  1.61s/it]
 36%|███▌      | 6804/19090 [3:18:45<5:10:07,  1.51s/it]
 36%|███▌      | 6805/19090 [3:18:47<5:10:06,  1.51s/it]
 36%|███▌      | 6806/19090 [3:18:48<5:15:41,  1.54s/it]
 36%|███▌      | 6807/19090 [3:18:49<4:59:38,  1.46s/it]
 36%|███▌      | 6808/19090 [3:18:51<5:34:39,  1.63s/it]
 36%|███▌      | 6809/19090 [3:18:53<5:25:35,  1.59s/it]
 36%|███▌      | 6810/19090 [3:18:54<5:06:54,  1.50s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.235954284667969, 'learning_rate': 4.061888002254634e-05, 'recon_loss': 0.0012549521721666678, 'recon_static': 0.00029637688494403845, 'recon_motion': 0.0009585753097780981, 'recon_norm': 1.2235503226518631, 'recon_norm_std': 0.06513256011530758, 'target_norm': 1.6208552777767182, 'recon_cos': 0.7119329854846, 'gradnorm/compressor.query': 0.007803239027254034, 'gradnorm/compressor.layer0': 0.16264730274059833, 'gradnorm/compressor.layer1': 3.9073775228194885, 'gradnorm/compressor.layer2': 1.2937051571800642, 'gradnorm/compressor.layer3': 0.7787555927669725, 'gradnorm/compressor.layer4': 0.3141897057796744, 'gradnorm/compressor.layer5': 0.21667685064505, 'gradnorm/compressor.layer6': 0.0668368055665158, 'gradnorm/compressor.layer7': 0.17280194815106234, 'gradnorm/decoder.layer0': 0.3031506537367315, 'gradnorm/decoder.layer1': 0.282943409185804, 'gradnorm/decoder.layer2': 0.13079894506730091, 'gradnorm/max_param': 2.753481149673462, 'epoch': 3.57}

 36%|███▌      | 6810/19090 [3:18:54<5:06:54,  1.50s/it]
 36%|███▌      | 6811/19090 [3:18:56<5:34:49,  1.64s/it]
 36%|███▌      | 6812/19090 [3:18:57<5:13:36,  1.53s/it]
 36%|███▌      | 6813/19090 [3:18:59<4:58:39,  1.46s/it]
 36%|███▌      | 6814/19090 [3:19:01<5:40:01,  1.66s/it]
 36%|███▌      | 6815/19090 [3:19:02<5:16:55,  1.55s/it]
 36%|███▌      | 6816/19090 [3:19:03<5:00:52,  1.47s/it]
 36%|███▌      | 6817/19090 [3:19:05<5:29:58,  1.61s/it]
 36%|███▌      | 6818/19090 [3:19:07<5:09:45,  1.51s/it]
 36%|███▌      | 6819/19090 [3:19:08<4:55:47,  1.45s/it]
 36%|███▌      | 6820/19090 [3:19:10<5:27:07,  1.60s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.31824254989624, 'learning_rate': 4.0583160134250804e-05, 'recon_loss': 0.0012492713227402418, 'recon_static': 0.00029952086333651096, 'recon_motion': 0.000949750434665475, 'recon_norm': 1.2232482582330704, 'recon_norm_std': 0.06746391346678138, 'target_norm': 1.6186587423086167, 'recon_cos': 0.7123110085725785, 'gradnorm/compressor.query': 0.008254413442525156, 'gradnorm/compressor.layer0': 0.16187680463447746, 'gradnorm/compressor.layer1': 3.970234270462326, 'gradnorm/compressor.layer2': 1.3755451768856082, 'gradnorm/compressor.layer3': 0.7797533576993328, 'gradnorm/compressor.layer4': 0.31048213835801347, 'gradnorm/compressor.layer5': 0.21053177172502896, 'gradnorm/compressor.layer6': 0.060464362479997494, 'gradnorm/compressor.layer7': 0.17080811429735604, 'gradnorm/decoder.layer0': 0.2961830558065107, 'gradnorm/decoder.layer1': 0.2760924061740884, 'gradnorm/decoder.layer2': 0.1433190403359638, 'gradnorm/max_param': 2.7982280254364014, 'epoch': 3.57}

 36%|███▌      | 6820/19090 [3:19:10<5:27:07,  1.60s/it]
 36%|███▌      | 6821/19090 [3:19:11<5:07:52,  1.51s/it]
 36%|███▌      | 6822/19090 [3:19:12<4:54:58,  1.44s/it]
 36%|███▌      | 6823/19090 [3:19:14<5:18:51,  1.56s/it]
 36%|███▌      | 6824/19090 [3:19:16<5:02:36,  1.48s/it]
 36%|███▌      | 6825/19090 [3:19:17<4:51:05,  1.42s/it]
 36%|███▌      | 6826/19090 [3:19:19<5:27:27,  1.60s/it]
 36%|███▌      | 6827/19090 [3:19:20<5:09:24,  1.51s/it]
 36%|███▌      | 6828/19090 [3:19:22<4:58:10,  1.46s/it]
 36%|███▌      | 6829/19090 [3:19:23<5:21:42,  1.57s/it]
 36%|███▌      | 6830/19090 [3:19:25<5:05:35,  1.50s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.810819625854492, 'learning_rate': 4.0547388143488856e-05, 'recon_loss': 0.0012623584392713384, 'recon_static': 0.0002983037826197688, 'recon_motion': 0.0009640546617447398, 'recon_norm': 1.2245917290449142, 'recon_norm_std': 0.0640983410179615, 'target_norm': 1.6209599792957305, 'recon_cos': 0.7098692446947098, 'gradnorm/compressor.query': 0.008907358957723882, 'gradnorm/compressor.layer0': 0.1855665999518828, 'gradnorm/compressor.layer1': 4.450555213838581, 'gradnorm/compressor.layer2': 1.4835078868958458, 'gradnorm/compressor.layer3': 0.8457513315323946, 'gradnorm/compressor.layer4': 0.3188225238236066, 'gradnorm/compressor.layer5': 0.23363993931219879, 'gradnorm/compressor.layer6': 0.06803819613617468, 'gradnorm/compressor.layer7': 0.17250860533984083, 'gradnorm/decoder.layer0': 0.3059004822995018, 'gradnorm/decoder.layer1': 0.28476433407002943, 'gradnorm/decoder.layer2': 0.1437029818681073, 'gradnorm/max_param': 3.1367666721343994, 'epoch': 3.58}

 36%|███▌      | 6830/19090 [3:19:25<5:05:35,  1.50s/it]
 36%|███▌      | 6831/19090 [3:19:26<4:54:23,  1.44s/it]
 36%|███▌      | 6832/19090 [3:19:28<5:41:10,  1.67s/it]
 36%|███▌      | 6833/19090 [3:19:30<5:18:37,  1.56s/it]
 36%|███▌      | 6834/19090 [3:19:31<5:03:07,  1.48s/it]
 36%|███▌      | 6835/19090 [3:19:33<5:37:22,  1.65s/it]
 36%|███▌      | 6836/19090 [3:19:34<5:16:09,  1.55s/it]
 36%|███▌      | 6837/19090 [3:19:36<5:01:02,  1.47s/it]
 36%|███▌      | 6838/19090 [3:19:37<5:14:02,  1.54s/it]
 36%|███▌      | 6839/19090 [3:19:38<4:59:41,  1.47s/it]
 36%|███▌      | 6840/19090 [3:19:40<4:49:59,  1.42s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.803824424743652, 'learning_rate': 4.051156416986454e-05, 'recon_loss': 0.0012679718987783418, 'recon_static': 0.0003012279703398235, 'recon_motion': 0.0009667438600445167, 'recon_norm': 1.2235478788614274, 'recon_norm_std': 0.06680876472964883, 'target_norm': 1.623776689171791, 'recon_cos': 0.7096615791320801, 'gradnorm/compressor.query': 0.009390278888416816, 'gradnorm/compressor.layer0': 0.19153045011251596, 'gradnorm/compressor.layer1': 4.444911301782398, 'gradnorm/compressor.layer2': 1.4657824140982645, 'gradnorm/compressor.layer3': 0.8370089927791846, 'gradnorm/compressor.layer4': 0.3339918369998883, 'gradnorm/compressor.layer5': 0.25502851859669085, 'gradnorm/compressor.layer6': 0.06924647953904886, 'gradnorm/compressor.layer7': 0.1801790583449488, 'gradnorm/decoder.layer0': 0.3405036001239187, 'gradnorm/decoder.layer1': 0.29039867982117673, 'gradnorm/decoder.layer2': 0.1396347119181842, 'gradnorm/max_param': 3.1327550411224365, 'epoch': 3.58}

 36%|███▌      | 6840/19090 [3:19:40<4:49:59,  1.42s/it]
 36%|███▌      | 6841/19090 [3:19:42<5:45:58,  1.69s/it]
 36%|███▌      | 6842/19090 [3:19:43<5:22:03,  1.58s/it]
 36%|███▌      | 6843/19090 [3:19:45<5:05:11,  1.50s/it]
 36%|███▌      | 6844/19090 [3:19:47<5:50:49,  1.72s/it]
 36%|███▌      | 6845/19090 [3:19:48<5:24:48,  1.59s/it]
 36%|███▌      | 6846/19090 [3:19:50<5:07:28,  1.51s/it]
 36%|███▌      | 6847/19090 [3:19:52<5:33:10,  1.63s/it]
 36%|███▌      | 6848/19090 [3:19:53<5:12:58,  1.53s/it]
 36%|███▌      | 6849/19090 [3:19:54<4:58:42,  1.46s/it]
 36%|███▌      | 6850/19090 [3:19:56<5:40:57,  1.67s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.5804362297058105, 'learning_rate': 4.0475688333155714e-05, 'recon_loss': 0.0012872995226643979, 'recon_static': 0.000311519540991867, 'recon_motion': 0.0009757799271028489, 'recon_norm': 1.22509663105011, 'recon_norm_std': 0.06479777311906218, 'target_norm': 1.6245606005191804, 'recon_cos': 0.7046807765960693, 'gradnorm/compressor.query': 0.010247751997526165, 'gradnorm/compressor.layer0': 0.20373402589946088, 'gradnorm/compressor.layer1': 4.224087877399818, 'gradnorm/compressor.layer2': 1.4208991925738974, 'gradnorm/compressor.layer3': 0.7908903799863211, 'gradnorm/compressor.layer4': 0.34663819615474134, 'gradnorm/compressor.layer5': 0.25397327712145196, 'gradnorm/compressor.layer6': 0.06924297588557657, 'gradnorm/compressor.layer7': 0.17957596841949577, 'gradnorm/decoder.layer0': 0.3466521346197758, 'gradnorm/decoder.layer1': 0.29951618491003096, 'gradnorm/decoder.layer2': 0.14078091217025884, 'gradnorm/max_param': 2.976464033126831, 'epoch': 3.59}

 36%|███▌      | 6850/19090 [3:19:56<5:40:57,  1.67s/it]
 36%|███▌      | 6851/19090 [3:19:58<5:18:33,  1.56s/it]
 36%|███▌      | 6852/19090 [3:19:59<5:03:10,  1.49s/it]
 36%|███▌      | 6853/19090 [3:20:01<5:38:23,  1.66s/it]
 36%|███▌      | 6854/19090 [3:20:02<5:16:46,  1.55s/it]
 36%|███▌      | 6855/19090 [3:20:04<5:01:40,  1.48s/it]
 36%|███▌      | 6856/19090 [3:20:05<5:28:02,  1.61s/it]
 36%|███▌      | 6857/19090 [3:20:07<5:09:11,  1.52s/it]
 36%|███▌      | 6858/19090 [3:20:08<4:56:16,  1.45s/it]
 36%|███▌      | 6859/19090 [3:20:10<5:43:55,  1.69s/it]
 36%|███▌      | 6860/19090 [3:20:12<5:20:40,  1.57s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.773075103759766, 'learning_rate': 4.043976075331361e-05, 'recon_loss': 0.0012755931413266808, 'recon_static': 0.0003153894336719532, 'recon_motion': 0.0009602036938304081, 'recon_norm': 1.229366382956505, 'recon_norm_std': 0.06606640703976155, 'target_norm': 1.6207797169685363, 'recon_cos': 0.7061884686350822, 'gradnorm/compressor.query': 0.009462438651068935, 'gradnorm/compressor.layer0': 0.19318039635081477, 'gradnorm/compressor.layer1': 4.417775548729269, 'gradnorm/compressor.layer2': 1.4482389772859312, 'gradnorm/compressor.layer3': 0.8328614369054789, 'gradnorm/compressor.layer4': 0.34039168197681363, 'gradnorm/compressor.layer5': 0.23803325943401032, 'gradnorm/compressor.layer6': 0.06895695796014656, 'gradnorm/compressor.layer7': 0.16954439810997673, 'gradnorm/decoder.layer0': 0.3419704214649753, 'gradnorm/decoder.layer1': 0.30725016421429285, 'gradnorm/decoder.layer2': 0.14025101337177254, 'gradnorm/max_param': 3.1134395599365234, 'epoch': 3.59}

 36%|███▌      | 6860/19090 [3:20:12<5:20:40,  1.57s/it]
 36%|███▌      | 6861/19090 [3:20:13<5:04:19,  1.49s/it]
 36%|███▌      | 6862/19090 [3:20:15<5:34:35,  1.64s/it]
 36%|███▌      | 6863/19090 [3:20:16<5:12:53,  1.54s/it]
 36%|███▌      | 6864/19090 [3:20:18<4:58:43,  1.47s/it]
 36%|███▌      | 6865/19090 [3:20:19<5:28:25,  1.61s/it]
 36%|███▌      | 6866/19090 [3:20:21<5:09:16,  1.52s/it]
 36%|███▌      | 6867/19090 [3:20:22<4:56:41,  1.46s/it]
 36%|███▌      | 6868/19090 [3:20:24<5:48:03,  1.71s/it]
 36%|███▌      | 6869/19090 [3:20:26<5:23:16,  1.59s/it]
 36%|███▌      | 6870/19090 [3:20:27<5:06:13,  1.50s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.186460494995117, 'learning_rate': 4.040378155046251e-05, 'recon_loss': 0.0012712303228909149, 'recon_static': 0.0003070776292588562, 'recon_motion': 0.0009641526863561012, 'recon_norm': 1.2271197319030762, 'recon_norm_std': 0.06533350860700012, 'target_norm': 1.6207405239343644, 'recon_cos': 0.707130677998066, 'gradnorm/compressor.query': 0.010183834307298236, 'gradnorm/compressor.layer0': 0.21293420083177636, 'gradnorm/compressor.layer1': 4.794158663483032, 'gradnorm/compressor.layer2': 1.594751006276037, 'gradnorm/compressor.layer3': 0.9484719058085449, 'gradnorm/compressor.layer4': 0.32341501277575074, 'gradnorm/compressor.layer5': 0.23566480818898755, 'gradnorm/compressor.layer6': 0.07305166259416732, 'gradnorm/compressor.layer7': 0.1832309852189611, 'gradnorm/decoder.layer0': 0.3391997384815119, 'gradnorm/decoder.layer1': 0.30439183977764966, 'gradnorm/decoder.layer2': 0.1427815175411918, 'gradnorm/max_param': 3.3786699771881104, 'epoch': 3.6}

 36%|███▌      | 6870/19090 [3:20:27<5:06:13,  1.50s/it]
 36%|███▌      | 6871/19090 [3:20:29<5:33:03,  1.64s/it]
 36%|███▌      | 6872/19090 [3:20:30<5:11:26,  1.53s/it]
 36%|███▌      | 6873/19090 [3:20:32<4:57:02,  1.46s/it]
 36%|███▌      | 6874/19090 [3:20:34<5:37:32,  1.66s/it]
 36%|███▌      | 6875/19090 [3:20:35<5:15:07,  1.55s/it]
 36%|███▌      | 6876/19090 [3:20:36<4:59:15,  1.47s/it]
 36%|███▌      | 6877/19090 [3:20:38<5:36:25,  1.65s/it]
 36%|███▌      | 6878/19090 [3:20:40<5:14:21,  1.54s/it]
 36%|███▌      | 6879/19090 [3:20:41<4:58:34,  1.47s/it]
 36%|███▌      | 6880/19090 [3:20:43<5:56:24,  1.75s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.358037948608398, 'learning_rate': 4.036775084489926e-05, 'recon_loss': 0.0012536615860881284, 'recon_static': 0.0003001310265972279, 'recon_motion': 0.0009535306046018377, 'recon_norm': 1.2226537972688676, 'recon_norm_std': 0.06543904906138778, 'target_norm': 1.6193748652935027, 'recon_cos': 0.7115452721714973, 'gradnorm/compressor.query': 0.009424159218352985, 'gradnorm/compressor.layer0': 0.19286845724500887, 'gradnorm/compressor.layer1': 4.974527514847182, 'gradnorm/compressor.layer2': 1.6328086305417293, 'gradnorm/compressor.layer3': 0.8919010056444767, 'gradnorm/compressor.layer4': 0.354495970650655, 'gradnorm/compressor.layer5': 0.2587747533908893, 'gradnorm/compressor.layer6': 0.06789933104383057, 'gradnorm/compressor.layer7': 0.18664302287870738, 'gradnorm/decoder.layer0': 0.3441899583596082, 'gradnorm/decoder.layer1': 0.30413969630156684, 'gradnorm/decoder.layer2': 0.14407416183743188, 'gradnorm/max_param': 3.506492853164673, 'epoch': 3.6}

 36%|███▌      | 6880/19090 [3:20:43<5:56:24,  1.75s/it]
 36%|███▌      | 6881/19090 [3:20:45<5:28:27,  1.61s/it]
 36%|███▌      | 6882/19090 [3:20:46<5:08:14,  1.51s/it]
 36%|███▌      | 6883/19090 [3:20:48<6:14:02,  1.84s/it]
 36%|███▌      | 6884/19090 [3:20:50<5:40:20,  1.67s/it]
 36%|███▌      | 6885/19090 [3:20:51<5:17:04,  1.56s/it]
 36%|███▌      | 6886/19090 [3:20:53<6:11:57,  1.83s/it]
 36%|███▌      | 6887/19090 [3:20:55<5:39:30,  1.67s/it]
 36%|███▌      | 6888/19090 [3:20:56<5:16:20,  1.56s/it]
 36%|███▌      | 6889/19090 [3:20:59<6:13:57,  1.84s/it]
 36%|███▌      | 6890/19090 [3:21:00<5:41:50,  1.68s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.006300926208496, 'learning_rate': 4.033166875709291e-05, 'recon_loss': 0.0012666009657550602, 'recon_static': 0.00031750730631756595, 'recon_motion': 0.0009490936528891325, 'recon_norm': 1.2306132644414902, 'recon_norm_std': 0.06415901863947511, 'target_norm': 1.6185277312994004, 'recon_cos': 0.7077364772558212, 'gradnorm/compressor.query': 0.012480695517028989, 'gradnorm/compressor.layer0': 0.241241036209042, 'gradnorm/compressor.layer1': 4.652620680451716, 'gradnorm/compressor.layer2': 1.4962879981089576, 'gradnorm/compressor.layer3': 0.8318249083927797, 'gradnorm/compressor.layer4': 0.3345043974802602, 'gradnorm/compressor.layer5': 0.23629081053173165, 'gradnorm/compressor.layer6': 0.07560052004075032, 'gradnorm/compressor.layer7': 0.19735534440606672, 'gradnorm/decoder.layer0': 0.3291891869957499, 'gradnorm/decoder.layer1': 0.29171950707442085, 'gradnorm/decoder.layer2': 0.14542615122341895, 'gradnorm/max_param': 3.2786660194396973, 'epoch': 3.61}

 36%|███▌      | 6890/19090 [3:21:00<5:41:50,  1.68s/it]
 36%|███▌      | 6891/19090 [3:21:01<5:18:41,  1.57s/it]
 36%|███▌      | 6892/19090 [3:21:04<6:08:17,  1.81s/it]
 36%|███▌      | 6893/19090 [3:21:05<5:36:40,  1.66s/it]
 36%|███▌      | 6894/19090 [3:21:06<5:14:26,  1.55s/it]
 36%|███▌      | 6895/19090 [3:21:09<6:22:38,  1.88s/it]
 36%|███▌      | 6896/19090 [3:21:10<5:46:39,  1.71s/it]
 36%|███▌      | 6897/19090 [3:21:11<5:21:52,  1.58s/it]
 36%|███▌      | 6898/19090 [3:21:14<6:13:54,  1.84s/it]
 36%|███▌      | 6899/19090 [3:21:15<5:40:25,  1.68s/it]
 36%|███▌      | 6900/19090 [3:21:16<5:17:09,  1.56s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.112444877624512, 'learning_rate': 4.0295535407684326e-05, 'recon_loss': 0.0012937238934682682, 'recon_static': 0.00033024828153429555, 'recon_motion': 0.0009634756221203134, 'recon_norm': 1.2344850331544877, 'recon_norm_std': 0.0642257109284401, 'target_norm': 1.6222201615571976, 'recon_cos': 0.7020276784896851, 'gradnorm/compressor.query': 0.010994389675734703, 'gradnorm/compressor.layer0': 0.2209492382994206, 'gradnorm/compressor.layer1': 4.75840357223664, 'gradnorm/compressor.layer2': 1.4985363495077115, 'gradnorm/compressor.layer3': 0.8691270805760826, 'gradnorm/compressor.layer4': 0.35202441070455304, 'gradnorm/compressor.layer5': 0.2251403255112516, 'gradnorm/compressor.layer6': 0.07396816313703776, 'gradnorm/compressor.layer7': 0.18696517962860007, 'gradnorm/decoder.layer0': 0.3499122777918582, 'gradnorm/decoder.layer1': 0.29431315626896315, 'gradnorm/decoder.layer2': 0.1437363869503121, 'gradnorm/max_param': 3.3531861305236816, 'epoch': 3.61}

 36%|███▌      | 6900/19090 [3:21:16<5:17:09,  1.56s/it]
 36%|███▌      | 6901/19090 [3:21:19<6:48:08,  2.01s/it]
 36%|███▌      | 6902/19090 [3:21:21<6:04:06,  1.79s/it]
 36%|███▌      | 6903/19090 [3:21:22<5:33:02,  1.64s/it]
 36%|███▌      | 6904/19090 [3:21:25<6:33:47,  1.94s/it]
 36%|███▌      | 6905/19090 [3:21:26<5:54:38,  1.75s/it]
 36%|███▌      | 6906/19090 [3:21:27<5:27:07,  1.61s/it]
 36%|███▌      | 6907/19090 [3:21:30<6:59:45,  2.07s/it]
 36%|███▌      | 6908/19090 [3:21:32<6:12:26,  1.83s/it]
 36%|███▌      | 6909/19090 [3:21:33<5:39:24,  1.67s/it]
 36%|███▌      | 6910/19090 [3:21:36<6:48:42,  2.01s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.444936275482178, 'learning_rate': 4.025935091748575e-05, 'recon_loss': 0.0012712397234281524, 'recon_static': 0.00030680813069920986, 'recon_motion': 0.0009644316276535391, 'recon_norm': 1.2240004032850265, 'recon_norm_std': 0.06482473369687795, 'target_norm': 1.6209104299545287, 'recon_cos': 0.7074527695775032, 'gradnorm/compressor.query': 0.01361704265912217, 'gradnorm/compressor.layer0': 0.26578514096272965, 'gradnorm/compressor.layer1': 5.05867085637422, 'gradnorm/compressor.layer2': 1.6066671757413657, 'gradnorm/compressor.layer3': 0.9590625614527947, 'gradnorm/compressor.layer4': 0.36038402214757764, 'gradnorm/compressor.layer5': 0.2597451739757823, 'gradnorm/compressor.layer6': 0.07320261522011773, 'gradnorm/compressor.layer7': 0.19767090739308185, 'gradnorm/decoder.layer0': 0.3729386097651931, 'gradnorm/decoder.layer1': 0.28765994111551885, 'gradnorm/decoder.layer2': 0.14683338549710812, 'gradnorm/max_param': 3.564955234527588, 'epoch': 3.62}

 36%|███▌      | 6910/19090 [3:21:36<6:48:42,  2.01s/it]
 36%|███▌      | 6911/19090 [3:21:37<6:05:37,  1.80s/it]
 36%|███▌      | 6912/19090 [3:21:38<5:34:50,  1.65s/it]
 36%|███▌      | 6913/19090 [3:21:41<6:38:51,  1.97s/it]
 36%|███▌      | 6914/19090 [3:21:42<5:57:35,  1.76s/it]
 36%|███▌      | 6915/19090 [3:21:44<5:28:41,  1.62s/it]
 36%|███▌      | 6916/19090 [3:21:46<6:35:43,  1.95s/it]
 36%|███▌      | 6917/19090 [3:21:48<5:55:18,  1.75s/it]
 36%|███▌      | 6918/19090 [3:21:49<5:27:26,  1.61s/it]
 36%|███▌      | 6919/19090 [3:21:52<6:30:14,  1.92s/it]
 36%|███▌      | 6920/19090 [3:21:53<5:51:31,  1.73s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.241596221923828, 'learning_rate': 4.0223115407480424e-05, 'recon_loss': 0.001275451364926994, 'recon_static': 0.0003109930134087335, 'recon_motion': 0.0009644583318731747, 'recon_norm': 1.2238499224185944, 'recon_norm_std': 0.06609209747985005, 'target_norm': 1.6214014559984207, 'recon_cos': 0.7064878985285759, 'gradnorm/compressor.query': 0.009005387353383422, 'gradnorm/compressor.layer0': 0.20235178574324825, 'gradnorm/compressor.layer1': 4.881767661648733, 'gradnorm/compressor.layer2': 1.5251220779334675, 'gradnorm/compressor.layer3': 0.9024767426746093, 'gradnorm/compressor.layer4': 0.3576650782237609, 'gradnorm/compressor.layer5': 0.26239191850197907, 'gradnorm/compressor.layer6': 0.07500074344008255, 'gradnorm/compressor.layer7': 0.19636736204869207, 'gradnorm/decoder.layer0': 0.35349286253763496, 'gradnorm/decoder.layer1': 0.2705636225117754, 'gradnorm/decoder.layer2': 0.14854333266928585, 'gradnorm/max_param': 3.4408535957336426, 'epoch': 3.62}

 36%|███▌      | 6920/19090 [3:21:53<5:51:31,  1.73s/it]
 36%|███▋      | 6921/19090 [3:21:54<5:24:36,  1.60s/it]
 36%|███▋      | 6922/19090 [3:21:57<6:33:38,  1.94s/it]
 36%|███▋      | 6923/19090 [3:21:58<5:54:02,  1.75s/it]
 36%|███▋      | 6924/19090 [3:22:00<5:26:49,  1.61s/it]
 36%|███▋      | 6925/19090 [3:22:02<6:40:18,  1.97s/it]
 36%|███▋      | 6926/19090 [3:22:04<5:58:33,  1.77s/it]
 36%|███▋      | 6927/19090 [3:22:05<5:29:08,  1.62s/it]
 36%|███▋      | 6928/19090 [3:22:08<6:41:48,  1.98s/it]
 36%|███▋      | 6929/19090 [3:22:09<5:59:36,  1.77s/it]
 36%|███▋      | 6930/19090 [3:22:10<5:29:56,  1.63s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.5305280685424805, 'learning_rate': 4.018682899882218e-05, 'recon_loss': 0.0012619160494068637, 'recon_static': 0.0003065317519940436, 'recon_motion': 0.0009553843134199269, 'recon_norm': 1.2231545567512512, 'recon_norm_std': 0.06413696529343724, 'target_norm': 1.6189999610185624, 'recon_cos': 0.7092951714992524, 'gradnorm/compressor.query': 0.012005771690245174, 'gradnorm/compressor.layer0': 0.240231436424581, 'gradnorm/compressor.layer1': 5.151059399128456, 'gradnorm/compressor.layer2': 1.633646044995546, 'gradnorm/compressor.layer3': 0.8940354018253134, 'gradnorm/compressor.layer4': 0.38345681677352633, 'gradnorm/compressor.layer5': 0.2666910880549001, 'gradnorm/compressor.layer6': 0.07932161405666435, 'gradnorm/compressor.layer7': 0.2049734211619298, 'gradnorm/decoder.layer0': 0.3733016736721408, 'gradnorm/decoder.layer1': 0.3088198690471318, 'gradnorm/decoder.layer2': 0.16179187544977786, 'gradnorm/max_param': 3.6305606365203857, 'epoch': 3.63}

 36%|███▋      | 6930/19090 [3:22:10<5:29:56,  1.63s/it]
 36%|███▋      | 6931/19090 [3:22:13<6:36:08,  1.95s/it]
 36%|███▋      | 6932/19090 [3:22:14<5:55:46,  1.76s/it]
 36%|███▋      | 6933/19090 [3:22:16<5:27:21,  1.62s/it]
 36%|███▋      | 6934/19090 [3:22:18<6:30:12,  1.93s/it]
 36%|███▋      | 6935/19090 [3:22:20<5:52:08,  1.74s/it]
 36%|███▋      | 6936/19090 [3:22:21<5:26:15,  1.61s/it]
 36%|███▋      | 6937/19090 [3:22:24<6:33:32,  1.94s/it]
 36%|███▋      | 6938/19090 [3:22:25<5:53:44,  1.75s/it]
 36%|███▋      | 6939/19090 [3:22:26<5:26:00,  1.61s/it]
 36%|███▋      | 6940/19090 [3:22:29<6:39:05,  1.97s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 4.939999103546143, 'learning_rate': 4.0150491812835025e-05, 'recon_loss': 0.0012720008904580026, 'recon_static': 0.00030206867013475857, 'recon_motion': 0.0009699322225060314, 'recon_norm': 1.2235381573438644, 'recon_norm_std': 0.06327783297747373, 'target_norm': 1.6221182763576507, 'recon_cos': 0.7078786164522171, 'gradnorm/compressor.query': 0.0086424136550254, 'gradnorm/compressor.layer0': 0.1871544216300458, 'gradnorm/compressor.layer1': 4.550153356629605, 'gradnorm/compressor.layer2': 1.5244893673094047, 'gradnorm/compressor.layer3': 0.9394636252722541, 'gradnorm/compressor.layer4': 0.33959400383541455, 'gradnorm/compressor.layer5': 0.23893348354004215, 'gradnorm/compressor.layer6': 0.07577797634388306, 'gradnorm/compressor.layer7': 0.20301915125168796, 'gradnorm/decoder.layer0': 0.3534375204563557, 'gradnorm/decoder.layer1': 0.30488941642456396, 'gradnorm/decoder.layer2': 0.14383288435181676, 'gradnorm/max_param': 3.2067291736602783, 'epoch': 3.64}

 36%|███▋      | 6940/19090 [3:22:29<6:39:05,  1.97s/it]
 36%|███▋      | 6941/19090 [3:22:30<5:57:43,  1.77s/it]
 36%|███▋      | 6942/19090 [3:22:32<5:28:41,  1.62s/it]
 36%|███▋      | 6943/19090 [3:22:34<6:45:30,  2.00s/it]
 36%|███▋      | 6944/19090 [3:22:36<6:02:24,  1.79s/it]
 36%|███▋      | 6945/19090 [3:22:37<5:31:44,  1.64s/it]
 36%|███▋      | 6946/19090 [3:22:40<6:56:11,  2.06s/it]
 36%|███▋      | 6947/19090 [3:22:41<6:09:26,  1.83s/it]
 36%|███▋      | 6948/19090 [3:22:43<5:37:04,  1.67s/it]
 36%|███▋      | 6949/19090 [3:22:45<6:34:34,  1.95s/it]
 36%|███▋      | 6950/19090 [3:22:47<5:54:55,  1.75s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.394214630126953, 'learning_rate': 4.01141039710127e-05, 'recon_loss': 0.0012543016084237025, 'recon_static': 0.00030665185986435974, 'recon_motion': 0.0009476497449213639, 'recon_norm': 1.2215564966201782, 'recon_norm_std': 0.06690160483121872, 'target_norm': 1.6194453686475754, 'recon_cos': 0.7113643288612366, 'gradnorm/compressor.query': 0.009708599021177801, 'gradnorm/compressor.layer0': 0.20396501819661947, 'gradnorm/compressor.layer1': 4.9968648618174205, 'gradnorm/compressor.layer2': 1.6587622680440368, 'gradnorm/compressor.layer3': 0.9467961042413471, 'gradnorm/compressor.layer4': 0.3285426932669408, 'gradnorm/compressor.layer5': 0.2528663909538448, 'gradnorm/compressor.layer6': 0.07118214467195053, 'gradnorm/compressor.layer7': 0.1905199766078963, 'gradnorm/decoder.layer0': 0.3401720719168902, 'gradnorm/decoder.layer1': 0.29659301767609514, 'gradnorm/decoder.layer2': 0.1494895021562735, 'gradnorm/max_param': 3.522184133529663, 'epoch': 3.64}

 36%|███▋      | 6950/19090 [3:22:47<5:54:55,  1.75s/it]
 36%|███▋      | 6951/19090 [3:22:48<5:27:04,  1.62s/it]
 36%|███▋      | 6952/19090 [3:22:50<6:26:09,  1.91s/it]
 36%|███▋      | 6953/19090 [3:22:52<5:48:51,  1.72s/it]
 36%|███▋      | 6954/19090 [3:22:53<5:22:10,  1.59s/it]
 36%|███▋      | 6955/19090 [3:22:56<6:55:09,  2.05s/it]
 36%|███▋      | 6956/19090 [3:22:57<6:08:46,  1.82s/it]
 36%|███▋      | 6957/19090 [3:22:59<5:36:08,  1.66s/it]
 36%|███▋      | 6958/19090 [3:23:01<6:30:00,  1.93s/it]
 36%|███▋      | 6959/19090 [3:23:03<5:51:13,  1.74s/it]
 36%|███▋      | 6960/19090 [3:23:04<5:24:08,  1.60s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.718081951141357, 'learning_rate': 4.007766559501841e-05, 'recon_loss': 0.0012587043427629396, 'recon_static': 0.00029984103675815275, 'recon_motion': 0.0009588633285602555, 'recon_norm': 1.2195977866649628, 'recon_norm_std': 0.06599765401333571, 'target_norm': 1.6209611415863037, 'recon_cos': 0.7108075678348541, 'gradnorm/compressor.query': 0.009677343011168954, 'gradnorm/compressor.layer0': 0.19761284400583762, 'gradnorm/compressor.layer1': 5.369897949272007, 'gradnorm/compressor.layer2': 1.570188602854815, 'gradnorm/compressor.layer3': 0.9241374427456187, 'gradnorm/compressor.layer4': 0.36118899579389446, 'gradnorm/compressor.layer5': 0.265137533674867, 'gradnorm/compressor.layer6': 0.07729495898575413, 'gradnorm/compressor.layer7': 0.1954752911205834, 'gradnorm/decoder.layer0': 0.35271600525464786, 'gradnorm/decoder.layer1': 0.33478730709051996, 'gradnorm/decoder.layer2': 0.14338721672805302, 'gradnorm/max_param': 3.785470485687256, 'epoch': 3.65}

 36%|███▋      | 6960/19090 [3:23:04<5:24:08,  1.60s/it]
 36%|███▋      | 6961/19090 [3:23:06<6:10:13,  1.83s/it]
 36%|███▋      | 6962/19090 [3:23:08<5:37:30,  1.67s/it]
 36%|███▋      | 6963/19090 [3:23:09<5:14:02,  1.55s/it]
 36%|███▋      | 6964/19090 [3:23:12<6:29:52,  1.93s/it]
 36%|███▋      | 6965/19090 [3:23:13<5:50:43,  1.74s/it]
 36%|███▋      | 6966/19090 [3:23:14<5:24:16,  1.60s/it]
 36%|███▋      | 6967/19090 [3:23:17<6:35:31,  1.96s/it]
 37%|███▋      | 6968/19090 [3:23:18<5:54:39,  1.76s/it]
 37%|███▋      | 6969/19090 [3:23:20<5:26:32,  1.62s/it]
 37%|███▋      | 6970/19090 [3:23:22<6:16:34,  1.86s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.400325298309326, 'learning_rate': 4.004117680668422e-05, 'recon_loss': 0.0012616842112038284, 'recon_static': 0.00030728503552381883, 'recon_motion': 0.000954399217152968, 'recon_norm': 1.221276804804802, 'recon_norm_std': 0.06543649826198816, 'target_norm': 1.6201071083545684, 'recon_cos': 0.7094798654317855, 'gradnorm/compressor.query': 0.012393350842015129, 'gradnorm/compressor.layer0': 0.23364829412182903, 'gradnorm/compressor.layer1': 5.032572886303489, 'gradnorm/compressor.layer2': 1.5384227371824786, 'gradnorm/compressor.layer3': 0.9532572545081416, 'gradnorm/compressor.layer4': 0.3640648788570894, 'gradnorm/compressor.layer5': 0.2560172140137918, 'gradnorm/compressor.layer6': 0.07344293739094751, 'gradnorm/compressor.layer7': 0.20274011203241107, 'gradnorm/decoder.layer0': 0.3578869238335023, 'gradnorm/decoder.layer1': 0.33096016178580934, 'gradnorm/decoder.layer2': 0.15597104563176817, 'gradnorm/max_param': 3.546984910964966, 'epoch': 3.65}

 37%|███▋      | 6970/19090 [3:23:22<6:16:34,  1.86s/it]
 37%|███▋      | 6971/19090 [3:23:23<5:41:21,  1.69s/it]
 37%|███▋      | 6972/19090 [3:23:25<5:16:42,  1.57s/it]
 37%|███▋      | 6973/19090 [3:23:27<6:28:23,  1.92s/it]
 37%|███▋      | 6974/19090 [3:23:29<5:50:01,  1.73s/it]
 37%|███▋      | 6975/19090 [3:23:30<5:23:21,  1.60s/it]
 37%|███▋      | 6976/19090 [3:23:33<6:51:11,  2.04s/it]
 37%|███▋      | 6977/19090 [3:23:34<6:05:29,  1.81s/it]
 37%|███▋      | 6978/19090 [3:23:36<5:33:25,  1.65s/it]
 37%|███▋      | 6979/19090 [3:23:38<6:43:48,  2.00s/it]
 37%|███▋      | 6980/19090 [3:23:40<6:00:24,  1.79s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.501504898071289, 'learning_rate': 4.0004637728010806e-05, 'recon_loss': 0.001270356131135486, 'recon_static': 0.0003181219624821097, 'recon_motion': 0.0009522341497358866, 'recon_norm': 1.228438663482666, 'recon_norm_std': 0.06545690177008509, 'target_norm': 1.62188960313797, 'recon_cos': 0.7079965218901634, 'gradnorm/compressor.query': 0.011382070324790633, 'gradnorm/compressor.layer0': 0.23249317325803642, 'gradnorm/compressor.layer1': 5.103902423272712, 'gradnorm/compressor.layer2': 1.6130118895658325, 'gradnorm/compressor.layer3': 1.0099195588059422, 'gradnorm/compressor.layer4': 0.369847443452998, 'gradnorm/compressor.layer5': 0.26671782253754434, 'gradnorm/compressor.layer6': 0.08337026450758138, 'gradnorm/compressor.layer7': 0.20342451717771343, 'gradnorm/decoder.layer0': 0.39660116645960786, 'gradnorm/decoder.layer1': 0.3206655951969945, 'gradnorm/decoder.layer2': 0.15631608161689514, 'gradnorm/max_param': 3.5975611209869385, 'epoch': 3.66}

 37%|███▋      | 6980/19090 [3:23:40<6:00:24,  1.79s/it]
 37%|███▋      | 6981/19090 [3:23:41<5:30:10,  1.64s/it]
 37%|███▋      | 6982/19090 [3:23:44<6:27:55,  1.92s/it]
 37%|███▋      | 6983/19090 [3:23:45<5:49:15,  1.73s/it]
 37%|███▋      | 6984/19090 [3:23:46<5:22:11,  1.60s/it]
 37%|███▋      | 6985/19090 [3:23:49<6:54:54,  2.06s/it]
 37%|███▋      | 6986/19090 [3:23:50<6:08:16,  1.83s/it]
 37%|███▋      | 6987/19090 [3:23:52<5:35:22,  1.66s/it]
 37%|███▋      | 6988/19090 [3:23:54<6:37:14,  1.97s/it]
 37%|███▋      | 6989/19090 [3:23:56<5:55:47,  1.76s/it]
 37%|███▋      | 6990/19090 [3:23:57<5:26:55,  1.62s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.837891101837158, 'learning_rate': 3.9968048481166985e-05, 'recon_loss': 0.0012706696114037185, 'recon_static': 0.000314317498850869, 'recon_motion': 0.0009563521394738928, 'recon_norm': 1.225853717327118, 'recon_norm_std': 0.06481361500918865, 'target_norm': 1.6221619427204133, 'recon_cos': 0.708212585747242, 'gradnorm/compressor.query': 0.011273210759916235, 'gradnorm/compressor.layer0': 0.23323665582309971, 'gradnorm/compressor.layer1': 5.428018129446385, 'gradnorm/compressor.layer2': 1.7317292249074154, 'gradnorm/compressor.layer3': 1.0262301250954, 'gradnorm/compressor.layer4': 0.3790195138494854, 'gradnorm/compressor.layer5': 0.2553373792285646, 'gradnorm/compressor.layer6': 0.07824438926638418, 'gradnorm/compressor.layer7': 0.20615717070407039, 'gradnorm/decoder.layer0': 0.3725074742994643, 'gradnorm/decoder.layer1': 0.29570474359631216, 'gradnorm/decoder.layer2': 0.1653728101340992, 'gradnorm/max_param': 3.826052188873291, 'epoch': 3.66}

 37%|███▋      | 6990/19090 [3:23:57<5:26:55,  1.62s/it]
 37%|███▋      | 6991/19090 [3:24:00<6:33:37,  1.95s/it]
 37%|███▋      | 6992/19090 [3:24:01<5:53:09,  1.75s/it]
 37%|███▋      | 6993/19090 [3:24:02<5:24:35,  1.61s/it]
 37%|███▋      | 6994/19090 [3:24:05<6:40:32,  1.99s/it]
 37%|███▋      | 6995/19090 [3:24:06<5:57:43,  1.77s/it]
 37%|███▋      | 6996/19090 [3:24:08<5:28:15,  1.63s/it]
 37%|███▋      | 6997/19090 [3:24:11<6:39:29,  1.98s/it]
 37%|███▋      | 6998/19090 [3:24:12<5:57:25,  1.77s/it]
 37%|███▋      | 6999/19090 [3:24:13<5:27:54,  1.63s/it]
 37%|███▋      | 7000/19090 [3:24:16<6:35:12,  1.96s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.733008861541748, 'learning_rate': 3.99314091884893e-05, 'recon_loss': 0.0012592308718012645, 'recon_static': 0.0003049060396733694, 'recon_motion': 0.0009543248394038528, 'recon_norm': 1.2220224857330322, 'recon_norm_std': 0.06500478759407997, 'target_norm': 1.6194973468780518, 'recon_cos': 0.7102438122034073, 'gradnorm/compressor.query': 0.009938871436639452, 'gradnorm/compressor.layer0': 0.21559571537239478, 'gradnorm/compressor.layer1': 5.3755352729588886, 'gradnorm/compressor.layer2': 1.5972299916543748, 'gradnorm/compressor.layer3': 0.9202555316591946, 'gradnorm/compressor.layer4': 0.38414708878217024, 'gradnorm/compressor.layer5': 0.2759986145107421, 'gradnorm/compressor.layer6': 0.08142922514698472, 'gradnorm/compressor.layer7': 0.20837025116610539, 'gradnorm/decoder.layer0': 0.3688005162244254, 'gradnorm/decoder.layer1': 0.3075033310190299, 'gradnorm/decoder.layer2': 0.14850672563260195, 'gradnorm/max_param': 3.7893054485321045, 'epoch': 3.67}

 37%|███▋      | 7000/19090 [3:24:16<6:35:12,  1.96s/it]
 37%|███▋      | 7001/19090 [3:24:20<8:20:23,  2.48s/it]
 37%|███▋      | 7002/19090 [3:24:21<7:07:48,  2.12s/it]
 37%|███▋      | 7003/19090 [3:24:22<6:17:42,  1.87s/it]
 37%|███▋      | 7004/19090 [3:24:23<5:42:30,  1.70s/it]
 37%|███▋      | 7005/19090 [3:24:25<5:17:32,  1.58s/it]
 37%|███▋      | 7006/19090 [3:24:26<5:00:10,  1.49s/it]
 37%|███▋      | 7007/19090 [3:24:27<4:47:57,  1.43s/it]
 37%|███▋      | 7008/19090 [3:24:29<4:39:26,  1.39s/it]
 37%|███▋      | 7009/19090 [3:24:30<4:33:39,  1.36s/it]
 37%|███▋      | 7010/19090 [3:24:31<4:29:19,  1.34s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 6.088304042816162, 'learning_rate': 3.989471997248161e-05, 'recon_loss': 0.0012632919853786007, 'recon_static': 0.0003059246446355246, 'recon_motion': 0.0009573673829436302, 'recon_norm': 1.220778015255928, 'recon_norm_std': 0.06409209240227938, 'target_norm': 1.6212760776281356, 'recon_cos': 0.7098101809620857, 'gradnorm/compressor.query': 0.01077804892164521, 'gradnorm/compressor.layer0': 0.21683282811803356, 'gradnorm/compressor.layer1': 5.748695557285284, 'gradnorm/compressor.layer2': 1.593396615137714, 'gradnorm/compressor.layer3': 0.9439557009219445, 'gradnorm/compressor.layer4': 0.38886313788366494, 'gradnorm/compressor.layer5': 0.2911486737562551, 'gradnorm/compressor.layer6': 0.08174928957917733, 'gradnorm/compressor.layer7': 0.2087601769273704, 'gradnorm/decoder.layer0': 0.36593342756811764, 'gradnorm/decoder.layer1': 0.3117593452005837, 'gradnorm/decoder.layer2': 0.15971538294671, 'gradnorm/max_param': 4.0526347160339355, 'epoch': 3.67}

 37%|███▋      | 7010/19090 [3:24:31<4:29:19,  1.34s/it]
 37%|███▋      | 7011/19090 [3:24:32<4:26:25,  1.32s/it]
 37%|███▋      | 7012/19090 [3:24:34<4:24:31,  1.31s/it]
 37%|███▋      | 7013/19090 [3:24:35<4:22:57,  1.31s/it]
 37%|███▋      | 7014/19090 [3:24:37<4:33:23,  1.36s/it]
 37%|███▋      | 7015/19090 [3:24:39<5:25:49,  1.62s/it]
 37%|███▋      | 7016/19090 [3:24:40<5:05:59,  1.52s/it]
 37%|███▋      | 7017/19090 [3:24:42<5:18:52,  1.58s/it]
 37%|███▋      | 7018/19090 [3:24:44<6:11:43,  1.85s/it]
 37%|███▋      | 7019/19090 [3:24:46<5:38:02,  1.68s/it]
 37%|███▋      | 7020/19090 [3:24:47<5:45:11,  1.72s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 6.174981594085693, 'learning_rate': 3.985798095581474e-05, 'recon_loss': 0.0012519094481831417, 'recon_static': 0.00029860787399229596, 'recon_motion': 0.0009533015923807398, 'recon_norm': 1.2182048171758653, 'recon_norm_std': 0.06557018756866455, 'target_norm': 1.6201200634241104, 'recon_cos': 0.7122370839118958, 'gradnorm/compressor.query': 0.011497278597442102, 'gradnorm/compressor.layer0': 0.21669075920574052, 'gradnorm/compressor.layer1': 5.774485326514633, 'gradnorm/compressor.layer2': 1.755632600153344, 'gradnorm/compressor.layer3': 1.0626180246660673, 'gradnorm/compressor.layer4': 0.3920481506450963, 'gradnorm/compressor.layer5': 0.2583877907287459, 'gradnorm/compressor.layer6': 0.07991094991399195, 'gradnorm/compressor.layer7': 0.21505427882714548, 'gradnorm/decoder.layer0': 0.357794187318701, 'gradnorm/decoder.layer1': 0.3199397802873396, 'gradnorm/decoder.layer2': 0.1541093580162227, 'gradnorm/max_param': 4.070819854736328, 'epoch': 3.68}

 37%|███▋      | 7020/19090 [3:24:47<5:45:11,  1.72s/it]
 37%|███▋      | 7021/19090 [3:24:49<6:08:22,  1.83s/it]
 37%|███▋      | 7022/19090 [3:24:51<5:35:40,  1.67s/it]
 37%|███▋      | 7023/19090 [3:24:53<5:53:49,  1.76s/it]
 37%|███▋      | 7024/19090 [3:24:55<6:06:24,  1.82s/it]
 37%|███▋      | 7025/19090 [3:24:56<5:34:18,  1.66s/it]
 37%|███▋      | 7026/19090 [3:24:58<5:56:15,  1.77s/it]
 37%|███▋      | 7027/19090 [3:25:00<6:03:15,  1.81s/it]
 37%|███▋      | 7028/19090 [3:25:01<5:33:13,  1.66s/it]
 37%|███▋      | 7029/19090 [3:25:03<6:09:56,  1.84s/it]
 37%|███▋      | 7030/19090 [3:25:05<5:50:01,  1.74s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.7633867263793945, 'learning_rate': 3.982119226132598e-05, 'recon_loss': 0.0012589676131028681, 'recon_static': 0.0002983842212415766, 'recon_motion': 0.0009605833911336958, 'recon_norm': 1.2146056234836577, 'recon_norm_std': 0.0632914842106402, 'target_norm': 1.6202531307935715, 'recon_cos': 0.7105061769485473, 'gradnorm/compressor.query': 0.010159021194929193, 'gradnorm/compressor.layer0': 0.20842584503051026, 'gradnorm/compressor.layer1': 5.32714834798923, 'gradnorm/compressor.layer2': 1.7991613686174226, 'gradnorm/compressor.layer3': 1.0137078016084966, 'gradnorm/compressor.layer4': 0.3639895394690246, 'gradnorm/compressor.layer5': 0.2649393066817999, 'gradnorm/compressor.layer6': 0.07888081147624604, 'gradnorm/compressor.layer7': 0.21459466192902485, 'gradnorm/decoder.layer0': 0.3979943185537343, 'gradnorm/decoder.layer1': 0.30383305685641426, 'gradnorm/decoder.layer2': 0.15594270409861713, 'gradnorm/max_param': 3.755120038986206, 'epoch': 3.68}

 37%|███▋      | 7030/19090 [3:25:05<5:50:01,  1.74s/it]
 37%|███▋      | 7031/19090 [3:25:07<5:38:49,  1.69s/it]
 37%|███▋      | 7032/19090 [3:25:09<6:15:37,  1.87s/it]
 37%|███▋      | 7033/19090 [3:25:10<6:04:46,  1.82s/it]
 37%|███▋      | 7034/19090 [3:25:12<5:35:26,  1.67s/it]
 37%|███▋      | 7035/19090 [3:25:14<6:16:24,  1.87s/it]
 37%|███▋      | 7036/19090 [3:25:16<5:58:33,  1.78s/it]
 37%|███▋      | 7037/19090 [3:25:17<5:41:59,  1.70s/it]
 37%|███▋      | 7038/19090 [3:25:19<6:04:22,  1.81s/it]
 37%|███▋      | 7039/19090 [3:25:21<6:08:00,  1.83s/it]
 37%|███▋      | 7040/19090 [3:25:23<5:37:14,  1.68s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 6.476541519165039, 'learning_rate': 3.978435401201872e-05, 'recon_loss': 0.0012742533639539034, 'recon_static': 0.000305781918723369, 'recon_motion': 0.0009684714445029385, 'recon_norm': 1.2205131202936172, 'recon_norm_std': 0.06265878649428487, 'target_norm': 1.6217803329229354, 'recon_cos': 0.7070857405662536, 'gradnorm/compressor.query': 0.012656477414511931, 'gradnorm/compressor.layer0': 0.25446104911667533, 'gradnorm/compressor.layer1': 6.034804171566809, 'gradnorm/compressor.layer2': 1.9501481220035328, 'gradnorm/compressor.layer3': 1.0237868782825625, 'gradnorm/compressor.layer4': 0.3802952035749417, 'gradnorm/compressor.layer5': 0.3022479548728228, 'gradnorm/compressor.layer6': 0.08720633035529779, 'gradnorm/compressor.layer7': 0.212387794785595, 'gradnorm/decoder.layer0': 0.420186247884328, 'gradnorm/decoder.layer1': 0.3473546294903631, 'gradnorm/decoder.layer2': 0.15717433275166395, 'gradnorm/max_param': 4.253770351409912, 'epoch': 3.69}

 37%|███▋      | 7040/19090 [3:25:23<5:37:14,  1.68s/it]
 37%|███▋      | 7041/19090 [3:25:25<6:10:10,  1.84s/it]
 37%|███▋      | 7042/19090 [3:25:27<6:06:43,  1.83s/it]
 37%|███▋      | 7043/19090 [3:25:28<5:34:15,  1.66s/it]
 37%|███▋      | 7044/19090 [3:25:30<6:27:07,  1.93s/it]
 37%|███▋      | 7045/19090 [3:25:32<5:56:08,  1.77s/it]
 37%|███▋      | 7046/19090 [3:25:33<5:46:04,  1.72s/it]
 37%|███▋      | 7047/19090 [3:25:36<6:24:53,  1.92s/it]
 37%|███▋      | 7048/19090 [3:25:37<5:57:31,  1.78s/it]
 37%|███▋      | 7049/19090 [3:25:39<5:49:33,  1.74s/it]
 37%|███▋      | 7050/19090 [3:25:42<6:54:58,  2.07s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 5.850460529327393, 'learning_rate': 3.9747466331062074e-05, 'recon_loss': 0.0012620612105820328, 'recon_static': 0.00030663773795822633, 'recon_motion': 0.0009554234842653386, 'recon_norm': 1.22004791200161, 'recon_norm_std': 0.06345647452399135, 'target_norm': 1.6198074579238892, 'recon_cos': 0.7095330521464348, 'gradnorm/compressor.query': 0.010384255850674576, 'gradnorm/compressor.layer0': 0.21816110532117344, 'gradnorm/compressor.layer1': 5.478185916437903, 'gradnorm/compressor.layer2': 1.643610825829353, 'gradnorm/compressor.layer3': 0.9417970408151133, 'gradnorm/compressor.layer4': 0.39806054553215287, 'gradnorm/compressor.layer5': 0.2783189070816695, 'gradnorm/compressor.layer6': 0.0871805194526519, 'gradnorm/compressor.layer7': 0.2184301149159514, 'gradnorm/decoder.layer0': 0.3882116343609105, 'gradnorm/decoder.layer1': 0.338574109601448, 'gradnorm/decoder.layer2': 0.1570367971076317, 'gradnorm/max_param': 3.861459493637085, 'epoch': 3.69}

 37%|███▋      | 7050/19090 [3:25:42<6:54:58,  2.07s/it]
 37%|███▋      | 7051/19090 [3:25:43<6:08:42,  1.84s/it]
 37%|███▋      | 7052/19090 [3:25:44<5:35:34,  1.67s/it]
 37%|███▋      | 7053/19090 [3:25:47<6:19:40,  1.89s/it]
 37%|███▋      | 7054/19090 [3:25:48<5:43:59,  1.71s/it]
 37%|███▋      | 7055/19090 [3:25:50<5:36:45,  1.68s/it]
 37%|███▋      | 7056/19090 [3:25:52<6:18:28,  1.89s/it]
 37%|███▋      | 7057/19090 [3:25:53<5:50:04,  1.75s/it]
 37%|███▋      | 7058/19090 [3:25:55<5:41:04,  1.70s/it]
 37%|███▋      | 7059/19090 [3:25:57<6:21:49,  1.90s/it]
 37%|███▋      | 7060/19090 [3:25:59<6:01:27,  1.80s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 6.522822380065918, 'learning_rate': 3.971052934179038e-05, 'recon_loss': 0.0012754281022353097, 'recon_static': 0.0003016333415871486, 'recon_motion': 0.0009737947591929696, 'recon_norm': 1.2170757681131363, 'recon_norm_std': 0.06345610963180662, 'target_norm': 1.624184674024582, 'recon_cos': 0.7079664915800095, 'gradnorm/compressor.query': 0.010674830677667209, 'gradnorm/compressor.layer0': 0.2308304744387344, 'gradnorm/compressor.layer1': 6.093645456835611, 'gradnorm/compressor.layer2': 1.8587245925572655, 'gradnorm/compressor.layer3': 1.1172050528047714, 'gradnorm/compressor.layer4': 0.4426844665227371, 'gradnorm/compressor.layer5': 0.2998958039137435, 'gradnorm/compressor.layer6': 0.09134052350717496, 'gradnorm/compressor.layer7': 0.22964470036400872, 'gradnorm/decoder.layer0': 0.4174865543747329, 'gradnorm/decoder.layer1': 0.3336885715059438, 'gradnorm/decoder.layer2': 0.16019252622125837, 'gradnorm/max_param': 4.295967102050781, 'epoch': 3.7}

 37%|███▋      | 7060/19090 [3:25:59<6:01:27,  1.80s/it]
 37%|███▋      | 7061/19090 [3:26:00<5:31:27,  1.65s/it]
 37%|███▋      | 7062/19090 [3:26:03<6:18:34,  1.89s/it]
 37%|███▋      | 7063/19090 [3:26:05<6:29:18,  1.94s/it]
 37%|███▋      | 7064/19090 [3:26:06<5:51:06,  1.75s/it]
 37%|███▋      | 7065/19090 [3:26:08<6:10:16,  1.85s/it]
 37%|███▋      | 7066/19090 [3:26:10<6:07:26,  1.83s/it]
 37%|███▋      | 7067/19090 [3:26:11<5:34:37,  1.67s/it]
 37%|███▋      | 7068/19090 [3:26:14<6:15:40,  1.87s/it]
 37%|███▋      | 7069/19090 [3:26:15<6:18:52,  1.89s/it]
 37%|███▋      | 7070/19090 [3:26:17<5:42:25,  1.71s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 6.949941158294678, 'learning_rate': 3.967354316770289e-05, 'recon_loss': 0.001264266439829953, 'recon_static': 0.000309105304768309, 'recon_motion': 0.0009551611103233882, 'recon_norm': 1.220872402191162, 'recon_norm_std': 0.06481155781075358, 'target_norm': 1.61990966796875, 'recon_cos': 0.7087617173790932, 'gradnorm/compressor.query': 0.012012069200815323, 'gradnorm/compressor.layer0': 0.25913274723714486, 'gradnorm/compressor.layer1': 6.486682159677151, 'gradnorm/compressor.layer2': 2.059524007451252, 'gradnorm/compressor.layer3': 1.1332539191658992, 'gradnorm/compressor.layer4': 0.41679452682734597, 'gradnorm/compressor.layer5': 0.285845967531178, 'gradnorm/compressor.layer6': 0.09124734425599304, 'gradnorm/compressor.layer7': 0.23014748536544277, 'gradnorm/decoder.layer0': 0.4345986346327378, 'gradnorm/decoder.layer1': 0.3201155056772513, 'gradnorm/decoder.layer2': 0.15286787666940985, 'gradnorm/max_param': 4.572545528411865, 'epoch': 3.7}

 37%|███▋      | 7070/19090 [3:26:17<5:42:25,  1.71s/it]
 37%|███▋      | 7071/19090 [3:26:19<6:08:16,  1.84s/it]
 37%|███▋      | 7072/19090 [3:26:20<5:42:50,  1.71s/it]
 37%|███▋      | 7073/19090 [3:26:22<5:56:08,  1.78s/it]
 37%|███▋      | 7074/19090 [3:26:24<6:13:52,  1.87s/it]
 37%|███▋      | 7075/19090 [3:26:26<5:49:36,  1.75s/it]
 37%|███▋      | 7076/19090 [3:26:27<5:45:03,  1.72s/it]
 37%|███▋      | 7077/19090 [3:26:30<6:11:46,  1.86s/it]
 37%|███▋      | 7078/19090 [3:26:31<5:45:57,  1.73s/it]
 37%|███▋      | 7079/19090 [3:26:33<5:32:17,  1.66s/it]
 37%|███▋      | 7080/19090 [3:26:35<6:19:19,  1.90s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 6.403879642486572, 'learning_rate': 3.963650793246327e-05, 'recon_loss': 0.0012539320421637967, 'recon_static': 0.0002988096748595126, 'recon_motion': 0.000955122361483518, 'recon_norm': 1.217062348127365, 'recon_norm_std': 0.06371061746031045, 'target_norm': 1.6198443084955216, 'recon_cos': 0.7118471875786782, 'gradnorm/compressor.query': 0.010580599434097887, 'gradnorm/compressor.layer0': 0.24269047161132862, 'gradnorm/compressor.layer1': 6.003660039376664, 'gradnorm/compressor.layer2': 1.8119938900151933, 'gradnorm/compressor.layer3': 1.037035463504758, 'gradnorm/compressor.layer4': 0.3860980360224721, 'gradnorm/compressor.layer5': 0.283838992048614, 'gradnorm/compressor.layer6': 0.0862752282004211, 'gradnorm/compressor.layer7': 0.2032025354296892, 'gradnorm/decoder.layer0': 0.38583645212067813, 'gradnorm/decoder.layer1': 0.3087389881857584, 'gradnorm/decoder.layer2': 0.15939401461734068, 'gradnorm/max_param': 4.232261657714844, 'epoch': 3.71}

 37%|███▋      | 7080/19090 [3:26:35<6:19:19,  1.90s/it]
 37%|███▋      | 7081/19090 [3:26:36<5:53:23,  1.77s/it]
 37%|███▋      | 7082/19090 [3:26:38<5:25:03,  1.62s/it]
 37%|███▋      | 7083/19090 [3:26:40<6:25:06,  1.92s/it]
 37%|███▋      | 7084/19090 [3:26:42<6:02:45,  1.81s/it]
 37%|███▋      | 7085/19090 [3:26:43<5:31:26,  1.66s/it]
 37%|███▋      | 7086/19090 [3:26:46<6:14:26,  1.87s/it]
 37%|███▋      | 7087/19090 [3:26:47<6:06:39,  1.83s/it]
 37%|███▋      | 7088/19090 [3:26:49<5:34:11,  1.67s/it]
 37%|███▋      | 7089/19090 [3:26:51<6:27:47,  1.94s/it]
 37%|███▋      | 7090/19090 [3:26:53<5:48:59,  1.74s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 6.110903263092041, 'learning_rate': 3.959942375989924e-05, 'recon_loss': 0.001253673163591884, 'recon_static': 0.00030571840397897176, 'recon_motion': 0.0009479547457885929, 'recon_norm': 1.219147226214409, 'recon_norm_std': 0.0640158912166953, 'target_norm': 1.6173605024814606, 'recon_cos': 0.7103898450732231, 'gradnorm/compressor.query': 0.011189774455445401, 'gradnorm/compressor.layer0': 0.24030040205774236, 'gradnorm/compressor.layer1': 5.679086351654708, 'gradnorm/compressor.layer2': 1.832839754882471, 'gradnorm/compressor.layer3': 1.0377931801072524, 'gradnorm/compressor.layer4': 0.40257867417389664, 'gradnorm/compressor.layer5': 0.31274238961151724, 'gradnorm/compressor.layer6': 0.0870361996900561, 'gradnorm/compressor.layer7': 0.21012315028577497, 'gradnorm/decoder.layer0': 0.3989960551377611, 'gradnorm/decoder.layer1': 0.3140077457808969, 'gradnorm/decoder.layer2': 0.16588512789423154, 'gradnorm/max_param': 4.0027666091918945, 'epoch': 3.71}

 37%|███▋      | 7090/19090 [3:26:53<5:48:59,  1.74s/it]
 37%|███▋      | 7091/19090 [3:26:54<5:21:47,  1.61s/it]
 37%|███▋      | 7092/19090 [3:26:57<6:43:21,  2.02s/it]
 37%|███▋      | 7093/19090 [3:26:58<6:02:06,  1.81s/it]
 37%|███▋      | 7094/19090 [3:26:59<5:30:57,  1.66s/it]
 37%|███▋      | 7095/19090 [3:27:02<6:46:30,  2.03s/it]
 37%|███▋      | 7096/19090 [3:27:04<6:01:58,  1.81s/it]
 37%|███▋      | 7097/19090 [3:27:05<5:30:31,  1.65s/it]
 37%|███▋      | 7098/19090 [3:27:07<6:16:04,  1.88s/it]
 37%|███▋      | 7099/19090 [3:27:09<6:00:21,  1.80s/it]
 37%|███▋      | 7100/19090 [3:27:10<5:29:39,  1.65s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 6.431605815887451, 'learning_rate': 3.956229077400213e-05, 'recon_loss': 0.0012791881541488691, 'recon_static': 0.00030495495375362227, 'recon_motion': 0.0009742331749293953, 'recon_norm': 1.2175042480230331, 'recon_norm_std': 0.0633232107385993, 'target_norm': 1.625583615899086, 'recon_cos': 0.707247456908226, 'gradnorm/compressor.query': 0.011598324659027435, 'gradnorm/compressor.layer0': 0.24771023983110757, 'gradnorm/compressor.layer1': 5.982234402669819, 'gradnorm/compressor.layer2': 1.9209057604636088, 'gradnorm/compressor.layer3': 1.078350541116555, 'gradnorm/compressor.layer4': 0.44007265290363257, 'gradnorm/compressor.layer5': 0.32488817024101646, 'gradnorm/compressor.layer6': 0.09555080027560774, 'gradnorm/compressor.layer7': 0.22117679539270096, 'gradnorm/decoder.layer0': 0.416700256503708, 'gradnorm/decoder.layer1': 0.3243990706919072, 'gradnorm/decoder.layer2': 0.16755728447532997, 'gradnorm/max_param': 4.217225551605225, 'epoch': 3.72}

 37%|███▋      | 7100/19090 [3:27:10<5:29:39,  1.65s/it]
 37%|███▋      | 7101/19090 [3:27:12<6:07:20,  1.84s/it]
 37%|███▋      | 7102/19090 [3:27:14<5:57:13,  1.79s/it]
 37%|███▋      | 7103/19090 [3:27:15<5:27:15,  1.64s/it]
 37%|███▋      | 7104/19090 [3:27:18<6:08:01,  1.84s/it]
 37%|███▋      | 7105/19090 [3:27:20<6:03:37,  1.82s/it]
 37%|███▋      | 7106/19090 [3:27:21<5:31:33,  1.66s/it]
 37%|███▋      | 7107/19090 [3:27:23<5:57:43,  1.79s/it]
 37%|███▋      | 7108/19090 [3:27:25<6:13:47,  1.87s/it]
 37%|███▋      | 7109/19090 [3:27:26<5:38:37,  1.70s/it]
 37%|███▋      | 7110/19090 [3:27:28<5:51:20,  1.76s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 6.5927252769470215, 'learning_rate': 3.9525109098926486e-05, 'recon_loss': 0.0012706799490842969, 'recon_static': 0.00030430266124312764, 'recon_motion': 0.0009663773147622123, 'recon_norm': 1.2183804392814637, 'recon_norm_std': 0.06257094824686646, 'target_norm': 1.622730368375778, 'recon_cos': 0.7086819365620614, 'gradnorm/compressor.query': 0.011084890643641238, 'gradnorm/compressor.layer0': 0.24828283845982724, 'gradnorm/compressor.layer1': 6.136344835143395, 'gradnorm/compressor.layer2': 1.954724137826571, 'gradnorm/compressor.layer3': 1.14372907525605, 'gradnorm/compressor.layer4': 0.4154129498422926, 'gradnorm/compressor.layer5': 0.3007902292670014, 'gradnorm/compressor.layer6': 0.09414707968969402, 'gradnorm/compressor.layer7': 0.22159216745392757, 'gradnorm/decoder.layer0': 0.4112571292769417, 'gradnorm/decoder.layer1': 0.3167070913027403, 'gradnorm/decoder.layer2': 0.16739933395305936, 'gradnorm/max_param': 4.3259100914001465, 'epoch': 3.72}

 37%|███▋      | 7110/19090 [3:27:28<5:51:20,  1.76s/it]
 37%|███▋      | 7111/19090 [3:27:30<6:19:42,  1.90s/it]
 37%|███▋      | 7112/19090 [3:27:32<5:42:49,  1.72s/it]
 37%|███▋      | 7113/19090 [3:27:34<5:56:25,  1.79s/it]
 37%|███▋      | 7114/19090 [3:27:36<6:25:33,  1.93s/it]
 37%|███▋      | 7115/19090 [3:27:37<5:46:52,  1.74s/it]
 37%|███▋      | 7116/19090 [3:27:39<5:37:14,  1.69s/it]
 37%|███▋      | 7117/19090 [3:27:41<6:03:05,  1.82s/it]
 37%|███▋      | 7118/19090 [3:27:42<5:35:26,  1.68s/it]
 37%|███▋      | 7119/19090 [3:27:44<5:38:37,  1.70s/it]
 37%|███▋      | 7120/19090 [3:27:47<6:39:37,  2.00s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 6.5148773193359375, 'learning_rate': 3.948787885898965e-05, 'recon_loss': 0.0012739715079078452, 'recon_static': 0.00030890797352185474, 'recon_motion': 0.0009650635489379056, 'recon_norm': 1.2179457813501358, 'recon_norm_std': 0.06463453136384487, 'target_norm': 1.6229855179786683, 'recon_cos': 0.7074817165732383, 'gradnorm/compressor.query': 0.013547252724011307, 'gradnorm/compressor.layer0': 0.30241782343128043, 'gradnorm/compressor.layer1': 6.035955461341546, 'gradnorm/compressor.layer2': 2.0083421535482926, 'gradnorm/compressor.layer3': 1.1222812331508774, 'gradnorm/compressor.layer4': 0.41205677944304925, 'gradnorm/compressor.layer5': 0.29829034222932566, 'gradnorm/compressor.layer6': 0.08956562405945571, 'gradnorm/compressor.layer7': 0.2311451409994114, 'gradnorm/decoder.layer0': 0.39505810071888303, 'gradnorm/decoder.layer1': 0.3466095974413353, 'gradnorm/decoder.layer2': 0.17277132747568288, 'gradnorm/max_param': 4.254278659820557, 'epoch': 3.73}

 37%|███▋      | 7120/19090 [3:27:47<6:39:37,  2.00s/it]
 37%|███▋      | 7121/19090 [3:27:48<5:56:53,  1.79s/it]
 37%|███▋      | 7122/19090 [3:27:49<5:27:52,  1.64s/it]
 37%|███▋      | 7123/19090 [3:27:52<6:19:02,  1.90s/it]
 37%|███▋      | 7124/19090 [3:27:53<5:42:41,  1.72s/it]
 37%|███▋      | 7125/19090 [3:27:55<5:46:06,  1.74s/it]
 37%|███▋      | 7126/19090 [3:27:57<6:01:30,  1.81s/it]
 37%|███▋      | 7127/19090 [3:27:59<5:52:47,  1.77s/it]
 37%|███▋      | 7128/19090 [3:28:00<5:25:19,  1.63s/it]
 37%|███▋      | 7129/19090 [3:28:02<6:05:17,  1.83s/it]
 37%|███▋      | 7130/19090 [3:28:04<6:18:58,  1.90s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 6.623069763183594, 'learning_rate': 3.9450600178671325e-05, 'recon_loss': 0.0012887607619632036, 'recon_static': 0.00033662353162071667, 'recon_motion': 0.0009521372470771894, 'recon_norm': 1.233128610253334, 'recon_norm_std': 0.06239042803645134, 'target_norm': 1.6189410656690597, 'recon_cos': 0.7016975298523903, 'gradnorm/compressor.query': 0.014337160456378317, 'gradnorm/compressor.layer0': 0.3065726575422724, 'gradnorm/compressor.layer1': 6.182358271929106, 'gradnorm/compressor.layer2': 1.9443965731888995, 'gradnorm/compressor.layer3': 1.05209920260607, 'gradnorm/compressor.layer4': 0.4331021830230935, 'gradnorm/compressor.layer5': 0.3182471609985348, 'gradnorm/compressor.layer6': 0.10203882083436325, 'gradnorm/compressor.layer7': 0.23083141854733755, 'gradnorm/decoder.layer0': 0.41531833004501206, 'gradnorm/decoder.layer1': 0.3300771543951267, 'gradnorm/decoder.layer2': 0.16650373285457945, 'gradnorm/max_param': 4.357672691345215, 'epoch': 3.73}

 37%|███▋      | 7130/19090 [3:28:04<6:18:58,  1.90s/it]
 37%|███▋      | 7131/19090 [3:28:05<5:42:34,  1.72s/it]
 37%|███▋      | 7132/19090 [3:28:08<6:09:13,  1.85s/it]
 37%|███▋      | 7133/19090 [3:28:10<6:30:40,  1.96s/it]
 37%|███▋      | 7134/19090 [3:28:11<5:50:03,  1.76s/it]
 37%|███▋      | 7135/19090 [3:28:13<5:50:50,  1.76s/it]
 37%|███▋      | 7136/19090 [3:28:15<6:09:48,  1.86s/it]
 37%|███▋      | 7137/19090 [3:28:16<5:46:29,  1.74s/it]
 37%|███▋      | 7138/19090 [3:28:18<5:40:04,  1.71s/it]
 37%|███▋      | 7139/19090 [3:28:21<6:33:55,  1.98s/it]
 37%|███▋      | 7140/19090 [3:28:22<5:52:27,  1.77s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.025731563568115, 'learning_rate': 3.9413273182613206e-05, 'recon_loss': 0.0012964381516212598, 'recon_static': 0.0003345041950524319, 'recon_motion': 0.0009619339456548914, 'recon_norm': 1.2297583490610122, 'recon_norm_std': 0.061273529473692176, 'target_norm': 1.6206501424312592, 'recon_cos': 0.7003976821899414, 'gradnorm/compressor.query': 0.013627683904003927, 'gradnorm/compressor.layer0': 0.3102299494584997, 'gradnorm/compressor.layer1': 6.562053265579512, 'gradnorm/compressor.layer2': 2.084040775983794, 'gradnorm/compressor.layer3': 1.1214760583710701, 'gradnorm/compressor.layer4': 0.41944291944872164, 'gradnorm/compressor.layer5': 0.29115787332214704, 'gradnorm/compressor.layer6': 0.09710593615678421, 'gradnorm/compressor.layer7': 0.21780029268140016, 'gradnorm/decoder.layer0': 0.3789603007983612, 'gradnorm/decoder.layer1': 0.3355344528012695, 'gradnorm/decoder.layer2': 0.1709740771068766, 'gradnorm/max_param': 4.625547409057617, 'epoch': 3.74}

 37%|███▋      | 7140/19090 [3:28:22<5:52:27,  1.77s/it]
 37%|███▋      | 7141/19090 [3:28:24<5:37:02,  1.69s/it]
 37%|███▋      | 7142/19090 [3:28:26<6:48:38,  2.05s/it]
 37%|███▋      | 7143/19090 [3:28:28<6:04:28,  1.83s/it]
 37%|███▋      | 7144/19090 [3:28:29<5:32:13,  1.67s/it]
 37%|███▋      | 7145/19090 [3:28:32<6:27:04,  1.94s/it]
 37%|███▋      | 7146/19090 [3:28:33<5:48:03,  1.75s/it]
 37%|███▋      | 7147/19090 [3:28:34<5:27:31,  1.65s/it]
 37%|███▋      | 7148/19090 [3:28:37<6:32:17,  1.97s/it]
 37%|███▋      | 7149/19090 [3:28:38<5:52:02,  1.77s/it]
 37%|███▋      | 7150/19090 [3:28:40<5:29:05,  1.65s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.1767401695251465, 'learning_rate': 3.937589799561849e-05, 'recon_loss': 0.0012684230314334854, 'recon_static': 0.0003157734150590841, 'recon_motion': 0.00095264957199106, 'recon_norm': 1.2193606555461884, 'recon_norm_std': 0.06355808740481735, 'target_norm': 1.621431165933609, 'recon_cos': 0.7083781003952027, 'gradnorm/compressor.query': 0.012440557724149861, 'gradnorm/compressor.layer0': 0.26291601843434104, 'gradnorm/compressor.layer1': 6.718474058650358, 'gradnorm/compressor.layer2': 2.0823317090895217, 'gradnorm/compressor.layer3': 1.153990264292859, 'gradnorm/compressor.layer4': 0.44321526693505786, 'gradnorm/compressor.layer5': 0.2933373738863121, 'gradnorm/compressor.layer6': 0.09120766778840045, 'gradnorm/compressor.layer7': 0.23475068500223856, 'gradnorm/decoder.layer0': 0.3963018476398857, 'gradnorm/decoder.layer1': 0.31914414178598494, 'gradnorm/decoder.layer2': 0.16063276786983544, 'gradnorm/max_param': 4.736443996429443, 'epoch': 3.75}

 37%|███▋      | 7150/19090 [3:28:40<5:29:05,  1.65s/it]
 37%|███▋      | 7151/19090 [3:28:42<6:32:44,  1.97s/it]
 37%|███▋      | 7152/19090 [3:28:44<5:52:03,  1.77s/it]
 37%|███▋      | 7153/19090 [3:28:45<5:41:40,  1.72s/it]
 37%|███▋      | 7154/19090 [3:28:47<5:50:14,  1.76s/it]
 37%|███▋      | 7155/19090 [3:28:48<5:22:26,  1.62s/it]
 37%|███▋      | 7156/19090 [3:28:51<6:03:36,  1.83s/it]
 37%|███▋      | 7157/19090 [3:28:52<5:47:05,  1.75s/it]
 37%|███▋      | 7158/19090 [3:28:54<5:19:54,  1.61s/it]
 38%|███▊      | 7159/19090 [3:28:56<6:10:00,  1.86s/it]
 38%|███▊      | 7160/19090 [3:28:58<5:52:56,  1.78s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.1789445877075195, 'learning_rate': 3.933847474265155e-05, 'recon_loss': 0.00125942480226513, 'recon_static': 0.00031232586843543686, 'recon_motion': 0.000947098909819033, 'recon_norm': 1.220128658413887, 'recon_norm_std': 0.06576286582276225, 'target_norm': 1.6196345269680024, 'recon_cos': 0.7099021092057228, 'gradnorm/compressor.query': 0.013340937161560436, 'gradnorm/compressor.layer0': 0.2828299991188782, 'gradnorm/compressor.layer1': 6.732716792115352, 'gradnorm/compressor.layer2': 2.0228393141797523, 'gradnorm/compressor.layer3': 1.1591944448532778, 'gradnorm/compressor.layer4': 0.4461184954084887, 'gradnorm/compressor.layer5': 0.3059782783638433, 'gradnorm/compressor.layer6': 0.10197619400068889, 'gradnorm/compressor.layer7': 0.22422416660531644, 'gradnorm/decoder.layer0': 0.43764844147247683, 'gradnorm/decoder.layer1': 0.3406172860873232, 'gradnorm/decoder.layer2': 0.17638277869945065, 'gradnorm/max_param': 4.746371746063232, 'epoch': 3.75}

 38%|███▊      | 7160/19090 [3:28:58<5:52:56,  1.78s/it]
 38%|███▊      | 7161/19090 [3:28:59<5:30:20,  1.66s/it]
 38%|███▊      | 7162/19090 [3:29:02<6:30:31,  1.96s/it]
 38%|███▊      | 7163/19090 [3:29:03<5:49:57,  1.76s/it]
 38%|███▊      | 7164/19090 [3:29:04<5:21:22,  1.62s/it]
 38%|███▊      | 7165/19090 [3:29:07<6:51:49,  2.07s/it]
 38%|███▊      | 7166/19090 [3:29:09<6:04:42,  1.84s/it]
 38%|███▊      | 7167/19090 [3:29:10<5:31:50,  1.67s/it]
 38%|███▊      | 7168/19090 [3:29:13<6:23:57,  1.93s/it]
 38%|███▊      | 7169/19090 [3:29:14<5:45:01,  1.74s/it]
 38%|███▊      | 7170/19090 [3:29:15<5:23:01,  1.63s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.009421348571777, 'learning_rate': 3.9301003548837414e-05, 'recon_loss': 0.0012533842324046417, 'recon_static': 0.0003080819849856198, 'recon_motion': 0.0009453022503294051, 'recon_norm': 1.2182622611522675, 'recon_norm_std': 0.06323104472830891, 'target_norm': 1.6180159449577332, 'recon_cos': 0.7111812755465508, 'gradnorm/compressor.query': 0.01182963393446379, 'gradnorm/compressor.layer0': 0.27082837387581477, 'gradnorm/compressor.layer1': 6.527379983816248, 'gradnorm/compressor.layer2': 2.093662089309989, 'gradnorm/compressor.layer3': 1.187112042146971, 'gradnorm/compressor.layer4': 0.4209766713517724, 'gradnorm/compressor.layer5': 0.31712315428185056, 'gradnorm/compressor.layer6': 0.10357150782649056, 'gradnorm/compressor.layer7': 0.22239084487579333, 'gradnorm/decoder.layer0': 0.42685527256053707, 'gradnorm/decoder.layer1': 0.32931274068131833, 'gradnorm/decoder.layer2': 0.17470433108874975, 'gradnorm/max_param': 4.6013922691345215, 'epoch': 3.76}

 38%|███▊      | 7170/19090 [3:29:15<5:23:01,  1.63s/it]
 38%|███▊      | 7171/19090 [3:29:18<6:28:45,  1.96s/it]
 38%|███▊      | 7172/19090 [3:29:19<5:48:44,  1.76s/it]
 38%|███▊      | 7173/19090 [3:29:21<5:28:28,  1.65s/it]
 38%|███▊      | 7174/19090 [3:29:23<6:12:16,  1.87s/it]
 38%|███▊      | 7175/19090 [3:29:24<5:37:30,  1.70s/it]
 38%|███▊      | 7176/19090 [3:29:26<5:54:24,  1.78s/it]
 38%|███▊      | 7177/19090 [3:29:28<5:54:05,  1.78s/it]
 38%|███▊      | 7178/19090 [3:29:29<5:24:44,  1.64s/it]
 38%|███▊      | 7179/19090 [3:29:32<6:14:16,  1.89s/it]
 38%|███▊      | 7180/19090 [3:29:33<5:55:30,  1.79s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.610073089599609, 'learning_rate': 3.926348453946145e-05, 'recon_loss': 0.0012680842366535217, 'recon_static': 0.00030708293270436114, 'recon_motion': 0.0009610012828488834, 'recon_norm': 1.2169571608304977, 'recon_norm_std': 0.06300199162214995, 'target_norm': 1.6199753999710083, 'recon_cos': 0.7079976111650467, 'gradnorm/compressor.query': 0.013209838374553435, 'gradnorm/compressor.layer0': 0.2826057447583215, 'gradnorm/compressor.layer1': 7.130470986238967, 'gradnorm/compressor.layer2': 2.2146935790484354, 'gradnorm/compressor.layer3': 1.1506458251991434, 'gradnorm/compressor.layer4': 0.45818034421936904, 'gradnorm/compressor.layer5': 0.3547424819948091, 'gradnorm/compressor.layer6': 0.10574054576527976, 'gradnorm/compressor.layer7': 0.2290413262676018, 'gradnorm/decoder.layer0': 0.4551436413931468, 'gradnorm/decoder.layer1': 0.3514227165377292, 'gradnorm/decoder.layer2': 0.17503677177675447, 'gradnorm/max_param': 5.02677059173584, 'epoch': 3.76}

 38%|███▊      | 7180/19090 [3:29:33<5:55:30,  1.79s/it]
 38%|███▊      | 7181/19090 [3:29:35<5:25:33,  1.64s/it]
 38%|███▊      | 7182/19090 [3:29:37<6:06:59,  1.85s/it]
 38%|███▊      | 7183/19090 [3:29:39<6:24:58,  1.94s/it]
 38%|███▊      | 7184/19090 [3:29:40<5:46:10,  1.74s/it]
 38%|███▊      | 7185/19090 [3:29:42<5:51:59,  1.77s/it]
 38%|███▊      | 7186/19090 [3:29:45<6:32:24,  1.98s/it]
 38%|███▊      | 7187/19090 [3:29:46<5:51:24,  1.77s/it]
 38%|███▊      | 7188/19090 [3:29:47<5:22:30,  1.63s/it]
 38%|███▊      | 7189/19090 [3:29:50<6:24:15,  1.94s/it]
 38%|███▊      | 7190/19090 [3:29:51<5:46:05,  1.74s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.800946235656738, 'learning_rate': 3.922591783996886e-05, 'recon_loss': 0.0012718226265860722, 'recon_static': 0.0003120269364444539, 'recon_motion': 0.000959795706148725, 'recon_norm': 1.2193617433309556, 'recon_norm_std': 0.061397457309067247, 'target_norm': 1.621940904855728, 'recon_cos': 0.7077993676066399, 'gradnorm/compressor.query': 0.015295820124156013, 'gradnorm/compressor.layer0': 0.29825583366147074, 'gradnorm/compressor.layer1': 7.340993550590714, 'gradnorm/compressor.layer2': 2.1257114341455954, 'gradnorm/compressor.layer3': 1.2529899124811097, 'gradnorm/compressor.layer4': 0.46954334199032005, 'gradnorm/compressor.layer5': 0.35062271180315724, 'gradnorm/compressor.layer6': 0.10078013755864514, 'gradnorm/compressor.layer7': 0.23910603282721274, 'gradnorm/decoder.layer0': 0.46365626308057795, 'gradnorm/decoder.layer1': 0.3578680828184132, 'gradnorm/decoder.layer2': 0.18156324809933047, 'gradnorm/max_param': 5.175050258636475, 'epoch': 3.77}

 38%|███▊      | 7190/19090 [3:29:51<5:46:05,  1.74s/it]
 38%|███▊      | 7191/19090 [3:29:53<5:19:05,  1.61s/it]
 38%|███▊      | 7192/19090 [3:29:55<6:04:51,  1.84s/it]
 38%|███▊      | 7193/19090 [3:29:56<5:32:01,  1.67s/it]
 38%|███▊      | 7194/19090 [3:29:58<5:09:03,  1.56s/it]
 38%|███▊      | 7195/19090 [3:30:00<6:27:55,  1.96s/it]
 38%|███▊      | 7196/19090 [3:30:02<5:48:05,  1.76s/it]
 38%|███▊      | 7197/19090 [3:30:03<5:20:12,  1.62s/it]
 38%|███▊      | 7198/19090 [3:30:06<6:15:13,  1.89s/it]
 38%|███▊      | 7199/19090 [3:30:07<5:39:31,  1.71s/it]
 38%|███▊      | 7200/19090 [3:30:08<5:26:50,  1.65s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.397149085998535, 'learning_rate': 3.918830357596434e-05, 'recon_loss': 0.0012646778573980554, 'recon_static': 0.00031043235067045315, 'recon_motion': 0.0009542455372866243, 'recon_norm': 1.219075345993042, 'recon_norm_std': 0.06308288909494877, 'target_norm': 1.619818878173828, 'recon_cos': 0.7087836250662803, 'gradnorm/compressor.query': 0.013833479856400974, 'gradnorm/compressor.layer0': 0.28293758966433125, 'gradnorm/compressor.layer1': 6.910896774817124, 'gradnorm/compressor.layer2': 2.1406537353368984, 'gradnorm/compressor.layer3': 1.2454792767779646, 'gradnorm/compressor.layer4': 0.4716670348561381, 'gradnorm/compressor.layer5': 0.31157272570787853, 'gradnorm/compressor.layer6': 0.10072265980801696, 'gradnorm/compressor.layer7': 0.23600099321889795, 'gradnorm/decoder.layer0': 0.43277422463739457, 'gradnorm/decoder.layer1': 0.3729946039937057, 'gradnorm/decoder.layer2': 0.17789890722362114, 'gradnorm/max_param': 4.871884346008301, 'epoch': 3.77}

 38%|███▊      | 7200/19090 [3:30:08<5:26:50,  1.65s/it]
 38%|███▊      | 7201/19090 [3:30:11<6:13:31,  1.89s/it]
 38%|███▊      | 7202/19090 [3:30:12<5:38:25,  1.71s/it]
 38%|███▊      | 7203/19090 [3:30:13<5:18:53,  1.61s/it]
 38%|███▊      | 7204/19090 [3:30:16<6:08:15,  1.86s/it]
 38%|███▊      | 7205/19090 [3:30:17<5:38:55,  1.71s/it]
 38%|███▊      | 7206/19090 [3:30:19<5:37:13,  1.70s/it]
 38%|███▊      | 7207/19090 [3:30:21<6:13:15,  1.88s/it]
 38%|███▊      | 7208/19090 [3:30:23<5:37:52,  1.71s/it]
 38%|███▊      | 7209/19090 [3:30:25<6:03:48,  1.84s/it]
 38%|███▊      | 7210/19090 [3:30:26<6:04:10,  1.84s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.158912181854248, 'learning_rate': 3.915064187321156e-05, 'recon_loss': 0.0012674855068325996, 'recon_static': 0.0003171634692989755, 'recon_motion': 0.0009503219975158572, 'recon_norm': 1.2214423775672913, 'recon_norm_std': 0.06131614428013563, 'target_norm': 1.6194039732217789, 'recon_cos': 0.7076529279351235, 'gradnorm/compressor.query': 0.01377384812664025, 'gradnorm/compressor.layer0': 0.2879773986146908, 'gradnorm/compressor.layer1': 6.713820525406028, 'gradnorm/compressor.layer2': 2.0417451129816073, 'gradnorm/compressor.layer3': 1.0956899060762606, 'gradnorm/compressor.layer4': 0.4401801813002766, 'gradnorm/compressor.layer5': 0.31558929751317116, 'gradnorm/compressor.layer6': 0.10537794407310078, 'gradnorm/compressor.layer7': 0.21845118515248033, 'gradnorm/decoder.layer0': 0.44728309897058816, 'gradnorm/decoder.layer1': 0.3707282625605181, 'gradnorm/decoder.layer2': 0.18049516773491997, 'gradnorm/max_param': 4.732743740081787, 'epoch': 3.78}

 38%|███▊      | 7210/19090 [3:30:27<6:04:10,  1.84s/it]
 38%|███▊      | 7211/19090 [3:30:28<5:31:24,  1.67s/it]
 38%|███▊      | 7212/19090 [3:30:30<5:55:56,  1.80s/it]
 38%|███▊      | 7213/19090 [3:30:32<6:21:05,  1.93s/it]
 38%|███▊      | 7214/19090 [3:30:33<5:43:04,  1.73s/it]
 38%|███▊      | 7215/19090 [3:30:35<5:45:18,  1.74s/it]
 38%|███▊      | 7216/19090 [3:30:37<6:15:54,  1.90s/it]
 38%|███▊      | 7217/19090 [3:30:39<5:42:30,  1.73s/it]
 38%|███▊      | 7218/19090 [3:30:40<5:43:03,  1.73s/it]
 38%|███▊      | 7219/19090 [3:30:43<6:33:46,  1.99s/it]
 38%|███▊      | 7220/19090 [3:30:44<5:56:48,  1.80s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.448819160461426, 'learning_rate': 3.911293285763285e-05, 'recon_loss': 0.0012715776334516703, 'recon_static': 0.00031066316296346486, 'recon_motion': 0.0009609144515707157, 'recon_norm': 1.217142117023468, 'recon_norm_std': 0.06227182177826762, 'target_norm': 1.6202720165252686, 'recon_cos': 0.7069398134946823, 'gradnorm/compressor.query': 0.012862509239311799, 'gradnorm/compressor.layer0': 0.2824681841606907, 'gradnorm/compressor.layer1': 6.993564099388513, 'gradnorm/compressor.layer2': 2.0650980220926147, 'gradnorm/compressor.layer3': 1.2287548420559389, 'gradnorm/compressor.layer4': 0.45426204675814685, 'gradnorm/compressor.layer5': 0.31240815958880125, 'gradnorm/compressor.layer6': 0.10669790710196055, 'gradnorm/compressor.layer7': 0.24183737919371082, 'gradnorm/decoder.layer0': 0.442930846299828, 'gradnorm/decoder.layer1': 0.3427513963182584, 'gradnorm/decoder.layer2': 0.18183127006989835, 'gradnorm/max_param': 4.930747032165527, 'epoch': 3.78}

 38%|███▊      | 7220/19090 [3:30:44<5:56:48,  1.80s/it]
 38%|███▊      | 7221/19090 [3:30:46<5:40:37,  1.72s/it]
 38%|███▊      | 7222/19090 [3:30:49<6:30:27,  1.97s/it]
 38%|███▊      | 7223/19090 [3:30:50<5:53:23,  1.79s/it]
 38%|███▊      | 7224/19090 [3:30:51<5:24:52,  1.64s/it]
 38%|███▊      | 7225/19090 [3:30:54<6:20:12,  1.92s/it]
 38%|███▊      | 7226/19090 [3:30:55<5:42:28,  1.73s/it]
 38%|███▊      | 7227/19090 [3:30:57<5:27:59,  1.66s/it]
 38%|███▊      | 7228/19090 [3:30:59<6:23:32,  1.94s/it]
 38%|███▊      | 7229/19090 [3:31:00<5:44:37,  1.74s/it]
 38%|███▊      | 7230/19090 [3:31:02<5:30:51,  1.67s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.40976619720459, 'learning_rate': 3.907517665530872e-05, 'recon_loss': 0.0012619617977179588, 'recon_static': 0.00031388313800562173, 'recon_motion': 0.0009480786480708048, 'recon_norm': 1.2195154666900634, 'recon_norm_std': 0.06246944572776556, 'target_norm': 1.6178582459688187, 'recon_cos': 0.7085330799221993, 'gradnorm/compressor.query': 0.013875868885479726, 'gradnorm/compressor.layer0': 0.3065767768428551, 'gradnorm/compressor.layer1': 6.938012456439028, 'gradnorm/compressor.layer2': 2.1452812175873257, 'gradnorm/compressor.layer3': 1.1608099845588273, 'gradnorm/compressor.layer4': 0.4783952522685736, 'gradnorm/compressor.layer5': 0.32183770176602683, 'gradnorm/compressor.layer6': 0.12043457433366875, 'gradnorm/compressor.layer7': 0.23083781740387083, 'gradnorm/decoder.layer0': 0.4242562974643376, 'gradnorm/decoder.layer1': 0.34193082092210264, 'gradnorm/decoder.layer2': 0.1661064180170698, 'gradnorm/max_param': 4.890784740447998, 'epoch': 3.79}

 38%|███▊      | 7230/19090 [3:31:02<5:30:51,  1.67s/it]
 38%|███▊      | 7231/19090 [3:31:05<6:38:50,  2.02s/it]
 38%|███▊      | 7232/19090 [3:31:06<5:56:04,  1.80s/it]
 38%|███▊      | 7233/19090 [3:31:07<5:25:45,  1.65s/it]
 38%|███▊      | 7234/19090 [3:31:10<5:59:45,  1.82s/it]
 38%|███▊      | 7235/19090 [3:31:11<5:28:21,  1.66s/it]
 38%|███▊      | 7236/19090 [3:31:12<5:11:42,  1.58s/it]
 38%|███▊      | 7237/19090 [3:31:15<6:13:34,  1.89s/it]
 38%|███▊      | 7238/19090 [3:31:16<5:37:57,  1.71s/it]
 38%|███▊      | 7239/19090 [3:31:18<5:17:05,  1.61s/it]
 38%|███▊      | 7240/19090 [3:31:20<6:32:30,  1.99s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 6.502119064331055, 'learning_rate': 3.9037373392477445e-05, 'recon_loss': 0.0012681893829721957, 'recon_static': 0.0003160278793075122, 'recon_motion': 0.0009521614978439174, 'recon_norm': 1.221029469370842, 'recon_norm_std': 0.06193113224580884, 'target_norm': 1.6189488023519516, 'recon_cos': 0.707210224866867, 'gradnorm/compressor.query': 0.01299795889538052, 'gradnorm/compressor.layer0': 0.27042179599231153, 'gradnorm/compressor.layer1': 6.0515093725479066, 'gradnorm/compressor.layer2': 1.9093447718748289, 'gradnorm/compressor.layer3': 1.12224787941382, 'gradnorm/compressor.layer4': 0.43550831577012933, 'gradnorm/compressor.layer5': 0.31860318823318906, 'gradnorm/compressor.layer6': 0.11332477769203984, 'gradnorm/compressor.layer7': 0.22832920752200644, 'gradnorm/decoder.layer0': 0.4329901437419504, 'gradnorm/decoder.layer1': 0.32055783761840373, 'gradnorm/decoder.layer2': 0.1790631684695244, 'gradnorm/max_param': 4.266115665435791, 'epoch': 3.79}

 38%|███▊      | 7240/19090 [3:31:20<6:32:30,  1.99s/it]
 38%|███▊      | 7241/19090 [3:31:22<5:51:22,  1.78s/it]
 38%|███▊      | 7242/19090 [3:31:23<5:22:31,  1.63s/it]
 38%|███▊      | 7243/19090 [3:31:26<6:22:49,  1.94s/it]
 38%|███▊      | 7244/19090 [3:31:27<5:44:27,  1.74s/it]
 38%|███▊      | 7245/19090 [3:31:28<5:17:32,  1.61s/it]
 38%|███▊      | 7246/19090 [3:31:31<6:03:59,  1.84s/it]
 38%|███▊      | 7247/19090 [3:31:32<5:31:06,  1.68s/it]
 38%|███▊      | 7248/19090 [3:31:33<5:08:03,  1.56s/it]
 38%|███▊      | 7249/19090 [3:31:36<6:47:54,  2.07s/it]
 38%|███▊      | 7250/19090 [3:31:38<6:01:41,  1.83s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.552533149719238, 'learning_rate': 3.899952319553465e-05, 'recon_loss': 0.0012891314952867105, 'recon_static': 0.0003215275653928984, 'recon_motion': 0.0009676039262558334, 'recon_norm': 1.2203792810440064, 'recon_norm_std': 0.06189182214438915, 'target_norm': 1.620801991224289, 'recon_cos': 0.7024053201079369, 'gradnorm/compressor.query': 0.015586495453723717, 'gradnorm/compressor.layer0': 0.3353172053587333, 'gradnorm/compressor.layer1': 7.0202814294639895, 'gradnorm/compressor.layer2': 2.299653222022581, 'gradnorm/compressor.layer3': 1.2462728246131103, 'gradnorm/compressor.layer4': 0.48347204399355215, 'gradnorm/compressor.layer5': 0.3330275342364765, 'gradnorm/compressor.layer6': 0.11181552753353084, 'gradnorm/compressor.layer7': 0.25520846146301374, 'gradnorm/decoder.layer0': 0.47350289047430466, 'gradnorm/decoder.layer1': 0.3485472651649727, 'gradnorm/decoder.layer2': 0.18481172184204642, 'gradnorm/max_param': 4.948816299438477, 'epoch': 3.8}

 38%|███▊      | 7250/19090 [3:31:38<6:01:41,  1.83s/it]
 38%|███▊      | 7251/19090 [3:31:39<5:29:47,  1.67s/it]
 38%|███▊      | 7252/19090 [3:31:42<6:52:12,  2.09s/it]
 38%|███▊      | 7253/19090 [3:31:43<6:05:55,  1.85s/it]
 38%|███▊      | 7254/19090 [3:31:45<5:32:51,  1.69s/it]
 38%|███▊      | 7255/19090 [3:31:47<6:22:38,  1.94s/it]
 38%|███▊      | 7256/19090 [3:31:49<5:44:36,  1.75s/it]
 38%|███▊      | 7257/19090 [3:31:50<5:18:12,  1.61s/it]
 38%|███▊      | 7258/19090 [3:31:53<6:28:33,  1.97s/it]
 38%|███▊      | 7259/19090 [3:31:54<5:48:23,  1.77s/it]
 38%|███▊      | 7260/19090 [3:31:55<5:20:45,  1.63s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.448709964752197, 'learning_rate': 3.896162619103287e-05, 'recon_loss': 0.001274859221302904, 'recon_static': 0.0003151260665617883, 'recon_motion': 0.0009597331576514989, 'recon_norm': 1.219540497660637, 'recon_norm_std': 0.06280975006520748, 'target_norm': 1.6193803995847702, 'recon_cos': 0.7055156514048576, 'gradnorm/compressor.query': 0.013951772732321044, 'gradnorm/compressor.layer0': 0.30420010047650103, 'gradnorm/compressor.layer1': 6.913358456386041, 'gradnorm/compressor.layer2': 2.265458620473541, 'gradnorm/compressor.layer3': 1.2881815356452933, 'gradnorm/compressor.layer4': 0.47333092261034054, 'gradnorm/compressor.layer5': 0.3446730553271187, 'gradnorm/compressor.layer6': 0.1223423634739206, 'gradnorm/compressor.layer7': 0.24617715209466912, 'gradnorm/decoder.layer0': 0.46787611246651944, 'gradnorm/decoder.layer1': 0.3604965331867199, 'gradnorm/decoder.layer2': 0.19244484458057767, 'gradnorm/max_param': 4.873475551605225, 'epoch': 3.8}

 38%|███▊      | 7260/19090 [3:31:55<5:20:45,  1.63s/it]
 38%|███▊      | 7261/19090 [3:31:58<6:38:08,  2.02s/it]
 38%|███▊      | 7262/19090 [3:31:59<5:54:55,  1.80s/it]
 38%|███▊      | 7263/19090 [3:32:01<5:24:50,  1.65s/it]
 38%|███▊      | 7264/19090 [3:32:04<6:38:30,  2.02s/it]
 38%|███▊      | 7265/19090 [3:32:05<5:55:26,  1.80s/it]
 38%|███▊      | 7266/19090 [3:32:06<5:25:34,  1.65s/it]
 38%|███▊      | 7267/19090 [3:32:09<6:27:01,  1.96s/it]
 38%|███▊      | 7268/19090 [3:32:10<5:46:51,  1.76s/it]
 38%|███▊      | 7269/19090 [3:32:11<5:19:04,  1.62s/it]
 38%|███▊      | 7270/19090 [3:32:14<6:30:11,  1.98s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.32090950012207, 'learning_rate': 3.892368250568117e-05, 'recon_loss': 0.0012874459876911714, 'recon_static': 0.00033383441477781163, 'recon_motion': 0.0009536115132505075, 'recon_norm': 1.2270685583353043, 'recon_norm_std': 0.06065738908946514, 'target_norm': 1.619238057732582, 'recon_cos': 0.7023251235485077, 'gradnorm/compressor.query': 0.01697876457616354, 'gradnorm/compressor.layer0': 0.37688763696920385, 'gradnorm/compressor.layer1': 6.821878120257389, 'gradnorm/compressor.layer2': 2.17451601891879, 'gradnorm/compressor.layer3': 1.1793790502433186, 'gradnorm/compressor.layer4': 0.4899555188298851, 'gradnorm/compressor.layer5': 0.3492361772699033, 'gradnorm/compressor.layer6': 0.12882333313986785, 'gradnorm/compressor.layer7': 0.25237125624741885, 'gradnorm/decoder.layer0': 0.4311986533294343, 'gradnorm/decoder.layer1': 0.3607021167023042, 'gradnorm/decoder.layer2': 0.19379489670752326, 'gradnorm/max_param': 4.807795524597168, 'epoch': 3.81}

 38%|███▊      | 7270/19090 [3:32:14<6:30:11,  1.98s/it]
 38%|███▊      | 7271/19090 [3:32:16<5:49:50,  1.78s/it]
 38%|███▊      | 7272/19090 [3:32:17<5:21:33,  1.63s/it]
 38%|███▊      | 7273/19090 [3:32:19<6:15:55,  1.91s/it]
 38%|███▊      | 7274/19090 [3:32:21<5:39:23,  1.72s/it]
 38%|███▊      | 7275/19090 [3:32:22<5:14:05,  1.60s/it]
 38%|███▊      | 7276/19090 [3:32:25<6:26:12,  1.96s/it]
 38%|███▊      | 7277/19090 [3:32:26<5:46:17,  1.76s/it]
 38%|███▊      | 7278/19090 [3:32:27<5:18:35,  1.62s/it]
 38%|███▊      | 7279/19090 [3:32:30<6:26:37,  1.96s/it]
 38%|███▊      | 7280/19090 [3:32:31<5:46:42,  1.76s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.115129470825195, 'learning_rate': 3.888569226634466e-05, 'recon_loss': 0.0013037726632319391, 'recon_static': 0.0003455670230323449, 'recon_motion': 0.0009582056460203603, 'recon_norm': 1.2327207326889038, 'recon_norm_std': 0.061467396654188634, 'target_norm': 1.6207023411989212, 'recon_cos': 0.698620681464672, 'gradnorm/compressor.query': 0.016994640293892957, 'gradnorm/compressor.layer0': 0.38417230161656446, 'gradnorm/compressor.layer1': 6.604135154223746, 'gradnorm/compressor.layer2': 2.1948948798935586, 'gradnorm/compressor.layer3': 1.1433689608365942, 'gradnorm/compressor.layer4': 0.4836004124245648, 'gradnorm/compressor.layer5': 0.3274625721867562, 'gradnorm/compressor.layer6': 0.12283400636050358, 'gradnorm/compressor.layer7': 0.24224968081218057, 'gradnorm/decoder.layer0': 0.4112058017482839, 'gradnorm/decoder.layer1': 0.34499140391836025, 'gradnorm/decoder.layer2': 0.18641002062863377, 'gradnorm/max_param': 4.653957366943359, 'epoch': 3.81}

 38%|███▊      | 7280/19090 [3:32:32<5:46:42,  1.76s/it]
 38%|███▊      | 7281/19090 [3:32:33<5:18:54,  1.62s/it]
 38%|███▊      | 7282/19090 [3:32:36<6:35:17,  2.01s/it]
 38%|███▊      | 7283/19090 [3:32:37<5:52:31,  1.79s/it]
 38%|███▊      | 7284/19090 [3:32:38<5:22:47,  1.64s/it]
 38%|███▊      | 7285/19090 [3:32:41<6:26:52,  1.97s/it]
 38%|███▊      | 7286/19090 [3:32:42<5:47:00,  1.76s/it]
 38%|███▊      | 7287/19090 [3:32:44<5:19:02,  1.62s/it]
 38%|███▊      | 7288/19090 [3:32:46<6:20:35,  1.93s/it]
 38%|███▊      | 7289/19090 [3:32:48<5:42:30,  1.74s/it]
 38%|███▊      | 7290/19090 [3:32:49<5:16:38,  1.61s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.004261493682861, 'learning_rate': 3.8847655600044145e-05, 'recon_loss': 0.0013246642163721844, 'recon_static': 0.0003626070603786502, 'recon_motion': 0.0009620571392588318, 'recon_norm': 1.2387098133563996, 'recon_norm_std': 0.060464908182621, 'target_norm': 1.6223171710968018, 'recon_cos': 0.6936311110854149, 'gradnorm/compressor.query': 0.01675186680168539, 'gradnorm/compressor.layer0': 0.3566010226503536, 'gradnorm/compressor.layer1': 6.473560847884998, 'gradnorm/compressor.layer2': 2.1949787361314215, 'gradnorm/compressor.layer3': 1.2152658201567907, 'gradnorm/compressor.layer4': 0.459650749044473, 'gradnorm/compressor.layer5': 0.3324367403321857, 'gradnorm/compressor.layer6': 0.12221729965807605, 'gradnorm/compressor.layer7': 0.239159108257373, 'gradnorm/decoder.layer0': 0.4250876525641272, 'gradnorm/decoder.layer1': 0.35115569207877956, 'gradnorm/decoder.layer2': 0.1802842555411329, 'gradnorm/max_param': 4.5622758865356445, 'epoch': 3.82}

 38%|███▊      | 7290/19090 [3:32:49<5:16:38,  1.61s/it]
 38%|███▊      | 7291/19090 [3:32:51<6:10:56,  1.89s/it]
 38%|███▊      | 7292/19090 [3:32:53<5:35:41,  1.71s/it]
 38%|███▊      | 7293/19090 [3:32:54<5:11:01,  1.58s/it]
 38%|███▊      | 7294/19090 [3:32:56<6:04:32,  1.85s/it]
 38%|███▊      | 7295/19090 [3:32:58<5:31:47,  1.69s/it]
 38%|███▊      | 7296/19090 [3:32:59<5:08:29,  1.57s/it]
 38%|███▊      | 7297/19090 [3:33:02<6:32:07,  2.00s/it]
 38%|███▊      | 7298/19090 [3:33:03<5:50:48,  1.78s/it]
 38%|███▊      | 7299/19090 [3:33:05<5:21:44,  1.64s/it]
 38%|███▊      | 7300/19090 [3:33:07<6:35:08,  2.01s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 6.993708610534668, 'learning_rate': 3.88095726339556e-05, 'recon_loss': 0.0013324364233994856, 'recon_static': 0.0003709904049173929, 'recon_motion': 0.0009614460141165182, 'recon_norm': 1.246129259467125, 'recon_norm_std': 0.0593513997271657, 'target_norm': 1.6201200306415557, 'recon_cos': 0.6910201996564865, 'gradnorm/compressor.query': 0.016976128894837347, 'gradnorm/compressor.layer0': 0.375013212130178, 'gradnorm/compressor.layer1': 6.500119428271077, 'gradnorm/compressor.layer2': 2.111472549227519, 'gradnorm/compressor.layer3': 1.1395050717015405, 'gradnorm/compressor.layer4': 0.4689925043334694, 'gradnorm/compressor.layer5': 0.33037797298818494, 'gradnorm/compressor.layer6': 0.1269636663315339, 'gradnorm/compressor.layer7': 0.250941960662112, 'gradnorm/decoder.layer0': 0.4398588723383103, 'gradnorm/decoder.layer1': 0.3479374915969763, 'gradnorm/decoder.layer2': 0.19989120832549112, 'gradnorm/max_param': 4.580260753631592, 'epoch': 3.82}

 38%|███▊      | 7300/19090 [3:33:07<6:35:08,  2.01s/it]
 38%|███▊      | 7301/19090 [3:33:09<5:52:50,  1.80s/it]
 38%|███▊      | 7302/19090 [3:33:10<5:23:17,  1.65s/it]
 38%|███▊      | 7303/19090 [3:33:13<6:19:04,  1.93s/it]
 38%|███▊      | 7304/19090 [3:33:14<5:42:15,  1.74s/it]
 38%|███▊      | 7305/19090 [3:33:15<5:15:11,  1.60s/it]
 38%|███▊      | 7306/19090 [3:33:18<6:24:47,  1.96s/it]
 38%|███▊      | 7307/19090 [3:33:19<5:45:25,  1.76s/it]
 38%|███▊      | 7308/19090 [3:33:21<5:17:47,  1.62s/it]
 38%|███▊      | 7309/19090 [3:33:23<6:14:24,  1.91s/it]
 38%|███▊      | 7310/19090 [3:33:25<5:38:08,  1.72s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 8.04019546508789, 'learning_rate': 3.877144349540988e-05, 'recon_loss': 0.0013313994626514614, 'recon_static': 0.00036840076791122556, 'recon_motion': 0.0009629986918298528, 'recon_norm': 1.2403602004051208, 'recon_norm_std': 0.05987965865060687, 'target_norm': 1.6226608783006669, 'recon_cos': 0.6920450612902641, 'gradnorm/compressor.query': 0.018112521662507455, 'gradnorm/compressor.layer0': 0.39007793265303015, 'gradnorm/compressor.layer1': 7.495599854674507, 'gradnorm/compressor.layer2': 2.396745414165273, 'gradnorm/compressor.layer3': 1.2983165220968587, 'gradnorm/compressor.layer4': 0.510133895267402, 'gradnorm/compressor.layer5': 0.3619671712118208, 'gradnorm/compressor.layer6': 0.1320260125876416, 'gradnorm/compressor.layer7': 0.2533893164752464, 'gradnorm/decoder.layer0': 0.47859443482671615, 'gradnorm/decoder.layer1': 0.3688756607409111, 'gradnorm/decoder.layer2': 0.20050672175967052, 'gradnorm/max_param': 5.283064365386963, 'epoch': 3.83}

 38%|███▊      | 7310/19090 [3:33:25<5:38:08,  1.72s/it]
 38%|███▊      | 7311/19090 [3:33:26<5:12:48,  1.59s/it]
 38%|███▊      | 7312/19090 [3:33:29<6:20:45,  1.94s/it]
 38%|███▊      | 7313/19090 [3:33:30<5:42:31,  1.75s/it]
 38%|███▊      | 7314/19090 [3:33:31<5:15:35,  1.61s/it]
 38%|███▊      | 7315/19090 [3:33:34<6:10:58,  1.89s/it]
 38%|███▊      | 7316/19090 [3:33:35<5:35:34,  1.71s/it]
 38%|███▊      | 7317/19090 [3:33:36<5:11:05,  1.59s/it]
 38%|███▊      | 7318/19090 [3:33:39<6:17:01,  1.92s/it]
 38%|███▊      | 7319/19090 [3:33:40<5:39:46,  1.73s/it]
 38%|███▊      | 7320/19090 [3:33:42<5:14:14,  1.60s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.467677116394043, 'learning_rate': 3.8733268311892144e-05, 'recon_loss': 0.0013043454644503071, 'recon_static': 0.0003546163745340891, 'recon_motion': 0.0009497291262960062, 'recon_norm': 1.2355467587709428, 'recon_norm_std': 0.05977412527427077, 'target_norm': 1.6172008633613586, 'recon_cos': 0.6967119514942169, 'gradnorm/compressor.query': 0.015684316036259745, 'gradnorm/compressor.layer0': 0.3568568532041124, 'gradnorm/compressor.layer1': 6.966140600660029, 'gradnorm/compressor.layer2': 2.2189212786690238, 'gradnorm/compressor.layer3': 1.1624908590470115, 'gradnorm/compressor.layer4': 0.4935771760394952, 'gradnorm/compressor.layer5': 0.3350967710131671, 'gradnorm/compressor.layer6': 0.13318453600318428, 'gradnorm/compressor.layer7': 0.2548507582579677, 'gradnorm/decoder.layer0': 0.4786763758110779, 'gradnorm/decoder.layer1': 0.36331713919611847, 'gradnorm/decoder.layer2': 0.1917381134452136, 'gradnorm/max_param': 4.910116672515869, 'epoch': 3.83}

 38%|███▊      | 7320/19090 [3:33:42<5:14:14,  1.60s/it]
 38%|███▊      | 7321/19090 [3:33:44<6:07:24,  1.87s/it]
 38%|███▊      | 7322/19090 [3:33:45<5:33:03,  1.70s/it]
 38%|███▊      | 7323/19090 [3:33:47<5:09:14,  1.58s/it]
 38%|███▊      | 7324/19090 [3:33:49<6:21:18,  1.94s/it]
 38%|███▊      | 7325/19090 [3:33:51<5:43:16,  1.75s/it]
 38%|███▊      | 7326/19090 [3:33:52<5:16:31,  1.61s/it]
 38%|███▊      | 7327/19090 [3:33:55<6:23:27,  1.96s/it]
 38%|███▊      | 7328/19090 [3:33:56<5:43:50,  1.75s/it]
 38%|███▊      | 7329/19090 [3:33:57<5:16:29,  1.61s/it]
 38%|███▊      | 7330/19090 [3:34:00<6:17:23,  1.93s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.691984176635742, 'learning_rate': 3.8695047211041555e-05, 'recon_loss': 0.0013066157815046608, 'recon_static': 0.00035659605928231033, 'recon_motion': 0.0009500197353190743, 'recon_norm': 1.2368116915225982, 'recon_norm_std': 0.059547951351851224, 'target_norm': 1.6163177907466888, 'recon_cos': 0.6958195596933365, 'gradnorm/compressor.query': 0.01775308877937768, 'gradnorm/compressor.layer0': 0.4012665634064088, 'gradnorm/compressor.layer1': 7.146539621694763, 'gradnorm/compressor.layer2': 2.328787280617158, 'gradnorm/compressor.layer3': 1.2717115723683547, 'gradnorm/compressor.layer4': 0.5119956628645542, 'gradnorm/compressor.layer5': 0.3577431449140061, 'gradnorm/compressor.layer6': 0.12849968315662477, 'gradnorm/compressor.layer7': 0.2500560220891399, 'gradnorm/decoder.layer0': 0.47625452001934804, 'gradnorm/decoder.layer1': 0.39613663405514266, 'gradnorm/decoder.layer2': 0.19710790353135818, 'gradnorm/max_param': 5.036955833435059, 'epoch': 3.84}

 38%|███▊      | 7330/19090 [3:34:00<6:17:23,  1.93s/it]
 38%|███▊      | 7331/19090 [3:34:01<5:39:53,  1.73s/it]
 38%|███▊      | 7332/19090 [3:34:03<5:13:40,  1.60s/it]
 38%|███▊      | 7333/19090 [3:34:05<6:19:27,  1.94s/it]
 38%|███▊      | 7334/19090 [3:34:07<5:41:21,  1.74s/it]
 38%|███▊      | 7335/19090 [3:34:08<5:14:29,  1.61s/it]
 38%|███▊      | 7336/19090 [3:34:11<6:18:40,  1.93s/it]
 38%|███▊      | 7337/19090 [3:34:12<5:40:39,  1.74s/it]
 38%|███▊      | 7338/19090 [3:34:13<5:14:05,  1.60s/it]
 38%|███▊      | 7339/19090 [3:34:16<6:18:39,  1.93s/it]
 38%|███▊      | 7340/19090 [3:34:17<5:40:51,  1.74s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.999471187591553, 'learning_rate': 3.865678032065078e-05, 'recon_loss': 0.0013088163337670266, 'recon_static': 0.00035575279689510354, 'recon_motion': 0.0009530635114060714, 'recon_norm': 1.2362461894750596, 'recon_norm_std': 0.0596371166408062, 'target_norm': 1.6182318329811096, 'recon_cos': 0.6962104439735413, 'gradnorm/compressor.query': 0.021456695873517024, 'gradnorm/compressor.layer0': 0.43584435400999955, 'gradnorm/compressor.layer1': 7.463366495558254, 'gradnorm/compressor.layer2': 2.3350068857226245, 'gradnorm/compressor.layer3': 1.3554275462459267, 'gradnorm/compressor.layer4': 0.48536034313218696, 'gradnorm/compressor.layer5': 0.3301642491770291, 'gradnorm/compressor.layer6': 0.130309451099561, 'gradnorm/compressor.layer7': 0.25645450363481226, 'gradnorm/decoder.layer0': 0.4356020392136793, 'gradnorm/decoder.layer1': 0.39657123910760317, 'gradnorm/decoder.layer2': 0.18840305100466592, 'gradnorm/max_param': 5.260162353515625, 'epoch': 3.84}

 38%|███▊      | 7340/19090 [3:34:17<5:40:51,  1.74s/it]
 38%|███▊      | 7341/19090 [3:34:18<5:14:28,  1.61s/it]
 38%|███▊      | 7342/19090 [3:34:21<6:17:26,  1.93s/it]
 38%|███▊      | 7343/19090 [3:34:22<5:39:50,  1.74s/it]
 38%|███▊      | 7344/19090 [3:34:24<5:13:37,  1.60s/it]
 38%|███▊      | 7345/19090 [3:34:26<6:02:01,  1.85s/it]
 38%|███▊      | 7346/19090 [3:34:27<5:29:16,  1.68s/it]
 38%|███▊      | 7347/19090 [3:34:29<5:06:25,  1.57s/it]
 38%|███▊      | 7348/19090 [3:34:32<6:27:37,  1.98s/it]
 38%|███▊      | 7349/19090 [3:34:33<5:46:48,  1.77s/it]
 39%|███▊      | 7350/19090 [3:34:34<5:18:32,  1.63s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.825565814971924, 'learning_rate': 3.861846776866557e-05, 'recon_loss': 0.0012958479637745768, 'recon_static': 0.0003464027315203566, 'recon_motion': 0.0009494451922364533, 'recon_norm': 1.231252419948578, 'recon_norm_std': 0.06114083528518677, 'target_norm': 1.6187738537788392, 'recon_cos': 0.699893505871296, 'gradnorm/compressor.query': 0.017999244714504453, 'gradnorm/compressor.layer0': 0.37560142528798746, 'gradnorm/compressor.layer1': 7.348128681316591, 'gradnorm/compressor.layer2': 2.213703487689508, 'gradnorm/compressor.layer3': 1.172855780575653, 'gradnorm/compressor.layer4': 0.4979394385452237, 'gradnorm/compressor.layer5': 0.339514660717458, 'gradnorm/compressor.layer6': 0.13411545444563633, 'gradnorm/compressor.layer7': 0.25368231655256507, 'gradnorm/decoder.layer0': 0.46290776667664807, 'gradnorm/decoder.layer1': 0.3650751225686607, 'gradnorm/decoder.layer2': 0.1841181206172293, 'gradnorm/max_param': 5.179720401763916, 'epoch': 3.85}

 39%|███▊      | 7350/19090 [3:34:34<5:18:32,  1.63s/it]
 39%|███▊      | 7351/19090 [3:34:37<6:21:09,  1.95s/it]
 39%|███▊      | 7352/19090 [3:34:38<5:42:35,  1.75s/it]
 39%|███▊      | 7353/19090 [3:34:40<5:15:34,  1.61s/it]
 39%|███▊      | 7354/19090 [3:34:43<6:36:41,  2.03s/it]
 39%|███▊      | 7355/19090 [3:34:44<5:53:26,  1.81s/it]
 39%|███▊      | 7356/19090 [3:34:45<5:23:14,  1.65s/it]
 39%|███▊      | 7357/19090 [3:34:48<6:33:43,  2.01s/it]
 39%|███▊      | 7358/19090 [3:34:49<5:51:03,  1.80s/it]
 39%|███▊      | 7359/19090 [3:34:51<5:21:07,  1.64s/it]
 39%|███▊      | 7360/19090 [3:34:53<6:12:00,  1.90s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 8.865967750549316, 'learning_rate': 3.8580109683184376e-05, 'recon_loss': 0.0012998344493098557, 'recon_static': 0.00033691327189444564, 'recon_motion': 0.0009629211708670482, 'recon_norm': 1.2275241255760192, 'recon_norm_std': 0.05864258427172899, 'target_norm': 1.6200873792171477, 'recon_cos': 0.6990893751382827, 'gradnorm/compressor.query': 0.019314138118519716, 'gradnorm/compressor.layer0': 0.40367611392072966, 'gradnorm/compressor.layer1': 8.29356073826184, 'gradnorm/compressor.layer2': 2.647146174603975, 'gradnorm/compressor.layer3': 1.323561198501102, 'gradnorm/compressor.layer4': 0.4994146675998332, 'gradnorm/compressor.layer5': 0.36852548197474155, 'gradnorm/compressor.layer6': 0.1383389519599845, 'gradnorm/compressor.layer7': 0.2689290575754204, 'gradnorm/decoder.layer0': 0.49566359907223756, 'gradnorm/decoder.layer1': 0.3648111471765716, 'gradnorm/decoder.layer2': 0.21039469072257225, 'gradnorm/max_param': 5.846192359924316, 'epoch': 3.86}

 39%|███▊      | 7360/19090 [3:34:53<6:12:00,  1.90s/it]
 39%|███▊      | 7361/19090 [3:34:54<5:36:01,  1.72s/it]
 39%|███▊      | 7362/19090 [3:34:56<5:10:53,  1.59s/it]
 39%|███▊      | 7363/19090 [3:34:58<6:11:06,  1.90s/it]
 39%|███▊      | 7364/19090 [3:35:00<5:35:22,  1.72s/it]
 39%|███▊      | 7365/19090 [3:35:01<5:10:10,  1.59s/it]
 39%|███▊      | 7366/19090 [3:35:04<6:35:41,  2.03s/it]
 39%|███▊      | 7367/19090 [3:35:05<5:52:20,  1.80s/it]
 39%|███▊      | 7368/19090 [3:35:06<5:22:03,  1.65s/it]
 39%|███▊      | 7369/19090 [3:35:09<6:25:55,  1.98s/it]
 39%|███▊      | 7370/19090 [3:35:10<5:45:25,  1.77s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.439833164215088, 'learning_rate': 3.854170619245787e-05, 'recon_loss': 0.0012969542352948339, 'recon_static': 0.00034564669913379477, 'recon_motion': 0.0009513075521681458, 'recon_norm': 1.2279097527265548, 'recon_norm_std': 0.06079830462113023, 'target_norm': 1.6208702236413957, 'recon_cos': 0.7005100026726723, 'gradnorm/compressor.query': 0.020733200042691392, 'gradnorm/compressor.layer0': 0.37902201034527255, 'gradnorm/compressor.layer1': 6.913497229276149, 'gradnorm/compressor.layer2': 2.220274280753372, 'gradnorm/compressor.layer3': 1.2784752923027354, 'gradnorm/compressor.layer4': 0.4987925613140793, 'gradnorm/compressor.layer5': 0.34739914506003794, 'gradnorm/compressor.layer6': 0.1351774874302414, 'gradnorm/compressor.layer7': 0.27420449994221974, 'gradnorm/decoder.layer0': 0.46676348323379807, 'gradnorm/decoder.layer1': 0.35925917210082636, 'gradnorm/decoder.layer2': 0.19130133653538775, 'gradnorm/max_param': 4.872949600219727, 'epoch': 3.86}

 39%|███▊      | 7370/19090 [3:35:10<5:45:25,  1.77s/it]
 39%|███▊      | 7371/19090 [3:35:12<5:17:33,  1.63s/it]
 39%|███▊      | 7372/19090 [3:35:15<6:28:46,  1.99s/it]
 39%|███▊      | 7373/19090 [3:35:16<5:47:25,  1.78s/it]
 39%|███▊      | 7374/19090 [3:35:17<5:18:08,  1.63s/it]
 39%|███▊      | 7375/19090 [3:35:20<6:27:59,  1.99s/it]
 39%|███▊      | 7376/19090 [3:35:21<5:46:52,  1.78s/it]
 39%|███▊      | 7377/19090 [3:35:23<5:18:06,  1.63s/it]
 39%|███▊      | 7378/19090 [3:35:25<6:28:00,  1.99s/it]
 39%|███▊      | 7379/19090 [3:35:27<5:46:47,  1.78s/it]
 39%|███▊      | 7380/19090 [3:35:28<5:18:19,  1.63s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 8.45647144317627, 'learning_rate': 3.850325742488854e-05, 'recon_loss': 0.0013236102618975564, 'recon_static': 0.00035660492430906744, 'recon_motion': 0.0009670053361332976, 'recon_norm': 1.237130454182625, 'recon_norm_std': 0.05813991446048021, 'target_norm': 1.620184764266014, 'recon_cos': 0.6933004662394524, 'gradnorm/compressor.query': 0.01993394429403959, 'gradnorm/compressor.layer0': 0.42850809360169373, 'gradnorm/compressor.layer1': 7.868659964145588, 'gradnorm/compressor.layer2': 2.5402326029806983, 'gradnorm/compressor.layer3': 1.4292992236589066, 'gradnorm/compressor.layer4': 0.5121363908405466, 'gradnorm/compressor.layer5': 0.3672100628352899, 'gradnorm/compressor.layer6': 0.14108872117409427, 'gradnorm/compressor.layer7': 0.2689851813036966, 'gradnorm/decoder.layer0': 0.47335108607316567, 'gradnorm/decoder.layer1': 0.40113387554687957, 'gradnorm/decoder.layer2': 0.20509317433678814, 'gradnorm/max_param': 5.545292377471924, 'epoch': 3.87}

 39%|███▊      | 7380/19090 [3:35:28<5:18:19,  1.63s/it]
 39%|███▊      | 7381/19090 [3:35:31<6:24:28,  1.97s/it]
 39%|███▊      | 7382/19090 [3:35:32<5:44:21,  1.76s/it]
 39%|███▊      | 7383/19090 [3:35:33<5:17:13,  1.63s/it]
 39%|███▊      | 7384/19090 [3:35:35<5:40:33,  1.75s/it]
 39%|███▊      | 7385/19090 [3:35:37<5:13:39,  1.61s/it]
 39%|███▊      | 7386/19090 [3:35:38<4:54:58,  1.51s/it]
 39%|███▊      | 7387/19090 [3:35:40<5:18:51,  1.63s/it]
 39%|███▊      | 7388/19090 [3:35:41<4:58:36,  1.53s/it]
 39%|███▊      | 7389/19090 [3:35:42<4:44:09,  1.46s/it]
 39%|███▊      | 7390/19090 [3:35:45<5:28:09,  1.68s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 8.6488676071167, 'learning_rate': 3.846476350903028e-05, 'recon_loss': 0.001326214984874241, 'recon_static': 0.00036189270686008966, 'recon_motion': 0.000964322236541193, 'recon_norm': 1.2363950937986374, 'recon_norm_std': 0.06029488518834114, 'target_norm': 1.6207093447446823, 'recon_cos': 0.6926115795969963, 'gradnorm/compressor.query': 0.019514050211955253, 'gradnorm/compressor.layer0': 0.4300982679315337, 'gradnorm/compressor.layer1': 8.12475726425272, 'gradnorm/compressor.layer2': 2.457481452196664, 'gradnorm/compressor.layer3': 1.2809181248766972, 'gradnorm/compressor.layer4': 0.5378901407520826, 'gradnorm/compressor.layer5': 0.3756837769982149, 'gradnorm/compressor.layer6': 0.14045816697765554, 'gradnorm/compressor.layer7': 0.2647613200189169, 'gradnorm/decoder.layer0': 0.4748981994159403, 'gradnorm/decoder.layer1': 0.3783993033451095, 'gradnorm/decoder.layer2': 0.19205137448435994, 'gradnorm/max_param': 5.726747512817383, 'epoch': 3.87}

 39%|███▊      | 7390/19090 [3:35:45<5:28:09,  1.68s/it]
 39%|███▊      | 7391/19090 [3:35:46<5:06:09,  1.57s/it]
 39%|███▊      | 7392/19090 [3:35:47<4:49:54,  1.49s/it]
 39%|███▊      | 7393/19090 [3:35:50<5:44:41,  1.77s/it]
 39%|███▊      | 7394/19090 [3:35:51<5:16:59,  1.63s/it]
 39%|███▊      | 7395/19090 [3:35:52<4:57:21,  1.53s/it]
 39%|███▊      | 7396/19090 [3:35:54<4:58:31,  1.53s/it]
 39%|███▊      | 7397/19090 [3:35:55<4:44:28,  1.46s/it]
 39%|███▉      | 7398/19090 [3:35:56<4:34:36,  1.41s/it]
 39%|███▉      | 7399/19090 [3:35:59<5:23:38,  1.66s/it]
 39%|███▉      | 7400/19090 [3:36:00<5:02:07,  1.55s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 7.95681095123291, 'learning_rate': 3.8426224573587903e-05, 'recon_loss': 0.00132569877023343, 'recon_static': 0.00035988289819215423, 'recon_motion': 0.0009658159091486595, 'recon_norm': 1.2371517032384873, 'recon_norm_std': 0.05869135800749063, 'target_norm': 1.6217462569475174, 'recon_cos': 0.693226021528244, 'gradnorm/compressor.query': 0.020111972517144742, 'gradnorm/compressor.layer0': 0.4357758677011053, 'gradnorm/compressor.layer1': 7.447266649455739, 'gradnorm/compressor.layer2': 2.201233938267958, 'gradnorm/compressor.layer3': 1.3750269193615123, 'gradnorm/compressor.layer4': 0.5180730504680412, 'gradnorm/compressor.layer5': 0.38328268409333194, 'gradnorm/compressor.layer6': 0.14107968415520364, 'gradnorm/compressor.layer7': 0.2887444034669841, 'gradnorm/decoder.layer0': 0.484344850647821, 'gradnorm/decoder.layer1': 0.3603792670953632, 'gradnorm/decoder.layer2': 0.19893626212149293, 'gradnorm/max_param': 5.248484134674072, 'epoch': 3.88}

 39%|███▉      | 7400/19090 [3:36:00<5:02:07,  1.55s/it]
 39%|███▉      | 7401/19090 [3:36:01<4:47:09,  1.47s/it]
 39%|███▉      | 7402/19090 [3:36:03<5:17:06,  1.63s/it]
 39%|███▉      | 7403/19090 [3:36:04<4:57:36,  1.53s/it]
 39%|███▉      | 7404/19090 [3:36:06<4:43:33,  1.46s/it]
 39%|███▉      | 7405/19090 [3:36:08<5:21:12,  1.65s/it]
 39%|███▉      | 7406/19090 [3:36:09<5:00:05,  1.54s/it]
 39%|███▉      | 7407/19090 [3:36:10<4:45:03,  1.46s/it]
 39%|███▉      | 7408/19090 [3:36:13<5:32:20,  1.71s/it]
 39%|███▉      | 7409/19090 [3:36:14<5:07:33,  1.58s/it]
 39%|███▉      | 7410/19090 [3:36:15<4:50:31,  1.49s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 8.064882278442383, 'learning_rate': 3.8387640747416764e-05, 'recon_loss': 0.0013086093036690726, 'recon_static': 0.00034982307261088863, 'recon_motion': 0.000958786258706823, 'recon_norm': 1.232967260479927, 'recon_norm_std': 0.05983813060447574, 'target_norm': 1.6196485370397569, 'recon_cos': 0.69670120626688, 'gradnorm/compressor.query': 0.017175484549803423, 'gradnorm/compressor.layer0': 0.3812523404033929, 'gradnorm/compressor.layer1': 7.526587851657662, 'gradnorm/compressor.layer2': 2.365712023889291, 'gradnorm/compressor.layer3': 1.3414285701256663, 'gradnorm/compressor.layer4': 0.4996932250906244, 'gradnorm/compressor.layer5': 0.3582591479970942, 'gradnorm/compressor.layer6': 0.14306335966752703, 'gradnorm/compressor.layer7': 0.2631690706057089, 'gradnorm/decoder.layer0': 0.44631320211956, 'gradnorm/decoder.layer1': 0.37559144744560824, 'gradnorm/decoder.layer2': 0.20757683355513304, 'gradnorm/max_param': 5.305325031280518, 'epoch': 3.88}

 39%|███▉      | 7410/19090 [3:36:15<4:50:31,  1.49s/it]
 39%|███▉      | 7411/19090 [3:36:17<4:49:54,  1.49s/it]
 39%|███▉      | 7412/19090 [3:36:18<4:37:53,  1.43s/it]
 39%|███▉      | 7413/19090 [3:36:19<4:29:32,  1.38s/it]
 39%|███▉      | 7414/19090 [3:36:21<4:51:10,  1.50s/it]
 39%|███▉      | 7415/19090 [3:36:22<4:39:04,  1.43s/it]
 39%|███▉      | 7416/19090 [3:36:24<4:30:19,  1.39s/it]
 39%|███▉      | 7417/19090 [3:36:26<5:23:06,  1.66s/it]
 39%|███▉      | 7418/19090 [3:36:27<5:01:30,  1.55s/it]
 39%|███▉      | 7419/19090 [3:36:29<4:47:24,  1.48s/it]
 39%|███▉      | 7420/19090 [3:36:30<5:09:39,  1.59s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 8.766715049743652, 'learning_rate': 3.834901215952231e-05, 'recon_loss': 0.0013039048993960024, 'recon_static': 0.00034350465793977494, 'recon_motion': 0.0009604002130799927, 'recon_norm': 1.2266351461410523, 'recon_norm_std': 0.05996754849329591, 'target_norm': 1.622140794992447, 'recon_cos': 0.6988341152668, 'gradnorm/compressor.query': 0.018505977545942187, 'gradnorm/compressor.layer0': 0.42106234021075506, 'gradnorm/compressor.layer1': 8.158541419622942, 'gradnorm/compressor.layer2': 2.6557488361375126, 'gradnorm/compressor.layer3': 1.4412821894654808, 'gradnorm/compressor.layer4': 0.5403544658648245, 'gradnorm/compressor.layer5': 0.35936488261701005, 'gradnorm/compressor.layer6': 0.1432591724754546, 'gradnorm/compressor.layer7': 0.2970542617464527, 'gradnorm/decoder.layer0': 0.5128637999748827, 'gradnorm/decoder.layer1': 0.3909582814365615, 'gradnorm/decoder.layer2': 0.19948205190485974, 'gradnorm/max_param': 5.750102996826172, 'epoch': 3.89}

 39%|███▉      | 7420/19090 [3:36:30<5:09:39,  1.59s/it]
 39%|███▉      | 7421/19090 [3:36:32<4:52:15,  1.50s/it]
 39%|███▉      | 7422/19090 [3:36:33<4:39:48,  1.44s/it]
 39%|███▉      | 7423/19090 [3:36:35<5:26:15,  1.68s/it]
 39%|███▉      | 7424/19090 [3:36:37<5:03:32,  1.56s/it]
 39%|███▉      | 7425/19090 [3:36:38<4:47:39,  1.48s/it]
 39%|███▉      | 7426/19090 [3:36:40<5:23:15,  1.66s/it]
 39%|███▉      | 7427/19090 [3:36:41<5:01:30,  1.55s/it]
 39%|███▉      | 7428/19090 [3:36:42<4:45:54,  1.47s/it]
 39%|███▉      | 7429/19090 [3:36:45<5:42:11,  1.76s/it]
 39%|███▉      | 7430/19090 [3:36:46<5:14:59,  1.62s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 8.154970169067383, 'learning_rate': 3.8310338939059644e-05, 'recon_loss': 0.0012950625241501257, 'recon_static': 0.000340897618298186, 'recon_motion': 0.0009541649036691524, 'recon_norm': 1.2261242300271988, 'recon_norm_std': 0.06003332920372486, 'target_norm': 1.6207567483186722, 'recon_cos': 0.700986123085022, 'gradnorm/compressor.query': 0.018707040471006787, 'gradnorm/compressor.layer0': 0.3861385771456647, 'gradnorm/compressor.layer1': 7.5541264859235735, 'gradnorm/compressor.layer2': 2.568055620108091, 'gradnorm/compressor.layer3': 1.324124718312324, 'gradnorm/compressor.layer4': 0.5453182405973023, 'gradnorm/compressor.layer5': 0.342879015538493, 'gradnorm/compressor.layer6': 0.1514645355733052, 'gradnorm/compressor.layer7': 0.2718663581483173, 'gradnorm/decoder.layer0': 0.49638113573691933, 'gradnorm/decoder.layer1': 0.3768482951924483, 'gradnorm/decoder.layer2': 0.2020767656140613, 'gradnorm/max_param': 5.324570178985596, 'epoch': 3.89}

 39%|███▉      | 7430/19090 [3:36:46<5:14:59,  1.62s/it]
 39%|███▉      | 7431/19090 [3:36:47<4:55:17,  1.52s/it]
 39%|███▉      | 7432/19090 [3:36:50<5:47:05,  1.79s/it]
 39%|███▉      | 7433/19090 [3:36:51<5:17:57,  1.64s/it]
 39%|███▉      | 7434/19090 [3:36:52<4:57:20,  1.53s/it]
 39%|███▉      | 7435/19090 [3:36:55<6:00:03,  1.85s/it]
 39%|███▉      | 7436/19090 [3:36:56<5:26:49,  1.68s/it]
 39%|███▉      | 7437/19090 [3:36:58<5:03:35,  1.56s/it]
 39%|███▉      | 7438/19090 [3:37:00<5:34:34,  1.72s/it]
 39%|███▉      | 7439/19090 [3:37:01<5:09:29,  1.59s/it]
 39%|███▉      | 7440/19090 [3:37:02<4:52:17,  1.51s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 9.09360408782959, 'learning_rate': 3.82716212153331e-05, 'recon_loss': 0.001304262742633, 'recon_static': 0.00033981309534283357, 'recon_motion': 0.000964449651655741, 'recon_norm': 1.2250394910573958, 'recon_norm_std': 0.05840337248519063, 'target_norm': 1.6220703303813935, 'recon_cos': 0.6992521971464157, 'gradnorm/compressor.query': 0.018973987455527412, 'gradnorm/compressor.layer0': 0.4108659870597077, 'gradnorm/compressor.layer1': 8.54480022234831, 'gradnorm/compressor.layer2': 2.561758259674781, 'gradnorm/compressor.layer3': 1.373381615935018, 'gradnorm/compressor.layer4': 0.5503833767567303, 'gradnorm/compressor.layer5': 0.3941017494651161, 'gradnorm/compressor.layer6': 0.15487522068154205, 'gradnorm/compressor.layer7': 0.2872723486650088, 'gradnorm/decoder.layer0': 0.5249957050420282, 'gradnorm/decoder.layer1': 0.4204501178459123, 'gradnorm/decoder.layer2': 0.21195390212538393, 'gradnorm/max_param': 6.023525238037109, 'epoch': 3.9}

 39%|███▉      | 7440/19090 [3:37:02<4:52:17,  1.51s/it]
 39%|███▉      | 7441/19090 [3:37:05<5:41:32,  1.76s/it]
 39%|███▉      | 7442/19090 [3:37:06<5:15:13,  1.62s/it]
 39%|███▉      | 7443/19090 [3:37:07<4:56:46,  1.53s/it]
 39%|███▉      | 7444/19090 [3:37:09<5:28:03,  1.69s/it]
 39%|███▉      | 7445/19090 [3:37:11<5:05:38,  1.57s/it]
 39%|███▉      | 7446/19090 [3:37:12<4:49:52,  1.49s/it]
 39%|███▉      | 7447/19090 [3:37:14<5:12:22,  1.61s/it]
 39%|███▉      | 7448/19090 [3:37:15<4:54:37,  1.52s/it]
 39%|███▉      | 7449/19090 [3:37:16<4:42:17,  1.45s/it]
 39%|███▉      | 7450/19090 [3:37:19<5:26:58,  1.69s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 8.072537422180176, 'learning_rate': 3.823285911779582e-05, 'recon_loss': 0.0012982365005882457, 'recon_static': 0.0003429178657825105, 'recon_motion': 0.0009553186086122878, 'recon_norm': 1.2267561823129653, 'recon_norm_std': 0.06001627808436751, 'target_norm': 1.6202210485935211, 'recon_cos': 0.699543434381485, 'gradnorm/compressor.query': 0.018034281392780783, 'gradnorm/compressor.layer0': 0.38841192859497997, 'gradnorm/compressor.layer1': 7.439713861849218, 'gradnorm/compressor.layer2': 2.583897641803063, 'gradnorm/compressor.layer3': 1.3986147975946706, 'gradnorm/compressor.layer4': 0.5675252712605522, 'gradnorm/compressor.layer5': 0.3907444008494999, 'gradnorm/compressor.layer6': 0.14707268514150468, 'gradnorm/compressor.layer7': 0.3023622379022549, 'gradnorm/decoder.layer0': 0.5064016550638926, 'gradnorm/decoder.layer1': 0.380047407016508, 'gradnorm/decoder.layer2': 0.2096854243054959, 'gradnorm/max_param': 5.243072509765625, 'epoch': 3.9}

 39%|███▉      | 7450/19090 [3:37:19<5:26:58,  1.69s/it]
 39%|███▉      | 7451/19090 [3:37:20<5:05:21,  1.57s/it]
 39%|███▉      | 7452/19090 [3:37:21<4:50:48,  1.50s/it]
 39%|███▉      | 7453/19090 [3:37:23<5:12:25,  1.61s/it]
 39%|███▉      | 7454/19090 [3:37:24<4:54:11,  1.52s/it]
 39%|███▉      | 7455/19090 [3:37:26<4:41:08,  1.45s/it]
 39%|███▉      | 7456/19090 [3:37:28<5:05:19,  1.57s/it]
 39%|███▉      | 7457/19090 [3:37:29<4:48:49,  1.49s/it]
 39%|███▉      | 7458/19090 [3:37:30<4:36:54,  1.43s/it]
 39%|███▉      | 7459/19090 [3:37:32<5:04:09,  1.57s/it]
 39%|███▉      | 7460/19090 [3:37:33<4:48:08,  1.49s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 8.492840766906738, 'learning_rate': 3.819405277604929e-05, 'recon_loss': 0.001311362854903564, 'recon_static': 0.00035508737710188143, 'recon_motion': 0.0009562754901708103, 'recon_norm': 1.234294417500496, 'recon_norm_std': 0.05969798294827342, 'target_norm': 1.6200573235750197, 'recon_cos': 0.6964567959308624, 'gradnorm/compressor.query': 0.019069928409612646, 'gradnorm/compressor.layer0': 0.4116189035739756, 'gradnorm/compressor.layer1': 7.913418742712265, 'gradnorm/compressor.layer2': 2.5249435607343385, 'gradnorm/compressor.layer3': 1.3942593007494841, 'gradnorm/compressor.layer4': 0.5747958287273546, 'gradnorm/compressor.layer5': 0.39586805726718577, 'gradnorm/compressor.layer6': 0.15724824080673397, 'gradnorm/compressor.layer7': 0.29329124502751536, 'gradnorm/decoder.layer0': 0.49102549152278774, 'gradnorm/decoder.layer1': 0.36379136397621836, 'gradnorm/decoder.layer2': 0.21390202208103132, 'gradnorm/max_param': 5.577728748321533, 'epoch': 3.91}

 39%|███▉      | 7460/19090 [3:37:33<4:48:08,  1.49s/it]
 39%|███▉      | 7461/19090 [3:37:35<4:36:42,  1.43s/it]
 39%|███▉      | 7462/19090 [3:37:37<5:24:50,  1.68s/it]
 39%|███▉      | 7463/19090 [3:37:38<5:02:22,  1.56s/it]
 39%|███▉      | 7464/19090 [3:37:40<4:46:35,  1.48s/it]
 39%|███▉      | 7465/19090 [3:37:41<5:06:12,  1.58s/it]
 39%|███▉      | 7466/19090 [3:37:43<4:49:12,  1.49s/it]
 39%|███▉      | 7467/19090 [3:37:44<4:37:05,  1.43s/it]
 39%|███▉      | 7468/19090 [3:37:46<5:14:09,  1.62s/it]
 39%|███▉      | 7469/19090 [3:37:47<4:54:50,  1.52s/it]
 39%|███▉      | 7470/19090 [3:37:49<4:41:28,  1.45s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 8.022507667541504, 'learning_rate': 3.815520231984294e-05, 'recon_loss': 0.001318612473551184, 'recon_static': 0.00035826564489980226, 'recon_motion': 0.000960346836654935, 'recon_norm': 1.2337610572576523, 'recon_norm_std': 0.05869328398257494, 'target_norm': 1.6200634747743607, 'recon_cos': 0.6940493330359458, 'gradnorm/compressor.query': 0.017292795641573166, 'gradnorm/compressor.layer0': 0.41859535236913425, 'gradnorm/compressor.layer1': 7.348505566277893, 'gradnorm/compressor.layer2': 2.68143959708879, 'gradnorm/compressor.layer3': 1.4222263279415182, 'gradnorm/compressor.layer4': 0.5253584047102113, 'gradnorm/compressor.layer5': 0.41297931061547993, 'gradnorm/compressor.layer6': 0.15814358568091405, 'gradnorm/compressor.layer7': 0.2720921526137228, 'gradnorm/decoder.layer0': 0.49237962030982224, 'gradnorm/decoder.layer1': 0.3759170088728746, 'gradnorm/decoder.layer2': 0.20588861797020666, 'gradnorm/max_param': 5.178179740905762, 'epoch': 3.91}

 39%|███▉      | 7470/19090 [3:37:49<4:41:28,  1.45s/it]
 39%|███▉      | 7471/19090 [3:37:51<5:20:47,  1.66s/it]
 39%|███▉      | 7472/19090 [3:37:52<4:59:44,  1.55s/it]
 39%|███▉      | 7473/19090 [3:37:53<4:44:59,  1.47s/it]
 39%|███▉      | 7474/19090 [3:37:55<5:13:48,  1.62s/it]
 39%|███▉      | 7475/19090 [3:37:57<4:54:30,  1.52s/it]
 39%|███▉      | 7476/19090 [3:37:58<4:40:53,  1.45s/it]
 39%|███▉      | 7477/19090 [3:38:00<5:16:43,  1.64s/it]
 39%|███▉      | 7478/19090 [3:38:01<4:56:25,  1.53s/it]
 39%|███▉      | 7479/19090 [3:38:02<4:41:48,  1.46s/it]
 39%|███▉      | 7480/19090 [3:38:05<5:43:27,  1.78s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 9.103825569152832, 'learning_rate': 3.811630787907372e-05, 'recon_loss': 0.0013205023773480208, 'recon_static': 0.0003651412014733069, 'recon_motion': 0.0009553611322189681, 'recon_norm': 1.2356917411088943, 'recon_norm_std': 0.05896203415468335, 'target_norm': 1.6212078660726548, 'recon_cos': 0.6942711889743804, 'gradnorm/compressor.query': 0.022634945247433483, 'gradnorm/compressor.layer0': 0.4558607747008238, 'gradnorm/compressor.layer1': 8.46513666861672, 'gradnorm/compressor.layer2': 2.8050472273225213, 'gradnorm/compressor.layer3': 1.4687726445177176, 'gradnorm/compressor.layer4': 0.5542125280185322, 'gradnorm/compressor.layer5': 0.4158340021821162, 'gradnorm/compressor.layer6': 0.14943748489995645, 'gradnorm/compressor.layer7': 0.28580727139113, 'gradnorm/decoder.layer0': 0.4696738992778956, 'gradnorm/decoder.layer1': 0.3715899111981179, 'gradnorm/decoder.layer2': 0.21078147239192602, 'gradnorm/max_param': 5.966458797454834, 'epoch': 3.92}

 39%|███▉      | 7480/19090 [3:38:05<5:43:27,  1.78s/it]
 39%|███▉      | 7481/19090 [3:38:06<5:15:02,  1.63s/it]
 39%|███▉      | 7482/19090 [3:38:08<4:55:06,  1.53s/it]
 39%|███▉      | 7483/19090 [3:38:10<5:54:40,  1.83s/it]
 39%|███▉      | 7484/19090 [3:38:11<5:22:34,  1.67s/it]
 39%|███▉      | 7485/19090 [3:38:13<5:00:14,  1.55s/it]
 39%|███▉      | 7486/19090 [3:38:15<6:06:12,  1.89s/it]
 39%|███▉      | 7487/19090 [3:38:17<5:30:55,  1.71s/it]
 39%|███▉      | 7488/19090 [3:38:18<5:06:23,  1.58s/it]
 39%|███▉      | 7489/19090 [3:38:20<5:41:23,  1.77s/it]
 39%|███▉      | 7490/19090 [3:38:21<5:13:32,  1.62s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 9.190781593322754, 'learning_rate': 3.807736958378558e-05, 'recon_loss': 0.0013204465736635028, 'recon_static': 0.0003556113377271686, 'recon_motion': 0.0009648351580835879, 'recon_norm': 1.2319502174854278, 'recon_norm_std': 0.057843003142625096, 'target_norm': 1.619895339012146, 'recon_cos': 0.6937068298459053, 'gradnorm/compressor.query': 0.02088471119525844, 'gradnorm/compressor.layer0': 0.44371905420420393, 'gradnorm/compressor.layer1': 8.54534039153658, 'gradnorm/compressor.layer2': 2.821848358670656, 'gradnorm/compressor.layer3': 1.4927380729458188, 'gradnorm/compressor.layer4': 0.5662235096762408, 'gradnorm/compressor.layer5': 0.39568040637579666, 'gradnorm/compressor.layer6': 0.1470510077503257, 'gradnorm/compressor.layer7': 0.29057056522232755, 'gradnorm/decoder.layer0': 0.5360378327166483, 'gradnorm/decoder.layer1': 0.38381591004094656, 'gradnorm/decoder.layer2': 0.20323318571486337, 'gradnorm/max_param': 6.022993564605713, 'epoch': 3.92}

 39%|███▉      | 7490/19090 [3:38:21<5:13:32,  1.62s/it]
 39%|███▉      | 7491/19090 [3:38:23<4:54:08,  1.52s/it]
 39%|███▉      | 7492/19090 [3:38:25<5:43:09,  1.78s/it]
 39%|███▉      | 7493/19090 [3:38:26<5:14:58,  1.63s/it]
 39%|███▉      | 7494/19090 [3:38:28<4:55:19,  1.53s/it]
 39%|███▉      | 7495/19090 [3:38:30<5:43:30,  1.78s/it]
 39%|███▉      | 7496/19090 [3:38:31<5:15:09,  1.63s/it]
 39%|███▉      | 7497/19090 [3:38:33<4:55:21,  1.53s/it]
 39%|███▉      | 7498/19090 [3:38:35<5:53:35,  1.83s/it]
 39%|███▉      | 7499/19090 [3:38:36<5:22:17,  1.67s/it]
 39%|███▉      | 7500/19090 [3:38:38<5:00:34,  1.56s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 8.438814163208008, 'learning_rate': 3.803838756416915e-05, 'recon_loss': 0.001317557558650151, 'recon_static': 0.0003514366275339853, 'recon_motion': 0.0009661209740443155, 'recon_norm': 1.2290347158908843, 'recon_norm_std': 0.058029474783688785, 'target_norm': 1.6240121215581893, 'recon_cos': 0.6961495235562325, 'gradnorm/compressor.query': 0.018709631424566792, 'gradnorm/compressor.layer0': 0.4265417429306699, 'gradnorm/compressor.layer1': 7.856471471464374, 'gradnorm/compressor.layer2': 2.537717492314852, 'gradnorm/compressor.layer3': 1.3702274639752048, 'gradnorm/compressor.layer4': 0.5465625415276347, 'gradnorm/compressor.layer5': 0.37885364678575945, 'gradnorm/compressor.layer6': 0.15331210582175514, 'gradnorm/compressor.layer7': 0.29072391047157525, 'gradnorm/decoder.layer0': 0.48951425132618387, 'gradnorm/decoder.layer1': 0.3978262307553565, 'gradnorm/decoder.layer2': 0.20356228416204208, 'gradnorm/max_param': 5.5369720458984375, 'epoch': 3.93}

 39%|███▉      | 7500/19090 [3:38:38<5:00:34,  1.56s/it]
 39%|███▉      | 7501/19090 [3:38:42<7:17:25,  2.26s/it]
 39%|███▉      | 7502/19090 [3:38:43<6:20:41,  1.97s/it]
 39%|███▉      | 7503/19090 [3:38:44<5:41:03,  1.77s/it]
 39%|███▉      | 7504/19090 [3:38:45<5:13:31,  1.62s/it]
 39%|███▉      | 7505/19090 [3:38:47<4:54:24,  1.52s/it]
 39%|███▉      | 7506/19090 [3:38:48<4:40:34,  1.45s/it]
 39%|███▉      | 7507/19090 [3:38:49<4:31:12,  1.40s/it]
 39%|███▉      | 7508/19090 [3:38:51<4:24:40,  1.37s/it]
 39%|███▉      | 7509/19090 [3:38:52<4:20:02,  1.35s/it]
 39%|███▉      | 7510/19090 [3:38:53<4:17:00,  1.33s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 9.445114135742188, 'learning_rate': 3.799936195056126e-05, 'recon_loss': 0.0013071905501419677, 'recon_static': 0.00034599937644088643, 'recon_motion': 0.0009611911984393374, 'recon_norm': 1.2271704256534577, 'recon_norm_std': 0.05858559338375926, 'target_norm': 1.618879309296608, 'recon_cos': 0.6967474445700645, 'gradnorm/compressor.query': 0.018417690245665665, 'gradnorm/compressor.layer0': 0.42434773038686663, 'gradnorm/compressor.layer1': 8.815149220173518, 'gradnorm/compressor.layer2': 2.83987649219218, 'gradnorm/compressor.layer3': 1.4509402843730435, 'gradnorm/compressor.layer4': 0.6004199038869863, 'gradnorm/compressor.layer5': 0.40371745509850837, 'gradnorm/compressor.layer6': 0.15549699416544221, 'gradnorm/compressor.layer7': 0.31333646882209715, 'gradnorm/decoder.layer0': 0.5308927569023332, 'gradnorm/decoder.layer1': 0.4220853563670072, 'gradnorm/decoder.layer2': 0.21661729506103583, 'gradnorm/max_param': 6.214766979217529, 'epoch': 3.93}

 39%|███▉      | 7510/19090 [3:38:53<4:17:00,  1.33s/it]
 39%|███▉      | 7511/19090 [3:38:55<4:15:32,  1.32s/it]
 39%|███▉      | 7512/19090 [3:38:56<4:14:05,  1.32s/it]
 39%|███▉      | 7513/19090 [3:38:57<4:23:16,  1.36s/it]
 39%|███▉      | 7514/19090 [3:38:59<4:18:58,  1.34s/it]
 39%|███▉      | 7515/19090 [3:39:00<4:15:52,  1.33s/it]
 39%|███▉      | 7516/19090 [3:39:02<5:01:20,  1.56s/it]
 39%|███▉      | 7517/19090 [3:39:03<4:45:30,  1.48s/it]
 39%|███▉      | 7518/19090 [3:39:05<4:35:00,  1.43s/it]
 39%|███▉      | 7519/19090 [3:39:06<4:58:10,  1.55s/it]
 39%|███▉      | 7520/19090 [3:39:08<4:43:44,  1.47s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 8.670929908752441, 'learning_rate': 3.796029287344444e-05, 'recon_loss': 0.0013191770703997463, 'recon_static': 0.000355640024645254, 'recon_motion': 0.0009635370413889177, 'recon_norm': 1.2319239497184753, 'recon_norm_std': 0.05823655212298036, 'target_norm': 1.6213720351457597, 'recon_cos': 0.6947680324316025, 'gradnorm/compressor.query': 0.016977310073521684, 'gradnorm/compressor.layer0': 0.3823952444465666, 'gradnorm/compressor.layer1': 8.06436017482459, 'gradnorm/compressor.layer2': 2.6051339959591053, 'gradnorm/compressor.layer3': 1.4384028046157362, 'gradnorm/compressor.layer4': 0.5816947937406978, 'gradnorm/compressor.layer5': 0.39200873385158663, 'gradnorm/compressor.layer6': 0.16105917201528527, 'gradnorm/compressor.layer7': 0.30352621808707436, 'gradnorm/decoder.layer0': 0.5640790287037105, 'gradnorm/decoder.layer1': 0.4224943982830798, 'gradnorm/decoder.layer2': 0.20515093646173954, 'gradnorm/max_param': 5.684536457061768, 'epoch': 3.94}

 39%|███▉      | 7520/19090 [3:39:08<4:43:44,  1.47s/it]
 39%|███▉      | 7521/19090 [3:39:09<4:33:25,  1.42s/it]
 39%|███▉      | 7522/19090 [3:39:11<5:09:21,  1.60s/it]
 39%|███▉      | 7523/19090 [3:39:12<4:51:04,  1.51s/it]
 39%|███▉      | 7524/19090 [3:39:14<4:38:30,  1.44s/it]
 39%|███▉      | 7525/19090 [3:39:15<4:50:36,  1.51s/it]
 39%|███▉      | 7526/19090 [3:39:17<4:38:33,  1.45s/it]
 39%|███▉      | 7527/19090 [3:39:18<4:30:10,  1.40s/it]
 39%|███▉      | 7528/19090 [3:39:20<4:59:59,  1.56s/it]
 39%|███▉      | 7529/19090 [3:39:21<5:00:24,  1.56s/it]
 39%|███▉      | 7530/19090 [3:39:23<4:45:37,  1.48s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 9.322575569152832, 'learning_rate': 3.792118046344661e-05, 'recon_loss': 0.0013087365950923413, 'recon_static': 0.00035505485357134603, 'recon_motion': 0.0009536817538901232, 'recon_norm': 1.2309128224849701, 'recon_norm_std': 0.05889273742213845, 'target_norm': 1.620209464430809, 'recon_cos': 0.696986836194992, 'gradnorm/compressor.query': 0.02201518406897684, 'gradnorm/compressor.layer0': 0.43470520326283424, 'gradnorm/compressor.layer1': 8.703547085674817, 'gradnorm/compressor.layer2': 2.7345145053698516, 'gradnorm/compressor.layer3': 1.5441038181850397, 'gradnorm/compressor.layer4': 0.5870520608811057, 'gradnorm/compressor.layer5': 0.3916020000959506, 'gradnorm/compressor.layer6': 0.1672359995595113, 'gradnorm/compressor.layer7': 0.3082591186058001, 'gradnorm/decoder.layer0': 0.5322507240340655, 'gradnorm/decoder.layer1': 0.40087420406209007, 'gradnorm/decoder.layer2': 0.20600793859603883, 'gradnorm/max_param': 6.134688377380371, 'epoch': 3.94}

 39%|███▉      | 7530/19090 [3:39:23<4:45:37,  1.48s/it]
 39%|███▉      | 7531/19090 [3:39:25<5:24:29,  1.68s/it]
 39%|███▉      | 7532/19090 [3:39:26<5:01:47,  1.57s/it]
 39%|███▉      | 7533/19090 [3:39:27<4:45:35,  1.48s/it]
 39%|███▉      | 7534/19090 [3:39:30<5:25:16,  1.69s/it]
 39%|███▉      | 7535/19090 [3:39:31<5:02:12,  1.57s/it]
 39%|███▉      | 7536/19090 [3:39:32<4:45:48,  1.48s/it]
 39%|███▉      | 7537/19090 [3:39:35<5:37:18,  1.75s/it]
 39%|███▉      | 7538/19090 [3:39:36<5:10:11,  1.61s/it]
 39%|███▉      | 7539/19090 [3:39:37<4:51:11,  1.51s/it]
 39%|███▉      | 7540/19090 [3:39:40<6:15:13,  1.95s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 10.218048095703125, 'learning_rate': 3.788202485134052e-05, 'recon_loss': 0.0013036904158070684, 'recon_static': 0.0003531557282258291, 'recon_motion': 0.0009505346999503673, 'recon_norm': 1.231096538901329, 'recon_norm_std': 0.05913525614887476, 'target_norm': 1.6172829031944276, 'recon_cos': 0.6970110341906548, 'gradnorm/compressor.query': 0.020840944884042197, 'gradnorm/compressor.layer0': 0.4749738761343631, 'gradnorm/compressor.layer1': 9.465609089336455, 'gradnorm/compressor.layer2': 3.2550542118462644, 'gradnorm/compressor.layer3': 1.6456766738051618, 'gradnorm/compressor.layer4': 0.6424970256795948, 'gradnorm/compressor.layer5': 0.45629619926238274, 'gradnorm/compressor.layer6': 0.17023933834072383, 'gradnorm/compressor.layer7': 0.3353024957271875, 'gradnorm/decoder.layer0': 0.5231881757756222, 'gradnorm/decoder.layer1': 0.4346827500863338, 'gradnorm/decoder.layer2': 0.2366592829827403, 'gradnorm/max_param': 6.672278881072998, 'epoch': 3.95}

 39%|███▉      | 7540/19090 [3:39:40<6:15:13,  1.95s/it]
 40%|███▉      | 7541/19090 [3:39:41<5:36:57,  1.75s/it]
 40%|███▉      | 7542/19090 [3:39:43<5:10:01,  1.61s/it]
 40%|███▉      | 7543/19090 [3:39:46<6:28:24,  2.02s/it]
 40%|███▉      | 7544/19090 [3:39:47<5:45:55,  1.80s/it]
 40%|███▉      | 7545/19090 [3:39:48<5:16:30,  1.64s/it]
 40%|███▉      | 7546/19090 [3:39:51<6:07:07,  1.91s/it]
 40%|███▉      | 7547/19090 [3:39:52<5:31:03,  1.72s/it]
 40%|███▉      | 7548/19090 [3:39:53<5:05:51,  1.59s/it]
 40%|███▉      | 7549/19090 [3:39:56<5:52:12,  1.83s/it]
 40%|███▉      | 7550/19090 [3:39:57<5:20:53,  1.67s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 9.880849838256836, 'learning_rate': 3.78428261680434e-05, 'recon_loss': 0.001320529551594518, 'recon_static': 0.0003484651322651189, 'recon_motion': 0.0009720643938635476, 'recon_norm': 1.2276070147752762, 'recon_norm_std': 0.05774599788710475, 'target_norm': 1.6212809592485429, 'recon_cos': 0.694147863984108, 'gradnorm/compressor.query': 0.020522752922453324, 'gradnorm/compressor.layer0': 0.47435734968558185, 'gradnorm/compressor.layer1': 9.207767889230979, 'gradnorm/compressor.layer2': 3.0300402242678484, 'gradnorm/compressor.layer3': 1.5113766305277516, 'gradnorm/compressor.layer4': 0.5977353561008405, 'gradnorm/compressor.layer5': 0.43433234739496646, 'gradnorm/compressor.layer6': 0.1616017886153998, 'gradnorm/compressor.layer7': 0.3200309958571014, 'gradnorm/decoder.layer0': 0.5159431112020112, 'gradnorm/decoder.layer1': 0.40879073126717025, 'gradnorm/decoder.layer2': 0.22268449035375165, 'gradnorm/max_param': 6.490538120269775, 'epoch': 3.95}

 40%|███▉      | 7550/19090 [3:39:57<5:20:53,  1.67s/it]
 40%|███▉      | 7551/19090 [3:39:58<4:58:41,  1.55s/it]
 40%|███▉      | 7552/19090 [3:40:01<5:59:28,  1.87s/it]
 40%|███▉      | 7553/19090 [3:40:02<5:25:38,  1.69s/it]
 40%|███▉      | 7554/19090 [3:40:03<5:01:41,  1.57s/it]
 40%|███▉      | 7555/19090 [3:40:06<6:16:56,  1.96s/it]
 40%|███▉      | 7556/19090 [3:40:08<5:37:51,  1.76s/it]
 40%|███▉      | 7557/19090 [3:40:09<5:10:16,  1.61s/it]
 40%|███▉      | 7558/19090 [3:40:12<6:30:20,  2.03s/it]
 40%|███▉      | 7559/19090 [3:40:13<5:46:59,  1.81s/it]
 40%|███▉      | 7560/19090 [3:40:14<5:17:01,  1.65s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 9.612464904785156, 'learning_rate': 3.7803584544616464e-05, 'recon_loss': 0.0013221830537077039, 'recon_static': 0.00035221626530983484, 'recon_motion': 0.0009699667760287411, 'recon_norm': 1.2286969035863877, 'recon_norm_std': 0.058394478913396594, 'target_norm': 1.623492133617401, 'recon_cos': 0.6948955655097961, 'gradnorm/compressor.query': 0.0225160805559664, 'gradnorm/compressor.layer0': 0.4773001668110158, 'gradnorm/compressor.layer1': 8.997259782497668, 'gradnorm/compressor.layer2': 2.801809210054878, 'gradnorm/compressor.layer3': 1.5042668690508258, 'gradnorm/compressor.layer4': 0.5951983705389371, 'gradnorm/compressor.layer5': 0.43347642359370836, 'gradnorm/compressor.layer6': 0.1672367847742682, 'gradnorm/compressor.layer7': 0.3094996302445318, 'gradnorm/decoder.layer0': 0.49093035956810727, 'gradnorm/decoder.layer1': 0.38887122327931906, 'gradnorm/decoder.layer2': 0.2221918784889137, 'gradnorm/max_param': 6.341689586639404, 'epoch': 3.96}

 40%|███▉      | 7560/19090 [3:40:14<5:17:01,  1.65s/it]
 40%|███▉      | 7561/19090 [3:40:18<6:49:28,  2.13s/it]
 40%|███▉      | 7562/19090 [3:40:19<6:00:41,  1.88s/it]
 40%|███▉      | 7563/19090 [3:40:20<5:26:32,  1.70s/it]
 40%|███▉      | 7564/19090 [3:40:23<6:26:54,  2.01s/it]
 40%|███▉      | 7565/19090 [3:40:24<5:44:56,  1.80s/it]
 40%|███▉      | 7566/19090 [3:40:26<5:15:14,  1.64s/it]
 40%|███▉      | 7567/19090 [3:40:28<6:16:22,  1.96s/it]
 40%|███▉      | 7568/19090 [3:40:30<5:37:10,  1.76s/it]
 40%|███▉      | 7569/19090 [3:40:31<5:09:48,  1.61s/it]
 40%|███▉      | 7570/19090 [3:40:33<6:08:50,  1.92s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 9.770397186279297, 'learning_rate': 3.776430011226452e-05, 'recon_loss': 0.001321766444016248, 'recon_static': 0.0003580200951546431, 'recon_motion': 0.0009637463866965845, 'recon_norm': 1.2306274890899658, 'recon_norm_std': 0.05783338416367769, 'target_norm': 1.6226982533931733, 'recon_cos': 0.6945464685559273, 'gradnorm/compressor.query': 0.025328879748582618, 'gradnorm/compressor.layer0': 0.5129209766573556, 'gradnorm/compressor.layer1': 9.091161405356424, 'gradnorm/compressor.layer2': 2.983617396834071, 'gradnorm/compressor.layer3': 1.553917085313491, 'gradnorm/compressor.layer4': 0.6385659179521352, 'gradnorm/compressor.layer5': 0.4435533611726269, 'gradnorm/compressor.layer6': 0.1761176228778681, 'gradnorm/compressor.layer7': 0.29886700495688284, 'gradnorm/decoder.layer0': 0.5245444989796727, 'gradnorm/decoder.layer1': 0.4234334452811498, 'gradnorm/decoder.layer2': 0.22757042107996442, 'gradnorm/max_param': 6.406753063201904, 'epoch': 3.97}

 40%|███▉      | 7570/19090 [3:40:33<6:08:50,  1.92s/it]
 40%|███▉      | 7571/19090 [3:40:35<5:32:02,  1.73s/it]
 40%|███▉      | 7572/19090 [3:40:36<5:06:33,  1.60s/it]
 40%|███▉      | 7573/19090 [3:40:38<5:54:42,  1.85s/it]
 40%|███▉      | 7574/19090 [3:40:40<5:35:22,  1.75s/it]
 40%|███▉      | 7575/19090 [3:40:41<5:08:41,  1.61s/it]
 40%|███▉      | 7576/19090 [3:40:43<5:41:10,  1.78s/it]
 40%|███▉      | 7577/19090 [3:40:45<5:38:22,  1.76s/it]
 40%|███▉      | 7578/19090 [3:40:46<5:10:39,  1.62s/it]
 40%|███▉      | 7579/19090 [3:40:49<5:55:34,  1.85s/it]
 40%|███▉      | 7580/19090 [3:40:50<5:37:44,  1.76s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 10.998764038085938, 'learning_rate': 3.772497300233548e-05, 'recon_loss': 0.0013254648278234527, 'recon_static': 0.00036054599186172707, 'recon_motion': 0.0009649188272305764, 'recon_norm': 1.2319595485925674, 'recon_norm_std': 0.05789202321320772, 'target_norm': 1.6212832391262055, 'recon_cos': 0.6928280189633369, 'gradnorm/compressor.query': 0.025262332002626432, 'gradnorm/compressor.layer0': 0.5295431261033756, 'gradnorm/compressor.layer1': 10.37593056055979, 'gradnorm/compressor.layer2': 3.0246090849194296, 'gradnorm/compressor.layer3': 1.619554271249149, 'gradnorm/compressor.layer4': 0.6502001142568027, 'gradnorm/compressor.layer5': 0.43940219160131566, 'gradnorm/compressor.layer6': 0.1674738957752919, 'gradnorm/compressor.layer7': 0.3301834863746961, 'gradnorm/decoder.layer0': 0.562951674846933, 'gradnorm/decoder.layer1': 0.3827923944604134, 'gradnorm/decoder.layer2': 0.210373088050536, 'gradnorm/max_param': 7.314912796020508, 'epoch': 3.97}

 40%|███▉      | 7580/19090 [3:40:50<5:37:44,  1.76s/it]
 40%|███▉      | 7581/19090 [3:40:52<5:21:01,  1.67s/it]
 40%|███▉      | 7582/19090 [3:40:54<5:48:43,  1.82s/it]
 40%|███▉      | 7583/19090 [3:40:56<5:36:37,  1.76s/it]
 40%|███▉      | 7584/19090 [3:40:57<5:18:29,  1.66s/it]
 40%|███▉      | 7585/19090 [3:40:59<5:53:00,  1.84s/it]
 40%|███▉      | 7586/19090 [3:41:01<5:20:44,  1.67s/it]
 40%|███▉      | 7587/19090 [3:41:02<4:59:31,  1.56s/it]
 40%|███▉      | 7588/19090 [3:41:04<5:41:52,  1.78s/it]
 40%|███▉      | 7589/19090 [3:41:06<5:13:57,  1.64s/it]
 40%|███▉      | 7590/19090 [3:41:07<5:22:09,  1.68s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 9.8739595413208, 'learning_rate': 3.7685603346320004e-05, 'recon_loss': 0.001320104583282955, 'recon_static': 0.00035034740285482257, 'recon_motion': 0.0009697571949800476, 'recon_norm': 1.2260392785072327, 'recon_norm_std': 0.056966683734208344, 'target_norm': 1.6232797652482986, 'recon_cos': 0.6953030824661255, 'gradnorm/compressor.query': 0.022565481756811185, 'gradnorm/compressor.layer0': 0.48385144210010006, 'gradnorm/compressor.layer1': 9.218600617831006, 'gradnorm/compressor.layer2': 2.881830808143097, 'gradnorm/compressor.layer3': 1.6520025383465642, 'gradnorm/compressor.layer4': 0.6058581668883526, 'gradnorm/compressor.layer5': 0.4414633148575982, 'gradnorm/compressor.layer6': 0.17517121211964237, 'gradnorm/compressor.layer7': 0.3180019853360503, 'gradnorm/decoder.layer0': 0.5782710483726218, 'gradnorm/decoder.layer1': 0.4147179472093916, 'gradnorm/decoder.layer2': 0.2089051247896693, 'gradnorm/max_param': 6.497997283935547, 'epoch': 3.98}

 40%|███▉      | 7590/19090 [3:41:07<5:22:09,  1.68s/it]
 40%|███▉      | 7591/19090 [3:41:09<5:27:41,  1.71s/it]
 40%|███▉      | 7592/19090 [3:41:11<5:15:54,  1.65s/it]
 40%|███▉      | 7593/19090 [3:41:12<4:57:29,  1.55s/it]
 40%|███▉      | 7594/19090 [3:41:14<5:34:46,  1.75s/it]
 40%|███▉      | 7595/19090 [3:41:15<5:11:59,  1.63s/it]
 40%|███▉      | 7596/19090 [3:41:17<5:04:02,  1.59s/it]
 40%|███▉      | 7597/19090 [3:41:19<5:44:47,  1.80s/it]
 40%|███▉      | 7598/19090 [3:41:21<5:15:14,  1.65s/it]
 40%|███▉      | 7599/19090 [3:41:22<5:03:24,  1.58s/it]
 40%|███▉      | 7600/19090 [3:41:24<5:43:39,  1.79s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 10.082406997680664, 'learning_rate': 3.764619127585093e-05, 'recon_loss': 0.0013270375988213346, 'recon_static': 0.0003658145149529446, 'recon_motion': 0.0009612231209757738, 'recon_norm': 1.2345796823501587, 'recon_norm_std': 0.05732184164226055, 'target_norm': 1.621002396941185, 'recon_cos': 0.6923257872462273, 'gradnorm/compressor.query': 0.022482942862211943, 'gradnorm/compressor.layer0': 0.5229894229027241, 'gradnorm/compressor.layer1': 9.424694948799653, 'gradnorm/compressor.layer2': 2.9529496398207304, 'gradnorm/compressor.layer3': 1.5863915906739994, 'gradnorm/compressor.layer4': 0.6249172632767594, 'gradnorm/compressor.layer5': 0.47272557152220257, 'gradnorm/compressor.layer6': 0.1888451110655706, 'gradnorm/compressor.layer7': 0.3288568781315104, 'gradnorm/decoder.layer0': 0.5734983583425158, 'gradnorm/decoder.layer1': 0.4315462138633839, 'gradnorm/decoder.layer2': 0.21571299961730975, 'gradnorm/max_param': 6.642577171325684, 'epoch': 3.98}

 40%|███▉      | 7600/19090 [3:41:24<5:43:39,  1.79s/it]
 40%|███▉      | 7601/19090 [3:41:26<5:24:48,  1.70s/it]
 40%|███▉      | 7602/19090 [3:41:27<5:01:38,  1.58s/it]
 40%|███▉      | 7603/19090 [3:41:30<6:13:09,  1.95s/it]
 40%|███▉      | 7604/19090 [3:41:31<5:36:06,  1.76s/it]
 40%|███▉      | 7605/19090 [3:41:32<5:09:18,  1.62s/it]
 40%|███▉      | 7606/19090 [3:41:35<5:45:30,  1.81s/it]
 40%|███▉      | 7607/19090 [3:41:36<5:15:47,  1.65s/it]
 40%|███▉      | 7608/19090 [3:41:37<5:01:05,  1.57s/it]
 40%|███▉      | 7609/19090 [3:41:40<6:10:30,  1.94s/it]
 40%|███▉      | 7610/19090 [3:41:41<5:33:41,  1.74s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 10.343974113464355, 'learning_rate': 3.760673692270296e-05, 'recon_loss': 0.0013183692062739283, 'recon_static': 0.0003620471397880465, 'recon_motion': 0.0009563220301060937, 'recon_norm': 1.2337694853544234, 'recon_norm_std': 0.05775479367002845, 'target_norm': 1.6192888379096986, 'recon_cos': 0.6942527532577515, 'gradnorm/compressor.query': 0.02315072357056107, 'gradnorm/compressor.layer0': 0.5132155040979073, 'gradnorm/compressor.layer1': 9.624406276077096, 'gradnorm/compressor.layer2': 3.1868138790310554, 'gradnorm/compressor.layer3': 1.6287250372261142, 'gradnorm/compressor.layer4': 0.6349251884035096, 'gradnorm/compressor.layer5': 0.45935314866080235, 'gradnorm/compressor.layer6': 0.18182549246875143, 'gradnorm/compressor.layer7': 0.33627478722674853, 'gradnorm/decoder.layer0': 0.5474301109385822, 'gradnorm/decoder.layer1': 0.4235945554617974, 'gradnorm/decoder.layer2': 0.23823188833281328, 'gradnorm/max_param': 6.783638000488281, 'epoch': 3.99}

 40%|███▉      | 7610/19090 [3:41:41<5:33:41,  1.74s/it]
 40%|███▉      | 7611/19090 [3:41:43<5:09:48,  1.62s/it]
 40%|███▉      | 7612/19090 [3:41:45<6:00:53,  1.89s/it]
 40%|███▉      | 7613/19090 [3:41:47<5:30:59,  1.73s/it]
 40%|███▉      | 7614/19090 [3:41:48<5:17:05,  1.66s/it]
 40%|███▉      | 7615/19090 [3:41:50<5:54:25,  1.85s/it]
 40%|███▉      | 7616/19090 [3:41:52<5:36:36,  1.76s/it]
 40%|███▉      | 7617/19090 [3:41:54<5:27:34,  1.71s/it]
 40%|███▉      | 7618/19090 [3:41:56<6:06:11,  1.92s/it]
 40%|███▉      | 7619/19090 [3:41:57<5:32:22,  1.74s/it]
 40%|███▉      | 7620/19090 [3:41:59<5:27:12,  1.71s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 10.290082931518555, 'learning_rate': 3.756724041879216e-05, 'recon_loss': 0.0013082850578939541, 'recon_static': 0.00035060005393461324, 'recon_motion': 0.0009576850279700011, 'recon_norm': 1.2266545295715332, 'recon_norm_std': 0.055976287182420495, 'target_norm': 1.6189794659614563, 'recon_cos': 0.6963450074195862, 'gradnorm/compressor.query': 0.02365989645388153, 'gradnorm/compressor.layer0': 0.5131787736341485, 'gradnorm/compressor.layer1': 9.567906665078844, 'gradnorm/compressor.layer2': 3.2125167490006747, 'gradnorm/compressor.layer3': 1.5864910794454257, 'gradnorm/compressor.layer4': 0.6452627250557634, 'gradnorm/compressor.layer5': 0.41499333746858896, 'gradnorm/compressor.layer6': 0.183494723783356, 'gradnorm/compressor.layer7': 0.34901631523597243, 'gradnorm/decoder.layer0': 0.5290997972314619, 'gradnorm/decoder.layer1': 0.40623978455244264, 'gradnorm/decoder.layer2': 0.22535707167809338, 'gradnorm/max_param': 6.744203567504883, 'epoch': 3.99}

 40%|███▉      | 7620/19090 [3:41:59<5:27:12,  1.71s/it]
 40%|███▉      | 7621/19090 [3:42:01<5:56:47,  1.87s/it]
 40%|███▉      | 7622/19090 [3:42:03<5:31:01,  1.73s/it]
 40%|███▉      | 7623/19090 [3:42:04<5:23:04,  1.69s/it]
 40%|███▉      | 7624/19090 [3:42:07<6:06:31,  1.92s/it]
 40%|███▉      | 7625/19090 [3:42:08<5:39:00,  1.77s/it]
 40%|███▉      | 7626/19090 [3:42:10<5:19:46,  1.67s/it]
 40%|███▉      | 7627/19090 [3:42:12<6:15:43,  1.97s/it]
 40%|███▉      | 7628/19090 [3:42:13<5:37:11,  1.77s/it]
 40%|███▉      | 7629/19090 [3:42:15<5:30:03,  1.73s/it]
 40%|███▉      | 7630/19090 [3:42:18<6:10:50,  1.94s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 10.094834327697754, 'learning_rate': 3.752770189617551e-05, 'recon_loss': 0.0013291241863043978, 'recon_static': 0.0003666278236778453, 'recon_motion': 0.0009624963713577017, 'recon_norm': 1.23518248796463, 'recon_norm_std': 0.05709848068654537, 'target_norm': 1.6220762014389039, 'recon_cos': 0.6925415441393852, 'gradnorm/compressor.query': 0.0218755117690729, 'gradnorm/compressor.layer0': 0.5053288686315202, 'gradnorm/compressor.layer1': 9.304052015715133, 'gradnorm/compressor.layer2': 3.3045196851322083, 'gradnorm/compressor.layer3': 1.6894447987963181, 'gradnorm/compressor.layer4': 0.6320716147152702, 'gradnorm/compressor.layer5': 0.4573144605802327, 'gradnorm/compressor.layer6': 0.18439266192976936, 'gradnorm/compressor.layer7': 0.35366179674138165, 'gradnorm/decoder.layer0': 0.5441472827424663, 'gradnorm/decoder.layer1': 0.446734133439177, 'gradnorm/decoder.layer2': 0.216783763729598, 'gradnorm/max_param': 6.557187557220459, 'epoch': 4.0}

 40%|███▉      | 7630/19090 [3:42:18<6:10:50,  1.94s/it]
 40%|███▉      | 7631/19090 [3:42:19<5:33:36,  1.75s/it]
 40%|███▉      | 7632/19090 [3:42:20<5:12:08,  1.63s/it]
 40%|███▉      | 7633/19090 [3:42:22<5:40:07,  1.78s/it]
 40%|███▉      | 7634/19090 [3:42:24<5:10:38,  1.63s/it]
 40%|███▉      | 7635/19090 [3:42:25<4:50:09,  1.52s/it]
 40%|████      | 7636/19090 [3:42:27<5:03:22,  1.59s/it]
 40%|████      | 7637/19090 [3:42:35<11:13:03,  3.53s/it]
 40%|████      | 7638/19090 [3:42:36<9:05:01,  2.86s/it] 
 40%|████      | 7639/19090 [3:42:40<9:46:27,  3.07s/it]
 40%|████      | 7640/19090 [3:42:41<8:04:25,  2.54s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 10.97084903717041, 'learning_rate': 3.74881214870505e-05, 'recon_loss': 0.0013158497778931633, 'recon_static': 0.0003593380592064932, 'recon_motion': 0.0009565117536112667, 'recon_norm': 1.2290847688913344, 'recon_norm_std': 0.0575870331376791, 'target_norm': 1.619644170999527, 'recon_cos': 0.6948348239064217, 'gradnorm/compressor.query': 0.026882527547823807, 'gradnorm/compressor.layer0': 0.533631987200038, 'gradnorm/compressor.layer1': 10.310722799041388, 'gradnorm/compressor.layer2': 3.0918943446417657, 'gradnorm/compressor.layer3': 1.7117240874339137, 'gradnorm/compressor.layer4': 0.6423283471237452, 'gradnorm/compressor.layer5': 0.43610455856678604, 'gradnorm/compressor.layer6': 0.19286905601487325, 'gradnorm/compressor.layer7': 0.34053933002456427, 'gradnorm/decoder.layer0': 0.5408279148392683, 'gradnorm/decoder.layer1': 0.4223207308994753, 'gradnorm/decoder.layer2': 0.21547636181869836, 'gradnorm/max_param': 7.267874240875244, 'epoch': 4.0}

 40%|████      | 7640/19090 [3:42:41<8:04:25,  2.54s/it]
 40%|████      | 7641/19090 [3:42:42<6:53:25,  2.17s/it]
 40%|████      | 7642/19090 [3:42:44<6:54:42,  2.17s/it]
 40%|████      | 7643/19090 [3:42:46<6:04:34,  1.91s/it]
 40%|████      | 7644/19090 [3:42:47<5:29:18,  1.73s/it]
 40%|████      | 7645/19090 [3:42:49<5:35:07,  1.76s/it]
 40%|████      | 7646/19090 [3:42:50<5:08:13,  1.62s/it]
 40%|████      | 7647/19090 [3:42:51<4:49:41,  1.52s/it]
 40%|████      | 7648/19090 [3:42:53<5:25:17,  1.71s/it]
 40%|████      | 7649/19090 [3:42:55<5:01:32,  1.58s/it]
 40%|████      | 7650/19090 [3:42:56<4:45:33,  1.50s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 11.070377349853516, 'learning_rate': 3.744849932375467e-05, 'recon_loss': 0.001334657080587931, 'recon_static': 0.00036396236828295514, 'recon_motion': 0.0009706947224913165, 'recon_norm': 1.233302116394043, 'recon_norm_std': 0.056672589294612406, 'target_norm': 1.622033965587616, 'recon_cos': 0.6907239764928818, 'gradnorm/compressor.query': 0.02571348524112388, 'gradnorm/compressor.layer0': 0.5345107870860107, 'gradnorm/compressor.layer1': 10.359433181162034, 'gradnorm/compressor.layer2': 3.2225460149233465, 'gradnorm/compressor.layer3': 1.7929304530223698, 'gradnorm/compressor.layer4': 0.6841022017430268, 'gradnorm/compressor.layer5': 0.45434053757327725, 'gradnorm/compressor.layer6': 0.18459441012549752, 'gradnorm/compressor.layer7': 0.3457913151554727, 'gradnorm/decoder.layer0': 0.5387003691220759, 'gradnorm/decoder.layer1': 0.4193791808248274, 'gradnorm/decoder.layer2': 0.23558757550545045, 'gradnorm/max_param': 7.302602767944336, 'epoch': 4.01}

 40%|████      | 7650/19090 [3:42:56<4:45:33,  1.50s/it]
 40%|████      | 7651/19090 [3:42:58<5:15:48,  1.66s/it]
 40%|████      | 7652/19090 [3:42:59<4:55:23,  1.55s/it]
 40%|████      | 7653/19090 [3:43:01<4:42:23,  1.48s/it]
 40%|████      | 7654/19090 [3:43:03<5:22:58,  1.69s/it]
 40%|████      | 7655/19090 [3:43:04<4:59:27,  1.57s/it]
 40%|████      | 7656/19090 [3:43:05<4:43:13,  1.49s/it]
 40%|████      | 7657/19090 [3:43:07<5:12:26,  1.64s/it]
 40%|████      | 7658/19090 [3:43:09<5:05:18,  1.60s/it]
 40%|████      | 7659/19090 [3:43:10<4:47:00,  1.51s/it]
 40%|████      | 7660/19090 [3:43:12<5:11:49,  1.64s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 10.778714179992676, 'learning_rate': 3.740883553876515e-05, 'recon_loss': 0.0013198026979807764, 'recon_static': 0.0003625256853410974, 'recon_motion': 0.0009572770155500621, 'recon_norm': 1.2314115852117538, 'recon_norm_std': 0.058639452792704105, 'target_norm': 1.6204147815704346, 'recon_cos': 0.6940012395381927, 'gradnorm/compressor.query': 0.02112012162259001, 'gradnorm/compressor.layer0': 0.5021747737997767, 'gradnorm/compressor.layer1': 10.099322124432257, 'gradnorm/compressor.layer2': 3.1356430394439014, 'gradnorm/compressor.layer3': 1.6662372830729113, 'gradnorm/compressor.layer4': 0.6290111098923058, 'gradnorm/compressor.layer5': 0.455830913266459, 'gradnorm/compressor.layer6': 0.1896893709520959, 'gradnorm/compressor.layer7': 0.3123631773875568, 'gradnorm/decoder.layer0': 0.5764810023821105, 'gradnorm/decoder.layer1': 0.4448952141896433, 'gradnorm/decoder.layer2': 0.23636990958067838, 'gradnorm/max_param': 7.11965799331665, 'epoch': 4.01}

 40%|████      | 7660/19090 [3:43:12<5:11:49,  1.64s/it]
 40%|████      | 7661/19090 [3:43:14<5:12:03,  1.64s/it]
 40%|████      | 7662/19090 [3:43:15<4:51:52,  1.53s/it]
 40%|████      | 7663/19090 [3:43:17<4:59:35,  1.57s/it]
 40%|████      | 7664/19090 [3:43:18<5:03:21,  1.59s/it]
 40%|████      | 7665/19090 [3:43:20<4:45:53,  1.50s/it]
 40%|████      | 7666/19090 [3:43:22<5:33:06,  1.75s/it]
 40%|████      | 7667/19090 [3:43:24<5:23:55,  1.70s/it]
 40%|████      | 7668/19090 [3:43:25<5:00:02,  1.58s/it]
 40%|████      | 7669/19090 [3:43:27<5:21:39,  1.69s/it]
 40%|████      | 7670/19090 [3:43:29<5:42:46,  1.80s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 11.317459106445312, 'learning_rate': 3.736913026469823e-05, 'recon_loss': 0.0013255051715532317, 'recon_static': 0.00035986885632155463, 'recon_motion': 0.0009656363050453365, 'recon_norm': 1.2307964980602264, 'recon_norm_std': 0.05722835473716259, 'target_norm': 1.6233072847127914, 'recon_cos': 0.6937580019235611, 'gradnorm/compressor.query': 0.024801787541045672, 'gradnorm/compressor.layer0': 0.5329977066015749, 'gradnorm/compressor.layer1': 10.642665762565853, 'gradnorm/compressor.layer2': 3.2021688383289595, 'gradnorm/compressor.layer3': 1.7201508102171585, 'gradnorm/compressor.layer4': 0.6645251289478629, 'gradnorm/compressor.layer5': 0.43059734770599833, 'gradnorm/compressor.layer6': 0.18919965167255476, 'gradnorm/compressor.layer7': 0.33524647323053364, 'gradnorm/decoder.layer0': 0.5530617489184042, 'gradnorm/decoder.layer1': 0.4357365592890578, 'gradnorm/decoder.layer2': 0.223830498022046, 'gradnorm/max_param': 7.50244665145874, 'epoch': 4.02}

 40%|████      | 7670/19090 [3:43:29<5:42:46,  1.80s/it]
 40%|████      | 7671/19090 [3:43:30<5:13:24,  1.65s/it]
 40%|████      | 7672/19090 [3:43:32<5:31:28,  1.74s/it]
 40%|████      | 7673/19090 [3:43:34<5:16:09,  1.66s/it]
 40%|████      | 7674/19090 [3:43:35<4:54:31,  1.55s/it]
 40%|████      | 7675/19090 [3:43:37<5:45:35,  1.82s/it]
 40%|████      | 7676/19090 [3:43:39<5:20:10,  1.68s/it]
 40%|████      | 7677/19090 [3:43:40<4:57:38,  1.56s/it]
 40%|████      | 7678/19090 [3:43:43<5:50:20,  1.84s/it]
 40%|████      | 7679/19090 [3:43:44<5:30:08,  1.74s/it]
 40%|████      | 7680/19090 [3:43:45<5:05:00,  1.60s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 11.244040489196777, 'learning_rate': 3.7329383634308946e-05, 'recon_loss': 0.0013278812170028687, 'recon_static': 0.00036263924412196504, 'recon_motion': 0.0009652419874328188, 'recon_norm': 1.2304392158985138, 'recon_norm_std': 0.05645719338208437, 'target_norm': 1.619928812980652, 'recon_cos': 0.6915351212024688, 'gradnorm/compressor.query': 0.02834190992083457, 'gradnorm/compressor.layer0': 0.5725583314122394, 'gradnorm/compressor.layer1': 10.463066539670587, 'gradnorm/compressor.layer2': 3.4510909083421146, 'gradnorm/compressor.layer3': 1.8110495057904148, 'gradnorm/compressor.layer4': 0.6950202284757572, 'gradnorm/compressor.layer5': 0.42976534745138584, 'gradnorm/compressor.layer6': 0.18526771023427682, 'gradnorm/compressor.layer7': 0.37488377279921997, 'gradnorm/decoder.layer0': 0.5648939266051899, 'gradnorm/decoder.layer1': 0.45363308932086444, 'gradnorm/decoder.layer2': 0.25811613871255884, 'gradnorm/max_param': 7.374279975891113, 'epoch': 4.02}

 40%|████      | 7680/19090 [3:43:45<5:05:00,  1.60s/it]
 40%|████      | 7681/19090 [3:43:47<5:35:40,  1.77s/it]
 40%|████      | 7682/19090 [3:43:49<5:20:23,  1.69s/it]
 40%|████      | 7683/19090 [3:43:50<4:57:48,  1.57s/it]
 40%|████      | 7684/19090 [3:43:53<5:42:14,  1.80s/it]
 40%|████      | 7685/19090 [3:43:54<5:38:15,  1.78s/it]
 40%|████      | 7686/19090 [3:43:56<5:10:01,  1.63s/it]
 40%|████      | 7687/19090 [3:43:57<5:19:01,  1.68s/it]
 40%|████      | 7688/19090 [3:43:59<5:06:54,  1.61s/it]
 40%|████      | 7689/19090 [3:44:00<4:48:08,  1.52s/it]
 40%|████      | 7690/19090 [3:44:02<5:31:50,  1.75s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 10.316421508789062, 'learning_rate': 3.728959578049057e-05, 'recon_loss': 0.0013255203346488997, 'recon_static': 0.000365323168807663, 'recon_motion': 0.0009601971731171943, 'recon_norm': 1.2342741906642913, 'recon_norm_std': 0.05711558135226369, 'target_norm': 1.6212997257709503, 'recon_cos': 0.6932116001844406, 'gradnorm/compressor.query': 0.024301073159738153, 'gradnorm/compressor.layer0': 0.4832699229173778, 'gradnorm/compressor.layer1': 9.543769393731319, 'gradnorm/compressor.layer2': 3.288309550237196, 'gradnorm/compressor.layer3': 1.7077856426448124, 'gradnorm/compressor.layer4': 0.6643776538469309, 'gradnorm/compressor.layer5': 0.4684851056899774, 'gradnorm/compressor.layer6': 0.19214876335040415, 'gradnorm/compressor.layer7': 0.3710941917015258, 'gradnorm/decoder.layer0': 0.5493378359535245, 'gradnorm/decoder.layer1': 0.4422744887824444, 'gradnorm/decoder.layer2': 0.2201895350357594, 'gradnorm/max_param': 6.727038383483887, 'epoch': 4.03}

 40%|████      | 7690/19090 [3:44:02<5:31:50,  1.75s/it]
 40%|████      | 7691/19090 [3:44:04<5:05:37,  1.61s/it]
 40%|████      | 7692/19090 [3:44:05<4:47:23,  1.51s/it]
 40%|████      | 7693/19090 [3:44:07<5:15:21,  1.66s/it]
 40%|████      | 7694/19090 [3:44:09<5:17:13,  1.67s/it]
 40%|████      | 7695/19090 [3:44:10<4:55:18,  1.55s/it]
 40%|████      | 7696/19090 [3:44:12<5:28:03,  1.73s/it]
 40%|████      | 7697/19090 [3:44:13<5:02:59,  1.60s/it]
 40%|████      | 7698/19090 [3:44:15<4:45:05,  1.50s/it]
 40%|████      | 7699/19090 [3:44:17<5:25:04,  1.71s/it]
 40%|████      | 7700/19090 [3:44:18<5:02:01,  1.59s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 11.087188720703125, 'learning_rate': 3.7249766836274244e-05, 'recon_loss': 0.0013261722109746189, 'recon_static': 0.00036654577634180895, 'recon_motion': 0.0009596264004358091, 'recon_norm': 1.2305810302495956, 'recon_norm_std': 0.057880209386348726, 'target_norm': 1.6210082948207856, 'recon_cos': 0.6923904061317444, 'gradnorm/compressor.query': 0.026325689880234496, 'gradnorm/compressor.layer0': 0.5303480682404039, 'gradnorm/compressor.layer1': 10.385521391120895, 'gradnorm/compressor.layer2': 3.2280695538407462, 'gradnorm/compressor.layer3': 1.6855619392296488, 'gradnorm/compressor.layer4': 0.7075816588573106, 'gradnorm/compressor.layer5': 0.4742063687262069, 'gradnorm/compressor.layer6': 0.20797025314578862, 'gradnorm/compressor.layer7': 0.3682521036855843, 'gradnorm/decoder.layer0': 0.5697291800798788, 'gradnorm/decoder.layer1': 0.4767453983976974, 'gradnorm/decoder.layer2': 0.25862340843481646, 'gradnorm/max_param': 7.320175647735596, 'epoch': 4.03}

 40%|████      | 7700/19090 [3:44:18<5:02:01,  1.59s/it]
 40%|████      | 7701/19090 [3:44:20<4:44:36,  1.50s/it]
 40%|████      | 7702/19090 [3:44:22<5:30:43,  1.74s/it]
 40%|████      | 7703/19090 [3:44:24<5:38:02,  1.78s/it]
 40%|████      | 7704/19090 [3:44:25<5:09:42,  1.63s/it]
 40%|████      | 7705/19090 [3:44:27<5:24:49,  1.71s/it]
 40%|████      | 7706/19090 [3:44:29<5:20:46,  1.69s/it]
 40%|████      | 7707/19090 [3:44:30<4:57:26,  1.57s/it]
 40%|████      | 7708/19090 [3:44:32<5:23:09,  1.70s/it]
 40%|████      | 7709/19090 [3:44:34<5:27:00,  1.72s/it]
 40%|████      | 7710/19090 [3:44:35<5:02:21,  1.59s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 10.652615547180176, 'learning_rate': 3.720989693482848e-05, 'recon_loss': 0.001309076667530462, 'recon_static': 0.00034660301171243193, 'recon_motion': 0.0009624736616387963, 'recon_norm': 1.2255463689565658, 'recon_norm_std': 0.05787023594602943, 'target_norm': 1.6217913508415223, 'recon_cos': 0.6977690428495407, 'gradnorm/compressor.query': 0.023314671978447003, 'gradnorm/compressor.layer0': 0.506905611574347, 'gradnorm/compressor.layer1': 9.936752579130047, 'gradnorm/compressor.layer2': 3.172339494641564, 'gradnorm/compressor.layer3': 1.7331960984457946, 'gradnorm/compressor.layer4': 0.6820016904807185, 'gradnorm/compressor.layer5': 0.47208791249381143, 'gradnorm/compressor.layer6': 0.20272945442181897, 'gradnorm/compressor.layer7': 0.3392767945350443, 'gradnorm/decoder.layer0': 0.5611618868302586, 'gradnorm/decoder.layer1': 0.4504553649903812, 'gradnorm/decoder.layer2': 0.22787158228706664, 'gradnorm/max_param': 7.0046067237854, 'epoch': 4.04}

 40%|████      | 7710/19090 [3:44:35<5:02:21,  1.59s/it]
 40%|████      | 7711/19090 [3:44:37<5:23:28,  1.71s/it]
 40%|████      | 7712/19090 [3:44:38<5:19:40,  1.69s/it]
 40%|████      | 7713/19090 [3:44:40<4:56:50,  1.57s/it]
 40%|████      | 7714/19090 [3:44:42<5:30:39,  1.74s/it]
 40%|████      | 7715/19090 [3:44:44<5:33:02,  1.76s/it]
 40%|████      | 7716/19090 [3:44:45<5:06:14,  1.62s/it]
 40%|████      | 7717/19090 [3:44:47<5:47:00,  1.83s/it]
 40%|████      | 7718/19090 [3:44:49<5:27:30,  1.73s/it]
 40%|████      | 7719/19090 [3:44:50<5:02:13,  1.59s/it]
 40%|████      | 7720/19090 [3:44:52<5:44:38,  1.82s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 12.74747371673584, 'learning_rate': 3.716998620945871e-05, 'recon_loss': 0.0013047453947365284, 'recon_static': 0.00034509405013523066, 'recon_motion': 0.0009596513802534901, 'recon_norm': 1.2219255179166795, 'recon_norm_std': 0.05751418052241206, 'target_norm': 1.6199263036251068, 'recon_cos': 0.6977968484163284, 'gradnorm/compressor.query': 0.02773561844254606, 'gradnorm/compressor.layer0': 0.6037632209009499, 'gradnorm/compressor.layer1': 11.988091599998764, 'gradnorm/compressor.layer2': 3.6306840465417793, 'gradnorm/compressor.layer3': 1.9489354480056962, 'gradnorm/compressor.layer4': 0.6567808456524927, 'gradnorm/compressor.layer5': 0.5060933871619775, 'gradnorm/compressor.layer6': 0.1961420182174277, 'gradnorm/compressor.layer7': 0.35045153663598794, 'gradnorm/decoder.layer0': 0.5799578146967019, 'gradnorm/decoder.layer1': 0.4464340475291492, 'gradnorm/decoder.layer2': 0.23197298217401094, 'gradnorm/max_param': 8.451086044311523, 'epoch': 4.04}

 40%|████      | 7720/19090 [3:44:52<5:44:38,  1.82s/it]
 40%|████      | 7721/19090 [3:44:54<5:14:27,  1.66s/it]
 40%|████      | 7722/19090 [3:44:55<4:53:21,  1.55s/it]
 40%|████      | 7723/19090 [3:44:57<5:33:10,  1.76s/it]
 40%|████      | 7724/19090 [3:44:59<5:06:17,  1.62s/it]
 40%|████      | 7725/19090 [3:45:00<4:47:28,  1.52s/it]
 40%|████      | 7726/19090 [3:45:02<5:35:19,  1.77s/it]
 40%|████      | 7727/19090 [3:45:04<5:26:32,  1.72s/it]
 40%|████      | 7728/19090 [3:45:05<5:01:24,  1.59s/it]
 40%|████      | 7729/19090 [3:45:07<5:42:44,  1.81s/it]
 40%|████      | 7730/19090 [3:45:09<5:43:08,  1.81s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 11.132966041564941, 'learning_rate': 3.7130034793606905e-05, 'recon_loss': 0.001313551946077496, 'recon_static': 0.0003515015952871181, 'recon_motion': 0.0009620503667974844, 'recon_norm': 1.2250496923923493, 'recon_norm_std': 0.05702088586986065, 'target_norm': 1.6200891196727754, 'recon_cos': 0.6954857736825943, 'gradnorm/compressor.query': 0.022874045773980462, 'gradnorm/compressor.layer0': 0.5276054580637525, 'gradnorm/compressor.layer1': 10.423642764945992, 'gradnorm/compressor.layer2': 3.2714246146594017, 'gradnorm/compressor.layer3': 1.6979631366884431, 'gradnorm/compressor.layer4': 0.6558120946785654, 'gradnorm/compressor.layer5': 0.485399393530185, 'gradnorm/compressor.layer6': 0.20556497460091028, 'gradnorm/compressor.layer7': 0.36262163144098364, 'gradnorm/decoder.layer0': 0.5863073580404277, 'gradnorm/decoder.layer1': 0.43595702698429306, 'gradnorm/decoder.layer2': 0.230571578410915, 'gradnorm/max_param': 7.347883701324463, 'epoch': 4.05}

 40%|████      | 7730/19090 [3:45:09<5:43:08,  1.81s/it]
 40%|████      | 7731/19090 [3:45:11<5:13:33,  1.66s/it]
 41%|████      | 7732/19090 [3:45:12<5:22:40,  1.70s/it]
 41%|████      | 7733/19090 [3:45:14<5:22:28,  1.70s/it]
 41%|████      | 7734/19090 [3:45:15<4:59:19,  1.58s/it]
 41%|████      | 7735/19090 [3:45:18<5:33:37,  1.76s/it]
 41%|████      | 7736/19090 [3:45:19<5:31:11,  1.75s/it]
 41%|████      | 7737/19090 [3:45:21<5:05:25,  1.61s/it]
 41%|████      | 7738/19090 [3:45:22<5:24:13,  1.71s/it]
 41%|████      | 7739/19090 [3:45:24<5:26:48,  1.73s/it]
 41%|████      | 7740/19090 [3:45:26<5:01:50,  1.60s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 12.363758087158203, 'learning_rate': 3.709004282085105e-05, 'recon_loss': 0.0013126166275469586, 'recon_static': 0.000356146703416016, 'recon_motion': 0.0009564699066686444, 'recon_norm': 1.2293505936861038, 'recon_norm_std': 0.05742627140134573, 'target_norm': 1.6193862110376358, 'recon_cos': 0.6952923506498336, 'gradnorm/compressor.query': 0.02422456633036439, 'gradnorm/compressor.layer0': 0.550253784625961, 'gradnorm/compressor.layer1': 11.671777604806163, 'gradnorm/compressor.layer2': 3.4225640147584717, 'gradnorm/compressor.layer3': 1.777475272541838, 'gradnorm/compressor.layer4': 0.6452784051270358, 'gradnorm/compressor.layer5': 0.4951013265322684, 'gradnorm/compressor.layer6': 0.20496197090448215, 'gradnorm/compressor.layer7': 0.36044395110136984, 'gradnorm/decoder.layer0': 0.6207947641164785, 'gradnorm/decoder.layer1': 0.42955290680453384, 'gradnorm/decoder.layer2': 0.22813448135059763, 'gradnorm/max_param': 8.228690147399902, 'epoch': 4.05}

 41%|████      | 7740/19090 [3:45:26<5:01:50,  1.60s/it]
 41%|████      | 7741/19090 [3:45:27<5:09:29,  1.64s/it]
 41%|████      | 7742/19090 [3:45:29<5:39:55,  1.80s/it]
 41%|████      | 7743/19090 [3:45:31<5:10:50,  1.64s/it]
 41%|████      | 7744/19090 [3:45:33<5:33:14,  1.76s/it]
 41%|████      | 7745/19090 [3:45:35<5:37:04,  1.78s/it]
 41%|████      | 7746/19090 [3:45:36<5:09:03,  1.63s/it]
 41%|████      | 7747/19090 [3:45:37<4:59:35,  1.58s/it]
 41%|████      | 7748/19090 [3:45:40<5:45:06,  1.83s/it]
 41%|████      | 7749/19090 [3:45:41<5:14:21,  1.66s/it]
 41%|████      | 7750/19090 [3:45:43<5:09:13,  1.64s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 12.401994705200195, 'learning_rate': 3.7050010424904734e-05, 'recon_loss': 0.0013073184964014217, 'recon_static': 0.0003440993546973914, 'recon_motion': 0.0009632191446144134, 'recon_norm': 1.220126861333847, 'recon_norm_std': 0.05728848222643137, 'target_norm': 1.6224010080099105, 'recon_cos': 0.6980143100023269, 'gradnorm/compressor.query': 0.025656803059149053, 'gradnorm/compressor.layer0': 0.5355598153207182, 'gradnorm/compressor.layer1': 11.644716667636498, 'gradnorm/compressor.layer2': 3.5789135726124335, 'gradnorm/compressor.layer3': 1.8976475206208108, 'gradnorm/compressor.layer4': 0.6732948072391584, 'gradnorm/compressor.layer5': 0.48722821937193067, 'gradnorm/compressor.layer6': 0.1939292082770458, 'gradnorm/compressor.layer7': 0.3611186477086595, 'gradnorm/decoder.layer0': 0.6092674698467507, 'gradnorm/decoder.layer1': 0.4800628870756508, 'gradnorm/decoder.layer2': 0.2290748372470053, 'gradnorm/max_param': 8.209513664245605, 'epoch': 4.06}

 41%|████      | 7750/19090 [3:45:43<5:09:13,  1.64s/it]
 41%|████      | 7751/19090 [3:45:45<5:32:27,  1.76s/it]
 41%|████      | 7752/19090 [3:45:46<5:05:40,  1.62s/it]
 41%|████      | 7753/19090 [3:45:48<5:24:47,  1.72s/it]
 41%|████      | 7754/19090 [3:45:50<5:37:49,  1.79s/it]
 41%|████      | 7755/19090 [3:45:51<5:09:28,  1.64s/it]
 41%|████      | 7756/19090 [3:45:53<5:21:35,  1.70s/it]
 41%|████      | 7757/19090 [3:45:55<5:51:16,  1.86s/it]
 41%|████      | 7758/19090 [3:45:56<5:18:49,  1.69s/it]
 41%|████      | 7759/19090 [3:45:58<4:56:09,  1.57s/it]
 41%|████      | 7760/19090 [3:46:00<6:00:31,  1.91s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 11.71937084197998, 'learning_rate': 3.700993773961673e-05, 'recon_loss': 0.001301251770928502, 'recon_static': 0.000336554571549641, 'recon_motion': 0.0009646972248447128, 'recon_norm': 1.2183580458164216, 'recon_norm_std': 0.05650701289996505, 'target_norm': 1.6205977648496628, 'recon_cos': 0.6992949977517128, 'gradnorm/compressor.query': 0.02303984810703101, 'gradnorm/compressor.layer0': 0.4928715357618805, 'gradnorm/compressor.layer1': 11.060769845613674, 'gradnorm/compressor.layer2': 3.2283014594536166, 'gradnorm/compressor.layer3': 1.7395213986029345, 'gradnorm/compressor.layer4': 0.6513277925534182, 'gradnorm/compressor.layer5': 0.4271974023593549, 'gradnorm/compressor.layer6': 0.2002578451605802, 'gradnorm/compressor.layer7': 0.35409927661997465, 'gradnorm/decoder.layer0': 0.5389163053995155, 'gradnorm/decoder.layer1': 0.4425360980369373, 'gradnorm/decoder.layer2': 0.230705672862668, 'gradnorm/max_param': 7.797813892364502, 'epoch': 4.06}

 41%|████      | 7760/19090 [3:46:00<6:00:31,  1.91s/it]
 41%|████      | 7761/19090 [3:46:02<5:26:27,  1.73s/it]
 41%|████      | 7762/19090 [3:46:03<5:01:38,  1.60s/it]
 41%|████      | 7763/19090 [3:46:06<5:49:34,  1.85s/it]
 41%|████      | 7764/19090 [3:46:07<5:18:16,  1.69s/it]
 41%|████      | 7765/19090 [3:46:08<4:56:21,  1.57s/it]
 41%|████      | 7766/19090 [3:46:11<5:58:41,  1.90s/it]
 41%|████      | 7767/19090 [3:46:12<5:24:13,  1.72s/it]
 41%|████      | 7768/19090 [3:46:13<5:00:29,  1.59s/it]
 41%|████      | 7769/19090 [3:46:16<5:36:38,  1.78s/it]
 41%|████      | 7770/19090 [3:46:17<5:09:24,  1.64s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 11.9207763671875, 'learning_rate': 3.6969824898970485e-05, 'recon_loss': 0.001301156633417122, 'recon_static': 0.00034049263049382714, 'recon_motion': 0.0009606639956473373, 'recon_norm': 1.2174037247896194, 'recon_norm_std': 0.0551949979737401, 'target_norm': 1.6213591337203979, 'recon_cos': 0.6993402108550072, 'gradnorm/compressor.query': 0.021693520278416367, 'gradnorm/compressor.layer0': 0.46102299947938113, 'gradnorm/compressor.layer1': 11.209889092419937, 'gradnorm/compressor.layer2': 3.4414419979875723, 'gradnorm/compressor.layer3': 1.7226500911126794, 'gradnorm/compressor.layer4': 0.636173810800153, 'gradnorm/compressor.layer5': 0.46283879372179515, 'gradnorm/compressor.layer6': 0.2065871441722249, 'gradnorm/compressor.layer7': 0.3719979964338204, 'gradnorm/decoder.layer0': 0.5711936414534345, 'gradnorm/decoder.layer1': 0.4798292540804923, 'gradnorm/decoder.layer2': 0.2504988104854826, 'gradnorm/max_param': 7.9034342765808105, 'epoch': 4.07}

 41%|████      | 7770/19090 [3:46:17<5:09:24,  1.64s/it]
 41%|████      | 7771/19090 [3:46:18<4:49:47,  1.54s/it]
 41%|████      | 7772/19090 [3:46:21<6:01:34,  1.92s/it]
 41%|████      | 7773/19090 [3:46:22<5:26:03,  1.73s/it]
 41%|████      | 7774/19090 [3:46:24<5:00:43,  1.59s/it]
 41%|████      | 7775/19090 [3:46:26<5:41:03,  1.81s/it]
 41%|████      | 7776/19090 [3:46:27<5:11:45,  1.65s/it]
 41%|████      | 7777/19090 [3:46:28<4:51:27,  1.55s/it]
 41%|████      | 7778/19090 [3:46:31<6:05:10,  1.94s/it]
 41%|████      | 7779/19090 [3:46:33<5:28:37,  1.74s/it]
 41%|████      | 7780/19090 [3:46:34<5:02:57,  1.61s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 12.173102378845215, 'learning_rate': 3.692967203708372e-05, 'recon_loss': 0.001313037020736374, 'recon_static': 0.0003458362312812824, 'recon_motion': 0.0009672008163761348, 'recon_norm': 1.2217264115810393, 'recon_norm_std': 0.05550564434379339, 'target_norm': 1.6211741358041762, 'recon_cos': 0.6963298469781876, 'gradnorm/compressor.query': 0.025499777512348182, 'gradnorm/compressor.layer0': 0.5406812605497286, 'gradnorm/compressor.layer1': 11.470161538221275, 'gradnorm/compressor.layer2': 3.399196355873501, 'gradnorm/compressor.layer3': 1.8416631494318774, 'gradnorm/compressor.layer4': 0.6571005123612469, 'gradnorm/compressor.layer5': 0.4568884047591307, 'gradnorm/compressor.layer6': 0.21341475440958171, 'gradnorm/compressor.layer7': 0.3529537224409689, 'gradnorm/decoder.layer0': 0.5402348251506395, 'gradnorm/decoder.layer1': 0.4687013600862993, 'gradnorm/decoder.layer2': 0.24161922017164242, 'gradnorm/max_param': 8.086915969848633, 'epoch': 4.08}

 41%|████      | 7780/19090 [3:46:34<5:02:57,  1.61s/it]
 41%|████      | 7781/19090 [3:46:36<5:51:30,  1.86s/it]
 41%|████      | 7782/19090 [3:46:38<5:18:29,  1.69s/it]
 41%|████      | 7783/19090 [3:46:39<4:55:58,  1.57s/it]
 41%|████      | 7784/19090 [3:46:42<6:06:56,  1.95s/it]
 41%|████      | 7785/19090 [3:46:43<5:29:29,  1.75s/it]
 41%|████      | 7786/19090 [3:46:44<5:03:21,  1.61s/it]
 41%|████      | 7787/19090 [3:46:47<5:52:18,  1.87s/it]
 41%|████      | 7788/19090 [3:46:48<5:20:13,  1.70s/it]
 41%|████      | 7789/19090 [3:46:49<4:57:49,  1.58s/it]
 41%|████      | 7790/19090 [3:46:52<6:13:13,  1.98s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 13.31105899810791, 'learning_rate': 3.688947928820798e-05, 'recon_loss': 0.0013188606506446376, 'recon_static': 0.0003561073157470673, 'recon_motion': 0.0009627533523598686, 'recon_norm': 1.2277407497167587, 'recon_norm_std': 0.0563053141348064, 'target_norm': 1.621738103032112, 'recon_cos': 0.6948506906628609, 'gradnorm/compressor.query': 0.026331303161302128, 'gradnorm/compressor.layer0': 0.5742320623606343, 'gradnorm/compressor.layer1': 12.487957455921222, 'gradnorm/compressor.layer2': 3.9196200632362967, 'gradnorm/compressor.layer3': 1.99004293462549, 'gradnorm/compressor.layer4': 0.7399770646793898, 'gradnorm/compressor.layer5': 0.4708956407853961, 'gradnorm/compressor.layer6': 0.22050971443430503, 'gradnorm/compressor.layer7': 0.3475731203795428, 'gradnorm/decoder.layer0': 0.5798057918721824, 'gradnorm/decoder.layer1': 0.5008410294629605, 'gradnorm/decoder.layer2': 0.23516574821426867, 'gradnorm/max_param': 8.804484367370605, 'epoch': 4.08}

 41%|████      | 7790/19090 [3:46:52<6:13:13,  1.98s/it]
 41%|████      | 7791/19090 [3:46:54<5:34:28,  1.78s/it]
 41%|████      | 7792/19090 [3:46:55<5:06:44,  1.63s/it]
 41%|████      | 7793/19090 [3:46:57<5:53:30,  1.88s/it]
 41%|████      | 7794/19090 [3:46:59<5:20:27,  1.70s/it]
 41%|████      | 7795/19090 [3:47:00<4:57:03,  1.58s/it]
 41%|████      | 7796/19090 [3:47:03<5:55:30,  1.89s/it]
 41%|████      | 7797/19090 [3:47:04<5:21:29,  1.71s/it]
 41%|████      | 7798/19090 [3:47:05<4:57:54,  1.58s/it]
 41%|████      | 7799/19090 [3:47:08<5:56:19,  1.89s/it]
 41%|████      | 7800/19090 [3:47:09<5:22:21,  1.71s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 12.402430534362793, 'learning_rate': 3.684924678672815e-05, 'recon_loss': 0.001315353118116036, 'recon_static': 0.00035464163884171287, 'recon_motion': 0.0009607114669051952, 'recon_norm': 1.2252253472805024, 'recon_norm_std': 0.05593450060114265, 'target_norm': 1.620965486764908, 'recon_cos': 0.6953889399766922, 'gradnorm/compressor.query': 0.026482410908411275, 'gradnorm/compressor.layer0': 0.5954602356069009, 'gradnorm/compressor.layer1': 11.637601195980196, 'gradnorm/compressor.layer2': 3.5331190446135046, 'gradnorm/compressor.layer3': 1.9758814667505218, 'gradnorm/compressor.layer4': 0.7261386168088538, 'gradnorm/compressor.layer5': 0.4944667719890823, 'gradnorm/compressor.layer6': 0.22610098595893507, 'gradnorm/compressor.layer7': 0.3883878350520422, 'gradnorm/decoder.layer0': 0.5990750793346181, 'gradnorm/decoder.layer1': 0.5044270271402334, 'gradnorm/decoder.layer2': 0.2390586880772677, 'gradnorm/max_param': 8.203726768493652, 'epoch': 4.09}

 41%|████      | 7800/19090 [3:47:09<5:22:21,  1.71s/it]
 41%|████      | 7801/19090 [3:47:10<4:58:24,  1.59s/it]
 41%|████      | 7802/19090 [3:47:13<6:10:24,  1.97s/it]
 41%|████      | 7803/19090 [3:47:15<5:32:09,  1.77s/it]
 41%|████      | 7804/19090 [3:47:16<5:11:12,  1.65s/it]
 41%|████      | 7805/19090 [3:47:19<6:25:39,  2.05s/it]
 41%|████      | 7806/19090 [3:47:20<5:43:58,  1.83s/it]
 41%|████      | 7807/19090 [3:47:22<5:15:34,  1.68s/it]
 41%|████      | 7808/19090 [3:47:24<6:10:30,  1.97s/it]
 41%|████      | 7809/19090 [3:47:25<5:32:34,  1.77s/it]
 41%|████      | 7810/19090 [3:47:27<5:06:27,  1.63s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 12.35370922088623, 'learning_rate': 3.680897466716205e-05, 'recon_loss': 0.0013202035770518705, 'recon_static': 0.0003682006208691746, 'recon_motion': 0.0009520029838313348, 'recon_norm': 1.2327137231826781, 'recon_norm_std': 0.05672499649226666, 'target_norm': 1.619607502222061, 'recon_cos': 0.6935052245855331, 'gradnorm/compressor.query': 0.02606430869957539, 'gradnorm/compressor.layer0': 0.5653367303787065, 'gradnorm/compressor.layer1': 11.650649783632645, 'gradnorm/compressor.layer2': 3.3829574025331026, 'gradnorm/compressor.layer3': 1.8671810517139897, 'gradnorm/compressor.layer4': 0.718237071907812, 'gradnorm/compressor.layer5': 0.5066882673582179, 'gradnorm/compressor.layer6': 0.21997980571038273, 'gradnorm/compressor.layer7': 0.3998988224754826, 'gradnorm/decoder.layer0': 0.6126618385595141, 'gradnorm/decoder.layer1': 0.46894560547374264, 'gradnorm/decoder.layer2': 0.22194830126893259, 'gradnorm/max_param': 8.214599609375, 'epoch': 4.09}

 41%|████      | 7810/19090 [3:47:27<5:06:27,  1.63s/it]
 41%|████      | 7811/19090 [3:47:29<5:43:26,  1.83s/it]
 41%|████      | 7812/19090 [3:47:30<5:13:15,  1.67s/it]
 41%|████      | 7813/19090 [3:47:32<4:52:18,  1.56s/it]
 41%|████      | 7814/19090 [3:47:34<5:43:07,  1.83s/it]
 41%|████      | 7815/19090 [3:47:35<5:13:49,  1.67s/it]
 41%|████      | 7816/19090 [3:47:37<5:02:15,  1.61s/it]
 41%|████      | 7817/19090 [3:47:39<5:35:33,  1.79s/it]
 41%|████      | 7818/19090 [3:47:40<5:07:47,  1.64s/it]
 41%|████      | 7819/19090 [3:47:42<5:12:31,  1.66s/it]
 41%|████      | 7820/19090 [3:47:45<6:05:37,  1.95s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 13.383015632629395, 'learning_rate': 3.6768663064159945e-05, 'recon_loss': 0.0013105210149660707, 'recon_static': 0.00035536826399038544, 'recon_motion': 0.0009551527793519199, 'recon_norm': 1.223773154616356, 'recon_norm_std': 0.05774492900818586, 'target_norm': 1.6202903985977173, 'recon_cos': 0.6962846145033836, 'gradnorm/compressor.query': 0.026692956442571504, 'gradnorm/compressor.layer0': 0.5749494613330447, 'gradnorm/compressor.layer1': 12.64869807483289, 'gradnorm/compressor.layer2': 3.6179562239985064, 'gradnorm/compressor.layer3': 2.0139796912524575, 'gradnorm/compressor.layer4': 0.7067062727007832, 'gradnorm/compressor.layer5': 0.5082290115038252, 'gradnorm/compressor.layer6': 0.22305357767646494, 'gradnorm/compressor.layer7': 0.35937835857646383, 'gradnorm/decoder.layer0': 0.6265819080419608, 'gradnorm/decoder.layer1': 0.49745789651648925, 'gradnorm/decoder.layer2': 0.24858507205962704, 'gradnorm/max_param': 8.918002128601074, 'epoch': 4.1}

 41%|████      | 7820/19090 [3:47:45<6:05:37,  1.95s/it]
 41%|████      | 7821/19090 [3:47:46<5:28:33,  1.75s/it]
 41%|████      | 7822/19090 [3:47:48<5:19:49,  1.70s/it]
 41%|████      | 7823/19090 [3:47:50<6:25:23,  2.05s/it]
 41%|████      | 7824/19090 [3:47:52<5:42:46,  1.83s/it]
 41%|████      | 7825/19090 [3:47:53<5:12:25,  1.66s/it]
 41%|████      | 7826/19090 [3:47:56<6:27:32,  2.06s/it]
 41%|████      | 7827/19090 [3:47:57<5:43:37,  1.83s/it]
 41%|████      | 7828/19090 [3:47:59<5:12:51,  1.67s/it]
 41%|████      | 7829/19090 [3:48:01<6:20:22,  2.03s/it]
 41%|████      | 7830/19090 [3:48:03<5:38:59,  1.81s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 13.29182243347168, 'learning_rate': 3.6728312112504114e-05, 'recon_loss': 0.0013124892982887104, 'recon_static': 0.0003544871062331367, 'recon_motion': 0.0009580021622241474, 'recon_norm': 1.2234253406524658, 'recon_norm_std': 0.0574964283965528, 'target_norm': 1.6208178132772446, 'recon_cos': 0.6959340140223503, 'gradnorm/compressor.query': 0.027364211218113468, 'gradnorm/compressor.layer0': 0.6133689331740398, 'gradnorm/compressor.layer1': 12.52007299432408, 'gradnorm/compressor.layer2': 3.7579006890558126, 'gradnorm/compressor.layer3': 1.9424060112443822, 'gradnorm/compressor.layer4': 0.6707263042730315, 'gradnorm/compressor.layer5': 0.5253732387686557, 'gradnorm/compressor.layer6': 0.2339747667330005, 'gradnorm/compressor.layer7': 0.41947155248897544, 'gradnorm/decoder.layer0': 0.6296174665935298, 'gradnorm/decoder.layer1': 0.48452261244459843, 'gradnorm/decoder.layer2': 0.24648520202014337, 'gradnorm/max_param': 8.826706886291504, 'epoch': 4.1}

 41%|████      | 7830/19090 [3:48:03<5:38:59,  1.81s/it]
 41%|████      | 7831/19090 [3:48:04<5:09:38,  1.65s/it]
 41%|████      | 7832/19090 [3:48:07<6:01:28,  1.93s/it]
 41%|████      | 7833/19090 [3:48:08<5:25:07,  1.73s/it]
 41%|████      | 7834/19090 [3:48:09<5:00:05,  1.60s/it]
 41%|████      | 7835/19090 [3:48:12<6:15:38,  2.00s/it]
 41%|████      | 7836/19090 [3:48:13<5:35:38,  1.79s/it]
 41%|████      | 7837/19090 [3:48:15<5:08:08,  1.64s/it]
 41%|████      | 7838/19090 [3:48:17<5:57:45,  1.91s/it]
 41%|████      | 7839/19090 [3:48:19<5:23:37,  1.73s/it]
 41%|████      | 7840/19090 [3:48:20<4:59:47,  1.60s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 12.273550987243652, 'learning_rate': 3.6687921947108415e-05, 'recon_loss': 0.0013298573030624539, 'recon_static': 0.00036825383649556896, 'recon_motion': 0.0009616034178179689, 'recon_norm': 1.2302688747644424, 'recon_norm_std': 0.05703814271837473, 'target_norm': 1.6216286540031433, 'recon_cos': 0.6918411955237389, 'gradnorm/compressor.query': 0.030187721960834284, 'gradnorm/compressor.layer0': 0.5802955862034211, 'gradnorm/compressor.layer1': 11.586503141746906, 'gradnorm/compressor.layer2': 3.390250892919779, 'gradnorm/compressor.layer3': 1.7772065832845094, 'gradnorm/compressor.layer4': 0.6290529921031549, 'gradnorm/compressor.layer5': 0.4685633923706659, 'gradnorm/compressor.layer6': 0.21209195137469708, 'gradnorm/compressor.layer7': 0.3737814297107564, 'gradnorm/decoder.layer0': 0.5764121637534764, 'gradnorm/decoder.layer1': 0.467100866444102, 'gradnorm/decoder.layer2': 0.22953946864278918, 'gradnorm/max_param': 8.168399810791016, 'epoch': 4.11}

 41%|████      | 7840/19090 [3:48:20<4:59:47,  1.60s/it]
 41%|████      | 7841/19090 [3:48:23<6:09:09,  1.97s/it]
 41%|████      | 7842/19090 [3:48:24<5:30:55,  1.77s/it]
 41%|████      | 7843/19090 [3:48:25<5:04:22,  1.62s/it]
 41%|████      | 7844/19090 [3:48:28<6:20:08,  2.03s/it]
 41%|████      | 7845/19090 [3:48:30<5:38:26,  1.81s/it]
 41%|████      | 7846/19090 [3:48:31<5:09:25,  1.65s/it]
 41%|████      | 7847/19090 [3:48:34<6:12:41,  1.99s/it]
 41%|████      | 7848/19090 [3:48:35<5:33:09,  1.78s/it]
 41%|████      | 7849/19090 [3:48:36<5:05:31,  1.63s/it]
 41%|████      | 7850/19090 [3:48:39<5:45:48,  1.85s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 12.092144966125488, 'learning_rate': 3.664749270301781e-05, 'recon_loss': 0.0013222965033492073, 'recon_static': 0.0003618665330577642, 'recon_motion': 0.0009604300168575719, 'recon_norm': 1.2309685975313187, 'recon_norm_std': 0.05629551243036986, 'target_norm': 1.6200469195842744, 'recon_cos': 0.6934582144021988, 'gradnorm/compressor.query': 0.024423957250101998, 'gradnorm/compressor.layer0': 0.555737240871799, 'gradnorm/compressor.layer1': 11.362098839521794, 'gradnorm/compressor.layer2': 3.346576529906349, 'gradnorm/compressor.layer3': 2.01445328122727, 'gradnorm/compressor.layer4': 0.6968702803652934, 'gradnorm/compressor.layer5': 0.47447659554257704, 'gradnorm/compressor.layer6': 0.2531513288721478, 'gradnorm/compressor.layer7': 0.4057712117724609, 'gradnorm/decoder.layer0': 0.5687405071933531, 'gradnorm/decoder.layer1': 0.47618386595401163, 'gradnorm/decoder.layer2': 0.2567703685537056, 'gradnorm/max_param': 8.010488510131836, 'epoch': 4.11}

 41%|████      | 7850/19090 [3:48:39<5:45:48,  1.85s/it]
 41%|████      | 7851/19090 [3:48:40<5:14:21,  1.68s/it]
 41%|████      | 7852/19090 [3:48:41<4:52:23,  1.56s/it]
 41%|████      | 7853/19090 [3:48:44<6:06:54,  1.96s/it]
 41%|████      | 7854/19090 [3:48:45<5:29:00,  1.76s/it]
 41%|████      | 7855/19090 [3:48:47<5:02:38,  1.62s/it]
 41%|████      | 7856/19090 [3:48:49<5:42:25,  1.83s/it]
 41%|████      | 7857/19090 [3:48:50<5:11:59,  1.67s/it]
 41%|████      | 7858/19090 [3:48:51<4:50:41,  1.55s/it]
 41%|████      | 7859/19090 [3:48:55<6:14:49,  2.00s/it]
 41%|████      | 7860/19090 [3:48:56<5:34:53,  1.79s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 12.535566329956055, 'learning_rate': 3.660702451540791e-05, 'recon_loss': 0.0013270065566757694, 'recon_static': 0.00035767073059105313, 'recon_motion': 0.0009693358180811629, 'recon_norm': 1.2236664086580276, 'recon_norm_std': 0.054835092183202504, 'target_norm': 1.620698508620262, 'recon_cos': 0.6921248868107795, 'gradnorm/compressor.query': 0.02409226411869682, 'gradnorm/compressor.layer0': 0.572706970216922, 'gradnorm/compressor.layer1': 11.746350959521761, 'gradnorm/compressor.layer2': 3.6475693121031547, 'gradnorm/compressor.layer3': 1.9651275490479236, 'gradnorm/compressor.layer4': 0.7244639700624246, 'gradnorm/compressor.layer5': 0.5115496232189573, 'gradnorm/compressor.layer6': 0.24237646913796354, 'gradnorm/compressor.layer7': 0.37009156070452404, 'gradnorm/decoder.layer0': 0.6232194332701887, 'gradnorm/decoder.layer1': 0.4861898876876449, 'gradnorm/decoder.layer2': 0.24817310335397744, 'gradnorm/max_param': 8.280817985534668, 'epoch': 4.12}

 41%|████      | 7860/19090 [3:48:56<5:34:53,  1.79s/it]
 41%|████      | 7861/19090 [3:48:57<5:07:22,  1.64s/it]
 41%|████      | 7862/19090 [3:49:00<6:05:59,  1.96s/it]
 41%|████      | 7863/19090 [3:49:01<5:28:48,  1.76s/it]
 41%|████      | 7864/19090 [3:49:02<5:03:18,  1.62s/it]
 41%|████      | 7865/19090 [3:49:05<6:24:00,  2.05s/it]
 41%|████      | 7866/19090 [3:49:07<5:41:20,  1.82s/it]
 41%|████      | 7867/19090 [3:49:08<5:11:47,  1.67s/it]
 41%|████      | 7868/19090 [3:49:11<6:07:25,  1.96s/it]
 41%|████      | 7869/19090 [3:49:12<5:29:33,  1.76s/it]
 41%|████      | 7870/19090 [3:49:13<5:03:07,  1.62s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 12.629449844360352, 'learning_rate': 3.656651751958454e-05, 'recon_loss': 0.001325349576654844, 'recon_static': 0.00036149351217318326, 'recon_motion': 0.0009638560732128099, 'recon_norm': 1.227872058749199, 'recon_norm_std': 0.05425059227272868, 'target_norm': 1.619804212450981, 'recon_cos': 0.6924793273210526, 'gradnorm/compressor.query': 0.02655294777762326, 'gradnorm/compressor.layer0': 0.5741849357354225, 'gradnorm/compressor.layer1': 11.859234111687833, 'gradnorm/compressor.layer2': 3.565154308173612, 'gradnorm/compressor.layer3': 2.0098559004339176, 'gradnorm/compressor.layer4': 0.7424621475409763, 'gradnorm/compressor.layer5': 0.509383573635473, 'gradnorm/compressor.layer6': 0.24810264741353044, 'gradnorm/compressor.layer7': 0.4158725044604437, 'gradnorm/decoder.layer0': 0.6337325389796621, 'gradnorm/decoder.layer1': 0.5197453452081212, 'gradnorm/decoder.layer2': 0.25363774407482054, 'gradnorm/max_param': 8.360822677612305, 'epoch': 4.12}

 41%|████      | 7870/19090 [3:49:13<5:03:07,  1.62s/it]
 41%|████      | 7871/19090 [3:49:16<6:07:04,  1.96s/it]
 41%|████      | 7872/19090 [3:49:17<5:29:13,  1.76s/it]
 41%|████      | 7873/19090 [3:49:19<5:02:59,  1.62s/it]
 41%|████      | 7874/19090 [3:49:22<6:14:36,  2.00s/it]
 41%|████▏     | 7875/19090 [3:49:23<5:34:38,  1.79s/it]
 41%|████▏     | 7876/19090 [3:49:24<5:06:29,  1.64s/it]
 41%|████▏     | 7877/19090 [3:49:27<5:58:37,  1.92s/it]
 41%|████▏     | 7878/19090 [3:49:28<5:23:39,  1.73s/it]
 41%|████▏     | 7879/19090 [3:49:29<4:59:02,  1.60s/it]
 41%|████▏     | 7880/19090 [3:49:32<6:13:59,  2.00s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 13.028328895568848, 'learning_rate': 3.652597185098331e-05, 'recon_loss': 0.0013323554652743042, 'recon_static': 0.00037903834527242, 'recon_motion': 0.0009533171163639053, 'recon_norm': 1.234052023291588, 'recon_norm_std': 0.055883658770471814, 'target_norm': 1.6177719235420227, 'recon_cos': 0.689592395722866, 'gradnorm/compressor.query': 0.029617517894108516, 'gradnorm/compressor.layer0': 0.6326448194847165, 'gradnorm/compressor.layer1': 12.26001003545005, 'gradnorm/compressor.layer2': 3.692199557200346, 'gradnorm/compressor.layer3': 1.9196699565102593, 'gradnorm/compressor.layer4': 0.7739566255313588, 'gradnorm/compressor.layer5': 0.48819577804167436, 'gradnorm/compressor.layer6': 0.21671841894343144, 'gradnorm/compressor.layer7': 0.4185799381473577, 'gradnorm/decoder.layer0': 0.5862730210019905, 'gradnorm/decoder.layer1': 0.4945055251338388, 'gradnorm/decoder.layer2': 0.2513005432845974, 'gradnorm/max_param': 8.642291069030762, 'epoch': 4.13}

 41%|████▏     | 7880/19090 [3:49:32<6:13:59,  2.00s/it]
 41%|████▏     | 7881/19090 [3:49:34<5:34:02,  1.79s/it]
 41%|████▏     | 7882/19090 [3:49:35<5:05:47,  1.64s/it]
 41%|████▏     | 7883/19090 [3:49:37<5:55:34,  1.90s/it]
 41%|████▏     | 7884/19090 [3:49:39<5:20:54,  1.72s/it]
 41%|████▏     | 7885/19090 [3:49:40<4:56:28,  1.59s/it]
 41%|████▏     | 7886/19090 [3:49:43<6:10:19,  1.98s/it]
 41%|████▏     | 7887/19090 [3:49:44<5:32:17,  1.78s/it]
 41%|████▏     | 7888/19090 [3:49:45<5:04:40,  1.63s/it]
 41%|████▏     | 7889/19090 [3:49:48<6:26:31,  2.07s/it]
 41%|████▏     | 7890/19090 [3:49:50<5:42:57,  1.84s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 13.808439254760742, 'learning_rate': 3.6485387645169064e-05, 'recon_loss': 0.0013425009703496471, 'recon_static': 0.00038347938025253827, 'recon_motion': 0.0009590216344804503, 'recon_norm': 1.2377974569797516, 'recon_norm_std': 0.05656704623252153, 'target_norm': 1.622717598080635, 'recon_cos': 0.6892513707280159, 'gradnorm/compressor.query': 0.02999776642526277, 'gradnorm/compressor.layer0': 0.6350021625842418, 'gradnorm/compressor.layer1': 13.02498983137499, 'gradnorm/compressor.layer2': 3.8567738280798776, 'gradnorm/compressor.layer3': 2.0101542194473323, 'gradnorm/compressor.layer4': 0.748486501726181, 'gradnorm/compressor.layer5': 0.530546788808394, 'gradnorm/compressor.layer6': 0.22843737770526082, 'gradnorm/compressor.layer7': 0.3650100753647738, 'gradnorm/decoder.layer0': 0.62540777732111, 'gradnorm/decoder.layer1': 0.48211222915904894, 'gradnorm/decoder.layer2': 0.22880359990999236, 'gradnorm/max_param': 9.183481216430664, 'epoch': 4.13}

 41%|████▏     | 7890/19090 [3:49:50<5:42:57,  1.84s/it]
 41%|████▏     | 7891/19090 [3:49:51<5:12:06,  1.67s/it]
 41%|████▏     | 7892/19090 [3:49:54<6:17:54,  2.02s/it]
 41%|████▏     | 7893/19090 [3:49:55<5:37:09,  1.81s/it]
 41%|████▏     | 7894/19090 [3:49:56<5:07:50,  1.65s/it]
 41%|████▏     | 7895/19090 [3:50:00<6:25:55,  2.07s/it]
 41%|████▏     | 7896/19090 [3:50:01<5:42:27,  1.84s/it]
 41%|████▏     | 7897/19090 [3:50:02<5:12:28,  1.67s/it]
 41%|████▏     | 7898/19090 [3:50:05<5:59:35,  1.93s/it]
 41%|████▏     | 7899/19090 [3:50:06<5:23:46,  1.74s/it]
 41%|████▏     | 7900/19090 [3:50:07<4:58:37,  1.60s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 14.647189140319824, 'learning_rate': 3.6444765037835584e-05, 'recon_loss': 0.001339036546414718, 'recon_static': 0.00037041921241325325, 'recon_motion': 0.0009686173114459961, 'recon_norm': 1.2286532133817674, 'recon_norm_std': 0.0556508082896471, 'target_norm': 1.6228149622678756, 'recon_cos': 0.6897376388311386, 'gradnorm/compressor.query': 0.0289496236958001, 'gradnorm/compressor.layer0': 0.6313608025098946, 'gradnorm/compressor.layer1': 13.786595698259438, 'gradnorm/compressor.layer2': 4.156623307292301, 'gradnorm/compressor.layer3': 2.205623211318102, 'gradnorm/compressor.layer4': 0.789985494601197, 'gradnorm/compressor.layer5': 0.5432509642556029, 'gradnorm/compressor.layer6': 0.2582949113088707, 'gradnorm/compressor.layer7': 0.43605256766726924, 'gradnorm/decoder.layer0': 0.6465413278807168, 'gradnorm/decoder.layer1': 0.5159124382397028, 'gradnorm/decoder.layer2': 0.26070193719407536, 'gradnorm/max_param': 9.720369338989258, 'epoch': 4.14}

 41%|████▏     | 7900/19090 [3:50:07<4:58:37,  1.60s/it]
 41%|████▏     | 7901/19090 [3:50:10<5:56:33,  1.91s/it]
 41%|████▏     | 7902/19090 [3:50:11<5:21:37,  1.72s/it]
 41%|████▏     | 7903/19090 [3:50:12<4:57:06,  1.59s/it]
 41%|████▏     | 7904/19090 [3:50:15<6:15:10,  2.01s/it]
 41%|████▏     | 7905/19090 [3:50:17<5:34:56,  1.80s/it]
 41%|████▏     | 7906/19090 [3:50:18<5:06:16,  1.64s/it]
 41%|████▏     | 7907/19090 [3:50:21<6:05:00,  1.96s/it]
 41%|████▏     | 7908/19090 [3:50:22<5:27:46,  1.76s/it]
 41%|████▏     | 7909/19090 [3:50:23<5:01:32,  1.62s/it]
 41%|████▏     | 7910/19090 [3:50:26<6:18:29,  2.03s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 13.534956932067871, 'learning_rate': 3.6404104164804975e-05, 'recon_loss': 0.0013313260511495172, 'recon_static': 0.0003564611768524628, 'recon_motion': 0.0009748648619279266, 'recon_norm': 1.2256074845790863, 'recon_norm_std': 0.053953283187001946, 'target_norm': 1.621104297041893, 'recon_cos': 0.6913798406720162, 'gradnorm/compressor.query': 0.0268977417722932, 'gradnorm/compressor.layer0': 0.6293188838318603, 'gradnorm/compressor.layer1': 12.735661840655602, 'gradnorm/compressor.layer2': 3.7983581409143596, 'gradnorm/compressor.layer3': 2.1118014472916724, 'gradnorm/compressor.layer4': 0.724929862082719, 'gradnorm/compressor.layer5': 0.5161414594365502, 'gradnorm/compressor.layer6': 0.24437681598671526, 'gradnorm/compressor.layer7': 0.3720148400633242, 'gradnorm/decoder.layer0': 0.6177409141486783, 'gradnorm/decoder.layer1': 0.5243679999610612, 'gradnorm/decoder.layer2': 0.2596032488430452, 'gradnorm/max_param': 8.977843284606934, 'epoch': 4.14}

 41%|████▏     | 7910/19090 [3:50:26<6:18:29,  2.03s/it]
 41%|████▏     | 7911/19090 [3:50:28<5:37:01,  1.81s/it]
 41%|████▏     | 7912/19090 [3:50:29<5:07:53,  1.65s/it]
 41%|████▏     | 7913/19090 [3:50:31<5:54:44,  1.90s/it]
 41%|████▏     | 7914/19090 [3:50:33<5:20:02,  1.72s/it]
 41%|████▏     | 7915/19090 [3:50:34<4:56:02,  1.59s/it]
 41%|████▏     | 7916/19090 [3:50:37<6:00:33,  1.94s/it]
 41%|████▏     | 7917/19090 [3:50:38<5:24:30,  1.74s/it]
 41%|████▏     | 7918/19090 [3:50:39<4:58:55,  1.61s/it]
 41%|████▏     | 7919/19090 [3:50:42<6:01:25,  1.94s/it]
 41%|████▏     | 7920/19090 [3:50:43<5:24:51,  1.75s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 12.584371566772461, 'learning_rate': 3.6363405162027305e-05, 'recon_loss': 0.0013304788706591352, 'recon_static': 0.000373912895884132, 'recon_motion': 0.0009565659347572364, 'recon_norm': 1.2302209496498109, 'recon_norm_std': 0.054560516774654386, 'target_norm': 1.618701633810997, 'recon_cos': 0.6903153136372566, 'gradnorm/compressor.query': 0.03147689953626179, 'gradnorm/compressor.layer0': 0.6016420831923924, 'gradnorm/compressor.layer1': 11.78641467634885, 'gradnorm/compressor.layer2': 3.623566047918195, 'gradnorm/compressor.layer3': 2.074647675902098, 'gradnorm/compressor.layer4': 0.7141400401546303, 'gradnorm/compressor.layer5': 0.5693161766326346, 'gradnorm/compressor.layer6': 0.2242399827028556, 'gradnorm/compressor.layer7': 0.36719750329859074, 'gradnorm/decoder.layer0': 0.6036247969465295, 'gradnorm/decoder.layer1': 0.4562180612437253, 'gradnorm/decoder.layer2': 0.2402902398632357, 'gradnorm/max_param': 8.309615135192871, 'epoch': 4.15}

 41%|████▏     | 7920/19090 [3:50:43<5:24:51,  1.75s/it]
 41%|████▏     | 7921/19090 [3:50:45<4:59:27,  1.61s/it]
 41%|████▏     | 7922/19090 [3:50:47<6:02:00,  1.94s/it]
 42%|████▏     | 7923/19090 [3:50:49<5:25:57,  1.75s/it]
 42%|████▏     | 7924/19090 [3:50:50<5:00:02,  1.61s/it]
 42%|████▏     | 7925/19090 [3:50:53<6:07:31,  1.98s/it]
 42%|████▏     | 7926/19090 [3:50:54<5:29:21,  1.77s/it]
 42%|████▏     | 7927/19090 [3:50:55<5:02:16,  1.62s/it]
 42%|████▏     | 7928/19090 [3:50:58<5:42:02,  1.84s/it]
 42%|████▏     | 7929/19090 [3:50:59<5:11:04,  1.67s/it]
 42%|████▏     | 7930/19090 [3:51:00<4:49:37,  1.56s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 13.55837345123291, 'learning_rate': 3.632266816558015e-05, 'recon_loss': 0.001353953787474893, 'recon_static': 0.0003862105688313022, 'recon_motion': 0.0009677432550233789, 'recon_norm': 1.2396239906549453, 'recon_norm_std': 0.054320373199880125, 'target_norm': 1.6212983071804046, 'recon_cos': 0.68548404276371, 'gradnorm/compressor.query': 0.028289208514114067, 'gradnorm/compressor.layer0': 0.6213875801287887, 'gradnorm/compressor.layer1': 12.823060981853589, 'gradnorm/compressor.layer2': 3.672107575538288, 'gradnorm/compressor.layer3': 1.9719985923181735, 'gradnorm/compressor.layer4': 0.7474371115597933, 'gradnorm/compressor.layer5': 0.5535510873573174, 'gradnorm/compressor.layer6': 0.24075682360841547, 'gradnorm/compressor.layer7': 0.38968348524752716, 'gradnorm/decoder.layer0': 0.5243305729447586, 'gradnorm/decoder.layer1': 0.4790056016680065, 'gradnorm/decoder.layer2': 0.24342768523247543, 'gradnorm/max_param': 9.041421890258789, 'epoch': 4.15}

 42%|████▏     | 7930/19090 [3:51:00<4:49:37,  1.56s/it]
 42%|████▏     | 7931/19090 [3:51:03<5:40:10,  1.83s/it]
 42%|████▏     | 7932/19090 [3:51:04<5:10:21,  1.67s/it]
 42%|████▏     | 7933/19090 [3:51:05<4:49:27,  1.56s/it]
 42%|████▏     | 7934/19090 [3:51:08<6:03:52,  1.96s/it]
 42%|████▏     | 7935/19090 [3:51:09<5:26:31,  1.76s/it]
 42%|████▏     | 7936/19090 [3:51:11<5:00:32,  1.62s/it]
 42%|████▏     | 7937/19090 [3:51:14<6:23:05,  2.06s/it]
 42%|████▏     | 7938/19090 [3:51:15<5:39:57,  1.83s/it]
 42%|████▏     | 7939/19090 [3:51:16<5:09:36,  1.67s/it]
 42%|████▏     | 7940/19090 [3:51:19<6:14:54,  2.02s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 14.554825782775879, 'learning_rate': 3.628189331166811e-05, 'recon_loss': 0.0013382399745751172, 'recon_static': 0.00038552816404262557, 'recon_motion': 0.0009527118250844069, 'recon_norm': 1.237735304236412, 'recon_norm_std': 0.05615172814577818, 'target_norm': 1.6210034400224687, 'recon_cos': 0.6894164472818375, 'gradnorm/compressor.query': 0.02832218792888048, 'gradnorm/compressor.layer0': 0.6661820189843466, 'gradnorm/compressor.layer1': 13.784800377065563, 'gradnorm/compressor.layer2': 3.822475203439285, 'gradnorm/compressor.layer3': 2.1734928170685057, 'gradnorm/compressor.layer4': 0.889858417875725, 'gradnorm/compressor.layer5': 0.5423137913471255, 'gradnorm/compressor.layer6': 0.26830539576937346, 'gradnorm/compressor.layer7': 0.3929688326644052, 'gradnorm/decoder.layer0': 0.6261929332978319, 'gradnorm/decoder.layer1': 0.5123788783533187, 'gradnorm/decoder.layer2': 0.2742388019675006, 'gradnorm/max_param': 9.718454360961914, 'epoch': 4.16}

 42%|████▏     | 7940/19090 [3:51:19<6:14:54,  2.02s/it]
 42%|████▏     | 7941/19090 [3:51:20<5:34:14,  1.80s/it]
 42%|████▏     | 7942/19090 [3:51:22<5:05:38,  1.64s/it]
 42%|████▏     | 7943/19090 [3:51:25<6:12:08,  2.00s/it]
 42%|████▏     | 7944/19090 [3:51:26<5:32:40,  1.79s/it]
 42%|████▏     | 7945/19090 [3:51:27<5:04:52,  1.64s/it]
 42%|████▏     | 7946/19090 [3:51:30<6:03:31,  1.96s/it]
 42%|████▏     | 7947/19090 [3:51:31<5:26:09,  1.76s/it]
 42%|████▏     | 7948/19090 [3:51:32<5:00:20,  1.62s/it]
 42%|████▏     | 7949/19090 [3:51:35<6:00:01,  1.94s/it]
 42%|████▏     | 7950/19090 [3:51:36<5:24:11,  1.75s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 14.615724563598633, 'learning_rate': 3.624108073662233e-05, 'recon_loss': 0.001352714459062554, 'recon_static': 0.00038724980840925126, 'recon_motion': 0.0009654645851696842, 'recon_norm': 1.2376282960176468, 'recon_norm_std': 0.05356547199189663, 'target_norm': 1.6205266773700715, 'recon_cos': 0.6855632424354553, 'gradnorm/compressor.query': 0.031172867668929013, 'gradnorm/compressor.layer0': 0.6998543008672032, 'gradnorm/compressor.layer1': 13.837622532909375, 'gradnorm/compressor.layer2': 3.9242141860306554, 'gradnorm/compressor.layer3': 2.107706724189319, 'gradnorm/compressor.layer4': 0.7906744704016875, 'gradnorm/compressor.layer5': 0.5416260453422526, 'gradnorm/compressor.layer6': 0.23707487807479405, 'gradnorm/compressor.layer7': 0.40492744843822054, 'gradnorm/decoder.layer0': 0.6153900918235469, 'gradnorm/decoder.layer1': 0.4773712643499412, 'gradnorm/decoder.layer2': 0.2482934312130913, 'gradnorm/max_param': 9.755824089050293, 'epoch': 4.16}

 42%|████▏     | 7950/19090 [3:51:36<5:24:11,  1.75s/it]
 42%|████▏     | 7951/19090 [3:51:38<4:59:15,  1.61s/it]
 42%|████▏     | 7952/19090 [3:51:40<6:00:36,  1.94s/it]
 42%|████▏     | 7953/19090 [3:51:42<5:24:04,  1.75s/it]
 42%|████▏     | 7954/19090 [3:51:43<4:58:49,  1.61s/it]
 42%|████▏     | 7955/19090 [3:51:46<5:58:06,  1.93s/it]
 42%|████▏     | 7956/19090 [3:51:47<5:22:26,  1.74s/it]
 42%|████▏     | 7957/19090 [3:51:48<4:57:42,  1.60s/it]
 42%|████▏     | 7958/19090 [3:51:51<5:38:33,  1.82s/it]
 42%|████▏     | 7959/19090 [3:51:52<5:08:41,  1.66s/it]
 42%|████▏     | 7960/19090 [3:51:53<4:48:07,  1.55s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 13.676859855651855, 'learning_rate': 3.6200230576900115e-05, 'recon_loss': 0.0013630915404064582, 'recon_static': 0.00039080506830941887, 'recon_motion': 0.0009722864997456782, 'recon_norm': 1.2368225127458572, 'recon_norm_std': 0.053526630625128746, 'target_norm': 1.6215525925159455, 'recon_cos': 0.6829073131084442, 'gradnorm/compressor.query': 0.02493622872667891, 'gradnorm/compressor.layer0': 0.6095749453656101, 'gradnorm/compressor.layer1': 12.920754915745961, 'gradnorm/compressor.layer2': 3.6805716908099217, 'gradnorm/compressor.layer3': 2.0883954029444007, 'gradnorm/compressor.layer4': 0.8015969564605657, 'gradnorm/compressor.layer5': 0.5127170812435851, 'gradnorm/compressor.layer6': 0.24206134371403776, 'gradnorm/compressor.layer7': 0.4220694341125245, 'gradnorm/decoder.layer0': 0.593103885455695, 'gradnorm/decoder.layer1': 0.5165507935237725, 'gradnorm/decoder.layer2': 0.26391958680678906, 'gradnorm/max_param': 9.110045433044434, 'epoch': 4.17}

 42%|████▏     | 7960/19090 [3:51:53<4:48:07,  1.55s/it]
 42%|████▏     | 7961/19090 [3:51:56<6:05:49,  1.97s/it]
 42%|████▏     | 7962/19090 [3:51:57<5:27:40,  1.77s/it]
 42%|████▏     | 7963/19090 [3:51:59<5:01:17,  1.62s/it]
 42%|████▏     | 7964/19090 [3:52:02<6:13:00,  2.01s/it]
 42%|████▏     | 7965/19090 [3:52:03<5:34:36,  1.80s/it]
 42%|████▏     | 7966/19090 [3:52:04<5:06:01,  1.65s/it]
 42%|████▏     | 7967/19090 [3:52:07<5:57:20,  1.93s/it]
 42%|████▏     | 7968/19090 [3:52:08<5:21:38,  1.74s/it]
 42%|████▏     | 7969/19090 [3:52:09<4:57:25,  1.60s/it]
 42%|████▏     | 7970/19090 [3:52:12<6:05:06,  1.97s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 14.670944213867188, 'learning_rate': 3.615934296908443e-05, 'recon_loss': 0.001352797297295183, 'recon_static': 0.0003907497433829121, 'recon_motion': 0.0009620475510018877, 'recon_norm': 1.2393167108297347, 'recon_norm_std': 0.0551379781216383, 'target_norm': 1.6222625225782394, 'recon_cos': 0.6860460579395294, 'gradnorm/compressor.query': 0.027388158780014105, 'gradnorm/compressor.layer0': 0.6442829925532787, 'gradnorm/compressor.layer1': 13.866275278094353, 'gradnorm/compressor.layer2': 3.872212053088264, 'gradnorm/compressor.layer3': 2.3555079214909864, 'gradnorm/compressor.layer4': 0.8346547061557825, 'gradnorm/compressor.layer5': 0.5716907100051469, 'gradnorm/compressor.layer6': 0.24633336890071547, 'gradnorm/compressor.layer7': 0.40430039482611646, 'gradnorm/decoder.layer0': 0.6569683820513605, 'gradnorm/decoder.layer1': 0.4999489311125574, 'gradnorm/decoder.layer2': 0.2747799480610857, 'gradnorm/max_param': 9.776557922363281, 'epoch': 4.17}

 42%|████▏     | 7970/19090 [3:52:12<6:05:06,  1.97s/it]
 42%|████▏     | 7971/19090 [3:52:14<5:27:03,  1.76s/it]
 42%|████▏     | 7972/19090 [3:52:15<5:00:48,  1.62s/it]
 42%|████▏     | 7973/19090 [3:52:18<6:23:27,  2.07s/it]
 42%|████▏     | 7974/19090 [3:52:19<5:40:15,  1.84s/it]
 42%|████▏     | 7975/19090 [3:52:21<5:10:52,  1.68s/it]
 42%|████▏     | 7976/19090 [3:52:23<6:01:06,  1.95s/it]
 42%|████▏     | 7977/19090 [3:52:24<5:24:34,  1.75s/it]
 42%|████▏     | 7978/19090 [3:52:26<4:59:00,  1.61s/it]
 42%|████▏     | 7979/19090 [3:52:28<6:00:14,  1.95s/it]
 42%|████▏     | 7980/19090 [3:52:30<5:23:46,  1.75s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 13.570192337036133, 'learning_rate': 3.611841804988343e-05, 'recon_loss': 0.0013317034143256023, 'recon_static': 0.000381501760421088, 'recon_motion': 0.0009502016589976847, 'recon_norm': 1.23557206094265, 'recon_norm_std': 0.055257436633110044, 'target_norm': 1.6179743379354476, 'recon_cos': 0.6897924348711968, 'gradnorm/compressor.query': 0.02894585548598153, 'gradnorm/compressor.layer0': 0.667492345273135, 'gradnorm/compressor.layer1': 12.81121730431419, 'gradnorm/compressor.layer2': 3.6665254623504873, 'gradnorm/compressor.layer3': 2.0527253155277014, 'gradnorm/compressor.layer4': 0.813710484753854, 'gradnorm/compressor.layer5': 0.5433256499453196, 'gradnorm/compressor.layer6': 0.2558498235186396, 'gradnorm/compressor.layer7': 0.3995244589457075, 'gradnorm/decoder.layer0': 0.6188010117666805, 'gradnorm/decoder.layer1': 0.5269665534681695, 'gradnorm/decoder.layer2': 0.27634870649355636, 'gradnorm/max_param': 9.030854225158691, 'epoch': 4.18}

 42%|████▏     | 7980/19090 [3:52:30<5:23:46,  1.75s/it]
 42%|████▏     | 7981/19090 [3:52:31<4:58:02,  1.61s/it]
 42%|████▏     | 7982/19090 [3:52:34<6:13:49,  2.02s/it]
 42%|████▏     | 7983/19090 [3:52:35<5:33:29,  1.80s/it]
 42%|████▏     | 7984/19090 [3:52:37<5:05:19,  1.65s/it]
 42%|████▏     | 7985/19090 [3:52:39<5:59:29,  1.94s/it]
 42%|████▏     | 7986/19090 [3:52:41<5:23:20,  1.75s/it]
 42%|████▏     | 7987/19090 [3:52:42<4:57:59,  1.61s/it]
 42%|████▏     | 7988/19090 [3:52:45<6:00:15,  1.95s/it]
 42%|████▏     | 7989/19090 [3:52:46<5:23:36,  1.75s/it]
 42%|████▏     | 7990/19090 [3:52:47<4:58:14,  1.61s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 13.429587364196777, 'learning_rate': 3.607745595613003e-05, 'recon_loss': 0.0013560749706812204, 'recon_static': 0.0003978017659392208, 'recon_motion': 0.0009582731858245097, 'recon_norm': 1.2414267927408218, 'recon_norm_std': 0.05507359160110355, 'target_norm': 1.6207603216171265, 'recon_cos': 0.6843972414731979, 'gradnorm/compressor.query': 0.033846221313943235, 'gradnorm/compressor.layer0': 0.7057318575949705, 'gradnorm/compressor.layer1': 12.628641641985082, 'gradnorm/compressor.layer2': 3.7526650177863963, 'gradnorm/compressor.layer3': 2.068066339384266, 'gradnorm/compressor.layer4': 0.8392851635313691, 'gradnorm/compressor.layer5': 0.5563580247375883, 'gradnorm/compressor.layer6': 0.25790662203371, 'gradnorm/compressor.layer7': 0.4327032722608164, 'gradnorm/decoder.layer0': 0.6495357083995101, 'gradnorm/decoder.layer1': 0.498982496877719, 'gradnorm/decoder.layer2': 0.27210455663217076, 'gradnorm/max_param': 8.902629852294922, 'epoch': 4.19}

 42%|████▏     | 7990/19090 [3:52:47<4:58:14,  1.61s/it]
 42%|████▏     | 7991/19090 [3:52:50<5:53:35,  1.91s/it]
 42%|████▏     | 7992/19090 [3:52:51<5:19:18,  1.73s/it]
 42%|████▏     | 7993/19090 [3:52:52<4:55:10,  1.60s/it]
 42%|████▏     | 7994/19090 [3:52:55<5:55:06,  1.92s/it]
 42%|████▏     | 7995/19090 [3:52:56<5:19:59,  1.73s/it]
 42%|████▏     | 7996/19090 [3:52:58<4:55:17,  1.60s/it]
 42%|████▏     | 7997/19090 [3:53:00<5:59:18,  1.94s/it]
 42%|████▏     | 7998/19090 [3:53:02<5:22:53,  1.75s/it]
 42%|████▏     | 7999/19090 [3:53:03<4:57:54,  1.61s/it]
 42%|████▏     | 8000/19090 [3:53:06<6:09:57,  2.00s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 13.742136001586914, 'learning_rate': 3.603645682478144e-05, 'recon_loss': 0.0013601907790871336, 'recon_static': 0.0003986022049502935, 'recon_motion': 0.0009615885574021377, 'recon_norm': 1.2410184741020203, 'recon_norm_std': 0.055290553439408544, 'target_norm': 1.6220736473798751, 'recon_cos': 0.6839975982904434, 'gradnorm/compressor.query': 0.03882054256801713, 'gradnorm/compressor.layer0': 0.7243758491070094, 'gradnorm/compressor.layer1': 12.907774458809396, 'gradnorm/compressor.layer2': 3.8016178351285754, 'gradnorm/compressor.layer3': 2.2457593586002758, 'gradnorm/compressor.layer4': 0.8167496830402806, 'gradnorm/compressor.layer5': 0.6050214284065486, 'gradnorm/compressor.layer6': 0.28518928702248286, 'gradnorm/compressor.layer7': 0.4426048008151175, 'gradnorm/decoder.layer0': 0.7157088917696767, 'gradnorm/decoder.layer1': 0.5573832861435616, 'gradnorm/decoder.layer2': 0.2835270311884083, 'gradnorm/max_param': 9.09823226928711, 'epoch': 4.19}

 42%|████▏     | 8000/19090 [3:53:06<6:09:57,  2.00s/it]
 42%|████▏     | 8001/19090 [3:53:10<8:19:16,  2.70s/it]
 42%|████▏     | 8002/19090 [3:53:11<7:00:30,  2.28s/it]
 42%|████▏     | 8003/19090 [3:53:13<6:05:28,  1.98s/it]
 42%|████▏     | 8004/19090 [3:53:14<5:27:14,  1.77s/it]
 42%|████▏     | 8005/19090 [3:53:15<5:00:13,  1.63s/it]
 42%|████▏     | 8006/19090 [3:53:17<4:41:28,  1.52s/it]
 42%|████▏     | 8007/19090 [3:53:18<4:28:12,  1.45s/it]
 42%|████▏     | 8008/19090 [3:53:19<4:19:04,  1.40s/it]
 42%|████▏     | 8009/19090 [3:53:20<4:12:37,  1.37s/it]
 42%|████▏     | 8010/19090 [3:53:22<4:08:05,  1.34s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 13.374052047729492, 'learning_rate': 3.59954207929187e-05, 'recon_loss': 0.0013446495431708172, 'recon_static': 0.0003958854933443945, 'recon_motion': 0.0009487640883889981, 'recon_norm': 1.2438917487859726, 'recon_norm_std': 0.056194229330867526, 'target_norm': 1.6188818484544754, 'recon_cos': 0.6871236458420753, 'gradnorm/compressor.query': 0.028965078479615707, 'gradnorm/compressor.layer0': 0.6654744116709569, 'gradnorm/compressor.layer1': 12.66403538996944, 'gradnorm/compressor.layer2': 3.425231402299961, 'gradnorm/compressor.layer3': 2.093516322518496, 'gradnorm/compressor.layer4': 0.8054867451060852, 'gradnorm/compressor.layer5': 0.5833120909864872, 'gradnorm/compressor.layer6': 0.276292109991067, 'gradnorm/compressor.layer7': 0.3838281587289725, 'gradnorm/decoder.layer0': 0.626107736254708, 'gradnorm/decoder.layer1': 0.5126342446424508, 'gradnorm/decoder.layer2': 0.2474608068307774, 'gradnorm/max_param': 8.92748737335205, 'epoch': 4.2}

 42%|████▏     | 8010/19090 [3:53:22<4:08:05,  1.34s/it]
 42%|████▏     | 8011/19090 [3:53:23<4:04:57,  1.33s/it]
 42%|████▏     | 8012/19090 [3:53:24<4:02:37,  1.31s/it]
 42%|████▏     | 8013/19090 [3:53:26<4:11:26,  1.36s/it]
 42%|████▏     | 8014/19090 [3:53:27<4:16:11,  1.39s/it]
 42%|████▏     | 8015/19090 [3:53:29<4:45:06,  1.54s/it]
 42%|████▏     | 8016/19090 [3:53:31<5:05:04,  1.65s/it]
 42%|████▏     | 8017/19090 [3:53:33<4:56:31,  1.61s/it]
 42%|████▏     | 8018/19090 [3:53:34<5:15:53,  1.71s/it]
 42%|████▏     | 8019/19090 [3:53:36<5:29:47,  1.79s/it]
 42%|████▏     | 8020/19090 [3:53:38<5:18:02,  1.72s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 15.002412796020508, 'learning_rate': 3.595434799774624e-05, 'recon_loss': 0.0013445305201457813, 'recon_static': 0.00038911123483558184, 'recon_motion': 0.0009554192612995394, 'recon_norm': 1.2367014229297637, 'recon_norm_std': 0.05388206029310823, 'target_norm': 1.6183218270540238, 'recon_cos': 0.6863909080624581, 'gradnorm/compressor.query': 0.029249517590095975, 'gradnorm/compressor.layer0': 0.7030819985697795, 'gradnorm/compressor.layer1': 14.253591519887902, 'gradnorm/compressor.layer2': 3.8242774779629407, 'gradnorm/compressor.layer3': 2.1582098637126133, 'gradnorm/compressor.layer4': 0.8622875449527635, 'gradnorm/compressor.layer5': 0.6103557861181989, 'gradnorm/compressor.layer6': 0.25479734018070116, 'gradnorm/compressor.layer7': 0.44827183976506374, 'gradnorm/decoder.layer0': 0.6299712599526408, 'gradnorm/decoder.layer1': 0.5355525811557077, 'gradnorm/decoder.layer2': 0.25253201302987954, 'gradnorm/max_param': 10.049920082092285, 'epoch': 4.2}

 42%|████▏     | 8020/19090 [3:53:38<5:18:02,  1.72s/it]
 42%|████▏     | 8021/19090 [3:53:40<5:22:34,  1.75s/it]
 42%|████▏     | 8022/19090 [3:53:41<5:14:57,  1.71s/it]
 42%|████▏     | 8023/19090 [3:53:43<5:20:27,  1.74s/it]
 42%|████▏     | 8024/19090 [3:53:45<5:19:29,  1.73s/it]
 42%|████▏     | 8025/19090 [3:53:47<5:25:44,  1.77s/it]
 42%|████▏     | 8026/19090 [3:53:49<5:29:54,  1.79s/it]
 42%|████▏     | 8027/19090 [3:53:50<5:27:49,  1.78s/it]
 42%|████▏     | 8028/19090 [3:53:52<5:31:09,  1.80s/it]
 42%|████▏     | 8029/19090 [3:53:54<5:25:49,  1.77s/it]
 42%|████▏     | 8030/19090 [3:53:56<5:16:43,  1.72s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 14.25008487701416, 'learning_rate': 3.591323857659139e-05, 'recon_loss': 0.0013552793709095568, 'recon_static': 0.0003923608259356115, 'recon_motion': 0.0009629185617086478, 'recon_norm': 1.239290463924408, 'recon_norm_std': 0.05428346609696746, 'target_norm': 1.6202757984399796, 'recon_cos': 0.6844204723834991, 'gradnorm/compressor.query': 0.032534774509759355, 'gradnorm/compressor.layer0': 0.739767002527838, 'gradnorm/compressor.layer1': 13.306440953523575, 'gradnorm/compressor.layer2': 4.195304998465709, 'gradnorm/compressor.layer3': 2.3891502269458242, 'gradnorm/compressor.layer4': 0.8548729271738799, 'gradnorm/compressor.layer5': 0.5785236788233519, 'gradnorm/compressor.layer6': 0.28196016033862775, 'gradnorm/compressor.layer7': 0.42473452635404774, 'gradnorm/decoder.layer0': 0.6833804424453639, 'gradnorm/decoder.layer1': 0.532397013676434, 'gradnorm/decoder.layer2': 0.26600762465849964, 'gradnorm/max_param': 9.378877639770508, 'epoch': 4.21}

 42%|████▏     | 8030/19090 [3:53:56<5:16:43,  1.72s/it]
 42%|████▏     | 8031/19090 [3:53:57<5:25:17,  1.76s/it]
 42%|████▏     | 8032/19090 [3:53:59<5:14:01,  1.70s/it]
 42%|████▏     | 8033/19090 [3:54:00<5:00:51,  1.63s/it]
 42%|████▏     | 8034/19090 [3:54:02<5:12:18,  1.69s/it]
 42%|████▏     | 8035/19090 [3:54:04<5:20:04,  1.74s/it]
 42%|████▏     | 8036/19090 [3:54:05<4:55:39,  1.60s/it]
 42%|████▏     | 8037/19090 [3:54:08<5:22:51,  1.75s/it]
 42%|████▏     | 8038/19090 [3:54:09<5:15:12,  1.71s/it]
 42%|████▏     | 8039/19090 [3:54:10<4:52:18,  1.59s/it]
 42%|████▏     | 8040/19090 [3:54:12<5:06:50,  1.67s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 12.766249656677246, 'learning_rate': 3.587209266690396e-05, 'recon_loss': 0.0013651123736053705, 'recon_static': 0.00040081954721244984, 'recon_motion': 0.0009642928183893673, 'recon_norm': 1.243647065758705, 'recon_norm_std': 0.05305632762610912, 'target_norm': 1.619212144613266, 'recon_cos': 0.6816561818122864, 'gradnorm/compressor.query': 0.03161748125035508, 'gradnorm/compressor.layer0': 0.7028322458101247, 'gradnorm/compressor.layer1': 12.00712882325017, 'gradnorm/compressor.layer2': 3.614394755579102, 'gradnorm/compressor.layer3': 1.8577609710348875, 'gradnorm/compressor.layer4': 0.7897121285816118, 'gradnorm/compressor.layer5': 0.520637109668299, 'gradnorm/compressor.layer6': 0.27231388840968207, 'gradnorm/compressor.layer7': 0.41800354429503694, 'gradnorm/decoder.layer0': 0.5879252984350282, 'gradnorm/decoder.layer1': 0.48817618193306284, 'gradnorm/decoder.layer2': 0.2613635855936672, 'gradnorm/max_param': 8.463709831237793, 'epoch': 4.21}

 42%|████▏     | 8040/19090 [3:54:12<5:06:50,  1.67s/it]
 42%|████▏     | 8041/19090 [3:54:14<5:07:39,  1.67s/it]
 42%|████▏     | 8042/19090 [3:54:15<4:46:54,  1.56s/it]
 42%|████▏     | 8043/19090 [3:54:17<4:59:54,  1.63s/it]
 42%|████▏     | 8044/19090 [3:54:19<5:00:12,  1.63s/it]
 42%|████▏     | 8045/19090 [3:54:20<4:41:35,  1.53s/it]
 42%|████▏     | 8046/19090 [3:54:22<4:56:34,  1.61s/it]
 42%|████▏     | 8047/19090 [3:54:23<4:53:01,  1.59s/it]
 42%|████▏     | 8048/19090 [3:54:25<4:36:36,  1.50s/it]
 42%|████▏     | 8049/19090 [3:54:26<4:49:17,  1.57s/it]
 42%|████▏     | 8050/19090 [3:54:28<4:55:25,  1.61s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 15.073751449584961, 'learning_rate': 3.583091040625575e-05, 'recon_loss': 0.0013791684148600325, 'recon_static': 0.00041283689279225655, 'recon_motion': 0.0009663315315265209, 'recon_norm': 1.2458188027143478, 'recon_norm_std': 0.05446740314364433, 'target_norm': 1.6224127292633057, 'recon_cos': 0.6792425096035004, 'gradnorm/compressor.query': 0.034256343372361615, 'gradnorm/compressor.layer0': 0.7784133147192801, 'gradnorm/compressor.layer1': 14.140839194682995, 'gradnorm/compressor.layer2': 4.305850001832536, 'gradnorm/compressor.layer3': 2.4275814631236, 'gradnorm/compressor.layer4': 0.8981837370373409, 'gradnorm/compressor.layer5': 0.6096697119107883, 'gradnorm/compressor.layer6': 0.2765438337546242, 'gradnorm/compressor.layer7': 0.42828600076784257, 'gradnorm/decoder.layer0': 0.6670780377969066, 'gradnorm/decoder.layer1': 0.4992166432558598, 'gradnorm/decoder.layer2': 0.28546361165884887, 'gradnorm/max_param': 9.96875286102295, 'epoch': 4.22}

 42%|████▏     | 8050/19090 [3:54:28<4:55:25,  1.61s/it]
 42%|████▏     | 8051/19090 [3:54:29<4:37:55,  1.51s/it]
 42%|████▏     | 8052/19090 [3:54:31<4:33:50,  1.49s/it]
 42%|████▏     | 8053/19090 [3:54:33<4:48:50,  1.57s/it]
 42%|████▏     | 8054/19090 [3:54:34<4:33:21,  1.49s/it]
 42%|████▏     | 8055/19090 [3:54:35<4:31:51,  1.48s/it]
 42%|████▏     | 8056/19090 [3:54:37<4:55:56,  1.61s/it]
 42%|████▏     | 8057/19090 [3:54:38<4:38:26,  1.51s/it]
 42%|████▏     | 8058/19090 [3:54:40<4:26:05,  1.45s/it]
 42%|████▏     | 8059/19090 [3:54:42<5:07:09,  1.67s/it]
 42%|████▏     | 8060/19090 [3:54:43<4:46:26,  1.56s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 14.492603302001953, 'learning_rate': 3.578969193234008e-05, 'recon_loss': 0.0013612721755634992, 'recon_static': 0.00040241131791844963, 'recon_motion': 0.0009588608809281141, 'recon_norm': 1.2459380626678467, 'recon_norm_std': 0.053032868821173904, 'target_norm': 1.6198009997606277, 'recon_cos': 0.6832253351807595, 'gradnorm/compressor.query': 0.031799424403733914, 'gradnorm/compressor.layer0': 0.7200480225892905, 'gradnorm/compressor.layer1': 13.68343317240922, 'gradnorm/compressor.layer2': 3.9093380128130764, 'gradnorm/compressor.layer3': 2.209979061873315, 'gradnorm/compressor.layer4': 0.8407448126626926, 'gradnorm/compressor.layer5': 0.5658453169786407, 'gradnorm/compressor.layer6': 0.28072805451625193, 'gradnorm/compressor.layer7': 0.4182481455821661, 'gradnorm/decoder.layer0': 0.7183328745000426, 'gradnorm/decoder.layer1': 0.49360981049572333, 'gradnorm/decoder.layer2': 0.26895554189959553, 'gradnorm/max_param': 9.645941734313965, 'epoch': 4.22}

 42%|████▏     | 8060/19090 [3:54:43<4:46:26,  1.56s/it]
 42%|████▏     | 8061/19090 [3:54:45<4:31:31,  1.48s/it]
 42%|████▏     | 8062/19090 [3:54:47<5:06:52,  1.67s/it]
 42%|████▏     | 8063/19090 [3:54:48<4:46:07,  1.56s/it]
 42%|████▏     | 8064/19090 [3:54:49<4:32:02,  1.48s/it]
 42%|████▏     | 8065/19090 [3:54:52<5:20:06,  1.74s/it]
 42%|████▏     | 8066/19090 [3:54:53<4:55:39,  1.61s/it]
 42%|████▏     | 8067/19090 [3:54:54<4:38:34,  1.52s/it]
 42%|████▏     | 8068/19090 [3:54:56<4:59:44,  1.63s/it]
 42%|████▏     | 8069/19090 [3:54:57<4:41:16,  1.53s/it]
 42%|████▏     | 8070/19090 [3:54:59<4:28:44,  1.46s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 15.121838569641113, 'learning_rate': 3.5748437382971384e-05, 'recon_loss': 0.0013694669672986493, 'recon_static': 0.00041408646502532064, 'recon_motion': 0.0009553805415634997, 'recon_norm': 1.2458838641643524, 'recon_norm_std': 0.05417433315888047, 'target_norm': 1.6204632639884948, 'recon_cos': 0.6808864116668701, 'gradnorm/compressor.query': 0.03358670738885782, 'gradnorm/compressor.layer0': 0.7535080327451922, 'gradnorm/compressor.layer1': 14.296081454240847, 'gradnorm/compressor.layer2': 4.129845056837054, 'gradnorm/compressor.layer3': 2.142209100370903, 'gradnorm/compressor.layer4': 0.8024996671943331, 'gradnorm/compressor.layer5': 0.6148663101032018, 'gradnorm/compressor.layer6': 0.2828108076181936, 'gradnorm/compressor.layer7': 0.4149671784026524, 'gradnorm/decoder.layer0': 0.7015962087098712, 'gradnorm/decoder.layer1': 0.48571780411614, 'gradnorm/decoder.layer2': 0.275788878208381, 'gradnorm/max_param': 10.078486442565918, 'epoch': 4.23}

 42%|████▏     | 8070/19090 [3:54:59<4:28:44,  1.46s/it]
 42%|████▏     | 8071/19090 [3:55:01<4:49:18,  1.58s/it]
 42%|████▏     | 8072/19090 [3:55:02<4:33:43,  1.49s/it]
 42%|████▏     | 8073/19090 [3:55:03<4:22:56,  1.43s/it]
 42%|████▏     | 8074/19090 [3:55:05<5:09:36,  1.69s/it]
 42%|████▏     | 8075/19090 [3:55:07<4:48:08,  1.57s/it]
 42%|████▏     | 8076/19090 [3:55:08<4:32:25,  1.48s/it]
 42%|████▏     | 8077/19090 [3:55:10<5:14:39,  1.71s/it]
 42%|████▏     | 8078/19090 [3:55:12<4:51:45,  1.59s/it]
 42%|████▏     | 8079/19090 [3:55:13<4:36:08,  1.50s/it]
 42%|████▏     | 8080/19090 [3:55:16<5:42:07,  1.86s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 16.324920654296875, 'learning_rate': 3.57071468960847e-05, 'recon_loss': 0.0013600988808320836, 'recon_static': 0.00039444109061150814, 'recon_motion': 0.0009656577865825966, 'recon_norm': 1.240938600897789, 'recon_norm_std': 0.05305030355229974, 'target_norm': 1.6207997441291808, 'recon_cos': 0.6834884837269783, 'gradnorm/compressor.query': 0.03531298864393661, 'gradnorm/compressor.layer0': 0.7960142835313173, 'gradnorm/compressor.layer1': 15.494692478236448, 'gradnorm/compressor.layer2': 4.1396653700777195, 'gradnorm/compressor.layer3': 2.5119419977574435, 'gradnorm/compressor.layer4': 0.8794010380277044, 'gradnorm/compressor.layer5': 0.6532241454002685, 'gradnorm/compressor.layer6': 0.28813809714680544, 'gradnorm/compressor.layer7': 0.4447956264447639, 'gradnorm/decoder.layer0': 0.7128586017261717, 'gradnorm/decoder.layer1': 0.5269965547687901, 'gradnorm/decoder.layer2': 0.2628976120948445, 'gradnorm/max_param': 10.92469310760498, 'epoch': 4.23}

 42%|████▏     | 8080/19090 [3:55:16<5:42:07,  1.86s/it]
 42%|████▏     | 8081/19090 [3:55:17<5:10:21,  1.69s/it]
 42%|████▏     | 8082/19090 [3:55:18<4:48:26,  1.57s/it]
 42%|████▏     | 8083/19090 [3:55:21<5:38:38,  1.85s/it]
 42%|████▏     | 8084/19090 [3:55:22<5:08:36,  1.68s/it]
 42%|████▏     | 8085/19090 [3:55:23<4:46:56,  1.56s/it]
 42%|████▏     | 8086/19090 [3:55:26<5:52:57,  1.92s/it]
 42%|████▏     | 8087/19090 [3:55:27<5:18:00,  1.73s/it]
 42%|████▏     | 8088/19090 [3:55:29<4:53:43,  1.60s/it]
 42%|████▏     | 8089/19090 [3:55:31<5:56:34,  1.94s/it]
 42%|████▏     | 8090/19090 [3:55:33<5:20:53,  1.75s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 15.313017845153809, 'learning_rate': 3.566582060973521e-05, 'recon_loss': 0.0013608128007035702, 'recon_static': 0.00038973045739112424, 'recon_motion': 0.0009710823389468715, 'recon_norm': 1.2336132735013963, 'recon_norm_std': 0.05468734847381711, 'target_norm': 1.6231409072875977, 'recon_cos': 0.6841722756624222, 'gradnorm/compressor.query': 0.03309009576257274, 'gradnorm/compressor.layer0': 0.787714269037759, 'gradnorm/compressor.layer1': 14.437798038725685, 'gradnorm/compressor.layer2': 4.1673010279635445, 'gradnorm/compressor.layer3': 2.38128342892869, 'gradnorm/compressor.layer4': 0.9165441293888124, 'gradnorm/compressor.layer5': 0.6147674241732275, 'gradnorm/compressor.layer6': 0.27850231443809764, 'gradnorm/compressor.layer7': 0.46797164826936494, 'gradnorm/decoder.layer0': 0.6748802485027868, 'gradnorm/decoder.layer1': 0.5780773658717555, 'gradnorm/decoder.layer2': 0.2753492013064553, 'gradnorm/max_param': 10.178297996520996, 'epoch': 4.24}

 42%|████▏     | 8090/19090 [3:55:33<5:20:53,  1.75s/it]
 42%|████▏     | 8091/19090 [3:55:34<4:55:42,  1.61s/it]
 42%|████▏     | 8092/19090 [3:55:36<5:20:43,  1.75s/it]
 42%|████▏     | 8093/19090 [3:55:37<4:55:32,  1.61s/it]
 42%|████▏     | 8094/19090 [3:55:39<4:38:00,  1.52s/it]
 42%|████▏     | 8095/19090 [3:55:41<5:38:32,  1.85s/it]
 42%|████▏     | 8096/19090 [3:55:42<5:08:10,  1.68s/it]
 42%|████▏     | 8097/19090 [3:55:44<4:46:59,  1.57s/it]
 42%|████▏     | 8098/19090 [3:55:46<5:37:58,  1.84s/it]
 42%|████▏     | 8099/19090 [3:55:48<5:08:24,  1.68s/it]
 42%|████▏     | 8100/19090 [3:55:49<4:47:24,  1.57s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 15.933384895324707, 'learning_rate': 3.562445866209783e-05, 'recon_loss': 0.0013512185250874609, 'recon_static': 0.0003924834672943689, 'recon_motion': 0.0009587350257788784, 'recon_norm': 1.2402195543050767, 'recon_norm_std': 0.054361823573708536, 'target_norm': 1.6196850568056107, 'recon_cos': 0.685250186920166, 'gradnorm/compressor.query': 0.035714872076817505, 'gradnorm/compressor.layer0': 0.8143839635231284, 'gradnorm/compressor.layer1': 15.120161307145555, 'gradnorm/compressor.layer2': 4.124029565581227, 'gradnorm/compressor.layer3': 2.3222889301487775, 'gradnorm/compressor.layer4': 0.8687598262203712, 'gradnorm/compressor.layer5': 0.6128370291991215, 'gradnorm/compressor.layer6': 0.2789771243744618, 'gradnorm/compressor.layer7': 0.45569464053203673, 'gradnorm/decoder.layer0': 0.6347810948760951, 'gradnorm/decoder.layer1': 0.5520718381392737, 'gradnorm/decoder.layer2': 0.2545704297564033, 'gradnorm/max_param': 10.660439491271973, 'epoch': 4.24}

 42%|████▏     | 8100/19090 [3:55:49<4:47:24,  1.57s/it]
 42%|████▏     | 8101/19090 [3:55:51<5:37:08,  1.84s/it]
 42%|████▏     | 8102/19090 [3:55:53<5:06:47,  1.68s/it]
 42%|████▏     | 8103/19090 [3:55:54<4:45:45,  1.56s/it]
 42%|████▏     | 8104/19090 [3:55:57<5:43:25,  1.88s/it]
 42%|████▏     | 8105/19090 [3:55:58<5:12:08,  1.70s/it]
 42%|████▏     | 8106/19090 [3:55:59<4:49:23,  1.58s/it]
 42%|████▏     | 8107/19090 [3:56:02<5:39:36,  1.86s/it]
 42%|████▏     | 8108/19090 [3:56:03<5:08:56,  1.69s/it]
 42%|████▏     | 8109/19090 [3:56:04<4:47:23,  1.57s/it]
 42%|████▏     | 8110/19090 [3:56:06<5:21:22,  1.76s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 15.605998992919922, 'learning_rate': 3.558306119146666e-05, 'recon_loss': 0.0013775312254438177, 'recon_static': 0.0003997721149062272, 'recon_motion': 0.000977759060333483, 'recon_norm': 1.2387132972478867, 'recon_norm_std': 0.05293113086372614, 'target_norm': 1.6229689031839372, 'recon_cos': 0.6799199968576431, 'gradnorm/compressor.query': 0.04650699431793695, 'gradnorm/compressor.layer0': 0.8362039515727883, 'gradnorm/compressor.layer1': 14.713786006912924, 'gradnorm/compressor.layer2': 4.300118183020026, 'gradnorm/compressor.layer3': 2.3379490589306133, 'gradnorm/compressor.layer4': 0.8895264928259354, 'gradnorm/compressor.layer5': 0.6421009831495709, 'gradnorm/compressor.layer6': 0.2764968163676066, 'gradnorm/compressor.layer7': 0.4478741633039831, 'gradnorm/decoder.layer0': 0.7430905520040143, 'gradnorm/decoder.layer1': 0.5346419375976232, 'gradnorm/decoder.layer2': 0.2732649531229944, 'gradnorm/max_param': 10.371747016906738, 'epoch': 4.25}

 42%|████▏     | 8110/19090 [3:56:06<5:21:22,  1.76s/it]
 42%|████▏     | 8111/19090 [3:56:08<4:56:01,  1.62s/it]
 42%|████▏     | 8112/19090 [3:56:09<4:38:35,  1.52s/it]
 42%|████▏     | 8113/19090 [3:56:12<5:43:48,  1.88s/it]
 43%|████▎     | 8114/19090 [3:56:13<5:11:46,  1.70s/it]
 43%|████▎     | 8115/19090 [3:56:14<4:49:21,  1.58s/it]
 43%|████▎     | 8116/19090 [3:56:17<5:47:13,  1.90s/it]
 43%|████▎     | 8117/19090 [3:56:18<5:13:53,  1.72s/it]
 43%|████▎     | 8118/19090 [3:56:20<4:50:35,  1.59s/it]
 43%|████▎     | 8119/19090 [3:56:22<5:24:36,  1.78s/it]
 43%|████▎     | 8120/19090 [3:56:23<4:58:03,  1.63s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 14.899496078491211, 'learning_rate': 3.55416283362546e-05, 'recon_loss': 0.0013709363265661523, 'recon_static': 0.00041267848864663393, 'recon_motion': 0.0009582578117260709, 'recon_norm': 1.2469926446676254, 'recon_norm_std': 0.05363795608282089, 'target_norm': 1.620527446269989, 'recon_cos': 0.6806455686688423, 'gradnorm/compressor.query': 0.032489143957535785, 'gradnorm/compressor.layer0': 0.7555395824911538, 'gradnorm/compressor.layer1': 14.103329399346963, 'gradnorm/compressor.layer2': 3.9068457797612615, 'gradnorm/compressor.layer3': 2.2264845449620974, 'gradnorm/compressor.layer4': 0.856255087176059, 'gradnorm/compressor.layer5': 0.6159825211497478, 'gradnorm/compressor.layer6': 0.2785950893769987, 'gradnorm/compressor.layer7': 0.4621569134210082, 'gradnorm/decoder.layer0': 0.7031416996986037, 'gradnorm/decoder.layer1': 0.5656701401881292, 'gradnorm/decoder.layer2': 0.28340992447480284, 'gradnorm/max_param': 9.9423828125, 'epoch': 4.25}

 43%|████▎     | 8120/19090 [3:56:23<4:58:03,  1.63s/it]
 43%|████▎     | 8121/19090 [3:56:24<4:39:54,  1.53s/it]
 43%|████▎     | 8122/19090 [3:56:27<5:31:51,  1.82s/it]
 43%|████▎     | 8123/19090 [3:56:28<5:03:01,  1.66s/it]
 43%|████▎     | 8124/19090 [3:56:29<4:44:18,  1.56s/it]
 43%|████▎     | 8125/19090 [3:56:32<6:01:52,  1.98s/it]
 43%|████▎     | 8126/19090 [3:56:34<5:24:16,  1.77s/it]
 43%|████▎     | 8127/19090 [3:56:35<4:57:53,  1.63s/it]
 43%|████▎     | 8128/19090 [3:56:38<6:08:11,  2.02s/it]
 43%|████▎     | 8129/19090 [3:56:39<5:28:45,  1.80s/it]
 43%|████▎     | 8130/19090 [3:56:40<5:01:42,  1.65s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 15.69885540008545, 'learning_rate': 3.5500160234992866e-05, 'recon_loss': 0.0013675042107934132, 'recon_static': 0.00040441097735310905, 'recon_motion': 0.000963093203608878, 'recon_norm': 1.2454038232564926, 'recon_norm_std': 0.05315378550440073, 'target_norm': 1.6198692560195922, 'recon_cos': 0.6815577805042267, 'gradnorm/compressor.query': 0.03584341354937652, 'gradnorm/compressor.layer0': 0.7609712111507813, 'gradnorm/compressor.layer1': 14.729260880934685, 'gradnorm/compressor.layer2': 4.474965927591023, 'gradnorm/compressor.layer3': 2.551822467368814, 'gradnorm/compressor.layer4': 0.9289043878868695, 'gradnorm/compressor.layer5': 0.5838117312397789, 'gradnorm/compressor.layer6': 0.28261576285065776, 'gradnorm/compressor.layer7': 0.4480427483233087, 'gradnorm/decoder.layer0': 0.7550074386861801, 'gradnorm/decoder.layer1': 0.5075079958619372, 'gradnorm/decoder.layer2': 0.2712876959066664, 'gradnorm/max_param': 10.3845853805542, 'epoch': 4.26}

 43%|████▎     | 8130/19090 [3:56:41<5:01:42,  1.65s/it]
 43%|████▎     | 8131/19090 [3:56:44<6:22:09,  2.09s/it]
 43%|████▎     | 8132/19090 [3:56:45<5:38:46,  1.85s/it]
 43%|████▎     | 8133/19090 [3:56:46<5:08:06,  1.69s/it]
 43%|████▎     | 8134/19090 [3:56:49<6:02:59,  1.99s/it]
 43%|████▎     | 8135/19090 [3:56:50<5:24:44,  1.78s/it]
 43%|████▎     | 8136/19090 [3:56:51<4:57:54,  1.63s/it]
 43%|████▎     | 8137/19090 [3:56:54<6:02:50,  1.99s/it]
 43%|████▎     | 8138/19090 [3:56:56<5:24:41,  1.78s/it]
 43%|████▎     | 8139/19090 [3:56:57<4:57:59,  1.63s/it]
 43%|████▎     | 8140/19090 [3:57:00<6:13:00,  2.04s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 15.617435455322266, 'learning_rate': 3.54586570263305e-05, 'recon_loss': 0.0013638899370562285, 'recon_static': 0.00040615024627186356, 'recon_motion': 0.0009577396980603225, 'recon_norm': 1.2409439653158187, 'recon_norm_std': 0.05454563656821847, 'target_norm': 1.620220896601677, 'recon_cos': 0.6822050184011459, 'gradnorm/compressor.query': 0.03807727244426089, 'gradnorm/compressor.layer0': 0.84577011301209, 'gradnorm/compressor.layer1': 14.729391928123361, 'gradnorm/compressor.layer2': 4.322177808501263, 'gradnorm/compressor.layer3': 2.2538312506139175, 'gradnorm/compressor.layer4': 0.9043757031483477, 'gradnorm/compressor.layer5': 0.5982011791574736, 'gradnorm/compressor.layer6': 0.2897216910026884, 'gradnorm/compressor.layer7': 0.44035181342464386, 'gradnorm/decoder.layer0': 0.799435522190797, 'gradnorm/decoder.layer1': 0.5332268789959815, 'gradnorm/decoder.layer2': 0.30745295424451047, 'gradnorm/max_param': 10.38306713104248, 'epoch': 4.26}

 43%|████▎     | 8140/19090 [3:57:00<6:13:00,  2.04s/it]
 43%|████▎     | 8141/19090 [3:57:01<5:32:15,  1.82s/it]
 43%|████▎     | 8142/19090 [3:57:02<5:03:12,  1.66s/it]
 43%|████▎     | 8143/19090 [3:57:05<5:53:33,  1.94s/it]
 43%|████▎     | 8144/19090 [3:57:06<5:18:15,  1.74s/it]
 43%|████▎     | 8145/19090 [3:57:08<4:53:22,  1.61s/it]
 43%|████▎     | 8146/19090 [3:57:10<5:56:50,  1.96s/it]
 43%|████▎     | 8147/19090 [3:57:12<5:20:28,  1.76s/it]
 43%|████▎     | 8148/19090 [3:57:13<4:55:00,  1.62s/it]
 43%|████▎     | 8149/19090 [3:57:16<5:53:45,  1.94s/it]
 43%|████▎     | 8150/19090 [3:57:17<5:18:12,  1.75s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 14.499128341674805, 'learning_rate': 3.5417118849033937e-05, 'recon_loss': 0.0013621748046716675, 'recon_static': 0.00040566762472735716, 'recon_motion': 0.0009565071493852883, 'recon_norm': 1.2433339565992356, 'recon_norm_std': 0.05408426029607653, 'target_norm': 1.6187506049871445, 'recon_cos': 0.6822414606809616, 'gradnorm/compressor.query': 0.033142488522880285, 'gradnorm/compressor.layer0': 0.7152910761126807, 'gradnorm/compressor.layer1': 13.639934788880842, 'gradnorm/compressor.layer2': 4.119210775762893, 'gradnorm/compressor.layer3': 2.1725794246715924, 'gradnorm/compressor.layer4': 0.7733217403211581, 'gradnorm/compressor.layer5': 0.5785738517106662, 'gradnorm/compressor.layer6': 0.284923157726985, 'gradnorm/compressor.layer7': 0.43000112705304794, 'gradnorm/decoder.layer0': 0.6686358773371366, 'gradnorm/decoder.layer1': 0.5075850637677075, 'gradnorm/decoder.layer2': 0.2694772445090129, 'gradnorm/max_param': 9.616106986999512, 'epoch': 4.27}

 43%|████▎     | 8150/19090 [3:57:17<5:18:12,  1.75s/it]
 43%|████▎     | 8151/19090 [3:57:18<4:53:31,  1.61s/it]
 43%|████▎     | 8152/19090 [3:57:21<5:54:53,  1.95s/it]
 43%|████▎     | 8153/19090 [3:57:22<5:18:55,  1.75s/it]
 43%|████▎     | 8154/19090 [3:57:24<4:54:26,  1.62s/it]
 43%|████▎     | 8155/19090 [3:57:26<5:38:59,  1.86s/it]
 43%|████▎     | 8156/19090 [3:57:27<5:08:03,  1.69s/it]
 43%|████▎     | 8157/19090 [3:57:29<4:46:38,  1.57s/it]
 43%|████▎     | 8158/19090 [3:57:31<5:40:50,  1.87s/it]
 43%|████▎     | 8159/19090 [3:57:32<5:09:19,  1.70s/it]
 43%|████▎     | 8160/19090 [3:57:34<4:47:24,  1.58s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 15.865509033203125, 'learning_rate': 3.5375545841986525e-05, 'recon_loss': 0.001347760934731923, 'recon_static': 0.0003937841269362252, 'recon_motion': 0.000953976834716741, 'recon_norm': 1.2373215138912201, 'recon_norm_std': 0.05388701390475035, 'target_norm': 1.6178339183330537, 'recon_cos': 0.6855871617794037, 'gradnorm/compressor.query': 0.0351130697193282, 'gradnorm/compressor.layer0': 0.7750802752235532, 'gradnorm/compressor.layer1': 15.053145800341142, 'gradnorm/compressor.layer2': 4.120595024991579, 'gradnorm/compressor.layer3': 2.3244376456139166, 'gradnorm/compressor.layer4': 0.813495539596137, 'gradnorm/compressor.layer5': 0.6477454525542443, 'gradnorm/compressor.layer6': 0.2764957654578735, 'gradnorm/compressor.layer7': 0.4404741252125985, 'gradnorm/decoder.layer0': 0.6554062732843311, 'gradnorm/decoder.layer1': 0.5217158914483729, 'gradnorm/decoder.layer2': 0.28163551029634987, 'gradnorm/max_param': 10.613739013671875, 'epoch': 4.27}

 43%|████▎     | 8160/19090 [3:57:34<4:47:24,  1.58s/it]
 43%|████▎     | 8161/19090 [3:57:36<5:41:39,  1.88s/it]
 43%|████▎     | 8162/19090 [3:57:38<5:09:53,  1.70s/it]
 43%|████▎     | 8163/19090 [3:57:39<4:47:09,  1.58s/it]
 43%|████▎     | 8164/19090 [3:57:42<5:50:15,  1.92s/it]
 43%|████▎     | 8165/19090 [3:57:43<5:15:37,  1.73s/it]
 43%|████▎     | 8166/19090 [3:57:44<4:51:24,  1.60s/it]
 43%|████▎     | 8167/19090 [3:57:47<5:58:31,  1.97s/it]
 43%|████▎     | 8168/19090 [3:57:48<5:21:13,  1.76s/it]
 43%|████▎     | 8169/19090 [3:57:50<4:55:31,  1.62s/it]
 43%|████▎     | 8170/19090 [3:57:52<5:50:18,  1.92s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 15.287534713745117, 'learning_rate': 3.533393814418806e-05, 'recon_loss': 0.0013610411842819303, 'recon_static': 0.0004049043927807361, 'recon_motion': 0.0009561368016875349, 'recon_norm': 1.2423414498567582, 'recon_norm_std': 0.05331621700897813, 'target_norm': 1.6191405266523362, 'recon_cos': 0.6827109828591347, 'gradnorm/compressor.query': 0.0334209805251187, 'gradnorm/compressor.layer0': 0.7961393214126697, 'gradnorm/compressor.layer1': 14.403620349578945, 'gradnorm/compressor.layer2': 4.106528336891084, 'gradnorm/compressor.layer3': 2.4919171321187967, 'gradnorm/compressor.layer4': 1.016662947590597, 'gradnorm/compressor.layer5': 0.6839496013313636, 'gradnorm/compressor.layer6': 0.2825320217669275, 'gradnorm/compressor.layer7': 0.4444226358530102, 'gradnorm/decoder.layer0': 0.6391279825923987, 'gradnorm/decoder.layer1': 0.5162482685124731, 'gradnorm/decoder.layer2': 0.28727664097466743, 'gradnorm/max_param': 10.154219627380371, 'epoch': 4.28}

 43%|████▎     | 8170/19090 [3:57:52<5:50:18,  1.92s/it]
 43%|████▎     | 8171/19090 [3:57:54<5:15:34,  1.73s/it]
 43%|████▎     | 8172/19090 [3:57:55<4:51:29,  1.60s/it]
 43%|████▎     | 8173/19090 [3:57:58<5:51:39,  1.93s/it]
 43%|████▎     | 8174/19090 [3:57:59<5:17:05,  1.74s/it]
 43%|████▎     | 8175/19090 [3:58:00<4:52:32,  1.61s/it]
 43%|████▎     | 8176/19090 [3:58:03<5:58:36,  1.97s/it]
 43%|████▎     | 8177/19090 [3:58:04<5:21:32,  1.77s/it]
 43%|████▎     | 8178/19090 [3:58:06<4:55:21,  1.62s/it]
 43%|████▎     | 8179/19090 [3:58:08<5:58:57,  1.97s/it]
 43%|████▎     | 8180/19090 [3:58:10<5:21:29,  1.77s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 15.22900390625, 'learning_rate': 3.529229589475436e-05, 'recon_loss': 0.0014082523295655847, 'recon_static': 0.00043699955131160095, 'recon_motion': 0.0009712528131785803, 'recon_norm': 1.2568197697401047, 'recon_norm_std': 0.052547957934439185, 'target_norm': 1.6232580453157426, 'recon_cos': 0.6726834833621979, 'gradnorm/compressor.query': 0.03653057355066812, 'gradnorm/compressor.layer0': 0.8688230813113883, 'gradnorm/compressor.layer1': 14.377467200809452, 'gradnorm/compressor.layer2': 4.077842264130373, 'gradnorm/compressor.layer3': 2.3413432670627286, 'gradnorm/compressor.layer4': 0.8932125249411595, 'gradnorm/compressor.layer5': 0.6178396135862891, 'gradnorm/compressor.layer6': 0.29910233327012264, 'gradnorm/compressor.layer7': 0.44513225253741284, 'gradnorm/decoder.layer0': 0.7014770693684809, 'gradnorm/decoder.layer1': 0.5607157329934613, 'gradnorm/decoder.layer2': 0.2655507292415882, 'gradnorm/max_param': 10.134530067443848, 'epoch': 4.28}

 43%|████▎     | 8180/19090 [3:58:10<5:21:29,  1.77s/it]
 43%|████▎     | 8181/19090 [3:58:11<4:55:34,  1.63s/it]
 43%|████▎     | 8182/19090 [3:58:13<5:42:32,  1.88s/it]
 43%|████▎     | 8183/19090 [3:58:15<5:10:09,  1.71s/it]
 43%|████▎     | 8184/19090 [3:58:16<4:47:21,  1.58s/it]
 43%|████▎     | 8185/19090 [3:58:19<5:51:43,  1.94s/it]
 43%|████▎     | 8186/19090 [3:58:20<5:16:28,  1.74s/it]
 43%|████▎     | 8187/19090 [3:58:21<4:51:49,  1.61s/it]
 43%|████▎     | 8188/19090 [3:58:24<5:46:04,  1.90s/it]
 43%|████▎     | 8189/19090 [3:58:25<5:12:38,  1.72s/it]
 43%|████▎     | 8190/19090 [3:58:27<4:49:21,  1.59s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 17.677698135375977, 'learning_rate': 3.5250619232916716e-05, 'recon_loss': 0.0013938916847109795, 'recon_static': 0.00043901560638914816, 'recon_motion': 0.0009548760441248305, 'recon_norm': 1.2544024914503098, 'recon_norm_std': 0.053741992078721525, 'target_norm': 1.6205167382955552, 'recon_cos': 0.6745602533221244, 'gradnorm/compressor.query': 0.03838174823983617, 'gradnorm/compressor.layer0': 0.9086104597291937, 'gradnorm/compressor.layer1': 16.732607793593285, 'gradnorm/compressor.layer2': 4.762690606896127, 'gradnorm/compressor.layer3': 2.4764030718104566, 'gradnorm/compressor.layer4': 0.9963434243042697, 'gradnorm/compressor.layer5': 0.64997839249863, 'gradnorm/compressor.layer6': 0.34690862269303613, 'gradnorm/compressor.layer7': 0.500418368783174, 'gradnorm/decoder.layer0': 0.7711670041676807, 'gradnorm/decoder.layer1': 0.6391966643955391, 'gradnorm/decoder.layer2': 0.29788081541399675, 'gradnorm/max_param': 11.79660701751709, 'epoch': 4.29}

 43%|████▎     | 8190/19090 [3:58:27<4:49:21,  1.59s/it]
 43%|████▎     | 8191/19090 [3:58:29<5:47:09,  1.91s/it]
 43%|████▎     | 8192/19090 [3:58:30<5:13:11,  1.72s/it]
 43%|████▎     | 8193/19090 [3:58:32<4:49:41,  1.60s/it]
 43%|████▎     | 8194/19090 [3:58:34<5:39:37,  1.87s/it]
 43%|████▎     | 8195/19090 [3:58:36<5:07:56,  1.70s/it]
 43%|████▎     | 8196/19090 [3:58:37<4:48:34,  1.59s/it]
 43%|████▎     | 8197/19090 [3:58:39<5:34:05,  1.84s/it]
 43%|████▎     | 8198/19090 [3:58:41<5:04:15,  1.68s/it]
 43%|████▎     | 8199/19090 [3:58:42<5:06:59,  1.69s/it]
 43%|████▎     | 8200/19090 [3:58:44<5:30:05,  1.82s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 14.816134452819824, 'learning_rate': 3.520890829802151e-05, 'recon_loss': 0.001385184348328039, 'recon_static': 0.0004193277862214018, 'recon_motion': 0.000965856590482872, 'recon_norm': 1.2480970650911332, 'recon_norm_std': 0.05227773524820804, 'target_norm': 1.6214043349027634, 'recon_cos': 0.6772478893399239, 'gradnorm/compressor.query': 0.034400150597121044, 'gradnorm/compressor.layer0': 0.8090083215872569, 'gradnorm/compressor.layer1': 13.906507853999845, 'gradnorm/compressor.layer2': 4.145092666381612, 'gradnorm/compressor.layer3': 2.3628357018021666, 'gradnorm/compressor.layer4': 0.9799332695254195, 'gradnorm/compressor.layer5': 0.673919759960331, 'gradnorm/compressor.layer6': 0.3178804223351549, 'gradnorm/compressor.layer7': 0.4657912015315713, 'gradnorm/decoder.layer0': 0.7403805872667549, 'gradnorm/decoder.layer1': 0.5925559032802459, 'gradnorm/decoder.layer2': 0.2732317968196612, 'gradnorm/max_param': 9.801877975463867, 'epoch': 4.3}

 43%|████▎     | 8200/19090 [3:58:44<5:30:05,  1.82s/it]
 43%|████▎     | 8201/19090 [3:58:46<5:01:08,  1.66s/it]
 43%|████▎     | 8202/19090 [3:58:48<5:53:31,  1.95s/it]
 43%|████▎     | 8203/19090 [3:58:50<5:17:31,  1.75s/it]
 43%|████▎     | 8204/19090 [3:58:51<4:52:25,  1.61s/it]
 43%|████▎     | 8205/19090 [3:58:54<6:09:24,  2.04s/it]
 43%|████▎     | 8206/19090 [3:58:55<5:28:45,  1.81s/it]
 43%|████▎     | 8207/19090 [3:58:57<5:00:14,  1.66s/it]
 43%|████▎     | 8208/19090 [3:58:59<6:04:12,  2.01s/it]
 43%|████▎     | 8209/19090 [3:59:01<5:25:15,  1.79s/it]
 43%|████▎     | 8210/19090 [3:59:02<4:58:04,  1.64s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 14.647950172424316, 'learning_rate': 3.5167163229529716e-05, 'recon_loss': 0.0013820292049786076, 'recon_static': 0.00042564665636746213, 'recon_motion': 0.0009563825544319116, 'recon_norm': 1.253312736749649, 'recon_norm_std': 0.052297910675406456, 'target_norm': 1.6196287721395493, 'recon_cos': 0.6777926951646804, 'gradnorm/compressor.query': 0.04532095906207909, 'gradnorm/compressor.layer0': 0.9113773128285019, 'gradnorm/compressor.layer1': 13.765790161111916, 'gradnorm/compressor.layer2': 4.10234228078554, 'gradnorm/compressor.layer3': 2.16434341976029, 'gradnorm/compressor.layer4': 0.9858159614530756, 'gradnorm/compressor.layer5': 0.6091538786122049, 'gradnorm/compressor.layer6': 0.3308726122301252, 'gradnorm/compressor.layer7': 0.5050279419788699, 'gradnorm/decoder.layer0': 0.7949574530510469, 'gradnorm/decoder.layer1': 0.558210094453777, 'gradnorm/decoder.layer2': 0.2611657606179962, 'gradnorm/max_param': 9.701743125915527, 'epoch': 4.3}

 43%|████▎     | 8210/19090 [3:59:02<4:58:04,  1.64s/it]
 43%|████▎     | 8211/19090 [3:59:04<5:36:39,  1.86s/it]
 43%|████▎     | 8212/19090 [3:59:06<5:20:51,  1.77s/it]
 43%|████▎     | 8213/19090 [3:59:07<4:54:51,  1.63s/it]
 43%|████▎     | 8214/19090 [3:59:10<5:40:54,  1.88s/it]
 43%|████▎     | 8215/19090 [3:59:11<5:29:42,  1.82s/it]
 43%|████▎     | 8216/19090 [3:59:13<5:00:53,  1.66s/it]
 43%|████▎     | 8217/19090 [3:59:15<5:31:11,  1.83s/it]
 43%|████▎     | 8218/19090 [3:59:17<5:31:55,  1.83s/it]
 43%|████▎     | 8219/19090 [3:59:18<5:02:17,  1.67s/it]
 43%|████▎     | 8220/19090 [3:59:20<5:23:12,  1.78s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 17.064300537109375, 'learning_rate': 3.5125384167016415e-05, 'recon_loss': 0.0014013677573530003, 'recon_static': 0.00043570859925239346, 'recon_motion': 0.0009656591559178196, 'recon_norm': 1.2539164781570435, 'recon_norm_std': 0.052326601929962635, 'target_norm': 1.6223285913467407, 'recon_cos': 0.6738849312067032, 'gradnorm/compressor.query': 0.0399736867979832, 'gradnorm/compressor.layer0': 0.9241673368985286, 'gradnorm/compressor.layer1': 15.968470501926552, 'gradnorm/compressor.layer2': 5.077028126730181, 'gradnorm/compressor.layer3': 2.5995790177427818, 'gradnorm/compressor.layer4': 0.9863951405335123, 'gradnorm/compressor.layer5': 0.5940929978120952, 'gradnorm/compressor.layer6': 0.31509191184475777, 'gradnorm/compressor.layer7': 0.45471678046368263, 'gradnorm/decoder.layer0': 0.8639925619727971, 'gradnorm/decoder.layer1': 0.5884580693471118, 'gradnorm/decoder.layer2': 0.2895093891477529, 'gradnorm/max_param': 11.256429672241211, 'epoch': 4.31}

 43%|████▎     | 8220/19090 [3:59:20<5:23:12,  1.78s/it]
 43%|████▎     | 8221/19090 [3:59:22<5:16:20,  1.75s/it]
 43%|████▎     | 8222/19090 [3:59:23<4:51:16,  1.61s/it]
 43%|████▎     | 8223/19090 [3:59:25<5:08:56,  1.71s/it]
 43%|████▎     | 8224/19090 [3:59:27<5:48:50,  1.93s/it]
 43%|████▎     | 8225/19090 [3:59:29<5:14:01,  1.73s/it]
 43%|████▎     | 8226/19090 [3:59:31<5:28:52,  1.82s/it]
 43%|████▎     | 8227/19090 [3:59:32<5:29:48,  1.82s/it]
 43%|████▎     | 8228/19090 [3:59:34<5:00:28,  1.66s/it]
 43%|████▎     | 8229/19090 [3:59:36<5:24:40,  1.79s/it]
 43%|████▎     | 8230/19090 [3:59:38<5:46:00,  1.91s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 18.64939308166504, 'learning_rate': 3.508357125017038e-05, 'recon_loss': 0.001375556958373636, 'recon_static': 0.00041289782820967956, 'recon_motion': 0.0009626590996049345, 'recon_norm': 1.2464026629924774, 'recon_norm_std': 0.05281181987375021, 'target_norm': 1.6204507946968079, 'recon_cos': 0.6794654220342636, 'gradnorm/compressor.query': 0.03718139536222102, 'gradnorm/compressor.layer0': 0.8982764058190668, 'gradnorm/compressor.layer1': 17.711811086196832, 'gradnorm/compressor.layer2': 4.866700174050837, 'gradnorm/compressor.layer3': 2.605939053435388, 'gradnorm/compressor.layer4': 0.9770586813207445, 'gradnorm/compressor.layer5': 0.6520302070410052, 'gradnorm/compressor.layer6': 0.3192180450924612, 'gradnorm/compressor.layer7': 0.4833785698401022, 'gradnorm/decoder.layer0': 0.8130701705426521, 'gradnorm/decoder.layer1': 0.580086191037802, 'gradnorm/decoder.layer2': 0.3078089040943147, 'gradnorm/max_param': 12.488190650939941, 'epoch': 4.31}

 43%|████▎     | 8230/19090 [3:59:38<5:46:00,  1.91s/it]
 43%|████▎     | 8231/19090 [3:59:39<5:12:08,  1.72s/it]
 43%|████▎     | 8232/19090 [3:59:41<5:05:05,  1.69s/it]
 43%|████▎     | 8233/19090 [3:59:43<5:45:36,  1.91s/it]
 43%|████▎     | 8234/19090 [3:59:45<5:11:42,  1.72s/it]
 43%|████▎     | 8235/19090 [3:59:46<4:52:02,  1.61s/it]
 43%|████▎     | 8236/19090 [3:59:49<5:45:00,  1.91s/it]
 43%|████▎     | 8237/19090 [3:59:50<5:11:03,  1.72s/it]
 43%|████▎     | 8238/19090 [3:59:51<4:49:33,  1.60s/it]
 43%|████▎     | 8239/19090 [3:59:54<5:57:03,  1.97s/it]
 43%|████▎     | 8240/19090 [3:59:55<5:19:29,  1.77s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 17.11338233947754, 'learning_rate': 3.504172461879353e-05, 'recon_loss': 0.0013886310276575386, 'recon_static': 0.00041359059177921155, 'recon_motion': 0.0009750404380611144, 'recon_norm': 1.24254352748394, 'recon_norm_std': 0.05305840224027634, 'target_norm': 1.6242168635129928, 'recon_cos': 0.6776569753885269, 'gradnorm/compressor.query': 0.034609641451918624, 'gradnorm/compressor.layer0': 0.8123059040930907, 'gradnorm/compressor.layer1': 16.235949266488372, 'gradnorm/compressor.layer2': 4.391041583189885, 'gradnorm/compressor.layer3': 2.598829695602073, 'gradnorm/compressor.layer4': 0.922887440140474, 'gradnorm/compressor.layer5': 0.6823899050354967, 'gradnorm/compressor.layer6': 0.3053761671210279, 'gradnorm/compressor.layer7': 0.47417089041665095, 'gradnorm/decoder.layer0': 0.7665367147631978, 'gradnorm/decoder.layer1': 0.5097348726431111, 'gradnorm/decoder.layer2': 0.2875607504100178, 'gradnorm/max_param': 11.447138786315918, 'epoch': 4.32}

 43%|████▎     | 8240/19090 [3:59:55<5:19:29,  1.77s/it]
 43%|████▎     | 8241/19090 [3:59:57<4:53:16,  1.62s/it]
 43%|████▎     | 8242/19090 [3:59:59<5:56:53,  1.97s/it]
 43%|████▎     | 8243/19090 [4:00:01<5:19:14,  1.77s/it]
 43%|████▎     | 8244/19090 [4:00:02<4:53:07,  1.62s/it]
 43%|████▎     | 8245/19090 [4:00:05<5:56:47,  1.97s/it]
 43%|████▎     | 8246/19090 [4:00:06<5:19:37,  1.77s/it]
 43%|████▎     | 8247/19090 [4:00:07<4:58:41,  1.65s/it]
 43%|████▎     | 8248/19090 [4:00:10<5:55:54,  1.97s/it]
 43%|████▎     | 8249/19090 [4:00:11<5:18:56,  1.77s/it]
 43%|████▎     | 8250/19090 [4:00:13<4:56:39,  1.64s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 17.42650032043457, 'learning_rate': 3.4999844412800555e-05, 'recon_loss': 0.0013662732992088423, 'recon_static': 0.00039994479520828464, 'recon_motion': 0.000966328484355472, 'recon_norm': 1.242261615395546, 'recon_norm_std': 0.0511981438845396, 'target_norm': 1.6187929421663285, 'recon_cos': 0.6811786442995071, 'gradnorm/compressor.query': 0.0365139649979568, 'gradnorm/compressor.layer0': 0.8193721649563159, 'gradnorm/compressor.layer1': 16.510787674097823, 'gradnorm/compressor.layer2': 4.6426730276637915, 'gradnorm/compressor.layer3': 2.4714846373923125, 'gradnorm/compressor.layer4': 0.9557688892781908, 'gradnorm/compressor.layer5': 0.7344858613441966, 'gradnorm/compressor.layer6': 0.3183326709991741, 'gradnorm/compressor.layer7': 0.4602692172334441, 'gradnorm/decoder.layer0': 0.7827893921873604, 'gradnorm/decoder.layer1': 0.5288765846847202, 'gradnorm/decoder.layer2': 0.2877546830562649, 'gradnorm/max_param': 11.641469955444336, 'epoch': 4.32}

 43%|████▎     | 8250/19090 [4:00:13<4:56:39,  1.64s/it]
 43%|████▎     | 8251/19090 [4:00:15<5:48:55,  1.93s/it]
 43%|████▎     | 8252/19090 [4:00:17<5:14:23,  1.74s/it]
 43%|████▎     | 8253/19090 [4:00:18<5:14:46,  1.74s/it]
 43%|████▎     | 8254/19090 [4:00:21<5:49:16,  1.93s/it]
 43%|████▎     | 8255/19090 [4:00:22<5:14:38,  1.74s/it]
 43%|████▎     | 8256/19090 [4:00:24<5:05:57,  1.69s/it]
 43%|████▎     | 8257/19090 [4:00:26<5:48:23,  1.93s/it]
 43%|████▎     | 8258/19090 [4:00:27<5:14:00,  1.74s/it]
 43%|████▎     | 8259/19090 [4:00:30<5:37:43,  1.87s/it]
 43%|████▎     | 8260/19090 [4:00:31<5:25:02,  1.80s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 18.331363677978516, 'learning_rate': 3.495793077221837e-05, 'recon_loss': 0.0013546855218010024, 'recon_static': 0.0003908047183358576, 'recon_motion': 0.0009638808114686981, 'recon_norm': 1.2340138226747512, 'recon_norm_std': 0.052914219442754984, 'target_norm': 1.6203513950109483, 'recon_cos': 0.6846170380711556, 'gradnorm/compressor.query': 0.03377769484336938, 'gradnorm/compressor.layer0': 0.8263934404228236, 'gradnorm/compressor.layer1': 17.378205147832624, 'gradnorm/compressor.layer2': 4.9068511788948, 'gradnorm/compressor.layer3': 2.557249318773815, 'gradnorm/compressor.layer4': 1.0224776681079544, 'gradnorm/compressor.layer5': 0.6333131105139002, 'gradnorm/compressor.layer6': 0.32755360352113927, 'gradnorm/compressor.layer7': 0.44985838423387936, 'gradnorm/decoder.layer0': 0.7642635385214475, 'gradnorm/decoder.layer1': 0.553864284588216, 'gradnorm/decoder.layer2': 0.2983795364372229, 'gradnorm/max_param': 12.254051208496094, 'epoch': 4.33}

 43%|████▎     | 8260/19090 [4:00:31<5:25:02,  1.80s/it]
 43%|████▎     | 8261/19090 [4:00:33<4:57:24,  1.65s/it]
 43%|████▎     | 8262/19090 [4:00:34<4:50:40,  1.61s/it]
 43%|████▎     | 8263/19090 [4:00:37<5:34:12,  1.85s/it]
 43%|████▎     | 8264/19090 [4:00:38<5:03:59,  1.68s/it]
 43%|████▎     | 8265/19090 [4:00:39<4:46:15,  1.59s/it]
 43%|████▎     | 8266/19090 [4:00:42<5:52:56,  1.96s/it]
 43%|████▎     | 8267/19090 [4:00:43<5:16:54,  1.76s/it]
 43%|████▎     | 8268/19090 [4:00:45<4:51:59,  1.62s/it]
 43%|████▎     | 8269/19090 [4:00:48<6:09:08,  2.05s/it]
 43%|████▎     | 8270/19090 [4:00:49<5:28:15,  1.82s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 17.753950119018555, 'learning_rate': 3.4915983837185704e-05, 'recon_loss': 0.0013495326827978715, 'recon_static': 0.0003913325737812556, 'recon_motion': 0.0009582000653608702, 'recon_norm': 1.2341286540031433, 'recon_norm_std': 0.05409463467076421, 'target_norm': 1.6197488844394683, 'recon_cos': 0.6855130225419999, 'gradnorm/compressor.query': 0.039582941437389545, 'gradnorm/compressor.layer0': 0.8172262718674611, 'gradnorm/compressor.layer1': 16.901624684987755, 'gradnorm/compressor.layer2': 4.419661220495874, 'gradnorm/compressor.layer3': 2.617746048724045, 'gradnorm/compressor.layer4': 0.9275012556617648, 'gradnorm/compressor.layer5': 0.6172151438642638, 'gradnorm/compressor.layer6': 0.34049395043856956, 'gradnorm/compressor.layer7': 0.47840165318523187, 'gradnorm/decoder.layer0': 0.7231841350170863, 'gradnorm/decoder.layer1': 0.5427480290173728, 'gradnorm/decoder.layer2': 0.2806865558357812, 'gradnorm/max_param': 11.917672157287598, 'epoch': 4.33}

 43%|████▎     | 8270/19090 [4:00:49<5:28:15,  1.82s/it]
 43%|████▎     | 8271/19090 [4:00:50<4:59:45,  1.66s/it]
 43%|████▎     | 8272/19090 [4:00:53<6:00:42,  2.00s/it]
 43%|████▎     | 8273/19090 [4:00:54<5:22:29,  1.79s/it]
 43%|████▎     | 8274/19090 [4:00:56<4:55:32,  1.64s/it]
 43%|████▎     | 8275/19090 [4:00:58<5:55:11,  1.97s/it]
 43%|████▎     | 8276/19090 [4:01:00<5:18:57,  1.77s/it]
 43%|████▎     | 8277/19090 [4:01:01<4:53:12,  1.63s/it]
 43%|████▎     | 8278/19090 [4:01:04<5:49:47,  1.94s/it]
 43%|████▎     | 8279/19090 [4:01:05<5:15:01,  1.75s/it]
 43%|████▎     | 8280/19090 [4:01:06<4:50:32,  1.61s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 15.318028450012207, 'learning_rate': 3.487400374795258e-05, 'recon_loss': 0.0013692615932086483, 'recon_static': 0.0004055352503200993, 'recon_motion': 0.0009637263603508473, 'recon_norm': 1.2397716492414474, 'recon_norm_std': 0.052763853967189786, 'target_norm': 1.622502726316452, 'recon_cos': 0.6819048449397087, 'gradnorm/compressor.query': 0.037121709543026085, 'gradnorm/compressor.layer0': 0.7762191151419562, 'gradnorm/compressor.layer1': 14.387598504359797, 'gradnorm/compressor.layer2': 4.375920879994705, 'gradnorm/compressor.layer3': 2.3320498569583274, 'gradnorm/compressor.layer4': 0.9684429297305761, 'gradnorm/compressor.layer5': 0.5967760523321891, 'gradnorm/compressor.layer6': 0.33754763531953513, 'gradnorm/compressor.layer7': 0.47374069715540457, 'gradnorm/decoder.layer0': 0.6634864103424704, 'gradnorm/decoder.layer1': 0.5499615894256593, 'gradnorm/decoder.layer2': 0.2699334557529061, 'gradnorm/max_param': 10.142925262451172, 'epoch': 4.34}

 43%|████▎     | 8280/19090 [4:01:06<4:50:32,  1.61s/it]
 43%|████▎     | 8281/19090 [4:01:09<5:35:16,  1.86s/it]
 43%|████▎     | 8282/19090 [4:01:10<5:04:34,  1.69s/it]
 43%|████▎     | 8283/19090 [4:01:11<4:42:59,  1.57s/it]
 43%|████▎     | 8284/19090 [4:01:14<5:50:17,  1.95s/it]
 43%|████▎     | 8285/19090 [4:01:15<5:14:56,  1.75s/it]
 43%|████▎     | 8286/19090 [4:01:17<4:50:17,  1.61s/it]
 43%|████▎     | 8287/19090 [4:01:19<5:44:44,  1.91s/it]
 43%|████▎     | 8288/19090 [4:01:21<5:11:37,  1.73s/it]
 43%|████▎     | 8289/19090 [4:01:22<4:47:34,  1.60s/it]
 43%|████▎     | 8290/19090 [4:01:24<5:43:38,  1.91s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 15.619978904724121, 'learning_rate': 3.4831990644879895e-05, 'recon_loss': 0.001417118264362216, 'recon_static': 0.00045531984360422937, 'recon_motion': 0.0009617984120268375, 'recon_norm': 1.2604003489017486, 'recon_norm_std': 0.05167797487229109, 'target_norm': 1.6220383167266845, 'recon_cos': 0.6696172758936882, 'gradnorm/compressor.query': 0.047522579299970315, 'gradnorm/compressor.layer0': 0.917653877960626, 'gradnorm/compressor.layer1': 14.596692983169916, 'gradnorm/compressor.layer2': 4.579010444917448, 'gradnorm/compressor.layer3': 2.553117929026268, 'gradnorm/compressor.layer4': 0.9879118639924304, 'gradnorm/compressor.layer5': 0.6157422588231943, 'gradnorm/compressor.layer6': 0.35207637741433223, 'gradnorm/compressor.layer7': 0.47480080907784056, 'gradnorm/decoder.layer0': 0.683774455572873, 'gradnorm/decoder.layer1': 0.5733130961920012, 'gradnorm/decoder.layer2': 0.299078954505079, 'gradnorm/max_param': 10.28636360168457, 'epoch': 4.34}

 43%|████▎     | 8290/19090 [4:01:24<5:43:38,  1.91s/it]
 43%|████▎     | 8291/19090 [4:01:26<5:10:15,  1.72s/it]
 43%|████▎     | 8292/19090 [4:01:27<4:47:12,  1.60s/it]
 43%|████▎     | 8293/19090 [4:01:30<5:45:34,  1.92s/it]
 43%|████▎     | 8294/19090 [4:01:31<5:11:40,  1.73s/it]
 43%|████▎     | 8295/19090 [4:01:32<4:48:06,  1.60s/it]
 43%|████▎     | 8296/19090 [4:01:35<5:37:16,  1.87s/it]
 43%|████▎     | 8297/19090 [4:01:36<5:05:48,  1.70s/it]
 43%|████▎     | 8298/19090 [4:01:38<4:50:13,  1.61s/it]
 43%|████▎     | 8299/19090 [4:01:40<5:48:41,  1.94s/it]
 43%|████▎     | 8300/19090 [4:01:42<5:13:53,  1.75s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 17.918880462646484, 'learning_rate': 3.4789944668438914e-05, 'recon_loss': 0.001406373255304061, 'recon_static': 0.00044414090589270925, 'recon_motion': 0.0009622323457733728, 'recon_norm': 1.2574767142534256, 'recon_norm_std': 0.05231099789962172, 'target_norm': 1.6204277336597444, 'recon_cos': 0.6716343984007835, 'gradnorm/compressor.query': 0.04669698051727143, 'gradnorm/compressor.layer0': 0.9353104453887957, 'gradnorm/compressor.layer1': 17.009944026973788, 'gradnorm/compressor.layer2': 4.667142071804564, 'gradnorm/compressor.layer3': 2.51436370126422, 'gradnorm/compressor.layer4': 0.9939172340125556, 'gradnorm/compressor.layer5': 0.6339543719193513, 'gradnorm/compressor.layer6': 0.3303458014404585, 'gradnorm/compressor.layer7': 0.48861129287636584, 'gradnorm/decoder.layer0': 0.789436969091572, 'gradnorm/decoder.layer1': 0.5632564967797093, 'gradnorm/decoder.layer2': 0.2978835792218489, 'gradnorm/max_param': 11.992594718933105, 'epoch': 4.35}

 43%|████▎     | 8300/19090 [4:01:42<5:13:53,  1.75s/it]
 43%|████▎     | 8301/19090 [4:01:43<4:55:00,  1.64s/it]
 43%|████▎     | 8302/19090 [4:01:46<5:48:44,  1.94s/it]
 43%|████▎     | 8303/19090 [4:01:47<5:13:41,  1.74s/it]
 43%|████▎     | 8304/19090 [4:01:48<4:50:32,  1.62s/it]
 44%|████▎     | 8305/19090 [4:01:51<5:55:00,  1.97s/it]
 44%|████▎     | 8306/19090 [4:01:52<5:18:15,  1.77s/it]
 44%|████▎     | 8307/19090 [4:01:54<4:52:24,  1.63s/it]
 44%|████▎     | 8308/19090 [4:01:56<5:48:11,  1.94s/it]
 44%|████▎     | 8309/19090 [4:01:58<5:13:25,  1.74s/it]
 44%|████▎     | 8310/19090 [4:01:59<4:49:32,  1.61s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 18.34478187561035, 'learning_rate': 3.474786595921081e-05, 'recon_loss': 0.001387826245627366, 'recon_static': 0.0004350742601673119, 'recon_motion': 0.0009527519650873728, 'recon_norm': 1.2549178153276443, 'recon_norm_std': 0.05233096992596984, 'target_norm': 1.6181356817483903, 'recon_cos': 0.6758111983537674, 'gradnorm/compressor.query': 0.041304957171141174, 'gradnorm/compressor.layer0': 0.9266942347112472, 'gradnorm/compressor.layer1': 17.4382682822462, 'gradnorm/compressor.layer2': 4.583354163845006, 'gradnorm/compressor.layer3': 2.778352149284858, 'gradnorm/compressor.layer4': 0.9897752709550643, 'gradnorm/compressor.layer5': 0.6599653746813348, 'gradnorm/compressor.layer6': 0.3383495213463928, 'gradnorm/compressor.layer7': 0.4745056133109075, 'gradnorm/decoder.layer0': 0.818189034225451, 'gradnorm/decoder.layer1': 0.5823338381305262, 'gradnorm/decoder.layer2': 0.29595827057968593, 'gradnorm/max_param': 12.294779777526855, 'epoch': 4.35}

 44%|████▎     | 8310/19090 [4:01:59<4:49:32,  1.61s/it]
 44%|████▎     | 8311/19090 [4:02:01<5:41:29,  1.90s/it]
 44%|████▎     | 8312/19090 [4:02:03<5:08:43,  1.72s/it]
 44%|████▎     | 8313/19090 [4:02:04<4:45:42,  1.59s/it]
 44%|████▎     | 8314/19090 [4:02:07<5:49:53,  1.95s/it]
 44%|████▎     | 8315/19090 [4:02:08<5:14:32,  1.75s/it]
 44%|████▎     | 8316/19090 [4:02:10<5:04:30,  1.70s/it]
 44%|████▎     | 8317/19090 [4:02:12<5:37:09,  1.88s/it]
 44%|████▎     | 8318/19090 [4:02:13<5:05:34,  1.70s/it]
 44%|████▎     | 8319/19090 [4:02:15<5:03:11,  1.69s/it]
 44%|████▎     | 8320/19090 [4:02:17<5:30:37,  1.84s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 18.123838424682617, 'learning_rate': 3.470575465788623e-05, 'recon_loss': 0.0013970300875371321, 'recon_static': 0.00043127109602210113, 'recon_motion': 0.0009657589689595625, 'recon_norm': 1.2498481094837188, 'recon_norm_std': 0.051509345136582854, 'target_norm': 1.6210300534963609, 'recon_cos': 0.6739328563213348, 'gradnorm/compressor.query': 0.04161039043839122, 'gradnorm/compressor.layer0': 0.9528114553554035, 'gradnorm/compressor.layer1': 17.08857182766819, 'gradnorm/compressor.layer2': 5.022503755681033, 'gradnorm/compressor.layer3': 2.742696339925952, 'gradnorm/compressor.layer4': 0.9936814062568825, 'gradnorm/compressor.layer5': 0.6895717270288344, 'gradnorm/compressor.layer6': 0.3466788815771057, 'gradnorm/compressor.layer7': 0.48346242770117953, 'gradnorm/decoder.layer0': 0.7649076565626973, 'gradnorm/decoder.layer1': 0.5572043525770208, 'gradnorm/decoder.layer2': 0.29032379200289127, 'gradnorm/max_param': 12.047574996948242, 'epoch': 4.36}

 44%|████▎     | 8320/19090 [4:02:17<5:30:37,  1.84s/it]
 44%|████▎     | 8321/19090 [4:02:18<5:04:09,  1.69s/it]
 44%|████▎     | 8322/19090 [4:02:20<5:16:39,  1.76s/it]
 44%|████▎     | 8323/19090 [4:02:22<5:34:40,  1.87s/it]
 44%|████▎     | 8324/19090 [4:02:24<5:12:00,  1.74s/it]
 44%|████▎     | 8325/19090 [4:02:26<5:28:00,  1.83s/it]
 44%|████▎     | 8326/19090 [4:02:28<5:41:51,  1.91s/it]
 44%|████▎     | 8327/19090 [4:02:29<5:16:14,  1.76s/it]
 44%|████▎     | 8328/19090 [4:02:31<5:02:34,  1.69s/it]
 44%|████▎     | 8329/19090 [4:02:34<6:01:26,  2.02s/it]
 44%|████▎     | 8330/19090 [4:02:35<5:22:59,  1.80s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 19.460498809814453, 'learning_rate': 3.466361090526475e-05, 'recon_loss': 0.0013848776899976656, 'recon_static': 0.0004229412530548871, 'recon_motion': 0.0009619364151149057, 'recon_norm': 1.249224853515625, 'recon_norm_std': 0.052346847113221885, 'target_norm': 1.620645385980606, 'recon_cos': 0.6772456467151642, 'gradnorm/compressor.query': 0.037021998384530436, 'gradnorm/compressor.layer0': 0.8902194454988369, 'gradnorm/compressor.layer1': 18.50428186300741, 'gradnorm/compressor.layer2': 5.007713471622272, 'gradnorm/compressor.layer3': 2.7422193215634945, 'gradnorm/compressor.layer4': 1.0045185522528035, 'gradnorm/compressor.layer5': 0.6990451917010513, 'gradnorm/compressor.layer6': 0.350973430485833, 'gradnorm/compressor.layer7': 0.5179291884099608, 'gradnorm/decoder.layer0': 0.7979918488490141, 'gradnorm/decoder.layer1': 0.5420890652790641, 'gradnorm/decoder.layer2': 0.3030723653892009, 'gradnorm/max_param': 13.048344612121582, 'epoch': 4.36}

 44%|████▎     | 8330/19090 [4:02:35<5:22:59,  1.80s/it]
 44%|████▎     | 8331/19090 [4:02:36<5:00:38,  1.68s/it]
 44%|████▎     | 8332/19090 [4:02:39<5:46:15,  1.93s/it]
 44%|████▎     | 8333/19090 [4:02:40<5:12:26,  1.74s/it]
 44%|████▎     | 8334/19090 [4:02:42<4:50:12,  1.62s/it]
 44%|████▎     | 8335/19090 [4:02:44<5:35:49,  1.87s/it]
 44%|████▎     | 8336/19090 [4:02:45<5:05:20,  1.70s/it]
 44%|████▎     | 8337/19090 [4:02:47<4:55:48,  1.65s/it]
 44%|████▎     | 8338/19090 [4:02:49<5:37:03,  1.88s/it]
 44%|████▎     | 8339/19090 [4:02:51<5:05:23,  1.70s/it]
 44%|████▎     | 8340/19090 [4:02:52<5:01:43,  1.68s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 18.232891082763672, 'learning_rate': 3.462143484225448e-05, 'recon_loss': 0.001361887771054171, 'recon_static': 0.00039671358390478415, 'recon_motion': 0.0009651741871493869, 'recon_norm': 1.2328389137983322, 'recon_norm_std': 0.05187898715957999, 'target_norm': 1.6235746771097184, 'recon_cos': 0.6839519038796424, 'gradnorm/compressor.query': 0.03636659244230809, 'gradnorm/compressor.layer0': 0.81026955881732, 'gradnorm/compressor.layer1': 17.317345116805036, 'gradnorm/compressor.layer2': 4.770073087440221, 'gradnorm/compressor.layer3': 2.4934935779722136, 'gradnorm/compressor.layer4': 0.9668704842953317, 'gradnorm/compressor.layer5': 0.7002388316826306, 'gradnorm/compressor.layer6': 0.3483907783727979, 'gradnorm/compressor.layer7': 0.5243150273015573, 'gradnorm/decoder.layer0': 0.8240886784468681, 'gradnorm/decoder.layer1': 0.5601531716319348, 'gradnorm/decoder.layer2': 0.3233784967313794, 'gradnorm/max_param': 12.211356163024902, 'epoch': 4.37}

 44%|████▎     | 8340/19090 [4:02:52<5:01:43,  1.68s/it]
 44%|████▎     | 8341/19090 [4:02:55<5:52:26,  1.97s/it]
 44%|████▎     | 8342/19090 [4:02:56<5:16:53,  1.77s/it]
 44%|████▎     | 8343/19090 [4:02:58<4:57:06,  1.66s/it]
 44%|████▎     | 8344/19090 [4:03:00<5:42:23,  1.91s/it]
 44%|████▎     | 8345/19090 [4:03:01<5:08:18,  1.72s/it]
 44%|████▎     | 8346/19090 [4:03:03<4:46:56,  1.60s/it]
 44%|████▎     | 8347/19090 [4:03:05<5:41:48,  1.91s/it]
 44%|████▎     | 8348/19090 [4:03:07<5:08:05,  1.72s/it]
 44%|████▎     | 8349/19090 [4:03:08<4:46:09,  1.60s/it]
 44%|████▎     | 8350/19090 [4:03:11<6:02:04,  2.02s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 19.889123916625977, 'learning_rate': 3.457922660987155e-05, 'recon_loss': 0.001354773301864043, 'recon_static': 0.00039893411012599246, 'recon_motion': 0.0009558392150211148, 'recon_norm': 1.2393027603626252, 'recon_norm_std': 0.0531718342565, 'target_norm': 1.6200878381729127, 'recon_cos': 0.6845837458968163, 'gradnorm/compressor.query': 0.034853487549180225, 'gradnorm/compressor.layer0': 0.8387414840288662, 'gradnorm/compressor.layer1': 18.9714582794272, 'gradnorm/compressor.layer2': 4.881343529033528, 'gradnorm/compressor.layer3': 2.8552324481675475, 'gradnorm/compressor.layer4': 1.0280206575173525, 'gradnorm/compressor.layer5': 0.6967990927987593, 'gradnorm/compressor.layer6': 0.3661523390617499, 'gradnorm/compressor.layer7': 0.5068876207052817, 'gradnorm/decoder.layer0': 0.8073489794331986, 'gradnorm/decoder.layer1': 0.5519993616217916, 'gradnorm/decoder.layer2': 0.2942465065997493, 'gradnorm/max_param': 13.37770938873291, 'epoch': 4.37}

 44%|████▎     | 8350/19090 [4:03:11<6:02:04,  2.02s/it]
 44%|████▎     | 8351/19090 [4:03:12<5:22:42,  1.80s/it]
 44%|████▍     | 8352/19090 [4:03:14<4:55:15,  1.65s/it]
 44%|████▍     | 8353/19090 [4:03:16<5:47:36,  1.94s/it]
 44%|████▍     | 8354/19090 [4:03:17<5:14:27,  1.76s/it]
 44%|████▍     | 8355/19090 [4:03:19<4:58:44,  1.67s/it]
 44%|████▍     | 8356/19090 [4:03:21<5:39:04,  1.90s/it]
 44%|████▍     | 8357/19090 [4:03:23<5:18:06,  1.78s/it]
 44%|████▍     | 8358/19090 [4:03:24<4:52:04,  1.63s/it]
 44%|████▍     | 8359/19090 [4:03:26<5:24:38,  1.82s/it]
 44%|████▍     | 8360/19090 [4:03:28<5:14:42,  1.76s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 20.06691551208496, 'learning_rate': 3.453698634923963e-05, 'recon_loss': 0.001368444980471395, 'recon_static': 0.00039991500161704605, 'recon_motion': 0.0009685299592092633, 'recon_norm': 1.2344140857458115, 'recon_norm_std': 0.05146495895460248, 'target_norm': 1.620837339758873, 'recon_cos': 0.68108219653368, 'gradnorm/compressor.query': 0.03839762925008499, 'gradnorm/compressor.layer0': 0.9183397233386458, 'gradnorm/compressor.layer1': 19.06667074634713, 'gradnorm/compressor.layer2': 5.173137591352321, 'gradnorm/compressor.layer3': 2.903375167662074, 'gradnorm/compressor.layer4': 1.028105652693774, 'gradnorm/compressor.layer5': 0.6802543162118178, 'gradnorm/compressor.layer6': 0.3969849306580811, 'gradnorm/compressor.layer7': 0.5238444895396331, 'gradnorm/decoder.layer0': 0.8288635343570638, 'gradnorm/decoder.layer1': 0.6032121012355404, 'gradnorm/decoder.layer2': 0.32356992397151013, 'gradnorm/max_param': 13.445140838623047, 'epoch': 4.38}

 44%|████▍     | 8360/19090 [4:03:28<5:14:42,  1.76s/it]
 44%|████▍     | 8361/19090 [4:03:29<4:49:46,  1.62s/it]
 44%|████▍     | 8362/19090 [4:03:32<5:32:25,  1.86s/it]
 44%|████▍     | 8363/19090 [4:03:33<5:14:40,  1.76s/it]
 44%|████▍     | 8364/19090 [4:03:35<4:49:23,  1.62s/it]
 44%|████▍     | 8365/19090 [4:03:37<5:38:29,  1.89s/it]
 44%|████▍     | 8366/19090 [4:03:39<5:23:59,  1.81s/it]
 44%|████▍     | 8367/19090 [4:03:40<4:56:41,  1.66s/it]
 44%|████▍     | 8368/19090 [4:03:42<5:22:50,  1.81s/it]
 44%|████▍     | 8369/19090 [4:03:44<5:18:41,  1.78s/it]
 44%|████▍     | 8370/19090 [4:03:45<4:53:33,  1.64s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 18.816299438476562, 'learning_rate': 3.449471420158951e-05, 'recon_loss': 0.0013596000731922685, 'recon_static': 0.00039357356799882834, 'recon_motion': 0.0009660265423008241, 'recon_norm': 1.2351965725421905, 'recon_norm_std': 0.05145878558978438, 'target_norm': 1.6194235324859618, 'recon_cos': 0.6829969167709351, 'gradnorm/compressor.query': 0.03515988900636456, 'gradnorm/compressor.layer0': 0.8668539419552463, 'gradnorm/compressor.layer1': 17.819153088653653, 'gradnorm/compressor.layer2': 5.064069339849978, 'gradnorm/compressor.layer3': 2.7268078537865788, 'gradnorm/compressor.layer4': 0.9808725991947082, 'gradnorm/compressor.layer5': 0.6451942270806427, 'gradnorm/compressor.layer6': 0.3442435298005521, 'gradnorm/compressor.layer7': 0.5228497920898743, 'gradnorm/decoder.layer0': 0.7265564702922052, 'gradnorm/decoder.layer1': 0.5566713023516183, 'gradnorm/decoder.layer2': 0.2997783313809165, 'gradnorm/max_param': 12.564737319946289, 'epoch': 4.38}

 44%|████▍     | 8370/19090 [4:03:45<4:53:33,  1.64s/it]
 44%|████▍     | 8371/19090 [4:03:48<5:31:25,  1.86s/it]
 44%|████▍     | 8372/19090 [4:03:49<5:07:18,  1.72s/it]
 44%|████▍     | 8373/19090 [4:03:51<5:28:27,  1.84s/it]
 44%|████▍     | 8374/19090 [4:03:53<5:33:55,  1.87s/it]
 44%|████▍     | 8375/19090 [4:03:55<5:21:54,  1.80s/it]
 44%|████▍     | 8376/19090 [4:03:56<5:09:32,  1.73s/it]
 44%|████▍     | 8377/19090 [4:03:58<5:14:37,  1.76s/it]
 44%|████▍     | 8378/19090 [4:04:00<5:13:52,  1.76s/it]
 44%|████▍     | 8379/19090 [4:04:02<5:14:04,  1.76s/it]
 44%|████▍     | 8380/19090 [4:04:04<5:25:37,  1.82s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 18.653440475463867, 'learning_rate': 3.4452410308258576e-05, 'recon_loss': 0.0013646303239511326, 'recon_static': 0.0004029814328532666, 'recon_motion': 0.0009616489318432286, 'recon_norm': 1.2391024947166442, 'recon_norm_std': 0.052462274022400376, 'target_norm': 1.6195681512355804, 'recon_cos': 0.6818092286586761, 'gradnorm/compressor.query': 0.03329306755053496, 'gradnorm/compressor.layer0': 0.8616923559025319, 'gradnorm/compressor.layer1': 17.687448238661013, 'gradnorm/compressor.layer2': 4.858689630877776, 'gradnorm/compressor.layer3': 2.8140334293573215, 'gradnorm/compressor.layer4': 0.9598009164335626, 'gradnorm/compressor.layer5': 0.6753992092277613, 'gradnorm/compressor.layer6': 0.33649495912605054, 'gradnorm/compressor.layer7': 0.52832685596976, 'gradnorm/decoder.layer0': 0.7898016008097619, 'gradnorm/decoder.layer1': 0.589563062968653, 'gradnorm/decoder.layer2': 0.30749065885079285, 'gradnorm/max_param': 12.47171401977539, 'epoch': 4.39}

 44%|████▍     | 8380/19090 [4:04:04<5:25:37,  1.82s/it]
 44%|████▍     | 8381/19090 [4:04:05<5:06:16,  1.72s/it]
 44%|████▍     | 8382/19090 [4:04:07<5:06:57,  1.72s/it]
 44%|████▍     | 8383/19090 [4:04:09<5:49:56,  1.96s/it]
 44%|████▍     | 8384/19090 [4:04:11<5:14:46,  1.76s/it]
 44%|████▍     | 8385/19090 [4:04:12<4:51:20,  1.63s/it]
 44%|████▍     | 8386/19090 [4:04:15<5:58:07,  2.01s/it]
 44%|████▍     | 8387/19090 [4:04:16<5:19:45,  1.79s/it]
 44%|████▍     | 8388/19090 [4:04:17<4:53:30,  1.65s/it]
 44%|████▍     | 8389/19090 [4:04:20<5:39:51,  1.91s/it]
 44%|████▍     | 8390/19090 [4:04:21<5:07:07,  1.72s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 19.1865177154541, 'learning_rate': 3.441007481069035e-05, 'recon_loss': 0.001346233717049472, 'recon_static': 0.0004023379544378258, 'recon_motion': 0.000943895745149348, 'recon_norm': 1.2352499455213546, 'recon_norm_std': 0.0532976932823658, 'target_norm': 1.6171086221933364, 'recon_cos': 0.6852003544569015, 'gradnorm/compressor.query': 0.03854488965978598, 'gradnorm/compressor.layer0': 0.9333266829923573, 'gradnorm/compressor.layer1': 18.19161552716749, 'gradnorm/compressor.layer2': 5.034875357334241, 'gradnorm/compressor.layer3': 2.817227692193058, 'gradnorm/compressor.layer4': 1.0434697123191183, 'gradnorm/compressor.layer5': 0.7010416744132039, 'gradnorm/compressor.layer6': 0.3545462406867766, 'gradnorm/compressor.layer7': 0.5274660886919663, 'gradnorm/decoder.layer0': 0.7609290308549568, 'gradnorm/decoder.layer1': 0.6162879520901198, 'gradnorm/decoder.layer2': 0.2919195205952073, 'gradnorm/max_param': 12.826957702636719, 'epoch': 4.39}

 44%|████▍     | 8390/19090 [4:04:21<5:07:07,  1.72s/it]
 44%|████▍     | 8391/19090 [4:04:22<4:44:18,  1.59s/it]
 44%|████▍     | 8392/19090 [4:04:25<5:32:47,  1.87s/it]
 44%|████▍     | 8393/19090 [4:04:27<5:15:09,  1.77s/it]
 44%|████▍     | 8394/19090 [4:04:28<4:50:06,  1.63s/it]
 44%|████▍     | 8395/19090 [4:04:30<5:33:21,  1.87s/it]
 44%|████▍     | 8396/19090 [4:04:32<5:05:19,  1.71s/it]
 44%|████▍     | 8397/19090 [4:04:33<4:43:07,  1.59s/it]
 44%|████▍     | 8398/19090 [4:04:35<5:30:03,  1.85s/it]
 44%|████▍     | 8399/19090 [4:04:37<5:00:43,  1.69s/it]
 44%|████▍     | 8400/19090 [4:04:38<4:56:52,  1.67s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 20.795366287231445, 'learning_rate': 3.436770785043404e-05, 'recon_loss': 0.0013609241403173655, 'recon_static': 0.0004029854171676561, 'recon_motion': 0.0009579387318808585, 'recon_norm': 1.2362312614917754, 'recon_norm_std': 0.05087185818701982, 'target_norm': 1.6218551695346832, 'recon_cos': 0.6841415241360664, 'gradnorm/compressor.query': 0.03949634019372804, 'gradnorm/compressor.layer0': 0.8824748516011638, 'gradnorm/compressor.layer1': 19.85458304204441, 'gradnorm/compressor.layer2': 5.068022929584566, 'gradnorm/compressor.layer3': 2.931327503212718, 'gradnorm/compressor.layer4': 1.130966524045273, 'gradnorm/compressor.layer5': 0.6764766224545672, 'gradnorm/compressor.layer6': 0.3664928114421559, 'gradnorm/compressor.layer7': 0.5168358140088559, 'gradnorm/decoder.layer0': 0.7658441969384994, 'gradnorm/decoder.layer1': 0.6082341445333261, 'gradnorm/decoder.layer2': 0.3005140157793454, 'gradnorm/max_param': 14.002197265625, 'epoch': 4.4}

 44%|████▍     | 8400/19090 [4:04:38<4:56:52,  1.67s/it]
 44%|████▍     | 8401/19090 [4:04:41<5:31:26,  1.86s/it]
 44%|████▍     | 8402/19090 [4:04:42<5:01:21,  1.69s/it]
 44%|████▍     | 8403/19090 [4:04:44<5:05:06,  1.71s/it]
 44%|████▍     | 8404/19090 [4:04:46<5:30:29,  1.86s/it]
 44%|████▍     | 8405/19090 [4:04:47<5:13:57,  1.76s/it]
 44%|████▍     | 8406/19090 [4:04:49<5:03:14,  1.70s/it]
 44%|████▍     | 8407/19090 [4:04:51<5:41:16,  1.92s/it]
 44%|████▍     | 8408/19090 [4:04:53<5:14:39,  1.77s/it]
 44%|████▍     | 8409/19090 [4:04:54<4:53:54,  1.65s/it]
 44%|████▍     | 8410/19090 [4:04:57<5:56:20,  2.00s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 19.790063858032227, 'learning_rate': 3.432530956914403e-05, 'recon_loss': 0.0013598145422292873, 'recon_static': 0.00039818022269173526, 'recon_motion': 0.0009616343100788072, 'recon_norm': 1.234690311551094, 'recon_norm_std': 0.05266895415261388, 'target_norm': 1.6206779062747956, 'recon_cos': 0.6833979740738869, 'gradnorm/compressor.query': 0.05058567505477518, 'gradnorm/compressor.layer0': 0.9420701847455157, 'gradnorm/compressor.layer1': 18.88800934383124, 'gradnorm/compressor.layer2': 4.7391460286824625, 'gradnorm/compressor.layer3': 2.846027007617528, 'gradnorm/compressor.layer4': 1.1070454813944983, 'gradnorm/compressor.layer5': 0.7469029856270891, 'gradnorm/compressor.layer6': 0.3906358145170472, 'gradnorm/compressor.layer7': 0.5345509858378242, 'gradnorm/decoder.layer0': 0.8759370622691633, 'gradnorm/decoder.layer1': 0.5925616368292606, 'gradnorm/decoder.layer2': 0.3164224149554233, 'gradnorm/max_param': 13.31654167175293, 'epoch': 4.41}

 44%|████▍     | 8410/19090 [4:04:57<5:56:20,  2.00s/it]
 44%|████▍     | 8411/19090 [4:04:58<5:18:38,  1.79s/it]
 44%|████▍     | 8412/19090 [4:05:00<5:00:47,  1.69s/it]
 44%|████▍     | 8413/19090 [4:05:02<5:48:02,  1.96s/it]
 44%|████▍     | 8414/19090 [4:05:04<5:12:35,  1.76s/it]
 44%|████▍     | 8415/19090 [4:05:05<5:03:39,  1.71s/it]
 44%|████▍     | 8416/19090 [4:05:07<5:31:51,  1.87s/it]
 44%|████▍     | 8417/19090 [4:05:09<5:21:36,  1.81s/it]
 44%|████▍     | 8418/19090 [4:05:10<4:54:18,  1.65s/it]
 44%|████▍     | 8419/19090 [4:05:13<5:28:16,  1.85s/it]
 44%|████▍     | 8420/19090 [4:05:14<5:18:17,  1.79s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 18.770280838012695, 'learning_rate': 3.4282880108579455e-05, 'recon_loss': 0.0013725596159929409, 'recon_static': 0.00041506108609610235, 'recon_motion': 0.0009574985422659665, 'recon_norm': 1.2425560057163239, 'recon_norm_std': 0.051729265693575145, 'target_norm': 1.619422858953476, 'recon_cos': 0.6793787986040115, 'gradnorm/compressor.query': 0.039283376822691916, 'gradnorm/compressor.layer0': 0.8971420900551569, 'gradnorm/compressor.layer1': 17.865952556250324, 'gradnorm/compressor.layer2': 4.690171967954828, 'gradnorm/compressor.layer3': 2.6961802979126857, 'gradnorm/compressor.layer4': 1.060560194743273, 'gradnorm/compressor.layer5': 0.6676462967887546, 'gradnorm/compressor.layer6': 0.36842029670497956, 'gradnorm/compressor.layer7': 0.5239111906297077, 'gradnorm/decoder.layer0': 0.8259107388950281, 'gradnorm/decoder.layer1': 0.5574499036322328, 'gradnorm/decoder.layer2': 0.28991663240109256, 'gradnorm/max_param': 12.596304893493652, 'epoch': 4.41}

 44%|████▍     | 8420/19090 [4:05:14<5:18:17,  1.79s/it]
 44%|████▍     | 8421/19090 [4:05:16<4:52:09,  1.64s/it]
 44%|████▍     | 8422/19090 [4:05:18<5:41:07,  1.92s/it]
 44%|████▍     | 8423/19090 [4:05:20<5:22:26,  1.81s/it]
 44%|████▍     | 8424/19090 [4:05:21<4:54:58,  1.66s/it]
 44%|████▍     | 8425/19090 [4:05:24<5:38:36,  1.90s/it]
 44%|████▍     | 8426/19090 [4:05:25<5:36:43,  1.89s/it]
 44%|████▍     | 8427/19090 [4:05:27<5:05:11,  1.72s/it]
 44%|████▍     | 8428/19090 [4:05:29<5:30:56,  1.86s/it]
 44%|████▍     | 8429/19090 [4:05:31<5:20:25,  1.80s/it]
 44%|████▍     | 8430/19090 [4:05:32<4:53:40,  1.65s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 20.745662689208984, 'learning_rate': 3.4240419610603667e-05, 'recon_loss': 0.001378865956212394, 'recon_static': 0.0004185484278423246, 'recon_motion': 0.000960317520366516, 'recon_norm': 1.242683580517769, 'recon_norm_std': 0.05380208743736148, 'target_norm': 1.6211047977209092, 'recon_cos': 0.6786101341247559, 'gradnorm/compressor.query': 0.040224027518324, 'gradnorm/compressor.layer0': 0.9340647702334686, 'gradnorm/compressor.layer1': 19.826449612778553, 'gradnorm/compressor.layer2': 5.034953016347924, 'gradnorm/compressor.layer3': 2.8224761356395702, 'gradnorm/compressor.layer4': 1.0466495669743854, 'gradnorm/compressor.layer5': 0.6998879411103569, 'gradnorm/compressor.layer6': 0.3592699975571497, 'gradnorm/compressor.layer7': 0.48782524177777137, 'gradnorm/decoder.layer0': 0.8611348253150027, 'gradnorm/decoder.layer1': 0.5615567139426708, 'gradnorm/decoder.layer2': 0.30648322226496943, 'gradnorm/max_param': 13.981498718261719, 'epoch': 4.42}

 44%|████▍     | 8430/19090 [4:05:32<4:53:40,  1.65s/it]
 44%|████▍     | 8431/19090 [4:05:34<5:34:00,  1.88s/it]
 44%|████▍     | 8432/19090 [4:05:36<5:02:57,  1.71s/it]
 44%|████▍     | 8433/19090 [4:05:37<4:45:34,  1.61s/it]
 44%|████▍     | 8434/19090 [4:05:39<5:33:21,  1.88s/it]
 44%|████▍     | 8435/19090 [4:05:41<5:30:14,  1.86s/it]
 44%|████▍     | 8436/19090 [4:05:43<5:00:14,  1.69s/it]
 44%|████▍     | 8437/19090 [4:05:45<5:25:38,  1.83s/it]
 44%|████▍     | 8438/19090 [4:05:47<5:24:22,  1.83s/it]
 44%|████▍     | 8439/19090 [4:05:48<4:55:50,  1.67s/it]
 44%|████▍     | 8440/19090 [4:05:50<5:35:27,  1.89s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 19.834280014038086, 'learning_rate': 3.4197928217183805e-05, 'recon_loss': 0.0013853391719749197, 'recon_static': 0.0004195943365630228, 'recon_motion': 0.0009657448987127282, 'recon_norm': 1.2429137229919434, 'recon_norm_std': 0.052771631721407176, 'target_norm': 1.6225004374980927, 'recon_cos': 0.6776472210884095, 'gradnorm/compressor.query': 0.03738681237925766, 'gradnorm/compressor.layer0': 0.9285924537080937, 'gradnorm/compressor.layer1': 18.89502588444451, 'gradnorm/compressor.layer2': 4.861441709387702, 'gradnorm/compressor.layer3': 2.979274685553103, 'gradnorm/compressor.layer4': 1.0043097252989932, 'gradnorm/compressor.layer5': 0.7068981867393133, 'gradnorm/compressor.layer6': 0.37452293886488774, 'gradnorm/compressor.layer7': 0.5008330855419204, 'gradnorm/decoder.layer0': 0.8007629020263665, 'gradnorm/decoder.layer1': 0.5973373509732093, 'gradnorm/decoder.layer2': 0.3253135008159857, 'gradnorm/max_param': 13.323451042175293, 'epoch': 4.42}

 44%|████▍     | 8440/19090 [4:05:50<5:35:27,  1.89s/it]
 44%|████▍     | 8441/19090 [4:05:52<5:36:11,  1.89s/it]
 44%|████▍     | 8442/19090 [4:05:53<5:04:03,  1.71s/it]
 44%|████▍     | 8443/19090 [4:05:55<5:19:32,  1.80s/it]
 44%|████▍     | 8444/19090 [4:05:57<5:28:13,  1.85s/it]
 44%|████▍     | 8445/19090 [4:05:59<4:59:07,  1.69s/it]
 44%|████▍     | 8446/19090 [4:06:01<5:14:48,  1.77s/it]
 44%|████▍     | 8447/19090 [4:06:03<5:26:27,  1.84s/it]
 44%|████▍     | 8448/19090 [4:06:04<4:56:57,  1.67s/it]
 44%|████▍     | 8449/19090 [4:06:06<5:08:04,  1.74s/it]
 44%|████▍     | 8450/19090 [4:06:08<5:17:45,  1.79s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 21.254425048828125, 'learning_rate': 3.415540607039032e-05, 'recon_loss': 0.0013637818774441257, 'recon_static': 0.0004091801711183507, 'recon_motion': 0.0009546017070533707, 'recon_norm': 1.2404199242591858, 'recon_norm_std': 0.05223420104011893, 'target_norm': 1.6186948835849762, 'recon_cos': 0.6813274219632148, 'gradnorm/compressor.query': 0.04012942294246092, 'gradnorm/compressor.layer0': 0.9309885874211602, 'gradnorm/compressor.layer1': 20.269840269734413, 'gradnorm/compressor.layer2': 5.259549311776884, 'gradnorm/compressor.layer3': 3.0222512611539885, 'gradnorm/compressor.layer4': 1.0867301188901966, 'gradnorm/compressor.layer5': 0.6831221779558223, 'gradnorm/compressor.layer6': 0.40092432691737984, 'gradnorm/compressor.layer7': 0.5067853075438872, 'gradnorm/decoder.layer0': 0.8278421602735361, 'gradnorm/decoder.layer1': 0.6070401626777054, 'gradnorm/decoder.layer2': 0.31613993327601947, 'gradnorm/max_param': 14.293963432312012, 'epoch': 4.43}

 44%|████▍     | 8450/19090 [4:06:08<5:17:45,  1.79s/it]
 44%|████▍     | 8451/19090 [4:06:09<4:52:44,  1.65s/it]
 44%|████▍     | 8452/19090 [4:06:11<5:03:31,  1.71s/it]
 44%|████▍     | 8453/19090 [4:06:13<5:37:02,  1.90s/it]
 44%|████▍     | 8454/19090 [4:06:15<5:07:57,  1.74s/it]
 44%|████▍     | 8455/19090 [4:06:17<5:33:37,  1.88s/it]
 44%|████▍     | 8456/19090 [4:06:18<5:10:49,  1.75s/it]
 44%|████▍     | 8457/19090 [4:06:20<4:57:39,  1.68s/it]
 44%|████▍     | 8458/19090 [4:06:22<5:05:20,  1.72s/it]
 44%|████▍     | 8459/19090 [4:06:24<5:18:31,  1.80s/it]
 44%|████▍     | 8460/19090 [4:06:25<4:52:04,  1.65s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 17.34549331665039, 'learning_rate': 3.411285331239646e-05, 'recon_loss': 0.0013751449616393075, 'recon_static': 0.0004157286894042045, 'recon_motion': 0.0009594162504072302, 'recon_norm': 1.2443135172128676, 'recon_norm_std': 0.051206125412136316, 'target_norm': 1.6190984427928925, 'recon_cos': 0.6789696350693702, 'gradnorm/compressor.query': 0.03348381071676598, 'gradnorm/compressor.layer0': 0.8315773383948741, 'gradnorm/compressor.layer1': 16.462101732677734, 'gradnorm/compressor.layer2': 4.513793996481871, 'gradnorm/compressor.layer3': 2.5093229979641065, 'gradnorm/compressor.layer4': 0.84517087331693, 'gradnorm/compressor.layer5': 0.6718746451443467, 'gradnorm/compressor.layer6': 0.3325770436207291, 'gradnorm/compressor.layer7': 0.5046595775332262, 'gradnorm/decoder.layer0': 0.759216031361902, 'gradnorm/decoder.layer1': 0.555930692128209, 'gradnorm/decoder.layer2': 0.2918883301633637, 'gradnorm/max_param': 11.606061935424805, 'epoch': 4.43}

 44%|████▍     | 8460/19090 [4:06:25<4:52:04,  1.65s/it]
 44%|████▍     | 8461/19090 [4:06:27<5:35:51,  1.90s/it]
 44%|████▍     | 8462/19090 [4:06:29<5:15:48,  1.78s/it]
 44%|████▍     | 8463/19090 [4:06:30<4:49:12,  1.63s/it]
 44%|████▍     | 8464/19090 [4:06:33<5:39:58,  1.92s/it]
 44%|████▍     | 8465/19090 [4:06:35<5:35:40,  1.90s/it]
 44%|████▍     | 8466/19090 [4:06:36<5:03:29,  1.71s/it]
 44%|████▍     | 8467/19090 [4:06:38<5:17:40,  1.79s/it]
 44%|████▍     | 8468/19090 [4:06:40<5:20:55,  1.81s/it]
 44%|████▍     | 8469/19090 [4:06:41<4:52:42,  1.65s/it]
 44%|████▍     | 8470/19090 [4:06:43<5:28:25,  1.86s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 20.192564010620117, 'learning_rate': 3.4070270085477845e-05, 'recon_loss': 0.0013863081636372953, 'recon_static': 0.00042883864443865605, 'recon_motion': 0.0009574695621267892, 'recon_norm': 1.2475240737199784, 'recon_norm_std': 0.052011205814778805, 'target_norm': 1.6206544756889343, 'recon_cos': 0.6767121016979217, 'gradnorm/compressor.query': 0.04014072943462627, 'gradnorm/compressor.layer0': 0.9293719164778576, 'gradnorm/compressor.layer1': 19.333200563873763, 'gradnorm/compressor.layer2': 4.797411562455838, 'gradnorm/compressor.layer3': 2.7103043842051218, 'gradnorm/compressor.layer4': 1.0097844548194468, 'gradnorm/compressor.layer5': 0.6869747799918152, 'gradnorm/compressor.layer6': 0.33241857488946286, 'gradnorm/compressor.layer7': 0.5096269908657489, 'gradnorm/decoder.layer0': 0.6784371461871757, 'gradnorm/decoder.layer1': 0.5553241890367749, 'gradnorm/decoder.layer2': 0.33205483296735383, 'gradnorm/max_param': 13.633283615112305, 'epoch': 4.44}

 44%|████▍     | 8470/19090 [4:06:43<5:28:25,  1.86s/it]
 44%|████▍     | 8471/19090 [4:06:45<5:28:21,  1.86s/it]
 44%|████▍     | 8472/19090 [4:06:47<4:57:49,  1.68s/it]
 44%|████▍     | 8473/19090 [4:06:49<5:17:38,  1.80s/it]
 44%|████▍     | 8474/19090 [4:06:51<5:28:13,  1.86s/it]
 44%|████▍     | 8475/19090 [4:06:52<4:58:04,  1.68s/it]
 44%|████▍     | 8476/19090 [4:06:54<4:58:27,  1.69s/it]
 44%|████▍     | 8477/19090 [4:06:56<5:45:13,  1.95s/it]
 44%|████▍     | 8478/19090 [4:06:57<5:09:51,  1.75s/it]
 44%|████▍     | 8479/19090 [4:06:59<4:52:37,  1.65s/it]
 44%|████▍     | 8480/19090 [4:07:01<5:34:31,  1.89s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 20.27027130126953, 'learning_rate': 3.402765653201197e-05, 'recon_loss': 0.0013904182153055445, 'recon_static': 0.0004260431931470521, 'recon_motion': 0.0009643750687246211, 'recon_norm': 1.2485428720712661, 'recon_norm_std': 0.05170078538358212, 'target_norm': 1.6224252223968505, 'recon_cos': 0.6767576918005943, 'gradnorm/compressor.query': 0.0417875721864151, 'gradnorm/compressor.layer0': 0.9995548628462417, 'gradnorm/compressor.layer1': 19.21099548206988, 'gradnorm/compressor.layer2': 5.452142577772017, 'gradnorm/compressor.layer3': 2.742820210293349, 'gradnorm/compressor.layer4': 1.1341246273444208, 'gradnorm/compressor.layer5': 0.7625497051360256, 'gradnorm/compressor.layer6': 0.40131986761407296, 'gradnorm/compressor.layer7': 0.5624236743854236, 'gradnorm/decoder.layer0': 0.8521599729685291, 'gradnorm/decoder.layer1': 0.620128697566347, 'gradnorm/decoder.layer2': 0.34041952270109715, 'gradnorm/max_param': 13.545442581176758, 'epoch': 4.44}

 44%|████▍     | 8480/19090 [4:07:01<5:34:31,  1.89s/it]
 44%|████▍     | 8481/19090 [4:07:03<5:02:13,  1.71s/it]
 44%|████▍     | 8482/19090 [4:07:04<4:57:34,  1.68s/it]
 44%|████▍     | 8483/19090 [4:07:06<5:20:36,  1.81s/it]
 44%|████▍     | 8484/19090 [4:07:08<4:52:20,  1.65s/it]
 44%|████▍     | 8485/19090 [4:07:09<5:02:17,  1.71s/it]
 44%|████▍     | 8486/19090 [4:07:12<5:52:13,  1.99s/it]
 44%|████▍     | 8487/19090 [4:07:13<5:14:31,  1.78s/it]
 44%|████▍     | 8488/19090 [4:07:15<4:57:09,  1.68s/it]
 44%|████▍     | 8489/19090 [4:07:17<5:48:08,  1.97s/it]
 44%|████▍     | 8490/19090 [4:07:19<5:11:48,  1.76s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 20.482961654663086, 'learning_rate': 3.398501279447769e-05, 'recon_loss': 0.001377600451814942, 'recon_static': 0.0004296633866033517, 'recon_motion': 0.0009479370492044836, 'recon_norm': 1.246263888478279, 'recon_norm_std': 0.05300436988472938, 'target_norm': 1.619070255756378, 'recon_cos': 0.6780558109283448, 'gradnorm/compressor.query': 0.0411058891564534, 'gradnorm/compressor.layer0': 1.0015629832345447, 'gradnorm/compressor.layer1': 19.48105033157823, 'gradnorm/compressor.layer2': 5.096458791348564, 'gradnorm/compressor.layer3': 3.129193505901399, 'gradnorm/compressor.layer4': 1.1045131259758214, 'gradnorm/compressor.layer5': 0.7330556746646536, 'gradnorm/compressor.layer6': 0.37532458563029447, 'gradnorm/compressor.layer7': 0.49756882107661543, 'gradnorm/decoder.layer0': 0.8238926825590015, 'gradnorm/decoder.layer1': 0.5863549335000567, 'gradnorm/decoder.layer2': 0.31850626546485017, 'gradnorm/max_param': 13.736261367797852, 'epoch': 4.45}

 44%|████▍     | 8490/19090 [4:07:19<5:11:48,  1.76s/it]
 44%|████▍     | 8491/19090 [4:07:21<5:17:03,  1.79s/it]
 44%|████▍     | 8492/19090 [4:07:23<5:32:07,  1.88s/it]
 44%|████▍     | 8493/19090 [4:07:24<5:01:05,  1.70s/it]
 44%|████▍     | 8494/19090 [4:07:26<5:08:41,  1.75s/it]
 44%|████▍     | 8495/19090 [4:07:28<5:30:37,  1.87s/it]
 45%|████▍     | 8496/19090 [4:07:29<5:00:11,  1.70s/it]
 45%|████▍     | 8497/19090 [4:07:31<5:23:21,  1.83s/it]
 45%|████▍     | 8498/19090 [4:07:33<5:22:58,  1.83s/it]
 45%|████▍     | 8499/19090 [4:07:35<4:54:22,  1.67s/it]
 45%|████▍     | 8500/19090 [4:07:37<5:33:33,  1.89s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 22.04538345336914, 'learning_rate': 3.394233901545483e-05, 'recon_loss': 0.0013710856466786936, 'recon_static': 0.0004139246710110456, 'recon_motion': 0.0009571609378326684, 'recon_norm': 1.2406875431537627, 'recon_norm_std': 0.05269766869023442, 'target_norm': 1.622045174241066, 'recon_cos': 0.6809968575835228, 'gradnorm/compressor.query': 0.0387768598567444, 'gradnorm/compressor.layer0': 0.93269982096696, 'gradnorm/compressor.layer1': 21.100286169055046, 'gradnorm/compressor.layer2': 5.230239920136779, 'gradnorm/compressor.layer3': 3.027368163964057, 'gradnorm/compressor.layer4': 1.057533267875422, 'gradnorm/compressor.layer5': 0.7977563061107785, 'gradnorm/compressor.layer6': 0.4040620278387416, 'gradnorm/compressor.layer7': 0.5481016027052341, 'gradnorm/decoder.layer0': 0.8349819961479986, 'gradnorm/decoder.layer1': 0.6067667675837134, 'gradnorm/decoder.layer2': 0.31799141852253043, 'gradnorm/max_param': 14.88084602355957, 'epoch': 4.45}

 45%|████▍     | 8500/19090 [4:07:37<5:33:33,  1.89s/it]
 45%|████▍     | 8501/19090 [4:07:41<7:29:35,  2.55s/it]
 45%|████▍     | 8502/19090 [4:07:42<6:23:01,  2.17s/it]
 45%|████▍     | 8503/19090 [4:07:44<5:37:49,  1.91s/it]
 45%|████▍     | 8504/19090 [4:07:45<5:05:07,  1.73s/it]
 45%|████▍     | 8505/19090 [4:07:46<4:42:24,  1.60s/it]
 45%|████▍     | 8506/19090 [4:07:48<4:27:17,  1.52s/it]
 45%|████▍     | 8507/19090 [4:07:49<4:16:34,  1.45s/it]
 45%|████▍     | 8508/19090 [4:07:50<4:08:36,  1.41s/it]
 45%|████▍     | 8509/19090 [4:07:51<4:02:00,  1.37s/it]
 45%|████▍     | 8510/19090 [4:07:53<3:57:04,  1.34s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 21.7542667388916, 'learning_rate': 3.389963533762364e-05, 'recon_loss': 0.0013678418414201587, 'recon_static': 0.00040509210448362863, 'recon_motion': 0.0009627497696783393, 'recon_norm': 1.2366927415132523, 'recon_norm_std': 0.05132788019254804, 'target_norm': 1.6195227265357972, 'recon_cos': 0.680860523879528, 'gradnorm/compressor.query': 0.03583234954198989, 'gradnorm/compressor.layer0': 0.8974798655044585, 'gradnorm/compressor.layer1': 20.797054023459367, 'gradnorm/compressor.layer2': 5.182509998214359, 'gradnorm/compressor.layer3': 3.0886197179460213, 'gradnorm/compressor.layer4': 1.1431165681739714, 'gradnorm/compressor.layer5': 0.7675736599469568, 'gradnorm/compressor.layer6': 0.3819810635419313, 'gradnorm/compressor.layer7': 0.5482232504014503, 'gradnorm/decoder.layer0': 0.8802565311275182, 'gradnorm/decoder.layer1': 0.5651336843864965, 'gradnorm/decoder.layer2': 0.2988662196791732, 'gradnorm/max_param': 14.665606498718262, 'epoch': 4.46}

 45%|████▍     | 8510/19090 [4:07:53<3:57:04,  1.34s/it]
 45%|████▍     | 8511/19090 [4:07:54<3:54:02,  1.33s/it]
 45%|████▍     | 8512/19090 [4:07:55<3:51:36,  1.31s/it]
 45%|████▍     | 8513/19090 [4:07:57<4:24:24,  1.50s/it]
 45%|████▍     | 8514/19090 [4:07:59<4:12:50,  1.43s/it]
 45%|████▍     | 8515/19090 [4:08:01<5:06:39,  1.74s/it]
 45%|████▍     | 8516/19090 [4:08:03<5:02:46,  1.72s/it]
 45%|████▍     | 8517/19090 [4:08:04<4:40:52,  1.59s/it]
 45%|████▍     | 8518/19090 [4:08:06<5:17:31,  1.80s/it]
 45%|████▍     | 8519/19090 [4:08:08<5:19:39,  1.81s/it]
 45%|████▍     | 8520/19090 [4:08:09<4:51:46,  1.66s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 21.20908546447754, 'learning_rate': 3.385690190376431e-05, 'recon_loss': 0.001363507509813644, 'recon_static': 0.0003931042607291602, 'recon_motion': 0.0009704032403533347, 'recon_norm': 1.2295974761247634, 'recon_norm_std': 0.052668316103518006, 'target_norm': 1.6237131744623183, 'recon_cos': 0.6837635651230812, 'gradnorm/compressor.query': 0.0392315766814216, 'gradnorm/compressor.layer0': 0.8830501607836972, 'gradnorm/compressor.layer1': 20.176433541609487, 'gradnorm/compressor.layer2': 5.358652138409296, 'gradnorm/compressor.layer3': 3.1074445856728232, 'gradnorm/compressor.layer4': 1.1443273994217131, 'gradnorm/compressor.layer5': 0.7569737069037292, 'gradnorm/compressor.layer6': 0.4074631879420903, 'gradnorm/compressor.layer7': 0.5207245967527361, 'gradnorm/decoder.layer0': 0.9202551106562598, 'gradnorm/decoder.layer1': 0.5786293767951848, 'gradnorm/decoder.layer2': 0.2877027776077221, 'gradnorm/max_param': 14.228758811950684, 'epoch': 4.46}

 45%|████▍     | 8520/19090 [4:08:09<4:51:46,  1.66s/it]
 45%|████▍     | 8521/19090 [4:08:12<5:14:52,  1.79s/it]
 45%|████▍     | 8522/19090 [4:08:13<5:19:52,  1.82s/it]
 45%|████▍     | 8523/19090 [4:08:15<4:51:39,  1.66s/it]
 45%|████▍     | 8524/19090 [4:08:17<5:35:59,  1.91s/it]
 45%|████▍     | 8525/19090 [4:08:18<5:03:15,  1.72s/it]
 45%|████▍     | 8526/19090 [4:08:20<4:56:11,  1.68s/it]
 45%|████▍     | 8527/19090 [4:08:24<6:32:49,  2.23s/it]
 45%|████▍     | 8528/19090 [4:08:25<5:42:29,  1.95s/it]
 45%|████▍     | 8529/19090 [4:08:26<5:07:37,  1.75s/it]
 45%|████▍     | 8530/19090 [4:08:28<5:30:48,  1.88s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 22.354454040527344, 'learning_rate': 3.381413885675657e-05, 'recon_loss': 0.0013527333096135408, 'recon_static': 0.00039041164200170897, 'recon_motion': 0.0009623216697946191, 'recon_norm': 1.2309331119060516, 'recon_norm_std': 0.05199083695188165, 'target_norm': 1.6202630400657654, 'recon_cos': 0.6851223677396774, 'gradnorm/compressor.query': 0.037648100577072026, 'gradnorm/compressor.layer0': 0.9372722984992774, 'gradnorm/compressor.layer1': 21.3772406937316, 'gradnorm/compressor.layer2': 5.319682931271169, 'gradnorm/compressor.layer3': 3.1525452276459913, 'gradnorm/compressor.layer4': 1.1611563228604382, 'gradnorm/compressor.layer5': 0.7723968903698744, 'gradnorm/compressor.layer6': 0.38741157584203995, 'gradnorm/compressor.layer7': 0.533015543041872, 'gradnorm/decoder.layer0': 0.8546082510462601, 'gradnorm/decoder.layer1': 0.6453426917451056, 'gradnorm/decoder.layer2': 0.3031072105024847, 'gradnorm/max_param': 15.07639217376709, 'epoch': 4.47}

 45%|████▍     | 8530/19090 [4:08:28<5:30:48,  1.88s/it]
 45%|████▍     | 8531/19090 [4:08:30<4:59:17,  1.70s/it]
 45%|████▍     | 8532/19090 [4:08:31<4:37:05,  1.57s/it]
 45%|████▍     | 8533/19090 [4:08:33<5:18:26,  1.81s/it]
 45%|████▍     | 8534/19090 [4:08:35<4:54:10,  1.67s/it]
 45%|████▍     | 8535/19090 [4:08:36<4:53:33,  1.67s/it]
 45%|████▍     | 8536/19090 [4:08:39<5:41:42,  1.94s/it]
 45%|████▍     | 8537/19090 [4:08:40<5:06:53,  1.74s/it]
 45%|████▍     | 8538/19090 [4:08:42<4:56:15,  1.68s/it]
 45%|████▍     | 8539/19090 [4:08:44<5:52:47,  2.01s/it]
 45%|████▍     | 8540/19090 [4:08:46<5:14:40,  1.79s/it]
                                                        
{'loss': 0.0013, 'grad_norm': 22.70450210571289, 'learning_rate': 3.3771346339579126e-05, 'recon_loss': 0.001355551928281784, 'recon_static': 0.0003855584858683869, 'recon_motion': 0.0009699934438685886, 'recon_norm': 1.2275955438613892, 'recon_norm_std': 0.0515282591804862, 'target_norm': 1.6229296416044234, 'recon_cos': 0.6855760931968689, 'gradnorm/compressor.query': 0.035562098238017266, 'gradnorm/compressor.layer0': 0.8831472521743655, 'gradnorm/compressor.layer1': 21.605298897022635, 'gradnorm/compressor.layer2': 5.7876462452415485, 'gradnorm/compressor.layer3': 3.169554063255571, 'gradnorm/compressor.layer4': 1.2617669347295895, 'gradnorm/compressor.layer5': 0.8471728234055788, 'gradnorm/compressor.layer6': 0.4230822491143844, 'gradnorm/compressor.layer7': 0.5872374039050567, 'gradnorm/decoder.layer0': 0.9693057096842288, 'gradnorm/decoder.layer1': 0.6976289203412663, 'gradnorm/decoder.layer2': 0.3486020477996674, 'gradnorm/max_param': 15.236773490905762, 'epoch': 4.47}

 45%|████▍     | 8540/19090 [4:08:46<5:14:40,  1.79s/it]
 45%|████▍     | 8541/19090 [4:08:47<4:49:03,  1.64s/it]
 45%|████▍     | 8542/19090 [4:08:50<5:50:52,  2.00s/it]
 45%|████▍     | 8543/19090 [4:08:51<5:13:25,  1.78s/it]
 45%|████▍     | 8544/19090 [4:08:52<4:47:29,  1.64s/it]
 45%|████▍     | 8545/19090 [4:08:55<5:46:58,  1.97s/it]
 45%|████▍     | 8546/19090 [4:08:56<5:10:28,  1.77s/it]
 45%|████▍     | 8547/19090 [4:08:58<4:45:53,  1.63s/it]
 45%|████▍     | 8548/19090 [4:09:00<5:18:19,  1.81s/it]
 45%|████▍     | 8549/19090 [4:09:01<4:50:24,  1.65s/it]
 45%|████▍     | 8550/19090 [4:09:04<5:28:54,  1.87s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 21.343700408935547, 'learning_rate': 3.372852449530922e-05, 'recon_loss': 0.0013553961063735187, 'recon_static': 0.0003953804916818626, 'recon_motion': 0.0009600156015949323, 'recon_norm': 1.232015061378479, 'recon_norm_std': 0.05251172436401248, 'target_norm': 1.619432669878006, 'recon_cos': 0.6838944345712662, 'gradnorm/compressor.query': 0.03570226507330353, 'gradnorm/compressor.layer0': 0.9157195761727354, 'gradnorm/compressor.layer1': 20.345112041578314, 'gradnorm/compressor.layer2': 5.242526047714204, 'gradnorm/compressor.layer3': 3.0749893870596994, 'gradnorm/compressor.layer4': 1.2121753395613348, 'gradnorm/compressor.layer5': 0.7857470969990948, 'gradnorm/compressor.layer6': 0.4458675773190984, 'gradnorm/compressor.layer7': 0.6046819681708067, 'gradnorm/decoder.layer0': 0.8080805379934534, 'gradnorm/decoder.layer1': 0.6542689530976679, 'gradnorm/decoder.layer2': 0.3438302076571784, 'gradnorm/max_param': 14.346508979797363, 'epoch': 4.48}

 45%|████▍     | 8550/19090 [4:09:04<5:28:54,  1.87s/it]
 45%|████▍     | 8551/19090 [4:09:05<5:01:12,  1.71s/it]
 45%|████▍     | 8552/19090 [4:09:06<4:41:01,  1.60s/it]
 45%|████▍     | 8553/19090 [4:09:08<5:06:41,  1.75s/it]
 45%|████▍     | 8554/19090 [4:09:10<5:05:33,  1.74s/it]
 45%|████▍     | 8555/19090 [4:09:12<5:02:37,  1.72s/it]
 45%|████▍     | 8556/19090 [4:09:14<5:19:23,  1.82s/it]
 45%|████▍     | 8557/19090 [4:09:16<5:13:44,  1.79s/it]
 45%|████▍     | 8558/19090 [4:09:17<5:17:09,  1.81s/it]
 45%|████▍     | 8559/19090 [4:09:20<5:38:36,  1.93s/it]
 45%|████▍     | 8560/19090 [4:09:21<5:04:44,  1.74s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 20.547868728637695, 'learning_rate': 3.3685673467122165e-05, 'recon_loss': 0.001361237687524408, 'recon_static': 0.00040393439412582665, 'recon_motion': 0.0009573032686603256, 'recon_norm': 1.2362922996282577, 'recon_norm_std': 0.05172700509428978, 'target_norm': 1.6198839277029038, 'recon_cos': 0.6827012613415718, 'gradnorm/compressor.query': 0.043704542903774145, 'gradnorm/compressor.layer0': 0.9851475363579718, 'gradnorm/compressor.layer1': 19.441011068793546, 'gradnorm/compressor.layer2': 5.46554052724933, 'gradnorm/compressor.layer3': 3.1328321346894836, 'gradnorm/compressor.layer4': 1.247943473914339, 'gradnorm/compressor.layer5': 0.7133919403326731, 'gradnorm/compressor.layer6': 0.3805007615614399, 'gradnorm/compressor.layer7': 0.5594132021427298, 'gradnorm/decoder.layer0': 0.7648993576143859, 'gradnorm/decoder.layer1': 0.622180707349346, 'gradnorm/decoder.layer2': 0.3265354619856607, 'gradnorm/max_param': 13.707517623901367, 'epoch': 4.48}

 45%|████▍     | 8560/19090 [4:09:21<5:04:44,  1.74s/it]
 45%|████▍     | 8561/19090 [4:09:23<5:02:23,  1.72s/it]
 45%|████▍     | 8562/19090 [4:09:25<5:39:17,  1.93s/it]
 45%|████▍     | 8563/19090 [4:09:27<5:16:35,  1.80s/it]
 45%|████▍     | 8564/19090 [4:09:28<5:03:03,  1.73s/it]
 45%|████▍     | 8565/19090 [4:09:31<5:44:14,  1.96s/it]
 45%|████▍     | 8566/19090 [4:09:32<5:08:27,  1.76s/it]
 45%|████▍     | 8567/19090 [4:09:34<5:11:11,  1.77s/it]
 45%|████▍     | 8568/19090 [4:09:36<5:54:27,  2.02s/it]
 45%|████▍     | 8569/19090 [4:09:38<5:15:14,  1.80s/it]
 45%|████▍     | 8570/19090 [4:09:39<4:51:12,  1.66s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 19.998287200927734, 'learning_rate': 3.364279339829084e-05, 'recon_loss': 0.0013692170759895816, 'recon_static': 0.00041040755095309576, 'recon_motion': 0.0009588095388608053, 'recon_norm': 1.2383724004030228, 'recon_norm_std': 0.050653483252972364, 'target_norm': 1.6189557403326034, 'recon_cos': 0.6799625486135483, 'gradnorm/compressor.query': 0.03765755841793608, 'gradnorm/compressor.layer0': 0.963124028328536, 'gradnorm/compressor.layer1': 19.01464841242264, 'gradnorm/compressor.layer2': 5.012744682599513, 'gradnorm/compressor.layer3': 3.039471517854563, 'gradnorm/compressor.layer4': 1.1237802781045438, 'gradnorm/compressor.layer5': 0.72877612327224, 'gradnorm/compressor.layer6': 0.3590578047708509, 'gradnorm/compressor.layer7': 0.5213943575199288, 'gradnorm/decoder.layer0': 0.690507487833083, 'gradnorm/decoder.layer1': 0.5563545964514406, 'gradnorm/decoder.layer2': 0.31392623165594097, 'gradnorm/max_param': 13.407918930053711, 'epoch': 4.49}

 45%|████▍     | 8570/19090 [4:09:39<4:51:12,  1.66s/it]
 45%|████▍     | 8571/19090 [4:09:42<5:52:38,  2.01s/it]
 45%|████▍     | 8572/19090 [4:09:43<5:14:09,  1.79s/it]
 45%|████▍     | 8573/19090 [4:09:44<4:47:12,  1.64s/it]
 45%|████▍     | 8574/19090 [4:09:47<5:30:49,  1.89s/it]
 45%|████▍     | 8575/19090 [4:09:48<4:59:27,  1.71s/it]
 45%|████▍     | 8576/19090 [4:09:50<4:46:51,  1.64s/it]
 45%|████▍     | 8577/19090 [4:09:52<5:29:54,  1.88s/it]
 45%|████▍     | 8578/19090 [4:09:53<4:58:02,  1.70s/it]
 45%|████▍     | 8579/19090 [4:09:55<4:54:56,  1.68s/it]
 45%|████▍     | 8580/19090 [4:09:57<5:26:40,  1.86s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 20.989763259887695, 'learning_rate': 3.3599884432185225e-05, 'recon_loss': 0.001380805083317682, 'recon_static': 0.0004243077230057679, 'recon_motion': 0.000956497348670382, 'recon_norm': 1.2454728215932847, 'recon_norm_std': 0.05114385662600398, 'target_norm': 1.6196260094642638, 'recon_cos': 0.6775056332349777, 'gradnorm/compressor.query': 0.04543100924221453, 'gradnorm/compressor.layer0': 1.010181452059884, 'gradnorm/compressor.layer1': 20.046976994576855, 'gradnorm/compressor.layer2': 5.0616614506736175, 'gradnorm/compressor.layer3': 2.882025055479247, 'gradnorm/compressor.layer4': 1.217550274087072, 'gradnorm/compressor.layer5': 0.7735669576789687, 'gradnorm/compressor.layer6': 0.4131551012796917, 'gradnorm/compressor.layer7': 0.5505157534227488, 'gradnorm/decoder.layer0': 0.8637245731503199, 'gradnorm/decoder.layer1': 0.5741897368075582, 'gradnorm/decoder.layer2': 0.3309826505261594, 'gradnorm/max_param': 14.136065483093262, 'epoch': 4.49}

 45%|████▍     | 8580/19090 [4:09:57<5:26:40,  1.86s/it]
 45%|████▍     | 8581/19090 [4:09:59<5:28:55,  1.88s/it]
 45%|████▍     | 8582/19090 [4:10:00<4:57:41,  1.70s/it]
 45%|████▍     | 8583/19090 [4:10:03<5:20:56,  1.83s/it]
 45%|████▍     | 8584/19090 [4:10:05<5:39:04,  1.94s/it]
 45%|████▍     | 8585/19090 [4:10:06<5:04:54,  1.74s/it]
 45%|████▍     | 8586/19090 [4:10:08<4:57:44,  1.70s/it]
 45%|████▍     | 8587/19090 [4:10:10<5:30:39,  1.89s/it]
 45%|████▍     | 8588/19090 [4:10:11<4:58:50,  1.71s/it]
 45%|████▍     | 8589/19090 [4:10:13<5:08:39,  1.76s/it]
 45%|████▍     | 8590/19090 [4:10:16<5:41:27,  1.95s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 26.098369598388672, 'learning_rate': 3.355694671227192e-05, 'recon_loss': 0.0013741200906224548, 'recon_static': 0.00041181569031323306, 'recon_motion': 0.0009623044126783498, 'recon_norm': 1.2366809606552125, 'recon_norm_std': 0.049652593676000834, 'target_norm': 1.621930992603302, 'recon_cos': 0.6800438687205315, 'gradnorm/compressor.query': 0.04129004918925192, 'gradnorm/compressor.layer0': 1.0405909260621256, 'gradnorm/compressor.layer1': 25.03114758874181, 'gradnorm/compressor.layer2': 5.9284943396699, 'gradnorm/compressor.layer3': 3.6877218923241695, 'gradnorm/compressor.layer4': 1.4390083360781794, 'gradnorm/compressor.layer5': 0.8534442863068827, 'gradnorm/compressor.layer6': 0.4316657391314808, 'gradnorm/compressor.layer7': 0.573587602724259, 'gradnorm/decoder.layer0': 0.9272194487412971, 'gradnorm/decoder.layer1': 0.6533499854951587, 'gradnorm/decoder.layer2': 0.36657838824377215, 'gradnorm/max_param': 17.652681350708008, 'epoch': 4.5}

 45%|████▍     | 8590/19090 [4:10:16<5:41:27,  1.95s/it]
 45%|████▌     | 8591/19090 [4:10:17<5:06:20,  1.75s/it]
 45%|████▌     | 8592/19090 [4:10:19<5:15:58,  1.81s/it]
 45%|████▌     | 8593/19090 [4:10:21<5:30:49,  1.89s/it]
 45%|████▌     | 8594/19090 [4:10:22<4:59:00,  1.71s/it]
 45%|████▌     | 8595/19090 [4:10:24<5:20:05,  1.83s/it]
 45%|████▌     | 8596/19090 [4:10:26<5:08:53,  1.77s/it]
 45%|████▌     | 8597/19090 [4:10:27<4:43:39,  1.62s/it]
 45%|████▌     | 8598/19090 [4:10:30<5:32:56,  1.90s/it]
 45%|████▌     | 8599/19090 [4:10:32<5:36:11,  1.92s/it]
 45%|████▌     | 8600/19090 [4:10:33<5:02:58,  1.73s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 22.22845458984375, 'learning_rate': 3.3513980382113656e-05, 'recon_loss': 0.001368153898511082, 'recon_static': 0.00039363296382362024, 'recon_motion': 0.000974520863383077, 'recon_norm': 1.2316582024097442, 'recon_norm_std': 0.05088724810630083, 'target_norm': 1.6230313628911972, 'recon_cos': 0.6822737216949463, 'gradnorm/compressor.query': 0.03789220344559166, 'gradnorm/compressor.layer0': 0.8958483332479703, 'gradnorm/compressor.layer1': 21.158020576528347, 'gradnorm/compressor.layer2': 5.650302352242129, 'gradnorm/compressor.layer3': 3.145829931261526, 'gradnorm/compressor.layer4': 1.181384400891074, 'gradnorm/compressor.layer5': 0.8486630485951487, 'gradnorm/compressor.layer6': 0.41912505377778514, 'gradnorm/compressor.layer7': 0.5473664391704735, 'gradnorm/decoder.layer0': 0.8759005884673974, 'gradnorm/decoder.layer1': 0.6028076018613499, 'gradnorm/decoder.layer2': 0.3072956645296028, 'gradnorm/max_param': 14.921357154846191, 'epoch': 4.5}

 45%|████▌     | 8600/19090 [4:10:33<5:02:58,  1.73s/it]
 45%|████▌     | 8601/19090 [4:10:35<5:09:13,  1.77s/it]
 45%|████▌     | 8602/19090 [4:10:37<5:12:30,  1.79s/it]
 45%|████▌     | 8603/19090 [4:10:38<5:00:20,  1.72s/it]
 45%|████▌     | 8604/19090 [4:10:40<5:23:07,  1.85s/it]
 45%|████▌     | 8605/19090 [4:10:42<5:08:51,  1.77s/it]
 45%|████▌     | 8606/19090 [4:10:44<5:01:13,  1.72s/it]
 45%|████▌     | 8607/19090 [4:10:45<4:54:22,  1.68s/it]
 45%|████▌     | 8608/19090 [4:10:47<5:19:27,  1.83s/it]
 45%|████▌     | 8609/19090 [4:10:49<5:03:56,  1.74s/it]
 45%|████▌     | 8610/19090 [4:10:50<4:47:49,  1.65s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 25.824268341064453, 'learning_rate': 3.347098558536883e-05, 'recon_loss': 0.001362566332682036, 'recon_static': 0.00039721523571643045, 'recon_motion': 0.0009653510700445622, 'recon_norm': 1.2286030739545821, 'recon_norm_std': 0.05091760670766234, 'target_norm': 1.6222333043813706, 'recon_cos': 0.6830828651785851, 'gradnorm/compressor.query': 0.04163020076181519, 'gradnorm/compressor.layer0': 1.0224058469779764, 'gradnorm/compressor.layer1': 24.814674416688405, 'gradnorm/compressor.layer2': 6.000620492004348, 'gradnorm/compressor.layer3': 3.1350737898567362, 'gradnorm/compressor.layer4': 1.2980990321023587, 'gradnorm/compressor.layer5': 0.8450311188921135, 'gradnorm/compressor.layer6': 0.4235035228988061, 'gradnorm/compressor.layer7': 0.6088285686070821, 'gradnorm/decoder.layer0': 0.8824120009749239, 'gradnorm/decoder.layer1': 0.6429892387231213, 'gradnorm/decoder.layer2': 0.31678824265992106, 'gradnorm/max_param': 17.499977111816406, 'epoch': 4.51}

 45%|████▌     | 8610/19090 [4:10:50<4:47:49,  1.65s/it]
 45%|████▌     | 8611/19090 [4:10:53<5:25:23,  1.86s/it]
 45%|████▌     | 8612/19090 [4:10:54<5:14:14,  1.80s/it]
 45%|████▌     | 8613/19090 [4:10:56<4:47:18,  1.65s/it]
 45%|████▌     | 8614/19090 [4:10:58<5:30:37,  1.89s/it]
 45%|████▌     | 8615/19090 [4:11:00<5:13:51,  1.80s/it]
 45%|████▌     | 8616/19090 [4:11:01<4:50:01,  1.66s/it]
 45%|████▌     | 8617/19090 [4:11:03<5:28:53,  1.88s/it]
 45%|████▌     | 8618/19090 [4:11:05<5:09:40,  1.77s/it]
 45%|████▌     | 8619/19090 [4:11:06<4:56:37,  1.70s/it]
 45%|████▌     | 8620/19090 [4:11:09<5:23:06,  1.85s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 24.671052932739258, 'learning_rate': 3.3427962465791016e-05, 'recon_loss': 0.0013615114614367485, 'recon_static': 0.0004102436112589203, 'recon_motion': 0.000951267841446679, 'recon_norm': 1.2373566657304764, 'recon_norm_std': 0.05248642200604081, 'target_norm': 1.6206488192081452, 'recon_cos': 0.6832659140229225, 'gradnorm/compressor.query': 0.04111764911817951, 'gradnorm/compressor.layer0': 1.0336358186517993, 'gradnorm/compressor.layer1': 23.69400394095321, 'gradnorm/compressor.layer2': 5.733594628227553, 'gradnorm/compressor.layer3': 3.102066575605647, 'gradnorm/compressor.layer4': 1.1961331511124187, 'gradnorm/compressor.layer5': 0.7850781302234286, 'gradnorm/compressor.layer6': 0.4317557582670031, 'gradnorm/compressor.layer7': 0.5293621571952255, 'gradnorm/decoder.layer0': 0.8134058157941552, 'gradnorm/decoder.layer1': 0.634659696690303, 'gradnorm/decoder.layer2': 0.33148337771094644, 'gradnorm/max_param': 16.70932388305664, 'epoch': 4.52}

 45%|████▌     | 8620/19090 [4:11:09<5:23:06,  1.85s/it]
 45%|████▌     | 8621/19090 [4:11:10<4:59:04,  1.71s/it]
 45%|████▌     | 8622/19090 [4:11:12<5:10:46,  1.78s/it]
 45%|████▌     | 8623/19090 [4:11:14<5:12:32,  1.79s/it]
 45%|████▌     | 8624/19090 [4:11:15<4:46:08,  1.64s/it]
 45%|████▌     | 8625/19090 [4:11:17<5:15:05,  1.81s/it]
 45%|████▌     | 8626/19090 [4:11:19<5:32:10,  1.90s/it]
 45%|████▌     | 8627/19090 [4:11:21<5:00:20,  1.72s/it]
 45%|████▌     | 8628/19090 [4:11:23<5:34:59,  1.92s/it]
 45%|████▌     | 8629/19090 [4:11:25<5:26:52,  1.87s/it]
 45%|████▌     | 8630/19090 [4:11:26<4:56:07,  1.70s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 25.53531265258789, 'learning_rate': 3.338491116722849e-05, 'recon_loss': 0.0013678012386662886, 'recon_static': 0.0004018478422949556, 'recon_motion': 0.0009659533519879915, 'recon_norm': 1.2297154307365417, 'recon_norm_std': 0.05212121494114399, 'target_norm': 1.6226229041814804, 'recon_cos': 0.6819674015045166, 'gradnorm/compressor.query': 0.039214663373378596, 'gradnorm/compressor.layer0': 0.9987076514746513, 'gradnorm/compressor.layer1': 24.357844022607736, 'gradnorm/compressor.layer2': 6.454058932871963, 'gradnorm/compressor.layer3': 3.4314498370679942, 'gradnorm/compressor.layer4': 1.321575395977164, 'gradnorm/compressor.layer5': 0.8330072400716088, 'gradnorm/compressor.layer6': 0.42886084019148707, 'gradnorm/compressor.layer7': 0.6135665570074741, 'gradnorm/decoder.layer0': 0.8513759793735327, 'gradnorm/decoder.layer1': 0.6830629086000636, 'gradnorm/decoder.layer2': 0.3558371803908056, 'gradnorm/max_param': 17.17841911315918, 'epoch': 4.52}

 45%|████▌     | 8630/19090 [4:11:26<4:56:07,  1.70s/it]
 45%|████▌     | 8631/19090 [4:11:29<5:49:04,  2.00s/it]
 45%|████▌     | 8632/19090 [4:11:30<5:11:31,  1.79s/it]
 45%|████▌     | 8633/19090 [4:11:31<4:45:20,  1.64s/it]
 45%|████▌     | 8634/19090 [4:11:34<5:56:38,  2.05s/it]
 45%|████▌     | 8635/19090 [4:11:36<5:16:42,  1.82s/it]
 45%|████▌     | 8636/19090 [4:11:37<4:48:54,  1.66s/it]
 45%|████▌     | 8637/19090 [4:11:40<5:40:09,  1.95s/it]
 45%|████▌     | 8638/19090 [4:11:41<5:05:15,  1.75s/it]
 45%|████▌     | 8639/19090 [4:11:42<4:40:46,  1.61s/it]
 45%|████▌     | 8640/19090 [4:11:45<5:37:52,  1.94s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 24.42099952697754, 'learning_rate': 3.334183183362373e-05, 'recon_loss': 0.001365688053192571, 'recon_static': 0.000401257583143888, 'recon_motion': 0.0009644305100664496, 'recon_norm': 1.2349136650562287, 'recon_norm_std': 0.051594866812229155, 'target_norm': 1.6216717094182969, 'recon_cos': 0.6822150200605392, 'gradnorm/compressor.query': 0.036203746503435254, 'gradnorm/compressor.layer0': 0.9186421623002353, 'gradnorm/compressor.layer1': 23.405187361141895, 'gradnorm/compressor.layer2': 5.824167989487888, 'gradnorm/compressor.layer3': 3.1837143202341145, 'gradnorm/compressor.layer4': 1.2287128332654318, 'gradnorm/compressor.layer5': 0.7243989705817261, 'gradnorm/compressor.layer6': 0.40671639080147415, 'gradnorm/compressor.layer7': 0.5604560858277048, 'gradnorm/decoder.layer0': 0.808690346370611, 'gradnorm/decoder.layer1': 0.6475521846126736, 'gradnorm/decoder.layer2': 0.30488694827826035, 'gradnorm/max_param': 16.506641387939453, 'epoch': 4.53}

 45%|████▌     | 8640/19090 [4:11:45<5:37:52,  1.94s/it]
 45%|████▌     | 8641/19090 [4:11:46<5:07:24,  1.77s/it]
 45%|████▌     | 8642/19090 [4:11:48<4:42:08,  1.62s/it]
 45%|████▌     | 8643/19090 [4:11:50<5:37:31,  1.94s/it]
 45%|████▌     | 8644/19090 [4:11:51<5:03:36,  1.74s/it]
 45%|████▌     | 8645/19090 [4:11:53<4:40:10,  1.61s/it]
 45%|████▌     | 8646/19090 [4:11:55<5:11:34,  1.79s/it]
 45%|████▌     | 8647/19090 [4:11:57<5:07:31,  1.77s/it]
 45%|████▌     | 8648/19090 [4:11:58<4:42:12,  1.62s/it]
 45%|████▌     | 8649/19090 [4:12:00<5:24:20,  1.86s/it]
 45%|████▌     | 8650/19090 [4:12:02<5:09:49,  1.78s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 23.287734985351562, 'learning_rate': 3.329872460901297e-05, 'recon_loss': 0.0013807259761961176, 'recon_static': 0.000425572793756146, 'recon_motion': 0.0009551531853503548, 'recon_norm': 1.2433142751455306, 'recon_norm_std': 0.050696409121155736, 'target_norm': 1.6184127628803253, 'recon_cos': 0.6772141173481941, 'gradnorm/compressor.query': 0.04057231527837893, 'gradnorm/compressor.layer0': 0.9884560471904279, 'gradnorm/compressor.layer1': 22.190571832908777, 'gradnorm/compressor.layer2': 5.8406481338785765, 'gradnorm/compressor.layer3': 3.2732290415520073, 'gradnorm/compressor.layer4': 1.2918875054287988, 'gradnorm/compressor.layer5': 0.7421633666280895, 'gradnorm/compressor.layer6': 0.3959926882596227, 'gradnorm/compressor.layer7': 0.5838044058301227, 'gradnorm/decoder.layer0': 0.8843768278177196, 'gradnorm/decoder.layer1': 0.7040278466268682, 'gradnorm/decoder.layer2': 0.3099103678439059, 'gradnorm/max_param': 15.648493766784668, 'epoch': 4.53}

 45%|████▌     | 8650/19090 [4:12:02<5:09:49,  1.78s/it]
 45%|████▌     | 8651/19090 [4:12:03<4:43:48,  1.63s/it]
 45%|████▌     | 8652/19090 [4:12:06<5:42:42,  1.97s/it]
 45%|████▌     | 8653/19090 [4:12:07<5:14:52,  1.81s/it]
 45%|████▌     | 8654/19090 [4:12:09<4:47:13,  1.65s/it]
 45%|████▌     | 8655/19090 [4:12:11<5:43:36,  1.98s/it]
 45%|████▌     | 8656/19090 [4:12:13<5:10:29,  1.79s/it]
 45%|████▌     | 8657/19090 [4:12:14<4:44:19,  1.64s/it]
 45%|████▌     | 8658/19090 [4:12:17<5:28:58,  1.89s/it]
 45%|████▌     | 8659/19090 [4:12:18<5:06:43,  1.76s/it]
 45%|████▌     | 8660/19090 [4:12:19<4:41:38,  1.62s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 24.115636825561523, 'learning_rate': 3.3255589637525674e-05, 'recon_loss': 0.0013911599497077986, 'recon_static': 0.000432329619070515, 'recon_motion': 0.0009588303408236242, 'recon_norm': 1.2479041665792465, 'recon_norm_std': 0.05047772563993931, 'target_norm': 1.6203897416591644, 'recon_cos': 0.675109326839447, 'gradnorm/compressor.query': 0.04174453458455987, 'gradnorm/compressor.layer0': 1.0285254234951209, 'gradnorm/compressor.layer1': 23.142069027783982, 'gradnorm/compressor.layer2': 5.601173291590936, 'gradnorm/compressor.layer3': 3.1049199225132567, 'gradnorm/compressor.layer4': 1.164673101051111, 'gradnorm/compressor.layer5': 0.7844861335598681, 'gradnorm/compressor.layer6': 0.4319484010418435, 'gradnorm/compressor.layer7': 0.5257841146994738, 'gradnorm/decoder.layer0': 0.9389293900023834, 'gradnorm/decoder.layer1': 0.714332314231396, 'gradnorm/decoder.layer2': 0.3295242244248278, 'gradnorm/max_param': 16.321016311645508, 'epoch': 4.54}

 45%|████▌     | 8660/19090 [4:12:19<4:41:38,  1.62s/it]
 45%|████▌     | 8661/19090 [4:12:22<5:22:30,  1.86s/it]
 45%|████▌     | 8662/19090 [4:12:23<5:13:09,  1.80s/it]
 45%|████▌     | 8663/19090 [4:12:25<4:46:34,  1.65s/it]
 45%|████▌     | 8664/19090 [4:12:27<5:17:08,  1.83s/it]
 45%|████▌     | 8665/19090 [4:12:29<5:29:32,  1.90s/it]
 45%|████▌     | 8666/19090 [4:12:30<4:58:06,  1.72s/it]
 45%|████▌     | 8667/19090 [4:12:32<5:05:50,  1.76s/it]
 45%|████▌     | 8668/19090 [4:12:34<5:20:05,  1.84s/it]
 45%|████▌     | 8669/19090 [4:12:36<4:51:09,  1.68s/it]
 45%|████▌     | 8670/19090 [4:12:38<5:15:49,  1.82s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 25.343494415283203, 'learning_rate': 3.321242706338411e-05, 'recon_loss': 0.001376626334968023, 'recon_static': 0.0004120807010622229, 'recon_motion': 0.0009645457073929719, 'recon_norm': 1.2358351439237594, 'recon_norm_std': 0.05224157879129052, 'target_norm': 1.6213352978229523, 'recon_cos': 0.6790840283036232, 'gradnorm/compressor.query': 0.03989669301161527, 'gradnorm/compressor.layer0': 1.0298007839096917, 'gradnorm/compressor.layer1': 24.387231506402586, 'gradnorm/compressor.layer2': 5.473578539399753, 'gradnorm/compressor.layer3': 3.549944788543758, 'gradnorm/compressor.layer4': 1.2716182983935813, 'gradnorm/compressor.layer5': 0.7480401419792246, 'gradnorm/compressor.layer6': 0.3802980835250183, 'gradnorm/compressor.layer7': 0.5784891631200236, 'gradnorm/decoder.layer0': 0.8869492823519965, 'gradnorm/decoder.layer1': 0.6188929691252487, 'gradnorm/decoder.layer2': 0.3254753030088898, 'gradnorm/max_param': 17.199007034301758, 'epoch': 4.54}

 45%|████▌     | 8670/19090 [4:12:38<5:15:49,  1.82s/it]
 45%|████▌     | 8671/19090 [4:12:40<5:27:54,  1.89s/it]
 45%|████▌     | 8672/19090 [4:12:41<4:57:24,  1.71s/it]
 45%|████▌     | 8673/19090 [4:12:43<4:50:09,  1.67s/it]
 45%|████▌     | 8674/19090 [4:12:45<5:30:42,  1.91s/it]
 45%|████▌     | 8675/19090 [4:12:46<4:58:40,  1.72s/it]
 45%|████▌     | 8676/19090 [4:12:48<5:03:38,  1.75s/it]
 45%|████▌     | 8677/19090 [4:12:50<5:18:44,  1.84s/it]
 45%|████▌     | 8678/19090 [4:12:51<4:49:48,  1.67s/it]
 45%|████▌     | 8679/19090 [4:12:54<5:23:09,  1.86s/it]
 45%|████▌     | 8680/19090 [4:12:55<5:13:14,  1.81s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 24.562753677368164, 'learning_rate': 3.3169237030902786e-05, 'recon_loss': 0.0013537067628931255, 'recon_static': 0.0003976140083977953, 'recon_motion': 0.0009560927705024369, 'recon_norm': 1.2352618783712388, 'recon_norm_std': 0.051479931175708774, 'target_norm': 1.61856389939785, 'recon_cos': 0.6843540579080581, 'gradnorm/compressor.query': 0.04244633788822727, 'gradnorm/compressor.layer0': 1.0368725062565078, 'gradnorm/compressor.layer1': 23.332505711001257, 'gradnorm/compressor.layer2': 6.483186838150598, 'gradnorm/compressor.layer3': 3.3569869669052457, 'gradnorm/compressor.layer4': 1.2722975888803356, 'gradnorm/compressor.layer5': 0.8557524885519777, 'gradnorm/compressor.layer6': 0.47098570766145903, 'gradnorm/compressor.layer7': 0.6570298919482263, 'gradnorm/decoder.layer0': 1.0207338219751987, 'gradnorm/decoder.layer1': 0.611401221442071, 'gradnorm/decoder.layer2': 0.3533382181382504, 'gradnorm/max_param': 16.453826904296875, 'epoch': 4.55}

 45%|████▌     | 8680/19090 [4:12:55<5:13:14,  1.81s/it]
 45%|████▌     | 8681/19090 [4:12:57<4:45:57,  1.65s/it]
 45%|████▌     | 8682/19090 [4:12:59<5:32:48,  1.92s/it]
 45%|████▌     | 8683/19090 [4:13:01<5:22:27,  1.86s/it]
 45%|████▌     | 8684/19090 [4:13:02<4:52:34,  1.69s/it]
 45%|████▌     | 8685/19090 [4:13:04<4:56:15,  1.71s/it]
 46%|████▌     | 8686/19090 [4:13:06<5:09:14,  1.78s/it]
 46%|████▌     | 8687/19090 [4:13:07<4:43:09,  1.63s/it]
 46%|████▌     | 8688/19090 [4:13:09<5:09:24,  1.78s/it]
 46%|████▌     | 8689/19090 [4:13:11<5:11:08,  1.79s/it]
 46%|████▌     | 8690/19090 [4:13:13<4:44:35,  1.64s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 22.750720977783203, 'learning_rate': 3.312601968448808e-05, 'recon_loss': 0.0013607658067485318, 'recon_static': 0.00040261400645249525, 'recon_motion': 0.0009581518301274627, 'recon_norm': 1.233545058965683, 'recon_norm_std': 0.05194130223244429, 'target_norm': 1.6194719195365905, 'recon_cos': 0.6826625302433967, 'gradnorm/compressor.query': 0.039402399348313064, 'gradnorm/compressor.layer0': 0.9685980462294225, 'gradnorm/compressor.layer1': 21.65925557471404, 'gradnorm/compressor.layer2': 5.694814484934717, 'gradnorm/compressor.layer3': 3.3443230824155785, 'gradnorm/compressor.layer4': 1.180201891987869, 'gradnorm/compressor.layer5': 0.8817190301905851, 'gradnorm/compressor.layer6': 0.4391859810844833, 'gradnorm/compressor.layer7': 0.5730933754951614, 'gradnorm/decoder.layer0': 0.8269647711869419, 'gradnorm/decoder.layer1': 0.6551823332398251, 'gradnorm/decoder.layer2': 0.3347130461523704, 'gradnorm/max_param': 15.272991180419922, 'epoch': 4.55}

 46%|████▌     | 8690/19090 [4:13:13<4:44:35,  1.64s/it]
 46%|████▌     | 8691/19090 [4:13:15<5:32:34,  1.92s/it]
 46%|████▌     | 8692/19090 [4:13:17<5:23:44,  1.87s/it]
 46%|████▌     | 8693/19090 [4:13:18<4:53:23,  1.69s/it]
 46%|████▌     | 8694/19090 [4:13:20<5:22:45,  1.86s/it]
 46%|████▌     | 8695/19090 [4:13:23<5:37:21,  1.95s/it]
 46%|████▌     | 8696/19090 [4:13:24<5:03:00,  1.75s/it]
 46%|████▌     | 8697/19090 [4:13:26<5:05:22,  1.76s/it]
 46%|████▌     | 8698/19090 [4:13:28<5:20:44,  1.85s/it]
 46%|████▌     | 8699/19090 [4:13:29<4:51:03,  1.68s/it]
 46%|████▌     | 8700/19090 [4:13:31<4:55:54,  1.71s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 21.304302215576172, 'learning_rate': 3.308277516863765e-05, 'recon_loss': 0.0013630374014610425, 'recon_static': 0.000408377697749529, 'recon_motion': 0.0009546596993459389, 'recon_norm': 1.2351005792617797, 'recon_norm_std': 0.051254646945744754, 'target_norm': 1.6179153084754945, 'recon_cos': 0.6810369461774826, 'gradnorm/compressor.query': 0.0435697962338917, 'gradnorm/compressor.layer0': 1.0321418489252314, 'gradnorm/compressor.layer1': 20.204517667583342, 'gradnorm/compressor.layer2': 5.513315035886896, 'gradnorm/compressor.layer3': 3.189493640425077, 'gradnorm/compressor.layer4': 1.3138290669669843, 'gradnorm/compressor.layer5': 0.7805880985135603, 'gradnorm/compressor.layer6': 0.3865481877125228, 'gradnorm/compressor.layer7': 0.5470457355463103, 'gradnorm/decoder.layer0': 0.8935208340261438, 'gradnorm/decoder.layer1': 0.5747206175048183, 'gradnorm/decoder.layer2': 0.3180872684578805, 'gradnorm/max_param': 14.24698543548584, 'epoch': 4.56}

 46%|████▌     | 8700/19090 [4:13:31<4:55:54,  1.71s/it]
 46%|████▌     | 8701/19090 [4:13:33<5:19:01,  1.84s/it]
 46%|████▌     | 8702/19090 [4:13:34<4:50:05,  1.68s/it]
 46%|████▌     | 8703/19090 [4:13:36<5:02:46,  1.75s/it]
 46%|████▌     | 8704/19090 [4:13:38<5:24:02,  1.87s/it]
 46%|████▌     | 8705/19090 [4:13:40<4:53:32,  1.70s/it]
 46%|████▌     | 8706/19090 [4:13:41<5:00:58,  1.74s/it]
 46%|████▌     | 8707/19090 [4:13:44<5:29:19,  1.90s/it]
 46%|████▌     | 8708/19090 [4:13:45<4:57:08,  1.72s/it]
 46%|████▌     | 8709/19090 [4:13:47<5:19:37,  1.85s/it]
 46%|████▌     | 8710/19090 [4:13:49<5:27:22,  1.89s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 28.10040855407715, 'learning_rate': 3.3039503627940004e-05, 'recon_loss': 0.0013791512319585308, 'recon_static': 0.0004145914659602568, 'recon_motion': 0.0009645597994676792, 'recon_norm': 1.239766576886177, 'recon_norm_std': 0.05099871288985014, 'target_norm': 1.619106051325798, 'recon_cos': 0.6776767283678055, 'gradnorm/compressor.query': 0.05000417392043105, 'gradnorm/compressor.layer0': 1.2076298485514365, 'gradnorm/compressor.layer1': 26.95088985297494, 'gradnorm/compressor.layer2': 6.638216762312546, 'gradnorm/compressor.layer3': 3.5858067839822634, 'gradnorm/compressor.layer4': 1.4400839594365464, 'gradnorm/compressor.layer5': 0.8495519158237174, 'gradnorm/compressor.layer6': 0.4585494070368926, 'gradnorm/compressor.layer7': 0.6712109366830248, 'gradnorm/decoder.layer0': 0.9520767263159576, 'gradnorm/decoder.layer1': 0.6400370574817328, 'gradnorm/decoder.layer2': 0.35383839914705806, 'gradnorm/max_param': 19.008075714111328, 'epoch': 4.56}

 46%|████▌     | 8710/19090 [4:13:49<5:27:22,  1.89s/it]
 46%|████▌     | 8711/19090 [4:13:50<4:56:03,  1.71s/it]
 46%|████▌     | 8712/19090 [4:13:53<5:29:48,  1.91s/it]
 46%|████▌     | 8713/19090 [4:13:54<5:10:46,  1.80s/it]
 46%|████▌     | 8714/19090 [4:13:56<4:44:03,  1.64s/it]
 46%|████▌     | 8715/19090 [4:13:58<5:42:14,  1.98s/it]
 46%|████▌     | 8716/19090 [4:14:00<5:05:52,  1.77s/it]
 46%|████▌     | 8717/19090 [4:14:01<4:40:45,  1.62s/it]
 46%|████▌     | 8718/19090 [4:14:04<5:51:04,  2.03s/it]
 46%|████▌     | 8719/19090 [4:14:05<5:12:38,  1.81s/it]
 46%|████▌     | 8720/19090 [4:14:06<4:45:40,  1.65s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 24.240318298339844, 'learning_rate': 3.2996205207074024e-05, 'recon_loss': 0.0013836902275215835, 'recon_static': 0.0004179766139714047, 'recon_motion': 0.0009657136470195837, 'recon_norm': 1.2395243376493454, 'recon_norm_std': 0.04941123332828283, 'target_norm': 1.6211216747760773, 'recon_cos': 0.6771675005555153, 'gradnorm/compressor.query': 0.04144223764691086, 'gradnorm/compressor.layer0': 1.078083313951514, 'gradnorm/compressor.layer1': 23.25840426900833, 'gradnorm/compressor.layer2': 5.499680943338043, 'gradnorm/compressor.layer3': 3.327471601793498, 'gradnorm/compressor.layer4': 1.3489169650810504, 'gradnorm/compressor.layer5': 0.8066605387670358, 'gradnorm/compressor.layer6': 0.417925288712674, 'gradnorm/compressor.layer7': 0.564014145318659, 'gradnorm/decoder.layer0': 0.7945313115152731, 'gradnorm/decoder.layer1': 0.6653472365614562, 'gradnorm/decoder.layer2': 0.3465640990217794, 'gradnorm/max_param': 16.401247024536133, 'epoch': 4.57}

 46%|████▌     | 8720/19090 [4:14:06<4:45:40,  1.65s/it]
 46%|████▌     | 8721/19090 [4:14:09<5:36:20,  1.95s/it]
 46%|████▌     | 8722/19090 [4:14:10<5:02:10,  1.75s/it]
 46%|████▌     | 8723/19090 [4:14:12<4:38:16,  1.61s/it]
 46%|████▌     | 8724/19090 [4:14:14<5:41:25,  1.98s/it]
 46%|████▌     | 8725/19090 [4:14:16<5:05:45,  1.77s/it]
 46%|████▌     | 8726/19090 [4:14:17<4:40:54,  1.63s/it]
 46%|████▌     | 8727/19090 [4:14:20<5:41:23,  1.98s/it]
 46%|████▌     | 8728/19090 [4:14:21<5:06:14,  1.77s/it]
 46%|████▌     | 8729/19090 [4:14:22<4:41:27,  1.63s/it]
 46%|████▌     | 8730/19090 [4:14:25<5:34:02,  1.93s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 26.847206115722656, 'learning_rate': 3.2952880050808444e-05, 'recon_loss': 0.0013655503746122122, 'recon_static': 0.0004149491433054209, 'recon_motion': 0.0009506012254860252, 'recon_norm': 1.2389796674251556, 'recon_norm_std': 0.0515417542308569, 'target_norm': 1.6186909317970275, 'recon_cos': 0.6810407996177673, 'gradnorm/compressor.query': 0.04840948405984585, 'gradnorm/compressor.layer0': 1.1573037166065412, 'gradnorm/compressor.layer1': 25.803368826055205, 'gradnorm/compressor.layer2': 6.175746350609013, 'gradnorm/compressor.layer3': 3.347541584557144, 'gradnorm/compressor.layer4': 1.355435373736348, 'gradnorm/compressor.layer5': 0.7710479660896897, 'gradnorm/compressor.layer6': 0.4431241289407323, 'gradnorm/compressor.layer7': 0.5380580500249109, 'gradnorm/decoder.layer0': 0.8830865080288208, 'gradnorm/decoder.layer1': 0.6640241275247211, 'gradnorm/decoder.layer2': 0.36436951588974337, 'gradnorm/max_param': 18.19713592529297, 'epoch': 4.57}

 46%|████▌     | 8730/19090 [4:14:25<5:34:02,  1.93s/it]
 46%|████▌     | 8731/19090 [4:14:26<5:05:45,  1.77s/it]
 46%|████▌     | 8732/19090 [4:14:28<4:41:09,  1.63s/it]
 46%|████▌     | 8733/19090 [4:14:31<5:50:48,  2.03s/it]
 46%|████▌     | 8734/19090 [4:14:32<5:12:16,  1.81s/it]
 46%|████▌     | 8735/19090 [4:14:33<4:45:13,  1.65s/it]
 46%|████▌     | 8736/19090 [4:14:36<5:24:47,  1.88s/it]
 46%|████▌     | 8737/19090 [4:14:37<4:54:08,  1.70s/it]
 46%|████▌     | 8738/19090 [4:14:38<4:32:18,  1.58s/it]
 46%|████▌     | 8739/19090 [4:14:41<5:31:22,  1.92s/it]
 46%|████▌     | 8740/19090 [4:14:42<4:58:35,  1.73s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 26.430953979492188, 'learning_rate': 3.290952830400139e-05, 'recon_loss': 0.0013925293780630455, 'recon_static': 0.000419225017685676, 'recon_motion': 0.0009733043625601567, 'recon_norm': 1.24098758995533, 'recon_norm_std': 0.04955318616703153, 'target_norm': 1.6214400798082351, 'recon_cos': 0.6749032124876976, 'gradnorm/compressor.query': 0.046734933402469675, 'gradnorm/compressor.layer0': 1.1419688630009668, 'gradnorm/compressor.layer1': 25.290109937375348, 'gradnorm/compressor.layer2': 6.327799470793787, 'gradnorm/compressor.layer3': 3.7093605304747537, 'gradnorm/compressor.layer4': 1.1602763617918932, 'gradnorm/compressor.layer5': 0.7794813353706181, 'gradnorm/compressor.layer6': 0.4656471234384607, 'gradnorm/compressor.layer7': 0.5713622739130325, 'gradnorm/decoder.layer0': 0.9096594178908195, 'gradnorm/decoder.layer1': 0.6765940414370998, 'gradnorm/decoder.layer2': 0.3415009007309391, 'gradnorm/max_param': 17.834579467773438, 'epoch': 4.58}

 46%|████▌     | 8740/19090 [4:14:42<4:58:35,  1.73s/it]
 46%|████▌     | 8741/19090 [4:14:44<4:35:28,  1.60s/it]
 46%|████▌     | 8742/19090 [4:14:46<5:28:14,  1.90s/it]
 46%|████▌     | 8743/19090 [4:14:48<4:56:20,  1.72s/it]
 46%|████▌     | 8744/19090 [4:14:49<4:33:50,  1.59s/it]
 46%|████▌     | 8745/19090 [4:14:51<5:26:29,  1.89s/it]
 46%|████▌     | 8746/19090 [4:14:53<5:10:14,  1.80s/it]
 46%|████▌     | 8747/19090 [4:14:54<4:43:33,  1.64s/it]
 46%|████▌     | 8748/19090 [4:14:57<5:33:06,  1.93s/it]
 46%|████▌     | 8749/19090 [4:14:58<5:12:19,  1.81s/it]
 46%|████▌     | 8750/19090 [4:15:00<4:45:17,  1.66s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 26.23096466064453, 'learning_rate': 3.286615011159992e-05, 'recon_loss': 0.0013729853031691163, 'recon_static': 0.0004172326574916951, 'recon_motion': 0.0009557526325806975, 'recon_norm': 1.2392082959413528, 'recon_norm_std': 0.05043233474716544, 'target_norm': 1.6188688278198242, 'recon_cos': 0.679270176589489, 'gradnorm/compressor.query': 0.04439733146918954, 'gradnorm/compressor.layer0': 1.1283911879698785, 'gradnorm/compressor.layer1': 25.023533845037754, 'gradnorm/compressor.layer2': 6.627698528664776, 'gradnorm/compressor.layer3': 3.428540709335219, 'gradnorm/compressor.layer4': 1.3983979459315388, 'gradnorm/compressor.layer5': 0.9358890680127279, 'gradnorm/compressor.layer6': 0.4459269617676028, 'gradnorm/compressor.layer7': 0.6330298533217318, 'gradnorm/decoder.layer0': 0.9449301017160258, 'gradnorm/decoder.layer1': 0.6967178244455217, 'gradnorm/decoder.layer2': 0.34709526191300394, 'gradnorm/max_param': 17.64693260192871, 'epoch': 4.58}

 46%|████▌     | 8750/19090 [4:15:00<4:45:17,  1.66s/it]
 46%|████▌     | 8751/19090 [4:15:02<5:26:49,  1.90s/it]
 46%|████▌     | 8752/19090 [4:15:04<5:14:38,  1.83s/it]
 46%|████▌     | 8753/19090 [4:15:05<4:47:01,  1.67s/it]
 46%|████▌     | 8754/19090 [4:15:07<5:19:52,  1.86s/it]
 46%|████▌     | 8755/19090 [4:15:09<5:11:13,  1.81s/it]
 46%|████▌     | 8756/19090 [4:15:10<4:44:30,  1.65s/it]
 46%|████▌     | 8757/19090 [4:15:13<5:38:27,  1.97s/it]
 46%|████▌     | 8758/19090 [4:15:14<5:03:35,  1.76s/it]
 46%|████▌     | 8759/19090 [4:15:16<4:39:07,  1.62s/it]
 46%|████▌     | 8760/19090 [4:15:18<5:35:47,  1.95s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 27.864521026611328, 'learning_rate': 3.282274561863948e-05, 'recon_loss': 0.001359713595593348, 'recon_static': 0.0004034631245303899, 'recon_motion': 0.0009562505365465768, 'recon_norm': 1.2333553940057755, 'recon_norm_std': 0.051018132641911504, 'target_norm': 1.6184576213359834, 'recon_cos': 0.6826653942465782, 'gradnorm/compressor.query': 0.04296780715254746, 'gradnorm/compressor.layer0': 1.0750706849035847, 'gradnorm/compressor.layer1': 26.773063628116155, 'gradnorm/compressor.layer2': 6.356222585133234, 'gradnorm/compressor.layer3': 3.632095758427725, 'gradnorm/compressor.layer4': 1.3623736401680726, 'gradnorm/compressor.layer5': 0.8855956438969697, 'gradnorm/compressor.layer6': 0.4296159337873806, 'gradnorm/compressor.layer7': 0.6574457718421006, 'gradnorm/decoder.layer0': 1.0057103315130904, 'gradnorm/decoder.layer1': 0.7053876921089691, 'gradnorm/decoder.layer2': 0.3422555951510299, 'gradnorm/max_param': 18.882413864135742, 'epoch': 4.59}

 46%|████▌     | 8760/19090 [4:15:18<5:35:47,  1.95s/it]
 46%|████▌     | 8761/19090 [4:15:20<5:05:47,  1.78s/it]
 46%|████▌     | 8762/19090 [4:15:21<4:40:37,  1.63s/it]
 46%|████▌     | 8763/19090 [4:15:24<5:23:12,  1.88s/it]
 46%|████▌     | 8764/19090 [4:15:25<5:01:05,  1.75s/it]
 46%|████▌     | 8765/19090 [4:15:26<4:37:04,  1.61s/it]
 46%|████▌     | 8766/19090 [4:15:29<5:44:03,  2.00s/it]
 46%|████▌     | 8767/19090 [4:15:30<5:07:23,  1.79s/it]
 46%|████▌     | 8768/19090 [4:15:32<4:43:01,  1.65s/it]
 46%|████▌     | 8769/19090 [4:15:35<5:50:46,  2.04s/it]
 46%|████▌     | 8770/19090 [4:15:36<5:11:58,  1.81s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 26.454389572143555, 'learning_rate': 3.277931497024348e-05, 'recon_loss': 0.0013650705135660247, 'recon_static': 0.00040735221409704536, 'recon_motion': 0.0009577183023793623, 'recon_norm': 1.2322749674320221, 'recon_norm_std': 0.05057857017964125, 'target_norm': 1.6196665972471238, 'recon_cos': 0.681162329018116, 'gradnorm/compressor.query': 0.04157570070892268, 'gradnorm/compressor.layer0': 1.034936269742238, 'gradnorm/compressor.layer1': 25.30901789320854, 'gradnorm/compressor.layer2': 6.325710259767463, 'gradnorm/compressor.layer3': 3.660830766373176, 'gradnorm/compressor.layer4': 1.3495705097967596, 'gradnorm/compressor.layer5': 0.8845980319183867, 'gradnorm/compressor.layer6': 0.48633520282053117, 'gradnorm/compressor.layer7': 0.6472511469454013, 'gradnorm/decoder.layer0': 0.9619274777479025, 'gradnorm/decoder.layer1': 0.6897412311415497, 'gradnorm/decoder.layer2': 0.3724860924486307, 'gradnorm/max_param': 17.84882926940918, 'epoch': 4.59}

 46%|████▌     | 8770/19090 [4:15:36<5:11:58,  1.81s/it]
 46%|████▌     | 8771/19090 [4:15:37<4:45:02,  1.66s/it]
 46%|████▌     | 8772/19090 [4:15:40<5:48:07,  2.02s/it]
 46%|████▌     | 8773/19090 [4:15:41<5:10:09,  1.80s/it]
 46%|████▌     | 8774/19090 [4:15:43<4:43:38,  1.65s/it]
 46%|████▌     | 8775/19090 [4:15:45<5:31:06,  1.93s/it]
 46%|████▌     | 8776/19090 [4:15:47<4:58:07,  1.73s/it]
 46%|████▌     | 8777/19090 [4:15:48<4:36:24,  1.61s/it]
 46%|████▌     | 8778/19090 [4:15:51<5:29:38,  1.92s/it]
 46%|████▌     | 8779/19090 [4:15:52<4:56:58,  1.73s/it]
 46%|████▌     | 8780/19090 [4:15:54<4:52:49,  1.70s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 28.448699951171875, 'learning_rate': 3.273585831162276e-05, 'recon_loss': 0.0013705747755011544, 'recon_static': 0.0004045696827233769, 'recon_motion': 0.0009660050505772233, 'recon_norm': 1.2332149624824524, 'recon_norm_std': 0.05073493598029018, 'target_norm': 1.621093174815178, 'recon_cos': 0.6807712361216545, 'gradnorm/compressor.query': 0.048778043335625014, 'gradnorm/compressor.layer0': 1.056243144808763, 'gradnorm/compressor.layer1': 27.284338357909768, 'gradnorm/compressor.layer2': 6.7657191598167685, 'gradnorm/compressor.layer3': 3.580983848068767, 'gradnorm/compressor.layer4': 1.439580489881314, 'gradnorm/compressor.layer5': 0.9038457878286866, 'gradnorm/compressor.layer6': 0.5018586838042661, 'gradnorm/compressor.layer7': 0.6564711924578437, 'gradnorm/decoder.layer0': 1.0118473516201205, 'gradnorm/decoder.layer1': 0.6637371612643369, 'gradnorm/decoder.layer2': 0.374837224123567, 'gradnorm/max_param': 19.243257522583008, 'epoch': 4.6}

 46%|████▌     | 8780/19090 [4:15:54<4:52:49,  1.70s/it]
 46%|████▌     | 8781/19090 [4:15:56<5:40:19,  1.98s/it]
 46%|████▌     | 8782/19090 [4:15:57<5:04:37,  1.77s/it]
 46%|████▌     | 8783/19090 [4:15:59<4:47:16,  1.67s/it]
 46%|████▌     | 8784/19090 [4:16:01<5:36:50,  1.96s/it]
 46%|████▌     | 8785/19090 [4:16:03<5:02:31,  1.76s/it]
 46%|████▌     | 8786/19090 [4:16:04<4:38:08,  1.62s/it]
 46%|████▌     | 8787/19090 [4:16:07<5:38:22,  1.97s/it]
 46%|████▌     | 8788/19090 [4:16:08<5:03:20,  1.77s/it]
 46%|████▌     | 8789/19090 [4:16:09<4:38:45,  1.62s/it]
 46%|████▌     | 8790/19090 [4:16:12<5:35:18,  1.95s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 28.237794876098633, 'learning_rate': 3.2692375788075135e-05, 'recon_loss': 0.0013826687179971486, 'recon_static': 0.0004253559178323485, 'recon_motion': 0.0009573127870680764, 'recon_norm': 1.2408052265644074, 'recon_norm_std': 0.050390707422047855, 'target_norm': 1.620079469680786, 'recon_cos': 0.6768207177519798, 'gradnorm/compressor.query': 0.054657523420373166, 'gradnorm/compressor.layer0': 1.236593209001754, 'gradnorm/compressor.layer1': 26.949403049177754, 'gradnorm/compressor.layer2': 7.121691554721107, 'gradnorm/compressor.layer3': 3.666261838470501, 'gradnorm/compressor.layer4': 1.603698348345652, 'gradnorm/compressor.layer5': 0.9314968230069951, 'gradnorm/compressor.layer6': 0.4864738276734849, 'gradnorm/compressor.layer7': 0.5847817020588807, 'gradnorm/decoder.layer0': 0.9215710995733081, 'gradnorm/decoder.layer1': 0.6500332282816429, 'gradnorm/decoder.layer2': 0.3472812537505068, 'gradnorm/max_param': 19.00449562072754, 'epoch': 4.6}

 46%|████▌     | 8790/19090 [4:16:12<5:35:18,  1.95s/it]
 46%|████▌     | 8791/19090 [4:16:14<5:03:13,  1.77s/it]
 46%|████▌     | 8792/19090 [4:16:15<4:39:20,  1.63s/it]
 46%|████▌     | 8793/19090 [4:16:18<6:00:58,  2.10s/it]
 46%|████▌     | 8794/19090 [4:16:19<5:19:35,  1.86s/it]
 46%|████▌     | 8795/19090 [4:16:21<4:50:17,  1.69s/it]
 46%|████▌     | 8796/19090 [4:16:23<5:27:49,  1.91s/it]
 46%|████▌     | 8797/19090 [4:16:24<4:56:10,  1.73s/it]
 46%|████▌     | 8798/19090 [4:16:26<4:33:53,  1.60s/it]
 46%|████▌     | 8799/19090 [4:16:28<5:31:35,  1.93s/it]
 46%|████▌     | 8800/19090 [4:16:30<4:58:36,  1.74s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 28.463729858398438, 'learning_rate': 3.2648867544984896e-05, 'recon_loss': 0.0013598895398899912, 'recon_static': 0.00040257127548102287, 'recon_motion': 0.0009573182498570532, 'recon_norm': 1.2308652222156524, 'recon_norm_std': 0.05044431630522013, 'target_norm': 1.6200485110282898, 'recon_cos': 0.6828964591026306, 'gradnorm/compressor.query': 0.04731582200675294, 'gradnorm/compressor.layer0': 1.1256909897442902, 'gradnorm/compressor.layer1': 27.249089497045013, 'gradnorm/compressor.layer2': 6.924140783226284, 'gradnorm/compressor.layer3': 3.638816023882502, 'gradnorm/compressor.layer4': 1.456657955018508, 'gradnorm/compressor.layer5': 0.9023295149567947, 'gradnorm/compressor.layer6': 0.4822159609550493, 'gradnorm/compressor.layer7': 0.6203488855716733, 'gradnorm/decoder.layer0': 0.9845263652380656, 'gradnorm/decoder.layer1': 0.7328836841618529, 'gradnorm/decoder.layer2': 0.39625619796213724, 'gradnorm/max_param': 19.21802520751953, 'epoch': 4.61}

 46%|████▌     | 8800/19090 [4:16:30<4:58:36,  1.74s/it]
 46%|████▌     | 8801/19090 [4:16:31<4:35:11,  1.60s/it]
 46%|████▌     | 8802/19090 [4:16:34<5:38:28,  1.97s/it]
 46%|████▌     | 8803/19090 [4:16:35<5:10:03,  1.81s/it]
 46%|████▌     | 8804/19090 [4:16:36<4:43:02,  1.65s/it]
 46%|████▌     | 8805/19090 [4:16:39<5:23:05,  1.88s/it]
 46%|████▌     | 8806/19090 [4:16:40<5:04:08,  1.77s/it]
 46%|████▌     | 8807/19090 [4:16:42<4:39:07,  1.63s/it]
 46%|████▌     | 8808/19090 [4:16:44<5:21:23,  1.88s/it]
 46%|████▌     | 8809/19090 [4:16:46<5:09:14,  1.80s/it]
 46%|████▌     | 8810/19090 [4:16:47<4:42:27,  1.65s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 26.897491455078125, 'learning_rate': 3.260533372782234e-05, 'recon_loss': 0.0013509057083865629, 'recon_static': 0.000398459671123419, 'recon_motion': 0.0009524460372631438, 'recon_norm': 1.2284861385822297, 'recon_norm_std': 0.05020888801664114, 'target_norm': 1.6193484097719193, 'recon_cos': 0.685187204182148, 'gradnorm/compressor.query': 0.05313794367547117, 'gradnorm/compressor.layer0': 1.1348239798322608, 'gradnorm/compressor.layer1': 25.845684159031997, 'gradnorm/compressor.layer2': 6.1625665316752745, 'gradnorm/compressor.layer3': 3.439511490889291, 'gradnorm/compressor.layer4': 1.3900368121644704, 'gradnorm/compressor.layer5': 0.8410225335956737, 'gradnorm/compressor.layer6': 0.4219506337219951, 'gradnorm/compressor.layer7': 0.5981526775194632, 'gradnorm/decoder.layer0': 0.8207517265443423, 'gradnorm/decoder.layer1': 0.6529590879639652, 'gradnorm/decoder.layer2': 0.31968151140989703, 'gradnorm/max_param': 18.22543716430664, 'epoch': 4.61}

 46%|████▌     | 8810/19090 [4:16:47<4:42:27,  1.65s/it]
 46%|████▌     | 8811/19090 [4:16:49<5:17:15,  1.85s/it]
 46%|████▌     | 8812/19090 [4:16:52<5:39:56,  1.98s/it]
 46%|████▌     | 8813/19090 [4:16:53<5:04:05,  1.78s/it]
 46%|████▌     | 8814/19090 [4:16:54<4:45:11,  1.67s/it]
 46%|████▌     | 8815/19090 [4:16:57<5:42:19,  2.00s/it]
 46%|████▌     | 8816/19090 [4:16:58<5:05:42,  1.79s/it]
 46%|████▌     | 8817/19090 [4:17:00<4:51:40,  1.70s/it]
 46%|████▌     | 8818/19090 [4:17:02<5:25:02,  1.90s/it]
 46%|████▌     | 8819/19090 [4:17:04<4:53:23,  1.71s/it]
 46%|████▌     | 8820/19090 [4:17:05<4:48:51,  1.69s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 27.33159065246582, 'learning_rate': 3.2561774482143235e-05, 'recon_loss': 0.0013625186344143002, 'recon_static': 0.0004140251927310601, 'recon_motion': 0.0009484934314968995, 'recon_norm': 1.2373644798994063, 'recon_norm_std': 0.05077027939260006, 'target_norm': 1.6178200215101242, 'recon_cos': 0.6816989794373512, 'gradnorm/compressor.query': 0.04643904846258321, 'gradnorm/compressor.layer0': 1.1634563429559428, 'gradnorm/compressor.layer1': 26.141623141060286, 'gradnorm/compressor.layer2': 6.536894335143266, 'gradnorm/compressor.layer3': 3.8490768366196146, 'gradnorm/compressor.layer4': 1.4053824397711863, 'gradnorm/compressor.layer5': 0.8177208059214145, 'gradnorm/compressor.layer6': 0.47524650414017394, 'gradnorm/compressor.layer7': 0.676817656013815, 'gradnorm/decoder.layer0': 0.9210168756270535, 'gradnorm/decoder.layer1': 0.660685220101161, 'gradnorm/decoder.layer2': 0.3415138598794958, 'gradnorm/max_param': 18.436771392822266, 'epoch': 4.62}

 46%|████▌     | 8820/19090 [4:17:05<4:48:51,  1.69s/it]
 46%|████▌     | 8821/19090 [4:17:07<5:18:10,  1.86s/it]
 46%|████▌     | 8822/19090 [4:17:09<4:48:50,  1.69s/it]
 46%|████▌     | 8823/19090 [4:17:11<4:51:27,  1.70s/it]
 46%|████▌     | 8824/19090 [4:17:13<5:26:03,  1.91s/it]
 46%|████▌     | 8825/19090 [4:17:14<4:54:17,  1.72s/it]
 46%|████▌     | 8826/19090 [4:17:16<4:34:29,  1.60s/it]
 46%|████▌     | 8827/19090 [4:17:18<5:18:52,  1.86s/it]
 46%|████▌     | 8828/19090 [4:17:19<4:48:57,  1.69s/it]
 46%|████▌     | 8829/19090 [4:17:21<4:50:17,  1.70s/it]
 46%|████▋     | 8830/19090 [4:17:23<5:30:31,  1.93s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 25.892391204833984, 'learning_rate': 3.251818995358843e-05, 'recon_loss': 0.001382324195583351, 'recon_static': 0.00042247656747349536, 'recon_motion': 0.0009598476550308988, 'recon_norm': 1.2390457838773727, 'recon_norm_std': 0.050789172668009995, 'target_norm': 1.6201694518327714, 'recon_cos': 0.6770173698663712, 'gradnorm/compressor.query': 0.053853980237157296, 'gradnorm/compressor.layer0': 1.194872058180384, 'gradnorm/compressor.layer1': 24.776010736351655, 'gradnorm/compressor.layer2': 6.153953782387524, 'gradnorm/compressor.layer3': 3.4953994487781443, 'gradnorm/compressor.layer4': 1.4730326932812672, 'gradnorm/compressor.layer5': 0.9119102375001771, 'gradnorm/compressor.layer6': 0.48991198026519406, 'gradnorm/compressor.layer7': 0.6373895259577729, 'gradnorm/decoder.layer0': 0.9500464139441945, 'gradnorm/decoder.layer1': 0.6173427724480522, 'gradnorm/decoder.layer2': 0.33800899218414354, 'gradnorm/max_param': 17.47226333618164, 'epoch': 4.63}

 46%|████▋     | 8830/19090 [4:17:23<5:30:31,  1.93s/it]
 46%|████▋     | 8831/19090 [4:17:25<4:57:17,  1.74s/it]
 46%|████▋     | 8832/19090 [4:17:26<4:34:29,  1.61s/it]
 46%|████▋     | 8833/19090 [4:17:29<5:39:30,  1.99s/it]
 46%|████▋     | 8834/19090 [4:17:30<5:03:43,  1.78s/it]
 46%|████▋     | 8835/19090 [4:17:32<4:38:48,  1.63s/it]
 46%|████▋     | 8836/19090 [4:17:34<5:41:37,  2.00s/it]
 46%|████▋     | 8837/19090 [4:17:36<5:05:13,  1.79s/it]
 46%|████▋     | 8838/19090 [4:17:37<4:39:38,  1.64s/it]
 46%|████▋     | 8839/19090 [4:17:40<5:36:16,  1.97s/it]
 46%|████▋     | 8840/19090 [4:17:41<5:01:14,  1.76s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 28.488983154296875, 'learning_rate': 3.247458028788326e-05, 'recon_loss': 0.0013989105122163893, 'recon_static': 0.0004315957921789959, 'recon_motion': 0.0009673147447756492, 'recon_norm': 1.2471966981887816, 'recon_norm_std': 0.04961117682978511, 'target_norm': 1.6208373308181763, 'recon_cos': 0.673582473397255, 'gradnorm/compressor.query': 0.05582336377065771, 'gradnorm/compressor.layer0': 1.3509688644774445, 'gradnorm/compressor.layer1': 27.25729524444729, 'gradnorm/compressor.layer2': 6.906604878555478, 'gradnorm/compressor.layer3': 3.6930634507637676, 'gradnorm/compressor.layer4': 1.432412390967504, 'gradnorm/compressor.layer5': 1.0089082896705441, 'gradnorm/compressor.layer6': 0.5104419416540539, 'gradnorm/compressor.layer7': 0.668548198805538, 'gradnorm/decoder.layer0': 1.020644124979587, 'gradnorm/decoder.layer1': 0.713549500211239, 'gradnorm/decoder.layer2': 0.4071141369090386, 'gradnorm/max_param': 19.221643447875977, 'epoch': 4.63}

 46%|████▋     | 8840/19090 [4:17:41<5:01:14,  1.76s/it]
 46%|████▋     | 8841/19090 [4:17:42<4:37:13,  1.62s/it]
 46%|████▋     | 8842/19090 [4:17:45<5:28:20,  1.92s/it]
 46%|████▋     | 8843/19090 [4:17:46<4:55:47,  1.73s/it]
 46%|████▋     | 8844/19090 [4:17:47<4:32:57,  1.60s/it]
 46%|████▋     | 8845/19090 [4:17:50<5:37:10,  1.97s/it]
 46%|████▋     | 8846/19090 [4:17:52<5:02:08,  1.77s/it]
 46%|████▋     | 8847/19090 [4:17:53<4:37:25,  1.63s/it]
 46%|████▋     | 8848/19090 [4:17:55<5:15:23,  1.85s/it]
 46%|████▋     | 8849/19090 [4:17:57<4:46:39,  1.68s/it]
 46%|████▋     | 8850/19090 [4:17:58<4:44:12,  1.67s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 28.129610061645508, 'learning_rate': 3.2430945630837096e-05, 'recon_loss': 0.001395393832353875, 'recon_static': 0.0004369622329249978, 'recon_motion': 0.0009584315994288772, 'recon_norm': 1.2450137168169022, 'recon_norm_std': 0.0508099028840661, 'target_norm': 1.6227748930454253, 'recon_cos': 0.6749841749668122, 'gradnorm/compressor.query': 0.05559345627785404, 'gradnorm/compressor.layer0': 1.2966269290313597, 'gradnorm/compressor.layer1': 27.053618138698774, 'gradnorm/compressor.layer2': 6.142401440722055, 'gradnorm/compressor.layer3': 3.8785987878129045, 'gradnorm/compressor.layer4': 1.3227354761630186, 'gradnorm/compressor.layer5': 0.9619527896114357, 'gradnorm/compressor.layer6': 0.49230436421164064, 'gradnorm/compressor.layer7': 0.6750567527154091, 'gradnorm/decoder.layer0': 0.9518139686981438, 'gradnorm/decoder.layer1': 0.7233246470148073, 'gradnorm/decoder.layer2': 0.34294791277800746, 'gradnorm/max_param': 19.078344345092773, 'epoch': 4.64}

 46%|████▋     | 8850/19090 [4:17:58<4:44:12,  1.67s/it]
 46%|████▋     | 8851/19090 [4:18:01<5:35:28,  1.97s/it]
 46%|████▋     | 8852/19090 [4:18:02<5:00:38,  1.76s/it]
 46%|████▋     | 8853/19090 [4:18:04<4:46:34,  1.68s/it]
 46%|████▋     | 8854/19090 [4:18:06<5:29:42,  1.93s/it]
 46%|████▋     | 8855/19090 [4:18:07<4:56:37,  1.74s/it]
 46%|████▋     | 8856/19090 [4:18:09<4:34:04,  1.61s/it]
 46%|████▋     | 8857/19090 [4:18:11<5:23:51,  1.90s/it]
 46%|████▋     | 8858/19090 [4:18:13<4:52:35,  1.72s/it]
 46%|████▋     | 8859/19090 [4:18:14<4:41:01,  1.65s/it]
 46%|████▋     | 8860/19090 [4:18:17<5:26:37,  1.92s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 27.49380874633789, 'learning_rate': 3.238728612834293e-05, 'recon_loss': 0.0013882391300285234, 'recon_static': 0.0004210483333736192, 'recon_motion': 0.0009671907944721169, 'recon_norm': 1.2380121171474456, 'recon_norm_std': 0.05013644015416503, 'target_norm': 1.6213708609342574, 'recon_cos': 0.6761598095297814, 'gradnorm/compressor.query': 0.047275119509645504, 'gradnorm/compressor.layer0': 1.21364148865898, 'gradnorm/compressor.layer1': 26.25912927187776, 'gradnorm/compressor.layer2': 6.747292434761911, 'gradnorm/compressor.layer3': 3.786059512928582, 'gradnorm/compressor.layer4': 1.411356534019987, 'gradnorm/compressor.layer5': 0.8896572141915798, 'gradnorm/compressor.layer6': 0.49532196505194437, 'gradnorm/compressor.layer7': 0.6158579496962147, 'gradnorm/decoder.layer0': 0.984194935875267, 'gradnorm/decoder.layer1': 0.7213764302000425, 'gradnorm/decoder.layer2': 0.36760685693900974, 'gradnorm/max_param': 18.517696380615234, 'epoch': 4.64}

 46%|████▋     | 8860/19090 [4:18:17<5:26:37,  1.92s/it]
 46%|████▋     | 8861/19090 [4:18:18<4:54:30,  1.73s/it]
 46%|████▋     | 8862/19090 [4:18:19<4:37:11,  1.63s/it]
 46%|████▋     | 8863/19090 [4:18:22<5:24:31,  1.90s/it]
 46%|████▋     | 8864/19090 [4:18:23<4:52:32,  1.72s/it]
 46%|████▋     | 8865/19090 [4:18:25<4:58:57,  1.75s/it]
 46%|████▋     | 8866/19090 [4:18:27<5:22:17,  1.89s/it]
 46%|████▋     | 8867/19090 [4:18:28<4:51:36,  1.71s/it]
 46%|████▋     | 8868/19090 [4:18:30<4:42:37,  1.66s/it]
 46%|████▋     | 8869/19090 [4:18:32<5:22:21,  1.89s/it]
 46%|████▋     | 8870/19090 [4:18:34<4:51:23,  1.71s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 31.228282928466797, 'learning_rate': 3.234360192637676e-05, 'recon_loss': 0.001372602442279458, 'recon_static': 0.00041515728153171947, 'recon_motion': 0.0009574451134540141, 'recon_norm': 1.2368722170591355, 'recon_norm_std': 0.04974977541714907, 'target_norm': 1.6210298806428909, 'recon_cos': 0.6802111312747001, 'gradnorm/compressor.query': 0.05473667726567012, 'gradnorm/compressor.layer0': 1.2614947140794974, 'gradnorm/compressor.layer1': 29.92519461840171, 'gradnorm/compressor.layer2': 7.397741911433759, 'gradnorm/compressor.layer3': 4.225869282337807, 'gradnorm/compressor.layer4': 1.6035113131317635, 'gradnorm/compressor.layer5': 0.8470134143397628, 'gradnorm/compressor.layer6': 0.500964575432629, 'gradnorm/compressor.layer7': 0.6269699759907476, 'gradnorm/decoder.layer0': 0.9932208349754947, 'gradnorm/decoder.layer1': 0.6821616730844696, 'gradnorm/decoder.layer2': 0.3532891259207599, 'gradnorm/max_param': 21.105709075927734, 'epoch': 4.65}

 46%|████▋     | 8870/19090 [4:18:34<4:51:23,  1.71s/it]
 46%|████▋     | 8871/19090 [4:18:35<4:48:49,  1.70s/it]
 46%|████▋     | 8872/19090 [4:18:38<5:35:00,  1.97s/it]
 46%|████▋     | 8873/19090 [4:18:39<5:00:14,  1.76s/it]
 46%|████▋     | 8874/19090 [4:18:41<4:40:43,  1.65s/it]
 46%|████▋     | 8875/19090 [4:18:43<5:38:16,  1.99s/it]
 46%|████▋     | 8876/19090 [4:18:45<5:02:15,  1.78s/it]
 47%|████▋     | 8877/19090 [4:18:46<4:36:49,  1.63s/it]
 47%|████▋     | 8878/19090 [4:18:49<5:22:19,  1.89s/it]
 47%|████▋     | 8879/19090 [4:18:50<4:51:07,  1.71s/it]
 47%|████▋     | 8880/19090 [4:18:51<4:32:30,  1.60s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 31.121883392333984, 'learning_rate': 3.2299893170997195e-05, 'recon_loss': 0.0013702976226340978, 'recon_static': 0.0003999822787591256, 'recon_motion': 0.0009703153526061214, 'recon_norm': 1.2301323026418687, 'recon_norm_std': 0.04998167539015412, 'target_norm': 1.6213676691055299, 'recon_cos': 0.6806557089090347, 'gradnorm/compressor.query': 0.05252066329070814, 'gradnorm/compressor.layer0': 1.3115629302417922, 'gradnorm/compressor.layer1': 29.839718723147552, 'gradnorm/compressor.layer2': 7.320652517200082, 'gradnorm/compressor.layer3': 4.061794982770015, 'gradnorm/compressor.layer4': 1.7047472585803465, 'gradnorm/compressor.layer5': 1.0511958782584092, 'gradnorm/compressor.layer6': 0.5154690073040993, 'gradnorm/compressor.layer7': 0.6898851850946304, 'gradnorm/decoder.layer0': 1.013578377529646, 'gradnorm/decoder.layer1': 0.6606486814471452, 'gradnorm/decoder.layer2': 0.3647879253539086, 'gradnorm/max_param': 21.042692184448242, 'epoch': 4.65}

 47%|████▋     | 8880/19090 [4:18:51<4:32:30,  1.60s/it]
 47%|████▋     | 8881/19090 [4:18:54<5:35:48,  1.97s/it]
 47%|████▋     | 8882/19090 [4:18:55<5:00:25,  1.77s/it]
 47%|████▋     | 8883/19090 [4:18:57<4:39:03,  1.64s/it]
 47%|████▋     | 8884/19090 [4:18:59<5:23:51,  1.90s/it]
 47%|████▋     | 8885/19090 [4:19:00<4:52:54,  1.72s/it]
 47%|████▋     | 8886/19090 [4:19:02<4:56:05,  1.74s/it]
 47%|████▋     | 8887/19090 [4:19:05<5:34:20,  1.97s/it]
 47%|████▋     | 8888/19090 [4:19:06<4:59:51,  1.76s/it]
 47%|████▋     | 8889/19090 [4:19:07<4:41:23,  1.66s/it]
 47%|████▋     | 8890/19090 [4:19:10<5:36:08,  1.98s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 29.827463150024414, 'learning_rate': 3.225616000834496e-05, 'recon_loss': 0.0013974892353871837, 'recon_static': 0.00043302828999003395, 'recon_motion': 0.0009644609570386819, 'recon_norm': 1.2457276582717896, 'recon_norm_std': 0.05032969880849123, 'target_norm': 1.6209367156028747, 'recon_cos': 0.6738946735858917, 'gradnorm/compressor.query': 0.0483968211104989, 'gradnorm/compressor.layer0': 1.2554448275610297, 'gradnorm/compressor.layer1': 28.48717330604688, 'gradnorm/compressor.layer2': 7.490440125107621, 'gradnorm/compressor.layer3': 3.869903535453857, 'gradnorm/compressor.layer4': 1.6132284899344185, 'gradnorm/compressor.layer5': 0.8756049653401751, 'gradnorm/compressor.layer6': 0.49944256465111625, 'gradnorm/compressor.layer7': 0.6665388496377392, 'gradnorm/decoder.layer0': 0.9106066003012704, 'gradnorm/decoder.layer1': 0.6989864410987233, 'gradnorm/decoder.layer2': 0.34124676713355334, 'gradnorm/max_param': 20.089744567871094, 'epoch': 4.66}

 47%|████▋     | 8890/19090 [4:19:10<5:36:08,  1.98s/it]
 47%|████▋     | 8891/19090 [4:19:11<5:01:11,  1.77s/it]
 47%|████▋     | 8892/19090 [4:19:13<4:43:14,  1.67s/it]
 47%|████▋     | 8893/19090 [4:19:16<5:37:01,  1.98s/it]
 47%|████▋     | 8894/19090 [4:19:17<5:01:32,  1.77s/it]
 47%|████▋     | 8895/19090 [4:19:18<4:42:02,  1.66s/it]
 47%|████▋     | 8896/19090 [4:19:21<5:29:42,  1.94s/it]
 47%|████▋     | 8897/19090 [4:19:22<4:56:30,  1.75s/it]
 47%|████▋     | 8898/19090 [4:19:23<4:33:25,  1.61s/it]
 47%|████▋     | 8899/19090 [4:19:26<5:39:42,  2.00s/it]
 47%|████▋     | 8900/19090 [4:19:28<5:03:17,  1.79s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 30.352811813354492, 'learning_rate': 3.2212402584642336e-05, 'recon_loss': 0.0013747565797530114, 'recon_static': 0.0004153023786784615, 'recon_motion': 0.0009594542003469542, 'recon_norm': 1.2368710577487945, 'recon_norm_std': 0.049796685203909875, 'target_norm': 1.6183533936738967, 'recon_cos': 0.6784187734127045, 'gradnorm/compressor.query': 0.04814762614286272, 'gradnorm/compressor.layer0': 1.2097894456222342, 'gradnorm/compressor.layer1': 29.045525957107404, 'gradnorm/compressor.layer2': 7.397508572969564, 'gradnorm/compressor.layer3': 4.050949343336126, 'gradnorm/compressor.layer4': 1.425158890013983, 'gradnorm/compressor.layer5': 0.9036460825162136, 'gradnorm/compressor.layer6': 0.4723671746384328, 'gradnorm/compressor.layer7': 0.6335455569225502, 'gradnorm/decoder.layer0': 0.9703516486927131, 'gradnorm/decoder.layer1': 0.7101771302295765, 'gradnorm/decoder.layer2': 0.36490800116170186, 'gradnorm/max_param': 20.484121322631836, 'epoch': 4.66}

 47%|████▋     | 8900/19090 [4:19:28<5:03:17,  1.79s/it]
 47%|████▋     | 8901/19090 [4:19:29<4:38:07,  1.64s/it]
 47%|████▋     | 8902/19090 [4:19:32<5:31:18,  1.95s/it]
 47%|████▋     | 8903/19090 [4:19:33<4:57:33,  1.75s/it]
 47%|████▋     | 8904/19090 [4:19:34<4:34:13,  1.62s/it]
 47%|████▋     | 8905/19090 [4:19:37<5:17:25,  1.87s/it]
 47%|████▋     | 8906/19090 [4:19:38<4:48:28,  1.70s/it]
 47%|████▋     | 8907/19090 [4:19:39<4:27:12,  1.57s/it]
 47%|████▋     | 8908/19090 [4:19:42<5:20:09,  1.89s/it]
 47%|████▋     | 8909/19090 [4:19:43<4:49:41,  1.71s/it]
 47%|████▋     | 8910/19090 [4:19:45<4:43:29,  1.67s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 26.426197052001953, 'learning_rate': 3.216862104619275e-05, 'recon_loss': 0.0013725887954933568, 'recon_static': 0.00040555557352490724, 'recon_motion': 0.0009670332248788327, 'recon_norm': 1.233344739675522, 'recon_norm_std': 0.04859342454001307, 'target_norm': 1.6197086930274964, 'recon_cos': 0.6795729160308838, 'gradnorm/compressor.query': 0.04290484287206115, 'gradnorm/compressor.layer0': 1.0518918319506871, 'gradnorm/compressor.layer1': 25.212710804925607, 'gradnorm/compressor.layer2': 6.624557480338044, 'gradnorm/compressor.layer3': 3.5679152973966426, 'gradnorm/compressor.layer4': 1.3484658725622338, 'gradnorm/compressor.layer5': 0.9408962403635749, 'gradnorm/compressor.layer6': 0.4991957932329364, 'gradnorm/compressor.layer7': 0.6253253328454474, 'gradnorm/decoder.layer0': 0.9919584114947526, 'gradnorm/decoder.layer1': 0.6919799407383475, 'gradnorm/decoder.layer2': 0.3652482259643313, 'gradnorm/max_param': 17.78077507019043, 'epoch': 4.67}

 47%|████▋     | 8910/19090 [4:19:45<4:43:29,  1.67s/it]
 47%|████▋     | 8911/19090 [4:19:47<5:27:40,  1.93s/it]
 47%|████▋     | 8912/19090 [4:19:49<4:55:03,  1.74s/it]
 47%|████▋     | 8913/19090 [4:19:50<4:32:09,  1.60s/it]
 47%|████▋     | 8914/19090 [4:19:53<5:38:34,  2.00s/it]
 47%|████▋     | 8915/19090 [4:19:54<5:02:33,  1.78s/it]
 47%|████▋     | 8916/19090 [4:19:55<4:37:16,  1.64s/it]
 47%|████▋     | 8917/19090 [4:19:58<5:53:37,  2.09s/it]
 47%|████▋     | 8918/19090 [4:20:00<5:12:45,  1.84s/it]
 47%|████▋     | 8919/19090 [4:20:01<4:44:18,  1.68s/it]
 47%|████▋     | 8920/19090 [4:20:04<5:41:24,  2.01s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 30.591150283813477, 'learning_rate': 3.2124815539380276e-05, 'recon_loss': 0.001381680477061309, 'recon_static': 0.00042013019701698796, 'recon_motion': 0.0009615502669475973, 'recon_norm': 1.2387088596820832, 'recon_norm_std': 0.050714703649282454, 'target_norm': 1.6213579058647156, 'recon_cos': 0.6777406215667725, 'gradnorm/compressor.query': 0.04288439435729093, 'gradnorm/compressor.layer0': 1.1178008843197809, 'gradnorm/compressor.layer1': 29.347847324431225, 'gradnorm/compressor.layer2': 7.208690156722979, 'gradnorm/compressor.layer3': 3.947577281579547, 'gradnorm/compressor.layer4': 1.4804743059602623, 'gradnorm/compressor.layer5': 0.9470599393616773, 'gradnorm/compressor.layer6': 0.5309225363416147, 'gradnorm/compressor.layer7': 0.7041295610805304, 'gradnorm/decoder.layer0': 1.0830163981146386, 'gradnorm/decoder.layer1': 0.733850657172889, 'gradnorm/decoder.layer2': 0.3803666473246311, 'gradnorm/max_param': 20.69998550415039, 'epoch': 4.67}

 47%|████▋     | 8920/19090 [4:20:04<5:41:24,  2.01s/it]
 47%|████▋     | 8921/19090 [4:20:05<5:04:26,  1.80s/it]
 47%|████▋     | 8922/19090 [4:20:06<4:38:22,  1.64s/it]
 47%|████▋     | 8923/19090 [4:20:09<5:39:00,  2.00s/it]
 47%|████▋     | 8924/19090 [4:20:11<5:02:39,  1.79s/it]
 47%|████▋     | 8925/19090 [4:20:12<4:38:17,  1.64s/it]
 47%|████▋     | 8926/19090 [4:20:14<5:21:07,  1.90s/it]
 47%|████▋     | 8927/19090 [4:20:16<4:50:12,  1.71s/it]
 47%|████▋     | 8928/19090 [4:20:17<4:28:30,  1.59s/it]
 47%|████▋     | 8929/19090 [4:20:20<5:24:14,  1.91s/it]
 47%|████▋     | 8930/19090 [4:20:21<4:52:15,  1.73s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 30.56936264038086, 'learning_rate': 3.208098621066908e-05, 'recon_loss': 0.001361959395580925, 'recon_static': 0.0004011248172901105, 'recon_motion': 0.000960834497527685, 'recon_norm': 1.2297574877738953, 'recon_norm_std': 0.05002250773832202, 'target_norm': 1.6202982366085052, 'recon_cos': 0.6825436666607857, 'gradnorm/compressor.query': 0.05054176696102127, 'gradnorm/compressor.layer0': 1.1782304974582376, 'gradnorm/compressor.layer1': 29.517255488646487, 'gradnorm/compressor.layer2': 6.3307915991389665, 'gradnorm/compressor.layer3': 4.052802380830026, 'gradnorm/compressor.layer4': 1.5072813531522324, 'gradnorm/compressor.layer5': 0.9844207450289005, 'gradnorm/compressor.layer6': 0.5060116454016992, 'gradnorm/compressor.layer7': 0.5840340532587371, 'gradnorm/decoder.layer0': 0.95155539799688, 'gradnorm/decoder.layer1': 0.683361166487993, 'gradnorm/decoder.layer2': 0.3348931780641141, 'gradnorm/max_param': 20.81863784790039, 'epoch': 4.68}

 47%|████▋     | 8930/19090 [4:20:21<4:52:15,  1.73s/it]
 47%|████▋     | 8931/19090 [4:20:22<4:29:57,  1.59s/it]
 47%|████▋     | 8932/19090 [4:20:25<5:12:26,  1.85s/it]
 47%|████▋     | 8933/19090 [4:20:26<4:43:39,  1.68s/it]
 47%|████▋     | 8934/19090 [4:20:27<4:23:42,  1.56s/it]
 47%|████▋     | 8935/19090 [4:20:30<5:35:33,  1.98s/it]
 47%|████▋     | 8936/19090 [4:20:31<5:00:04,  1.77s/it]
 47%|████▋     | 8937/19090 [4:20:33<4:35:37,  1.63s/it]
 47%|████▋     | 8938/19090 [4:20:35<5:23:59,  1.91s/it]
 47%|████▋     | 8939/19090 [4:20:37<4:51:48,  1.72s/it]
 47%|████▋     | 8940/19090 [4:20:38<4:50:37,  1.72s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 31.997331619262695, 'learning_rate': 3.203713320660302e-05, 'recon_loss': 0.001358566779526882, 'recon_static': 0.0003938776564609725, 'recon_motion': 0.0009646891398006119, 'recon_norm': 1.2260682612657547, 'recon_norm_std': 0.05111015653237701, 'target_norm': 1.6215049147605896, 'recon_cos': 0.6841564401984215, 'gradnorm/compressor.query': 0.047279837438008454, 'gradnorm/compressor.layer0': 1.2002897667712151, 'gradnorm/compressor.layer1': 30.711758288446607, 'gradnorm/compressor.layer2': 7.62724546294734, 'gradnorm/compressor.layer3': 4.012617001357713, 'gradnorm/compressor.layer4': 1.370110009433598, 'gradnorm/compressor.layer5': 0.9077931452935059, 'gradnorm/compressor.layer6': 0.5206762617411963, 'gradnorm/compressor.layer7': 0.6745744414525953, 'gradnorm/decoder.layer0': 0.9577222548219743, 'gradnorm/decoder.layer1': 0.6542327621705017, 'gradnorm/decoder.layer2': 0.3542945878565945, 'gradnorm/max_param': 21.66165542602539, 'epoch': 4.68}

 47%|████▋     | 8940/19090 [4:20:38<4:50:37,  1.72s/it]
 47%|████▋     | 8941/19090 [4:20:40<5:14:05,  1.86s/it]
 47%|████▋     | 8942/19090 [4:20:42<4:44:59,  1.68s/it]
 47%|████▋     | 8943/19090 [4:20:44<5:10:01,  1.83s/it]
 47%|████▋     | 8944/19090 [4:20:46<5:00:47,  1.78s/it]
 47%|████▋     | 8945/19090 [4:20:47<4:35:31,  1.63s/it]
 47%|████▋     | 8946/19090 [4:20:49<5:19:16,  1.89s/it]
 47%|████▋     | 8947/19090 [4:20:51<4:55:17,  1.75s/it]
 47%|████▋     | 8948/19090 [4:20:52<4:31:33,  1.61s/it]
 47%|████▋     | 8949/19090 [4:20:55<5:23:05,  1.91s/it]
 47%|████▋     | 8950/19090 [4:20:56<5:00:45,  1.78s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 33.58758544921875, 'learning_rate': 3.199325667380509e-05, 'recon_loss': 0.0013653823029017077, 'recon_static': 0.00039366444252664223, 'recon_motion': 0.0009717178298160433, 'recon_norm': 1.2231352031230927, 'recon_norm_std': 0.049494359828531745, 'target_norm': 1.6228502362966537, 'recon_cos': 0.6827344670891762, 'gradnorm/compressor.query': 0.053848348754232105, 'gradnorm/compressor.layer0': 1.3487631412321976, 'gradnorm/compressor.layer1': 32.30117593520203, 'gradnorm/compressor.layer2': 7.810481184473067, 'gradnorm/compressor.layer3': 3.959783442384633, 'gradnorm/compressor.layer4': 1.581789426806307, 'gradnorm/compressor.layer5': 1.0224797084072692, 'gradnorm/compressor.layer6': 0.607020843691595, 'gradnorm/compressor.layer7': 0.7109738015768412, 'gradnorm/decoder.layer0': 1.0974007848849097, 'gradnorm/decoder.layer1': 0.6975594977840679, 'gradnorm/decoder.layer2': 0.3770403390508953, 'gradnorm/max_param': 22.78271484375, 'epoch': 4.69}

 47%|████▋     | 8950/19090 [4:20:56<5:00:45,  1.78s/it]
 47%|████▋     | 8951/19090 [4:20:57<4:35:40,  1.63s/it]
 47%|████▋     | 8952/19090 [4:21:00<5:30:54,  1.96s/it]
 47%|████▋     | 8953/19090 [4:21:01<5:00:45,  1.78s/it]
 47%|████▋     | 8954/19090 [4:21:03<4:35:19,  1.63s/it]
 47%|████▋     | 8955/19090 [4:21:05<5:10:11,  1.84s/it]
 47%|████▋     | 8956/19090 [4:21:07<4:52:11,  1.73s/it]
 47%|████▋     | 8957/19090 [4:21:08<4:29:27,  1.60s/it]
 47%|████▋     | 8958/19090 [4:21:10<5:05:51,  1.81s/it]
 47%|████▋     | 8959/19090 [4:21:12<4:48:36,  1.71s/it]
 47%|████▋     | 8960/19090 [4:21:13<4:26:53,  1.58s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 34.33567810058594, 'learning_rate': 3.1949356758976956e-05, 'recon_loss': 0.00138468743825797, 'recon_static': 0.0004117037831747439, 'recon_motion': 0.0009729836325277575, 'recon_norm': 1.2343790173530578, 'recon_norm_std': 0.04913696907460689, 'target_norm': 1.6224061459302903, 'recon_cos': 0.6776885181665421, 'gradnorm/compressor.query': 0.07305506645008893, 'gradnorm/compressor.layer0': 1.4934607062060405, 'gradnorm/compressor.layer1': 32.96215314628373, 'gradnorm/compressor.layer2': 7.893906679508881, 'gradnorm/compressor.layer3': 4.5171888342861655, 'gradnorm/compressor.layer4': 1.9297731202933894, 'gradnorm/compressor.layer5': 1.0026145373051647, 'gradnorm/compressor.layer6': 0.5922064544140483, 'gradnorm/compressor.layer7': 0.7322570615345237, 'gradnorm/decoder.layer0': 1.0428590181038975, 'gradnorm/decoder.layer1': 0.7842924006159957, 'gradnorm/decoder.layer2': 0.40242360998367327, 'gradnorm/max_param': 23.246931076049805, 'epoch': 4.69}

 47%|████▋     | 8960/19090 [4:21:13<4:26:53,  1.58s/it]
 47%|████▋     | 8961/19090 [4:21:16<5:29:01,  1.95s/it]
 47%|████▋     | 8962/19090 [4:21:17<5:03:36,  1.80s/it]
 47%|████▋     | 8963/19090 [4:21:18<4:37:19,  1.64s/it]
 47%|████▋     | 8964/19090 [4:21:21<5:31:31,  1.96s/it]
 47%|████▋     | 8965/19090 [4:21:23<5:05:29,  1.81s/it]
 47%|████▋     | 8966/19090 [4:21:24<4:39:02,  1.65s/it]
 47%|████▋     | 8967/19090 [4:21:26<5:21:57,  1.91s/it]
 47%|████▋     | 8968/19090 [4:21:28<5:00:16,  1.78s/it]
 47%|████▋     | 8969/19090 [4:21:29<4:34:59,  1.63s/it]
 47%|████▋     | 8970/19090 [4:21:32<5:30:27,  1.96s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 32.46660614013672, 'learning_rate': 3.190543360889846e-05, 'recon_loss': 0.0013835542922606692, 'recon_static': 0.0004136652118177153, 'recon_motion': 0.0009698890877189115, 'recon_norm': 1.235164687037468, 'recon_norm_std': 0.048374404199421406, 'target_norm': 1.6218790382146835, 'recon_cos': 0.677647240459919, 'gradnorm/compressor.query': 0.05833546415159672, 'gradnorm/compressor.layer0': 1.3571333544262212, 'gradnorm/compressor.layer1': 31.237372448719178, 'gradnorm/compressor.layer2': 7.2370969459162175, 'gradnorm/compressor.layer3': 4.200249918813542, 'gradnorm/compressor.layer4': 1.615108240592548, 'gradnorm/compressor.layer5': 1.0382991080258885, 'gradnorm/compressor.layer6': 0.5942928693487682, 'gradnorm/compressor.layer7': 0.7097459806118165, 'gradnorm/decoder.layer0': 1.0777836687761617, 'gradnorm/decoder.layer1': 0.7719129699757984, 'gradnorm/decoder.layer2': 0.378303855179613, 'gradnorm/max_param': 22.03078842163086, 'epoch': 4.7}

 47%|████▋     | 8970/19090 [4:21:32<5:30:27,  1.96s/it]
 47%|████▋     | 8971/19090 [4:21:33<5:11:17,  1.85s/it]
 47%|████▋     | 8972/19090 [4:21:35<4:42:37,  1.68s/it]
 47%|████▋     | 8973/19090 [4:21:37<5:21:32,  1.91s/it]
 47%|████▋     | 8974/19090 [4:21:39<5:11:49,  1.85s/it]
 47%|████▋     | 8975/19090 [4:21:40<4:43:01,  1.68s/it]
 47%|████▋     | 8976/19090 [4:21:43<5:25:08,  1.93s/it]
 47%|████▋     | 8977/19090 [4:21:44<4:53:58,  1.74s/it]
 47%|████▋     | 8978/19090 [4:21:45<4:30:34,  1.61s/it]
 47%|████▋     | 8979/19090 [4:21:48<5:04:56,  1.81s/it]
 47%|████▋     | 8980/19090 [4:21:50<5:25:29,  1.93s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 35.291038513183594, 'learning_rate': 3.186148737042714e-05, 'recon_loss': 0.0013802992063574493, 'recon_static': 0.00042587497737258675, 'recon_motion': 0.0009544242508127354, 'recon_norm': 1.2387006163597107, 'recon_norm_std': 0.049559625517576936, 'target_norm': 1.617575815320015, 'recon_cos': 0.6763929992914199, 'gradnorm/compressor.query': 0.07082816285074074, 'gradnorm/compressor.layer0': 1.6335443198243267, 'gradnorm/compressor.layer1': 34.04395997732197, 'gradnorm/compressor.layer2': 7.554821814020603, 'gradnorm/compressor.layer3': 4.409762882769432, 'gradnorm/compressor.layer4': 1.6857636313455133, 'gradnorm/compressor.layer5': 1.1154150607686861, 'gradnorm/compressor.layer6': 0.6286019720553913, 'gradnorm/compressor.layer7': 0.7484014880728799, 'gradnorm/decoder.layer0': 1.1502639363967466, 'gradnorm/decoder.layer1': 0.8302432984216899, 'gradnorm/decoder.layer2': 0.4667115233212719, 'gradnorm/max_param': 24.009235382080078, 'epoch': 4.7}

 47%|████▋     | 8980/19090 [4:21:50<5:25:29,  1.93s/it]
 47%|████▋     | 8981/19090 [4:21:51<4:52:44,  1.74s/it]
 47%|████▋     | 8982/19090 [4:21:53<4:52:41,  1.74s/it]
 47%|████▋     | 8983/19090 [4:21:55<5:20:32,  1.90s/it]
 47%|████▋     | 8984/19090 [4:21:56<4:49:29,  1.72s/it]
 47%|████▋     | 8985/19090 [4:21:58<4:42:11,  1.68s/it]
 47%|████▋     | 8986/19090 [4:22:00<5:14:37,  1.87s/it]
 47%|████▋     | 8987/19090 [4:22:02<4:45:56,  1.70s/it]
 47%|████▋     | 8988/19090 [4:22:03<4:49:25,  1.72s/it]
 47%|████▋     | 8989/19090 [4:22:06<5:23:06,  1.92s/it]
 47%|████▋     | 8990/19090 [4:22:07<4:51:12,  1.73s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 30.383188247680664, 'learning_rate': 3.181751819049772e-05, 'recon_loss': 0.001418337077484466, 'recon_static': 0.00044527136196848005, 'recon_motion': 0.0009730657009640708, 'recon_norm': 1.2469893395900726, 'recon_norm_std': 0.048837811686098576, 'target_norm': 1.6224323511123657, 'recon_cos': 0.6689634636044502, 'gradnorm/compressor.query': 0.05418976495102771, 'gradnorm/compressor.layer0': 1.3662576619707345, 'gradnorm/compressor.layer1': 28.967239291110182, 'gradnorm/compressor.layer2': 7.623432686239897, 'gradnorm/compressor.layer3': 4.260618072947326, 'gradnorm/compressor.layer4': 1.5742111773817553, 'gradnorm/compressor.layer5': 1.0045212223930853, 'gradnorm/compressor.layer6': 0.5530070518557941, 'gradnorm/compressor.layer7': 0.6873982094056308, 'gradnorm/decoder.layer0': 1.0019498531295568, 'gradnorm/decoder.layer1': 0.7071928819560305, 'gradnorm/decoder.layer2': 0.35816164015696467, 'gradnorm/max_param': 20.42701530456543, 'epoch': 4.71}

 47%|████▋     | 8990/19090 [4:22:07<4:51:12,  1.73s/it]
 47%|████▋     | 8991/19090 [4:22:09<4:46:58,  1.70s/it]
 47%|████▋     | 8992/19090 [4:22:11<5:27:18,  1.94s/it]
 47%|████▋     | 8993/19090 [4:22:13<4:54:44,  1.75s/it]
 47%|████▋     | 8994/19090 [4:22:14<4:31:35,  1.61s/it]
 47%|████▋     | 8995/19090 [4:22:16<5:25:12,  1.93s/it]
 47%|████▋     | 8996/19090 [4:22:18<4:52:46,  1.74s/it]
 47%|████▋     | 8997/19090 [4:22:19<4:35:54,  1.64s/it]
 47%|████▋     | 8998/19090 [4:22:22<5:18:09,  1.89s/it]
 47%|████▋     | 8999/19090 [4:22:23<4:47:41,  1.71s/it]
 47%|████▋     | 9000/19090 [4:22:24<4:29:31,  1.60s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 29.9147891998291, 'learning_rate': 3.177352621612165e-05, 'recon_loss': 0.0013947463390650227, 'recon_static': 0.0004374024909338914, 'recon_motion': 0.0009573438001098111, 'recon_norm': 1.2473191320896149, 'recon_norm_std': 0.04906520731747151, 'target_norm': 1.6189899444580078, 'recon_cos': 0.6739417910575867, 'gradnorm/compressor.query': 0.050354662804194025, 'gradnorm/compressor.layer0': 1.2390319762733504, 'gradnorm/compressor.layer1': 28.780209798995884, 'gradnorm/compressor.layer2': 6.748111319337717, 'gradnorm/compressor.layer3': 3.777474023412814, 'gradnorm/compressor.layer4': 1.458254875615741, 'gradnorm/compressor.layer5': 0.9279074095932447, 'gradnorm/compressor.layer6': 0.49147846983013604, 'gradnorm/compressor.layer7': 0.7472807544187677, 'gradnorm/decoder.layer0': 0.9168748221888041, 'gradnorm/decoder.layer1': 0.706278721892082, 'gradnorm/decoder.layer2': 0.3506491558913348, 'gradnorm/max_param': 20.299482345581055, 'epoch': 4.71}

 47%|████▋     | 9000/19090 [4:22:24<4:29:31,  1.60s/it]
 47%|████▋     | 9001/19090 [4:22:28<6:02:02,  2.15s/it]
 47%|████▋     | 9002/19090 [4:22:29<5:18:26,  1.89s/it]
 47%|████▋     | 9003/19090 [4:22:30<4:47:41,  1.71s/it]
 47%|████▋     | 9004/19090 [4:22:32<4:41:51,  1.68s/it]
 47%|████▋     | 9005/19090 [4:22:33<4:22:05,  1.56s/it]
 47%|████▋     | 9006/19090 [4:22:35<4:44:18,  1.69s/it]
 47%|████▋     | 9007/19090 [4:22:37<4:47:44,  1.71s/it]
 47%|████▋     | 9008/19090 [4:22:38<4:26:25,  1.59s/it]
 47%|████▋     | 9009/19090 [4:22:41<5:06:04,  1.82s/it]
 47%|████▋     | 9010/19090 [4:22:43<5:09:53,  1.84s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 33.284706115722656, 'learning_rate': 3.172951159438656e-05, 'recon_loss': 0.0013890657312003896, 'recon_static': 0.0004319674291764386, 'recon_motion': 0.000957098271464929, 'recon_norm': 1.2425349920988082, 'recon_norm_std': 0.048712652176618576, 'target_norm': 1.6177880585193634, 'recon_cos': 0.6742033377289772, 'gradnorm/compressor.query': 0.06622616809323872, 'gradnorm/compressor.layer0': 1.4550437155117688, 'gradnorm/compressor.layer1': 31.969451628114356, 'gradnorm/compressor.layer2': 7.538124334437931, 'gradnorm/compressor.layer3': 4.452586358939429, 'gradnorm/compressor.layer4': 1.7539267082125674, 'gradnorm/compressor.layer5': 1.094762279189876, 'gradnorm/compressor.layer6': 0.5555828283158971, 'gradnorm/compressor.layer7': 0.7332014893730507, 'gradnorm/decoder.layer0': 1.1011120571900568, 'gradnorm/decoder.layer1': 0.746938696343772, 'gradnorm/decoder.layer2': 0.40494135844162704, 'gradnorm/max_param': 22.544382095336914, 'epoch': 4.72}

 47%|████▋     | 9010/19090 [4:22:43<5:09:53,  1.84s/it]
 47%|████▋     | 9011/19090 [4:22:44<4:41:43,  1.68s/it]
 47%|████▋     | 9012/19090 [4:22:46<5:17:31,  1.89s/it]
 47%|████▋     | 9013/19090 [4:22:48<4:56:12,  1.76s/it]
 47%|████▋     | 9014/19090 [4:22:49<4:31:57,  1.62s/it]
 47%|████▋     | 9015/19090 [4:22:52<5:32:22,  1.98s/it]
 47%|████▋     | 9016/19090 [4:22:53<5:04:20,  1.81s/it]
 47%|████▋     | 9017/19090 [4:22:54<4:37:19,  1.65s/it]
 47%|████▋     | 9018/19090 [4:22:57<5:26:58,  1.95s/it]
 47%|████▋     | 9019/19090 [4:22:59<5:11:50,  1.86s/it]
 47%|████▋     | 9020/19090 [4:23:00<4:42:52,  1.69s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 36.746337890625, 'learning_rate': 3.168547447245581e-05, 'recon_loss': 0.0013912710128352046, 'recon_static': 0.00043879770673811437, 'recon_motion': 0.0009524733148282393, 'recon_norm': 1.2440756618976594, 'recon_norm_std': 0.04950987389311194, 'target_norm': 1.618981972336769, 'recon_cos': 0.6741775795817375, 'gradnorm/compressor.query': 0.06271713135644867, 'gradnorm/compressor.layer0': 1.5953264854184284, 'gradnorm/compressor.layer1': 35.309991960159984, 'gradnorm/compressor.layer2': 8.581490723191779, 'gradnorm/compressor.layer3': 4.385922774999218, 'gradnorm/compressor.layer4': 1.8140678201970328, 'gradnorm/compressor.layer5': 1.1734577934044375, 'gradnorm/compressor.layer6': 0.6479043262283672, 'gradnorm/compressor.layer7': 0.8499170557392675, 'gradnorm/decoder.layer0': 1.2162740940636114, 'gradnorm/decoder.layer1': 0.7890106033028508, 'gradnorm/decoder.layer2': 0.3981648241363459, 'gradnorm/max_param': 24.902528762817383, 'epoch': 4.72}

 47%|████▋     | 9020/19090 [4:23:00<4:42:52,  1.69s/it]
 47%|████▋     | 9021/19090 [4:23:02<5:10:43,  1.85s/it]
 47%|████▋     | 9022/19090 [4:23:04<5:01:40,  1.80s/it]
 47%|████▋     | 9023/19090 [4:23:05<4:35:31,  1.64s/it]
 47%|████▋     | 9024/19090 [4:23:08<5:12:43,  1.86s/it]
 47%|████▋     | 9025/19090 [4:23:09<4:52:57,  1.75s/it]
 47%|████▋     | 9026/19090 [4:23:10<4:30:05,  1.61s/it]
 47%|████▋     | 9027/19090 [4:23:12<4:53:18,  1.75s/it]
 47%|████▋     | 9028/19090 [4:23:14<4:57:39,  1.77s/it]
 47%|████▋     | 9029/19090 [4:23:16<4:44:25,  1.70s/it]
 47%|████▋     | 9030/19090 [4:23:18<5:00:25,  1.79s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 32.8927116394043, 'learning_rate': 3.164141499756804e-05, 'recon_loss': 0.0013956085807876662, 'recon_static': 0.0004306397830077913, 'recon_motion': 0.000964968791231513, 'recon_norm': 1.2416342467069625, 'recon_norm_std': 0.0491412696428597, 'target_norm': 1.6207488536834718, 'recon_cos': 0.6739233255386352, 'gradnorm/compressor.query': 0.05902409471427438, 'gradnorm/compressor.layer0': 1.4668640852767112, 'gradnorm/compressor.layer1': 31.514220887784337, 'gradnorm/compressor.layer2': 7.779345421013478, 'gradnorm/compressor.layer3': 4.444135620725955, 'gradnorm/compressor.layer4': 1.7104579305685577, 'gradnorm/compressor.layer5': 0.9714053586232018, 'gradnorm/compressor.layer6': 0.5791780572452655, 'gradnorm/compressor.layer7': 0.6906152068604591, 'gradnorm/decoder.layer0': 1.0335586599562487, 'gradnorm/decoder.layer1': 0.6938062275619413, 'gradnorm/decoder.layer2': 0.35937174505335356, 'gradnorm/max_param': 22.226545333862305, 'epoch': 4.73}

 47%|████▋     | 9030/19090 [4:23:18<5:00:25,  1.79s/it]
 47%|████▋     | 9031/19090 [4:23:20<4:56:32,  1.77s/it]
 47%|████▋     | 9032/19090 [4:23:21<4:51:23,  1.74s/it]
 47%|████▋     | 9033/19090 [4:23:23<5:10:52,  1.85s/it]
 47%|████▋     | 9034/19090 [4:23:25<4:57:19,  1.77s/it]
 47%|████▋     | 9035/19090 [4:23:26<4:43:24,  1.69s/it]
 47%|████▋     | 9036/19090 [4:23:29<5:08:18,  1.84s/it]
 47%|████▋     | 9037/19090 [4:23:30<5:05:42,  1.82s/it]
 47%|████▋     | 9038/19090 [4:23:32<4:56:23,  1.77s/it]
 47%|████▋     | 9039/19090 [4:23:34<5:23:01,  1.93s/it]
 47%|████▋     | 9040/19090 [4:23:36<5:07:58,  1.84s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 36.4096565246582, 'learning_rate': 3.1597333317036545e-05, 'recon_loss': 0.0014008144178660586, 'recon_static': 0.00042347187918494454, 'recon_motion': 0.0009773425379535183, 'recon_norm': 1.2398564249277115, 'recon_norm_std': 0.04920789794996381, 'target_norm': 1.623752436041832, 'recon_cos': 0.6741968438029289, 'gradnorm/compressor.query': 0.06130760990202874, 'gradnorm/compressor.layer0': 1.523033440631672, 'gradnorm/compressor.layer1': 35.18004056146397, 'gradnorm/compressor.layer2': 7.8015411419445275, 'gradnorm/compressor.layer3': 4.19413618649062, 'gradnorm/compressor.layer4': 1.8291427518053187, 'gradnorm/compressor.layer5': 1.051962659858472, 'gradnorm/compressor.layer6': 0.5942215552754052, 'gradnorm/compressor.layer7': 0.7321602722914563, 'gradnorm/decoder.layer0': 1.1137057381856754, 'gradnorm/decoder.layer1': 0.7313012260035511, 'gradnorm/decoder.layer2': 0.36461405170548206, 'gradnorm/max_param': 24.812191009521484, 'epoch': 4.74}

 47%|████▋     | 9040/19090 [4:23:36<5:07:58,  1.84s/it]
 47%|████▋     | 9041/19090 [4:23:37<4:40:52,  1.68s/it]
 47%|████▋     | 9042/19090 [4:23:40<5:19:42,  1.91s/it]
 47%|████▋     | 9043/19090 [4:23:41<4:51:28,  1.74s/it]
 47%|████▋     | 9044/19090 [4:23:42<4:28:14,  1.60s/it]
 47%|████▋     | 9045/19090 [4:23:45<5:18:44,  1.90s/it]
 47%|████▋     | 9046/19090 [4:23:46<4:49:50,  1.73s/it]
 47%|████▋     | 9047/19090 [4:23:48<4:33:44,  1.64s/it]
 47%|████▋     | 9048/19090 [4:23:50<5:31:10,  1.98s/it]
 47%|████▋     | 9049/19090 [4:23:52<4:56:22,  1.77s/it]
 47%|████▋     | 9050/19090 [4:23:53<4:36:01,  1.65s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 31.775487899780273, 'learning_rate': 3.155322957824891e-05, 'recon_loss': 0.0013790606695692985, 'recon_static': 0.00042069103074027225, 'recon_motion': 0.0009583696431946009, 'recon_norm': 1.2368365317583083, 'recon_norm_std': 0.04919444127008319, 'target_norm': 1.621100866794586, 'recon_cos': 0.6782633498311043, 'gradnorm/compressor.query': 0.04851562156769376, 'gradnorm/compressor.layer0': 1.2681056542798652, 'gradnorm/compressor.layer1': 30.525631999753504, 'gradnorm/compressor.layer2': 7.2914391473060505, 'gradnorm/compressor.layer3': 4.066667971584757, 'gradnorm/compressor.layer4': 1.727894058110293, 'gradnorm/compressor.layer5': 0.9694426429213211, 'gradnorm/compressor.layer6': 0.5972487862048903, 'gradnorm/compressor.layer7': 0.6960469770718957, 'gradnorm/decoder.layer0': 1.0708351251510493, 'gradnorm/decoder.layer1': 0.6897687108667209, 'gradnorm/decoder.layer2': 0.40609241143388, 'gradnorm/max_param': 21.530271530151367, 'epoch': 4.74}

 47%|████▋     | 9050/19090 [4:23:53<4:36:01,  1.65s/it]
 47%|████▋     | 9051/19090 [4:23:56<5:31:11,  1.98s/it]
 47%|████▋     | 9052/19090 [4:23:57<4:56:10,  1.77s/it]
 47%|████▋     | 9053/19090 [4:23:58<4:36:17,  1.65s/it]
 47%|████▋     | 9054/19090 [4:24:01<5:30:20,  1.97s/it]
 47%|████▋     | 9055/19090 [4:24:03<4:55:51,  1.77s/it]
 47%|████▋     | 9056/19090 [4:24:04<4:38:47,  1.67s/it]
 47%|████▋     | 9057/19090 [4:24:07<5:35:08,  2.00s/it]
 47%|████▋     | 9058/19090 [4:24:08<4:59:39,  1.79s/it]
 47%|████▋     | 9059/19090 [4:24:09<4:34:45,  1.64s/it]
 47%|████▋     | 9060/19090 [4:24:12<5:28:57,  1.97s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 32.86263656616211, 'learning_rate': 3.15091039286665e-05, 'recon_loss': 0.0013808812946081161, 'recon_static': 0.00040614604513393714, 'recon_motion': 0.0009747352887643501, 'recon_norm': 1.2289444148540496, 'recon_norm_std': 0.0474745255894959, 'target_norm': 1.622349840402603, 'recon_cos': 0.6783357605338096, 'gradnorm/compressor.query': 0.05550605533564859, 'gradnorm/compressor.layer0': 1.213979432002484, 'gradnorm/compressor.layer1': 31.423800871798893, 'gradnorm/compressor.layer2': 8.075658189150175, 'gradnorm/compressor.layer3': 4.348857484055551, 'gradnorm/compressor.layer4': 1.7766360708015358, 'gradnorm/compressor.layer5': 1.0753388824374897, 'gradnorm/compressor.layer6': 0.6007546040711363, 'gradnorm/compressor.layer7': 0.6978474601669868, 'gradnorm/decoder.layer0': 1.028281708343223, 'gradnorm/decoder.layer1': 0.7232351922405934, 'gradnorm/decoder.layer2': 0.38847411218697997, 'gradnorm/max_param': 22.16277313232422, 'epoch': 4.75}

 47%|████▋     | 9060/19090 [4:24:12<5:28:57,  1.97s/it]
 47%|████▋     | 9061/19090 [4:24:13<4:55:15,  1.77s/it]
 47%|████▋     | 9062/19090 [4:24:15<4:31:22,  1.62s/it]
 47%|████▋     | 9063/19090 [4:24:18<5:47:37,  2.08s/it]
 47%|████▋     | 9064/19090 [4:24:19<5:08:15,  1.84s/it]
 47%|████▋     | 9065/19090 [4:24:20<4:40:20,  1.68s/it]
 47%|████▋     | 9066/19090 [4:24:23<5:27:48,  1.96s/it]
 47%|████▋     | 9067/19090 [4:24:24<4:54:39,  1.76s/it]
 48%|████▊     | 9068/19090 [4:24:26<4:30:56,  1.62s/it]
 48%|████▊     | 9069/19090 [4:24:28<5:13:44,  1.88s/it]
 48%|████▊     | 9070/19090 [4:24:29<4:44:24,  1.70s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 36.10485076904297, 'learning_rate': 3.146495651582388e-05, 'recon_loss': 0.0013616491633001715, 'recon_static': 0.00040859720393200406, 'recon_motion': 0.0009530519484542311, 'recon_norm': 1.23190738260746, 'recon_norm_std': 0.04816659651696682, 'target_norm': 1.6174187242984772, 'recon_cos': 0.6813693180680275, 'gradnorm/compressor.query': 0.05659975023968557, 'gradnorm/compressor.layer0': 1.5355096851814816, 'gradnorm/compressor.layer1': 34.69514097992956, 'gradnorm/compressor.layer2': 8.511798972057708, 'gradnorm/compressor.layer3': 4.353575646597235, 'gradnorm/compressor.layer4': 1.634118976561782, 'gradnorm/compressor.layer5': 1.012728036703074, 'gradnorm/compressor.layer6': 0.5288415618510388, 'gradnorm/compressor.layer7': 0.655748584470919, 'gradnorm/decoder.layer0': 0.9668556889070866, 'gradnorm/decoder.layer1': 0.7650744346483686, 'gradnorm/decoder.layer2': 0.34062317007780685, 'gradnorm/max_param': 24.470775604248047, 'epoch': 4.75}

 48%|████▊     | 9070/19090 [4:24:29<4:44:24,  1.70s/it]
 48%|████▊     | 9071/19090 [4:24:31<4:23:56,  1.58s/it]
 48%|████▊     | 9072/19090 [4:24:33<5:22:01,  1.93s/it]
 48%|████▊     | 9073/19090 [4:24:35<4:49:42,  1.74s/it]
 48%|████▊     | 9074/19090 [4:24:36<4:26:59,  1.60s/it]
 48%|████▊     | 9075/19090 [4:24:39<5:30:42,  1.98s/it]
 48%|████▊     | 9076/19090 [4:24:40<4:55:50,  1.77s/it]
 48%|████▊     | 9077/19090 [4:24:41<4:31:27,  1.63s/it]
 48%|████▊     | 9078/19090 [4:24:44<5:30:32,  1.98s/it]
 48%|████▊     | 9079/19090 [4:24:45<4:55:50,  1.77s/it]
 48%|████▊     | 9080/19090 [4:24:47<4:31:21,  1.63s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 34.750770568847656, 'learning_rate': 3.1420787487328406e-05, 'recon_loss': 0.0013782895111944526, 'recon_static': 0.0004271203033567872, 'recon_motion': 0.0009511692522210069, 'recon_norm': 1.2411728084087372, 'recon_norm_std': 0.04925372265279293, 'target_norm': 1.617538270354271, 'recon_cos': 0.6771398261189461, 'gradnorm/compressor.query': 0.053602918464140656, 'gradnorm/compressor.layer0': 1.3859959827629098, 'gradnorm/compressor.layer1': 33.685793237144125, 'gradnorm/compressor.layer2': 7.10134521420543, 'gradnorm/compressor.layer3': 3.789344635517286, 'gradnorm/compressor.layer4': 1.6270455177331433, 'gradnorm/compressor.layer5': 1.0339061169900243, 'gradnorm/compressor.layer6': 0.5569367273715103, 'gradnorm/compressor.layer7': 0.7477343830351371, 'gradnorm/decoder.layer0': 0.9289916824567988, 'gradnorm/decoder.layer1': 0.7621622603938552, 'gradnorm/decoder.layer2': 0.3765330172508444, 'gradnorm/max_param': 23.76067352294922, 'epoch': 4.76}

 48%|████▊     | 9080/19090 [4:24:47<4:31:21,  1.63s/it]
 48%|████▊     | 9081/19090 [4:24:50<5:35:12,  2.01s/it]
 48%|████▊     | 9082/19090 [4:24:51<4:59:06,  1.79s/it]
 48%|████▊     | 9083/19090 [4:24:52<4:33:42,  1.64s/it]
 48%|████▊     | 9084/19090 [4:24:55<5:10:04,  1.86s/it]
 48%|████▊     | 9085/19090 [4:24:56<4:41:13,  1.69s/it]
 48%|████▊     | 9086/19090 [4:24:57<4:21:13,  1.57s/it]
 48%|████▊     | 9087/19090 [4:25:00<5:31:01,  1.99s/it]
 48%|████▊     | 9088/19090 [4:25:01<4:56:05,  1.78s/it]
 48%|████▊     | 9089/19090 [4:25:03<4:31:44,  1.63s/it]
 48%|████▊     | 9090/19090 [4:25:06<5:29:33,  1.98s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 34.43059539794922, 'learning_rate': 3.137659699085973e-05, 'recon_loss': 0.0013978928851429372, 'recon_static': 0.0004336155521741603, 'recon_motion': 0.0009642773569794372, 'recon_norm': 1.2420312851667403, 'recon_norm_std': 0.04856013301759958, 'target_norm': 1.6218577295541763, 'recon_cos': 0.674050061404705, 'gradnorm/compressor.query': 0.056332237838187485, 'gradnorm/compressor.layer0': 1.4528791978557651, 'gradnorm/compressor.layer1': 33.17946399589191, 'gradnorm/compressor.layer2': 7.604877372520495, 'gradnorm/compressor.layer3': 4.193811243920473, 'gradnorm/compressor.layer4': 1.7137620160038436, 'gradnorm/compressor.layer5': 1.1506154175897572, 'gradnorm/compressor.layer6': 0.5666376101767201, 'gradnorm/compressor.layer7': 0.8068248547633405, 'gradnorm/decoder.layer0': 1.0250712160631077, 'gradnorm/decoder.layer1': 0.7941649861499769, 'gradnorm/decoder.layer2': 0.3717860740636585, 'gradnorm/max_param': 23.401771545410156, 'epoch': 4.76}

 48%|████▊     | 9090/19090 [4:25:06<5:29:33,  1.98s/it]
 48%|████▊     | 9091/19090 [4:25:07<4:55:08,  1.77s/it]
 48%|████▊     | 9092/19090 [4:25:08<4:30:50,  1.63s/it]
 48%|████▊     | 9093/19090 [4:25:11<5:14:00,  1.88s/it]
 48%|████▊     | 9094/19090 [4:25:12<4:43:57,  1.70s/it]
 48%|████▊     | 9095/19090 [4:25:13<4:22:55,  1.58s/it]
 48%|████▊     | 9096/19090 [4:25:16<5:23:25,  1.94s/it]
 48%|████▊     | 9097/19090 [4:25:17<4:50:41,  1.75s/it]
 48%|████▊     | 9098/19090 [4:25:19<4:27:59,  1.61s/it]
 48%|████▊     | 9099/19090 [4:25:21<5:08:17,  1.85s/it]
 48%|████▊     | 9100/19090 [4:25:23<4:54:08,  1.77s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 40.016666412353516, 'learning_rate': 3.133238517416924e-05, 'recon_loss': 0.001412723257089965, 'recon_static': 0.00045396383575280197, 'recon_motion': 0.0009587594104232267, 'recon_norm': 1.2525844365358352, 'recon_norm_std': 0.0490714474581182, 'target_norm': 1.619697493314743, 'recon_cos': 0.6695465117692947, 'gradnorm/compressor.query': 0.060717235358798624, 'gradnorm/compressor.layer0': 1.600736600477504, 'gradnorm/compressor.layer1': 38.42342857269833, 'gradnorm/compressor.layer2': 9.318927364593122, 'gradnorm/compressor.layer3': 5.2539677839681875, 'gradnorm/compressor.layer4': 1.9240606022289954, 'gradnorm/compressor.layer5': 1.143915993990677, 'gradnorm/compressor.layer6': 0.5950008066957279, 'gradnorm/compressor.layer7': 0.7646859696922065, 'gradnorm/decoder.layer0': 1.1083903711504395, 'gradnorm/decoder.layer1': 0.7976889473754692, 'gradnorm/decoder.layer2': 0.38305387866724555, 'gradnorm/max_param': 27.09972381591797, 'epoch': 4.77}

 48%|████▊     | 9100/19090 [4:25:23<4:54:08,  1.77s/it]
 48%|████▊     | 9101/19090 [4:25:24<4:30:33,  1.63s/it]
 48%|████▊     | 9102/19090 [4:25:26<5:07:16,  1.85s/it]
 48%|████▊     | 9103/19090 [4:25:28<4:51:31,  1.75s/it]
 48%|████▊     | 9104/19090 [4:25:29<4:28:22,  1.61s/it]
 48%|████▊     | 9105/19090 [4:25:31<5:09:27,  1.86s/it]
 48%|████▊     | 9106/19090 [4:25:33<5:11:54,  1.87s/it]
 48%|████▊     | 9107/19090 [4:25:35<4:42:34,  1.70s/it]
 48%|████▊     | 9108/19090 [4:25:37<5:09:07,  1.86s/it]
 48%|████▊     | 9109/19090 [4:25:39<5:05:52,  1.84s/it]
 48%|████▊     | 9110/19090 [4:25:40<4:38:27,  1.67s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 33.43345260620117, 'learning_rate': 3.128815218507963e-05, 'recon_loss': 0.0013899917510570958, 'recon_static': 0.0004314520861953497, 'recon_motion': 0.0009585396532202139, 'recon_norm': 1.2399237006902695, 'recon_norm_std': 0.04821286667138338, 'target_norm': 1.619676199555397, 'recon_cos': 0.6746179327368736, 'gradnorm/compressor.query': 0.05568451881978304, 'gradnorm/compressor.layer0': 1.4338512553185354, 'gradnorm/compressor.layer1': 31.938283533686104, 'gradnorm/compressor.layer2': 8.316025982822122, 'gradnorm/compressor.layer3': 4.460883412475846, 'gradnorm/compressor.layer4': 1.6190658202336998, 'gradnorm/compressor.layer5': 1.0818927552316042, 'gradnorm/compressor.layer6': 0.582481926649926, 'gradnorm/compressor.layer7': 0.7825362869843385, 'gradnorm/decoder.layer0': 1.0894275814110639, 'gradnorm/decoder.layer1': 0.7468927708510755, 'gradnorm/decoder.layer2': 0.3726343600323795, 'gradnorm/max_param': 22.52387809753418, 'epoch': 4.77}

 48%|████▊     | 9110/19090 [4:25:40<4:38:27,  1.67s/it]
 48%|████▊     | 9111/19090 [4:25:42<5:11:27,  1.87s/it]
 48%|████▊     | 9112/19090 [4:25:44<5:03:03,  1.82s/it]
 48%|████▊     | 9113/19090 [4:25:45<4:35:57,  1.66s/it]
 48%|████▊     | 9114/19090 [4:25:48<5:15:46,  1.90s/it]
 48%|████▊     | 9115/19090 [4:25:50<5:21:22,  1.93s/it]
 48%|████▊     | 9116/19090 [4:25:51<4:48:48,  1.74s/it]
 48%|████▊     | 9117/19090 [4:25:53<5:10:07,  1.87s/it]
 48%|████▊     | 9118/19090 [4:25:55<4:45:33,  1.72s/it]
 48%|████▊     | 9119/19090 [4:25:56<4:23:42,  1.59s/it]
 48%|████▊     | 9120/19090 [4:25:59<5:21:28,  1.93s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 39.20893859863281, 'learning_rate': 3.124389817148441e-05, 'recon_loss': 0.0013987286249175668, 'recon_static': 0.00044480872238636946, 'recon_motion': 0.0009539199047139846, 'recon_norm': 1.2479158759117126, 'recon_norm_std': 0.04942393582314253, 'target_norm': 1.6189771205186845, 'recon_cos': 0.672630549967289, 'gradnorm/compressor.query': 0.05731042934910478, 'gradnorm/compressor.layer0': 1.4797035997878418, 'gradnorm/compressor.layer1': 37.74951818297652, 'gradnorm/compressor.layer2': 8.89590227754009, 'gradnorm/compressor.layer3': 4.716666260484514, 'gradnorm/compressor.layer4': 1.9962523634388405, 'gradnorm/compressor.layer5': 1.1748716061020839, 'gradnorm/compressor.layer6': 0.6583922705846699, 'gradnorm/compressor.layer7': 0.7984036014773799, 'gradnorm/decoder.layer0': 1.2118480089554908, 'gradnorm/decoder.layer1': 0.8307128992351437, 'gradnorm/decoder.layer2': 0.3781491413764431, 'gradnorm/max_param': 26.626909255981445, 'epoch': 4.78}

 48%|████▊     | 9120/19090 [4:25:59<5:21:28,  1.93s/it]
 48%|████▊     | 9121/19090 [4:26:00<4:58:11,  1.79s/it]
 48%|████▊     | 9122/19090 [4:26:01<4:32:29,  1.64s/it]
 48%|████▊     | 9123/19090 [4:26:04<5:29:07,  1.98s/it]
 48%|████▊     | 9124/19090 [4:26:05<4:58:13,  1.80s/it]
 48%|████▊     | 9125/19090 [4:26:07<4:32:24,  1.64s/it]
 48%|████▊     | 9126/19090 [4:26:09<5:15:49,  1.90s/it]
 48%|████▊     | 9127/19090 [4:26:11<4:55:52,  1.78s/it]
 48%|████▊     | 9128/19090 [4:26:12<4:30:50,  1.63s/it]
 48%|████▊     | 9129/19090 [4:26:15<5:22:26,  1.94s/it]
 48%|████▊     | 9130/19090 [4:26:16<4:49:27,  1.74s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 38.289878845214844, 'learning_rate': 3.119962328134732e-05, 'recon_loss': 0.001368557321256958, 'recon_static': 0.0004116479991353117, 'recon_motion': 0.0009569093308527954, 'recon_norm': 1.2317891240119934, 'recon_norm_std': 0.0484193341806531, 'target_norm': 1.6177190244197845, 'recon_cos': 0.6794071212410927, 'gradnorm/compressor.query': 0.061153959997962894, 'gradnorm/compressor.layer0': 1.4002105724829714, 'gradnorm/compressor.layer1': 36.94498952465087, 'gradnorm/compressor.layer2': 8.481849589620031, 'gradnorm/compressor.layer3': 4.4598517758496, 'gradnorm/compressor.layer4': 1.8302333444338497, 'gradnorm/compressor.layer5': 1.0707338158245872, 'gradnorm/compressor.layer6': 0.6461191903471386, 'gradnorm/compressor.layer7': 0.7398304504037012, 'gradnorm/decoder.layer0': 1.0901510154157763, 'gradnorm/decoder.layer1': 0.7581270096163362, 'gradnorm/decoder.layer2': 0.4023161202730525, 'gradnorm/max_param': 26.058822631835938, 'epoch': 4.78}

 48%|████▊     | 9130/19090 [4:26:16<4:49:27,  1.74s/it]
 48%|████▊     | 9131/19090 [4:26:17<4:26:24,  1.61s/it]
 48%|████▊     | 9132/19090 [4:26:20<5:36:38,  2.03s/it]
 48%|████▊     | 9133/19090 [4:26:22<4:59:13,  1.80s/it]
 48%|████▊     | 9134/19090 [4:26:23<4:33:46,  1.65s/it]
 48%|████▊     | 9135/19090 [4:26:26<5:26:50,  1.97s/it]
 48%|████▊     | 9136/19090 [4:26:27<4:52:24,  1.76s/it]
 48%|████▊     | 9137/19090 [4:26:28<4:28:20,  1.62s/it]
 48%|████▊     | 9138/19090 [4:26:31<5:27:24,  1.97s/it]
 48%|████▊     | 9139/19090 [4:26:32<4:53:00,  1.77s/it]
 48%|████▊     | 9140/19090 [4:26:34<4:44:10,  1.71s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 37.893348693847656, 'learning_rate': 3.115532766270196e-05, 'recon_loss': 0.0013678539136890322, 'recon_static': 0.0004121381600270979, 'recon_motion': 0.0009557157114613801, 'recon_norm': 1.2334412187337875, 'recon_norm_std': 0.04905572831630707, 'target_norm': 1.6179151326417922, 'recon_cos': 0.6798413708806038, 'gradnorm/compressor.query': 0.05912769145338891, 'gradnorm/compressor.layer0': 1.4599847798011376, 'gradnorm/compressor.layer1': 36.55431933053534, 'gradnorm/compressor.layer2': 8.435243205853316, 'gradnorm/compressor.layer3': 4.417740361443059, 'gradnorm/compressor.layer4': 1.7059414874615653, 'gradnorm/compressor.layer5': 1.040440899942663, 'gradnorm/compressor.layer6': 0.6134862374967102, 'gradnorm/compressor.layer7': 0.7154874144646174, 'gradnorm/decoder.layer0': 1.1286993662143177, 'gradnorm/decoder.layer1': 0.7613487714509402, 'gradnorm/decoder.layer2': 0.3849211441444998, 'gradnorm/max_param': 25.78365135192871, 'epoch': 4.79}

 48%|████▊     | 9140/19090 [4:26:34<4:44:10,  1.71s/it]
 48%|████▊     | 9141/19090 [4:26:36<5:10:58,  1.88s/it]
 48%|████▊     | 9142/19090 [4:26:37<4:47:21,  1.73s/it]
 48%|████▊     | 9143/19090 [4:26:39<4:47:35,  1.73s/it]
 48%|████▊     | 9144/19090 [4:26:41<5:11:45,  1.88s/it]
 48%|████▊     | 9145/19090 [4:26:43<4:42:06,  1.70s/it]
 48%|████▊     | 9146/19090 [4:26:45<4:55:44,  1.78s/it]
 48%|████▊     | 9147/19090 [4:26:47<5:15:24,  1.90s/it]
 48%|████▊     | 9148/19090 [4:26:48<4:53:13,  1.77s/it]
 48%|████▊     | 9149/19090 [4:26:50<4:59:35,  1.81s/it]
 48%|████▊     | 9150/19090 [4:26:52<5:05:21,  1.84s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 33.49277877807617, 'learning_rate': 3.1111011463651216e-05, 'recon_loss': 0.0013863402738934383, 'recon_static': 0.0004204278171528131, 'recon_motion': 0.0009659124829340726, 'recon_norm': 1.2344767272472381, 'recon_norm_std': 0.04792873589321971, 'target_norm': 1.6210195034742356, 'recon_cos': 0.6762845456600189, 'gradnorm/compressor.query': 0.0527670502979896, 'gradnorm/compressor.layer0': 1.2629700116291986, 'gradnorm/compressor.layer1': 32.21974833261718, 'gradnorm/compressor.layer2': 7.476241564265169, 'gradnorm/compressor.layer3': 4.426369260157868, 'gradnorm/compressor.layer4': 1.5307374991853937, 'gradnorm/compressor.layer5': 1.2068042726895638, 'gradnorm/compressor.layer6': 0.6396385366823802, 'gradnorm/compressor.layer7': 0.6991157029582571, 'gradnorm/decoder.layer0': 1.103899539568052, 'gradnorm/decoder.layer1': 0.7144337296395172, 'gradnorm/decoder.layer2': 0.3777512910481272, 'gradnorm/max_param': 22.726011276245117, 'epoch': 4.79}

 48%|████▊     | 9150/19090 [4:26:52<5:05:21,  1.84s/it]
 48%|████▊     | 9151/19090 [4:26:53<4:40:03,  1.69s/it]
 48%|████▊     | 9152/19090 [4:26:55<4:55:30,  1.78s/it]
 48%|████▊     | 9153/19090 [4:26:58<5:15:03,  1.90s/it]
 48%|████▊     | 9154/19090 [4:26:59<4:44:43,  1.72s/it]
 48%|████▊     | 9155/19090 [4:27:01<5:16:45,  1.91s/it]
 48%|████▊     | 9156/19090 [4:27:03<4:55:54,  1.79s/it]
 48%|████▊     | 9157/19090 [4:27:04<4:30:42,  1.64s/it]
 48%|████▊     | 9158/19090 [4:27:07<5:25:25,  1.97s/it]
 48%|████▊     | 9159/19090 [4:27:08<4:51:41,  1.76s/it]
 48%|████▊     | 9160/19090 [4:27:09<4:27:41,  1.62s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 37.45802688598633, 'learning_rate': 3.1066674832366776e-05, 'recon_loss': 0.0014016808738233522, 'recon_static': 0.0004365832865005359, 'recon_motion': 0.00096509760041954, 'recon_norm': 1.245604920387268, 'recon_norm_std': 0.04853286556899548, 'target_norm': 1.6212957113981248, 'recon_cos': 0.6728623896837235, 'gradnorm/compressor.query': 0.060185886798661895, 'gradnorm/compressor.layer0': 1.5977007199839421, 'gradnorm/compressor.layer1': 35.99464376657745, 'gradnorm/compressor.layer2': 8.674481854094324, 'gradnorm/compressor.layer3': 4.7100896119747055, 'gradnorm/compressor.layer4': 1.7676121155226288, 'gradnorm/compressor.layer5': 1.1061532964304077, 'gradnorm/compressor.layer6': 0.6622392195144196, 'gradnorm/compressor.layer7': 0.825993151079837, 'gradnorm/decoder.layer0': 1.1293552654085346, 'gradnorm/decoder.layer1': 0.7693742698461444, 'gradnorm/decoder.layer2': 0.4063732803798821, 'gradnorm/max_param': 25.389265060424805, 'epoch': 4.8}

 48%|████▊     | 9160/19090 [4:27:09<4:27:41,  1.62s/it]
 48%|████▊     | 9161/19090 [4:27:12<5:17:29,  1.92s/it]
 48%|████▊     | 9162/19090 [4:27:13<4:45:52,  1.73s/it]
 48%|████▊     | 9163/19090 [4:27:15<4:23:32,  1.59s/it]
 48%|████▊     | 9164/19090 [4:27:17<5:29:09,  1.99s/it]
 48%|████▊     | 9165/19090 [4:27:19<4:53:58,  1.78s/it]
 48%|████▊     | 9166/19090 [4:27:20<4:29:32,  1.63s/it]
 48%|████▊     | 9167/19090 [4:27:23<5:27:16,  1.98s/it]
 48%|████▊     | 9168/19090 [4:27:24<5:09:51,  1.87s/it]
 48%|████▊     | 9169/19090 [4:27:26<4:40:39,  1.70s/it]
 48%|████▊     | 9170/19090 [4:27:28<5:06:23,  1.85s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 37.110870361328125, 'learning_rate': 3.102231791708866e-05, 'recon_loss': 0.0014141164283500984, 'recon_static': 0.0004525597578322049, 'recon_motion': 0.0009615566450520419, 'recon_norm': 1.249164980649948, 'recon_norm_std': 0.049121803138405085, 'target_norm': 1.620284950733185, 'recon_cos': 0.6692007035017014, 'gradnorm/compressor.query': 0.06256949018704173, 'gradnorm/compressor.layer0': 1.5514382764983923, 'gradnorm/compressor.layer1': 35.5193712278508, 'gradnorm/compressor.layer2': 9.10235838459578, 'gradnorm/compressor.layer3': 4.877337164494794, 'gradnorm/compressor.layer4': 1.6033939961293764, 'gradnorm/compressor.layer5': 1.1101228584137457, 'gradnorm/compressor.layer6': 0.5825522734144774, 'gradnorm/compressor.layer7': 0.8122833770135895, 'gradnorm/decoder.layer0': 1.0407112635693814, 'gradnorm/decoder.layer1': 0.6881843108932246, 'gradnorm/decoder.layer2': 0.42377181492132265, 'gradnorm/max_param': 25.05032730102539, 'epoch': 4.8}

 48%|████▊     | 9170/19090 [4:27:28<5:06:23,  1.85s/it]
 48%|████▊     | 9171/19090 [4:27:30<5:00:43,  1.82s/it]
 48%|████▊     | 9172/19090 [4:27:31<4:34:03,  1.66s/it]
 48%|████▊     | 9173/19090 [4:27:33<4:57:11,  1.80s/it]
 48%|████▊     | 9174/19090 [4:27:35<4:50:57,  1.76s/it]
 48%|████▊     | 9175/19090 [4:27:36<4:27:12,  1.62s/it]
 48%|████▊     | 9176/19090 [4:27:38<4:50:02,  1.76s/it]
 48%|████▊     | 9177/19090 [4:27:40<4:58:38,  1.81s/it]
 48%|████▊     | 9178/19090 [4:27:41<4:32:57,  1.65s/it]
 48%|████▊     | 9179/19090 [4:27:44<5:15:53,  1.91s/it]
 48%|████▊     | 9180/19090 [4:27:46<5:01:31,  1.83s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 38.24005889892578, 'learning_rate': 3.09779408661247e-05, 'recon_loss': 0.0014167068206006662, 'recon_static': 0.0004576588413328864, 'recon_motion': 0.0009590479676262476, 'recon_norm': 1.2504651516675949, 'recon_norm_std': 0.04892097823321819, 'target_norm': 1.6207061797380446, 'recon_cos': 0.6684550896286965, 'gradnorm/compressor.query': 0.06592444988460795, 'gradnorm/compressor.layer0': 1.5849850806126757, 'gradnorm/compressor.layer1': 36.755714439326475, 'gradnorm/compressor.layer2': 8.947418471394199, 'gradnorm/compressor.layer3': 4.637246285873194, 'gradnorm/compressor.layer4': 1.750119273344536, 'gradnorm/compressor.layer5': 1.1620253588655483, 'gradnorm/compressor.layer6': 0.6036069239226223, 'gradnorm/compressor.layer7': 0.8315298154443597, 'gradnorm/decoder.layer0': 1.0435839492376477, 'gradnorm/decoder.layer1': 0.7297294669978974, 'gradnorm/decoder.layer2': 0.39028937226212224, 'gradnorm/max_param': 25.925201416015625, 'epoch': 4.81}

 48%|████▊     | 9180/19090 [4:27:46<5:01:31,  1.83s/it]
 48%|████▊     | 9181/19090 [4:27:47<4:34:37,  1.66s/it]
 48%|████▊     | 9182/19090 [4:27:49<5:07:25,  1.86s/it]
 48%|████▊     | 9183/19090 [4:27:51<5:00:45,  1.82s/it]
 48%|████▊     | 9184/19090 [4:27:52<4:34:11,  1.66s/it]
 48%|████▊     | 9185/19090 [4:27:55<5:10:50,  1.88s/it]
 48%|████▊     | 9186/19090 [4:27:56<5:06:15,  1.86s/it]
 48%|████▊     | 9187/19090 [4:27:58<4:38:14,  1.69s/it]
 48%|████▊     | 9188/19090 [4:28:00<5:14:21,  1.90s/it]
 48%|████▊     | 9189/19090 [4:28:02<5:03:12,  1.84s/it]
 48%|████▊     | 9190/19090 [4:28:03<4:43:58,  1.72s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 35.81488800048828, 'learning_rate': 3.0933543827850045e-05, 'recon_loss': 0.0014014210377354176, 'recon_static': 0.00044317966094240544, 'recon_motion': 0.0009582414058968425, 'recon_norm': 1.245562916994095, 'recon_norm_std': 0.04899887330830097, 'target_norm': 1.6212068110704423, 'recon_cos': 0.6729288429021836, 'gradnorm/compressor.query': 0.059908341660048846, 'gradnorm/compressor.layer0': 1.578382943101145, 'gradnorm/compressor.layer1': 34.09325052624511, 'gradnorm/compressor.layer2': 9.344537095820009, 'gradnorm/compressor.layer3': 4.788963727570337, 'gradnorm/compressor.layer4': 1.8643539827538798, 'gradnorm/compressor.layer5': 1.111524996892988, 'gradnorm/compressor.layer6': 0.5774158561804723, 'gradnorm/compressor.layer7': 0.7328104031335564, 'gradnorm/decoder.layer0': 1.1617336146539494, 'gradnorm/decoder.layer1': 0.7210503132178706, 'gradnorm/decoder.layer2': 0.39465086606103567, 'gradnorm/max_param': 24.04326057434082, 'epoch': 4.81}

 48%|████▊     | 9190/19090 [4:28:03<4:43:58,  1.72s/it]
 48%|████▊     | 9191/19090 [4:28:05<5:02:36,  1.83s/it]
 48%|████▊     | 9192/19090 [4:28:07<4:51:01,  1.76s/it]
 48%|████▊     | 9193/19090 [4:28:08<4:42:37,  1.71s/it]
 48%|████▊     | 9194/19090 [4:28:11<4:58:48,  1.81s/it]
 48%|████▊     | 9195/19090 [4:28:12<5:01:31,  1.83s/it]
 48%|████▊     | 9196/19090 [4:28:14<4:51:07,  1.77s/it]
 48%|████▊     | 9197/19090 [4:28:16<4:50:05,  1.76s/it]
 48%|████▊     | 9198/19090 [4:28:18<5:05:00,  1.85s/it]
 48%|████▊     | 9199/19090 [4:28:19<4:51:38,  1.77s/it]
 48%|████▊     | 9200/19090 [4:28:21<4:42:56,  1.72s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 38.4594612121582, 'learning_rate': 3.088912695070668e-05, 'recon_loss': 0.0014013444975716992, 'recon_static': 0.0004447943370905705, 'recon_motion': 0.0009565501983161085, 'recon_norm': 1.2459840208292008, 'recon_norm_std': 0.049309176299721, 'target_norm': 1.6209721863269806, 'recon_cos': 0.6726521506905556, 'gradnorm/compressor.query': 0.06699638200646385, 'gradnorm/compressor.layer0': 1.6439919935435223, 'gradnorm/compressor.layer1': 36.947469176056224, 'gradnorm/compressor.layer2': 8.980830652777458, 'gradnorm/compressor.layer3': 4.790472771999906, 'gradnorm/compressor.layer4': 1.8976342656576688, 'gradnorm/compressor.layer5': 1.1142917263476653, 'gradnorm/compressor.layer6': 0.6475693049585524, 'gradnorm/compressor.layer7': 0.7869852502984518, 'gradnorm/decoder.layer0': 1.0475306948441296, 'gradnorm/decoder.layer1': 0.7568299127154352, 'gradnorm/decoder.layer2': 0.3898624415685057, 'gradnorm/max_param': 26.058475494384766, 'epoch': 4.82}

 48%|████▊     | 9200/19090 [4:28:21<4:42:56,  1.72s/it]
 48%|████▊     | 9201/19090 [4:28:23<5:14:20,  1.91s/it]
 48%|████▊     | 9202/19090 [4:28:25<4:51:39,  1.77s/it]
 48%|████▊     | 9203/19090 [4:28:27<4:52:53,  1.78s/it]
 48%|████▊     | 9204/19090 [4:28:29<5:05:54,  1.86s/it]
 48%|████▊     | 9205/19090 [4:28:31<5:11:39,  1.89s/it]
 48%|████▊     | 9206/19090 [4:28:32<4:41:26,  1.71s/it]
 48%|████▊     | 9207/19090 [4:28:34<5:07:11,  1.86s/it]
 48%|████▊     | 9208/19090 [4:28:36<4:58:51,  1.81s/it]
 48%|████▊     | 9209/19090 [4:28:37<4:34:13,  1.67s/it]
 48%|████▊     | 9210/19090 [4:28:39<4:39:26,  1.70s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 38.97732162475586, 'learning_rate': 3.084469038320292e-05, 'recon_loss': 0.0014041947753867135, 'recon_static': 0.00044814366701757536, 'recon_motion': 0.0009560510952724144, 'recon_norm': 1.2465989887714386, 'recon_norm_std': 0.04939138758927584, 'target_norm': 1.620105516910553, 'recon_cos': 0.6716022044420242, 'gradnorm/compressor.query': 0.06270010772382877, 'gradnorm/compressor.layer0': 1.6179211560243614, 'gradnorm/compressor.layer1': 37.39625460388296, 'gradnorm/compressor.layer2': 9.32248843326603, 'gradnorm/compressor.layer3': 4.787321138153317, 'gradnorm/compressor.layer4': 1.9582608256883915, 'gradnorm/compressor.layer5': 1.1375274403374855, 'gradnorm/compressor.layer6': 0.6977288766887145, 'gradnorm/compressor.layer7': 0.7538142963524338, 'gradnorm/decoder.layer0': 1.146856452313778, 'gradnorm/decoder.layer1': 0.7946871099197325, 'gradnorm/decoder.layer2': 0.4150068023710136, 'gradnorm/max_param': 26.37723159790039, 'epoch': 4.82}

 48%|████▊     | 9210/19090 [4:28:39<4:39:26,  1.70s/it]
 48%|████▊     | 9211/19090 [4:28:41<5:05:56,  1.86s/it]
 48%|████▊     | 9212/19090 [4:28:42<4:37:20,  1.68s/it]
 48%|████▊     | 9213/19090 [4:28:44<4:35:02,  1.67s/it]
 48%|████▊     | 9214/19090 [4:28:46<5:07:09,  1.87s/it]
 48%|████▊     | 9215/19090 [4:28:48<4:39:00,  1.70s/it]
 48%|████▊     | 9216/19090 [4:28:50<4:52:41,  1.78s/it]
 48%|████▊     | 9217/19090 [4:28:51<4:48:13,  1.75s/it]
 48%|████▊     | 9218/19090 [4:28:53<4:29:24,  1.64s/it]
 48%|████▊     | 9219/19090 [4:28:55<5:14:31,  1.91s/it]
 48%|████▊     | 9220/19090 [4:28:57<4:43:46,  1.73s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 41.0673828125, 'learning_rate': 3.0800234273912915e-05, 'recon_loss': 0.00140015785000287, 'recon_static': 0.0004299163869291078, 'recon_motion': 0.0009702414492494427, 'recon_norm': 1.2394759386777878, 'recon_norm_std': 0.04872913649305701, 'target_norm': 1.6215556025505067, 'recon_cos': 0.6731060832738877, 'gradnorm/compressor.query': 0.06730510889210353, 'gradnorm/compressor.layer0': 1.5986205084810179, 'gradnorm/compressor.layer1': 39.6802143288023, 'gradnorm/compressor.layer2': 8.810021883951432, 'gradnorm/compressor.layer3': 4.963243996606316, 'gradnorm/compressor.layer4': 1.7144925881061779, 'gradnorm/compressor.layer5': 1.1522043952085987, 'gradnorm/compressor.layer6': 0.6155735836729593, 'gradnorm/compressor.layer7': 0.7765264383765907, 'gradnorm/decoder.layer0': 1.1156974939893776, 'gradnorm/decoder.layer1': 0.7385501674562811, 'gradnorm/decoder.layer2': 0.40114496395254945, 'gradnorm/max_param': 27.988094329833984, 'epoch': 4.83}

 48%|████▊     | 9220/19090 [4:28:57<4:43:46,  1.73s/it]
 48%|████▊     | 9221/19090 [4:28:58<4:38:44,  1.69s/it]
 48%|████▊     | 9222/19090 [4:29:00<4:51:53,  1.77s/it]
 48%|████▊     | 9223/19090 [4:29:02<4:38:47,  1.70s/it]
 48%|████▊     | 9224/19090 [4:29:03<4:32:35,  1.66s/it]
 48%|████▊     | 9225/19090 [4:29:05<5:00:20,  1.83s/it]
 48%|████▊     | 9226/19090 [4:29:07<4:48:54,  1.76s/it]
 48%|████▊     | 9227/19090 [4:29:08<4:28:39,  1.63s/it]
 48%|████▊     | 9228/19090 [4:29:11<5:27:01,  1.99s/it]
 48%|████▊     | 9229/19090 [4:29:12<4:52:03,  1.78s/it]
 48%|████▊     | 9230/19090 [4:29:14<4:27:40,  1.63s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 39.38520431518555, 'learning_rate': 3.0755758771476165e-05, 'recon_loss': 0.0014150591450743377, 'recon_static': 0.0004393637580506038, 'recon_motion': 0.0009756953557371161, 'recon_norm': 1.2439356118440628, 'recon_norm_std': 0.04823483033105731, 'target_norm': 1.6231934607028962, 'recon_cos': 0.6701968282461166, 'gradnorm/compressor.query': 0.0640350131636709, 'gradnorm/compressor.layer0': 1.5979099212199066, 'gradnorm/compressor.layer1': 37.76327988426306, 'gradnorm/compressor.layer2': 9.300025726610922, 'gradnorm/compressor.layer3': 5.208257303318764, 'gradnorm/compressor.layer4': 1.9907582260895074, 'gradnorm/compressor.layer5': 1.2359734338294168, 'gradnorm/compressor.layer6': 0.6472056304071968, 'gradnorm/compressor.layer7': 0.8147646081954488, 'gradnorm/decoder.layer0': 1.212611020304795, 'gradnorm/decoder.layer1': 0.8326697608336945, 'gradnorm/decoder.layer2': 0.46680448837136845, 'gradnorm/max_param': 26.632522583007812, 'epoch': 4.83}

 48%|████▊     | 9230/19090 [4:29:14<4:27:40,  1.63s/it]
 48%|████▊     | 9231/19090 [4:29:17<5:24:34,  1.98s/it]
 48%|████▊     | 9232/19090 [4:29:18<4:50:15,  1.77s/it]
 48%|████▊     | 9233/19090 [4:29:19<4:26:12,  1.62s/it]
 48%|████▊     | 9234/19090 [4:29:22<5:14:00,  1.91s/it]
 48%|████▊     | 9235/19090 [4:29:23<4:42:56,  1.72s/it]
 48%|████▊     | 9236/19090 [4:29:24<4:23:21,  1.60s/it]
 48%|████▊     | 9237/19090 [4:29:27<5:32:25,  2.02s/it]
 48%|████▊     | 9238/19090 [4:29:29<4:55:51,  1.80s/it]
 48%|████▊     | 9239/19090 [4:29:30<4:30:06,  1.65s/it]
 48%|████▊     | 9240/19090 [4:29:33<5:20:32,  1.95s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 35.551570892333984, 'learning_rate': 3.0711264024596976e-05, 'recon_loss': 0.0014011078717885539, 'recon_static': 0.00043785580564872364, 'recon_motion': 0.0009632521148887462, 'recon_norm': 1.2453563779592514, 'recon_norm_std': 0.048496019281446937, 'target_norm': 1.6200939178466798, 'recon_cos': 0.6724417075514794, 'gradnorm/compressor.query': 0.05815185970961569, 'gradnorm/compressor.layer0': 1.5709989195685796, 'gradnorm/compressor.layer1': 33.845008777375504, 'gradnorm/compressor.layer2': 9.161655327762913, 'gradnorm/compressor.layer3': 4.997484528067733, 'gradnorm/compressor.layer4': 1.6463557834571045, 'gradnorm/compressor.layer5': 1.1804403466999103, 'gradnorm/compressor.layer6': 0.6169358235281247, 'gradnorm/compressor.layer7': 0.8176621264059374, 'gradnorm/decoder.layer0': 1.09508656172691, 'gradnorm/decoder.layer1': 0.7419476974978062, 'gradnorm/decoder.layer2': 0.38042895130462656, 'gradnorm/max_param': 23.867528915405273, 'epoch': 4.84}

 48%|████▊     | 9240/19090 [4:29:33<5:20:32,  1.95s/it]
 48%|████▊     | 9241/19090 [4:29:34<4:47:31,  1.75s/it]
 48%|████▊     | 9242/19090 [4:29:35<4:24:38,  1.61s/it]
 48%|████▊     | 9243/19090 [4:29:38<5:20:03,  1.95s/it]
 48%|████▊     | 9244/19090 [4:29:39<4:47:06,  1.75s/it]
 48%|████▊     | 9245/19090 [4:29:40<4:24:20,  1.61s/it]
 48%|████▊     | 9246/19090 [4:29:43<5:35:12,  2.04s/it]
 48%|████▊     | 9247/19090 [4:29:45<4:57:35,  1.81s/it]
 48%|████▊     | 9248/19090 [4:29:46<4:31:22,  1.65s/it]
 48%|████▊     | 9249/19090 [4:29:49<5:25:48,  1.99s/it]
 48%|████▊     | 9250/19090 [4:29:50<4:51:13,  1.78s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 34.540283203125, 'learning_rate': 3.066675018204403e-05, 'recon_loss': 0.0014234841655706987, 'recon_static': 0.00045915700829937125, 'recon_motion': 0.000964327143447008, 'recon_norm': 1.251697912812233, 'recon_norm_std': 0.047730603255331516, 'target_norm': 1.6195459991693497, 'recon_cos': 0.666322872042656, 'gradnorm/compressor.query': 0.06907557576770858, 'gradnorm/compressor.layer0': 1.6879953434459585, 'gradnorm/compressor.layer1': 32.93041869739797, 'gradnorm/compressor.layer2': 8.896730709748235, 'gradnorm/compressor.layer3': 4.317333663802186, 'gradnorm/compressor.layer4': 1.9907573877504348, 'gradnorm/compressor.layer5': 1.08152461671436, 'gradnorm/compressor.layer6': 0.6077492619741047, 'gradnorm/compressor.layer7': 0.7609017710355918, 'gradnorm/decoder.layer0': 1.0260085796049532, 'gradnorm/decoder.layer1': 0.8152376782615868, 'gradnorm/decoder.layer2': 0.4182157410604116, 'gradnorm/max_param': 23.22006607055664, 'epoch': 4.85}

 48%|████▊     | 9250/19090 [4:29:50<4:51:13,  1.78s/it]
 48%|████▊     | 9251/19090 [4:29:51<4:27:07,  1.63s/it]
 48%|████▊     | 9252/19090 [4:29:54<5:36:34,  2.05s/it]
 48%|████▊     | 9253/19090 [4:29:56<4:58:41,  1.82s/it]
 48%|████▊     | 9254/19090 [4:29:57<4:32:20,  1.66s/it]
 48%|████▊     | 9255/19090 [4:30:00<5:37:55,  2.06s/it]
 48%|████▊     | 9256/19090 [4:30:01<4:59:37,  1.83s/it]
 48%|████▊     | 9257/19090 [4:30:03<4:32:45,  1.66s/it]
 48%|████▊     | 9258/19090 [4:30:05<5:32:36,  2.03s/it]
 49%|████▊     | 9259/19090 [4:30:07<4:55:47,  1.81s/it]
 49%|████▊     | 9260/19090 [4:30:08<4:30:11,  1.65s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 36.88338088989258, 'learning_rate': 3.0622217392649854e-05, 'recon_loss': 0.0014558796363417059, 'recon_static': 0.0004971222777385264, 'recon_motion': 0.0009587573746102862, 'recon_norm': 1.2696110337972641, 'recon_norm_std': 0.047614842653274536, 'target_norm': 1.6212823927402495, 'recon_cos': 0.6595308482646942, 'gradnorm/compressor.query': 0.070976479981143, 'gradnorm/compressor.layer0': 1.9239937493149384, 'gradnorm/compressor.layer1': 35.420291442878295, 'gradnorm/compressor.layer2': 8.584335663673754, 'gradnorm/compressor.layer3': 4.5479206573892865, 'gradnorm/compressor.layer4': 1.9171626928041166, 'gradnorm/compressor.layer5': 1.0883751822502463, 'gradnorm/compressor.layer6': 0.6155537578586734, 'gradnorm/compressor.layer7': 0.7521973844860543, 'gradnorm/decoder.layer0': 1.0482895247843136, 'gradnorm/decoder.layer1': 0.7981331560618804, 'gradnorm/decoder.layer2': 0.40580648274016606, 'gradnorm/max_param': 24.974353790283203, 'epoch': 4.85}

 49%|████▊     | 9260/19090 [4:30:08<4:30:11,  1.65s/it]
 49%|████▊     | 9261/19090 [4:30:11<5:34:07,  2.04s/it]
 49%|████▊     | 9262/19090 [4:30:12<4:57:14,  1.81s/it]
 49%|████▊     | 9263/19090 [4:30:13<4:31:31,  1.66s/it]
 49%|████▊     | 9264/19090 [4:30:16<5:17:32,  1.94s/it]
 49%|████▊     | 9265/19090 [4:30:17<4:45:36,  1.74s/it]
 49%|████▊     | 9266/19090 [4:30:19<4:22:56,  1.61s/it]
 49%|████▊     | 9267/19090 [4:30:21<5:11:08,  1.90s/it]
 49%|████▊     | 9268/19090 [4:30:23<4:40:52,  1.72s/it]
 49%|████▊     | 9269/19090 [4:30:24<4:19:48,  1.59s/it]
 49%|████▊     | 9270/19090 [4:30:27<5:28:13,  2.01s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 37.84507751464844, 'learning_rate': 3.057766580531031e-05, 'recon_loss': 0.0014530347194522618, 'recon_static': 0.000494364253245294, 'recon_motion': 0.0009586705171386711, 'recon_norm': 1.2696230083703994, 'recon_norm_std': 0.0475705562159419, 'target_norm': 1.6191672414541245, 'recon_cos': 0.659351247549057, 'gradnorm/compressor.query': 0.07245333559718488, 'gradnorm/compressor.layer0': 1.856544006389103, 'gradnorm/compressor.layer1': 36.15699950606037, 'gradnorm/compressor.layer2': 8.959112325702295, 'gradnorm/compressor.layer3': 5.741586915870908, 'gradnorm/compressor.layer4': 1.998670553378363, 'gradnorm/compressor.layer5': 1.1841969479512986, 'gradnorm/compressor.layer6': 0.6056939967390295, 'gradnorm/compressor.layer7': 0.7716686559847847, 'gradnorm/decoder.layer0': 1.0760752224313825, 'gradnorm/decoder.layer1': 0.7567081072742402, 'gradnorm/decoder.layer2': 0.38426286023000433, 'gradnorm/max_param': 25.49627113342285, 'epoch': 4.86}

 49%|████▊     | 9270/19090 [4:30:27<5:28:13,  2.01s/it]
 49%|████▊     | 9271/19090 [4:30:28<4:52:52,  1.79s/it]
 49%|████▊     | 9272/19090 [4:30:29<4:28:04,  1.64s/it]
 49%|████▊     | 9273/19090 [4:30:32<5:22:59,  1.97s/it]
 49%|████▊     | 9274/19090 [4:30:33<4:49:00,  1.77s/it]
 49%|████▊     | 9275/19090 [4:30:35<4:25:14,  1.62s/it]
 49%|████▊     | 9276/19090 [4:30:37<4:57:47,  1.82s/it]
 49%|████▊     | 9277/19090 [4:30:38<4:31:39,  1.66s/it]
 49%|████▊     | 9278/19090 [4:30:40<4:13:05,  1.55s/it]
 49%|████▊     | 9279/19090 [4:30:42<5:16:24,  1.94s/it]
 49%|████▊     | 9280/19090 [4:30:44<4:44:38,  1.74s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 42.481048583984375, 'learning_rate': 3.053309556898412e-05, 'recon_loss': 0.0014427293528569863, 'recon_static': 0.00048663024135748857, 'recon_motion': 0.0009560991311445832, 'recon_norm': 1.2645671218633652, 'recon_norm_std': 0.047085173707455394, 'target_norm': 1.6178347706794738, 'recon_cos': 0.6611499682068824, 'gradnorm/compressor.query': 0.07653115589460052, 'gradnorm/compressor.layer0': 1.8487844107767482, 'gradnorm/compressor.layer1': 40.747828139017116, 'gradnorm/compressor.layer2': 10.067241428295336, 'gradnorm/compressor.layer3': 5.43317162130713, 'gradnorm/compressor.layer4': 2.199826090614659, 'gradnorm/compressor.layer5': 1.2439220002826425, 'gradnorm/compressor.layer6': 0.7309394176860833, 'gradnorm/compressor.layer7': 0.8325775573734261, 'gradnorm/decoder.layer0': 1.1951296298692737, 'gradnorm/decoder.layer1': 0.8697039203964673, 'gradnorm/decoder.layer2': 0.40340402394823505, 'gradnorm/max_param': 28.740245819091797, 'epoch': 4.86}

 49%|████▊     | 9280/19090 [4:30:44<4:44:38,  1.74s/it]
 49%|████▊     | 9281/19090 [4:30:45<4:22:12,  1.60s/it]
 49%|████▊     | 9282/19090 [4:30:48<5:26:04,  1.99s/it]
 49%|████▊     | 9283/19090 [4:30:49<4:51:08,  1.78s/it]
 49%|████▊     | 9284/19090 [4:30:50<4:26:49,  1.63s/it]
 49%|████▊     | 9285/19090 [4:30:53<5:11:34,  1.91s/it]
 49%|████▊     | 9286/19090 [4:30:54<4:41:11,  1.72s/it]
 49%|████▊     | 9287/19090 [4:30:56<4:19:31,  1.59s/it]
 49%|████▊     | 9288/19090 [4:30:58<5:10:00,  1.90s/it]
 49%|████▊     | 9289/19090 [4:30:59<4:40:08,  1.71s/it]
 49%|████▊     | 9290/19090 [4:31:01<4:19:07,  1.59s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 45.51018524169922, 'learning_rate': 3.048850683269236e-05, 'recon_loss': 0.0014436232682783156, 'recon_static': 0.00047376252405229025, 'recon_motion': 0.0009698606983874925, 'recon_norm': 1.2612099260091783, 'recon_norm_std': 0.04685127846896649, 'target_norm': 1.6199584871530532, 'recon_cos': 0.6618795543909073, 'gradnorm/compressor.query': 0.07503968395946864, 'gradnorm/compressor.layer0': 1.9321194663695667, 'gradnorm/compressor.layer1': 43.79631616437977, 'gradnorm/compressor.layer2': 10.389936610772892, 'gradnorm/compressor.layer3': 5.660842969685823, 'gradnorm/compressor.layer4': 2.0792178385316817, 'gradnorm/compressor.layer5': 1.2605379797613063, 'gradnorm/compressor.layer6': 0.6563646125618086, 'gradnorm/compressor.layer7': 0.7837268476701772, 'gradnorm/decoder.layer0': 1.2786101150444358, 'gradnorm/decoder.layer1': 0.7574517876166507, 'gradnorm/decoder.layer2': 0.39953889271837395, 'gradnorm/max_param': 30.891530990600586, 'epoch': 4.87}

 49%|████▊     | 9290/19090 [4:31:01<4:19:07,  1.59s/it]
 49%|████▊     | 9291/19090 [4:31:03<5:09:35,  1.90s/it]
 49%|████▊     | 9292/19090 [4:31:05<4:39:23,  1.71s/it]
 49%|████▊     | 9293/19090 [4:31:06<4:18:13,  1.58s/it]
 49%|████▊     | 9294/19090 [4:31:09<5:13:43,  1.92s/it]
 49%|████▊     | 9295/19090 [4:31:10<4:42:34,  1.73s/it]
 49%|████▊     | 9296/19090 [4:31:11<4:20:42,  1.60s/it]
 49%|████▊     | 9297/19090 [4:31:14<5:09:55,  1.90s/it]
 49%|████▊     | 9298/19090 [4:31:15<4:39:56,  1.72s/it]
 49%|████▊     | 9299/19090 [4:31:16<4:18:49,  1.59s/it]
 49%|████▊     | 9300/19090 [4:31:19<5:04:35,  1.87s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 40.065006256103516, 'learning_rate': 3.0443899745517945e-05, 'recon_loss': 0.0014038362714927644, 'recon_static': 0.00043852385788341054, 'recon_motion': 0.000965312373591587, 'recon_norm': 1.2419755458831787, 'recon_norm_std': 0.0474423554725945, 'target_norm': 1.6194543600082398, 'recon_cos': 0.6710378423333168, 'gradnorm/compressor.query': 0.0773132953024675, 'gradnorm/compressor.layer0': 1.611147746769416, 'gradnorm/compressor.layer1': 38.35211552998087, 'gradnorm/compressor.layer2': 9.675774153459246, 'gradnorm/compressor.layer3': 5.449643485631276, 'gradnorm/compressor.layer4': 1.849763728854266, 'gradnorm/compressor.layer5': 1.2929656152600784, 'gradnorm/compressor.layer6': 0.6644654566065118, 'gradnorm/compressor.layer7': 0.8465172634944335, 'gradnorm/decoder.layer0': 1.171799415693969, 'gradnorm/decoder.layer1': 0.7658341569726588, 'gradnorm/decoder.layer2': 0.4332886730924247, 'gradnorm/max_param': 27.05047607421875, 'epoch': 4.87}

 49%|████▊     | 9300/19090 [4:31:19<5:04:35,  1.87s/it]
 49%|████▊     | 9301/19090 [4:31:20<4:43:03,  1.73s/it]
 49%|████▊     | 9302/19090 [4:31:22<4:21:27,  1.60s/it]
 49%|████▊     | 9303/19090 [4:31:24<4:56:16,  1.82s/it]
 49%|████▊     | 9304/19090 [4:31:26<4:45:46,  1.75s/it]
 49%|████▊     | 9305/19090 [4:31:27<4:23:01,  1.61s/it]
 49%|████▊     | 9306/19090 [4:31:29<4:57:38,  1.83s/it]
 49%|████▉     | 9307/19090 [4:31:31<5:00:23,  1.84s/it]
 49%|████▉     | 9308/19090 [4:31:32<4:32:53,  1.67s/it]
 49%|████▉     | 9309/19090 [4:31:34<4:53:58,  1.80s/it]
 49%|████▉     | 9310/19090 [4:31:37<5:10:01,  1.90s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 38.479793548583984, 'learning_rate': 3.039927445660517e-05, 'recon_loss': 0.0014146783942123875, 'recon_static': 0.0004528555095021147, 'recon_motion': 0.0009618229349143803, 'recon_norm': 1.247833278775215, 'recon_norm_std': 0.04864882584661245, 'target_norm': 1.6225437730550767, 'recon_cos': 0.6698310360312462, 'gradnorm/compressor.query': 0.05772605204630794, 'gradnorm/compressor.layer0': 1.5566946903350176, 'gradnorm/compressor.layer1': 36.956168941755344, 'gradnorm/compressor.layer2': 8.79564163478584, 'gradnorm/compressor.layer3': 5.158175022405461, 'gradnorm/compressor.layer4': 1.9074741872738323, 'gradnorm/compressor.layer5': 1.2791000098258203, 'gradnorm/compressor.layer6': 0.6636946837584906, 'gradnorm/compressor.layer7': 0.9043773508206113, 'gradnorm/decoder.layer0': 1.1397058101487545, 'gradnorm/decoder.layer1': 0.7503911031774122, 'gradnorm/decoder.layer2': 0.37954740255133157, 'gradnorm/max_param': 26.066768646240234, 'epoch': 4.88}

 49%|████▉     | 9310/19090 [4:31:37<5:10:01,  1.90s/it]
 49%|████▉     | 9311/19090 [4:31:38<4:39:47,  1.72s/it]
 49%|████▉     | 9312/19090 [4:31:40<4:45:06,  1.75s/it]
 49%|████▉     | 9313/19090 [4:31:42<5:08:10,  1.89s/it]
 49%|████▉     | 9314/19090 [4:31:43<4:47:19,  1.76s/it]
 49%|████▉     | 9315/19090 [4:31:45<4:53:17,  1.80s/it]
 49%|████▉     | 9316/19090 [4:31:47<5:02:49,  1.86s/it]
 49%|████▉     | 9317/19090 [4:31:49<4:34:32,  1.69s/it]
 49%|████▉     | 9318/19090 [4:31:50<4:41:11,  1.73s/it]
 49%|████▉     | 9319/19090 [4:31:52<5:00:49,  1.85s/it]
 49%|████▉     | 9320/19090 [4:31:54<4:36:48,  1.70s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 37.2083854675293, 'learning_rate': 3.035463111515918e-05, 'recon_loss': 0.001412391898338683, 'recon_static': 0.00046076130092842504, 'recon_motion': 0.0009516306221485138, 'recon_norm': 1.2536737769842148, 'recon_norm_std': 0.04834454944357276, 'target_norm': 1.6207601189613343, 'recon_cos': 0.6700551614165307, 'gradnorm/compressor.query': 0.06559086632852978, 'gradnorm/compressor.layer0': 1.6101861039401313, 'gradnorm/compressor.layer1': 35.28829196496605, 'gradnorm/compressor.layer2': 10.09809711017784, 'gradnorm/compressor.layer3': 5.165828247133448, 'gradnorm/compressor.layer4': 1.8769103808502765, 'gradnorm/compressor.layer5': 1.248451990991169, 'gradnorm/compressor.layer6': 0.6070492450889454, 'gradnorm/compressor.layer7': 0.8373272119744302, 'gradnorm/decoder.layer0': 1.0586395816805438, 'gradnorm/decoder.layer1': 0.7231146518425139, 'gradnorm/decoder.layer2': 0.38898773394174657, 'gradnorm/max_param': 24.88722038269043, 'epoch': 4.88}

 49%|████▉     | 9320/19090 [4:31:54<4:36:48,  1.70s/it]
 49%|████▉     | 9321/19090 [4:31:56<4:44:35,  1.75s/it]
 49%|████▉     | 9322/19090 [4:31:58<5:06:07,  1.88s/it]
 49%|████▉     | 9323/19090 [4:31:59<4:36:40,  1.70s/it]
 49%|████▉     | 9324/19090 [4:32:01<4:42:53,  1.74s/it]
 49%|████▉     | 9325/19090 [4:32:03<5:06:17,  1.88s/it]
 49%|████▉     | 9326/19090 [4:32:04<4:36:50,  1.70s/it]
 49%|████▉     | 9327/19090 [4:32:06<4:36:51,  1.70s/it]
 49%|████▉     | 9328/19090 [4:32:08<5:06:48,  1.89s/it]
 49%|████▉     | 9329/19090 [4:32:10<4:42:55,  1.74s/it]
 49%|████▉     | 9330/19090 [4:32:12<4:41:08,  1.73s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 40.25764465332031, 'learning_rate': 3.0309969870445464e-05, 'recon_loss': 0.001444006391102448, 'recon_static': 0.00048657516817911527, 'recon_motion': 0.0009574312483891845, 'recon_norm': 1.2634712129831314, 'recon_norm_std': 0.047085577994585036, 'target_norm': 1.6203164041042328, 'recon_cos': 0.6618603706359864, 'gradnorm/compressor.query': 0.06966688530124651, 'gradnorm/compressor.layer0': 1.802574539124791, 'gradnorm/compressor.layer1': 38.74262152534661, 'gradnorm/compressor.layer2': 9.146827823268328, 'gradnorm/compressor.layer3': 4.864961118003668, 'gradnorm/compressor.layer4': 2.0574285405335115, 'gradnorm/compressor.layer5': 1.1954906555250924, 'gradnorm/compressor.layer6': 0.7174961514652194, 'gradnorm/compressor.layer7': 0.8460726382227296, 'gradnorm/decoder.layer0': 1.150333163472991, 'gradnorm/decoder.layer1': 0.8317995778664474, 'gradnorm/decoder.layer2': 0.43925817615973994, 'gradnorm/max_param': 27.324729919433594, 'epoch': 4.89}

 49%|████▉     | 9330/19090 [4:32:12<4:41:08,  1.73s/it]
 49%|████▉     | 9331/19090 [4:32:14<5:21:57,  1.98s/it]
 49%|████▉     | 9332/19090 [4:32:15<4:47:45,  1.77s/it]
 49%|████▉     | 9333/19090 [4:32:17<4:37:45,  1.71s/it]
 49%|████▉     | 9334/19090 [4:32:20<5:17:56,  1.96s/it]
 49%|████▉     | 9335/19090 [4:32:21<4:44:51,  1.75s/it]
 49%|████▉     | 9336/19090 [4:32:22<4:32:42,  1.68s/it]
 49%|████▉     | 9337/19090 [4:32:25<5:16:03,  1.94s/it]
 49%|████▉     | 9338/19090 [4:32:26<4:43:34,  1.74s/it]
 49%|████▉     | 9339/19090 [4:32:28<4:34:37,  1.69s/it]
 49%|████▉     | 9340/19090 [4:32:30<5:17:02,  1.95s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 40.36994934082031, 'learning_rate': 3.02652908717894e-05, 'recon_loss': 0.0014340474299388006, 'recon_static': 0.0004851966725254897, 'recon_motion': 0.0009488507435889915, 'recon_norm': 1.2617223680019378, 'recon_norm_std': 0.04831884074956179, 'target_norm': 1.618345284461975, 'recon_cos': 0.6636881127953529, 'gradnorm/compressor.query': 0.06264807995661947, 'gradnorm/compressor.layer0': 1.7149854875178538, 'gradnorm/compressor.layer1': 38.91454950382874, 'gradnorm/compressor.layer2': 8.854517552024026, 'gradnorm/compressor.layer3': 5.169325462328101, 'gradnorm/compressor.layer4': 1.890231320086425, 'gradnorm/compressor.layer5': 1.0695141230146041, 'gradnorm/compressor.layer6': 0.5893521565594712, 'gradnorm/compressor.layer7': 0.768928407885843, 'gradnorm/decoder.layer0': 0.9760264336785912, 'gradnorm/decoder.layer1': 0.7525744360800983, 'gradnorm/decoder.layer2': 0.38254944852575584, 'gradnorm/max_param': 27.448759078979492, 'epoch': 4.89}

 49%|████▉     | 9340/19090 [4:32:30<5:17:02,  1.95s/it]
 49%|████▉     | 9341/19090 [4:32:32<4:44:27,  1.75s/it]
 49%|████▉     | 9342/19090 [4:32:33<4:27:37,  1.65s/it]
 49%|████▉     | 9343/19090 [4:32:36<5:29:08,  2.03s/it]
 49%|████▉     | 9344/19090 [4:32:37<4:53:14,  1.81s/it]
 49%|████▉     | 9345/19090 [4:32:38<4:28:10,  1.65s/it]
 49%|████▉     | 9346/19090 [4:32:41<5:22:29,  1.99s/it]
 49%|████▉     | 9347/19090 [4:32:43<4:48:28,  1.78s/it]
 49%|████▉     | 9348/19090 [4:32:44<4:24:21,  1.63s/it]
 49%|████▉     | 9349/19090 [4:32:46<5:08:45,  1.90s/it]
 49%|████▉     | 9350/19090 [4:32:48<4:38:48,  1.72s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 37.041168212890625, 'learning_rate': 3.0220594268575693e-05, 'recon_loss': 0.0014627689582994209, 'recon_static': 0.0004950460002874024, 'recon_motion': 0.0009677229711087421, 'recon_norm': 1.266207966208458, 'recon_norm_std': 0.04769291253760457, 'target_norm': 1.6225537836551667, 'recon_cos': 0.657889974117279, 'gradnorm/compressor.query': 0.0813912218906473, 'gradnorm/compressor.layer0': 1.7868220530855052, 'gradnorm/compressor.layer1': 35.36485678895257, 'gradnorm/compressor.layer2': 9.021944630971552, 'gradnorm/compressor.layer3': 5.318069769531836, 'gradnorm/compressor.layer4': 1.9826218076713542, 'gradnorm/compressor.layer5': 1.2086937081958518, 'gradnorm/compressor.layer6': 0.673105665263941, 'gradnorm/compressor.layer7': 0.7980721401074811, 'gradnorm/decoder.layer0': 1.1021724662480117, 'gradnorm/decoder.layer1': 0.8031900416654505, 'gradnorm/decoder.layer2': 0.3957711932278451, 'gradnorm/max_param': 24.938962936401367, 'epoch': 4.9}

 49%|████▉     | 9350/19090 [4:32:48<4:38:48,  1.72s/it]
 49%|████▉     | 9351/19090 [4:32:49<4:17:36,  1.59s/it]
 49%|████▉     | 9352/19090 [4:32:52<5:19:07,  1.97s/it]
 49%|████▉     | 9353/19090 [4:32:53<4:46:23,  1.76s/it]
 49%|████▉     | 9354/19090 [4:32:54<4:23:30,  1.62s/it]
 49%|████▉     | 9355/19090 [4:32:57<5:15:59,  1.95s/it]
 49%|████▉     | 9356/19090 [4:32:58<4:44:02,  1.75s/it]
 49%|████▉     | 9357/19090 [4:33:00<4:21:37,  1.61s/it]
 49%|████▉     | 9358/19090 [4:33:02<5:15:42,  1.95s/it]
 49%|████▉     | 9359/19090 [4:33:04<4:44:12,  1.75s/it]
 49%|████▉     | 9360/19090 [4:33:05<4:25:19,  1.64s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 38.428680419921875, 'learning_rate': 3.0175880210247915e-05, 'recon_loss': 0.0014466938271652907, 'recon_static': 0.0004816927612409927, 'recon_motion': 0.0009650011357734911, 'recon_norm': 1.2653556287288665, 'recon_norm_std': 0.046488668862730266, 'target_norm': 1.6198318719863891, 'recon_cos': 0.6614650070667267, 'gradnorm/compressor.query': 0.06023916956142242, 'gradnorm/compressor.layer0': 1.5471405417894601, 'gradnorm/compressor.layer1': 36.734340107916225, 'gradnorm/compressor.layer2': 9.50308458544822, 'gradnorm/compressor.layer3': 5.086383383992508, 'gradnorm/compressor.layer4': 1.9976429162682574, 'gradnorm/compressor.layer5': 1.2512897036995605, 'gradnorm/compressor.layer6': 0.7543415927521883, 'gradnorm/compressor.layer7': 0.7745900669663642, 'gradnorm/decoder.layer0': 1.1519286879117112, 'gradnorm/decoder.layer1': 0.7440704552915296, 'gradnorm/decoder.layer2': 0.4104366070926597, 'gradnorm/max_param': 25.90949821472168, 'epoch': 4.9}

 49%|████▉     | 9360/19090 [4:33:05<4:25:19,  1.64s/it]
 49%|████▉     | 9361/19090 [4:33:08<5:16:58,  1.95s/it]
 49%|████▉     | 9362/19090 [4:33:09<4:44:13,  1.75s/it]
 49%|████▉     | 9363/19090 [4:33:10<4:28:24,  1.66s/it]
 49%|████▉     | 9364/19090 [4:33:13<5:16:18,  1.95s/it]
 49%|████▉     | 9365/19090 [4:33:14<4:43:50,  1.75s/it]
 49%|████▉     | 9366/19090 [4:33:16<4:28:25,  1.66s/it]
 49%|████▉     | 9367/19090 [4:33:18<5:05:53,  1.89s/it]
 49%|████▉     | 9368/19090 [4:33:20<4:36:47,  1.71s/it]
 49%|████▉     | 9369/19090 [4:33:21<4:29:43,  1.66s/it]
 49%|████▉     | 9370/19090 [4:33:24<5:15:16,  1.95s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 38.39826583862305, 'learning_rate': 3.0131148846308034e-05, 'recon_loss': 0.0014281346957432106, 'recon_static': 0.0004578763997415081, 'recon_motion': 0.0009702582785394043, 'recon_norm': 1.2483261823654175, 'recon_norm_std': 0.04664646033197641, 'target_norm': 1.6212292790412903, 'recon_cos': 0.6655048668384552, 'gradnorm/compressor.query': 0.06586931693786596, 'gradnorm/compressor.layer0': 1.721126716426182, 'gradnorm/compressor.layer1': 36.65370735267188, 'gradnorm/compressor.layer2': 9.619498922578263, 'gradnorm/compressor.layer3': 5.168967360293286, 'gradnorm/compressor.layer4': 1.8616818135031543, 'gradnorm/compressor.layer5': 1.3247847796134782, 'gradnorm/compressor.layer6': 0.7259415875441726, 'gradnorm/compressor.layer7': 0.8521799771096101, 'gradnorm/decoder.layer0': 1.2059509974087856, 'gradnorm/decoder.layer1': 0.7689462752332564, 'gradnorm/decoder.layer2': 0.4393486916111993, 'gradnorm/max_param': 25.849973678588867, 'epoch': 4.91}

 49%|████▉     | 9370/19090 [4:33:24<5:15:16,  1.95s/it]
 49%|████▉     | 9371/19090 [4:33:25<4:42:59,  1.75s/it]
 49%|████▉     | 9372/19090 [4:33:26<4:31:30,  1.68s/it]
 49%|████▉     | 9373/19090 [4:33:29<5:12:02,  1.93s/it]
 49%|████▉     | 9374/19090 [4:33:30<4:40:33,  1.73s/it]
 49%|████▉     | 9375/19090 [4:33:32<4:47:48,  1.78s/it]
 49%|████▉     | 9376/19090 [4:33:35<5:26:33,  2.02s/it]
 49%|████▉     | 9377/19090 [4:33:36<4:50:56,  1.80s/it]
 49%|████▉     | 9378/19090 [4:33:38<4:44:03,  1.75s/it]
 49%|████▉     | 9379/19090 [4:33:40<5:14:21,  1.94s/it]
 49%|████▉     | 9380/19090 [4:33:41<4:42:26,  1.75s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 47.07691192626953, 'learning_rate': 3.008640032631585e-05, 'recon_loss': 0.0014149523485684767, 'recon_static': 0.00045482499699573965, 'recon_motion': 0.0009601273690350354, 'recon_norm': 1.2507689833641051, 'recon_norm_std': 0.0469293643720448, 'target_norm': 1.6193478643894195, 'recon_cos': 0.6686603486537933, 'gradnorm/compressor.query': 0.0811669594723853, 'gradnorm/compressor.layer0': 1.9170045623924, 'gradnorm/compressor.layer1': 45.46772443001464, 'gradnorm/compressor.layer2': 10.19453870387305, 'gradnorm/compressor.layer3': 5.58139327015949, 'gradnorm/compressor.layer4': 2.193401985060287, 'gradnorm/compressor.layer5': 1.3363463908187398, 'gradnorm/compressor.layer6': 0.6800504789479327, 'gradnorm/compressor.layer7': 0.8771989643474367, 'gradnorm/decoder.layer0': 1.235419543661249, 'gradnorm/decoder.layer1': 0.8002252276589648, 'gradnorm/decoder.layer2': 0.4049409904582721, 'gradnorm/max_param': 32.07040023803711, 'epoch': 4.91}

 49%|████▉     | 9380/19090 [4:33:41<4:42:26,  1.75s/it]
 49%|████▉     | 9381/19090 [4:33:43<4:38:03,  1.72s/it]
 49%|████▉     | 9382/19090 [4:33:45<4:53:43,  1.82s/it]
 49%|████▉     | 9383/19090 [4:33:46<4:28:10,  1.66s/it]
 49%|████▉     | 9384/19090 [4:33:48<4:51:54,  1.80s/it]
 49%|████▉     | 9385/19090 [4:33:50<4:55:22,  1.83s/it]
 49%|████▉     | 9386/19090 [4:33:52<4:29:11,  1.66s/it]
 49%|████▉     | 9387/19090 [4:33:54<4:41:50,  1.74s/it]
 49%|████▉     | 9388/19090 [4:33:56<5:19:14,  1.97s/it]
 49%|████▉     | 9389/19090 [4:33:57<4:45:51,  1.77s/it]
 49%|████▉     | 9390/19090 [4:33:59<4:34:18,  1.70s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 40.59842300415039, 'learning_rate': 3.0041634799888513e-05, 'recon_loss': 0.0014198986202245579, 'recon_static': 0.0004444137040991336, 'recon_motion': 0.0009754849219461903, 'recon_norm': 1.2454220652580261, 'recon_norm_std': 0.047461294941604136, 'target_norm': 1.6232101649045945, 'recon_cos': 0.6689562484622001, 'gradnorm/compressor.query': 0.05733228818385306, 'gradnorm/compressor.layer0': 1.5588273946458016, 'gradnorm/compressor.layer1': 38.725562480784255, 'gradnorm/compressor.layer2': 10.498874785668068, 'gradnorm/compressor.layer3': 5.136555368124117, 'gradnorm/compressor.layer4': 1.9478907523921882, 'gradnorm/compressor.layer5': 1.3419927595656709, 'gradnorm/compressor.layer6': 0.7548559265481183, 'gradnorm/compressor.layer7': 0.9221927612423483, 'gradnorm/decoder.layer0': 1.2787144390691614, 'gradnorm/decoder.layer1': 0.8337899903939884, 'gradnorm/decoder.layer2': 0.4156121101388351, 'gradnorm/max_param': 27.313772201538086, 'epoch': 4.92}

 49%|████▉     | 9390/19090 [4:33:59<4:34:18,  1.70s/it]
 49%|████▉     | 9391/19090 [4:34:01<5:09:29,  1.91s/it]
 49%|████▉     | 9392/19090 [4:34:03<4:38:57,  1.73s/it]
 49%|████▉     | 9393/19090 [4:34:04<4:26:39,  1.65s/it]
 49%|████▉     | 9394/19090 [4:34:06<4:57:36,  1.84s/it]
 49%|████▉     | 9395/19090 [4:34:08<4:30:56,  1.68s/it]
 49%|████▉     | 9396/19090 [4:34:09<4:38:13,  1.72s/it]
 49%|████▉     | 9397/19090 [4:34:12<5:10:45,  1.92s/it]
 49%|████▉     | 9398/19090 [4:34:13<4:39:51,  1.73s/it]
 49%|████▉     | 9399/19090 [4:34:15<4:37:33,  1.72s/it]
 49%|████▉     | 9400/19090 [4:34:17<4:55:12,  1.83s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 42.52909469604492, 'learning_rate': 2.9996852416700067e-05, 'recon_loss': 0.0014060507819522172, 'recon_static': 0.00044842565766884946, 'recon_motion': 0.0009576251410180703, 'recon_norm': 1.2467895179986954, 'recon_norm_std': 0.04788442347198725, 'target_norm': 1.6183484762907028, 'recon_cos': 0.6701712265610695, 'gradnorm/compressor.query': 0.08060856207552115, 'gradnorm/compressor.layer0': 1.7723379213293955, 'gradnorm/compressor.layer1': 40.840880733188165, 'gradnorm/compressor.layer2': 10.189139953796241, 'gradnorm/compressor.layer3': 4.964476661107065, 'gradnorm/compressor.layer4': 1.918103930027283, 'gradnorm/compressor.layer5': 1.352537696369459, 'gradnorm/compressor.layer6': 0.6893551894702229, 'gradnorm/compressor.layer7': 0.8428909025404832, 'gradnorm/decoder.layer0': 1.2928735519352386, 'gradnorm/decoder.layer1': 0.7627432950955759, 'gradnorm/decoder.layer2': 0.43198251777062313, 'gradnorm/max_param': 28.804988861083984, 'epoch': 4.92}

 49%|████▉     | 9400/19090 [4:34:17<4:55:12,  1.83s/it]
 49%|████▉     | 9401/19090 [4:34:18<4:28:48,  1.66s/it]
 49%|████▉     | 9402/19090 [4:34:20<4:51:26,  1.80s/it]
 49%|████▉     | 9403/19090 [4:34:22<4:54:32,  1.82s/it]
 49%|████▉     | 9404/19090 [4:34:24<4:32:54,  1.69s/it]
 49%|████▉     | 9405/19090 [4:34:26<4:50:20,  1.80s/it]
 49%|████▉     | 9406/19090 [4:34:28<5:00:37,  1.86s/it]
 49%|████▉     | 9407/19090 [4:34:29<4:32:23,  1.69s/it]
 49%|████▉     | 9408/19090 [4:34:31<4:52:06,  1.81s/it]
 49%|████▉     | 9409/19090 [4:34:33<5:09:56,  1.92s/it]
 49%|████▉     | 9410/19090 [4:34:34<4:39:20,  1.73s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 41.76736831665039, 'learning_rate': 2.9952053326480893e-05, 'recon_loss': 0.0014198354299878701, 'recon_static': 0.0004607203052728437, 'recon_motion': 0.000959115159639623, 'recon_norm': 1.2510359853506088, 'recon_norm_std': 0.04795835157856345, 'target_norm': 1.6190640538930894, 'recon_cos': 0.667089319229126, 'gradnorm/compressor.query': 0.06385073056058843, 'gradnorm/compressor.layer0': 1.6156722631864513, 'gradnorm/compressor.layer1': 40.17684296877369, 'gradnorm/compressor.layer2': 9.598313628534898, 'gradnorm/compressor.layer3': 5.161185936487475, 'gradnorm/compressor.layer4': 2.0907799450943165, 'gradnorm/compressor.layer5': 1.1826833616685366, 'gradnorm/compressor.layer6': 0.6716261114398472, 'gradnorm/compressor.layer7': 0.785844112482775, 'gradnorm/decoder.layer0': 1.1891160308778732, 'gradnorm/decoder.layer1': 0.7416079617601711, 'gradnorm/decoder.layer2': 0.3894447999177297, 'gradnorm/max_param': 28.337339401245117, 'epoch': 4.93}

 49%|████▉     | 9410/19090 [4:34:34<4:39:20,  1.73s/it]
 49%|████▉     | 9411/19090 [4:34:37<4:56:29,  1.84s/it]
 49%|████▉     | 9412/19090 [4:34:39<5:24:47,  2.01s/it]
 49%|████▉     | 9413/19090 [4:34:40<4:49:43,  1.80s/it]
 49%|████▉     | 9414/19090 [4:34:42<4:38:43,  1.73s/it]
 49%|████▉     | 9415/19090 [4:34:44<5:07:49,  1.91s/it]
 49%|████▉     | 9416/19090 [4:34:45<4:37:31,  1.72s/it]
 49%|████▉     | 9417/19090 [4:34:47<4:44:38,  1.77s/it]
 49%|████▉     | 9418/19090 [4:34:50<5:16:49,  1.97s/it]
 49%|████▉     | 9419/19090 [4:34:51<4:43:32,  1.76s/it]
 49%|████▉     | 9420/19090 [4:34:53<4:37:58,  1.72s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 44.030181884765625, 'learning_rate': 2.9907237679017235e-05, 'recon_loss': 0.0014203657949110493, 'recon_static': 0.0004527186967607122, 'recon_motion': 0.0009676470828708262, 'recon_norm': 1.2477483987808227, 'recon_norm_std': 0.047539519146084785, 'target_norm': 1.621575203537941, 'recon_cos': 0.6677931889891624, 'gradnorm/compressor.query': 0.08073254707652491, 'gradnorm/compressor.layer0': 1.8777824102090028, 'gradnorm/compressor.layer1': 42.57140285238804, 'gradnorm/compressor.layer2': 9.454470300709193, 'gradnorm/compressor.layer3': 5.02493837992526, 'gradnorm/compressor.layer4': 1.8104933449939704, 'gradnorm/compressor.layer5': 1.1921418467488571, 'gradnorm/compressor.layer6': 0.666068446953688, 'gradnorm/compressor.layer7': 0.8563416090917175, 'gradnorm/decoder.layer0': 1.1907461370084718, 'gradnorm/decoder.layer1': 0.7889979496604087, 'gradnorm/decoder.layer2': 0.4963993422880434, 'gradnorm/max_param': 30.02632713317871, 'epoch': 4.93}

 49%|████▉     | 9420/19090 [4:34:53<4:37:58,  1.72s/it]
 49%|████▉     | 9421/19090 [4:34:55<5:21:16,  1.99s/it]
 49%|████▉     | 9422/19090 [4:34:57<4:47:04,  1.78s/it]
 49%|████▉     | 9423/19090 [4:34:58<4:23:14,  1.63s/it]
 49%|████▉     | 9424/19090 [4:35:01<5:18:04,  1.97s/it]
 49%|████▉     | 9425/19090 [4:35:02<4:44:51,  1.77s/it]
 49%|████▉     | 9426/19090 [4:35:04<4:35:29,  1.71s/it]
 49%|████▉     | 9427/19090 [4:35:06<5:07:37,  1.91s/it]
 49%|████▉     | 9428/19090 [4:35:07<4:37:33,  1.72s/it]
 49%|████▉     | 9429/19090 [4:35:09<4:33:28,  1.70s/it]
 49%|████▉     | 9430/19090 [4:35:11<5:13:54,  1.95s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 43.93170166015625, 'learning_rate': 2.986240562415072e-05, 'recon_loss': 0.001410186811699532, 'recon_static': 0.0004351895637228154, 'recon_motion': 0.0009749972887220792, 'recon_norm': 1.2396434903144837, 'recon_norm_std': 0.04781062277033925, 'target_norm': 1.6225124835968017, 'recon_cos': 0.6707132399082184, 'gradnorm/compressor.query': 0.06634344258261893, 'gradnorm/compressor.layer0': 1.696357088805803, 'gradnorm/compressor.layer1': 42.087520418765465, 'gradnorm/compressor.layer2': 10.781660785312752, 'gradnorm/compressor.layer3': 5.461253057484678, 'gradnorm/compressor.layer4': 2.06547470584804, 'gradnorm/compressor.layer5': 1.3070239772012848, 'gradnorm/compressor.layer6': 0.7369368488217333, 'gradnorm/compressor.layer7': 0.8690397739833927, 'gradnorm/decoder.layer0': 1.223356588716024, 'gradnorm/decoder.layer1': 0.8489333697122157, 'gradnorm/decoder.layer2': 0.4397767125973586, 'gradnorm/max_param': 29.687135696411133, 'epoch': 4.94}

 49%|████▉     | 9430/19090 [4:35:11<5:13:54,  1.95s/it]
 49%|████▉     | 9431/19090 [4:35:13<4:41:43,  1.75s/it]
 49%|████▉     | 9432/19090 [4:35:14<4:36:00,  1.71s/it]
 49%|████▉     | 9433/19090 [4:35:17<5:14:00,  1.95s/it]
 49%|████▉     | 9434/19090 [4:35:18<4:41:41,  1.75s/it]
 49%|████▉     | 9435/19090 [4:35:19<4:23:41,  1.64s/it]
 49%|████▉     | 9436/19090 [4:35:22<5:17:35,  1.97s/it]
 49%|████▉     | 9437/19090 [4:35:23<4:43:56,  1.76s/it]
 49%|████▉     | 9438/19090 [4:35:25<4:20:57,  1.62s/it]
 49%|████▉     | 9439/19090 [4:35:28<5:26:58,  2.03s/it]
 49%|████▉     | 9440/19090 [4:35:29<4:50:54,  1.81s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 44.78761291503906, 'learning_rate': 2.9817557311777784e-05, 'recon_loss': 0.0014007583813508973, 'recon_static': 0.0004382370323583018, 'recon_motion': 0.0009625213497201912, 'recon_norm': 1.242149257659912, 'recon_norm_std': 0.046959247812628745, 'target_norm': 1.6206407964229583, 'recon_cos': 0.6726779863238335, 'gradnorm/compressor.query': 0.06333727225895477, 'gradnorm/compressor.layer0': 1.7575691563333813, 'gradnorm/compressor.layer1': 42.97096727233764, 'gradnorm/compressor.layer2': 10.541044395284528, 'gradnorm/compressor.layer3': 5.934415959944545, 'gradnorm/compressor.layer4': 2.016466659947526, 'gradnorm/compressor.layer5': 1.3011187782780185, 'gradnorm/compressor.layer6': 0.7927206243162846, 'gradnorm/compressor.layer7': 1.0260750366567835, 'gradnorm/decoder.layer0': 1.2572599346430582, 'gradnorm/decoder.layer1': 0.8642554331601252, 'gradnorm/decoder.layer2': 0.48467167105981435, 'gradnorm/max_param': 30.307422637939453, 'epoch': 4.94}

 49%|████▉     | 9440/19090 [4:35:29<4:50:54,  1.81s/it]
 49%|████▉     | 9441/19090 [4:35:30<4:26:03,  1.65s/it]
 49%|████▉     | 9442/19090 [4:35:33<5:01:03,  1.87s/it]
 49%|████▉     | 9443/19090 [4:35:34<4:32:39,  1.70s/it]
 49%|████▉     | 9444/19090 [4:35:35<4:12:59,  1.57s/it]
 49%|████▉     | 9445/19090 [4:35:38<5:25:48,  2.03s/it]
 49%|████▉     | 9446/19090 [4:35:40<4:49:47,  1.80s/it]
 49%|████▉     | 9447/19090 [4:35:41<4:24:34,  1.65s/it]
 49%|████▉     | 9448/19090 [4:35:44<5:28:32,  2.04s/it]
 49%|████▉     | 9449/19090 [4:35:45<4:51:53,  1.82s/it]
 50%|████▉     | 9450/19090 [4:35:46<4:26:26,  1.66s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 46.78269958496094, 'learning_rate': 2.977269289184927e-05, 'recon_loss': 0.0014009589416673408, 'recon_static': 0.00043540002225199713, 'recon_motion': 0.0009655589325120673, 'recon_norm': 1.2393978416919709, 'recon_norm_std': 0.04675781885161996, 'target_norm': 1.6195035874843597, 'recon_cos': 0.6718200027942658, 'gradnorm/compressor.query': 0.06612241264966436, 'gradnorm/compressor.layer0': 1.8028262890893265, 'gradnorm/compressor.layer1': 44.99229806050837, 'gradnorm/compressor.layer2': 10.572428173688056, 'gradnorm/compressor.layer3': 6.120687718365637, 'gradnorm/compressor.layer4': 2.349050504535325, 'gradnorm/compressor.layer5': 1.4606476358496874, 'gradnorm/compressor.layer6': 0.7689184081913408, 'gradnorm/compressor.layer7': 0.8895829780691065, 'gradnorm/decoder.layer0': 1.3445030142940781, 'gradnorm/decoder.layer1': 0.8830440183902297, 'gradnorm/decoder.layer2': 0.44534734957672845, 'gradnorm/max_param': 31.73629379272461, 'epoch': 4.95}

 50%|████▉     | 9450/19090 [4:35:46<4:26:26,  1.66s/it]
 50%|████▉     | 9451/19090 [4:35:49<5:16:24,  1.97s/it]
 50%|████▉     | 9452/19090 [4:35:50<4:43:30,  1.76s/it]
 50%|████▉     | 9453/19090 [4:35:52<4:20:12,  1.62s/it]
 50%|████▉     | 9454/19090 [4:35:54<5:14:02,  1.96s/it]
 50%|████▉     | 9455/19090 [4:35:56<4:41:54,  1.76s/it]
 50%|████▉     | 9456/19090 [4:35:57<4:24:03,  1.64s/it]
 50%|████▉     | 9457/19090 [4:36:00<5:20:16,  1.99s/it]
 50%|████▉     | 9458/19090 [4:36:01<4:46:01,  1.78s/it]
 50%|████▉     | 9459/19090 [4:36:03<4:22:07,  1.63s/it]
 50%|████▉     | 9460/19090 [4:36:05<5:10:12,  1.93s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 50.051300048828125, 'learning_rate': 2.9727812514369857e-05, 'recon_loss': 0.0014111564145423473, 'recon_static': 0.00044143560298834925, 'recon_motion': 0.0009697208224679343, 'recon_norm': 1.2445139974355697, 'recon_norm_std': 0.0474538529291749, 'target_norm': 1.622560828924179, 'recon_cos': 0.670982152223587, 'gradnorm/compressor.query': 0.07382940450195377, 'gradnorm/compressor.layer0': 2.0281022551073775, 'gradnorm/compressor.layer1': 48.01709442919131, 'gradnorm/compressor.layer2': 11.894420283759892, 'gradnorm/compressor.layer3': 6.490686346360158, 'gradnorm/compressor.layer4': 2.358503269053384, 'gradnorm/compressor.layer5': 1.3982250542409576, 'gradnorm/compressor.layer6': 0.7506803367348566, 'gradnorm/compressor.layer7': 0.9780931607752891, 'gradnorm/decoder.layer0': 1.3039851782459584, 'gradnorm/decoder.layer1': 0.9001115875473878, 'gradnorm/decoder.layer2': 0.46711034105720667, 'gradnorm/max_param': 33.86625289916992, 'epoch': 4.96}

 50%|████▉     | 9460/19090 [4:36:05<5:10:12,  1.93s/it]
 50%|████▉     | 9461/19090 [4:36:06<4:38:51,  1.74s/it]
 50%|████▉     | 9462/19090 [4:36:08<4:20:05,  1.62s/it]
 50%|████▉     | 9463/19090 [4:36:11<5:13:42,  1.96s/it]
 50%|████▉     | 9464/19090 [4:36:12<4:41:22,  1.75s/it]
 50%|████▉     | 9465/19090 [4:36:13<4:31:35,  1.69s/it]
 50%|████▉     | 9466/19090 [4:36:16<5:15:31,  1.97s/it]
 50%|████▉     | 9467/19090 [4:36:17<4:42:28,  1.76s/it]
 50%|████▉     | 9468/19090 [4:36:19<4:37:26,  1.73s/it]
 50%|████▉     | 9469/19090 [4:36:21<5:10:20,  1.94s/it]
 50%|████▉     | 9470/19090 [4:36:23<4:39:02,  1.74s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 45.884334564208984, 'learning_rate': 2.968291632939757e-05, 'recon_loss': 0.0014364933507749812, 'recon_static': 0.0004787944817508105, 'recon_motion': 0.0009576988581102341, 'recon_norm': 1.2591495543718338, 'recon_norm_std': 0.04730416825041175, 'target_norm': 1.6214040845632554, 'recon_cos': 0.6639326825737953, 'gradnorm/compressor.query': 0.07789807742174013, 'gradnorm/compressor.layer0': 1.976176951805843, 'gradnorm/compressor.layer1': 44.10145109042615, 'gradnorm/compressor.layer2': 10.655213962878763, 'gradnorm/compressor.layer3': 5.6202954433883345, 'gradnorm/compressor.layer4': 2.2797072955521185, 'gradnorm/compressor.layer5': 1.490668837184611, 'gradnorm/compressor.layer6': 0.7951533100166023, 'gradnorm/compressor.layer7': 0.8894701045128228, 'gradnorm/decoder.layer0': 1.2831567671718036, 'gradnorm/decoder.layer1': 0.858982239183214, 'gradnorm/decoder.layer2': 0.41728500341211294, 'gradnorm/max_param': 31.10504913330078, 'epoch': 4.96}

 50%|████▉     | 9470/19090 [4:36:23<4:39:02,  1.74s/it]
 50%|████▉     | 9471/19090 [4:36:24<4:26:01,  1.66s/it]
 50%|████▉     | 9472/19090 [4:36:27<5:12:01,  1.95s/it]
 50%|████▉     | 9473/19090 [4:36:28<4:40:34,  1.75s/it]
 50%|████▉     | 9474/19090 [4:36:29<4:18:21,  1.61s/it]
 50%|████▉     | 9475/19090 [4:36:32<5:32:03,  2.07s/it]
 50%|████▉     | 9476/19090 [4:36:34<4:53:59,  1.83s/it]
 50%|████▉     | 9477/19090 [4:36:35<4:27:29,  1.67s/it]
 50%|████▉     | 9478/19090 [4:36:38<5:36:56,  2.10s/it]
 50%|████▉     | 9479/19090 [4:36:39<4:58:26,  1.86s/it]
 50%|████▉     | 9480/19090 [4:36:41<4:30:37,  1.69s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 44.46669387817383, 'learning_rate': 2.9638004487043302e-05, 'recon_loss': 0.0014467051194515079, 'recon_static': 0.0004843296490435023, 'recon_motion': 0.0009623754929634742, 'recon_norm': 1.262199369072914, 'recon_norm_std': 0.0470631904900074, 'target_norm': 1.6197859913110733, 'recon_cos': 0.6608547523617745, 'gradnorm/compressor.query': 0.07096105062271403, 'gradnorm/compressor.layer0': 1.9458059887594996, 'gradnorm/compressor.layer1': 42.75385039540664, 'gradnorm/compressor.layer2': 10.026922986156446, 'gradnorm/compressor.layer3': 5.801680472705563, 'gradnorm/compressor.layer4': 2.277935180291114, 'gradnorm/compressor.layer5': 1.4089968662538463, 'gradnorm/compressor.layer6': 0.7439045368856709, 'gradnorm/compressor.layer7': 0.9280611767643501, 'gradnorm/decoder.layer0': 1.3876190271959485, 'gradnorm/decoder.layer1': 0.8551553300113424, 'gradnorm/decoder.layer2': 0.39928307285685416, 'gradnorm/max_param': 30.152355194091797, 'epoch': 4.97}

 50%|████▉     | 9480/19090 [4:36:41<4:30:37,  1.69s/it]
 50%|████▉     | 9481/19090 [4:36:43<5:10:39,  1.94s/it]
 50%|████▉     | 9482/19090 [4:36:45<4:39:05,  1.74s/it]
 50%|████▉     | 9483/19090 [4:36:46<4:17:46,  1.61s/it]
 50%|████▉     | 9484/19090 [4:36:48<5:02:43,  1.89s/it]
 50%|████▉     | 9485/19090 [4:36:50<4:33:57,  1.71s/it]
 50%|████▉     | 9486/19090 [4:36:51<4:14:56,  1.59s/it]
 50%|████▉     | 9487/19090 [4:36:54<5:02:22,  1.89s/it]
 50%|████▉     | 9488/19090 [4:36:55<4:33:25,  1.71s/it]
 50%|████▉     | 9489/19090 [4:36:57<4:33:42,  1.71s/it]
 50%|████▉     | 9490/19090 [4:36:59<5:19:57,  2.00s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 45.79759979248047, 'learning_rate': 2.9593077137470286e-05, 'recon_loss': 0.001459196986979805, 'recon_static': 0.0004905013964162208, 'recon_motion': 0.0009686955585493706, 'recon_norm': 1.2653458267450333, 'recon_norm_std': 0.04731378313153982, 'target_norm': 1.6229428082704545, 'recon_cos': 0.6590496808290481, 'gradnorm/compressor.query': 0.07660452227148552, 'gradnorm/compressor.layer0': 2.0950158690805227, 'gradnorm/compressor.layer1': 43.720351281879445, 'gradnorm/compressor.layer2': 11.525967677735485, 'gradnorm/compressor.layer3': 6.142928994351039, 'gradnorm/compressor.layer4': 2.1590207859680146, 'gradnorm/compressor.layer5': 1.3827818096656066, 'gradnorm/compressor.layer6': 0.822872925251746, 'gradnorm/compressor.layer7': 0.9258843215672967, 'gradnorm/decoder.layer0': 1.3755610361787785, 'gradnorm/decoder.layer1': 0.8798164910016406, 'gradnorm/decoder.layer2': 0.44528616442522767, 'gradnorm/max_param': 30.8338680267334, 'epoch': 4.97}

 50%|████▉     | 9490/19090 [4:36:59<5:19:57,  2.00s/it]
 50%|████▉     | 9491/19090 [4:37:01<4:45:33,  1.78s/it]
 50%|████▉     | 9492/19090 [4:37:02<4:23:09,  1.65s/it]
 50%|████▉     | 9493/19090 [4:37:04<5:10:46,  1.94s/it]
 50%|████▉     | 9494/19090 [4:37:06<4:39:14,  1.75s/it]
 50%|████▉     | 9495/19090 [4:37:07<4:24:07,  1.65s/it]
 50%|████▉     | 9496/19090 [4:37:10<5:27:11,  2.05s/it]
 50%|████▉     | 9497/19090 [4:37:11<4:50:49,  1.82s/it]
 50%|████▉     | 9498/19090 [4:37:13<4:25:04,  1.66s/it]
 50%|████▉     | 9499/19090 [4:37:16<5:19:46,  2.00s/it]
 50%|████▉     | 9500/19090 [4:37:17<4:45:44,  1.79s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 51.89029312133789, 'learning_rate': 2.9548134430893604e-05, 'recon_loss': 0.001446984420181252, 'recon_static': 0.0004784184173331596, 'recon_motion': 0.0009685660101240501, 'recon_norm': 1.258940726518631, 'recon_norm_std': 0.04635355016216636, 'target_norm': 1.622342038154602, 'recon_cos': 0.6619166299700737, 'gradnorm/compressor.query': 0.07506515025126137, 'gradnorm/compressor.layer0': 2.102415911323107, 'gradnorm/compressor.layer1': 50.05510879036849, 'gradnorm/compressor.layer2': 11.490600725344907, 'gradnorm/compressor.layer3': 6.338175913263271, 'gradnorm/compressor.layer4': 2.1591426961793423, 'gradnorm/compressor.layer5': 1.3455280250939299, 'gradnorm/compressor.layer6': 0.7438118275821339, 'gradnorm/compressor.layer7': 0.9213351269111988, 'gradnorm/decoder.layer0': 1.3295139454988578, 'gradnorm/decoder.layer1': 0.7988600596268056, 'gradnorm/decoder.layer2': 0.41982633701525673, 'gradnorm/max_param': 35.30584716796875, 'epoch': 4.98}

 50%|████▉     | 9500/19090 [4:37:17<4:45:44,  1.79s/it]
 50%|████▉     | 9501/19090 [4:37:21<6:24:56,  2.41s/it]
 50%|████▉     | 9502/19090 [4:37:22<5:31:01,  2.07s/it]
 50%|████▉     | 9503/19090 [4:37:23<4:53:13,  1.84s/it]
 50%|████▉     | 9504/19090 [4:37:25<4:26:49,  1.67s/it]
 50%|████▉     | 9505/19090 [4:37:26<4:08:16,  1.55s/it]
 50%|████▉     | 9506/19090 [4:37:27<3:55:07,  1.47s/it]
 50%|████▉     | 9507/19090 [4:37:28<3:46:00,  1.42s/it]
 50%|████▉     | 9508/19090 [4:37:30<3:39:51,  1.38s/it]
 50%|████▉     | 9509/19090 [4:37:31<3:35:12,  1.35s/it]
 50%|████▉     | 9510/19090 [4:37:32<3:32:06,  1.33s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 43.33146667480469, 'learning_rate': 2.950317651757969e-05, 'recon_loss': 0.0014145544380880892, 'recon_static': 0.00046172925431164913, 'recon_motion': 0.0009528252281597816, 'recon_norm': 1.2508097857236862, 'recon_norm_std': 0.047906233742833135, 'target_norm': 1.6200454860925675, 'recon_cos': 0.6687106415629387, 'gradnorm/compressor.query': 0.06985708276916616, 'gradnorm/compressor.layer0': 1.7428123824605344, 'gradnorm/compressor.layer1': 41.70313817206611, 'gradnorm/compressor.layer2': 9.941452296062321, 'gradnorm/compressor.layer3': 5.168029830929193, 'gradnorm/compressor.layer4': 2.0739683873002974, 'gradnorm/compressor.layer5': 1.394366257254132, 'gradnorm/compressor.layer6': 0.7309464305447916, 'gradnorm/compressor.layer7': 0.8657673384769572, 'gradnorm/decoder.layer0': 1.239156467872077, 'gradnorm/decoder.layer1': 0.8028785223446343, 'gradnorm/decoder.layer2': 0.4141104868306935, 'gradnorm/max_param': 29.415220260620117, 'epoch': 4.98}

 50%|████▉     | 9510/19090 [4:37:32<3:32:06,  1.33s/it]
 50%|████▉     | 9511/19090 [4:37:35<4:24:40,  1.66s/it]
 50%|████▉     | 9512/19090 [4:37:36<4:06:28,  1.54s/it]
 50%|████▉     | 9513/19090 [4:37:37<3:53:51,  1.47s/it]
 50%|████▉     | 9514/19090 [4:37:40<5:00:12,  1.88s/it]
 50%|████▉     | 9515/19090 [4:37:41<4:31:24,  1.70s/it]
 50%|████▉     | 9516/19090 [4:37:43<4:11:14,  1.57s/it]
 50%|████▉     | 9517/19090 [4:37:45<5:13:20,  1.96s/it]
 50%|████▉     | 9518/19090 [4:37:47<4:40:42,  1.76s/it]
 50%|████▉     | 9519/19090 [4:37:48<4:17:45,  1.62s/it]
 50%|████▉     | 9520/19090 [4:37:51<5:05:47,  1.92s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 50.66547775268555, 'learning_rate': 2.9458203547845826e-05, 'recon_loss': 0.001415797689696774, 'recon_static': 0.0004528120676695835, 'recon_motion': 0.0009629855849198066, 'recon_norm': 1.2455022543668748, 'recon_norm_std': 0.04763020789250731, 'target_norm': 1.6215200811624526, 'recon_cos': 0.6692488089203834, 'gradnorm/compressor.query': 0.07561396810956005, 'gradnorm/compressor.layer0': 1.902467696207619, 'gradnorm/compressor.layer1': 48.45356336467293, 'gradnorm/compressor.layer2': 12.498820745155571, 'gradnorm/compressor.layer3': 6.772847256578101, 'gradnorm/compressor.layer4': 2.5365284643066865, 'gradnorm/compressor.layer5': 1.583798398766453, 'gradnorm/compressor.layer6': 0.828098188721991, 'gradnorm/compressor.layer7': 1.0596653205712636, 'gradnorm/decoder.layer0': 1.323709485638055, 'gradnorm/decoder.layer1': 0.8764318943471446, 'gradnorm/decoder.layer2': 0.5040699892380477, 'gradnorm/max_param': 34.17665481567383, 'epoch': 4.99}

 50%|████▉     | 9520/19090 [4:37:51<5:05:47,  1.92s/it]
 50%|████▉     | 9521/19090 [4:37:52<4:35:28,  1.73s/it]
 50%|████▉     | 9522/19090 [4:37:53<4:14:01,  1.59s/it]
 50%|████▉     | 9523/19090 [4:37:56<5:16:54,  1.99s/it]
 50%|████▉     | 9524/19090 [4:37:57<4:43:28,  1.78s/it]
 50%|████▉     | 9525/19090 [4:37:59<4:19:59,  1.63s/it]
 50%|████▉     | 9526/19090 [4:38:02<5:25:48,  2.04s/it]
 50%|████▉     | 9527/19090 [4:38:03<4:49:35,  1.82s/it]
 50%|████▉     | 9528/19090 [4:38:04<4:24:11,  1.66s/it]
 50%|████▉     | 9529/19090 [4:38:07<5:20:50,  2.01s/it]
 50%|████▉     | 9530/19090 [4:38:08<4:46:08,  1.80s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 47.196022033691406, 'learning_rate': 2.9413215672059612e-05, 'recon_loss': 0.0014134335127891974, 'recon_static': 0.00044482036755653096, 'recon_motion': 0.0009686131408670918, 'recon_norm': 1.2439260631799698, 'recon_norm_std': 0.04783415785059333, 'target_norm': 1.6225368410348893, 'recon_cos': 0.6701266586780548, 'gradnorm/compressor.query': 0.07111657517685108, 'gradnorm/compressor.layer0': 1.757724742818082, 'gradnorm/compressor.layer1': 45.37482244283442, 'gradnorm/compressor.layer2': 11.06352754982348, 'gradnorm/compressor.layer3': 5.669983771837481, 'gradnorm/compressor.layer4': 2.2417538319862635, 'gradnorm/compressor.layer5': 1.430496903897148, 'gradnorm/compressor.layer6': 0.7015037919864412, 'gradnorm/compressor.layer7': 0.8762320291632688, 'gradnorm/decoder.layer0': 1.311366045519748, 'gradnorm/decoder.layer1': 0.841198843751041, 'gradnorm/decoder.layer2': 0.4309162085116421, 'gradnorm/max_param': 32.005184173583984, 'epoch': 4.99}

 50%|████▉     | 9530/19090 [4:38:08<4:46:08,  1.80s/it]
 50%|████▉     | 9531/19090 [4:38:10<4:21:59,  1.64s/it]
 50%|████▉     | 9532/19090 [4:38:12<5:07:53,  1.93s/it]
 50%|████▉     | 9533/19090 [4:38:14<4:36:58,  1.74s/it]
 50%|████▉     | 9534/19090 [4:38:15<4:15:30,  1.60s/it]
 50%|████▉     | 9535/19090 [4:38:18<5:07:16,  1.93s/it]
 50%|████▉     | 9536/19090 [4:38:19<4:36:36,  1.74s/it]
 50%|████▉     | 9537/19090 [4:38:20<4:14:48,  1.60s/it]
 50%|████▉     | 9538/19090 [4:38:23<5:12:55,  1.97s/it]
 50%|████▉     | 9539/19090 [4:38:24<4:42:43,  1.78s/it]
 50%|████▉     | 9540/19090 [4:38:26<4:19:06,  1.63s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 51.35111999511719, 'learning_rate': 2.9368213040638513e-05, 'recon_loss': 0.0013990458974149078, 'recon_static': 0.0004420562727318611, 'recon_motion': 0.0009569896472385153, 'recon_norm': 1.242492738366127, 'recon_norm_std': 0.04656722154468298, 'target_norm': 1.619599497318268, 'recon_cos': 0.6724022954702378, 'gradnorm/compressor.query': 0.07574582771379275, 'gradnorm/compressor.layer0': 1.9586940257358338, 'gradnorm/compressor.layer1': 49.256443899504156, 'gradnorm/compressor.layer2': 12.31115529419292, 'gradnorm/compressor.layer3': 6.561440227859241, 'gradnorm/compressor.layer4': 2.5338349506607454, 'gradnorm/compressor.layer5': 1.4661097231314648, 'gradnorm/compressor.layer6': 0.6938958899573486, 'gradnorm/compressor.layer7': 0.9323980743093746, 'gradnorm/decoder.layer0': 1.2511422660246811, 'gradnorm/decoder.layer1': 0.7712117159969503, 'gradnorm/decoder.layer2': 0.4475472742023039, 'gradnorm/max_param': 34.74291229248047, 'epoch': 5.0}

 50%|████▉     | 9540/19090 [4:38:26<4:19:06,  1.63s/it]
 50%|████▉     | 9541/19090 [4:38:28<5:01:55,  1.90s/it]
 50%|████▉     | 9542/19090 [4:38:29<4:32:22,  1.71s/it]
 50%|████▉     | 9543/19090 [4:38:31<4:11:10,  1.58s/it]
 50%|████▉     | 9544/19090 [4:38:32<3:56:17,  1.49s/it]
 50%|█████     | 9545/19090 [4:38:34<4:08:19,  1.56s/it]
 50%|█████     | 9546/19090 [4:38:42<9:25:33,  3.56s/it]
 50%|█████     | 9547/19090 [4:38:43<7:37:24,  2.88s/it]
 50%|█████     | 9548/19090 [4:38:47<8:04:08,  3.04s/it]
 50%|█████     | 9549/19090 [4:38:48<6:41:49,  2.53s/it]
 50%|█████     | 9550/19090 [4:38:49<5:42:36,  2.15s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 44.05537796020508, 'learning_rate': 2.932319580404933e-05, 'recon_loss': 0.0014082024077652021, 'recon_static': 0.00044009393022861334, 'recon_motion': 0.0009681084688054397, 'recon_norm': 1.2379581183195114, 'recon_norm_std': 0.04701715940609574, 'target_norm': 1.6224680095911026, 'recon_cos': 0.6712170481681824, 'gradnorm/compressor.query': 0.06568184346349183, 'gradnorm/compressor.layer0': 1.7214318593123263, 'gradnorm/compressor.layer1': 42.18739583320473, 'gradnorm/compressor.layer2': 10.615223773856252, 'gradnorm/compressor.layer3': 5.896936895210406, 'gradnorm/compressor.layer4': 2.1473420280644495, 'gradnorm/compressor.layer5': 1.4310035693312728, 'gradnorm/compressor.layer6': 0.7488379217630298, 'gradnorm/compressor.layer7': 0.937700854084252, 'gradnorm/decoder.layer0': 1.2533147730027963, 'gradnorm/decoder.layer1': 0.8876512640336448, 'gradnorm/decoder.layer2': 0.46658674801812816, 'gradnorm/max_param': 29.756223678588867, 'epoch': 5.0}

 50%|█████     | 9550/19090 [4:38:49<5:42:36,  2.15s/it]
 50%|█████     | 9551/19090 [4:38:51<5:46:01,  2.18s/it]
 50%|█████     | 9552/19090 [4:38:53<5:03:22,  1.91s/it]
 50%|█████     | 9553/19090 [4:38:54<4:33:30,  1.72s/it]
 50%|█████     | 9554/19090 [4:38:57<5:16:33,  1.99s/it]
 50%|█████     | 9555/19090 [4:38:58<4:42:38,  1.78s/it]
 50%|█████     | 9556/19090 [4:38:59<4:19:59,  1.64s/it]
 50%|█████     | 9557/19090 [4:39:02<4:55:30,  1.86s/it]
 50%|█████     | 9558/19090 [4:39:03<4:27:52,  1.69s/it]
 50%|█████     | 9559/19090 [4:39:04<4:09:23,  1.57s/it]
 50%|█████     | 9560/19090 [4:39:07<4:50:00,  1.83s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 51.05876922607422, 'learning_rate': 2.927816411280766e-05, 'recon_loss': 0.0014366442512255162, 'recon_static': 0.00047363564081024376, 'recon_motion': 0.0009630086278775707, 'recon_norm': 1.2559491962194442, 'recon_norm_std': 0.04674760932102799, 'target_norm': 1.6224871337413789, 'recon_cos': 0.6645306274294853, 'gradnorm/compressor.query': 0.09319117509444125, 'gradnorm/compressor.layer0': 2.270668257099938, 'gradnorm/compressor.layer1': 49.02955950320301, 'gradnorm/compressor.layer2': 12.069724651947455, 'gradnorm/compressor.layer3': 6.332762575364523, 'gradnorm/compressor.layer4': 2.477662332646044, 'gradnorm/compressor.layer5': 1.368849305683064, 'gradnorm/compressor.layer6': 0.761116219903301, 'gradnorm/compressor.layer7': 0.8647054945958991, 'gradnorm/decoder.layer0': 1.3521605474978615, 'gradnorm/decoder.layer1': 0.8734825122678893, 'gradnorm/decoder.layer2': 0.4716267368457559, 'gradnorm/max_param': 34.58149719238281, 'epoch': 5.01}

 50%|█████     | 9560/19090 [4:39:07<4:50:00,  1.83s/it]
 50%|█████     | 9561/19090 [4:39:08<4:23:57,  1.66s/it]
 50%|█████     | 9562/19090 [4:39:09<4:05:46,  1.55s/it]
 50%|█████     | 9563/19090 [4:39:11<4:41:40,  1.77s/it]
 50%|█████     | 9564/19090 [4:39:13<4:18:02,  1.63s/it]
 50%|█████     | 9565/19090 [4:39:14<4:01:53,  1.52s/it]
 50%|█████     | 9566/19090 [4:39:16<4:20:07,  1.64s/it]
 50%|█████     | 9567/19090 [4:39:17<4:02:59,  1.53s/it]
 50%|█████     | 9568/19090 [4:39:18<3:51:05,  1.46s/it]
 50%|█████     | 9569/19090 [4:39:21<4:47:16,  1.81s/it]
 50%|█████     | 9570/19090 [4:39:22<4:22:02,  1.65s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 49.16163635253906, 'learning_rate': 2.9233118117477475e-05, 'recon_loss': 0.0014248210994992406, 'recon_static': 0.00045703816722379996, 'recon_motion': 0.0009677829177235254, 'recon_norm': 1.2457647681236268, 'recon_norm_std': 0.047797322552651164, 'target_norm': 1.6222503691911698, 'recon_cos': 0.6671312391757965, 'gradnorm/compressor.query': 0.07167204444529245, 'gradnorm/compressor.layer0': 1.8783978509978392, 'gradnorm/compressor.layer1': 47.33170742054157, 'gradnorm/compressor.layer2': 11.26951094018967, 'gradnorm/compressor.layer3': 6.005095861020774, 'gradnorm/compressor.layer4': 2.138690269077577, 'gradnorm/compressor.layer5': 1.2807836847025036, 'gradnorm/compressor.layer6': 0.7970902021450786, 'gradnorm/compressor.layer7': 0.8855798271789189, 'gradnorm/decoder.layer0': 1.232571213465838, 'gradnorm/decoder.layer1': 0.809167520153985, 'gradnorm/decoder.layer2': 0.4142640720813562, 'gradnorm/max_param': 33.387542724609375, 'epoch': 5.01}

 50%|█████     | 9570/19090 [4:39:22<4:22:02,  1.65s/it]
 50%|█████     | 9571/19090 [4:39:24<4:04:54,  1.54s/it]
 50%|█████     | 9572/19090 [4:39:26<4:18:59,  1.63s/it]
 50%|█████     | 9573/19090 [4:39:27<4:02:40,  1.53s/it]
 50%|█████     | 9574/19090 [4:39:28<3:51:27,  1.46s/it]
 50%|█████     | 9575/19090 [4:39:30<4:23:56,  1.66s/it]
 50%|█████     | 9576/19090 [4:39:32<4:13:48,  1.60s/it]
 50%|█████     | 9577/19090 [4:39:33<3:59:15,  1.51s/it]
 50%|█████     | 9578/19090 [4:39:35<4:32:50,  1.72s/it]
 50%|█████     | 9579/19090 [4:39:37<4:42:58,  1.79s/it]
 50%|█████     | 9580/19090 [4:39:38<4:19:35,  1.64s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 48.73698425292969, 'learning_rate': 2.9188057968670556e-05, 'recon_loss': 0.0014472301700152457, 'recon_static': 0.00047954892797861246, 'recon_motion': 0.0009676812565885484, 'recon_norm': 1.2573574244976045, 'recon_norm_std': 0.046136739663779736, 'target_norm': 1.621422180533409, 'recon_cos': 0.661316953599453, 'gradnorm/compressor.query': 0.09491514331196914, 'gradnorm/compressor.layer0': 2.4532666985474, 'gradnorm/compressor.layer1': 46.8689032493494, 'gradnorm/compressor.layer2': 11.450624034151343, 'gradnorm/compressor.layer3': 5.525772572352422, 'gradnorm/compressor.layer4': 2.19775580469883, 'gradnorm/compressor.layer5': 1.4378155071821563, 'gradnorm/compressor.layer6': 0.8072052981208446, 'gradnorm/compressor.layer7': 0.9562411725503216, 'gradnorm/decoder.layer0': 1.285027030779265, 'gradnorm/decoder.layer1': 0.8143017305684852, 'gradnorm/decoder.layer2': 0.37814872761832224, 'gradnorm/max_param': 33.050777435302734, 'epoch': 5.02}

 50%|█████     | 9580/19090 [4:39:38<4:19:35,  1.64s/it]
 50%|█████     | 9581/19090 [4:39:40<4:18:05,  1.63s/it]
 50%|█████     | 9582/19090 [4:39:42<4:48:44,  1.82s/it]
 50%|█████     | 9583/19090 [4:39:44<4:23:19,  1.66s/it]
 50%|█████     | 9584/19090 [4:39:45<4:05:41,  1.55s/it]
 50%|█████     | 9585/19090 [4:39:47<4:42:19,  1.78s/it]
 50%|█████     | 9586/19090 [4:39:49<4:19:35,  1.64s/it]
 50%|█████     | 9587/19090 [4:39:50<4:02:44,  1.53s/it]
 50%|█████     | 9588/19090 [4:39:52<4:32:05,  1.72s/it]
 50%|█████     | 9589/19090 [4:39:53<4:11:49,  1.59s/it]
 50%|█████     | 9590/19090 [4:39:55<4:10:04,  1.58s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 46.49886703491211, 'learning_rate': 2.9142983817045987e-05, 'recon_loss': 0.001512709676171653, 'recon_static': 0.0005594654991000425, 'recon_motion': 0.0009532441414194182, 'recon_norm': 1.293702957034111, 'recon_norm_std': 0.045865587797015904, 'target_norm': 1.6173903495073318, 'recon_cos': 0.6450746804475784, 'gradnorm/compressor.query': 0.08056879040051469, 'gradnorm/compressor.layer0': 2.185797328422171, 'gradnorm/compressor.layer1': 44.44325461596366, 'gradnorm/compressor.layer2': 11.61386986609552, 'gradnorm/compressor.layer3': 6.099854214286344, 'gradnorm/compressor.layer4': 2.164635537637508, 'gradnorm/compressor.layer5': 1.3514777360917691, 'gradnorm/compressor.layer6': 0.7426106902822448, 'gradnorm/compressor.layer7': 0.9117287734133636, 'gradnorm/decoder.layer0': 1.1778001150126312, 'gradnorm/decoder.layer1': 0.7875799471525022, 'gradnorm/decoder.layer2': 0.4063018619005532, 'gradnorm/max_param': 31.34330177307129, 'epoch': 5.02}

 50%|█████     | 9590/19090 [4:39:55<4:10:04,  1.58s/it]
 50%|█████     | 9591/19090 [4:39:57<4:35:56,  1.74s/it]
 50%|█████     | 9592/19090 [4:39:58<4:14:48,  1.61s/it]
 50%|█████     | 9593/19090 [4:40:00<4:28:09,  1.69s/it]
 50%|█████     | 9594/19090 [4:40:02<4:29:09,  1.70s/it]
 50%|█████     | 9595/19090 [4:40:03<4:10:36,  1.58s/it]
 50%|█████     | 9596/19090 [4:40:05<4:18:05,  1.63s/it]
 50%|█████     | 9597/19090 [4:40:07<4:22:34,  1.66s/it]
 50%|█████     | 9598/19090 [4:40:08<4:05:02,  1.55s/it]
 50%|█████     | 9599/19090 [4:40:10<4:13:39,  1.60s/it]
 50%|█████     | 9600/19090 [4:40:12<4:52:30,  1.85s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 49.37689208984375, 'learning_rate': 2.9097895813309707e-05, 'recon_loss': 0.0014791091351071372, 'recon_static': 0.0005155950464541092, 'recon_motion': 0.0009635140988393687, 'recon_norm': 1.2790030896663667, 'recon_norm_std': 0.04626446794718504, 'target_norm': 1.619613564014435, 'recon_cos': 0.6537445113062859, 'gradnorm/compressor.query': 0.07527907152970822, 'gradnorm/compressor.layer0': 2.2076537208129556, 'gradnorm/compressor.layer1': 47.61038264030494, 'gradnorm/compressor.layer2': 11.084905488983974, 'gradnorm/compressor.layer3': 5.671409903128731, 'gradnorm/compressor.layer4': 2.349069687148631, 'gradnorm/compressor.layer5': 1.3769382773606629, 'gradnorm/compressor.layer6': 0.7877329587748693, 'gradnorm/compressor.layer7': 0.9203884294207012, 'gradnorm/decoder.layer0': 1.2890330455767245, 'gradnorm/decoder.layer1': 0.8303702165889941, 'gradnorm/decoder.layer2': 0.42038977525955185, 'gradnorm/max_param': 33.58292007446289, 'epoch': 5.03}

 50%|█████     | 9600/19090 [4:40:12<4:52:30,  1.85s/it]
 50%|█████     | 9601/19090 [4:40:13<4:25:49,  1.68s/it]
 50%|█████     | 9602/19090 [4:40:15<4:07:48,  1.57s/it]
 50%|█████     | 9603/19090 [4:40:17<4:57:35,  1.88s/it]
 50%|█████     | 9604/19090 [4:40:19<4:29:29,  1.70s/it]
 50%|█████     | 9605/19090 [4:40:20<4:09:45,  1.58s/it]
 50%|█████     | 9606/19090 [4:40:22<4:42:01,  1.78s/it]
 50%|█████     | 9607/19090 [4:40:23<4:18:26,  1.64s/it]
 50%|█████     | 9608/19090 [4:40:25<4:02:22,  1.53s/it]
 50%|█████     | 9609/19090 [4:40:27<4:34:34,  1.74s/it]
 50%|█████     | 9610/19090 [4:40:28<4:13:25,  1.60s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 49.83842468261719, 'learning_rate': 2.9052794108213932e-05, 'recon_loss': 0.0014622642484027892, 'recon_static': 0.0004933554359013214, 'recon_motion': 0.0009689088328741491, 'recon_norm': 1.265548327565193, 'recon_norm_std': 0.04549219608306885, 'target_norm': 1.619857805967331, 'recon_cos': 0.6570735067129135, 'gradnorm/compressor.query': 0.0686682907781369, 'gradnorm/compressor.layer0': 1.826438582420779, 'gradnorm/compressor.layer1': 48.033849860508525, 'gradnorm/compressor.layer2': 11.360480463317645, 'gradnorm/compressor.layer3': 5.844807411852047, 'gradnorm/compressor.layer4': 2.0872519705522423, 'gradnorm/compressor.layer5': 1.3484510648714796, 'gradnorm/compressor.layer6': 0.7116779621383295, 'gradnorm/compressor.layer7': 0.8880889993199124, 'gradnorm/decoder.layer0': 1.2872490351249737, 'gradnorm/decoder.layer1': 0.8449886907576883, 'gradnorm/decoder.layer2': 0.4619123366083722, 'gradnorm/max_param': 33.8818473815918, 'epoch': 5.03}

 50%|█████     | 9610/19090 [4:40:28<4:13:25,  1.60s/it]
 50%|█████     | 9611/19090 [4:40:30<4:07:12,  1.56s/it]
 50%|█████     | 9612/19090 [4:40:32<4:31:55,  1.72s/it]
 50%|█████     | 9613/19090 [4:40:33<4:11:38,  1.59s/it]
 50%|█████     | 9614/19090 [4:40:35<4:29:43,  1.71s/it]
 50%|█████     | 9615/19090 [4:40:37<4:29:42,  1.71s/it]
 50%|█████     | 9616/19090 [4:40:38<4:10:06,  1.58s/it]
 50%|█████     | 9617/19090 [4:40:40<4:30:10,  1.71s/it]
 50%|█████     | 9618/19090 [4:40:42<4:48:10,  1.83s/it]
 50%|█████     | 9619/19090 [4:40:43<4:22:58,  1.67s/it]
 50%|█████     | 9620/19090 [4:40:45<4:05:24,  1.55s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 48.19348907470703, 'learning_rate': 2.9007678852556713e-05, 'recon_loss': 0.0014349567412864418, 'recon_static': 0.000468279083725065, 'recon_motion': 0.0009666776852100157, 'recon_norm': 1.2549580752849578, 'recon_norm_std': 0.046134142018854615, 'target_norm': 1.620952644944191, 'recon_cos': 0.664427301287651, 'gradnorm/compressor.query': 0.07275971127079164, 'gradnorm/compressor.layer0': 2.0317235541365735, 'gradnorm/compressor.layer1': 46.311437761940894, 'gradnorm/compressor.layer2': 11.03328627684855, 'gradnorm/compressor.layer3': 6.336043463972114, 'gradnorm/compressor.layer4': 2.3107759389571885, 'gradnorm/compressor.layer5': 1.5056297038041648, 'gradnorm/compressor.layer6': 0.7739841571536864, 'gradnorm/compressor.layer7': 0.9583842775376673, 'gradnorm/decoder.layer0': 1.3287564291309555, 'gradnorm/decoder.layer1': 0.8518228045542888, 'gradnorm/decoder.layer2': 0.4854080812025148, 'gradnorm/max_param': 32.66410827636719, 'epoch': 5.04}

 50%|█████     | 9620/19090 [4:40:45<4:05:24,  1.55s/it]
 50%|█████     | 9621/19090 [4:40:47<4:24:55,  1.68s/it]
 50%|█████     | 9622/19090 [4:40:48<4:06:31,  1.56s/it]
 50%|█████     | 9623/19090 [4:40:49<3:54:06,  1.48s/it]
 50%|█████     | 9624/19090 [4:40:52<4:34:52,  1.74s/it]
 50%|█████     | 9625/19090 [4:40:53<4:13:04,  1.60s/it]
 50%|█████     | 9626/19090 [4:40:54<4:02:54,  1.54s/it]
 50%|█████     | 9627/19090 [4:40:57<4:37:43,  1.76s/it]
 50%|█████     | 9628/19090 [4:40:58<4:14:59,  1.62s/it]
 50%|█████     | 9629/19090 [4:40:59<4:09:51,  1.58s/it]
 50%|█████     | 9630/19090 [4:41:01<4:19:52,  1.65s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 47.333106994628906, 'learning_rate': 2.8962550197181406e-05, 'recon_loss': 0.0014300822542281822, 'recon_static': 0.00047290390721173027, 'recon_motion': 0.0009571783564751967, 'recon_norm': 1.253372785449028, 'recon_norm_std': 0.045712160598486665, 'target_norm': 1.618901827931404, 'recon_cos': 0.6643307834863663, 'gradnorm/compressor.query': 0.07817395881579957, 'gradnorm/compressor.layer0': 2.023332747763602, 'gradnorm/compressor.layer1': 45.36069904199146, 'gradnorm/compressor.layer2': 11.26644881750765, 'gradnorm/compressor.layer3': 6.422451763751507, 'gradnorm/compressor.layer4': 2.2971235906278307, 'gradnorm/compressor.layer5': 1.2188451191879042, 'gradnorm/compressor.layer6': 0.6925165561046535, 'gradnorm/compressor.layer7': 0.8646872622672617, 'gradnorm/decoder.layer0': 1.2987450135679595, 'gradnorm/decoder.layer1': 0.8061109748267384, 'gradnorm/decoder.layer2': 0.47215035848510023, 'gradnorm/max_param': 31.995620727539062, 'epoch': 5.04}

 50%|█████     | 9630/19090 [4:41:01<4:19:52,  1.65s/it]
 50%|█████     | 9631/19090 [4:41:02<4:02:45,  1.54s/it]
 50%|█████     | 9632/19090 [4:41:05<4:29:22,  1.71s/it]
 50%|█████     | 9633/19090 [4:41:06<4:09:12,  1.58s/it]
 50%|█████     | 9634/19090 [4:41:07<3:54:59,  1.49s/it]
 50%|█████     | 9635/19090 [4:41:10<4:46:14,  1.82s/it]
 50%|█████     | 9636/19090 [4:41:11<4:20:49,  1.66s/it]
 50%|█████     | 9637/19090 [4:41:12<4:03:18,  1.54s/it]
 50%|█████     | 9638/19090 [4:41:15<4:48:43,  1.83s/it]
 50%|█████     | 9639/19090 [4:41:16<4:24:09,  1.68s/it]
 50%|█████     | 9640/19090 [4:41:17<4:05:42,  1.56s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 49.69304275512695, 'learning_rate': 2.891740829297615e-05, 'recon_loss': 0.0014364051428856328, 'recon_static': 0.00047076401970116424, 'recon_motion': 0.0009656411290052347, 'recon_norm': 1.252871784567833, 'recon_norm_std': 0.047491357754915954, 'target_norm': 1.6221334606409072, 'recon_cos': 0.6640405401587486, 'gradnorm/compressor.query': 0.07537476689071732, 'gradnorm/compressor.layer0': 2.0331057947555156, 'gradnorm/compressor.layer1': 47.66376477636077, 'gradnorm/compressor.layer2': 11.904456937244266, 'gradnorm/compressor.layer3': 6.378341696542414, 'gradnorm/compressor.layer4': 2.257574471463048, 'gradnorm/compressor.layer5': 1.4810864611777426, 'gradnorm/compressor.layer6': 0.7073587744389384, 'gradnorm/compressor.layer7': 0.8800927185934605, 'gradnorm/decoder.layer0': 1.2297663061150865, 'gradnorm/decoder.layer1': 0.8549429265989319, 'gradnorm/decoder.layer2': 0.4656881628783099, 'gradnorm/max_param': 33.619544982910156, 'epoch': 5.05}

 50%|█████     | 9640/19090 [4:41:17<4:05:42,  1.56s/it]
 51%|█████     | 9641/19090 [4:41:20<4:55:11,  1.87s/it]
 51%|█████     | 9642/19090 [4:41:22<4:42:13,  1.79s/it]
 51%|█████     | 9643/19090 [4:41:23<4:18:11,  1.64s/it]
 51%|█████     | 9644/19090 [4:41:25<4:28:17,  1.70s/it]
 51%|█████     | 9645/19090 [4:41:26<4:23:16,  1.67s/it]
 51%|█████     | 9646/19090 [4:41:28<4:05:17,  1.56s/it]
 51%|█████     | 9647/19090 [4:41:29<3:56:08,  1.50s/it]
 51%|█████     | 9648/19090 [4:41:31<4:33:40,  1.74s/it]
 51%|█████     | 9649/19090 [4:41:33<4:12:30,  1.60s/it]
 51%|█████     | 9650/19090 [4:41:34<4:13:17,  1.61s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 49.66437530517578, 'learning_rate': 2.8872253290873414e-05, 'recon_loss': 0.0014409199764486402, 'recon_static': 0.00048479598335688935, 'recon_motion': 0.0009561240251059644, 'recon_norm': 1.2619200617074966, 'recon_norm_std': 0.046476691402494906, 'target_norm': 1.6190923541784286, 'recon_cos': 0.6622214064002037, 'gradnorm/compressor.query': 0.06897515807243429, 'gradnorm/compressor.layer0': 1.9155100428577607, 'gradnorm/compressor.layer1': 47.56772259023838, 'gradnorm/compressor.layer2': 12.361631011506873, 'gradnorm/compressor.layer3': 6.047550291851296, 'gradnorm/compressor.layer4': 2.2427701292783446, 'gradnorm/compressor.layer5': 1.3168536142025822, 'gradnorm/compressor.layer6': 0.7817142632992541, 'gradnorm/compressor.layer7': 0.9206762763904602, 'gradnorm/decoder.layer0': 1.3076098942571068, 'gradnorm/decoder.layer1': 0.8286794929373619, 'gradnorm/decoder.layer2': 0.4307498457194897, 'gradnorm/max_param': 33.55454635620117, 'epoch': 5.06}

 51%|█████     | 9650/19090 [4:41:34<4:13:17,  1.61s/it]
 51%|█████     | 9651/19090 [4:41:37<4:57:28,  1.89s/it]
 51%|█████     | 9652/19090 [4:41:38<4:29:21,  1.71s/it]
 51%|█████     | 9653/19090 [4:41:39<4:14:53,  1.62s/it]
 51%|█████     | 9654/19090 [4:41:42<4:41:36,  1.79s/it]
 51%|█████     | 9655/19090 [4:41:43<4:17:52,  1.64s/it]
 51%|█████     | 9656/19090 [4:41:45<4:17:28,  1.64s/it]
 51%|█████     | 9657/19090 [4:41:46<4:23:12,  1.67s/it]
 51%|█████     | 9658/19090 [4:41:48<4:05:04,  1.56s/it]
 51%|█████     | 9659/19090 [4:41:50<4:27:51,  1.70s/it]
 51%|█████     | 9660/19090 [4:41:51<4:25:45,  1.69s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 52.30977249145508, 'learning_rate': 2.8827085341849436e-05, 'recon_loss': 0.0014358734071720392, 'recon_static': 0.00048272809726768173, 'recon_motion': 0.0009531452844385058, 'recon_norm': 1.2573034107685088, 'recon_norm_std': 0.04707715418189764, 'target_norm': 1.6196133553981782, 'recon_cos': 0.6632208362221718, 'gradnorm/compressor.query': 0.08659355398838735, 'gradnorm/compressor.layer0': 2.1024137566802983, 'gradnorm/compressor.layer1': 50.328755623183504, 'gradnorm/compressor.layer2': 11.771182830933254, 'gradnorm/compressor.layer3': 6.952015306023538, 'gradnorm/compressor.layer4': 2.275170992618862, 'gradnorm/compressor.layer5': 1.602738409295282, 'gradnorm/compressor.layer6': 0.758326994865952, 'gradnorm/compressor.layer7': 1.0218451539326898, 'gradnorm/decoder.layer0': 1.2869732195743668, 'gradnorm/decoder.layer1': 0.8932095554226771, 'gradnorm/decoder.layer2': 0.4377368558329637, 'gradnorm/max_param': 35.49781799316406, 'epoch': 5.06}

 51%|█████     | 9660/19090 [4:41:51<4:25:45,  1.69s/it]
 51%|█████     | 9661/19090 [4:41:53<4:06:50,  1.57s/it]
 51%|█████     | 9662/19090 [4:41:55<4:25:18,  1.69s/it]
 51%|█████     | 9663/19090 [4:41:56<4:09:18,  1.59s/it]
 51%|█████     | 9664/19090 [4:41:57<3:55:02,  1.50s/it]
 51%|█████     | 9665/19090 [4:41:59<4:31:13,  1.73s/it]
 51%|█████     | 9666/19090 [4:42:01<4:33:55,  1.74s/it]
 51%|█████     | 9667/19090 [4:42:03<4:12:05,  1.61s/it]
 51%|█████     | 9668/19090 [4:42:05<4:34:01,  1.74s/it]
 51%|█████     | 9669/19090 [4:42:06<4:27:44,  1.71s/it]
 51%|█████     | 9670/19090 [4:42:07<4:08:03,  1.58s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 51.23625564575195, 'learning_rate': 2.8781904596923732e-05, 'recon_loss': 0.0014401549735339359, 'recon_static': 0.00047690810024505483, 'recon_motion': 0.0009632468354539014, 'recon_norm': 1.260486227273941, 'recon_norm_std': 0.04664286272600293, 'target_norm': 1.6196590811014175, 'recon_cos': 0.6625095292925834, 'gradnorm/compressor.query': 0.08177661259034828, 'gradnorm/compressor.layer0': 2.181856231827277, 'gradnorm/compressor.layer1': 49.07030690276198, 'gradnorm/compressor.layer2': 12.349641345295089, 'gradnorm/compressor.layer3': 6.78652341220555, 'gradnorm/compressor.layer4': 2.5630991165576935, 'gradnorm/compressor.layer5': 1.626575586447717, 'gradnorm/compressor.layer6': 0.8512174974865964, 'gradnorm/compressor.layer7': 1.0515297985235237, 'gradnorm/decoder.layer0': 1.3829533995488754, 'gradnorm/decoder.layer1': 0.8747323853102985, 'gradnorm/decoder.layer2': 0.44423535225054206, 'gradnorm/max_param': 34.61033248901367, 'epoch': 5.07}

 51%|█████     | 9670/19090 [4:42:08<4:08:03,  1.58s/it]
 51%|█████     | 9671/19090 [4:42:10<4:31:58,  1.73s/it]
 51%|█████     | 9672/19090 [4:42:11<4:38:07,  1.77s/it]
 51%|█████     | 9673/19090 [4:42:13<4:15:16,  1.63s/it]
 51%|█████     | 9674/19090 [4:42:14<4:18:07,  1.64s/it]
 51%|█████     | 9675/19090 [4:42:17<4:47:00,  1.83s/it]
 51%|█████     | 9676/19090 [4:42:18<4:21:22,  1.67s/it]
 51%|█████     | 9677/19090 [4:42:20<4:31:08,  1.73s/it]
 51%|█████     | 9678/19090 [4:42:22<4:50:57,  1.85s/it]
 51%|█████     | 9679/19090 [4:42:23<4:24:02,  1.68s/it]
 51%|█████     | 9680/19090 [4:42:25<4:34:18,  1.75s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 55.23395919799805, 'learning_rate': 2.8736711207158633e-05, 'recon_loss': 0.001464214880252257, 'recon_static': 0.0004910469811875373, 'recon_motion': 0.0009731678641401231, 'recon_norm': 1.2630217105150223, 'recon_norm_std': 0.04471116829663515, 'target_norm': 1.6220344632863999, 'recon_cos': 0.6572762548923492, 'gradnorm/compressor.query': 0.079631151116704, 'gradnorm/compressor.layer0': 2.193824779833571, 'gradnorm/compressor.layer1': 53.110710394361604, 'gradnorm/compressor.layer2': 13.131906781119328, 'gradnorm/compressor.layer3': 6.376795497190441, 'gradnorm/compressor.layer4': 2.236664672581579, 'gradnorm/compressor.layer5': 1.5660158655476166, 'gradnorm/compressor.layer6': 0.8031146408778541, 'gradnorm/compressor.layer7': 0.9796428469470202, 'gradnorm/decoder.layer0': 1.4397862539335453, 'gradnorm/decoder.layer1': 0.8971137536491912, 'gradnorm/decoder.layer2': 0.4193285098792812, 'gradnorm/max_param': 37.46178436279297, 'epoch': 5.07}

 51%|█████     | 9680/19090 [4:42:25<4:34:18,  1.75s/it]
 51%|█████     | 9681/19090 [4:42:28<5:01:10,  1.92s/it]
 51%|█████     | 9682/19090 [4:42:29<4:31:44,  1.73s/it]
 51%|█████     | 9683/19090 [4:42:30<4:26:19,  1.70s/it]
 51%|█████     | 9684/19090 [4:42:33<4:54:11,  1.88s/it]
 51%|█████     | 9685/19090 [4:42:34<4:26:15,  1.70s/it]
 51%|█████     | 9686/19090 [4:42:36<4:31:53,  1.73s/it]
 51%|█████     | 9687/19090 [4:42:37<4:28:41,  1.71s/it]
 51%|█████     | 9688/19090 [4:42:39<4:08:14,  1.58s/it]
 51%|█████     | 9689/19090 [4:42:41<4:25:19,  1.69s/it]
 51%|█████     | 9690/19090 [4:42:43<4:37:35,  1.77s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 46.335792541503906, 'learning_rate': 2.869150532365874e-05, 'recon_loss': 0.0014701729640364646, 'recon_static': 0.0004993683032807894, 'recon_motion': 0.0009708046680316329, 'recon_norm': 1.264767724275589, 'recon_norm_std': 0.04566427152603865, 'target_norm': 1.6211129575967789, 'recon_cos': 0.6554988294839859, 'gradnorm/compressor.query': 0.06374141751046504, 'gradnorm/compressor.layer0': 1.8505073470284439, 'gradnorm/compressor.layer1': 44.39880935859584, 'gradnorm/compressor.layer2': 11.084034794764177, 'gradnorm/compressor.layer3': 6.187040677987677, 'gradnorm/compressor.layer4': 2.28468263070672, 'gradnorm/compressor.layer5': 1.3580049209421297, 'gradnorm/compressor.layer6': 0.8397656426592551, 'gradnorm/compressor.layer7': 0.8890547713495218, 'gradnorm/decoder.layer0': 1.3421677433642616, 'gradnorm/decoder.layer1': 0.7824582484261319, 'gradnorm/decoder.layer2': 0.4626938284853787, 'gradnorm/max_param': 31.314905166625977, 'epoch': 5.08}

 51%|█████     | 9690/19090 [4:42:43<4:37:35,  1.77s/it]
 51%|█████     | 9691/19090 [4:42:44<4:14:34,  1.63s/it]
 51%|█████     | 9692/19090 [4:42:46<4:27:22,  1.71s/it]
 51%|█████     | 9693/19090 [4:42:48<4:45:00,  1.82s/it]
 51%|█████     | 9694/19090 [4:42:49<4:19:48,  1.66s/it]
 51%|█████     | 9695/19090 [4:42:51<4:33:51,  1.75s/it]
 51%|█████     | 9696/19090 [4:42:53<4:38:36,  1.78s/it]
 51%|█████     | 9697/19090 [4:42:54<4:15:13,  1.63s/it]
 51%|█████     | 9698/19090 [4:42:56<4:31:20,  1.73s/it]
 51%|█████     | 9699/19090 [4:42:58<4:38:51,  1.78s/it]
 51%|█████     | 9700/19090 [4:42:59<4:15:28,  1.63s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 57.85092544555664, 'learning_rate': 2.8646287097570394e-05, 'recon_loss': 0.0014404631045181304, 'recon_static': 0.000487566101946868, 'recon_motion': 0.000952897006936837, 'recon_norm': 1.2640549451112748, 'recon_norm_std': 0.046261485666036606, 'target_norm': 1.6177021980285644, 'recon_cos': 0.6617930069565773, 'gradnorm/compressor.query': 0.08375185004546648, 'gradnorm/compressor.layer0': 2.27667666262233, 'gradnorm/compressor.layer1': 55.473613343135945, 'gradnorm/compressor.layer2': 14.050635728673733, 'gradnorm/compressor.layer3': 7.292155600004551, 'gradnorm/compressor.layer4': 2.5348389225633814, 'gradnorm/compressor.layer5': 1.644890390356151, 'gradnorm/compressor.layer6': 0.8536420855033549, 'gradnorm/compressor.layer7': 1.001161199148875, 'gradnorm/decoder.layer0': 1.3267968661590128, 'gradnorm/decoder.layer1': 0.8808593411667642, 'gradnorm/decoder.layer2': 0.47248496743161317, 'gradnorm/max_param': 39.130496978759766, 'epoch': 5.08}

 51%|█████     | 9700/19090 [4:42:59<4:15:28,  1.63s/it]
 51%|█████     | 9701/19090 [4:43:02<4:38:31,  1.78s/it]
 51%|█████     | 9702/19090 [4:43:03<4:38:16,  1.78s/it]
 51%|█████     | 9703/19090 [4:43:05<4:15:02,  1.63s/it]
 51%|█████     | 9704/19090 [4:43:07<4:42:07,  1.80s/it]
 51%|█████     | 9705/19090 [4:43:09<4:41:50,  1.80s/it]
 51%|█████     | 9706/19090 [4:43:10<4:17:44,  1.65s/it]
 51%|█████     | 9707/19090 [4:43:12<4:57:51,  1.90s/it]
 51%|█████     | 9708/19090 [4:43:14<4:43:06,  1.81s/it]
 51%|█████     | 9709/19090 [4:43:15<4:18:42,  1.65s/it]
 51%|█████     | 9710/19090 [4:43:18<4:47:53,  1.84s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 49.99292755126953, 'learning_rate': 2.8601056680081245e-05, 'recon_loss': 0.0014581020892364903, 'recon_static': 0.00048482595375389794, 'recon_motion': 0.0009732761027407832, 'recon_norm': 1.2591765463352202, 'recon_norm_std': 0.04502831641584635, 'target_norm': 1.6209293395280837, 'recon_cos': 0.6580294832587242, 'gradnorm/compressor.query': 0.08194995723544181, 'gradnorm/compressor.layer0': 1.9735356249844809, 'gradnorm/compressor.layer1': 47.861560817698475, 'gradnorm/compressor.layer2': 12.466430754307872, 'gradnorm/compressor.layer3': 6.263643555537009, 'gradnorm/compressor.layer4': 2.178285865918353, 'gradnorm/compressor.layer5': 1.2684639537803588, 'gradnorm/compressor.layer6': 0.7234403325925537, 'gradnorm/compressor.layer7': 0.8921173375428535, 'gradnorm/decoder.layer0': 1.213970790624763, 'gradnorm/decoder.layer1': 0.8385823913225733, 'gradnorm/decoder.layer2': 0.40924712651851286, 'gradnorm/max_param': 33.7584342956543, 'epoch': 5.09}

 51%|█████     | 9710/19090 [4:43:18<4:47:53,  1.84s/it]
 51%|█████     | 9711/19090 [4:43:19<4:26:04,  1.70s/it]
 51%|█████     | 9712/19090 [4:43:20<4:06:45,  1.58s/it]
 51%|█████     | 9713/19090 [4:43:23<4:43:12,  1.81s/it]
 51%|█████     | 9714/19090 [4:43:24<4:31:36,  1.74s/it]
 51%|█████     | 9715/19090 [4:43:25<4:10:13,  1.60s/it]
 51%|█████     | 9716/19090 [4:43:28<4:50:22,  1.86s/it]
 51%|█████     | 9717/19090 [4:43:29<4:30:00,  1.73s/it]
 51%|█████     | 9718/19090 [4:43:31<4:08:52,  1.59s/it]
 51%|█████     | 9719/19090 [4:43:33<5:02:56,  1.94s/it]
 51%|█████     | 9720/19090 [4:43:35<4:41:19,  1.80s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 52.50255584716797, 'learning_rate': 2.855581422241969e-05, 'recon_loss': 0.001466702140169218, 'recon_static': 0.0005025028833188117, 'recon_motion': 0.0009641992568504065, 'recon_norm': 1.2678667515516282, 'recon_norm_std': 0.04537974931299686, 'target_norm': 1.6198609501123429, 'recon_cos': 0.6557201027870179, 'gradnorm/compressor.query': 0.08202162470432563, 'gradnorm/compressor.layer0': 2.0717321746527015, 'gradnorm/compressor.layer1': 50.456993599747896, 'gradnorm/compressor.layer2': 12.486353635053325, 'gradnorm/compressor.layer3': 6.190421291429272, 'gradnorm/compressor.layer4': 2.376051619755693, 'gradnorm/compressor.layer5': 1.3735305129703999, 'gradnorm/compressor.layer6': 0.7922788352906941, 'gradnorm/compressor.layer7': 0.9713478940432244, 'gradnorm/decoder.layer0': 1.4011414030328004, 'gradnorm/decoder.layer1': 0.8923201240266752, 'gradnorm/decoder.layer2': 0.46652229565108017, 'gradnorm/max_param': 35.59041976928711, 'epoch': 5.09}

 51%|█████     | 9720/19090 [4:43:35<4:41:19,  1.80s/it]
 51%|█████     | 9721/19090 [4:43:36<4:17:04,  1.65s/it]
 51%|█████     | 9722/19090 [4:43:39<4:50:55,  1.86s/it]
 51%|█████     | 9723/19090 [4:43:40<4:49:09,  1.85s/it]
 51%|█████     | 9724/19090 [4:43:42<4:22:15,  1.68s/it]
 51%|█████     | 9725/19090 [4:43:44<4:46:27,  1.84s/it]
 51%|█████     | 9726/19090 [4:43:46<4:42:02,  1.81s/it]
 51%|█████     | 9727/19090 [4:43:47<4:17:30,  1.65s/it]
 51%|█████     | 9728/19090 [4:43:49<4:39:31,  1.79s/it]
 51%|█████     | 9729/19090 [4:43:51<4:35:38,  1.77s/it]
 51%|█████     | 9730/19090 [4:43:52<4:13:12,  1.62s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 49.70810317993164, 'learning_rate': 2.8510559875854377e-05, 'recon_loss': 0.001466860980144702, 'recon_static': 0.0005022161203669384, 'recon_motion': 0.0009646448670537211, 'recon_norm': 1.267751345038414, 'recon_norm_std': 0.04612996885553002, 'target_norm': 1.6216090351343155, 'recon_cos': 0.6569282874464989, 'gradnorm/compressor.query': 0.08960803353520286, 'gradnorm/compressor.layer0': 2.2570033752051093, 'gradnorm/compressor.layer1': 47.5674839290849, 'gradnorm/compressor.layer2': 11.972228020481989, 'gradnorm/compressor.layer3': 6.925793918311877, 'gradnorm/compressor.layer4': 2.410952048712078, 'gradnorm/compressor.layer5': 1.4339517671249176, 'gradnorm/compressor.layer6': 0.7502420352765977, 'gradnorm/compressor.layer7': 0.929446373451903, 'gradnorm/decoder.layer0': 1.3068410038368812, 'gradnorm/decoder.layer1': 0.8054304304359547, 'gradnorm/decoder.layer2': 0.41854211216066284, 'gradnorm/max_param': 33.54704666137695, 'epoch': 5.1}

 51%|█████     | 9730/19090 [4:43:52<4:13:12,  1.62s/it]
 51%|█████     | 9731/19090 [4:43:54<4:52:31,  1.88s/it]
 51%|█████     | 9732/19090 [4:43:56<4:38:45,  1.79s/it]
 51%|█████     | 9733/19090 [4:43:57<4:15:20,  1.64s/it]
 51%|█████     | 9734/19090 [4:44:00<4:51:13,  1.87s/it]
 51%|█████     | 9735/19090 [4:44:01<4:46:16,  1.84s/it]
 51%|█████     | 9736/19090 [4:44:03<4:20:24,  1.67s/it]
 51%|█████     | 9737/19090 [4:44:05<4:40:52,  1.80s/it]
 51%|█████     | 9738/19090 [4:44:07<4:49:40,  1.86s/it]
 51%|█████     | 9739/19090 [4:44:08<4:22:52,  1.69s/it]
 51%|█████     | 9740/19090 [4:44:10<4:20:16,  1.67s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 57.11872100830078, 'learning_rate': 2.846529379169371e-05, 'recon_loss': 0.0014414254925213753, 'recon_static': 0.00048637184372637423, 'recon_motion': 0.0009550536109600216, 'recon_norm': 1.2636523842811584, 'recon_norm_std': 0.04580422313883901, 'target_norm': 1.6170764863491058, 'recon_cos': 0.6611055940389633, 'gradnorm/compressor.query': 0.08363274495880664, 'gradnorm/compressor.layer0': 2.2511364927594246, 'gradnorm/compressor.layer1': 54.88178213000763, 'gradnorm/compressor.layer2': 13.474375388650037, 'gradnorm/compressor.layer3': 7.042742615008713, 'gradnorm/compressor.layer4': 2.7339191165727477, 'gradnorm/compressor.layer5': 1.5234364044968016, 'gradnorm/compressor.layer6': 0.8065167348086826, 'gradnorm/compressor.layer7': 1.0554420315334663, 'gradnorm/decoder.layer0': 1.3333047277640404, 'gradnorm/decoder.layer1': 0.8755392048073365, 'gradnorm/decoder.layer2': 0.4476496284887246, 'gradnorm/max_param': 38.711517333984375, 'epoch': 5.1}

 51%|█████     | 9740/19090 [4:44:10<4:20:16,  1.67s/it]
 51%|█████     | 9741/19090 [4:44:13<5:11:54,  2.00s/it]
 51%|█████     | 9742/19090 [4:44:14<4:38:19,  1.79s/it]
 51%|█████     | 9743/19090 [4:44:15<4:14:46,  1.64s/it]
 51%|█████     | 9744/19090 [4:44:17<4:50:20,  1.86s/it]
 51%|█████     | 9745/19090 [4:44:19<4:23:31,  1.69s/it]
 51%|█████     | 9746/19090 [4:44:20<4:04:40,  1.57s/it]
 51%|█████     | 9747/19090 [4:44:23<4:54:35,  1.89s/it]
 51%|█████     | 9748/19090 [4:44:24<4:25:58,  1.71s/it]
 51%|█████     | 9749/19090 [4:44:25<4:06:17,  1.58s/it]
 51%|█████     | 9750/19090 [4:44:28<4:59:44,  1.93s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 54.087547302246094, 'learning_rate': 2.8420016121285333e-05, 'recon_loss': 0.001462160053779371, 'recon_static': 0.0004942951665725559, 'recon_motion': 0.0009678648944827728, 'recon_norm': 1.26364766061306, 'recon_norm_std': 0.046310001704841855, 'target_norm': 1.622337943315506, 'recon_cos': 0.6579911947250366, 'gradnorm/compressor.query': 0.07427496412519348, 'gradnorm/compressor.layer0': 1.9833534795341634, 'gradnorm/compressor.layer1': 52.0404857606376, 'gradnorm/compressor.layer2': 12.533834214719386, 'gradnorm/compressor.layer3': 6.512025712620916, 'gradnorm/compressor.layer4': 2.5338336333448823, 'gradnorm/compressor.layer5': 1.5789419646756244, 'gradnorm/compressor.layer6': 0.8516038307169184, 'gradnorm/compressor.layer7': 1.0371604087902757, 'gradnorm/decoder.layer0': 1.4057822827196202, 'gradnorm/decoder.layer1': 0.9481808989309335, 'gradnorm/decoder.layer2': 0.46765331087676865, 'gradnorm/max_param': 36.708534240722656, 'epoch': 5.11}

 51%|█████     | 9750/19090 [4:44:28<4:59:44,  1.93s/it]
 51%|█████     | 9751/19090 [4:44:29<4:29:54,  1.73s/it]
 51%|█████     | 9752/19090 [4:44:31<4:15:03,  1.64s/it]
 51%|█████     | 9753/19090 [4:44:33<4:59:22,  1.92s/it]
 51%|█████     | 9754/19090 [4:44:35<4:29:35,  1.73s/it]
 51%|█████     | 9755/19090 [4:44:36<4:24:15,  1.70s/it]
 51%|█████     | 9756/19090 [4:44:39<4:55:18,  1.90s/it]
 51%|█████     | 9757/19090 [4:44:40<4:26:51,  1.72s/it]
 51%|█████     | 9758/19090 [4:44:42<4:47:21,  1.85s/it]
 51%|█████     | 9759/19090 [4:44:44<4:51:54,  1.88s/it]
 51%|█████     | 9760/19090 [4:44:45<4:24:08,  1.70s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 58.037715911865234, 'learning_rate': 2.8374727016015633e-05, 'recon_loss': 0.0014841113268630579, 'recon_static': 0.000516607535246294, 'recon_motion': 0.0009675038381828927, 'recon_norm': 1.2719309419393539, 'recon_norm_std': 0.045404378324747086, 'target_norm': 1.6226186394691466, 'recon_cos': 0.6526641726493836, 'gradnorm/compressor.query': 0.09146142602530852, 'gradnorm/compressor.layer0': 2.503636766711846, 'gradnorm/compressor.layer1': 55.89164714609085, 'gradnorm/compressor.layer2': 13.199140139230247, 'gradnorm/compressor.layer3': 6.929429317167547, 'gradnorm/compressor.layer4': 2.750626752697248, 'gradnorm/compressor.layer5': 1.7049989802114172, 'gradnorm/compressor.layer6': 0.8875725958240366, 'gradnorm/compressor.layer7': 1.0553997319355433, 'gradnorm/decoder.layer0': 1.610562969370681, 'gradnorm/decoder.layer1': 0.9053263067549598, 'gradnorm/decoder.layer2': 0.4486290751771948, 'gradnorm/max_param': 39.41918182373047, 'epoch': 5.11}

 51%|█████     | 9760/19090 [4:44:45<4:24:08,  1.70s/it]
 51%|█████     | 9761/19090 [4:44:47<4:47:32,  1.85s/it]
 51%|█████     | 9762/19090 [4:44:49<4:38:00,  1.79s/it]
 51%|█████     | 9763/19090 [4:44:50<4:14:26,  1.64s/it]
 51%|█████     | 9764/19090 [4:44:53<4:50:57,  1.87s/it]
 51%|█████     | 9765/19090 [4:44:54<4:23:24,  1.69s/it]
 51%|█████     | 9766/19090 [4:44:55<4:04:10,  1.57s/it]
 51%|█████     | 9767/19090 [4:44:58<4:41:29,  1.81s/it]
 51%|█████     | 9768/19090 [4:44:59<4:17:43,  1.66s/it]
 51%|█████     | 9769/19090 [4:45:00<4:00:00,  1.54s/it]
 51%|█████     | 9770/19090 [4:45:03<4:51:40,  1.88s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 58.227149963378906, 'learning_rate': 2.8329426627309234e-05, 'recon_loss': 0.001475091726751998, 'recon_static': 0.000517164762277389, 'recon_motion': 0.0009579269972164184, 'recon_norm': 1.2765274107456208, 'recon_norm_std': 0.046583943907171486, 'target_norm': 1.6184475630521775, 'recon_cos': 0.6537410125136376, 'gradnorm/compressor.query': 0.08164179404119877, 'gradnorm/compressor.layer0': 2.247674588051387, 'gradnorm/compressor.layer1': 55.59359629686453, 'gradnorm/compressor.layer2': 14.787986849254057, 'gradnorm/compressor.layer3': 7.841658476751944, 'gradnorm/compressor.layer4': 2.611551439299058, 'gradnorm/compressor.layer5': 1.6059823432892366, 'gradnorm/compressor.layer6': 0.8397776732954328, 'gradnorm/compressor.layer7': 1.0176516339774146, 'gradnorm/decoder.layer0': 1.5491822885290294, 'gradnorm/decoder.layer1': 0.8688123104335918, 'gradnorm/decoder.layer2': 0.47255549657460616, 'gradnorm/max_param': 39.21176528930664, 'epoch': 5.12}

 51%|█████     | 9770/19090 [4:45:03<4:51:40,  1.88s/it]
 51%|█████     | 9771/19090 [4:45:04<4:27:55,  1.73s/it]
 51%|█████     | 9772/19090 [4:45:06<4:07:13,  1.59s/it]
 51%|█████     | 9773/19090 [4:45:08<4:55:30,  1.90s/it]
 51%|█████     | 9774/19090 [4:45:10<4:40:19,  1.81s/it]
 51%|█████     | 9775/19090 [4:45:11<4:16:06,  1.65s/it]
 51%|█████     | 9776/19090 [4:45:14<4:53:13,  1.89s/it]
 51%|█████     | 9777/19090 [4:45:15<4:39:01,  1.80s/it]
 51%|█████     | 9778/19090 [4:45:16<4:14:59,  1.64s/it]
 51%|█████     | 9779/19090 [4:45:19<4:42:43,  1.82s/it]
 51%|█████     | 9780/19090 [4:45:21<4:45:39,  1.84s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 58.4595832824707, 'learning_rate': 2.828411510662848e-05, 'recon_loss': 0.0014647849864559248, 'recon_static': 0.0005050122839747928, 'recon_motion': 0.0009597726981155575, 'recon_norm': 1.2666895776987075, 'recon_norm_std': 0.046557548362761735, 'target_norm': 1.6192956060171126, 'recon_cos': 0.6559957340359688, 'gradnorm/compressor.query': 0.1087920585712865, 'gradnorm/compressor.layer0': 2.5766451200675786, 'gradnorm/compressor.layer1': 55.93904720276459, 'gradnorm/compressor.layer2': 14.4162615827565, 'gradnorm/compressor.layer3': 7.7383648070167, 'gradnorm/compressor.layer4': 2.555395559010913, 'gradnorm/compressor.layer5': 1.6065004472633972, 'gradnorm/compressor.layer6': 0.9582745630791986, 'gradnorm/compressor.layer7': 1.0247122127976722, 'gradnorm/decoder.layer0': 1.3630120303965767, 'gradnorm/decoder.layer1': 0.8873034998461538, 'gradnorm/decoder.layer2': 0.5092728962982939, 'gradnorm/max_param': 39.453819274902344, 'epoch': 5.12}

 51%|█████     | 9780/19090 [4:45:21<4:45:39,  1.84s/it]
 51%|█████     | 9781/19090 [4:45:22<4:19:41,  1.67s/it]
 51%|█████     | 9782/19090 [4:45:24<4:57:19,  1.92s/it]
 51%|█████     | 9783/19090 [4:45:26<4:54:33,  1.90s/it]
 51%|█████▏    | 9784/19090 [4:45:27<4:26:21,  1.72s/it]
 51%|█████▏    | 9785/19090 [4:45:30<4:49:26,  1.87s/it]
 51%|█████▏    | 9786/19090 [4:45:32<4:52:48,  1.89s/it]
 51%|█████▏    | 9787/19090 [4:45:33<4:25:10,  1.71s/it]
 51%|█████▏    | 9788/19090 [4:45:35<4:44:29,  1.83s/it]
 51%|█████▏    | 9789/19090 [4:45:37<4:47:35,  1.86s/it]
 51%|█████▏    | 9790/19090 [4:45:38<4:21:08,  1.68s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 48.69066619873047, 'learning_rate': 2.823879260547293e-05, 'recon_loss': 0.0014520804397761822, 'recon_static': 0.00049812226643553, 'recon_motion': 0.0009539581777062267, 'recon_norm': 1.266334316134453, 'recon_norm_std': 0.04772891756147146, 'target_norm': 1.6206355005502702, 'recon_cos': 0.6602747321128846, 'gradnorm/compressor.query': 0.06709640559683029, 'gradnorm/compressor.layer0': 2.0059833432587113, 'gradnorm/compressor.layer1': 46.62130305383474, 'gradnorm/compressor.layer2': 11.497633317314223, 'gradnorm/compressor.layer3': 7.022866910133722, 'gradnorm/compressor.layer4': 2.3390321167709804, 'gradnorm/compressor.layer5': 1.4167235400433509, 'gradnorm/compressor.layer6': 0.7978145819583419, 'gradnorm/compressor.layer7': 0.9657569609906386, 'gradnorm/decoder.layer0': 1.3327842664690976, 'gradnorm/decoder.layer1': 0.8021596071381909, 'gradnorm/decoder.layer2': 0.47235308926580505, 'gradnorm/max_param': 32.8812141418457, 'epoch': 5.13}

 51%|█████▏    | 9790/19090 [4:45:38<4:21:08,  1.68s/it]
 51%|█████▏    | 9791/19090 [4:45:40<4:33:03,  1.76s/it]
 51%|█████▏    | 9792/19090 [4:45:42<4:47:29,  1.86s/it]
 51%|█████▏    | 9793/19090 [4:45:44<4:20:51,  1.68s/it]
 51%|█████▏    | 9794/19090 [4:45:45<4:32:57,  1.76s/it]
 51%|█████▏    | 9795/19090 [4:45:47<4:36:19,  1.78s/it]
 51%|█████▏    | 9796/19090 [4:45:49<4:12:56,  1.63s/it]
 51%|█████▏    | 9797/19090 [4:45:51<4:25:56,  1.72s/it]
 51%|█████▏    | 9798/19090 [4:45:53<4:44:50,  1.84s/it]
 51%|█████▏    | 9799/19090 [4:45:54<4:19:06,  1.67s/it]
 51%|█████▏    | 9800/19090 [4:45:56<4:27:29,  1.73s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 53.97389602661133, 'learning_rate': 2.8193459275378876e-05, 'recon_loss': 0.0014466669206740334, 'recon_static': 0.0004916374004096724, 'recon_motion': 0.0009550295595545322, 'recon_norm': 1.258605459332466, 'recon_norm_std': 0.04608631543815136, 'target_norm': 1.6190530568361283, 'recon_cos': 0.6601735904812813, 'gradnorm/compressor.query': 0.07918304202892612, 'gradnorm/compressor.layer0': 2.025527875879361, 'gradnorm/compressor.layer1': 51.99383977934309, 'gradnorm/compressor.layer2': 12.525581733153405, 'gradnorm/compressor.layer3': 5.9935850182382175, 'gradnorm/compressor.layer4': 2.507195702874226, 'gradnorm/compressor.layer5': 1.4392882295361393, 'gradnorm/compressor.layer6': 0.8023317167128201, 'gradnorm/compressor.layer7': 1.0366756013401397, 'gradnorm/decoder.layer0': 1.36218659357372, 'gradnorm/decoder.layer1': 0.8632166911707816, 'gradnorm/decoder.layer2': 0.47221973849403837, 'gradnorm/max_param': 36.67533493041992, 'epoch': 5.13}

 51%|█████▏    | 9800/19090 [4:45:56<4:27:29,  1.73s/it]
 51%|█████▏    | 9801/19090 [4:45:58<4:33:04,  1.76s/it]
 51%|█████▏    | 9802/19090 [4:45:59<4:10:55,  1.62s/it]
 51%|█████▏    | 9803/19090 [4:46:01<4:25:21,  1.71s/it]
 51%|█████▏    | 9804/19090 [4:46:04<5:09:32,  2.00s/it]
 51%|█████▏    | 9805/19090 [4:46:05<4:37:00,  1.79s/it]
 51%|█████▏    | 9806/19090 [4:46:06<4:23:17,  1.70s/it]
 51%|█████▏    | 9807/19090 [4:46:09<5:06:17,  1.98s/it]
 51%|█████▏    | 9808/19090 [4:46:10<4:33:56,  1.77s/it]
 51%|█████▏    | 9809/19090 [4:46:12<4:27:44,  1.73s/it]
 51%|█████▏    | 9810/19090 [4:46:14<5:03:40,  1.96s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 64.25648498535156, 'learning_rate': 2.8148115267918785e-05, 'recon_loss': 0.001447659233235754, 'recon_static': 0.00047511915690847675, 'recon_motion': 0.0009725401410833001, 'recon_norm': 1.2580988943576812, 'recon_norm_std': 0.045678882859647274, 'target_norm': 1.620015200972557, 'recon_cos': 0.6607472091913223, 'gradnorm/compressor.query': 0.07972869287529408, 'gradnorm/compressor.layer0': 2.1892889065031356, 'gradnorm/compressor.layer1': 62.07020777602871, 'gradnorm/compressor.layer2': 14.391396898407743, 'gradnorm/compressor.layer3': 7.01966819565914, 'gradnorm/compressor.layer4': 2.7356137330849664, 'gradnorm/compressor.layer5': 1.7139160149294626, 'gradnorm/compressor.layer6': 0.8613860787684082, 'gradnorm/compressor.layer7': 1.0627373822884838, 'gradnorm/decoder.layer0': 1.3551191159358036, 'gradnorm/decoder.layer1': 0.8254695639749788, 'gradnorm/decoder.layer2': 0.43247993996024553, 'gradnorm/max_param': 43.7885856628418, 'epoch': 5.14}

 51%|█████▏    | 9810/19090 [4:46:14<5:03:40,  1.96s/it]
 51%|█████▏    | 9811/19090 [4:46:16<4:32:14,  1.76s/it]
 51%|█████▏    | 9812/19090 [4:46:17<4:33:28,  1.77s/it]
 51%|█████▏    | 9813/19090 [4:46:20<5:09:26,  2.00s/it]
 51%|█████▏    | 9814/19090 [4:46:21<4:36:38,  1.79s/it]
 51%|█████▏    | 9815/19090 [4:46:23<4:21:10,  1.69s/it]
 51%|█████▏    | 9816/19090 [4:46:25<4:57:54,  1.93s/it]
 51%|█████▏    | 9817/19090 [4:46:26<4:27:55,  1.73s/it]
 51%|█████▏    | 9818/19090 [4:46:28<4:07:09,  1.60s/it]
 51%|█████▏    | 9819/19090 [4:46:31<5:22:25,  2.09s/it]
 51%|█████▏    | 9820/19090 [4:46:32<4:45:06,  1.85s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 51.56928634643555, 'learning_rate': 2.8102760734700856e-05, 'recon_loss': 0.0014207681873813272, 'recon_static': 0.0004551602665742394, 'recon_motion': 0.0009656078924308531, 'recon_norm': 1.2437574207782744, 'recon_norm_std': 0.047179513797163966, 'target_norm': 1.6196521669626236, 'recon_cos': 0.666807834804058, 'gradnorm/compressor.query': 0.07815039937078384, 'gradnorm/compressor.layer0': 2.037879454494555, 'gradnorm/compressor.layer1': 49.40824879337584, 'gradnorm/compressor.layer2': 12.654167098086269, 'gradnorm/compressor.layer3': 6.443346646224923, 'gradnorm/compressor.layer4': 2.4277908429110213, 'gradnorm/compressor.layer5': 1.5367335013848578, 'gradnorm/compressor.layer6': 0.7933459958366887, 'gradnorm/compressor.layer7': 0.9636820544821253, 'gradnorm/decoder.layer0': 1.3234013648206115, 'gradnorm/decoder.layer1': 0.8114608207994236, 'gradnorm/decoder.layer2': 0.43394071426311015, 'gradnorm/max_param': 34.849517822265625, 'epoch': 5.14}

 51%|█████▏    | 9820/19090 [4:46:32<4:45:06,  1.85s/it]
 51%|█████▏    | 9821/19090 [4:46:34<4:18:58,  1.68s/it]
 51%|█████▏    | 9822/19090 [4:46:36<4:51:53,  1.89s/it]
 51%|█████▏    | 9823/19090 [4:46:37<4:23:43,  1.71s/it]
 51%|█████▏    | 9824/19090 [4:46:39<4:12:36,  1.64s/it]
 51%|█████▏    | 9825/19090 [4:46:41<5:02:27,  1.96s/it]
 51%|█████▏    | 9826/19090 [4:46:43<4:31:04,  1.76s/it]
 51%|█████▏    | 9827/19090 [4:46:44<4:14:21,  1.65s/it]
 51%|█████▏    | 9828/19090 [4:46:47<5:01:13,  1.95s/it]
 51%|█████▏    | 9829/19090 [4:46:48<4:30:13,  1.75s/it]
 51%|█████▏    | 9830/19090 [4:46:49<4:08:31,  1.61s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 60.45108413696289, 'learning_rate': 2.805739582736846e-05, 'recon_loss': 0.0014346241019666194, 'recon_static': 0.0004705793005996384, 'recon_motion': 0.0009640448028221726, 'recon_norm': 1.256475669145584, 'recon_norm_std': 0.04595823287963867, 'target_norm': 1.6199293643236161, 'recon_cos': 0.6639113843441009, 'gradnorm/compressor.query': 0.1226003545306468, 'gradnorm/compressor.layer0': 2.4877732745019046, 'gradnorm/compressor.layer1': 57.60756328518201, 'gradnorm/compressor.layer2': 16.011294193413814, 'gradnorm/compressor.layer3': 7.696648589198758, 'gradnorm/compressor.layer4': 2.650777407642551, 'gradnorm/compressor.layer5': 1.6082967460832327, 'gradnorm/compressor.layer6': 0.8592931535066075, 'gradnorm/compressor.layer7': 1.002061923479709, 'gradnorm/decoder.layer0': 1.293509194635212, 'gradnorm/decoder.layer1': 0.826710122091736, 'gradnorm/decoder.layer2': 0.4309629583826401, 'gradnorm/max_param': 40.632781982421875, 'epoch': 5.15}

 51%|█████▏    | 9830/19090 [4:46:49<4:08:31,  1.61s/it]
 51%|█████▏    | 9831/19090 [4:46:52<4:53:37,  1.90s/it]
 52%|█████▏    | 9832/19090 [4:46:53<4:25:55,  1.72s/it]
 52%|█████▏    | 9833/19090 [4:46:55<4:06:59,  1.60s/it]
 52%|█████▏    | 9834/19090 [4:46:57<5:08:50,  2.00s/it]
 52%|█████▏    | 9835/19090 [4:46:59<4:35:37,  1.79s/it]
 52%|█████▏    | 9836/19090 [4:47:00<4:12:23,  1.64s/it]
 52%|█████▏    | 9837/19090 [4:47:03<5:00:51,  1.95s/it]
 52%|█████▏    | 9838/19090 [4:47:04<4:30:20,  1.75s/it]
 52%|█████▏    | 9839/19090 [4:47:05<4:08:43,  1.61s/it]
 52%|█████▏    | 9840/19090 [4:47:08<5:04:20,  1.97s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 59.305294036865234, 'learning_rate': 2.8012020697599666e-05, 'recon_loss': 0.001437873370014131, 'recon_static': 0.00046686825226061046, 'recon_motion': 0.0009710051381262019, 'recon_norm': 1.2489686399698257, 'recon_norm_std': 0.04643789026886225, 'target_norm': 1.6217936038970948, 'recon_cos': 0.6633282735943794, 'gradnorm/compressor.query': 0.08032914975704905, 'gradnorm/compressor.layer0': 2.0798131066501506, 'gradnorm/compressor.layer1': 56.93751286360996, 'gradnorm/compressor.layer2': 14.134919321095087, 'gradnorm/compressor.layer3': 7.581238286071432, 'gradnorm/compressor.layer4': 2.4280616742177146, 'gradnorm/compressor.layer5': 1.6563252845784069, 'gradnorm/compressor.layer6': 0.8577402692446867, 'gradnorm/compressor.layer7': 0.9885671149389291, 'gradnorm/decoder.layer0': 1.5349377192682674, 'gradnorm/decoder.layer1': 0.8422962309620657, 'gradnorm/decoder.layer2': 0.46555315909999234, 'gradnorm/max_param': 40.162315368652344, 'epoch': 5.15}

 52%|█████▏    | 9840/19090 [4:47:08<5:04:20,  1.97s/it]
 52%|█████▏    | 9841/19090 [4:47:09<4:32:39,  1.77s/it]
 52%|█████▏    | 9842/19090 [4:47:11<4:10:21,  1.62s/it]
 52%|█████▏    | 9843/19090 [4:47:13<4:54:42,  1.91s/it]
 52%|█████▏    | 9844/19090 [4:47:15<4:25:41,  1.72s/it]
 52%|█████▏    | 9845/19090 [4:47:16<4:05:25,  1.59s/it]
 52%|█████▏    | 9846/19090 [4:47:19<5:00:10,  1.95s/it]
 52%|█████▏    | 9847/19090 [4:47:20<4:29:22,  1.75s/it]
 52%|█████▏    | 9848/19090 [4:47:21<4:07:57,  1.61s/it]
 52%|█████▏    | 9849/19090 [4:47:24<5:07:12,  1.99s/it]
 52%|█████▏    | 9850/19090 [4:47:25<4:34:27,  1.78s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 65.80941009521484, 'learning_rate': 2.7966635497106713e-05, 'recon_loss': 0.0014356381085235625, 'recon_static': 0.00047548067013849506, 'recon_motion': 0.0009601573896361515, 'recon_norm': 1.255361557006836, 'recon_norm_std': 0.046672886796295644, 'target_norm': 1.6218619227409363, 'recon_cos': 0.6643530905246735, 'gradnorm/compressor.query': 0.08668007589685549, 'gradnorm/compressor.layer0': 2.2873146232159214, 'gradnorm/compressor.layer1': 63.40995399481851, 'gradnorm/compressor.layer2': 15.26038194810826, 'gradnorm/compressor.layer3': 7.447737884154118, 'gradnorm/compressor.layer4': 2.7112015037719304, 'gradnorm/compressor.layer5': 1.6956019703770138, 'gradnorm/compressor.layer6': 0.9658086176242309, 'gradnorm/compressor.layer7': 1.1849688608770144, 'gradnorm/decoder.layer0': 1.657959110202147, 'gradnorm/decoder.layer1': 0.9563423635717491, 'gradnorm/decoder.layer2': 0.49220419658617987, 'gradnorm/max_param': 44.7244873046875, 'epoch': 5.16}

 52%|█████▏    | 9850/19090 [4:47:25<4:34:27,  1.78s/it]
 52%|█████▏    | 9851/19090 [4:47:27<4:11:40,  1.63s/it]
 52%|█████▏    | 9852/19090 [4:47:30<5:13:31,  2.04s/it]
 52%|█████▏    | 9853/19090 [4:47:31<4:38:48,  1.81s/it]
 52%|█████▏    | 9854/19090 [4:47:32<4:14:40,  1.65s/it]
 52%|█████▏    | 9855/19090 [4:47:35<5:12:51,  2.03s/it]
 52%|█████▏    | 9856/19090 [4:47:36<4:38:20,  1.81s/it]
 52%|█████▏    | 9857/19090 [4:47:38<4:14:22,  1.65s/it]
 52%|█████▏    | 9858/19090 [4:47:41<5:08:46,  2.01s/it]
 52%|█████▏    | 9859/19090 [4:47:42<4:36:01,  1.79s/it]
 52%|█████▏    | 9860/19090 [4:47:43<4:12:32,  1.64s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 70.28141784667969, 'learning_rate': 2.7921240377635517e-05, 'recon_loss': 0.0014504380320431664, 'recon_static': 0.0004884546884568408, 'recon_motion': 0.0009619833988836035, 'recon_norm': 1.2595831006765366, 'recon_norm_std': 0.0471430990844965, 'target_norm': 1.6187470346689223, 'recon_cos': 0.6590372323989868, 'gradnorm/compressor.query': 0.10627731137014587, 'gradnorm/compressor.layer0': 2.66841436772562, 'gradnorm/compressor.layer1': 67.87456117491664, 'gradnorm/compressor.layer2': 15.684927558476272, 'gradnorm/compressor.layer3': 7.966828956174681, 'gradnorm/compressor.layer4': 2.779403845309466, 'gradnorm/compressor.layer5': 1.6002282516272197, 'gradnorm/compressor.layer6': 0.9473001690101212, 'gradnorm/compressor.layer7': 1.054291657939835, 'gradnorm/decoder.layer0': 1.5842893708995334, 'gradnorm/decoder.layer1': 0.924719866011823, 'gradnorm/decoder.layer2': 0.49262593435892404, 'gradnorm/max_param': 47.87986373901367, 'epoch': 5.17}

 52%|█████▏    | 9860/19090 [4:47:43<4:12:32,  1.64s/it]
 52%|█████▏    | 9861/19090 [4:47:46<5:04:40,  1.98s/it]
 52%|█████▏    | 9862/19090 [4:47:47<4:32:51,  1.77s/it]
 52%|█████▏    | 9863/19090 [4:47:48<4:10:18,  1.63s/it]
 52%|█████▏    | 9864/19090 [4:47:51<5:05:41,  1.99s/it]
 52%|█████▏    | 9865/19090 [4:47:53<4:33:16,  1.78s/it]
 52%|█████▏    | 9866/19090 [4:47:54<4:10:40,  1.63s/it]
 52%|█████▏    | 9867/19090 [4:47:56<4:45:14,  1.86s/it]
 52%|█████▏    | 9868/19090 [4:47:58<4:18:55,  1.68s/it]
 52%|█████▏    | 9869/19090 [4:47:59<4:00:37,  1.57s/it]
 52%|█████▏    | 9870/19090 [4:48:02<5:01:07,  1.96s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 63.96714782714844, 'learning_rate': 2.787583549096516e-05, 'recon_loss': 0.0014383404748514294, 'recon_static': 0.00047741058588144367, 'recon_motion': 0.0009609299464500509, 'recon_norm': 1.2547375798225402, 'recon_norm_std': 0.04496178245171904, 'target_norm': 1.6194444209337235, 'recon_cos': 0.6623833820223808, 'gradnorm/compressor.query': 0.08044630983098788, 'gradnorm/compressor.layer0': 2.20109670353286, 'gradnorm/compressor.layer1': 61.97544522265955, 'gradnorm/compressor.layer2': 13.699776774348996, 'gradnorm/compressor.layer3': 6.593182724690869, 'gradnorm/compressor.layer4': 2.6602843221567753, 'gradnorm/compressor.layer5': 1.6023450474801244, 'gradnorm/compressor.layer6': 1.0114230751206919, 'gradnorm/compressor.layer7': 1.0203003302837115, 'gradnorm/decoder.layer0': 1.4532730683392843, 'gradnorm/decoder.layer1': 0.9024844699802044, 'gradnorm/decoder.layer2': 0.44639202240785236, 'gradnorm/max_param': 43.72103500366211, 'epoch': 5.17}

 52%|█████▏    | 9870/19090 [4:48:02<5:01:07,  1.96s/it]
 52%|█████▏    | 9871/19090 [4:48:03<4:29:58,  1.76s/it]
 52%|█████▏    | 9872/19090 [4:48:04<4:08:18,  1.62s/it]
 52%|█████▏    | 9873/19090 [4:48:07<5:10:31,  2.02s/it]
 52%|█████▏    | 9874/19090 [4:48:09<4:36:38,  1.80s/it]
 52%|█████▏    | 9875/19090 [4:48:10<4:12:59,  1.65s/it]
 52%|█████▏    | 9876/19090 [4:48:13<5:09:00,  2.01s/it]
 52%|█████▏    | 9877/19090 [4:48:14<4:35:25,  1.79s/it]
 52%|█████▏    | 9878/19090 [4:48:15<4:11:55,  1.64s/it]
 52%|█████▏    | 9879/19090 [4:48:18<4:49:35,  1.89s/it]
 52%|█████▏    | 9880/19090 [4:48:19<4:21:47,  1.71s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 58.983428955078125, 'learning_rate': 2.783042098890737e-05, 'recon_loss': 0.001452367944875732, 'recon_static': 0.00048786618062877094, 'recon_motion': 0.0009645017358707264, 'recon_norm': 1.2606962025165558, 'recon_norm_std': 0.0461875194683671, 'target_norm': 1.6223826259374619, 'recon_cos': 0.6607156068086624, 'gradnorm/compressor.query': 0.07347794735286389, 'gradnorm/compressor.layer0': 2.1810489900763432, 'gradnorm/compressor.layer1': 56.866349788616205, 'gradnorm/compressor.layer2': 13.054322233631748, 'gradnorm/compressor.layer3': 7.304649638521531, 'gradnorm/compressor.layer4': 2.9039773566464047, 'gradnorm/compressor.layer5': 1.6802632609228472, 'gradnorm/compressor.layer6': 0.8621929783247824, 'gradnorm/compressor.layer7': 1.0311363041942871, 'gradnorm/decoder.layer0': 1.569431798128173, 'gradnorm/decoder.layer1': 0.9759305511432433, 'gradnorm/decoder.layer2': 0.49876452334012156, 'gradnorm/max_param': 40.116310119628906, 'epoch': 5.18}

 52%|█████▏    | 9880/19090 [4:48:19<4:21:47,  1.71s/it]
 52%|█████▏    | 9881/19090 [4:48:20<4:02:34,  1.58s/it]
 52%|█████▏    | 9882/19090 [4:48:23<4:55:42,  1.93s/it]
 52%|█████▏    | 9883/19090 [4:48:24<4:26:14,  1.74s/it]
 52%|█████▏    | 9884/19090 [4:48:26<4:05:41,  1.60s/it]
 52%|█████▏    | 9885/19090 [4:48:28<5:04:41,  1.99s/it]
 52%|█████▏    | 9886/19090 [4:48:30<4:32:35,  1.78s/it]
 52%|█████▏    | 9887/19090 [4:48:31<4:10:08,  1.63s/it]
 52%|█████▏    | 9888/19090 [4:48:34<5:10:58,  2.03s/it]
 52%|█████▏    | 9889/19090 [4:48:35<4:37:17,  1.81s/it]
 52%|█████▏    | 9890/19090 [4:48:37<4:13:37,  1.65s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 63.469879150390625, 'learning_rate': 2.7784997023306038e-05, 'recon_loss': 0.001442189159570262, 'recon_static': 0.0004772784923261497, 'recon_motion': 0.0009649106636061333, 'recon_norm': 1.2506888031959533, 'recon_norm_std': 0.04671641830354929, 'target_norm': 1.620509645342827, 'recon_cos': 0.6615875571966171, 'gradnorm/compressor.query': 0.07548024639386498, 'gradnorm/compressor.layer0': 2.11751787486633, 'gradnorm/compressor.layer1': 61.406351383504095, 'gradnorm/compressor.layer2': 13.910519336738417, 'gradnorm/compressor.layer3': 6.68413252138799, 'gradnorm/compressor.layer4': 2.679862147387026, 'gradnorm/compressor.layer5': 1.7038952546812174, 'gradnorm/compressor.layer6': 0.8341739626656534, 'gradnorm/compressor.layer7': 1.0966429462205833, 'gradnorm/decoder.layer0': 1.373095797756511, 'gradnorm/decoder.layer1': 0.9510727660382201, 'gradnorm/decoder.layer2': 0.49288749871598014, 'gradnorm/max_param': 43.32048797607422, 'epoch': 5.18}

 52%|█████▏    | 9890/19090 [4:48:37<4:13:37,  1.65s/it]
 52%|█████▏    | 9891/19090 [4:48:39<4:56:17,  1.93s/it]
 52%|█████▏    | 9892/19090 [4:48:40<4:26:24,  1.74s/it]
 52%|█████▏    | 9893/19090 [4:48:42<4:05:51,  1.60s/it]
 52%|█████▏    | 9894/19090 [4:48:45<5:06:05,  2.00s/it]
 52%|█████▏    | 9895/19090 [4:48:46<4:33:21,  1.78s/it]
 52%|█████▏    | 9896/19090 [4:48:47<4:10:27,  1.63s/it]
 52%|█████▏    | 9897/19090 [4:48:50<4:49:10,  1.89s/it]
 52%|█████▏    | 9898/19090 [4:48:51<4:21:33,  1.71s/it]
 52%|█████▏    | 9899/19090 [4:48:52<4:02:15,  1.58s/it]
 52%|█████▏    | 9900/19090 [4:48:55<4:54:37,  1.92s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 58.45686340332031, 'learning_rate': 2.773956374603668e-05, 'recon_loss': 0.0014471938600763678, 'recon_static': 0.0004790340048202779, 'recon_motion': 0.000968159845797345, 'recon_norm': 1.257390734553337, 'recon_norm_std': 0.046152699086815116, 'target_norm': 1.6213312774896622, 'recon_cos': 0.6613918110728264, 'gradnorm/compressor.query': 0.09624980954361935, 'gradnorm/compressor.layer0': 2.605507914693462, 'gradnorm/compressor.layer1': 55.933554330115655, 'gradnorm/compressor.layer2': 14.967330960921188, 'gradnorm/compressor.layer3': 6.661812763662433, 'gradnorm/compressor.layer4': 2.5865798567267566, 'gradnorm/compressor.layer5': 1.6398803474228008, 'gradnorm/compressor.layer6': 0.826558665102624, 'gradnorm/compressor.layer7': 0.9708175438236804, 'gradnorm/decoder.layer0': 1.2383246669919554, 'gradnorm/decoder.layer1': 0.8591306859303782, 'gradnorm/decoder.layer2': 0.40815716352390474, 'gradnorm/max_param': 39.44865417480469, 'epoch': 5.19}

 52%|█████▏    | 9900/19090 [4:48:55<4:54:37,  1.92s/it]
 52%|█████▏    | 9901/19090 [4:48:56<4:25:18,  1.73s/it]
 52%|█████▏    | 9902/19090 [4:48:58<4:04:30,  1.60s/it]
 52%|█████▏    | 9903/19090 [4:49:00<5:01:17,  1.97s/it]
 52%|█████▏    | 9904/19090 [4:49:02<4:29:39,  1.76s/it]
 52%|█████▏    | 9905/19090 [4:49:03<4:07:40,  1.62s/it]
 52%|█████▏    | 9906/19090 [4:49:06<5:06:52,  2.00s/it]
 52%|█████▏    | 9907/19090 [4:49:07<4:33:46,  1.79s/it]
 52%|█████▏    | 9908/19090 [4:49:08<4:10:26,  1.64s/it]
 52%|█████▏    | 9909/19090 [4:49:11<5:05:00,  1.99s/it]
 52%|█████▏    | 9910/19090 [4:49:13<4:32:31,  1.78s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 62.040283203125, 'learning_rate': 2.7694121309005965e-05, 'recon_loss': 0.001473743698443286, 'recon_static': 0.0005115170737553853, 'recon_motion': 0.0009622265817597509, 'recon_norm': 1.2684689551591872, 'recon_norm_std': 0.04515220569446683, 'target_norm': 1.6218430370092392, 'recon_cos': 0.6548015534877777, 'gradnorm/compressor.query': 0.11609882267924689, 'gradnorm/compressor.layer0': 2.8886378700880515, 'gradnorm/compressor.layer1': 59.379977176259715, 'gradnorm/compressor.layer2': 15.656323095824874, 'gradnorm/compressor.layer3': 7.298761226111663, 'gradnorm/compressor.layer4': 2.9599480498111053, 'gradnorm/compressor.layer5': 1.7180061811573535, 'gradnorm/compressor.layer6': 0.8981068873586614, 'gradnorm/compressor.layer7': 1.0201636797341844, 'gradnorm/decoder.layer0': 1.3117316585630379, 'gradnorm/decoder.layer1': 0.881287329133404, 'gradnorm/decoder.layer2': 0.45541930609156633, 'gradnorm/max_param': 41.879817962646484, 'epoch': 5.19}

 52%|█████▏    | 9910/19090 [4:49:13<4:32:31,  1.78s/it]
 52%|█████▏    | 9911/19090 [4:49:14<4:09:42,  1.63s/it]
 52%|█████▏    | 9912/19090 [4:49:16<4:48:16,  1.88s/it]
 52%|█████▏    | 9913/19090 [4:49:18<4:20:38,  1.70s/it]
 52%|█████▏    | 9914/19090 [4:49:19<4:01:17,  1.58s/it]
 52%|█████▏    | 9915/19090 [4:49:22<4:58:20,  1.95s/it]
 52%|█████▏    | 9916/19090 [4:49:23<4:27:46,  1.75s/it]
 52%|█████▏    | 9917/19090 [4:49:24<4:06:18,  1.61s/it]
 52%|█████▏    | 9918/19090 [4:49:27<5:05:05,  2.00s/it]
 52%|█████▏    | 9919/19090 [4:49:28<4:32:13,  1.78s/it]
 52%|█████▏    | 9920/19090 [4:49:30<4:09:36,  1.63s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 70.05060577392578, 'learning_rate': 2.7648669864151182e-05, 'recon_loss': 0.001474899353343062, 'recon_static': 0.0005191537551581859, 'recon_motion': 0.0009557455952744931, 'recon_norm': 1.2750411540269853, 'recon_norm_std': 0.04557628221809864, 'target_norm': 1.6183026105165481, 'recon_cos': 0.6535830497741699, 'gradnorm/compressor.query': 0.10304171565908606, 'gradnorm/compressor.layer0': 2.7938709855818193, 'gradnorm/compressor.layer1': 67.48327339052248, 'gradnorm/compressor.layer2': 16.175194028371614, 'gradnorm/compressor.layer3': 8.201529327691773, 'gradnorm/compressor.layer4': 2.779513985255728, 'gradnorm/compressor.layer5': 1.7416415192949948, 'gradnorm/compressor.layer6': 1.009926109915933, 'gradnorm/compressor.layer7': 1.1235050228874133, 'gradnorm/decoder.layer0': 1.5242714971021571, 'gradnorm/decoder.layer1': 0.8957108628296045, 'gradnorm/decoder.layer2': 0.45845790456180735, 'gradnorm/max_param': 47.60417175292969, 'epoch': 5.2}

 52%|█████▏    | 9920/19090 [4:49:30<4:09:36,  1.63s/it]
 52%|█████▏    | 9921/19090 [4:49:33<5:08:11,  2.02s/it]
 52%|█████▏    | 9922/19090 [4:49:34<4:34:34,  1.80s/it]
 52%|█████▏    | 9923/19090 [4:49:35<4:11:05,  1.64s/it]
 52%|█████▏    | 9924/19090 [4:49:38<4:58:10,  1.95s/it]
 52%|█████▏    | 9925/19090 [4:49:39<4:27:45,  1.75s/it]
 52%|█████▏    | 9926/19090 [4:49:40<4:06:12,  1.61s/it]
 52%|█████▏    | 9927/19090 [4:49:43<4:51:56,  1.91s/it]
 52%|█████▏    | 9928/19090 [4:49:44<4:23:14,  1.72s/it]
 52%|█████▏    | 9929/19090 [4:49:46<4:02:57,  1.59s/it]
 52%|█████▏    | 9930/19090 [4:49:48<4:49:41,  1.90s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 61.41581726074219, 'learning_rate': 2.7603209563439725e-05, 'recon_loss': 0.0014684595895232632, 'recon_static': 0.0004921913154248614, 'recon_motion': 0.0009762683024746366, 'recon_norm': 1.2633456230163573, 'recon_norm_std': 0.04458391722291708, 'target_norm': 1.6212243109941482, 'recon_cos': 0.6561401203274727, 'gradnorm/compressor.query': 0.09009143646622662, 'gradnorm/compressor.layer0': 2.5633427234406567, 'gradnorm/compressor.layer1': 58.81934487931033, 'gradnorm/compressor.layer2': 15.186989606907371, 'gradnorm/compressor.layer3': 7.691588764340341, 'gradnorm/compressor.layer4': 2.80563618971101, 'gradnorm/compressor.layer5': 1.640644618325784, 'gradnorm/compressor.layer6': 0.9542526391012602, 'gradnorm/compressor.layer7': 1.103744618110021, 'gradnorm/decoder.layer0': 1.4800122042745578, 'gradnorm/decoder.layer1': 0.8298738475089985, 'gradnorm/decoder.layer2': 0.4825560348466982, 'gradnorm/max_param': 41.49043655395508, 'epoch': 5.2}

 52%|█████▏    | 9930/19090 [4:49:48<4:49:41,  1.90s/it]
 52%|█████▏    | 9931/19090 [4:49:50<4:21:30,  1.71s/it]
 52%|█████▏    | 9932/19090 [4:49:51<4:01:40,  1.58s/it]
 52%|█████▏    | 9933/19090 [4:49:54<4:57:33,  1.95s/it]
 52%|█████▏    | 9934/19090 [4:49:55<4:27:14,  1.75s/it]
 52%|█████▏    | 9935/19090 [4:49:56<4:06:12,  1.61s/it]
 52%|█████▏    | 9936/19090 [4:49:59<5:09:24,  2.03s/it]
 52%|█████▏    | 9937/19090 [4:50:01<4:36:53,  1.82s/it]
 52%|█████▏    | 9938/19090 [4:50:02<4:13:09,  1.66s/it]
 52%|█████▏    | 9939/19090 [4:50:05<5:01:28,  1.98s/it]
 52%|█████▏    | 9940/19090 [4:50:06<4:30:24,  1.77s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 78.08135986328125, 'learning_rate': 2.755774055886861e-05, 'recon_loss': 0.0014331854647025466, 'recon_static': 0.0004788422695128247, 'recon_motion': 0.0009543431588099338, 'recon_norm': 1.254964780807495, 'recon_norm_std': 0.04719064189121127, 'target_norm': 1.6211642503738404, 'recon_cos': 0.6648704379796981, 'gradnorm/compressor.query': 0.10205724686578402, 'gradnorm/compressor.layer0': 2.858126007954119, 'gradnorm/compressor.layer1': 75.39932430379433, 'gradnorm/compressor.layer2': 17.731041468085532, 'gradnorm/compressor.layer3': 8.477667248717685, 'gradnorm/compressor.layer4': 2.821450973201067, 'gradnorm/compressor.layer5': 1.7481873525083673, 'gradnorm/compressor.layer6': 0.9760248764260395, 'gradnorm/compressor.layer7': 1.2186141794279752, 'gradnorm/decoder.layer0': 1.623790951418717, 'gradnorm/decoder.layer1': 0.9296261422562029, 'gradnorm/decoder.layer2': 0.4956365566181091, 'gradnorm/max_param': 53.187076568603516, 'epoch': 5.21}

 52%|█████▏    | 9940/19090 [4:50:06<4:30:24,  1.77s/it]
 52%|█████▏    | 9941/19090 [4:50:07<4:09:23,  1.64s/it]
 52%|█████▏    | 9942/19090 [4:50:10<5:01:24,  1.98s/it]
 52%|█████▏    | 9943/19090 [4:50:11<4:31:08,  1.78s/it]
 52%|█████▏    | 9944/19090 [4:50:13<4:09:47,  1.64s/it]
 52%|█████▏    | 9945/19090 [4:50:15<4:53:38,  1.93s/it]
 52%|█████▏    | 9946/19090 [4:50:16<4:25:39,  1.74s/it]
 52%|█████▏    | 9947/19090 [4:50:18<4:06:09,  1.62s/it]
 52%|█████▏    | 9948/19090 [4:50:21<5:02:19,  1.98s/it]
 52%|█████▏    | 9949/19090 [4:50:22<4:31:27,  1.78s/it]
 52%|█████▏    | 9950/19090 [4:50:23<4:09:48,  1.64s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 61.174991607666016, 'learning_rate': 2.751226300246395e-05, 'recon_loss': 0.001429823343642056, 'recon_static': 0.00047414094078703786, 'recon_motion': 0.0009556824385072105, 'recon_norm': 1.2519206404685974, 'recon_norm_std': 0.046738028340041635, 'target_norm': 1.6190504878759384, 'recon_cos': 0.6642684623599052, 'gradnorm/compressor.query': 0.08445841057229207, 'gradnorm/compressor.layer0': 2.214108407265447, 'gradnorm/compressor.layer1': 59.09621227022507, 'gradnorm/compressor.layer2': 13.772309116959507, 'gradnorm/compressor.layer3': 6.3644585313647015, 'gradnorm/compressor.layer4': 2.7861035462776784, 'gradnorm/compressor.layer5': 1.5283401244223058, 'gradnorm/compressor.layer6': 0.8825091625742976, 'gradnorm/compressor.layer7': 1.0027586912826087, 'gradnorm/decoder.layer0': 1.4585262534283356, 'gradnorm/decoder.layer1': 0.8377615733677725, 'gradnorm/decoder.layer2': 0.46241879459108504, 'gradnorm/max_param': 41.688541412353516, 'epoch': 5.21}

 52%|█████▏    | 9950/19090 [4:50:23<4:09:48,  1.64s/it]
 52%|█████▏    | 9951/19090 [4:50:26<4:59:05,  1.96s/it]
 52%|█████▏    | 9952/19090 [4:50:27<4:28:24,  1.76s/it]
 52%|█████▏    | 9953/19090 [4:50:29<4:06:35,  1.62s/it]
 52%|█████▏    | 9954/19090 [4:50:32<5:10:19,  2.04s/it]
 52%|█████▏    | 9955/19090 [4:50:33<4:35:59,  1.81s/it]
 52%|█████▏    | 9956/19090 [4:50:34<4:12:04,  1.66s/it]
 52%|█████▏    | 9957/19090 [4:50:36<4:34:53,  1.81s/it]
 52%|█████▏    | 9958/19090 [4:50:38<4:11:27,  1.65s/it]
 52%|█████▏    | 9959/19090 [4:50:39<3:54:52,  1.54s/it]
 52%|█████▏    | 9960/19090 [4:50:42<5:01:30,  1.98s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 68.59886932373047, 'learning_rate': 2.7466777046280457e-05, 'recon_loss': 0.0014460899983532727, 'recon_static': 0.00048337635889765805, 'recon_motion': 0.0009627135950722732, 'recon_norm': 1.257731631398201, 'recon_norm_std': 0.046530804317444566, 'target_norm': 1.6207179516553878, 'recon_cos': 0.6612667664885521, 'gradnorm/compressor.query': 0.11201469463994854, 'gradnorm/compressor.layer0': 2.5886262589466833, 'gradnorm/compressor.layer1': 66.35013069881626, 'gradnorm/compressor.layer2': 14.850216519097037, 'gradnorm/compressor.layer3': 7.612563700793105, 'gradnorm/compressor.layer4': 3.1037949375766383, 'gradnorm/compressor.layer5': 1.8183413462661893, 'gradnorm/compressor.layer6': 0.8595968480295125, 'gradnorm/compressor.layer7': 1.0116548260263478, 'gradnorm/decoder.layer0': 1.5948016119243436, 'gradnorm/decoder.layer1': 0.9027522429906109, 'gradnorm/decoder.layer2': 0.4585710978071771, 'gradnorm/max_param': 46.80288314819336, 'epoch': 5.22}

 52%|█████▏    | 9960/19090 [4:50:42<5:01:30,  1.98s/it]
 52%|█████▏    | 9961/19090 [4:50:43<4:30:00,  1.77s/it]
 52%|█████▏    | 9962/19090 [4:50:44<4:07:33,  1.63s/it]
 52%|█████▏    | 9963/19090 [4:50:47<5:00:39,  1.98s/it]
 52%|█████▏    | 9964/19090 [4:50:49<4:29:11,  1.77s/it]
 52%|█████▏    | 9965/19090 [4:50:50<4:07:25,  1.63s/it]
 52%|█████▏    | 9966/19090 [4:50:52<4:54:05,  1.93s/it]
 52%|█████▏    | 9967/19090 [4:50:54<4:25:07,  1.74s/it]
 52%|█████▏    | 9968/19090 [4:50:55<4:04:30,  1.61s/it]
 52%|█████▏    | 9969/19090 [4:50:58<4:51:10,  1.92s/it]
 52%|█████▏    | 9970/19090 [4:50:59<4:22:26,  1.73s/it]
                                                        
{'loss': 0.0015, 'grad_norm': 67.72441101074219, 'learning_rate': 2.742128284240092e-05, 'recon_loss': 0.0014659869018942117, 'recon_static': 0.0004931937248329632, 'recon_motion': 0.0009727931523229927, 'recon_norm': 1.2620349645614624, 'recon_norm_std': 0.045397387072443965, 'target_norm': 1.6220892369747162, 'recon_cos': 0.6569676876068116, 'gradnorm/compressor.query': 0.08868611406281061, 'gradnorm/compressor.layer0': 2.6066970332036283, 'gradnorm/compressor.layer1': 64.98778806557621, 'gradnorm/compressor.layer2': 16.417175491632136, 'gradnorm/compressor.layer3': 8.357528454055753, 'gradnorm/compressor.layer4': 2.837807451303959, 'gradnorm/compressor.layer5': 1.722586286512403, 'gradnorm/compressor.layer6': 0.9840515150157864, 'gradnorm/compressor.layer7': 1.196411226224122, 'gradnorm/decoder.layer0': 1.5257879660400464, 'gradnorm/decoder.layer1': 1.0006314906351184, 'gradnorm/decoder.layer2': 0.5093664601552095, 'gradnorm/max_param': 45.843929290771484, 'epoch': 5.22}

 52%|█████▏    | 9970/19090 [4:50:59<4:22:26,  1.73s/it]
 52%|█████▏    | 9971/19090 [4:51:00<4:02:18,  1.59s/it]
 52%|█████▏    | 9972/19090 [4:51:03<4:54:00,  1.93s/it]
 52%|█████▏    | 9973/19090 [4:51:04<4:24:24,  1.74s/it]
 52%|█████▏    | 9974/19090 [4:51:06<4:03:50,  1.60s/it]
 52%|█████▏    | 9975/19090 [4:51:08<4:47:39,  1.89s/it]
 52%|█████▏    | 9976/19090 [4:51:09<4:20:05,  1.71s/it]
 52%|█████▏    | 9977/19090 [4:51:11<4:01:06,  1.59s/it]
 52%|█████▏    | 9978/19090 [4:51:13<4:50:57,  1.92s/it]
 52%|█████▏    | 9979/19090 [4:51:15<4:22:34,  1.73s/it]
 52%|█████▏    | 9980/19090 [4:51:16<4:02:35,  1.60s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 78.65445709228516, 'learning_rate': 2.7375780542935707e-05, 'recon_loss': 0.0014519658201606945, 'recon_static': 0.00048647367148078045, 'recon_motion': 0.0009654921217588708, 'recon_norm': 1.2613687634468078, 'recon_norm_std': 0.045045322831720115, 'target_norm': 1.622120228409767, 'recon_cos': 0.6607125401496887, 'gradnorm/compressor.query': 0.09379578505528889, 'gradnorm/compressor.layer0': 2.89422828231615, 'gradnorm/compressor.layer1': 75.82340966433124, 'gradnorm/compressor.layer2': 18.467299144577492, 'gradnorm/compressor.layer3': 8.407778235852874, 'gradnorm/compressor.layer4': 2.913990991446264, 'gradnorm/compressor.layer5': 1.8204617418660636, 'gradnorm/compressor.layer6': 1.0307254034933921, 'gradnorm/compressor.layer7': 1.1466620598375203, 'gradnorm/decoder.layer0': 1.3875692848571313, 'gradnorm/decoder.layer1': 0.9309037211200456, 'gradnorm/decoder.layer2': 0.4921393749384566, 'gradnorm/max_param': 53.48600769042969, 'epoch': 5.23}

 52%|█████▏    | 9980/19090 [4:51:16<4:02:35,  1.60s/it]
 52%|█████▏    | 9981/19090 [4:51:19<4:55:47,  1.95s/it]
 52%|█████▏    | 9982/19090 [4:51:20<4:25:40,  1.75s/it]
 52%|█████▏    | 9983/19090 [4:51:21<4:04:46,  1.61s/it]
 52%|█████▏    | 9984/19090 [4:51:24<4:46:26,  1.89s/it]
 52%|█████▏    | 9985/19090 [4:51:25<4:19:27,  1.71s/it]
 52%|█████▏    | 9986/19090 [4:51:26<4:00:18,  1.58s/it]
 52%|█████▏    | 9987/19090 [4:51:29<4:40:40,  1.85s/it]
 52%|█████▏    | 9988/19090 [4:51:30<4:15:39,  1.69s/it]
 52%|█████▏    | 9989/19090 [4:51:31<3:57:26,  1.57s/it]
 52%|█████▏    | 9990/19090 [4:51:34<4:50:40,  1.92s/it]
                                                        
{'loss': 0.0014, 'grad_norm': 73.04936218261719, 'learning_rate': 2.7330270300022254e-05, 'recon_loss': 0.0014465991203906014, 'recon_static': 0.0004866165269049816, 'recon_motion': 0.0009599826225894503, 'recon_norm': 1.2551122605800629, 'recon_norm_std': 0.0468683491460979, 'target_norm': 1.620462316274643, 'recon_cos': 0.6605640619993209, 'gradnorm/compressor.query': 0.07567194879836636, 'gradnorm/compressor.layer0': 2.236998185301562, 'gradnorm/compressor.layer1': 70.40085767743707, 'gradnorm/compressor.layer2': 17.28495718458819, 'gradnorm/compressor.layer3': 7.691158013626827, 'gradnorm/compressor.layer4': 2.9660759376996206, 'gradnorm/compressor.layer5': 1.6504577261923057, 'gradnorm/compressor.layer6': 0.9707726313615033, 'gradnorm/compressor.layer7': 1.1645543128385916, 'gradnorm/decoder.layer0': 1.3903639473359612, 'gradnorm/decoder.layer1': 1.0012996691743727, 'gradnorm/decoder.layer2': 0.488682040711804, 'gradnorm/max_param': 49.66279983520508, 'epoch': 5.23}

 52%|█████▏    | 9990/19090 [4:51:34<4:50:40,  1.92s/it]
 52%|█████▏    | 9991/19090 [4:51:36<4:21:51,  1.73s/it]
 52%|█████▏    | 9992/19090 [4:51:37<4:01:42,  1.59s/it]
 52%|█████▏    | 9993/19090 [4:51:39<4:37:49,  1.83s/it]
 52%|█████▏    | 9994/19090 [4:51:40<4:13:02,  1.67s/it]
 52%|█████▏    | 9995/19090 [4:51:42<3:55:44,  1.56s/it]
 52%|█████▏    | 9996/19090 [4:51:44<4:39:50,  1.85s/it]
 52%|█████▏    | 9997/19090 [4:51:46<4:14:29,  1.68s/it]
 52%|█████▏    | 9998/19090 [4:51:47<3:56:51,  1.56s/it]
 52%|█████▏    | 9999/19090 [4:51:50<5:02:18,  2.00s/it]
 52%|█████▏    | 10000/19090 [4:51:51<4:29:55,  1.78s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 68.52470397949219, 'learning_rate': 2.728475226582456e-05, 'recon_loss': 0.0014127662754617632, 'recon_static': 0.00046226416525314563, 'recon_motion': 0.0009505021211225539, 'recon_norm': 1.2471682459115982, 'recon_norm_std': 0.04646658478304744, 'target_norm': 1.6168176710605622, 'recon_cos': 0.6678825736045837, 'gradnorm/compressor.query': 0.07831722393747781, 'gradnorm/compressor.layer0': 2.2817058369088863, 'gradnorm/compressor.layer1': 65.88611742051583, 'gradnorm/compressor.layer2': 16.138252903803107, 'gradnorm/compressor.layer3': 8.35895744065185, 'gradnorm/compressor.layer4': 3.0507776550975136, 'gradnorm/compressor.layer5': 1.8181233483363073, 'gradnorm/compressor.layer6': 1.1234359995867773, 'gradnorm/compressor.layer7': 1.1958859134150979, 'gradnorm/decoder.layer0': 1.5738116701145952, 'gradnorm/decoder.layer1': 1.0341833761147627, 'gradnorm/decoder.layer2': 0.5246078445741736, 'gradnorm/max_param': 46.478599548339844, 'epoch': 5.24}

 52%|█████▏    | 10000/19090 [4:51:51<4:29:55,  1.78s/it]
 52%|█████▏    | 10001/19090 [4:51:55<6:24:09,  2.54s/it]
 52%|█████▏    | 10002/19090 [4:51:57<5:27:36,  2.16s/it]
 52%|█████▏    | 10003/19090 [4:51:58<4:47:53,  1.90s/it]
 52%|█████▏    | 10004/19090 [4:51:59<4:19:51,  1.72s/it]
 52%|█████▏    | 10005/19090 [4:52:01<4:00:34,  1.59s/it]
 52%|█████▏    | 10006/19090 [4:52:02<3:46:54,  1.50s/it]
 52%|█████▏    | 10007/19090 [4:52:03<3:37:56,  1.44s/it]
 52%|█████▏    | 10008/19090 [4:52:04<3:30:38,  1.39s/it]
 52%|█████▏    | 10009/19090 [4:52:06<3:25:57,  1.36s/it]
 52%|█████▏    | 10010/19090 [4:52:07<3:22:39,  1.34s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 63.28404235839844, 'learning_rate': 2.723922659253266e-05, 'recon_loss': 0.0014419624145375564, 'recon_static': 0.00047533628458040766, 'recon_motion': 0.0009666261859820225, 'recon_norm': 1.2550402402877807, 'recon_norm_std': 0.044912563264369966, 'target_norm': 1.6212937265634537, 'recon_cos': 0.6624092474579811, 'gradnorm/compressor.query': 0.07322906694511706, 'gradnorm/compressor.layer0': 2.2818692555664777, 'gradnorm/compressor.layer1': 60.98204301523933, 'gradnorm/compressor.layer2': 14.409974119545689, 'gradnorm/compressor.layer3': 7.488929461788419, 'gradnorm/compressor.layer4': 2.8824732304278737, 'gradnorm/compressor.layer5': 1.717460981256183, 'gradnorm/compressor.layer6': 1.0007752155535954, 'gradnorm/compressor.layer7': 1.1722629159389515, 'gradnorm/decoder.layer0': 1.5239878907360878, 'gradnorm/decoder.layer1': 0.959332839834693, 'gradnorm/decoder.layer2': 0.48978818672606544, 'gradnorm/max_param': 43.016414642333984, 'epoch': 5.24}

 52%|█████▏    | 10010/19090 [4:52:07<3:22:39,  1.34s/it]
 52%|█████▏    | 10011/19090 [4:52:09<4:03:51,  1.61s/it]
 52%|█████▏    | 10012/19090 [4:52:11<3:49:06,  1.51s/it]
 52%|█████▏    | 10013/19090 [4:52:12<3:38:47,  1.45s/it]
 52%|█████▏    | 10014/19090 [4:52:15<4:55:13,  1.95s/it]
 52%|█████▏    | 10015/19090 [4:52:16<4:25:11,  1.75s/it]
 52%|█████▏    | 10016/19090 [4:52:18<4:04:04,  1.61s/it]
 52%|█████▏    | 10017/19090 [4:52:20<5:00:30,  1.99s/it]
 52%|█████▏    | 10018/19090 [4:52:22<4:28:38,  1.78s/it]
 52%|█████▏    | 10019/19090 [4:52:23<4:06:47,  1.63s/it]
 52%|█████▏    | 10020/19090 [4:52:26<5:14:31,  2.08s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 69.06857299804688, 'learning_rate': 2.7193693432362155e-05, 'recon_loss': 0.0014557395916199312, 'recon_static': 0.0004802220602869056, 'recon_motion': 0.0009755175342434086, 'recon_norm': 1.2566443353891372, 'recon_norm_std': 0.04509106511250138, 'target_norm': 1.6221777588129043, 'recon_cos': 0.659368485212326, 'gradnorm/compressor.query': 0.09719482606616309, 'gradnorm/compressor.layer0': 2.5339128592666587, 'gradnorm/compressor.layer1': 66.54783802719477, 'gradnorm/compressor.layer2': 16.02829148504879, 'gradnorm/compressor.layer3': 7.683031225436687, 'gradnorm/compressor.layer4': 3.225462793072056, 'gradnorm/compressor.layer5': 1.8023157429265628, 'gradnorm/compressor.layer6': 1.0144646104493675, 'gradnorm/compressor.layer7': 1.0200651093814155, 'gradnorm/decoder.layer0': 1.581267598774194, 'gradnorm/decoder.layer1': 1.020038697686731, 'gradnorm/decoder.layer2': 0.47395400335269294, 'gradnorm/max_param': 46.94657897949219, 'epoch': 5.25}

 52%|█████▏    | 10020/19090 [4:52:26<5:14:31,  2.08s/it]
 52%|█████▏    | 10021/19090 [4:52:27<4:38:41,  1.84s/it]
 52%|█████▏    | 10022/19090 [4:52:29<4:13:31,  1.68s/it]
 53%|█████▎    | 10023/19090 [4:52:31<4:58:28,  1.98s/it]
 53%|█████▎    | 10024/19090 [4:52:33<4:27:16,  1.77s/it]
 53%|█████▎    | 10025/19090 [4:52:34<4:05:42,  1.63s/it]
 53%|█████▎    | 10026/19090 [4:52:37<4:54:48,  1.95s/it]
 53%|█████▎    | 10027/19090 [4:52:38<4:25:33,  1.76s/it]
 53%|█████▎    | 10028/19090 [4:52:39<4:04:20,  1.62s/it]
 53%|█████▎    | 10029/19090 [4:52:42<4:59:10,  1.98s/it]
 53%|█████▎    | 10030/19090 [4:52:43<4:28:02,  1.78s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 66.67537689208984, 'learning_rate': 2.7148152937553655e-05, 'recon_loss': 0.0014469891553744673, 'recon_static': 0.0004834083163586911, 'recon_motion': 0.0009635808310122229, 'recon_norm': 1.254188558459282, 'recon_norm_std': 0.04538004994392395, 'target_norm': 1.6220795124769212, 'recon_cos': 0.6614428028464318, 'gradnorm/compressor.query': 0.08180941661830161, 'gradnorm/compressor.layer0': 2.291673960818618, 'gradnorm/compressor.layer1': 63.967400831522184, 'gradnorm/compressor.layer2': 16.587153673617046, 'gradnorm/compressor.layer3': 7.502714810167389, 'gradnorm/compressor.layer4': 2.9791648882524924, 'gradnorm/compressor.layer5': 1.700271315521943, 'gradnorm/compressor.layer6': 0.8565793079395817, 'gradnorm/compressor.layer7': 1.0693442427081874, 'gradnorm/decoder.layer0': 1.514129215183908, 'gradnorm/decoder.layer1': 0.9792456425951381, 'gradnorm/decoder.layer2': 0.44656515450352674, 'gradnorm/max_param': 45.123966217041016, 'epoch': 5.25}

 53%|█████▎    | 10030/19090 [4:52:43<4:28:02,  1.78s/it]
 53%|█████▎    | 10031/19090 [4:52:45<4:06:10,  1.63s/it]
 53%|█████▎    | 10032/19090 [4:52:48<5:02:48,  2.01s/it]
 53%|█████▎    | 10033/19090 [4:52:49<4:30:28,  1.79s/it]
 53%|█████▎    | 10034/19090 [4:52:50<4:07:55,  1.64s/it]
 53%|█████▎    | 10035/19090 [4:52:53<5:19:46,  2.12s/it]
 53%|█████▎    | 10036/19090 [4:52:55<4:42:01,  1.87s/it]
 53%|█████▎    | 10037/19090 [4:52:56<4:15:37,  1.69s/it]
 53%|█████▎    | 10038/19090 [4:52:59<5:07:26,  2.04s/it]
 53%|█████▎    | 10039/19090 [4:53:00<4:33:03,  1.81s/it]
 53%|█████▎    | 10040/19090 [4:53:01<4:09:35,  1.65s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 76.85368347167969, 'learning_rate': 2.7102605260372304e-05, 'recon_loss': 0.0014330728736240417, 'recon_static': 0.00047198575848597104, 'recon_motion': 0.0009610871071345173, 'recon_norm': 1.2528485655784607, 'recon_norm_std': 0.045077153109014036, 'target_norm': 1.6184334337711335, 'recon_cos': 0.6633472636342048, 'gradnorm/compressor.query': 0.09189697144822706, 'gradnorm/compressor.layer0': 2.5858162566283034, 'gradnorm/compressor.layer1': 74.47978765766757, 'gradnorm/compressor.layer2': 16.318399466026456, 'gradnorm/compressor.layer3': 8.078988182243375, 'gradnorm/compressor.layer4': 3.239280334891177, 'gradnorm/compressor.layer5': 1.951400789229857, 'gradnorm/compressor.layer6': 1.0973963310965973, 'gradnorm/compressor.layer7': 1.206560357839584, 'gradnorm/decoder.layer0': 1.6625394343238187, 'gradnorm/decoder.layer1': 0.9903140666127115, 'gradnorm/decoder.layer2': 0.5374856636442007, 'gradnorm/max_param': 52.54509735107422, 'epoch': 5.26}

 53%|█████▎    | 10040/19090 [4:53:01<4:09:35,  1.65s/it]
 53%|█████▎    | 10041/19090 [4:53:04<4:57:47,  1.97s/it]
 53%|█████▎    | 10042/19090 [4:53:05<4:26:48,  1.77s/it]
 53%|█████▎    | 10043/19090 [4:53:07<4:04:50,  1.62s/it]
 53%|█████▎    | 10044/19090 [4:53:10<5:03:56,  2.02s/it]
 53%|█████▎    | 10045/19090 [4:53:11<4:31:46,  1.80s/it]
 53%|█████▎    | 10046/19090 [4:53:12<4:08:54,  1.65s/it]
 53%|█████▎    | 10047/19090 [4:53:15<4:59:38,  1.99s/it]
 53%|█████▎    | 10048/19090 [4:53:16<4:28:31,  1.78s/it]
 53%|█████▎    | 10049/19090 [4:53:18<4:06:43,  1.64s/it]
 53%|█████▎    | 10050/19090 [4:53:20<5:00:30,  1.99s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 72.61474609375, 'learning_rate': 2.705705055310726e-05, 'recon_loss': 0.001450417676824145, 'recon_static': 0.00048427866495330817, 'recon_motion': 0.0009661390475230291, 'recon_norm': 1.2557859390974044, 'recon_norm_std': 0.044663292914628984, 'target_norm': 1.6213324069976807, 'recon_cos': 0.6606030702590943, 'gradnorm/compressor.query': 0.08860803902657673, 'gradnorm/compressor.layer0': 2.542647521653868, 'gradnorm/compressor.layer1': 69.7646154042291, 'gradnorm/compressor.layer2': 18.111477984019555, 'gradnorm/compressor.layer3': 7.606817861531563, 'gradnorm/compressor.layer4': 2.3659257705149828, 'gradnorm/compressor.layer5': 1.6657801257069667, 'gradnorm/compressor.layer6': 0.9409733326910457, 'gradnorm/compressor.layer7': 1.0866270003265817, 'gradnorm/decoder.layer0': 1.4574883692491623, 'gradnorm/decoder.layer1': 0.8271735861962377, 'gradnorm/decoder.layer2': 0.43626524182666887, 'gradnorm/max_param': 49.21723556518555, 'epoch': 5.26}

 53%|█████▎    | 10050/19090 [4:53:20<5:00:30,  1.99s/it]
 53%|█████▎    | 10051/19090 [4:53:22<4:28:31,  1.78s/it]
 53%|█████▎    | 10052/19090 [4:53:23<4:06:07,  1.63s/it]
 53%|█████▎    | 10053/19090 [4:53:26<4:56:43,  1.97s/it]
 53%|█████▎    | 10054/19090 [4:53:27<4:25:40,  1.76s/it]
 53%|█████▎    | 10055/19090 [4:53:28<4:04:19,  1.62s/it]
 53%|█████▎    | 10056/19090 [4:53:31<5:01:13,  2.00s/it]
 53%|█████▎    | 10057/19090 [4:53:33<4:29:10,  1.79s/it]
 53%|█████▎    | 10058/19090 [4:53:34<4:06:40,  1.64s/it]
 53%|█████▎    | 10059/19090 [4:53:37<5:06:51,  2.04s/it]
 53%|█████▎    | 10060/19090 [4:53:38<4:33:01,  1.81s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 76.81014251708984, 'learning_rate': 2.7011488968071185e-05, 'recon_loss': 0.0014373989863088355, 'recon_static': 0.0004693657778261695, 'recon_motion': 0.000968033223762177, 'recon_norm': 1.2509629011154175, 'recon_norm_std': 0.04486553240567446, 'target_norm': 1.620431712269783, 'recon_cos': 0.6630229070782662, 'gradnorm/compressor.query': 0.10383021038995127, 'gradnorm/compressor.layer0': 2.807853420768173, 'gradnorm/compressor.layer1': 73.56886853804909, 'gradnorm/compressor.layer2': 19.281901294459963, 'gradnorm/compressor.layer3': 9.388213750979578, 'gradnorm/compressor.layer4': 3.2181213653893614, 'gradnorm/compressor.layer5': 1.8815964693412612, 'gradnorm/compressor.layer6': 0.9687543069066937, 'gradnorm/compressor.layer7': 1.194867418986851, 'gradnorm/decoder.layer0': 1.5141245700328856, 'gradnorm/decoder.layer1': 0.9038752979989735, 'gradnorm/decoder.layer2': 0.4721946669311725, 'gradnorm/max_param': 51.899654388427734, 'epoch': 5.27}

 53%|█████▎    | 10060/19090 [4:53:38<4:33:01,  1.81s/it]
 53%|█████▎    | 10061/19090 [4:53:39<4:09:22,  1.66s/it]
 53%|█████▎    | 10062/19090 [4:53:42<4:47:24,  1.91s/it]
 53%|█████▎    | 10063/19090 [4:53:43<4:19:41,  1.73s/it]
 53%|█████▎    | 10064/19090 [4:53:44<4:00:04,  1.60s/it]
 53%|█████▎    | 10065/19090 [4:53:47<4:53:15,  1.95s/it]
 53%|█████▎    | 10066/19090 [4:53:49<4:23:50,  1.75s/it]
 53%|█████▎    | 10067/19090 [4:53:50<4:03:00,  1.62s/it]
 53%|█████▎    | 10068/19090 [4:53:52<4:46:25,  1.90s/it]
 53%|█████▎    | 10069/19090 [4:53:54<4:18:58,  1.72s/it]
 53%|█████▎    | 10070/19090 [4:53:55<3:59:48,  1.60s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 70.7295913696289, 'learning_rate': 2.6965920657599734e-05, 'recon_loss': 0.0014380934997461736, 'recon_static': 0.0004673120864026714, 'recon_motion': 0.0009707814009743742, 'recon_norm': 1.245227512717247, 'recon_norm_std': 0.046291059162467715, 'target_norm': 1.6230233550071715, 'recon_cos': 0.6638372451066971, 'gradnorm/compressor.query': 0.07631602158629872, 'gradnorm/compressor.layer0': 2.232865790871714, 'gradnorm/compressor.layer1': 68.47837321122816, 'gradnorm/compressor.layer2': 14.958045269557545, 'gradnorm/compressor.layer3': 8.2775543603751, 'gradnorm/compressor.layer4': 2.893430595814071, 'gradnorm/compressor.layer5': 1.5859941744895827, 'gradnorm/compressor.layer6': 0.978558842052984, 'gradnorm/compressor.layer7': 1.1178390097094093, 'gradnorm/decoder.layer0': 1.4652431909819175, 'gradnorm/decoder.layer1': 0.8277503551688623, 'gradnorm/decoder.layer2': 0.45800250930806696, 'gradnorm/max_param': 48.31184387207031, 'epoch': 5.28}

 53%|█████▎    | 10070/19090 [4:53:55<3:59:48,  1.60s/it]
 53%|█████▎    | 10071/19090 [4:53:58<4:50:22,  1.93s/it]
 53%|█████▎    | 10072/19090 [4:53:59<4:21:44,  1.74s/it]
 53%|█████▎    | 10073/19090 [4:54:00<4:01:53,  1.61s/it]
 53%|█████▎    | 10074/19090 [4:54:03<4:35:59,  1.84s/it]
 53%|█████▎    | 10075/19090 [4:54:04<4:11:59,  1.68s/it]
 53%|█████▎    | 10076/19090 [4:54:05<3:55:01,  1.56s/it]
 53%|█████▎    | 10077/19090 [4:54:07<4:15:33,  1.70s/it]
 53%|█████▎    | 10078/19090 [4:54:09<3:57:18,  1.58s/it]
 53%|█████▎    | 10079/19090 [4:54:10<3:44:46,  1.50s/it]
 53%|█████▎    | 10080/19090 [4:54:12<4:05:11,  1.63s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 73.05357360839844, 'learning_rate': 2.692034577405106e-05, 'recon_loss': 0.0014203754340996965, 'recon_static': 0.00046195254108170045, 'recon_motion': 0.000958422923577018, 'recon_norm': 1.2490866482257843, 'recon_norm_std': 0.046994432155042884, 'target_norm': 1.6194679319858551, 'recon_cos': 0.6669815078377723, 'gradnorm/compressor.query': 0.07337988942401212, 'gradnorm/compressor.layer0': 2.2424002622881165, 'gradnorm/compressor.layer1': 70.859156215208, 'gradnorm/compressor.layer2': 15.250933509145945, 'gradnorm/compressor.layer3': 7.90998191358811, 'gradnorm/compressor.layer4': 2.8079345099210586, 'gradnorm/compressor.layer5': 1.7386106714717837, 'gradnorm/compressor.layer6': 0.9237699268550805, 'gradnorm/compressor.layer7': 1.0334707104722376, 'gradnorm/decoder.layer0': 1.3900575658829766, 'gradnorm/decoder.layer1': 0.8050146826595682, 'gradnorm/decoder.layer2': 0.45129032974507693, 'gradnorm/max_param': 49.99070358276367, 'epoch': 5.28}

 53%|█████▎    | 10080/19090 [4:54:12<4:05:11,  1.63s/it]
 53%|█████▎    | 10081/19090 [4:54:13<3:50:23,  1.53s/it]
 53%|█████▎    | 10082/19090 [4:54:14<3:39:50,  1.46s/it]
 53%|█████▎    | 10083/19090 [4:54:17<4:09:32,  1.66s/it]
 53%|█████▎    | 10084/19090 [4:54:18<3:54:19,  1.56s/it]
 53%|█████▎    | 10085/19090 [4:54:19<3:42:22,  1.48s/it]
 53%|█████▎    | 10086/19090 [4:54:21<3:56:43,  1.58s/it]
 53%|█████▎    | 10087/19090 [4:54:22<3:44:13,  1.49s/it]
 53%|█████▎    | 10088/19090 [4:54:24<3:35:28,  1.44s/it]
 53%|█████▎    | 10089/19090 [4:54:26<4:03:26,  1.62s/it]
 53%|█████▎    | 10090/19090 [4:54:27<3:48:39,  1.52s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 68.3680191040039, 'learning_rate': 2.687476446980528e-05, 'recon_loss': 0.0014231449720682576, 'recon_static': 0.00046235249101300724, 'recon_motion': 0.0009607925167074427, 'recon_norm': 1.245975413918495, 'recon_norm_std': 0.04582606619223952, 'target_norm': 1.620297634601593, 'recon_cos': 0.6666235700249672, 'gradnorm/compressor.query': 0.07569695324044301, 'gradnorm/compressor.layer0': 2.1251724397592207, 'gradnorm/compressor.layer1': 65.99232761986427, 'gradnorm/compressor.layer2': 15.781499940007139, 'gradnorm/compressor.layer3': 7.046502553904845, 'gradnorm/compressor.layer4': 2.859933100364764, 'gradnorm/compressor.layer5': 1.6689566695003797, 'gradnorm/compressor.layer6': 0.9504863924379864, 'gradnorm/compressor.layer7': 1.140976525054349, 'gradnorm/decoder.layer0': 1.331370931156235, 'gradnorm/decoder.layer1': 0.9105152191721041, 'gradnorm/decoder.layer2': 0.4414700487330655, 'gradnorm/max_param': 46.556034088134766, 'epoch': 5.29}

 53%|█████▎    | 10090/19090 [4:54:27<3:48:39,  1.52s/it]
 53%|█████▎    | 10091/19090 [4:54:28<3:38:20,  1.46s/it]
 53%|█████▎    | 10092/19090 [4:54:31<4:18:11,  1.72s/it]
 53%|█████▎    | 10093/19090 [4:54:32<3:58:41,  1.59s/it]
 53%|█████▎    | 10094/19090 [4:54:33<3:45:15,  1.50s/it]
 53%|█████▎    | 10095/19090 [4:54:35<4:03:02,  1.62s/it]
 53%|█████▎    | 10096/19090 [4:54:36<3:48:11,  1.52s/it]
 53%|█████▎    | 10097/19090 [4:54:38<3:37:51,  1.45s/it]
 53%|█████▎    | 10098/19090 [4:54:40<3:58:09,  1.59s/it]
 53%|█████▎    | 10099/19090 [4:54:41<3:44:51,  1.50s/it]
 53%|█████▎    | 10100/19090 [4:54:42<3:35:28,  1.44s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 79.5674819946289, 'learning_rate': 2.682917689726398e-05, 'recon_loss': 0.0014506230625556781, 'recon_static': 0.00048348558630095795, 'recon_motion': 0.0009671374660683796, 'recon_norm': 1.2545281887054442, 'recon_norm_std': 0.04483687924221158, 'target_norm': 1.621225494146347, 'recon_cos': 0.6601725354790687, 'gradnorm/compressor.query': 0.09087719212876333, 'gradnorm/compressor.layer0': 2.5805590612330427, 'gradnorm/compressor.layer1': 77.10707158413715, 'gradnorm/compressor.layer2': 17.23449009764377, 'gradnorm/compressor.layer3': 8.108589854663531, 'gradnorm/compressor.layer4': 2.86589482658444, 'gradnorm/compressor.layer5': 1.6667094066226025, 'gradnorm/compressor.layer6': 0.9278615121561308, 'gradnorm/compressor.layer7': 1.052739414068558, 'gradnorm/decoder.layer0': 1.4192584769268117, 'gradnorm/decoder.layer1': 0.9155823876682239, 'gradnorm/decoder.layer2': 0.47536532319817537, 'gradnorm/max_param': 54.39453125, 'epoch': 5.29}

 53%|█████▎    | 10100/19090 [4:54:42<3:35:28,  1.44s/it]
 53%|█████▎    | 10101/19090 [4:54:44<4:15:18,  1.70s/it]
 53%|█████▎    | 10102/19090 [4:54:46<3:56:47,  1.58s/it]
 53%|█████▎    | 10103/19090 [4:54:47<3:43:55,  1.50s/it]
 53%|█████▎    | 10104/19090 [4:54:49<4:04:18,  1.63s/it]
 53%|█████▎    | 10105/19090 [4:54:50<3:49:22,  1.53s/it]
 53%|█████▎    | 10106/19090 [4:54:52<3:38:45,  1.46s/it]
 53%|█████▎    | 10107/19090 [4:54:53<3:55:48,  1.58s/it]
 53%|█████▎    | 10108/19090 [4:54:55<3:43:05,  1.49s/it]
 53%|█████▎    | 10109/19090 [4:54:56<3:34:05,  1.43s/it]
 53%|█████▎    | 10110/19090 [4:54:58<4:20:00,  1.74s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 75.64401245117188, 'learning_rate': 2.678358320884971e-05, 'recon_loss': 0.0014348606549901889, 'recon_static': 0.0004629173476132564, 'recon_motion': 0.000971943281183485, 'recon_norm': 1.2495790779590608, 'recon_norm_std': 0.045503266900777814, 'target_norm': 1.6205511838197708, 'recon_cos': 0.6638844475150109, 'gradnorm/compressor.query': 0.080323926566439, 'gradnorm/compressor.layer0': 2.373192350222026, 'gradnorm/compressor.layer1': 72.6831099544884, 'gradnorm/compressor.layer2': 18.64583042364968, 'gradnorm/compressor.layer3': 8.151247324598693, 'gradnorm/compressor.layer4': 3.2122194278962484, 'gradnorm/compressor.layer5': 1.8589881686921195, 'gradnorm/compressor.layer6': 1.0051729396986335, 'gradnorm/compressor.layer7': 1.0313217831683874, 'gradnorm/decoder.layer0': 1.6004314615720654, 'gradnorm/decoder.layer1': 0.8796836297337985, 'gradnorm/decoder.layer2': 0.5006446259231785, 'gradnorm/max_param': 51.27549362182617, 'epoch': 5.3}

 53%|█████▎    | 10110/19090 [4:54:58<4:20:00,  1.74s/it]
 53%|█████▎    | 10111/19090 [4:55:00<3:59:59,  1.60s/it]
 53%|█████▎    | 10112/19090 [4:55:01<3:46:03,  1.51s/it]
 53%|█████▎    | 10113/19090 [4:55:04<4:49:15,  1.93s/it]
 53%|█████▎    | 10114/19090 [4:55:05<4:19:59,  1.74s/it]
 53%|█████▎    | 10115/19090 [4:55:07<3:59:48,  1.60s/it]
 53%|█████▎    | 10116/19090 [4:55:09<4:45:35,  1.91s/it]
 53%|█████▎    | 10117/19090 [4:55:10<4:17:26,  1.72s/it]
 53%|█████▎    | 10118/19090 [4:55:12<3:57:49,  1.59s/it]
 53%|█████▎    | 10119/19090 [4:55:14<4:37:13,  1.85s/it]
 53%|█████▎    | 10120/19090 [4:55:16<4:11:36,  1.68s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 78.82376861572266, 'learning_rate': 2.6737983557005465e-05, 'recon_loss': 0.0014410669187782331, 'recon_static': 0.00047865700544207356, 'recon_motion': 0.0009624099548091181, 'recon_norm': 1.25286945104599, 'recon_norm_std': 0.04624815601855516, 'target_norm': 1.620854765176773, 'recon_cos': 0.6619002759456635, 'gradnorm/compressor.query': 0.09636777113763653, 'gradnorm/compressor.layer0': 2.720053349420531, 'gradnorm/compressor.layer1': 76.0664767602572, 'gradnorm/compressor.layer2': 18.148575386906234, 'gradnorm/compressor.layer3': 8.44976683526312, 'gradnorm/compressor.layer4': 3.0793803022620434, 'gradnorm/compressor.layer5': 1.8383095210179672, 'gradnorm/compressor.layer6': 0.9485834766649115, 'gradnorm/compressor.layer7': 1.1960348801231218, 'gradnorm/decoder.layer0': 1.581736520699227, 'gradnorm/decoder.layer1': 0.9783286945930565, 'gradnorm/decoder.layer2': 0.4985110195791903, 'gradnorm/max_param': 53.660072326660156, 'epoch': 5.3}

 53%|█████▎    | 10120/19090 [4:55:16<4:11:36,  1.68s/it]
 53%|█████▎    | 10121/19090 [4:55:17<3:53:50,  1.56s/it]
 53%|█████▎    | 10122/19090 [4:55:20<4:47:45,  1.93s/it]
 53%|█████▎    | 10123/19090 [4:55:21<4:18:53,  1.73s/it]
 53%|█████▎    | 10124/19090 [4:55:22<3:59:20,  1.60s/it]
 53%|█████▎    | 10125/19090 [4:55:25<4:58:10,  2.00s/it]
 53%|█████▎    | 10126/19090 [4:55:26<4:26:23,  1.78s/it]
 53%|█████▎    | 10127/19090 [4:55:28<4:04:13,  1.63s/it]
 53%|█████▎    | 10128/19090 [4:55:30<4:41:57,  1.89s/it]
 53%|█████▎    | 10129/19090 [4:55:31<4:15:07,  1.71s/it]
 53%|█████▎    | 10130/19090 [4:55:33<3:56:42,  1.59s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 95.39517974853516, 'learning_rate': 2.6692378094194182e-05, 'recon_loss': 0.0014647933363448828, 'recon_static': 0.0005009385669836775, 'recon_motion': 0.0009638547693612054, 'recon_norm': 1.2688905537128448, 'recon_norm_std': 0.04529621312394738, 'target_norm': 1.6194858074188232, 'recon_cos': 0.6564353749155998, 'gradnorm/compressor.query': 0.10317797293689514, 'gradnorm/compressor.layer0': 3.168050931906449, 'gradnorm/compressor.layer1': 92.2987358560506, 'gradnorm/compressor.layer2': 21.46079956499175, 'gradnorm/compressor.layer3': 9.371963619615814, 'gradnorm/compressor.layer4': 3.421057198401735, 'gradnorm/compressor.layer5': 2.0098614757876896, 'gradnorm/compressor.layer6': 1.089138609581463, 'gradnorm/compressor.layer7': 1.1890324191525545, 'gradnorm/decoder.layer0': 1.7243116387468074, 'gradnorm/decoder.layer1': 1.0553243339220728, 'gradnorm/decoder.layer2': 0.5537457598142839, 'gradnorm/max_param': 65.11351013183594, 'epoch': 5.31}

 53%|█████▎    | 10130/19090 [4:55:33<3:56:42,  1.59s/it]
 53%|█████▎    | 10131/19090 [4:55:36<4:53:36,  1.97s/it]
 53%|█████▎    | 10132/19090 [4:55:37<4:23:11,  1.76s/it]
 53%|█████▎    | 10133/19090 [4:55:38<4:02:02,  1.62s/it]
 53%|█████▎    | 10134/19090 [4:55:41<5:13:46,  2.10s/it]
 53%|█████▎    | 10135/19090 [4:55:43<4:37:20,  1.86s/it]
 53%|█████▎    | 10136/19090 [4:55:44<4:11:54,  1.69s/it]
 53%|█████▎    | 10137/19090 [4:55:47<4:56:39,  1.99s/it]
 53%|█████▎    | 10138/19090 [4:55:48<4:25:53,  1.78s/it]
 53%|█████▎    | 10139/19090 [4:55:49<4:03:22,  1.63s/it]
 53%|█████▎    | 10140/19090 [4:55:52<4:57:07,  1.99s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 72.54581451416016, 'learning_rate': 2.664676697289823e-05, 'recon_loss': 0.0014624054223531858, 'recon_static': 0.0004903927372652106, 'recon_motion': 0.0009720127200125717, 'recon_norm': 1.2605101019144058, 'recon_norm_std': 0.04480319311842322, 'target_norm': 1.6226858139038085, 'recon_cos': 0.6581792071461677, 'gradnorm/compressor.query': 0.09140187358975746, 'gradnorm/compressor.layer0': 2.585528568998923, 'gradnorm/compressor.layer1': 69.90264225715113, 'gradnorm/compressor.layer2': 16.834158294908804, 'gradnorm/compressor.layer3': 8.146003688842328, 'gradnorm/compressor.layer4': 3.1034968803516687, 'gradnorm/compressor.layer5': 1.951883089046958, 'gradnorm/compressor.layer6': 1.09642149415712, 'gradnorm/compressor.layer7': 1.2526427941767184, 'gradnorm/decoder.layer0': 1.607133285554509, 'gradnorm/decoder.layer1': 1.005937710276425, 'gradnorm/decoder.layer2': 0.5176098220044365, 'gradnorm/max_param': 49.31184387207031, 'epoch': 5.31}

 53%|█████▎    | 10140/19090 [4:55:52<4:57:07,  1.99s/it]
 53%|█████▎    | 10141/19090 [4:55:53<4:26:01,  1.78s/it]
 53%|█████▎    | 10142/19090 [4:55:55<4:03:43,  1.63s/it]
 53%|█████▎    | 10143/19090 [4:55:58<5:02:48,  2.03s/it]
 53%|█████▎    | 10144/19090 [4:55:59<4:29:34,  1.81s/it]
 53%|█████▎    | 10145/19090 [4:56:00<4:06:20,  1.65s/it]
 53%|█████▎    | 10146/19090 [4:56:03<4:53:04,  1.97s/it]
 53%|█████▎    | 10147/19090 [4:56:04<4:22:37,  1.76s/it]
 53%|█████▎    | 10148/19090 [4:56:05<4:01:28,  1.62s/it]
 53%|█████▎    | 10149/19090 [4:56:08<4:46:50,  1.92s/it]
 53%|█████▎    | 10150/19090 [4:56:09<4:18:44,  1.74s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 84.93441772460938, 'learning_rate': 2.660115034561888e-05, 'recon_loss': 0.0014808293082751334, 'recon_static': 0.0005064438184490428, 'recon_motion': 0.0009743855218403042, 'recon_norm': 1.26494420170784, 'recon_norm_std': 0.04453588053584099, 'target_norm': 1.621795842051506, 'recon_cos': 0.6529468923807145, 'gradnorm/compressor.query': 0.1066332391829153, 'gradnorm/compressor.layer0': 3.067458206799869, 'gradnorm/compressor.layer1': 82.24040894402975, 'gradnorm/compressor.layer2': 17.93241571019156, 'gradnorm/compressor.layer3': 9.942557721257009, 'gradnorm/compressor.layer4': 3.0067355519232515, 'gradnorm/compressor.layer5': 1.9808422090136022, 'gradnorm/compressor.layer6': 1.2180458871469069, 'gradnorm/compressor.layer7': 1.3221231719101136, 'gradnorm/decoder.layer0': 1.712484418143784, 'gradnorm/decoder.layer1': 1.0679217179608456, 'gradnorm/decoder.layer2': 0.5043475266718261, 'gradnorm/max_param': 58.019195556640625, 'epoch': 5.32}

 53%|█████▎    | 10150/19090 [4:56:09<4:18:44,  1.74s/it]
 53%|█████▎    | 10151/19090 [4:56:11<3:59:01,  1.60s/it]
 53%|█████▎    | 10152/19090 [4:56:14<4:56:04,  1.99s/it]
 53%|█████▎    | 10153/19090 [4:56:15<4:25:12,  1.78s/it]
 53%|█████▎    | 10154/19090 [4:56:16<4:03:38,  1.64s/it]
 53%|█████▎    | 10155/19090 [4:56:19<4:52:01,  1.96s/it]
 53%|█████▎    | 10156/19090 [4:56:20<4:22:01,  1.76s/it]
 53%|█████▎    | 10157/19090 [4:56:21<4:01:24,  1.62s/it]
 53%|█████▎    | 10158/19090 [4:56:24<4:51:20,  1.96s/it]
 53%|█████▎    | 10159/19090 [4:56:25<4:21:23,  1.76s/it]
 53%|█████▎    | 10160/19090 [4:56:27<4:00:46,  1.62s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 79.02440643310547, 'learning_rate': 2.6555528364875837e-05, 'recon_loss': 0.0014566452242434024, 'recon_static': 0.00048420772945974023, 'recon_motion': 0.0009724374845973216, 'recon_norm': 1.257770475745201, 'recon_norm_std': 0.04439636524766684, 'target_norm': 1.621566390991211, 'recon_cos': 0.658881177008152, 'gradnorm/compressor.query': 0.11096056695000789, 'gradnorm/compressor.layer0': 2.985296296920578, 'gradnorm/compressor.layer1': 76.21803889582176, 'gradnorm/compressor.layer2': 18.350785889804214, 'gradnorm/compressor.layer3': 8.36792558821494, 'gradnorm/compressor.layer4': 3.2939613810106887, 'gradnorm/compressor.layer5': 1.7554070862965079, 'gradnorm/compressor.layer6': 0.970023671529927, 'gradnorm/compressor.layer7': 1.1984230906031814, 'gradnorm/decoder.layer0': 1.5839812893725727, 'gradnorm/decoder.layer1': 0.9848586211427613, 'gradnorm/decoder.layer2': 0.42653458123139815, 'gradnorm/max_param': 53.77036666870117, 'epoch': 5.32}

 53%|█████▎    | 10160/19090 [4:56:27<4:00:46,  1.62s/it]
 53%|█████▎    | 10161/19090 [4:56:29<4:48:26,  1.94s/it]
 53%|█████▎    | 10162/19090 [4:56:31<4:19:23,  1.74s/it]
 53%|█████▎    | 10163/19090 [4:56:32<3:58:57,  1.61s/it]
 53%|█████▎    | 10164/19090 [4:56:35<5:13:50,  2.11s/it]
 53%|█████▎    | 10165/19090 [4:56:37<4:37:15,  1.86s/it]
 53%|█████▎    | 10166/19090 [4:56:38<4:11:39,  1.69s/it]
 53%|█████▎    | 10167/19090 [4:56:40<4:50:03,  1.95s/it]
 53%|█████▎    | 10168/19090 [4:56:42<4:20:35,  1.75s/it]
 53%|█████▎    | 10169/19090 [4:56:43<3:59:52,  1.61s/it]
 53%|█████▎    | 10170/19090 [4:56:46<4:54:11,  1.98s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 85.95184326171875, 'learning_rate': 2.6509901183206682e-05, 'recon_loss': 0.0014778602810110897, 'recon_static': 0.000507843287778087, 'recon_motion': 0.0009700169990537688, 'recon_norm': 1.2694204479455948, 'recon_norm_std': 0.04453919278457761, 'target_norm': 1.6206338912248612, 'recon_cos': 0.6537052169442177, 'gradnorm/compressor.query': 0.10853292954092958, 'gradnorm/compressor.layer0': 3.14628129679073, 'gradnorm/compressor.layer1': 82.98442648510924, 'gradnorm/compressor.layer2': 19.862901376038163, 'gradnorm/compressor.layer3': 8.5911275295507, 'gradnorm/compressor.layer4': 3.3820856383865947, 'gradnorm/compressor.layer5': 2.0169714636914073, 'gradnorm/compressor.layer6': 1.180797130825723, 'gradnorm/compressor.layer7': 1.3805496321657051, 'gradnorm/decoder.layer0': 1.6754865053704087, 'gradnorm/decoder.layer1': 1.0338964317516262, 'gradnorm/decoder.layer2': 0.6066105812429624, 'gradnorm/max_param': 58.54142379760742, 'epoch': 5.33}

 53%|█████▎    | 10170/19090 [4:56:46<4:54:11,  1.98s/it]
 53%|█████▎    | 10171/19090 [4:56:47<4:23:30,  1.77s/it]
 53%|█████▎    | 10172/19090 [4:56:48<4:02:03,  1.63s/it]
 53%|█████▎    | 10173/19090 [4:56:51<4:55:20,  1.99s/it]
 53%|█████▎    | 10174/19090 [4:56:53<4:24:23,  1.78s/it]
 53%|█████▎    | 10175/19090 [4:56:54<4:03:24,  1.64s/it]
 53%|█████▎    | 10176/19090 [4:56:57<4:49:18,  1.95s/it]
 53%|█████▎    | 10177/19090 [4:56:58<4:19:47,  1.75s/it]
 53%|█████▎    | 10178/19090 [4:56:59<3:59:15,  1.61s/it]
 53%|█████▎    | 10179/19090 [4:57:02<4:48:28,  1.94s/it]
 53%|█████▎    | 10180/19090 [4:57:03<4:19:07,  1.74s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 93.86680603027344, 'learning_rate': 2.64642689531664e-05, 'recon_loss': 0.0014494745148112998, 'recon_static': 0.0004910777723125647, 'recon_motion': 0.0009583967621438205, 'recon_norm': 1.2564745873212815, 'recon_norm_std': 0.04589588455855846, 'target_norm': 1.6188738286495208, 'recon_cos': 0.6591101959347725, 'gradnorm/compressor.query': 0.10168743701937553, 'gradnorm/compressor.layer0': 2.9477797477566363, 'gradnorm/compressor.layer1': 90.43619070838704, 'gradnorm/compressor.layer2': 22.371641791458053, 'gradnorm/compressor.layer3': 10.213089521334902, 'gradnorm/compressor.layer4': 3.2745120667837972, 'gradnorm/compressor.layer5': 1.6149171381617888, 'gradnorm/compressor.layer6': 0.9454362331151654, 'gradnorm/compressor.layer7': 1.1008942112247508, 'gradnorm/decoder.layer0': 1.5167865038607335, 'gradnorm/decoder.layer1': 0.9088276346409144, 'gradnorm/decoder.layer2': 0.4595853401836572, 'gradnorm/max_param': 63.802452087402344, 'epoch': 5.33}

 53%|█████▎    | 10180/19090 [4:57:03<4:19:07,  1.74s/it]
 53%|█████▎    | 10181/19090 [4:57:04<3:58:34,  1.61s/it]
 53%|█████▎    | 10182/19090 [4:57:07<4:58:44,  2.01s/it]
 53%|█████▎    | 10183/19090 [4:57:09<4:26:09,  1.79s/it]
 53%|█████▎    | 10184/19090 [4:57:10<4:03:24,  1.64s/it]
 53%|█████▎    | 10185/19090 [4:57:13<4:54:08,  1.98s/it]
 53%|█████▎    | 10186/19090 [4:57:14<4:22:52,  1.77s/it]
 53%|█████▎    | 10187/19090 [4:57:15<4:01:09,  1.63s/it]
 53%|█████▎    | 10188/19090 [4:57:18<4:52:15,  1.97s/it]
 53%|█████▎    | 10189/19090 [4:57:19<4:21:38,  1.76s/it]
 53%|█████▎    | 10190/19090 [4:57:21<4:00:34,  1.62s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 75.43873596191406, 'learning_rate': 2.6418631827326857e-05, 'recon_loss': 0.0014393300836673007, 'recon_static': 0.0004678736098867375, 'recon_motion': 0.0009714565036119893, 'recon_norm': 1.2501934707164764, 'recon_norm_std': 0.04537543039768934, 'target_norm': 1.6223849415779115, 'recon_cos': 0.6634764343500137, 'gradnorm/compressor.query': 0.09846795345203344, 'gradnorm/compressor.layer0': 2.543204722904501, 'gradnorm/compressor.layer1': 72.93791505880189, 'gradnorm/compressor.layer2': 16.516787225544387, 'gradnorm/compressor.layer3': 8.525128190763192, 'gradnorm/compressor.layer4': 3.1962766801528413, 'gradnorm/compressor.layer5': 1.9360052926645366, 'gradnorm/compressor.layer6': 0.9364705155329531, 'gradnorm/compressor.layer7': 1.0854183575305227, 'gradnorm/decoder.layer0': 1.4582666472673886, 'gradnorm/decoder.layer1': 0.8331456131899813, 'gradnorm/decoder.layer2': 0.5129796574781096, 'gradnorm/max_param': 51.45667266845703, 'epoch': 5.34}

 53%|█████▎    | 10190/19090 [4:57:21<4:00:34,  1.62s/it]
 53%|█████▎    | 10191/19090 [4:57:24<4:58:00,  2.01s/it]
 53%|█████▎    | 10192/19090 [4:57:25<4:25:56,  1.79s/it]
 53%|█████▎    | 10193/19090 [4:57:26<4:03:15,  1.64s/it]
 53%|█████▎    | 10194/19090 [4:57:29<5:01:30,  2.03s/it]
 53%|█████▎    | 10195/19090 [4:57:30<4:28:06,  1.81s/it]
 53%|█████▎    | 10196/19090 [4:57:32<4:04:45,  1.65s/it]
 53%|█████▎    | 10197/19090 [4:57:34<4:49:02,  1.95s/it]
 53%|█████▎    | 10198/19090 [4:57:36<4:19:40,  1.75s/it]
 53%|█████▎    | 10199/19090 [4:57:37<3:58:55,  1.61s/it]
 53%|█████▎    | 10200/19090 [4:57:40<4:56:56,  2.00s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 75.23663330078125, 'learning_rate': 2.6372989958276274e-05, 'recon_loss': 0.0014338774257339538, 'recon_static': 0.0004602485823852476, 'recon_motion': 0.0009736288629937917, 'recon_norm': 1.245570182800293, 'recon_norm_std': 0.044285082072019574, 'target_norm': 1.6214531391859055, 'recon_cos': 0.6643995270133018, 'gradnorm/compressor.query': 0.08177731868159509, 'gradnorm/compressor.layer0': 2.2797800841431317, 'gradnorm/compressor.layer1': 72.55513797709298, 'gradnorm/compressor.layer2': 17.330206564559827, 'gradnorm/compressor.layer3': 8.363289004112472, 'gradnorm/compressor.layer4': 3.3589201863270337, 'gradnorm/compressor.layer5': 2.101267967714626, 'gradnorm/compressor.layer6': 0.9237460206788781, 'gradnorm/compressor.layer7': 1.0303006860103727, 'gradnorm/decoder.layer0': 1.547094387333786, 'gradnorm/decoder.layer1': 0.7626305233915274, 'gradnorm/decoder.layer2': 0.48089828519715727, 'gradnorm/max_param': 51.18730926513672, 'epoch': 5.34}

 53%|█████▎    | 10200/19090 [4:57:40<4:56:56,  2.00s/it]
 53%|█████▎    | 10201/19090 [4:57:41<4:25:05,  1.79s/it]
 53%|█████▎    | 10202/19090 [4:57:42<4:02:34,  1.64s/it]
 53%|█████▎    | 10203/19090 [4:57:45<4:55:37,  2.00s/it]
 53%|█████▎    | 10204/19090 [4:57:46<4:24:16,  1.78s/it]
 53%|█████▎    | 10205/19090 [4:57:48<4:02:13,  1.64s/it]
 53%|█████▎    | 10206/19090 [4:57:50<4:47:06,  1.94s/it]
 53%|█████▎    | 10207/19090 [4:57:52<4:18:01,  1.74s/it]
 53%|█████▎    | 10208/19090 [4:57:53<3:58:00,  1.61s/it]
 53%|█████▎    | 10209/19090 [4:57:56<4:57:52,  2.01s/it]
 53%|█████▎    | 10210/19090 [4:57:57<4:25:47,  1.80s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 80.2720718383789, 'learning_rate': 2.632734349861874e-05, 'recon_loss': 0.0014221799734514206, 'recon_static': 0.0004520090042205993, 'recon_motion': 0.0009701710398076103, 'recon_norm': 1.2427109152078628, 'recon_norm_std': 0.044833794236183167, 'target_norm': 1.6211644381284713, 'recon_cos': 0.6672753781080246, 'gradnorm/compressor.query': 0.0856871317369323, 'gradnorm/compressor.layer0': 2.4323775434887986, 'gradnorm/compressor.layer1': 77.4878991308885, 'gradnorm/compressor.layer2': 18.292297676805802, 'gradnorm/compressor.layer3': 8.844612214346538, 'gradnorm/compressor.layer4': 3.282062684189877, 'gradnorm/compressor.layer5': 1.9953760697074516, 'gradnorm/compressor.layer6': 1.008878395578689, 'gradnorm/compressor.layer7': 1.1756784387800745, 'gradnorm/decoder.layer0': 1.5342134876279463, 'gradnorm/decoder.layer1': 0.8342925313134048, 'gradnorm/decoder.layer2': 0.5188291294000292, 'gradnorm/max_param': 54.665531158447266, 'epoch': 5.35}

 53%|█████▎    | 10210/19090 [4:57:57<4:25:47,  1.80s/it]
 53%|█████▎    | 10211/19090 [4:57:58<4:03:09,  1.64s/it]
 53%|█████▎    | 10212/19090 [4:58:01<4:59:50,  2.03s/it]
 53%|█████▎    | 10213/19090 [4:58:03<4:27:15,  1.81s/it]
 54%|█████▎    | 10214/19090 [4:58:04<4:04:14,  1.65s/it]
 54%|█████▎    | 10215/19090 [4:58:07<4:48:16,  1.95s/it]
 54%|█████▎    | 10216/19090 [4:58:08<4:18:55,  1.75s/it]
 54%|█████▎    | 10217/19090 [4:58:09<3:58:38,  1.61s/it]
 54%|█████▎    | 10218/19090 [4:58:12<4:58:32,  2.02s/it]
 54%|█████▎    | 10219/19090 [4:58:13<4:26:27,  1.80s/it]
 54%|█████▎    | 10220/19090 [4:58:15<4:04:09,  1.65s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 95.80914306640625, 'learning_rate': 2.6281692600973702e-05, 'recon_loss': 0.0014216225070413202, 'recon_static': 0.0004489705679588951, 'recon_motion': 0.0009726519187097437, 'recon_norm': 1.2402107775211335, 'recon_norm_std': 0.04546945113688707, 'target_norm': 1.6209996044635773, 'recon_cos': 0.6668393969535827, 'gradnorm/compressor.query': 0.09675077271830587, 'gradnorm/compressor.layer0': 2.9183473558450297, 'gradnorm/compressor.layer1': 92.22042844118108, 'gradnorm/compressor.layer2': 22.850352532069255, 'gradnorm/compressor.layer3': 10.781571623939104, 'gradnorm/compressor.layer4': 3.997930826013302, 'gradnorm/compressor.layer5': 2.160654738235072, 'gradnorm/compressor.layer6': 1.1104013772356798, 'gradnorm/compressor.layer7': 1.2352554947360872, 'gradnorm/decoder.layer0': 1.739739447192082, 'gradnorm/decoder.layer1': 1.0075254992720282, 'gradnorm/decoder.layer2': 0.6509551156050815, 'gradnorm/max_param': 65.06123352050781, 'epoch': 5.35}

 54%|█████▎    | 10220/19090 [4:58:15<4:04:09,  1.65s/it]
 54%|█████▎    | 10221/19090 [4:58:17<4:45:22,  1.93s/it]
 54%|█████▎    | 10222/19090 [4:58:19<4:16:57,  1.74s/it]
 54%|█████▎    | 10223/19090 [4:58:20<3:57:22,  1.61s/it]
 54%|█████▎    | 10224/19090 [4:58:23<4:50:12,  1.96s/it]
 54%|█████▎    | 10225/19090 [4:58:24<4:27:20,  1.81s/it]
 54%|█████▎    | 10226/19090 [4:58:25<4:04:08,  1.65s/it]
 54%|█████▎    | 10227/19090 [4:58:28<4:53:12,  1.98s/it]
 54%|█████▎    | 10228/19090 [4:58:30<4:22:09,  1.77s/it]
 54%|█████▎    | 10229/19090 [4:58:31<4:00:51,  1.63s/it]
 54%|█████▎    | 10230/19090 [4:58:33<4:45:45,  1.94s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 93.26847076416016, 'learning_rate': 2.6236037417975423e-05, 'recon_loss': 0.001410870865220204, 'recon_static': 0.0004570750868879259, 'recon_motion': 0.0009537957623251714, 'recon_norm': 1.2431618064641952, 'recon_norm_std': 0.046240666136145595, 'target_norm': 1.618113148212433, 'recon_cos': 0.6685903444886208, 'gradnorm/compressor.query': 0.10919780086196547, 'gradnorm/compressor.layer0': 3.0399074605609977, 'gradnorm/compressor.layer1': 89.9247189493801, 'gradnorm/compressor.layer2': 22.742674601877304, 'gradnorm/compressor.layer3': 8.136816265018158, 'gradnorm/compressor.layer4': 3.074997295595546, 'gradnorm/compressor.layer5': 1.9472299341051185, 'gradnorm/compressor.layer6': 1.1376435493186767, 'gradnorm/compressor.layer7': 1.3197384099695293, 'gradnorm/decoder.layer0': 1.5615849676647673, 'gradnorm/decoder.layer1': 0.9235880178816728, 'gradnorm/decoder.layer2': 0.5368472760273063, 'gradnorm/max_param': 63.442230224609375, 'epoch': 5.36}

 54%|█████▎    | 10230/19090 [4:58:33<4:45:45,  1.94s/it]
 54%|█████▎    | 10231/19090 [4:58:35<4:30:04,  1.83s/it]
 54%|█████▎    | 10232/19090 [4:58:36<4:06:04,  1.67s/it]
 54%|█████▎    | 10233/19090 [4:58:39<4:35:51,  1.87s/it]
 54%|█████▎    | 10234/19090 [4:58:40<4:30:44,  1.83s/it]
 54%|█████▎    | 10235/19090 [4:58:42<4:06:20,  1.67s/it]
 54%|█████▎    | 10236/19090 [4:58:44<4:37:21,  1.88s/it]
 54%|█████▎    | 10237/19090 [4:58:46<4:28:16,  1.82s/it]
 54%|█████▎    | 10238/19090 [4:58:47<4:04:47,  1.66s/it]
 54%|█████▎    | 10239/19090 [4:58:49<4:24:19,  1.79s/it]
 54%|█████▎    | 10240/19090 [4:58:51<4:29:45,  1.83s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 92.03244018554688, 'learning_rate': 2.6190378102272515e-05, 'recon_loss': 0.001451288696262054, 'recon_static': 0.00048156496486626563, 'recon_motion': 0.0009697237226646393, 'recon_norm': 1.2551708400249482, 'recon_norm_std': 0.04523364203050732, 'target_norm': 1.621123993396759, 'recon_cos': 0.6598633855581284, 'gradnorm/compressor.query': 0.1593071070682785, 'gradnorm/compressor.layer0': 3.7314150264162835, 'gradnorm/compressor.layer1': 88.28991384478043, 'gradnorm/compressor.layer2': 23.125732410276626, 'gradnorm/compressor.layer3': 10.140951894889978, 'gradnorm/compressor.layer4': 3.5327036068519457, 'gradnorm/compressor.layer5': 1.963175981570702, 'gradnorm/compressor.layer6': 1.2361579765884256, 'gradnorm/compressor.layer7': 1.3163551750802096, 'gradnorm/decoder.layer0': 1.5349894425715613, 'gradnorm/decoder.layer1': 1.011738662818334, 'gradnorm/decoder.layer2': 0.5460741173267151, 'gradnorm/max_param': 62.2853889465332, 'epoch': 5.36}

 54%|█████▎    | 10240/19090 [4:58:51<4:29:45,  1.83s/it]
 54%|█████▎    | 10241/19090 [4:58:52<4:05:39,  1.67s/it]
 54%|█████▎    | 10242/19090 [4:58:55<4:42:14,  1.91s/it]
 54%|█████▎    | 10243/19090 [4:58:57<4:38:07,  1.89s/it]
 54%|█████▎    | 10244/19090 [4:58:58<4:11:32,  1.71s/it]
 54%|█████▎    | 10245/19090 [4:59:00<4:34:10,  1.86s/it]
 54%|█████▎    | 10246/19090 [4:59:02<4:30:24,  1.83s/it]
 54%|█████▎    | 10247/19090 [4:59:03<4:06:03,  1.67s/it]
 54%|█████▎    | 10248/19090 [4:59:06<4:38:01,  1.89s/it]
 54%|█████▎    | 10249/19090 [4:59:07<4:24:15,  1.79s/it]
 54%|█████▎    | 10250/19090 [4:59:08<4:02:01,  1.64s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 85.25199890136719, 'learning_rate': 2.6144714806527404e-05, 'recon_loss': 0.0014492832735413686, 'recon_static': 0.0004904574678221251, 'recon_motion': 0.0009588258093572221, 'recon_norm': 1.2601136267185211, 'recon_norm_std': 0.04543630173429847, 'target_norm': 1.6189985483884812, 'recon_cos': 0.6598320603370667, 'gradnorm/compressor.query': 0.1065664826624936, 'gradnorm/compressor.layer0': 2.9670127202485497, 'gradnorm/compressor.layer1': 82.35424506173013, 'gradnorm/compressor.layer2': 19.257888564758503, 'gradnorm/compressor.layer3': 9.312678188180943, 'gradnorm/compressor.layer4': 3.184310963354605, 'gradnorm/compressor.layer5': 1.7878360459087912, 'gradnorm/compressor.layer6': 1.053640962527537, 'gradnorm/compressor.layer7': 1.2234349315319164, 'gradnorm/decoder.layer0': 1.4944748049743921, 'gradnorm/decoder.layer1': 0.9214703593172603, 'gradnorm/decoder.layer2': 0.49935279622565554, 'gradnorm/max_param': 58.100318908691406, 'epoch': 5.37}

 54%|█████▎    | 10250/19090 [4:59:09<4:02:01,  1.64s/it]
 54%|█████▎    | 10251/19090 [4:59:11<4:39:26,  1.90s/it]
 54%|█████▎    | 10252/19090 [4:59:13<4:30:44,  1.84s/it]
 54%|█████▎    | 10253/19090 [4:59:14<4:06:46,  1.68s/it]
 54%|█████▎    | 10254/19090 [4:59:16<4:38:22,  1.89s/it]
 54%|█████▎    | 10255/19090 [4:59:18<4:24:22,  1.80s/it]
 54%|█████▎    | 10256/19090 [4:59:19<4:02:02,  1.64s/it]
 54%|█████▎    | 10257/19090 [4:59:22<4:50:25,  1.97s/it]
 54%|█████▎    | 10258/19090 [4:59:23<4:20:23,  1.77s/it]
 54%|█████▎    | 10259/19090 [4:59:25<3:59:06,  1.62s/it]
 54%|█████▎    | 10260/19090 [4:59:27<4:29:14,  1.83s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 81.50544738769531, 'learning_rate': 2.6099047683415805e-05, 'recon_loss': 0.0014347126299981028, 'recon_static': 0.0004712585563538596, 'recon_motion': 0.0009634541042032651, 'recon_norm': 1.2466429859399795, 'recon_norm_std': 0.045640963967889546, 'target_norm': 1.6216441184282302, 'recon_cos': 0.6641599908471107, 'gradnorm/compressor.query': 0.08972985366644352, 'gradnorm/compressor.layer0': 2.6617556789692673, 'gradnorm/compressor.layer1': 78.55769239963868, 'gradnorm/compressor.layer2': 18.64882124788446, 'gradnorm/compressor.layer3': 9.718927019966433, 'gradnorm/compressor.layer4': 3.373122081348909, 'gradnorm/compressor.layer5': 2.1025409902235124, 'gradnorm/compressor.layer6': 1.159030560845097, 'gradnorm/compressor.layer7': 1.369810802872387, 'gradnorm/decoder.layer0': 1.5124879600108745, 'gradnorm/decoder.layer1': 0.981130278020186, 'gradnorm/decoder.layer2': 0.48076953622001334, 'gradnorm/max_param': 55.41999053955078, 'epoch': 5.37}

 54%|█████▎    | 10260/19090 [4:59:27<4:29:14,  1.83s/it]
 54%|█████▍    | 10261/19090 [4:59:28<4:10:15,  1.70s/it]
 54%|█████▍    | 10262/19090 [4:59:30<3:52:11,  1.58s/it]
 54%|█████▍    | 10263/19090 [4:59:32<4:25:48,  1.81s/it]
 54%|█████▍    | 10264/19090 [4:59:34<4:25:17,  1.80s/it]
 54%|█████▍    | 10265/19090 [4:59:35<4:02:35,  1.65s/it]
 54%|█████▍    | 10266/19090 [4:59:38<4:43:03,  1.92s/it]
 54%|█████▍    | 10267/19090 [4:59:39<4:15:29,  1.74s/it]
 54%|█████▍    | 10268/19090 [4:59:40<3:55:40,  1.60s/it]
 54%|█████▍    | 10269/19090 [4:59:43<4:31:46,  1.85s/it]
 54%|█████▍    | 10270/19090 [4:59:45<4:36:23,  1.88s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 84.65154266357422, 'learning_rate': 2.6053376885626257e-05, 'recon_loss': 0.0014537620590999722, 'recon_static': 0.0004867568386544008, 'recon_motion': 0.0009670052211731672, 'recon_norm': 1.2579822152853013, 'recon_norm_std': 0.045482829492539165, 'target_norm': 1.6230092525482178, 'recon_cos': 0.6604663699865341, 'gradnorm/compressor.query': 0.094338082551756, 'gradnorm/compressor.layer0': 2.729062833096916, 'gradnorm/compressor.layer1': 82.34002304184308, 'gradnorm/compressor.layer2': 16.944007665011046, 'gradnorm/compressor.layer3': 8.37745061292352, 'gradnorm/compressor.layer4': 3.2525535235446856, 'gradnorm/compressor.layer5': 2.118095400353818, 'gradnorm/compressor.layer6': 1.0281035076086347, 'gradnorm/compressor.layer7': 1.2002735779759983, 'gradnorm/decoder.layer0': 1.5959695806018424, 'gradnorm/decoder.layer1': 0.945205713971512, 'gradnorm/decoder.layer2': 0.5131294374355383, 'gradnorm/max_param': 58.092193603515625, 'epoch': 5.38}

 54%|█████▍    | 10270/19090 [4:59:45<4:36:23,  1.88s/it]
 54%|█████▍    | 10271/19090 [4:59:46<4:10:13,  1.70s/it]
 54%|█████▍    | 10272/19090 [4:59:48<4:19:52,  1.77s/it]
 54%|█████▍    | 10273/19090 [4:59:50<4:30:36,  1.84s/it]
 54%|█████▍    | 10274/19090 [4:59:51<4:06:07,  1.68s/it]
 54%|█████▍    | 10275/19090 [4:59:53<4:30:47,  1.84s/it]
 54%|█████▍    | 10276/19090 [4:59:55<4:29:52,  1.84s/it]
 54%|█████▍    | 10277/19090 [4:59:56<4:05:46,  1.67s/it]
 54%|█████▍    | 10278/19090 [4:59:59<4:36:57,  1.89s/it]
 54%|█████▍    | 10279/19090 [5:00:00<4:23:07,  1.79s/it]
 54%|█████▍    | 10280/19090 [5:00:02<4:00:48,  1.64s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 84.03999328613281, 'learning_rate': 2.6007702565859565e-05, 'recon_loss': 0.0014565787801984697, 'recon_static': 0.0004994641698431223, 'recon_motion': 0.0009571145914378576, 'recon_norm': 1.259548118710518, 'recon_norm_std': 0.04575387965887785, 'target_norm': 1.6189810037612915, 'recon_cos': 0.6575835719704628, 'gradnorm/compressor.query': 0.08655882686437444, 'gradnorm/compressor.layer0': 2.700205201722397, 'gradnorm/compressor.layer1': 80.78040981187827, 'gradnorm/compressor.layer2': 20.54883545780332, 'gradnorm/compressor.layer3': 9.272692913188957, 'gradnorm/compressor.layer4': 3.2636300888445557, 'gradnorm/compressor.layer5': 2.176037538730571, 'gradnorm/compressor.layer6': 0.9412888259782917, 'gradnorm/compressor.layer7': 1.202641080166168, 'gradnorm/decoder.layer0': 1.6339586241770203, 'gradnorm/decoder.layer1': 1.0132943628267104, 'gradnorm/decoder.layer2': 0.5425805549759414, 'gradnorm/max_param': 56.986785888671875, 'epoch': 5.39}

 54%|█████▍    | 10280/19090 [5:00:02<4:00:48,  1.64s/it]
 54%|█████▍    | 10281/19090 [5:00:04<4:29:39,  1.84s/it]
 54%|█████▍    | 10282/19090 [5:00:06<4:21:14,  1.78s/it]
 54%|█████▍    | 10283/19090 [5:00:07<4:00:12,  1.64s/it]
 54%|█████▍    | 10284/19090 [5:00:09<4:41:18,  1.92s/it]
 54%|█████▍    | 10285/19090 [5:00:11<4:29:39,  1.84s/it]
 54%|█████▍    | 10286/19090 [5:00:12<4:05:54,  1.68s/it]
 54%|█████▍    | 10287/19090 [5:00:15<4:46:11,  1.95s/it]
 54%|█████▍    | 10288/19090 [5:00:16<4:16:58,  1.75s/it]
 54%|█████▍    | 10289/19090 [5:00:18<3:56:26,  1.61s/it]
 54%|█████▍    | 10290/19090 [5:00:21<4:56:14,  2.02s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 93.30254364013672, 'learning_rate': 2.5962024876828312e-05, 'recon_loss': 0.0014566259662387893, 'recon_static': 0.0004888942712568678, 'recon_motion': 0.0009677316731540487, 'recon_norm': 1.2596838831901551, 'recon_norm_std': 0.04435267699882388, 'target_norm': 1.619141846895218, 'recon_cos': 0.6577263548970222, 'gradnorm/compressor.query': 0.116680921931804, 'gradnorm/compressor.layer0': 3.0581049619915155, 'gradnorm/compressor.layer1': 90.15690803486025, 'gradnorm/compressor.layer2': 21.569530103974216, 'gradnorm/compressor.layer3': 9.073871922854282, 'gradnorm/compressor.layer4': 3.1672716232141007, 'gradnorm/compressor.layer5': 2.0011156070147322, 'gradnorm/compressor.layer6': 1.082900211365278, 'gradnorm/compressor.layer7': 1.2059877199430271, 'gradnorm/decoder.layer0': 1.4611214098710432, 'gradnorm/decoder.layer1': 1.044354664520675, 'gradnorm/decoder.layer2': 0.5123947861297489, 'gradnorm/max_param': 63.606781005859375, 'epoch': 5.39}

 54%|█████▍    | 10290/19090 [5:00:21<4:56:14,  2.02s/it]
 54%|█████▍    | 10291/19090 [5:00:22<4:23:58,  1.80s/it]
 54%|█████▍    | 10292/19090 [5:00:23<4:01:21,  1.65s/it]
 54%|█████▍    | 10293/19090 [5:00:26<4:52:19,  1.99s/it]
 54%|█████▍    | 10294/19090 [5:00:27<4:21:33,  1.78s/it]
 54%|█████▍    | 10295/19090 [5:00:28<3:59:41,  1.64s/it]
 54%|█████▍    | 10296/19090 [5:00:31<4:38:20,  1.90s/it]
 54%|█████▍    | 10297/19090 [5:00:33<4:31:20,  1.85s/it]
 54%|█████▍    | 10298/19090 [5:00:34<4:06:29,  1.68s/it]
 54%|█████▍    | 10299/19090 [5:00:36<4:22:48,  1.79s/it]
 54%|█████▍    | 10300/19090 [5:00:38<4:41:10,  1.92s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 82.99024200439453, 'learning_rate': 2.591634397125637e-05, 'recon_loss': 0.0014422558509977535, 'recon_static': 0.0004768013925058767, 'recon_motion': 0.0009654544715886004, 'recon_norm': 1.2553684026002885, 'recon_norm_std': 0.0455731512978673, 'target_norm': 1.6199581414461135, 'recon_cos': 0.6619824856519699, 'gradnorm/compressor.query': 0.10140546248825237, 'gradnorm/compressor.layer0': 2.791246430001408, 'gradnorm/compressor.layer1': 79.11357663392864, 'gradnorm/compressor.layer2': 22.710753406976618, 'gradnorm/compressor.layer3': 9.17616538148689, 'gradnorm/compressor.layer4': 3.297068368746095, 'gradnorm/compressor.layer5': 1.871916269602089, 'gradnorm/compressor.layer6': 1.0675670747432817, 'gradnorm/compressor.layer7': 1.2396955626302233, 'gradnorm/decoder.layer0': 1.5269109790886708, 'gradnorm/decoder.layer1': 1.0119391830851603, 'gradnorm/decoder.layer2': 0.4821791095955468, 'gradnorm/max_param': 55.807395935058594, 'epoch': 5.4}

 54%|█████▍    | 10300/19090 [5:00:38<4:41:10,  1.92s/it]
 54%|█████▍    | 10301/19090 [5:00:40<4:13:41,  1.73s/it]
 54%|█████▍    | 10302/19090 [5:00:41<4:07:47,  1.69s/it]
 54%|█████▍    | 10303/19090 [5:00:44<4:48:58,  1.97s/it]
 54%|█████▍    | 10304/19090 [5:00:45<4:19:02,  1.77s/it]
 54%|█████▍    | 10305/19090 [5:00:46<3:59:14,  1.63s/it]
 54%|█████▍    | 10306/19090 [5:00:49<4:45:26,  1.95s/it]
 54%|█████▍    | 10307/19090 [5:00:50<4:16:17,  1.75s/it]
 54%|█████▍    | 10308/19090 [5:00:52<4:16:33,  1.75s/it]
 54%|█████▍    | 10309/19090 [5:00:55<4:42:43,  1.93s/it]
 54%|█████▍    | 10310/19090 [5:00:56<4:14:39,  1.74s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 79.90873718261719, 'learning_rate': 2.5870660001878326e-05, 'recon_loss': 0.0014729210757650434, 'recon_static': 0.0004965613858075813, 'recon_motion': 0.0009763597437995486, 'recon_norm': 1.2589241772890092, 'recon_norm_std': 0.045100854337215425, 'target_norm': 1.6224054247140884, 'recon_cos': 0.6546313807368278, 'gradnorm/compressor.query': 0.0922808671388522, 'gradnorm/compressor.layer0': 2.684890899999004, 'gradnorm/compressor.layer1': 76.98723387476329, 'gradnorm/compressor.layer2': 18.76941702943914, 'gradnorm/compressor.layer3': 8.851871926134026, 'gradnorm/compressor.layer4': 3.1921215586449736, 'gradnorm/compressor.layer5': 1.9796279838064044, 'gradnorm/compressor.layer6': 1.13427019712288, 'gradnorm/compressor.layer7': 1.2816116008826417, 'gradnorm/decoder.layer0': 1.515353542282632, 'gradnorm/decoder.layer1': 0.9595614559635268, 'gradnorm/decoder.layer2': 0.5031378750308789, 'gradnorm/max_param': 54.315208435058594, 'epoch': 5.4}

 54%|█████▍    | 10310/19090 [5:00:56<4:14:39,  1.74s/it]
 54%|█████▍    | 10311/19090 [5:00:57<4:12:01,  1.72s/it]
 54%|█████▍    | 10312/19090 [5:01:00<4:33:43,  1.87s/it]
 54%|█████▍    | 10313/19090 [5:01:01<4:08:12,  1.70s/it]
 54%|█████▍    | 10314/19090 [5:01:03<4:28:43,  1.84s/it]
 54%|█████▍    | 10315/19090 [5:01:05<4:38:32,  1.90s/it]
 54%|█████▍    | 10316/19090 [5:01:06<4:11:27,  1.72s/it]
 54%|█████▍    | 10317/19090 [5:01:09<4:24:23,  1.81s/it]
 54%|█████▍    | 10318/19090 [5:01:10<4:31:47,  1.86s/it]
 54%|█████▍    | 10319/19090 [5:01:12<4:06:50,  1.69s/it]
 54%|█████▍    | 10320/19090 [5:01:14<4:31:41,  1.86s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 101.04926300048828, 'learning_rate': 2.582497312143904e-05, 'recon_loss': 0.0014733286545379088, 'recon_static': 0.0004969587716914248, 'recon_motion': 0.0009763698515598662, 'recon_norm': 1.2611062616109847, 'recon_norm_std': 0.044313137978315355, 'target_norm': 1.6221644639968873, 'recon_cos': 0.6549731150269509, 'gradnorm/compressor.query': 0.1098789119197664, 'gradnorm/compressor.layer0': 3.212960230296554, 'gradnorm/compressor.layer1': 97.87393238562044, 'gradnorm/compressor.layer2': 21.852874552806618, 'gradnorm/compressor.layer3': 10.779007211239861, 'gradnorm/compressor.layer4': 3.954232524381003, 'gradnorm/compressor.layer5': 2.0761141118566773, 'gradnorm/compressor.layer6': 1.2372516479203723, 'gradnorm/compressor.layer7': 1.4092003284827528, 'gradnorm/decoder.layer0': 1.6861820018905895, 'gradnorm/decoder.layer1': 0.9762213149111991, 'gradnorm/decoder.layer2': 0.5625034703041687, 'gradnorm/max_param': 69.04583740234375, 'epoch': 5.41}

 54%|█████▍    | 10320/19090 [5:01:14<4:31:41,  1.86s/it]
 54%|█████▍    | 10321/19090 [5:01:16<4:24:42,  1.81s/it]
 54%|█████▍    | 10322/19090 [5:01:17<4:02:00,  1.66s/it]
 54%|█████▍    | 10323/19090 [5:01:19<4:19:52,  1.78s/it]
 54%|█████▍    | 10324/19090 [5:01:21<4:31:22,  1.86s/it]
 54%|█████▍    | 10325/19090 [5:01:22<4:06:27,  1.69s/it]
 54%|█████▍    | 10326/19090 [5:01:24<4:19:11,  1.77s/it]
 54%|█████▍    | 10327/19090 [5:01:27<4:37:12,  1.90s/it]
 54%|█████▍    | 10328/19090 [5:01:28<4:10:21,  1.71s/it]
 54%|█████▍    | 10329/19090 [5:01:30<4:17:04,  1.76s/it]
 54%|█████▍    | 10330/19090 [5:01:32<4:44:09,  1.95s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 93.59909057617188, 'learning_rate': 2.57792834826931e-05, 'recon_loss': 0.001450656686211005, 'recon_static': 0.0004836453008465469, 'recon_motion': 0.0009670113766333088, 'recon_norm': 1.2576563566923142, 'recon_norm_std': 0.04450082741677761, 'target_norm': 1.6199653267860412, 'recon_cos': 0.6596239462494851, 'gradnorm/compressor.query': 0.11560963692927369, 'gradnorm/compressor.layer0': 2.925471575942995, 'gradnorm/compressor.layer1': 90.95514218902085, 'gradnorm/compressor.layer2': 19.078813620096785, 'gradnorm/compressor.layer3': 9.801010110918973, 'gradnorm/compressor.layer4': 3.204830255402364, 'gradnorm/compressor.layer5': 1.788470974211811, 'gradnorm/compressor.layer6': 1.0515042905298222, 'gradnorm/compressor.layer7': 1.2101301886139257, 'gradnorm/decoder.layer0': 1.5022265280916425, 'gradnorm/decoder.layer1': 0.8881073552525817, 'gradnorm/decoder.layer2': 0.48505010011815664, 'gradnorm/max_param': 64.16231536865234, 'epoch': 5.41}

 54%|█████▍    | 10330/19090 [5:01:32<4:44:09,  1.95s/it]
 54%|█████▍    | 10331/19090 [5:01:33<4:15:35,  1.75s/it]
 54%|█████▍    | 10332/19090 [5:01:35<4:05:48,  1.68s/it]
 54%|█████▍    | 10333/19090 [5:01:37<4:42:50,  1.94s/it]
 54%|█████▍    | 10334/19090 [5:01:39<4:15:06,  1.75s/it]
 54%|█████▍    | 10335/19090 [5:01:40<3:55:02,  1.61s/it]
 54%|█████▍    | 10336/19090 [5:01:43<4:33:50,  1.88s/it]
 54%|█████▍    | 10337/19090 [5:01:44<4:07:54,  1.70s/it]
 54%|█████▍    | 10338/19090 [5:01:45<3:49:43,  1.57s/it]
 54%|█████▍    | 10339/19090 [5:01:48<4:46:05,  1.96s/it]
 54%|█████▍    | 10340/19090 [5:01:49<4:16:37,  1.76s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 100.71959686279297, 'learning_rate': 2.573359123840431e-05, 'recon_loss': 0.0014116271486273035, 'recon_static': 0.0004541710834018886, 'recon_motion': 0.0009574560477631167, 'recon_norm': 1.2399487733840941, 'recon_norm_std': 0.04528857860714197, 'target_norm': 1.6186523526906966, 'recon_cos': 0.6682864353060722, 'gradnorm/compressor.query': 0.11665221987695597, 'gradnorm/compressor.layer0': 3.085986636772568, 'gradnorm/compressor.layer1': 98.34411244769562, 'gradnorm/compressor.layer2': 19.00522581813848, 'gradnorm/compressor.layer3': 9.067900633521193, 'gradnorm/compressor.layer4': 3.228870222369247, 'gradnorm/compressor.layer5': 1.8956605042988561, 'gradnorm/compressor.layer6': 1.1051361338444297, 'gradnorm/compressor.layer7': 1.2402109699593975, 'gradnorm/decoder.layer0': 1.4819219296544675, 'gradnorm/decoder.layer1': 0.8403813969920314, 'gradnorm/decoder.layer2': 0.4800482359630562, 'gradnorm/max_param': 69.36914825439453, 'epoch': 5.42}

 54%|█████▍    | 10340/19090 [5:01:49<4:16:37,  1.76s/it]
 54%|█████▍    | 10341/19090 [5:01:51<3:55:50,  1.62s/it]
 54%|█████▍    | 10342/19090 [5:01:53<4:51:15,  2.00s/it]
 54%|█████▍    | 10343/19090 [5:01:55<4:20:25,  1.79s/it]
 54%|█████▍    | 10344/19090 [5:01:56<4:08:49,  1.71s/it]
 54%|█████▍    | 10345/19090 [5:01:59<4:48:14,  1.98s/it]
 54%|█████▍    | 10346/19090 [5:02:00<4:18:14,  1.77s/it]
 54%|█████▍    | 10347/19090 [5:02:01<3:56:58,  1.63s/it]
 54%|█████▍    | 10348/19090 [5:02:04<4:47:47,  1.98s/it]
 54%|█████▍    | 10349/19090 [5:02:06<4:17:28,  1.77s/it]
 54%|█████▍    | 10350/19090 [5:02:07<3:56:23,  1.62s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 115.4942398071289, 'learning_rate': 2.5687896541345185e-05, 'recon_loss': 0.0014127985312370583, 'recon_static': 0.00043985212323605083, 'recon_motion': 0.000972946391266305, 'recon_norm': 1.235345685482025, 'recon_norm_std': 0.04523746157065034, 'target_norm': 1.6224441528320312, 'recon_cos': 0.6697776734828949, 'gradnorm/compressor.query': 0.11695583014204164, 'gradnorm/compressor.layer0': 3.3126563269236997, 'gradnorm/compressor.layer1': 112.74595076570378, 'gradnorm/compressor.layer2': 22.83473739576599, 'gradnorm/compressor.layer3': 8.607755809908168, 'gradnorm/compressor.layer4': 3.25387474310414, 'gradnorm/compressor.layer5': 2.061625208436547, 'gradnorm/compressor.layer6': 0.9918987725142691, 'gradnorm/compressor.layer7': 1.334817651232074, 'gradnorm/decoder.layer0': 1.4671234701889186, 'gradnorm/decoder.layer1': 0.8852183886649925, 'gradnorm/decoder.layer2': 0.4726553830225536, 'gradnorm/max_param': 79.51644897460938, 'epoch': 5.42}

 54%|█████▍    | 10350/19090 [5:02:07<3:56:23,  1.62s/it]
 54%|█████▍    | 10351/19090 [5:02:09<4:31:47,  1.87s/it]
 54%|█████▍    | 10352/19090 [5:02:11<4:06:23,  1.69s/it]
 54%|█████▍    | 10353/19090 [5:02:12<4:01:11,  1.66s/it]
 54%|█████▍    | 10354/19090 [5:02:15<4:48:58,  1.98s/it]
 54%|█████▍    | 10355/19090 [5:02:16<4:18:23,  1.77s/it]
 54%|█████▍    | 10356/19090 [5:02:18<4:04:04,  1.68s/it]
 54%|█████▍    | 10357/19090 [5:02:20<4:44:38,  1.96s/it]
 54%|█████▍    | 10358/19090 [5:02:22<4:15:46,  1.76s/it]
 54%|█████▍    | 10359/19090 [5:02:23<3:55:34,  1.62s/it]
 54%|█████▍    | 10360/19090 [5:02:26<4:50:40,  2.00s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 106.25129699707031, 'learning_rate': 2.5642199544296447e-05, 'recon_loss': 0.0014223740174202249, 'recon_static': 0.0004536930195172317, 'recon_motion': 0.0009686810153652914, 'recon_norm': 1.2441646307706833, 'recon_norm_std': 0.04467710293829441, 'target_norm': 1.620461842417717, 'recon_cos': 0.6669065654277802, 'gradnorm/compressor.query': 0.13491986611731147, 'gradnorm/compressor.layer0': 4.021749731334517, 'gradnorm/compressor.layer1': 103.37314368798842, 'gradnorm/compressor.layer2': 22.358268065279706, 'gradnorm/compressor.layer3': 8.31630487769945, 'gradnorm/compressor.layer4': 2.9706678118792005, 'gradnorm/compressor.layer5': 2.0073582473056777, 'gradnorm/compressor.layer6': 0.9978916473013105, 'gradnorm/compressor.layer7': 0.988865615508947, 'gradnorm/decoder.layer0': 1.4861516659789444, 'gradnorm/decoder.layer1': 0.9105631692247499, 'gradnorm/decoder.layer2': 0.46771169744809576, 'gradnorm/max_param': 72.90657806396484, 'epoch': 5.43}

 54%|█████▍    | 10360/19090 [5:02:26<4:50:40,  2.00s/it]
 54%|█████▍    | 10361/19090 [5:02:27<4:20:01,  1.79s/it]
 54%|█████▍    | 10362/19090 [5:02:28<3:58:06,  1.64s/it]
 54%|█████▍    | 10363/19090 [5:02:31<4:41:50,  1.94s/it]
 54%|█████▍    | 10364/19090 [5:02:32<4:13:49,  1.75s/it]
 54%|█████▍    | 10365/19090 [5:02:34<3:54:50,  1.61s/it]
 54%|█████▍    | 10366/19090 [5:02:36<4:33:17,  1.88s/it]
 54%|█████▍    | 10367/19090 [5:02:37<4:07:26,  1.70s/it]
 54%|█████▍    | 10368/19090 [5:02:39<3:59:04,  1.64s/it]
 54%|█████▍    | 10369/19090 [5:02:41<4:35:32,  1.90s/it]
 54%|█████▍    | 10370/19090 [5:02:43<4:09:06,  1.71s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 109.04254913330078, 'learning_rate': 2.5596500400046507e-05, 'recon_loss': 0.0014155195356579497, 'recon_static': 0.0004575693026708905, 'recon_motion': 0.0009579502657288685, 'recon_norm': 1.2409049302339554, 'recon_norm_std': 0.0458976536989212, 'target_norm': 1.6188924372196198, 'recon_cos': 0.6673960521817207, 'gradnorm/compressor.query': 0.12529632343846817, 'gradnorm/compressor.layer0': 3.549130835715398, 'gradnorm/compressor.layer1': 105.7166900479177, 'gradnorm/compressor.layer2': 24.35974457505084, 'gradnorm/compressor.layer3': 9.226786585710228, 'gradnorm/compressor.layer4': 3.5126111484739244, 'gradnorm/compressor.layer5': 2.0414459212981764, 'gradnorm/compressor.layer6': 1.1111641546674635, 'gradnorm/compressor.layer7': 1.2025323869781064, 'gradnorm/decoder.layer0': 1.6253548748152489, 'gradnorm/decoder.layer1': 0.9969335568727739, 'gradnorm/decoder.layer2': 0.5427978576020512, 'gradnorm/max_param': 74.5686264038086, 'epoch': 5.43}

 54%|█████▍    | 10370/19090 [5:02:43<4:09:06,  1.71s/it]
 54%|█████▍    | 10371/19090 [5:02:44<4:05:11,  1.69s/it]
 54%|█████▍    | 10372/19090 [5:02:47<4:34:23,  1.89s/it]
 54%|█████▍    | 10373/19090 [5:02:48<4:08:14,  1.71s/it]
 54%|█████▍    | 10374/19090 [5:02:49<3:58:49,  1.64s/it]
 54%|█████▍    | 10375/19090 [5:02:52<4:41:45,  1.94s/it]
 54%|█████▍    | 10376/19090 [5:02:53<4:13:16,  1.74s/it]
 54%|█████▍    | 10377/19090 [5:02:55<4:10:23,  1.72s/it]
 54%|█████▍    | 10378/19090 [5:02:58<4:49:01,  1.99s/it]
 54%|█████▍    | 10379/19090 [5:02:59<4:18:39,  1.78s/it]
 54%|█████▍    | 10380/19090 [5:03:00<4:02:42,  1.67s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 108.50267791748047, 'learning_rate': 2.5550799261390952e-05, 'recon_loss': 0.0014357619220390916, 'recon_static': 0.0004617426071490627, 'recon_motion': 0.0009740193199831992, 'recon_norm': 1.2471941113471985, 'recon_norm_std': 0.044941456895321606, 'target_norm': 1.6234881371259688, 'recon_cos': 0.6651329070329666, 'gradnorm/compressor.query': 0.10645142371457224, 'gradnorm/compressor.layer0': 3.22422354762296, 'gradnorm/compressor.layer1': 105.92293410187428, 'gradnorm/compressor.layer2': 20.970248991689093, 'gradnorm/compressor.layer3': 9.084579615589936, 'gradnorm/compressor.layer4': 3.2326229344097306, 'gradnorm/compressor.layer5': 2.045302624310693, 'gradnorm/compressor.layer6': 1.091518414358208, 'gradnorm/compressor.layer7': 1.1312714716944283, 'gradnorm/decoder.layer0': 1.4882888593504209, 'gradnorm/decoder.layer1': 0.9358192955101051, 'gradnorm/decoder.layer2': 0.5298090634906754, 'gradnorm/max_param': 74.71630096435547, 'epoch': 5.44}

 54%|█████▍    | 10380/19090 [5:03:00<4:02:42,  1.67s/it]
 54%|█████▍    | 10381/19090 [5:03:03<4:46:46,  1.98s/it]
 54%|█████▍    | 10382/19090 [5:03:04<4:16:48,  1.77s/it]
 54%|█████▍    | 10383/19090 [5:03:06<4:12:14,  1.74s/it]
 54%|█████▍    | 10384/19090 [5:03:08<4:33:33,  1.89s/it]
 54%|█████▍    | 10385/19090 [5:03:09<4:07:33,  1.71s/it]
 54%|█████▍    | 10386/19090 [5:03:12<4:24:14,  1.82s/it]
 54%|█████▍    | 10387/19090 [5:03:14<4:38:43,  1.92s/it]
 54%|█████▍    | 10388/19090 [5:03:15<4:11:25,  1.73s/it]
 54%|█████▍    | 10389/19090 [5:03:17<4:19:54,  1.79s/it]
 54%|█████▍    | 10390/19090 [5:03:19<4:38:33,  1.92s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 107.14340209960938, 'learning_rate': 2.550509628113204e-05, 'recon_loss': 0.0014854304696200415, 'recon_static': 0.0005136750150995795, 'recon_motion': 0.000971755469799973, 'recon_norm': 1.2676535636186599, 'recon_norm_std': 0.04499744549393654, 'target_norm': 1.621812751889229, 'recon_cos': 0.6518221855163574, 'gradnorm/compressor.query': 0.1309899987719137, 'gradnorm/compressor.layer0': 3.802665031909814, 'gradnorm/compressor.layer1': 104.37324100314218, 'gradnorm/compressor.layer2': 21.378722463466378, 'gradnorm/compressor.layer3': 9.536294371471314, 'gradnorm/compressor.layer4': 3.496153898094676, 'gradnorm/compressor.layer5': 2.025239826647262, 'gradnorm/compressor.layer6': 1.0785387876041257, 'gradnorm/compressor.layer7': 1.2800913433175904, 'gradnorm/decoder.layer0': 1.7278528699403584, 'gradnorm/decoder.layer1': 1.014774729530513, 'gradnorm/decoder.layer2': 0.5706982810188903, 'gradnorm/max_param': 73.623779296875, 'epoch': 5.44}

 54%|█████▍    | 10390/19090 [5:03:19<4:38:33,  1.92s/it]
 54%|█████▍    | 10391/19090 [5:03:20<4:11:21,  1.73s/it]
 54%|█████▍    | 10392/19090 [5:03:22<4:02:36,  1.67s/it]
 54%|█████▍    | 10393/19090 [5:03:24<4:24:02,  1.82s/it]
 54%|█████▍    | 10394/19090 [5:03:25<4:00:55,  1.66s/it]
 54%|█████▍    | 10395/19090 [5:03:27<4:12:03,  1.74s/it]
 54%|█████▍    | 10396/19090 [5:03:29<4:12:14,  1.74s/it]
 54%|█████▍    | 10397/19090 [5:03:30<3:52:38,  1.61s/it]
 54%|█████▍    | 10398/19090 [5:03:33<4:21:50,  1.81s/it]
 54%|█████▍    | 10399/19090 [5:03:35<4:25:00,  1.83s/it]
 54%|█████▍    | 10400/19090 [5:03:36<4:01:46,  1.67s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 106.91939544677734, 'learning_rate': 2.5459391612078186e-05, 'recon_loss': 0.0014434141136007383, 'recon_static': 0.0004915315192192792, 'recon_motion': 0.0009518825580016709, 'recon_norm': 1.2595598101615906, 'recon_norm_std': 0.04456391502171755, 'target_norm': 1.6169909864664078, 'recon_cos': 0.6601249769330024, 'gradnorm/compressor.query': 0.09818648395340433, 'gradnorm/compressor.layer0': 3.2242133430526203, 'gradnorm/compressor.layer1': 104.16754739211008, 'gradnorm/compressor.layer2': 21.290130420462727, 'gradnorm/compressor.layer3': 9.761539770518084, 'gradnorm/compressor.layer4': 3.2898462951838736, 'gradnorm/compressor.layer5': 2.1113032816820065, 'gradnorm/compressor.layer6': 1.1165888856492163, 'gradnorm/compressor.layer7': 1.2772244272290916, 'gradnorm/decoder.layer0': 1.5424211906690886, 'gradnorm/decoder.layer1': 1.044993688664702, 'gradnorm/decoder.layer2': 0.5333399614528225, 'gradnorm/max_param': 73.48088073730469, 'epoch': 5.45}

 54%|█████▍    | 10400/19090 [5:03:36<4:01:46,  1.67s/it]
 54%|█████▍    | 10401/19090 [5:03:38<4:22:44,  1.81s/it]
 54%|█████▍    | 10402/19090 [5:03:40<4:25:08,  1.83s/it]
 54%|█████▍    | 10403/19090 [5:03:41<4:01:55,  1.67s/it]
 54%|█████▍    | 10404/19090 [5:03:43<4:16:34,  1.77s/it]
 55%|█████▍    | 10405/19090 [5:03:45<4:28:38,  1.86s/it]
 55%|█████▍    | 10406/19090 [5:03:46<4:04:11,  1.69s/it]
 55%|█████▍    | 10407/19090 [5:03:49<4:25:01,  1.83s/it]
 55%|█████▍    | 10408/19090 [5:03:51<4:32:51,  1.89s/it]
 55%|█████▍    | 10409/19090 [5:03:52<4:07:06,  1.71s/it]
 55%|█████▍    | 10410/19090 [5:03:54<4:22:48,  1.82s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 116.61878204345703, 'learning_rate': 2.5413685407043454e-05, 'recon_loss': 0.0014697637176141142, 'recon_static': 0.0005060937131929677, 'recon_motion': 0.0009636699658585712, 'recon_norm': 1.2644073069095612, 'recon_norm_std': 0.04461463568732142, 'target_norm': 1.622482094168663, 'recon_cos': 0.6561064004898072, 'gradnorm/compressor.query': 0.10788982942137305, 'gradnorm/compressor.layer0': 3.1799058394016853, 'gradnorm/compressor.layer1': 113.54842630872301, 'gradnorm/compressor.layer2': 24.113768181641973, 'gradnorm/compressor.layer3': 9.56552694445516, 'gradnorm/compressor.layer4': 3.4228723196143083, 'gradnorm/compressor.layer5': 2.0242813295020183, 'gradnorm/compressor.layer6': 1.230826383419126, 'gradnorm/compressor.layer7': 1.398659459406124, 'gradnorm/decoder.layer0': 1.7088800888940054, 'gradnorm/decoder.layer1': 1.0434438359198723, 'gradnorm/decoder.layer2': 0.5510905053670798, 'gradnorm/max_param': 80.09961700439453, 'epoch': 5.45}

 55%|█████▍    | 10410/19090 [5:03:54<4:22:48,  1.82s/it]
 55%|█████▍    | 10411/19090 [5:03:56<4:28:29,  1.86s/it]
 55%|█████▍    | 10412/19090 [5:03:57<4:03:59,  1.69s/it]
 55%|█████▍    | 10413/19090 [5:03:59<4:25:29,  1.84s/it]
 55%|█████▍    | 10414/19090 [5:04:01<4:24:28,  1.83s/it]
 55%|█████▍    | 10415/19090 [5:04:03<4:00:42,  1.66s/it]
 55%|█████▍    | 10416/19090 [5:04:05<4:25:34,  1.84s/it]
 55%|█████▍    | 10417/19090 [5:04:07<4:22:11,  1.81s/it]
 55%|█████▍    | 10418/19090 [5:04:08<3:59:19,  1.66s/it]
 55%|█████▍    | 10419/19090 [5:04:10<4:21:53,  1.81s/it]
 55%|█████▍    | 10420/19090 [5:04:12<4:21:10,  1.81s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 111.7408447265625, 'learning_rate': 2.5367977818847034e-05, 'recon_loss': 0.0014562323252903298, 'recon_static': 0.0004978144359483849, 'recon_motion': 0.0009584178769728169, 'recon_norm': 1.2621810972690581, 'recon_norm_std': 0.044893690198659894, 'target_norm': 1.6203152984380722, 'recon_cos': 0.6585546910762787, 'gradnorm/compressor.query': 0.12219431110579289, 'gradnorm/compressor.layer0': 3.679364038330877, 'gradnorm/compressor.layer1': 108.9345775568988, 'gradnorm/compressor.layer2': 22.27389306465889, 'gradnorm/compressor.layer3': 9.415770316434841, 'gradnorm/compressor.layer4': 3.1476450033926655, 'gradnorm/compressor.layer5': 2.0139218017110254, 'gradnorm/compressor.layer6': 1.1212753520978547, 'gradnorm/compressor.layer7': 1.406949314440807, 'gradnorm/decoder.layer0': 1.5477032226335712, 'gradnorm/decoder.layer1': 1.0458590004090285, 'gradnorm/decoder.layer2': 0.49835411019959697, 'gradnorm/max_param': 76.82837677001953, 'epoch': 5.46}

 55%|█████▍    | 10420/19090 [5:04:12<4:21:10,  1.81s/it]
 55%|█████▍    | 10421/19090 [5:04:13<4:01:31,  1.67s/it]
 55%|█████▍    | 10422/19090 [5:04:15<4:23:34,  1.82s/it]
 55%|█████▍    | 10423/19090 [5:04:17<4:28:26,  1.86s/it]
 55%|█████▍    | 10424/19090 [5:04:19<4:03:34,  1.69s/it]
 55%|█████▍    | 10425/19090 [5:04:21<4:45:54,  1.98s/it]
 55%|█████▍    | 10426/19090 [5:04:23<4:25:07,  1.84s/it]
 55%|█████▍    | 10427/19090 [5:04:24<4:01:14,  1.67s/it]
 55%|█████▍    | 10428/19090 [5:04:27<4:40:41,  1.94s/it]
 55%|█████▍    | 10429/19090 [5:04:28<4:20:09,  1.80s/it]
 55%|█████▍    | 10430/19090 [5:04:29<3:57:40,  1.65s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 144.11192321777344, 'learning_rate': 2.532226900031275e-05, 'recon_loss': 0.0014087356597883628, 'recon_static': 0.0004585704555211123, 'recon_motion': 0.000950165190442931, 'recon_norm': 1.2416272968053819, 'recon_norm_std': 0.04651588164269924, 'target_norm': 1.6186528712511064, 'recon_cos': 0.6695305123925209, 'gradnorm/compressor.query': 0.12660173791004, 'gradnorm/compressor.layer0': 3.934493233841064, 'gradnorm/compressor.layer1': 140.7009170079392, 'gradnorm/compressor.layer2': 27.88938088086519, 'gradnorm/compressor.layer3': 12.213429573157434, 'gradnorm/compressor.layer4': 3.973916483100852, 'gradnorm/compressor.layer5': 2.2348784833128086, 'gradnorm/compressor.layer6': 1.3418219284983652, 'gradnorm/compressor.layer7': 1.4216581745199663, 'gradnorm/decoder.layer0': 1.7035342783388852, 'gradnorm/decoder.layer1': 1.0792082168553314, 'gradnorm/decoder.layer2': 0.5839979750666721, 'gradnorm/max_param': 99.23465728759766, 'epoch': 5.46}

 55%|█████▍    | 10430/19090 [5:04:29<3:57:40,  1.65s/it]
 55%|█████▍    | 10431/19090 [5:04:32<4:35:00,  1.91s/it]
 55%|█████▍    | 10432/19090 [5:04:34<4:32:54,  1.89s/it]
 55%|█████▍    | 10433/19090 [5:04:35<4:06:30,  1.71s/it]
 55%|█████▍    | 10434/19090 [5:04:37<4:33:20,  1.89s/it]
 55%|█████▍    | 10435/19090 [5:04:39<4:28:45,  1.86s/it]
 55%|█████▍    | 10436/19090 [5:04:40<4:03:32,  1.69s/it]
 55%|█████▍    | 10437/19090 [5:04:43<4:37:57,  1.93s/it]
 55%|█████▍    | 10438/19090 [5:04:44<4:23:18,  1.83s/it]
 55%|█████▍    | 10439/19090 [5:04:46<3:59:49,  1.66s/it]
 55%|█████▍    | 10440/19090 [5:04:48<4:40:10,  1.94s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 112.50802612304688, 'learning_rate': 2.527655910426854e-05, 'recon_loss': 0.001431472465628758, 'recon_static': 0.0004663195453758817, 'recon_motion': 0.0009651529224356636, 'recon_norm': 1.2466866314411162, 'recon_norm_std': 0.04566761376336217, 'target_norm': 1.622064471244812, 'recon_cos': 0.6652845367789268, 'gradnorm/compressor.query': 0.1020713812738896, 'gradnorm/compressor.layer0': 3.0804841736057416, 'gradnorm/compressor.layer1': 109.84407005643273, 'gradnorm/compressor.layer2': 21.829477605506202, 'gradnorm/compressor.layer3': 9.311099772824486, 'gradnorm/compressor.layer4': 3.106616306005893, 'gradnorm/compressor.layer5': 2.0328603376760386, 'gradnorm/compressor.layer6': 1.1351327745356004, 'gradnorm/compressor.layer7': 1.1515374935338762, 'gradnorm/decoder.layer0': 1.436159504030613, 'gradnorm/decoder.layer1': 0.9693923174395281, 'gradnorm/decoder.layer2': 0.4605274477833624, 'gradnorm/max_param': 77.4906997680664, 'epoch': 5.47}

 55%|█████▍    | 10440/19090 [5:04:48<4:40:10,  1.94s/it]
 55%|█████▍    | 10441/19090 [5:04:50<4:14:09,  1.76s/it]
 55%|█████▍    | 10442/19090 [5:04:51<3:52:56,  1.62s/it]
 55%|█████▍    | 10443/19090 [5:04:54<4:41:34,  1.95s/it]
 55%|█████▍    | 10444/19090 [5:04:55<4:12:50,  1.75s/it]
 55%|█████▍    | 10445/19090 [5:04:56<3:52:27,  1.61s/it]
 55%|█████▍    | 10446/19090 [5:04:59<4:27:31,  1.86s/it]
 55%|█████▍    | 10447/19090 [5:05:00<4:13:53,  1.76s/it]
 55%|█████▍    | 10448/19090 [5:05:02<3:53:53,  1.62s/it]
 55%|█████▍    | 10449/19090 [5:05:04<4:33:31,  1.90s/it]
 55%|█████▍    | 10450/19090 [5:05:06<4:25:33,  1.84s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 117.96817779541016, 'learning_rate': 2.523084828354594e-05, 'recon_loss': 0.0014095436199568212, 'recon_static': 0.0004491010899073444, 'recon_motion': 0.0009604425722500309, 'recon_norm': 1.2369609534740449, 'recon_norm_std': 0.045195665955543515, 'target_norm': 1.6194848090410232, 'recon_cos': 0.6692546352744102, 'gradnorm/compressor.query': 0.1753917352093263, 'gradnorm/compressor.layer0': 3.70746792471368, 'gradnorm/compressor.layer1': 115.8113447729755, 'gradnorm/compressor.layer2': 19.36412875452637, 'gradnorm/compressor.layer3': 9.574618371275506, 'gradnorm/compressor.layer4': 3.6668002508856437, 'gradnorm/compressor.layer5': 2.0279146246987665, 'gradnorm/compressor.layer6': 1.091892790974236, 'gradnorm/compressor.layer7': 1.2634512990329607, 'gradnorm/decoder.layer0': 1.4900269370236332, 'gradnorm/decoder.layer1': 1.0004403813570204, 'gradnorm/decoder.layer2': 0.4859564099818214, 'gradnorm/max_param': 81.6838150024414, 'epoch': 5.47}

 55%|█████▍    | 10450/19090 [5:05:06<4:25:33,  1.84s/it]
 55%|█████▍    | 10451/19090 [5:05:07<4:01:39,  1.68s/it]
 55%|█████▍    | 10452/19090 [5:05:10<4:33:37,  1.90s/it]
 55%|█████▍    | 10453/19090 [5:05:11<4:25:28,  1.84s/it]
 55%|█████▍    | 10454/19090 [5:05:13<4:01:35,  1.68s/it]
 55%|█████▍    | 10455/19090 [5:05:15<4:27:13,  1.86s/it]
 55%|█████▍    | 10456/19090 [5:05:17<4:23:48,  1.83s/it]
 55%|█████▍    | 10457/19090 [5:05:18<4:00:16,  1.67s/it]
 55%|█████▍    | 10458/19090 [5:05:20<4:37:40,  1.93s/it]
 55%|█████▍    | 10459/19090 [5:05:22<4:20:03,  1.81s/it]
 55%|█████▍    | 10460/19090 [5:05:23<3:57:37,  1.65s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 96.34660339355469, 'learning_rate': 2.5185136690979567e-05, 'recon_loss': 0.001392439001938328, 'recon_static': 0.0004374328753328882, 'recon_motion': 0.0009550061004119925, 'recon_norm': 1.2346462339162827, 'recon_norm_std': 0.04594145147129893, 'target_norm': 1.6176780104637145, 'recon_cos': 0.6733511820435524, 'gradnorm/compressor.query': 0.08772630961449052, 'gradnorm/compressor.layer0': 2.4788605530401377, 'gradnorm/compressor.layer1': 94.20780927356819, 'gradnorm/compressor.layer2': 17.655247482600604, 'gradnorm/compressor.layer3': 8.210495665592122, 'gradnorm/compressor.layer4': 3.33369860237244, 'gradnorm/compressor.layer5': 2.170680424240498, 'gradnorm/compressor.layer6': 1.1223702424119864, 'gradnorm/compressor.layer7': 1.247949539711106, 'gradnorm/decoder.layer0': 1.5374037425605416, 'gradnorm/decoder.layer1': 0.9926113514929243, 'gradnorm/decoder.layer2': 0.5417493396415259, 'gradnorm/max_param': 66.4515380859375, 'epoch': 5.48}

 55%|█████▍    | 10460/19090 [5:05:23<3:57:37,  1.65s/it]
 55%|█████▍    | 10461/19090 [5:05:26<4:31:16,  1.89s/it]
 55%|█████▍    | 10462/19090 [5:05:27<4:23:23,  1.83s/it]
 55%|█████▍    | 10463/19090 [5:05:29<3:59:58,  1.67s/it]
 55%|█████▍    | 10464/19090 [5:05:31<4:31:13,  1.89s/it]
 55%|█████▍    | 10465/19090 [5:05:33<4:17:18,  1.79s/it]
 55%|█████▍    | 10466/19090 [5:05:34<3:55:38,  1.64s/it]
 55%|█████▍    | 10467/19090 [5:05:36<4:30:19,  1.88s/it]
 55%|█████▍    | 10468/19090 [5:05:38<4:24:08,  1.84s/it]
 55%|█████▍    | 10469/19090 [5:05:39<4:00:18,  1.67s/it]
 55%|█████▍    | 10470/19090 [5:05:42<4:37:39,  1.93s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 121.77206420898438, 'learning_rate': 2.5139424479406653e-05, 'recon_loss': 0.0014279726427048444, 'recon_static': 0.00045585434781969525, 'recon_motion': 0.0009721183552755975, 'recon_norm': 1.2418433904647828, 'recon_norm_std': 0.04482741085812449, 'target_norm': 1.6216614484786986, 'recon_cos': 0.6656240731477737, 'gradnorm/compressor.query': 0.1495656945727898, 'gradnorm/compressor.layer0': 3.9180216914029713, 'gradnorm/compressor.layer1': 118.50477558283252, 'gradnorm/compressor.layer2': 24.996292449690788, 'gradnorm/compressor.layer3': 11.051542038145556, 'gradnorm/compressor.layer4': 3.488487656179872, 'gradnorm/compressor.layer5': 1.995175622652871, 'gradnorm/compressor.layer6': 1.1260641680976813, 'gradnorm/compressor.layer7': 1.225111960140938, 'gradnorm/decoder.layer0': 1.5671559202258563, 'gradnorm/decoder.layer1': 1.0421737326837928, 'gradnorm/decoder.layer2': 0.5199555980083519, 'gradnorm/max_param': 83.58779907226562, 'epoch': 5.48}

 55%|█████▍    | 10470/19090 [5:05:42<4:37:39,  1.93s/it]
 55%|█████▍    | 10471/19090 [5:05:43<4:16:47,  1.79s/it]
 55%|█████▍    | 10472/19090 [5:05:45<3:55:15,  1.64s/it]
 55%|█████▍    | 10473/19090 [5:05:47<4:29:34,  1.88s/it]
 55%|█████▍    | 10474/19090 [5:05:49<4:32:09,  1.90s/it]
 55%|█████▍    | 10475/19090 [5:05:50<4:05:50,  1.71s/it]
 55%|█████▍    | 10476/19090 [5:05:52<4:09:40,  1.74s/it]
 55%|█████▍    | 10477/19090 [5:05:54<4:24:02,  1.84s/it]
 55%|█████▍    | 10478/19090 [5:05:55<4:00:15,  1.67s/it]
 55%|█████▍    | 10479/19090 [5:05:57<4:12:27,  1.76s/it]
 55%|█████▍    | 10480/19090 [5:06:00<4:31:50,  1.89s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 128.2342529296875, 'learning_rate': 2.5093711801666466e-05, 'recon_loss': 0.0014256264577852563, 'recon_static': 0.00046951406475272963, 'recon_motion': 0.0009561123792082071, 'recon_norm': 1.2479560166597365, 'recon_norm_std': 0.04592075953260064, 'target_norm': 1.6174716383218766, 'recon_cos': 0.6646478220820426, 'gradnorm/compressor.query': 0.1351501328401516, 'gradnorm/compressor.layer0': 3.7689628528313928, 'gradnorm/compressor.layer1': 125.22772225779322, 'gradnorm/compressor.layer2': 24.273637400609122, 'gradnorm/compressor.layer3': 11.429217565527233, 'gradnorm/compressor.layer4': 3.9775815006636113, 'gradnorm/compressor.layer5': 2.211228530115384, 'gradnorm/compressor.layer6': 1.1753945459292205, 'gradnorm/compressor.layer7': 1.3319691842125911, 'gradnorm/decoder.layer0': 1.684915116999165, 'gradnorm/decoder.layer1': 0.9727490239154406, 'gradnorm/decoder.layer2': 0.5739887321808353, 'gradnorm/max_param': 88.33472442626953, 'epoch': 5.49}

 55%|█████▍    | 10480/19090 [5:06:00<4:31:50,  1.89s/it]
 55%|█████▍    | 10481/19090 [5:06:01<4:05:44,  1.71s/it]
 55%|█████▍    | 10482/19090 [5:06:03<4:03:16,  1.70s/it]
 55%|█████▍    | 10483/19090 [5:06:05<4:28:53,  1.87s/it]
 55%|█████▍    | 10484/19090 [5:06:06<4:03:40,  1.70s/it]
 55%|█████▍    | 10485/19090 [5:06:08<4:04:45,  1.71s/it]
 55%|█████▍    | 10486/19090 [5:06:10<4:35:35,  1.92s/it]
 55%|█████▍    | 10487/19090 [5:06:12<4:08:15,  1.73s/it]
 55%|█████▍    | 10488/19090 [5:06:13<3:55:50,  1.65s/it]
 55%|█████▍    | 10489/19090 [5:06:16<4:45:08,  1.99s/it]
 55%|█████▍    | 10490/19090 [5:06:17<4:15:02,  1.78s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 126.8998031616211, 'learning_rate': 2.5047998810599837e-05, 'recon_loss': 0.0014177167759044096, 'recon_static': 0.0004564344242680818, 'recon_motion': 0.0009612823982024565, 'recon_norm': 1.2438048243522644, 'recon_norm_std': 0.04528724895790219, 'target_norm': 1.617970472574234, 'recon_cos': 0.666951110959053, 'gradnorm/compressor.query': 0.11272254413192402, 'gradnorm/compressor.layer0': 3.5358273706888603, 'gradnorm/compressor.layer1': 123.80937259427293, 'gradnorm/compressor.layer2': 25.008294522455444, 'gradnorm/compressor.layer3': 10.630068299441037, 'gradnorm/compressor.layer4': 3.66036872694871, 'gradnorm/compressor.layer5': 1.966911668656878, 'gradnorm/compressor.layer6': 1.0844732058685573, 'gradnorm/compressor.layer7': 1.292033246023002, 'gradnorm/decoder.layer0': 1.566244977987143, 'gradnorm/decoder.layer1': 1.0187827099188531, 'gradnorm/decoder.layer2': 0.5166383092030278, 'gradnorm/max_param': 87.32373809814453, 'epoch': 5.5}

 55%|█████▍    | 10490/19090 [5:06:17<4:15:02,  1.78s/it]
 55%|█████▍    | 10491/19090 [5:06:18<3:53:51,  1.63s/it]
 55%|█████▍    | 10492/19090 [5:06:21<4:47:05,  2.00s/it]
 55%|█████▍    | 10493/19090 [5:06:23<4:16:14,  1.79s/it]
 55%|█████▍    | 10494/19090 [5:06:24<3:54:54,  1.64s/it]
 55%|█████▍    | 10495/19090 [5:06:26<4:37:51,  1.94s/it]
 55%|█████▍    | 10496/19090 [5:06:28<4:09:37,  1.74s/it]
 55%|█████▍    | 10497/19090 [5:06:29<3:50:16,  1.61s/it]
 55%|█████▍    | 10498/19090 [5:06:32<4:41:54,  1.97s/it]
 55%|█████▍    | 10499/19090 [5:06:33<4:12:32,  1.76s/it]
 55%|█████▌    | 10500/19090 [5:06:34<3:52:06,  1.62s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 88.5378189086914, 'learning_rate': 2.5002285659048674e-05, 'recon_loss': 0.0014249637169996276, 'recon_static': 0.00045884285209467633, 'recon_motion': 0.0009661209289333784, 'recon_norm': 1.241689458489418, 'recon_norm_std': 0.044319081399589774, 'target_norm': 1.62066550552845, 'recon_cos': 0.6660337090492249, 'gradnorm/compressor.query': 0.09425371666240648, 'gradnorm/compressor.layer0': 2.5675440201147577, 'gradnorm/compressor.layer1': 85.1789218672598, 'gradnorm/compressor.layer2': 21.980026194562914, 'gradnorm/compressor.layer3': 8.646775331463893, 'gradnorm/compressor.layer4': 2.9569580391572883, 'gradnorm/compressor.layer5': 2.0107091055210216, 'gradnorm/compressor.layer6': 1.0470722638926568, 'gradnorm/compressor.layer7': 1.20352067262259, 'gradnorm/decoder.layer0': 1.5963232920048012, 'gradnorm/decoder.layer1': 0.9187138881203714, 'gradnorm/decoder.layer2': 0.5327387873428692, 'gradnorm/max_param': 60.0916633605957, 'epoch': 5.5}

 55%|█████▌    | 10500/19090 [5:06:34<3:52:06,  1.62s/it]
 55%|█████▌    | 10501/19090 [5:06:38<5:31:54,  2.32s/it]
 55%|█████▌    | 10502/19090 [5:06:40<4:47:21,  2.01s/it]
 55%|█████▌    | 10503/19090 [5:06:41<4:16:14,  1.79s/it]
 55%|█████▌    | 10504/19090 [5:06:42<3:54:28,  1.64s/it]
 55%|█████▌    | 10505/19090 [5:06:44<3:39:16,  1.53s/it]
 55%|█████▌    | 10506/19090 [5:06:45<3:28:38,  1.46s/it]
 55%|█████▌    | 10507/19090 [5:06:47<4:11:16,  1.76s/it]
 55%|█████▌    | 10508/19090 [5:06:49<3:51:00,  1.62s/it]
 55%|█████▌    | 10509/19090 [5:06:50<3:36:42,  1.52s/it]
 55%|█████▌    | 10510/19090 [5:06:52<4:21:24,  1.83s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 106.2225570678711, 'learning_rate': 2.4956572499855373e-05, 'recon_loss': 0.0014411423093406484, 'recon_static': 0.00047387116428581065, 'recon_motion': 0.0009672711676103063, 'recon_norm': 1.2493365466594697, 'recon_norm_std': 0.04515899838879704, 'target_norm': 1.6215436607599258, 'recon_cos': 0.6626529142260551, 'gradnorm/compressor.query': 0.09997145908580778, 'gradnorm/compressor.layer0': 2.9999386463249325, 'gradnorm/compressor.layer1': 104.00034273531266, 'gradnorm/compressor.layer2': 19.282172726604067, 'gradnorm/compressor.layer3': 8.317065369166006, 'gradnorm/compressor.layer4': 2.8480200934076314, 'gradnorm/compressor.layer5': 1.7702611634638223, 'gradnorm/compressor.layer6': 1.17568558718825, 'gradnorm/compressor.layer7': 1.1915769861172627, 'gradnorm/decoder.layer0': 1.4927439825358033, 'gradnorm/decoder.layer1': 0.806257032023673, 'gradnorm/decoder.layer2': 0.5011380476324678, 'gradnorm/max_param': 73.33683013916016, 'epoch': 5.51}

 55%|█████▌    | 10510/19090 [5:06:52<4:21:24,  1.83s/it]
 55%|█████▌    | 10511/19090 [5:06:54<3:58:00,  1.66s/it]
 55%|█████▌    | 10512/19090 [5:06:55<3:55:52,  1.65s/it]
 55%|█████▌    | 10513/19090 [5:06:58<4:55:40,  2.07s/it]
 55%|█████▌    | 10514/19090 [5:07:00<4:22:44,  1.84s/it]
 55%|█████▌    | 10515/19090 [5:07:01<3:59:26,  1.68s/it]
 55%|█████▌    | 10516/19090 [5:07:04<4:40:11,  1.96s/it]
 55%|█████▌    | 10517/19090 [5:07:05<4:11:26,  1.76s/it]
 55%|█████▌    | 10518/19090 [5:07:06<3:51:07,  1.62s/it]
 55%|█████▌    | 10519/19090 [5:07:09<4:53:16,  2.05s/it]
 55%|█████▌    | 10520/19090 [5:07:10<4:20:40,  1.83s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 108.40856170654297, 'learning_rate': 2.4910859485862395e-05, 'recon_loss': 0.0015431191626703368, 'recon_static': 0.0005830438807606697, 'recon_motion': 0.0009600753168342635, 'recon_norm': 1.3014635175466538, 'recon_norm_std': 0.04355060923844576, 'target_norm': 1.619385102391243, 'recon_cos': 0.6388744413852692, 'gradnorm/compressor.query': 0.1026941625052597, 'gradnorm/compressor.layer0': 3.292140652939262, 'gradnorm/compressor.layer1': 105.68546253949737, 'gradnorm/compressor.layer2': 21.38614118797754, 'gradnorm/compressor.layer3': 9.629071055891844, 'gradnorm/compressor.layer4': 3.481245063490595, 'gradnorm/compressor.layer5': 2.0034784823414804, 'gradnorm/compressor.layer6': 1.14019861482149, 'gradnorm/compressor.layer7': 1.2584542005902382, 'gradnorm/decoder.layer0': 1.430680643968543, 'gradnorm/decoder.layer1': 0.8686899795656025, 'gradnorm/decoder.layer2': 0.49626069751615, 'gradnorm/max_param': 74.5374984741211, 'epoch': 5.51}

 55%|█████▌    | 10520/19090 [5:07:10<4:20:40,  1.83s/it]
 55%|█████▌    | 10521/19090 [5:07:12<3:58:04,  1.67s/it]
 55%|█████▌    | 10522/19090 [5:07:15<4:48:01,  2.02s/it]
 55%|█████▌    | 10523/19090 [5:07:16<4:16:55,  1.80s/it]
 55%|█████▌    | 10524/19090 [5:07:17<3:54:54,  1.65s/it]
 55%|█████▌    | 10525/19090 [5:07:20<4:42:59,  1.98s/it]
 55%|█████▌    | 10526/19090 [5:07:21<4:13:14,  1.77s/it]
 55%|█████▌    | 10527/19090 [5:07:23<3:52:20,  1.63s/it]
 55%|█████▌    | 10528/19090 [5:07:26<4:49:47,  2.03s/it]
 55%|█████▌    | 10529/19090 [5:07:27<4:17:49,  1.81s/it]
 55%|█████▌    | 10530/19090 [5:07:28<3:55:48,  1.65s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 112.78032684326172, 'learning_rate': 2.4865146769911708e-05, 'recon_loss': 0.0014752853923710064, 'recon_static': 0.0005109035140776541, 'recon_motion': 0.000964381854282692, 'recon_norm': 1.2670103430747985, 'recon_norm_std': 0.045922210440039635, 'target_norm': 1.6235181927680968, 'recon_cos': 0.655260780453682, 'gradnorm/compressor.query': 0.1538777634839669, 'gradnorm/compressor.layer0': 4.495504889218159, 'gradnorm/compressor.layer1': 109.53947009097908, 'gradnorm/compressor.layer2': 23.874307192991225, 'gradnorm/compressor.layer3': 10.33172258518145, 'gradnorm/compressor.layer4': 3.48103192671962, 'gradnorm/compressor.layer5': 2.081888562550225, 'gradnorm/compressor.layer6': 1.2024019719255512, 'gradnorm/compressor.layer7': 1.3931082509652546, 'gradnorm/decoder.layer0': 1.5527538754180368, 'gradnorm/decoder.layer1': 1.0265285027621163, 'gradnorm/decoder.layer2': 0.49478148567031704, 'gradnorm/max_param': 77.2698974609375, 'epoch': 5.52}

 55%|█████▌    | 10530/19090 [5:07:28<3:55:48,  1.65s/it]
 55%|█████▌    | 10531/19090 [5:07:31<5:00:45,  2.11s/it]
 55%|█████▌    | 10532/19090 [5:07:33<4:25:23,  1.86s/it]
 55%|█████▌    | 10533/19090 [5:07:34<4:00:39,  1.69s/it]
 55%|█████▌    | 10534/19090 [5:07:37<4:45:07,  2.00s/it]
 55%|█████▌    | 10535/19090 [5:07:38<4:14:24,  1.78s/it]
 55%|█████▌    | 10536/19090 [5:07:39<3:52:56,  1.63s/it]
 55%|█████▌    | 10537/19090 [5:07:42<4:36:42,  1.94s/it]
 55%|█████▌    | 10538/19090 [5:07:43<4:08:33,  1.74s/it]
 55%|█████▌    | 10539/19090 [5:07:44<3:48:53,  1.61s/it]
 55%|█████▌    | 10540/19090 [5:07:47<4:31:47,  1.91s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 109.0300064086914, 'learning_rate': 2.481943450484427e-05, 'recon_loss': 0.001491174686816521, 'recon_static': 0.0005303841520799324, 'recon_motion': 0.0009607905289158225, 'recon_norm': 1.2720009952783584, 'recon_norm_std': 0.04483415726572275, 'target_norm': 1.6208814918994903, 'recon_cos': 0.6503751441836357, 'gradnorm/compressor.query': 0.12690982080042057, 'gradnorm/compressor.layer0': 3.69170695699803, 'gradnorm/compressor.layer1': 105.03494582161692, 'gradnorm/compressor.layer2': 26.27455255216359, 'gradnorm/compressor.layer3': 11.241688010099118, 'gradnorm/compressor.layer4': 3.6003117903159154, 'gradnorm/compressor.layer5': 2.162093165060339, 'gradnorm/compressor.layer6': 1.2339769035323283, 'gradnorm/compressor.layer7': 1.3819284145292408, 'gradnorm/decoder.layer0': 1.4973881552679056, 'gradnorm/decoder.layer1': 1.0979837492415965, 'gradnorm/decoder.layer2': 0.5592031022957877, 'gradnorm/max_param': 74.08792114257812, 'epoch': 5.52}

 55%|█████▌    | 10540/19090 [5:07:47<4:31:47,  1.91s/it]
 55%|█████▌    | 10541/19090 [5:07:48<4:05:16,  1.72s/it]
 55%|█████▌    | 10542/19090 [5:07:50<3:46:29,  1.59s/it]
 55%|█████▌    | 10543/19090 [5:07:52<4:35:27,  1.93s/it]
 55%|█████▌    | 10544/19090 [5:07:54<4:07:25,  1.74s/it]
 55%|█████▌    | 10545/19090 [5:07:55<3:48:04,  1.60s/it]
 55%|█████▌    | 10546/19090 [5:07:58<4:41:56,  1.98s/it]
 55%|█████▌    | 10547/19090 [5:07:59<4:11:58,  1.77s/it]
 55%|█████▌    | 10548/19090 [5:08:00<3:51:05,  1.62s/it]
 55%|█████▌    | 10549/19090 [5:08:03<4:42:55,  1.99s/it]
 55%|█████▌    | 10550/19090 [5:08:04<4:13:07,  1.78s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 109.1819839477539, 'learning_rate': 2.4773722843499532e-05, 'recon_loss': 0.001458846963942051, 'recon_static': 0.0004920815554214641, 'recon_motion': 0.000966765389603097, 'recon_norm': 1.2608815371990203, 'recon_norm_std': 0.04407818401232362, 'target_norm': 1.6202729284763335, 'recon_cos': 0.6580490678548813, 'gradnorm/compressor.query': 0.12445214286124516, 'gradnorm/compressor.layer0': 3.8080545992583428, 'gradnorm/compressor.layer1': 106.11433894477928, 'gradnorm/compressor.layer2': 23.50192837209388, 'gradnorm/compressor.layer3': 8.442197947025278, 'gradnorm/compressor.layer4': 3.48948042229122, 'gradnorm/compressor.layer5': 2.0067465954869674, 'gradnorm/compressor.layer6': 1.1262797388348347, 'gradnorm/compressor.layer7': 1.2103188195569607, 'gradnorm/decoder.layer0': 1.4807142391340211, 'gradnorm/decoder.layer1': 0.9682410349057796, 'gradnorm/decoder.layer2': 0.5075671556980527, 'gradnorm/max_param': 74.8437728881836, 'epoch': 5.53}

 55%|█████▌    | 10550/19090 [5:08:04<4:13:07,  1.78s/it]
 55%|█████▌    | 10551/19090 [5:08:06<3:52:59,  1.64s/it]
 55%|█████▌    | 10552/19090 [5:08:09<4:45:57,  2.01s/it]
 55%|█████▌    | 10553/19090 [5:08:10<4:15:06,  1.79s/it]
 55%|█████▌    | 10554/19090 [5:08:11<3:53:21,  1.64s/it]
 55%|█████▌    | 10555/19090 [5:08:14<4:33:43,  1.92s/it]
 55%|█████▌    | 10556/19090 [5:08:15<4:06:26,  1.73s/it]
 55%|█████▌    | 10557/19090 [5:08:16<3:47:22,  1.60s/it]
 55%|█████▌    | 10558/19090 [5:08:19<4:30:39,  1.90s/it]
 55%|█████▌    | 10559/19090 [5:08:20<4:04:18,  1.72s/it]
 55%|█████▌    | 10560/19090 [5:08:21<3:45:51,  1.59s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 127.5169906616211, 'learning_rate': 2.472801193871495e-05, 'recon_loss': 0.0014503551123198123, 'recon_static': 0.0004904238703602459, 'recon_motion': 0.0009599311713827773, 'recon_norm': 1.2570812314748765, 'recon_norm_std': 0.04504742417484522, 'target_norm': 1.6207498848438262, 'recon_cos': 0.6599646359682083, 'gradnorm/compressor.query': 0.12579007354731828, 'gradnorm/compressor.layer0': 3.9012181947066176, 'gradnorm/compressor.layer1': 124.55749017772878, 'gradnorm/compressor.layer2': 24.290754856207638, 'gradnorm/compressor.layer3': 10.644962362256038, 'gradnorm/compressor.layer4': 3.9630468045401113, 'gradnorm/compressor.layer5': 2.167526832423281, 'gradnorm/compressor.layer6': 1.0719075256395227, 'gradnorm/compressor.layer7': 1.184384575958593, 'gradnorm/decoder.layer0': 1.7257567722144023, 'gradnorm/decoder.layer1': 1.09277698287866, 'gradnorm/decoder.layer2': 0.5400025560618764, 'gradnorm/max_param': 87.83014678955078, 'epoch': 5.53}

 55%|█████▌    | 10560/19090 [5:08:21<3:45:51,  1.59s/it]
 55%|█████▌    | 10561/19090 [5:08:24<4:40:15,  1.97s/it]
 55%|█████▌    | 10562/19090 [5:08:26<4:11:03,  1.77s/it]
 55%|█████▌    | 10563/19090 [5:08:27<3:50:41,  1.62s/it]
 55%|█████▌    | 10564/19090 [5:08:30<4:53:47,  2.07s/it]
 55%|█████▌    | 10565/19090 [5:08:31<4:20:36,  1.83s/it]
 55%|█████▌    | 10566/19090 [5:08:33<3:57:07,  1.67s/it]
 55%|█████▌    | 10567/19090 [5:08:35<4:26:32,  1.88s/it]
 55%|█████▌    | 10568/19090 [5:08:36<4:01:49,  1.70s/it]
 55%|█████▌    | 10569/19090 [5:08:38<3:44:18,  1.58s/it]
 55%|█████▌    | 10570/19090 [5:08:40<4:33:13,  1.92s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 116.8908920288086, 'learning_rate': 2.4682301943325427e-05, 'recon_loss': 0.0014646502997493372, 'recon_static': 0.0004917650105198846, 'recon_motion': 0.0009728853066917509, 'recon_norm': 1.256752121448517, 'recon_norm_std': 0.04515023287385702, 'target_norm': 1.6225932151079179, 'recon_cos': 0.6569633886218071, 'gradnorm/compressor.query': 0.13732541905869944, 'gradnorm/compressor.layer0': 4.20938019274283, 'gradnorm/compressor.layer1': 113.76098418806643, 'gradnorm/compressor.layer2': 23.79380468948929, 'gradnorm/compressor.layer3': 10.805825703787056, 'gradnorm/compressor.layer4': 3.204523888274965, 'gradnorm/compressor.layer5': 1.9941180878952653, 'gradnorm/compressor.layer6': 1.0252296539386307, 'gradnorm/compressor.layer7': 1.277433526386856, 'gradnorm/decoder.layer0': 1.688768263463245, 'gradnorm/decoder.layer1': 1.1088601650483723, 'gradnorm/decoder.layer2': 0.5112852050859568, 'gradnorm/max_param': 80.26189422607422, 'epoch': 5.54}

 55%|█████▌    | 10570/19090 [5:08:40<4:33:13,  1.92s/it]
 55%|█████▌    | 10571/19090 [5:08:42<4:06:05,  1.73s/it]
 55%|█████▌    | 10572/19090 [5:08:44<4:23:45,  1.86s/it]
 55%|█████▌    | 10573/19090 [5:08:46<4:23:56,  1.86s/it]
 55%|█████▌    | 10574/19090 [5:08:47<3:59:24,  1.69s/it]
 55%|█████▌    | 10575/19090 [5:08:48<3:42:08,  1.57s/it]
 55%|█████▌    | 10576/19090 [5:08:51<4:47:45,  2.03s/it]
 55%|█████▌    | 10577/19090 [5:08:53<4:16:27,  1.81s/it]
 55%|█████▌    | 10578/19090 [5:08:54<3:54:23,  1.65s/it]
 55%|█████▌    | 10579/19090 [5:08:57<4:47:52,  2.03s/it]
 55%|█████▌    | 10580/19090 [5:08:58<4:16:56,  1.81s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 140.2811279296875, 'learning_rate': 2.4636593010162825e-05, 'recon_loss': 0.001464714578469284, 'recon_static': 0.000495603332092287, 'recon_motion': 0.0009691112456494011, 'recon_norm': 1.259987187385559, 'recon_norm_std': 0.04375531990081072, 'target_norm': 1.6217710763216018, 'recon_cos': 0.6567082971334457, 'gradnorm/compressor.query': 0.1348714965483728, 'gradnorm/compressor.layer0': 3.8957906846484645, 'gradnorm/compressor.layer1': 136.42800155897248, 'gradnorm/compressor.layer2': 29.829645308950834, 'gradnorm/compressor.layer3': 11.484973544138894, 'gradnorm/compressor.layer4': 3.829944909797654, 'gradnorm/compressor.layer5': 2.4820679322997057, 'gradnorm/compressor.layer6': 1.1431713162023731, 'gradnorm/compressor.layer7': 1.267334572268875, 'gradnorm/decoder.layer0': 1.9303419725324344, 'gradnorm/decoder.layer1': 1.1890655537589883, 'gradnorm/decoder.layer2': 0.6377278855608253, 'gradnorm/max_param': 96.2394790649414, 'epoch': 5.54}

 55%|█████▌    | 10580/19090 [5:08:58<4:16:56,  1.81s/it]
 55%|█████▌    | 10581/19090 [5:08:59<3:54:56,  1.66s/it]
 55%|█████▌    | 10582/19090 [5:09:02<4:20:16,  1.84s/it]
 55%|█████▌    | 10583/19090 [5:09:03<3:56:56,  1.67s/it]
 55%|█████▌    | 10584/19090 [5:09:04<3:40:31,  1.56s/it]
 55%|█████▌    | 10585/19090 [5:09:07<4:29:14,  1.90s/it]
 55%|█████▌    | 10586/19090 [5:09:08<4:03:10,  1.72s/it]
 55%|█████▌    | 10587/19090 [5:09:09<3:44:57,  1.59s/it]
 55%|█████▌    | 10588/19090 [5:09:13<4:48:44,  2.04s/it]
 55%|█████▌    | 10589/19090 [5:09:14<4:17:04,  1.81s/it]
 55%|█████▌    | 10590/19090 [5:09:15<3:54:43,  1.66s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 114.19467163085938, 'learning_rate': 2.459088529205548e-05, 'recon_loss': 0.0014251951826736331, 'recon_static': 0.000463977951585548, 'recon_motion': 0.0009612172623747028, 'recon_norm': 1.2417874872684478, 'recon_norm_std': 0.04498672941699624, 'target_norm': 1.6206049293279647, 'recon_cos': 0.6658901840448379, 'gradnorm/compressor.query': 0.10398957328356265, 'gradnorm/compressor.layer0': 3.048691271802404, 'gradnorm/compressor.layer1': 111.54945214315264, 'gradnorm/compressor.layer2': 21.680649612660435, 'gradnorm/compressor.layer3': 9.745441251329026, 'gradnorm/compressor.layer4': 3.3107764600517227, 'gradnorm/compressor.layer5': 2.1955498943736353, 'gradnorm/compressor.layer6': 1.1050747010712962, 'gradnorm/compressor.layer7': 1.3041457004913395, 'gradnorm/decoder.layer0': 1.6736768478350306, 'gradnorm/decoder.layer1': 1.0137960545324425, 'gradnorm/decoder.layer2': 0.529050650106214, 'gradnorm/max_param': 78.68109130859375, 'epoch': 5.55}

 55%|█████▌    | 10590/19090 [5:09:15<3:54:43,  1.66s/it]
 55%|█████▌    | 10591/19090 [5:09:18<4:45:32,  2.02s/it]
 55%|█████▌    | 10592/19090 [5:09:19<4:14:17,  1.80s/it]
 55%|█████▌    | 10593/19090 [5:09:21<3:52:26,  1.64s/it]
 55%|█████▌    | 10594/19090 [5:09:23<4:31:29,  1.92s/it]
 56%|█████▌    | 10595/19090 [5:09:24<4:04:32,  1.73s/it]
 56%|█████▌    | 10596/19090 [5:09:26<3:45:44,  1.59s/it]
 56%|█████▌    | 10597/19090 [5:09:28<4:32:12,  1.92s/it]
 56%|█████▌    | 10598/19090 [5:09:30<4:05:10,  1.73s/it]
 56%|█████▌    | 10599/19090 [5:09:31<3:46:00,  1.60s/it]
 56%|█████▌    | 10600/19090 [5:09:33<4:23:17,  1.86s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 118.18921661376953, 'learning_rate': 2.4545178941827633e-05, 'recon_loss': 0.0014151779498206452, 'recon_static': 0.0004500221482885536, 'recon_motion': 0.0009651558313635178, 'recon_norm': 1.2415150493383407, 'recon_norm_std': 0.04557118872180581, 'target_norm': 1.6205401867628098, 'recon_cos': 0.6686318337917327, 'gradnorm/compressor.query': 0.10335059326159161, 'gradnorm/compressor.layer0': 3.0563140987235182, 'gradnorm/compressor.layer1': 115.48015843506407, 'gradnorm/compressor.layer2': 22.644861092237214, 'gradnorm/compressor.layer3': 9.391120986407728, 'gradnorm/compressor.layer4': 3.2464605277990257, 'gradnorm/compressor.layer5': 2.3035289891453523, 'gradnorm/compressor.layer6': 1.13374526984964, 'gradnorm/compressor.layer7': 1.3354115482575417, 'gradnorm/decoder.layer0': 1.5359188447364585, 'gradnorm/decoder.layer1': 1.0482515423156324, 'gradnorm/decoder.layer2': 0.5634956554966332, 'gradnorm/max_param': 81.45640563964844, 'epoch': 5.55}

 56%|█████▌    | 10600/19090 [5:09:33<4:23:17,  1.86s/it]
 56%|█████▌    | 10601/19090 [5:09:35<3:58:52,  1.69s/it]
 56%|█████▌    | 10602/19090 [5:09:36<3:41:42,  1.57s/it]
 56%|█████▌    | 10603/19090 [5:09:39<4:26:00,  1.88s/it]
 56%|█████▌    | 10604/19090 [5:09:40<4:00:36,  1.70s/it]
 56%|█████▌    | 10605/19090 [5:09:42<3:59:38,  1.69s/it]
 56%|█████▌    | 10606/19090 [5:09:44<4:37:03,  1.96s/it]
 56%|█████▌    | 10607/19090 [5:09:45<4:08:24,  1.76s/it]
 56%|█████▌    | 10608/19090 [5:09:47<4:05:17,  1.74s/it]
 56%|█████▌    | 10609/19090 [5:09:50<4:34:09,  1.94s/it]
 56%|█████▌    | 10610/19090 [5:09:51<4:06:34,  1.74s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 119.18902587890625, 'learning_rate': 2.4499474112298964e-05, 'recon_loss': 0.001443212534650229, 'recon_static': 0.0004755285233841278, 'recon_motion': 0.0009676840374595486, 'recon_norm': 1.2524985373020172, 'recon_norm_std': 0.04487498635426164, 'target_norm': 1.6222766518592835, 'recon_cos': 0.6624547421932221, 'gradnorm/compressor.query': 0.10691137823373584, 'gradnorm/compressor.layer0': 3.032311322537168, 'gradnorm/compressor.layer1': 116.16921336056511, 'gradnorm/compressor.layer2': 23.358417361223207, 'gradnorm/compressor.layer3': 11.2965305639919, 'gradnorm/compressor.layer4': 3.842133841308604, 'gradnorm/compressor.layer5': 2.2843173582055685, 'gradnorm/compressor.layer6': 1.3481644708747036, 'gradnorm/compressor.layer7': 1.4599655100262843, 'gradnorm/decoder.layer0': 1.6740637739104394, 'gradnorm/decoder.layer1': 1.1343979362922318, 'gradnorm/decoder.layer2': 0.5351412972916034, 'gradnorm/max_param': 81.94906616210938, 'epoch': 5.56}

 56%|█████▌    | 10610/19090 [5:09:51<4:06:34,  1.74s/it]
 56%|█████▌    | 10611/19090 [5:09:53<4:05:39,  1.74s/it]
 56%|█████▌    | 10612/19090 [5:09:55<4:46:11,  2.03s/it]
 56%|█████▌    | 10613/19090 [5:09:57<4:14:48,  1.80s/it]
 56%|█████▌    | 10614/19090 [5:09:58<3:52:36,  1.65s/it]
 56%|█████▌    | 10615/19090 [5:10:00<4:32:10,  1.93s/it]
 56%|█████▌    | 10616/19090 [5:10:02<4:04:57,  1.73s/it]
 56%|█████▌    | 10617/19090 [5:10:03<3:55:39,  1.67s/it]
 56%|█████▌    | 10618/19090 [5:10:06<4:35:40,  1.95s/it]
 56%|█████▌    | 10619/19090 [5:10:07<4:07:28,  1.75s/it]
 56%|█████▌    | 10620/19090 [5:10:09<4:08:57,  1.76s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 124.33097076416016, 'learning_rate': 2.4453770956284088e-05, 'recon_loss': 0.001462567734415643, 'recon_static': 0.0004997031479433645, 'recon_motion': 0.0009628645770135335, 'recon_norm': 1.2594239473342896, 'recon_norm_std': 0.044998154323548076, 'target_norm': 1.6213975995779037, 'recon_cos': 0.6573407858610153, 'gradnorm/compressor.query': 0.10478343215618997, 'gradnorm/compressor.layer0': 3.2968349363957175, 'gradnorm/compressor.layer1': 121.74460973694072, 'gradnorm/compressor.layer2': 22.815959176945913, 'gradnorm/compressor.layer3': 8.949599818654603, 'gradnorm/compressor.layer4': 3.6462782815353636, 'gradnorm/compressor.layer5': 2.1664226345649125, 'gradnorm/compressor.layer6': 1.1872424800118149, 'gradnorm/compressor.layer7': 1.3540418518450055, 'gradnorm/decoder.layer0': 1.5319055983290133, 'gradnorm/decoder.layer1': 1.014828061165357, 'gradnorm/decoder.layer2': 0.5373937235603048, 'gradnorm/max_param': 85.87569427490234, 'epoch': 5.56}

 56%|█████▌    | 10620/19090 [5:10:09<4:08:57,  1.76s/it]
 56%|█████▌    | 10621/19090 [5:10:11<4:36:21,  1.96s/it]
 56%|█████▌    | 10622/19090 [5:10:13<4:07:57,  1.76s/it]
 56%|█████▌    | 10623/19090 [5:10:14<4:09:21,  1.77s/it]
 56%|█████▌    | 10624/19090 [5:10:17<4:33:20,  1.94s/it]
 56%|█████▌    | 10625/19090 [5:10:18<4:05:44,  1.74s/it]
 56%|█████▌    | 10626/19090 [5:10:20<4:02:30,  1.72s/it]
 56%|█████▌    | 10627/19090 [5:10:22<4:29:26,  1.91s/it]
 56%|█████▌    | 10628/19090 [5:10:23<4:02:56,  1.72s/it]
 56%|█████▌    | 10629/19090 [5:10:25<4:03:43,  1.73s/it]
 56%|█████▌    | 10630/19090 [5:10:28<4:49:08,  2.05s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 117.74188232421875, 'learning_rate': 2.4408069626591992e-05, 'recon_loss': 0.0014385137154022231, 'recon_static': 0.0004760906158480793, 'recon_motion': 0.0009624230980989524, 'recon_norm': 1.2493893295526504, 'recon_norm_std': 0.0448767869733274, 'target_norm': 1.6187426298856735, 'recon_cos': 0.6617642432451248, 'gradnorm/compressor.query': 0.09682512903156887, 'gradnorm/compressor.layer0': 3.1162680127862314, 'gradnorm/compressor.layer1': 114.85351574124755, 'gradnorm/compressor.layer2': 23.444189149612587, 'gradnorm/compressor.layer3': 9.468815265090669, 'gradnorm/compressor.layer4': 3.6273210921227923, 'gradnorm/compressor.layer5': 1.9772545729972655, 'gradnorm/compressor.layer6': 1.0701903670302144, 'gradnorm/compressor.layer7': 1.2641484630953026, 'gradnorm/decoder.layer0': 1.3650466287774101, 'gradnorm/decoder.layer1': 0.9397005639596895, 'gradnorm/decoder.layer2': 0.504091982656243, 'gradnorm/max_param': 81.00796508789062, 'epoch': 5.57}

 56%|█████▌    | 10630/19090 [5:10:28<4:49:08,  2.05s/it]
 56%|█████▌    | 10631/19090 [5:10:29<4:16:57,  1.82s/it]
 56%|█████▌    | 10632/19090 [5:10:30<3:54:15,  1.66s/it]
 56%|█████▌    | 10633/19090 [5:10:33<4:42:13,  2.00s/it]
 56%|█████▌    | 10634/19090 [5:10:34<4:11:54,  1.79s/it]
 56%|█████▌    | 10635/19090 [5:10:36<3:50:43,  1.64s/it]
 56%|█████▌    | 10636/19090 [5:10:39<4:48:07,  2.04s/it]
 56%|█████▌    | 10637/19090 [5:10:40<4:15:52,  1.82s/it]
 56%|█████▌    | 10638/19090 [5:10:41<3:53:14,  1.66s/it]
 56%|█████▌    | 10639/19090 [5:10:44<4:47:38,  2.04s/it]
 56%|█████▌    | 10640/19090 [5:10:46<4:15:47,  1.82s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 142.87860107421875, 'learning_rate': 2.436237027602557e-05, 'recon_loss': 0.0014339766785269603, 'recon_static': 0.0004637631027435418, 'recon_motion': 0.0009702135415864177, 'recon_norm': 1.2473018020391464, 'recon_norm_std': 0.04486590437591076, 'target_norm': 1.6201398491859436, 'recon_cos': 0.6637011915445328, 'gradnorm/compressor.query': 0.12414606670002525, 'gradnorm/compressor.layer0': 3.6907879669876382, 'gradnorm/compressor.layer1': 140.28163636057286, 'gradnorm/compressor.layer2': 24.10915221867258, 'gradnorm/compressor.layer3': 10.520436698059216, 'gradnorm/compressor.layer4': 4.096732647340935, 'gradnorm/compressor.layer5': 2.3120995252102663, 'gradnorm/compressor.layer6': 1.2592140587135796, 'gradnorm/compressor.layer7': 1.453942345894199, 'gradnorm/decoder.layer0': 1.648831578015885, 'gradnorm/decoder.layer1': 0.9803890180654395, 'gradnorm/decoder.layer2': 0.5269155954127553, 'gradnorm/max_param': 98.93282318115234, 'epoch': 5.57}

 56%|█████▌    | 10640/19090 [5:10:46<4:15:47,  1.82s/it]
 56%|█████▌    | 10641/19090 [5:10:47<3:53:19,  1.66s/it]
 56%|█████▌    | 10642/19090 [5:10:50<4:45:40,  2.03s/it]
 56%|█████▌    | 10643/19090 [5:10:51<4:14:07,  1.81s/it]
 56%|█████▌    | 10644/19090 [5:10:52<3:52:10,  1.65s/it]
 56%|█████▌    | 10645/19090 [5:10:55<4:36:55,  1.97s/it]
 56%|█████▌    | 10646/19090 [5:10:56<4:08:02,  1.76s/it]
 56%|█████▌    | 10647/19090 [5:10:58<3:48:13,  1.62s/it]
 56%|█████▌    | 10648/19090 [5:11:00<4:40:22,  1.99s/it]
 56%|█████▌    | 10649/19090 [5:11:02<4:10:20,  1.78s/it]
 56%|█████▌    | 10650/19090 [5:11:03<3:49:29,  1.63s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 119.21968841552734, 'learning_rate': 2.431667305738112e-05, 'recon_loss': 0.0014443920692428947, 'recon_static': 0.00046690751623827963, 'recon_motion': 0.0009774845078936779, 'recon_norm': 1.2470501959323883, 'recon_norm_std': 0.04425961710512638, 'target_norm': 1.6210255295038223, 'recon_cos': 0.6616358950734138, 'gradnorm/compressor.query': 0.11114147920118031, 'gradnorm/compressor.layer0': 3.6122452853996627, 'gradnorm/compressor.layer1': 116.81930491147429, 'gradnorm/compressor.layer2': 21.34518333972891, 'gradnorm/compressor.layer3': 8.595761483342567, 'gradnorm/compressor.layer4': 3.518003888074365, 'gradnorm/compressor.layer5': 2.206610553446738, 'gradnorm/compressor.layer6': 1.2042589664834191, 'gradnorm/compressor.layer7': 1.298836064135158, 'gradnorm/decoder.layer0': 1.6171339827818294, 'gradnorm/decoder.layer1': 0.8528769261677928, 'gradnorm/decoder.layer2': 0.5416222486868484, 'gradnorm/max_param': 82.38618469238281, 'epoch': 5.58}

 56%|█████▌    | 10650/19090 [5:11:03<3:49:29,  1.63s/it]
 56%|█████▌    | 10651/19090 [5:11:06<4:48:36,  2.05s/it]
 56%|█████▌    | 10652/19090 [5:11:07<4:16:08,  1.82s/it]
 56%|█████▌    | 10653/19090 [5:11:09<3:53:22,  1.66s/it]
 56%|█████▌    | 10654/19090 [5:11:11<4:44:35,  2.02s/it]
 56%|█████▌    | 10655/19090 [5:11:13<4:13:22,  1.80s/it]
 56%|█████▌    | 10656/19090 [5:11:14<3:51:28,  1.65s/it]
 56%|█████▌    | 10657/19090 [5:11:17<4:29:25,  1.92s/it]
 56%|█████▌    | 10658/19090 [5:11:18<4:02:43,  1.73s/it]
 56%|█████▌    | 10659/19090 [5:11:19<3:44:04,  1.59s/it]
 56%|█████▌    | 10660/19090 [5:11:22<4:31:33,  1.93s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 103.56961822509766, 'learning_rate': 2.4270978123447766e-05, 'recon_loss': 0.001444934358005412, 'recon_static': 0.0004797048313776031, 'recon_motion': 0.0009652295295381919, 'recon_norm': 1.2498997092247008, 'recon_norm_std': 0.044587038457393646, 'target_norm': 1.620279112458229, 'recon_cos': 0.6608591452240944, 'gradnorm/compressor.query': 0.10889286400256595, 'gradnorm/compressor.layer0': 3.417016050261418, 'gradnorm/compressor.layer1': 101.33749113420463, 'gradnorm/compressor.layer2': 18.89650446139113, 'gradnorm/compressor.layer3': 8.227156948192693, 'gradnorm/compressor.layer4': 3.2172348752639657, 'gradnorm/compressor.layer5': 2.079562041685048, 'gradnorm/compressor.layer6': 1.1350023347771783, 'gradnorm/compressor.layer7': 1.226867734991514, 'gradnorm/decoder.layer0': 1.5331427884226514, 'gradnorm/decoder.layer1': 0.9233035317981308, 'gradnorm/decoder.layer2': 0.5043764803556252, 'gradnorm/max_param': 71.47674560546875, 'epoch': 5.58}

 56%|█████▌    | 10660/19090 [5:11:22<4:31:33,  1.93s/it]
 56%|█████▌    | 10661/19090 [5:11:23<4:04:16,  1.74s/it]
 56%|█████▌    | 10662/19090 [5:11:24<3:45:04,  1.60s/it]
 56%|█████▌    | 10663/19090 [5:11:27<4:34:12,  1.95s/it]
 56%|█████▌    | 10664/19090 [5:11:29<4:05:58,  1.75s/it]
 56%|█████▌    | 10665/19090 [5:11:30<3:46:12,  1.61s/it]
 56%|█████▌    | 10666/19090 [5:11:32<4:15:42,  1.82s/it]
 56%|█████▌    | 10667/19090 [5:11:33<3:52:56,  1.66s/it]
 56%|█████▌    | 10668/19090 [5:11:35<3:37:05,  1.55s/it]
 56%|█████▌    | 10669/19090 [5:11:38<4:38:22,  1.98s/it]
 56%|█████▌    | 10670/19090 [5:11:39<4:09:10,  1.78s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 112.1635971069336, 'learning_rate': 2.4225285627007034e-05, 'recon_loss': 0.0014736348064616322, 'recon_static': 0.000498710697866045, 'recon_motion': 0.0009749240809469484, 'recon_norm': 1.261090052127838, 'recon_norm_std': 0.044171866960823536, 'target_norm': 1.6240723967552184, 'recon_cos': 0.6560388058423996, 'gradnorm/compressor.query': 0.12267265483465963, 'gradnorm/compressor.layer0': 3.4654120803795934, 'gradnorm/compressor.layer1': 109.03716127569766, 'gradnorm/compressor.layer2': 24.049675712064577, 'gradnorm/compressor.layer3': 8.977538584469828, 'gradnorm/compressor.layer4': 3.2648941025202314, 'gradnorm/compressor.layer5': 2.0431952560635596, 'gradnorm/compressor.layer6': 0.9774452187292955, 'gradnorm/compressor.layer7': 1.1528481107088875, 'gradnorm/decoder.layer0': 1.5344480477715152, 'gradnorm/decoder.layer1': 0.9066776713470109, 'gradnorm/decoder.layer2': 0.5153740358827072, 'gradnorm/max_param': 76.90204620361328, 'epoch': 5.59}

 56%|█████▌    | 10670/19090 [5:11:39<4:09:10,  1.78s/it]
 56%|█████▌    | 10671/19090 [5:11:40<3:48:38,  1.63s/it]
 56%|█████▌    | 10672/19090 [5:11:43<4:43:24,  2.02s/it]
 56%|█████▌    | 10673/19090 [5:11:44<4:12:45,  1.80s/it]
 56%|█████▌    | 10674/19090 [5:11:47<4:32:16,  1.94s/it]
 56%|█████▌    | 10675/19090 [5:11:49<4:29:52,  1.92s/it]
 56%|█████▌    | 10676/19090 [5:11:50<4:03:06,  1.73s/it]
 56%|█████▌    | 10677/19090 [5:11:52<4:25:48,  1.90s/it]
 56%|█████▌    | 10678/19090 [5:11:54<4:35:03,  1.96s/it]
 56%|█████▌    | 10679/19090 [5:11:56<4:06:37,  1.76s/it]
 56%|█████▌    | 10680/19090 [5:11:57<3:46:43,  1.62s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 111.20796966552734, 'learning_rate': 2.4179595720832293e-05, 'recon_loss': 0.0014553303422871976, 'recon_static': 0.00048189087319769894, 'recon_motion': 0.0009734394290717319, 'recon_norm': 1.2516414374113083, 'recon_norm_std': 0.04433523174375296, 'target_norm': 1.622663077712059, 'recon_cos': 0.6590461656451225, 'gradnorm/compressor.query': 0.10864949150805807, 'gradnorm/compressor.layer0': 3.326039121781947, 'gradnorm/compressor.layer1': 108.30604423922979, 'gradnorm/compressor.layer2': 22.843799429238587, 'gradnorm/compressor.layer3': 9.174101940971653, 'gradnorm/compressor.layer4': 3.132936546552471, 'gradnorm/compressor.layer5': 2.001280255634022, 'gradnorm/compressor.layer6': 1.0847360015549672, 'gradnorm/compressor.layer7': 1.2112405705355578, 'gradnorm/decoder.layer0': 1.5883644964090449, 'gradnorm/decoder.layer1': 0.857379781954119, 'gradnorm/decoder.layer2': 0.461957401234553, 'gradnorm/max_param': 76.40509033203125, 'epoch': 5.59}

 56%|█████▌    | 10680/19090 [5:11:57<3:46:43,  1.62s/it]
 56%|█████▌    | 10681/19090 [5:12:00<4:31:25,  1.94s/it]
 56%|█████▌    | 10682/19090 [5:12:01<4:03:55,  1.74s/it]
 56%|█████▌    | 10683/19090 [5:12:02<3:44:29,  1.60s/it]
 56%|█████▌    | 10684/19090 [5:12:05<4:17:54,  1.84s/it]
 56%|█████▌    | 10685/19090 [5:12:06<3:54:27,  1.67s/it]
 56%|█████▌    | 10686/19090 [5:12:07<3:38:12,  1.56s/it]
 56%|█████▌    | 10687/19090 [5:12:10<4:15:31,  1.82s/it]
 56%|█████▌    | 10688/19090 [5:12:11<3:52:42,  1.66s/it]
 56%|█████▌    | 10689/19090 [5:12:12<3:36:51,  1.55s/it]
 56%|█████▌    | 10690/19090 [5:12:15<4:28:06,  1.92s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 107.45085906982422, 'learning_rate': 2.413390855768823e-05, 'recon_loss': 0.0014338250155560672, 'recon_static': 0.00048020253889262675, 'recon_motion': 0.000953622511588037, 'recon_norm': 1.2489128947257995, 'recon_norm_std': 0.04559801500290632, 'target_norm': 1.6203656584024428, 'recon_cos': 0.6636265978217125, 'gradnorm/compressor.query': 0.09748509667580804, 'gradnorm/compressor.layer0': 3.0800818405177335, 'gradnorm/compressor.layer1': 104.77187830890024, 'gradnorm/compressor.layer2': 21.605110326872392, 'gradnorm/compressor.layer3': 8.526225751980208, 'gradnorm/compressor.layer4': 2.898766676314839, 'gradnorm/compressor.layer5': 1.9252675824156904, 'gradnorm/compressor.layer6': 1.1016237498901436, 'gradnorm/compressor.layer7': 1.4173879563657608, 'gradnorm/decoder.layer0': 1.7420544081686706, 'gradnorm/decoder.layer1': 0.9651114829798785, 'gradnorm/decoder.layer2': 0.5172400468022623, 'gradnorm/max_param': 73.89167022705078, 'epoch': 5.6}

 56%|█████▌    | 10690/19090 [5:12:15<4:28:06,  1.92s/it]
 56%|█████▌    | 10691/19090 [5:12:16<4:01:41,  1.73s/it]
 56%|█████▌    | 10692/19090 [5:12:17<3:43:21,  1.60s/it]
 56%|█████▌    | 10693/19090 [5:12:20<4:31:11,  1.94s/it]
 56%|█████▌    | 10694/19090 [5:12:21<4:04:14,  1.75s/it]
 56%|█████▌    | 10695/19090 [5:12:23<3:45:03,  1.61s/it]
 56%|█████▌    | 10696/19090 [5:12:26<4:40:40,  2.01s/it]
 56%|█████▌    | 10697/19090 [5:12:27<4:10:20,  1.79s/it]
 56%|█████▌    | 10698/19090 [5:12:28<3:48:46,  1.64s/it]
 56%|█████▌    | 10699/19090 [5:12:31<4:33:29,  1.96s/it]
 56%|█████▌    | 10700/19090 [5:12:32<4:05:11,  1.75s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 125.94438171386719, 'learning_rate': 2.4088224290330375e-05, 'recon_loss': 0.001477893887204118, 'recon_static': 0.0004941784347465728, 'recon_motion': 0.0009837154255365021, 'recon_norm': 1.2592541813850402, 'recon_norm_std': 0.04442321034148335, 'target_norm': 1.6244365334510804, 'recon_cos': 0.6548536494374275, 'gradnorm/compressor.query': 0.13476651647521679, 'gradnorm/compressor.layer0': 3.9015996488942153, 'gradnorm/compressor.layer1': 123.10731947928603, 'gradnorm/compressor.layer2': 24.141834769444436, 'gradnorm/compressor.layer3': 9.135286568759591, 'gradnorm/compressor.layer4': 3.797575834922176, 'gradnorm/compressor.layer5': 1.9219097351408825, 'gradnorm/compressor.layer6': 1.0965748411556568, 'gradnorm/compressor.layer7': 1.3738775006343424, 'gradnorm/decoder.layer0': 1.6505651979832785, 'gradnorm/decoder.layer1': 0.9212722424186194, 'gradnorm/decoder.layer2': 0.5232621084125448, 'gradnorm/max_param': 86.84503936767578, 'epoch': 5.61}

 56%|█████▌    | 10700/19090 [5:12:32<4:05:11,  1.75s/it]
 56%|█████▌    | 10701/19090 [5:12:34<3:45:26,  1.61s/it]
 56%|█████▌    | 10702/19090 [5:12:36<4:37:50,  1.99s/it]
 56%|█████▌    | 10703/19090 [5:12:38<4:08:43,  1.78s/it]
 56%|█████▌    | 10704/19090 [5:12:39<3:47:44,  1.63s/it]
 56%|█████▌    | 10705/19090 [5:12:42<4:27:16,  1.91s/it]
 56%|█████▌    | 10706/19090 [5:12:43<4:00:40,  1.72s/it]
 56%|█████▌    | 10707/19090 [5:12:44<3:42:16,  1.59s/it]
 56%|█████▌    | 10708/19090 [5:12:47<4:37:57,  1.99s/it]
 56%|█████▌    | 10709/19090 [5:12:48<4:08:11,  1.78s/it]
 56%|█████▌    | 10710/19090 [5:12:50<3:47:24,  1.63s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 142.55459594726562, 'learning_rate': 2.4042543071504593e-05, 'recon_loss': 0.0015055294119520114, 'recon_static': 0.0005410032092186157, 'recon_motion': 0.0009645261452533305, 'recon_norm': 1.2757722705602645, 'recon_norm_std': 0.0435274357907474, 'target_norm': 1.6203110814094543, 'recon_cos': 0.6466853350400925, 'gradnorm/compressor.query': 0.1335561283856142, 'gradnorm/compressor.layer0': 4.260514660044844, 'gradnorm/compressor.layer1': 139.27368094196763, 'gradnorm/compressor.layer2': 27.504644937974795, 'gradnorm/compressor.layer3': 10.992122603864047, 'gradnorm/compressor.layer4': 3.761584127994154, 'gradnorm/compressor.layer5': 2.415971612182474, 'gradnorm/compressor.layer6': 1.2879088354515404, 'gradnorm/compressor.layer7': 1.5404224873932306, 'gradnorm/decoder.layer0': 1.8684332614256614, 'gradnorm/decoder.layer1': 1.1481236658311795, 'gradnorm/decoder.layer2': 0.5707935499217812, 'gradnorm/max_param': 98.22457885742188, 'epoch': 5.61}

 56%|█████▌    | 10710/19090 [5:12:50<3:47:24,  1.63s/it]
 56%|█████▌    | 10711/19090 [5:12:52<4:39:20,  2.00s/it]
 56%|█████▌    | 10712/19090 [5:12:54<4:09:07,  1.78s/it]
 56%|█████▌    | 10713/19090 [5:12:55<3:47:57,  1.63s/it]
 56%|█████▌    | 10714/19090 [5:12:58<4:51:43,  2.09s/it]
 56%|█████▌    | 10715/19090 [5:12:59<4:17:54,  1.85s/it]
 56%|█████▌    | 10716/19090 [5:13:01<3:54:11,  1.68s/it]
 56%|█████▌    | 10717/19090 [5:13:04<4:52:34,  2.10s/it]
 56%|█████▌    | 10718/19090 [5:13:05<4:18:29,  1.85s/it]
 56%|█████▌    | 10719/19090 [5:13:06<3:54:49,  1.68s/it]
 56%|█████▌    | 10720/19090 [5:13:09<4:40:23,  2.01s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 139.7239227294922, 'learning_rate': 2.399686505394652e-05, 'recon_loss': 0.0015210551413474605, 'recon_static': 0.000555579004867468, 'recon_motion': 0.0009654761495767161, 'recon_norm': 1.2898870617151261, 'recon_norm_std': 0.044592518080025914, 'target_norm': 1.621134403347969, 'recon_cos': 0.644431258738041, 'gradnorm/compressor.query': 0.17508380067027, 'gradnorm/compressor.layer0': 5.573071496066469, 'gradnorm/compressor.layer1': 136.36810391592675, 'gradnorm/compressor.layer2': 26.86030325802896, 'gradnorm/compressor.layer3': 11.921489744348623, 'gradnorm/compressor.layer4': 4.185074516553612, 'gradnorm/compressor.layer5': 2.3193690375927556, 'gradnorm/compressor.layer6': 1.412850466435136, 'gradnorm/compressor.layer7': 1.585242809867104, 'gradnorm/decoder.layer0': 1.7266950599564836, 'gradnorm/decoder.layer1': 1.0792254484705242, 'gradnorm/decoder.layer2': 0.5290950939505444, 'gradnorm/max_param': 96.18112182617188, 'epoch': 5.62}

 56%|█████▌    | 10720/19090 [5:13:09<4:40:23,  2.01s/it]
 56%|█████▌    | 10721/19090 [5:13:10<4:09:50,  1.79s/it]
 56%|█████▌    | 10722/19090 [5:13:12<3:48:37,  1.64s/it]
 56%|█████▌    | 10723/19090 [5:13:15<4:44:06,  2.04s/it]
 56%|█████▌    | 10724/19090 [5:13:16<4:12:27,  1.81s/it]
 56%|█████▌    | 10725/19090 [5:13:17<3:50:20,  1.65s/it]
 56%|█████▌    | 10726/19090 [5:13:20<4:33:10,  1.96s/it]
 56%|█████▌    | 10727/19090 [5:13:21<4:04:52,  1.76s/it]
 56%|█████▌    | 10728/19090 [5:13:23<3:45:06,  1.62s/it]
 56%|█████▌    | 10729/19090 [5:13:25<4:40:56,  2.02s/it]
 56%|█████▌    | 10730/19090 [5:13:27<4:10:40,  1.80s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 143.6840057373047, 'learning_rate': 2.395119039038111e-05, 'recon_loss': 0.0015363554208306595, 'recon_static': 0.0005713762933737598, 'recon_motion': 0.0009649791900301352, 'recon_norm': 1.2909096032381058, 'recon_norm_std': 0.04358292911201715, 'target_norm': 1.620623105764389, 'recon_cos': 0.6400314301252366, 'gradnorm/compressor.query': 0.14682520950758082, 'gradnorm/compressor.layer0': 4.6810555472432664, 'gradnorm/compressor.layer1': 140.1959426793443, 'gradnorm/compressor.layer2': 28.015809092047142, 'gradnorm/compressor.layer3': 12.107252838239669, 'gradnorm/compressor.layer4': 4.45686178194471, 'gradnorm/compressor.layer5': 2.6350223311576437, 'gradnorm/compressor.layer6': 1.4630410457023795, 'gradnorm/compressor.layer7': 1.5819418045943312, 'gradnorm/decoder.layer0': 1.8613890310692016, 'gradnorm/decoder.layer1': 1.205699197988567, 'gradnorm/decoder.layer2': 0.6420107252452094, 'gradnorm/max_param': 98.88500213623047, 'epoch': 5.62}

 56%|█████▌    | 10730/19090 [5:13:27<4:10:40,  1.80s/it]
 56%|█████▌    | 10731/19090 [5:13:28<3:49:39,  1.65s/it]
 56%|█████▌    | 10732/19090 [5:13:31<4:41:49,  2.02s/it]
 56%|█████▌    | 10733/19090 [5:13:32<4:11:02,  1.80s/it]
 56%|█████▌    | 10734/19090 [5:13:34<3:49:27,  1.65s/it]
 56%|█████▌    | 10735/19090 [5:13:36<4:26:03,  1.91s/it]
 56%|█████▌    | 10736/19090 [5:13:37<3:59:56,  1.72s/it]
 56%|█████▌    | 10737/19090 [5:13:39<3:41:32,  1.59s/it]
 56%|█████▌    | 10738/19090 [5:13:42<4:37:26,  1.99s/it]
 56%|█████▋    | 10739/19090 [5:13:43<4:07:52,  1.78s/it]
 56%|█████▋    | 10740/19090 [5:13:44<3:47:02,  1.63s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 129.53945922851562, 'learning_rate': 2.390551923352211e-05, 'recon_loss': 0.001473789382725954, 'recon_static': 0.000508298823115183, 'recon_motion': 0.0009654905588831753, 'recon_norm': 1.2649265438318253, 'recon_norm_std': 0.04406664986163378, 'target_norm': 1.6204578131437302, 'recon_cos': 0.6541765749454498, 'gradnorm/compressor.query': 0.12272041398016266, 'gradnorm/compressor.layer0': 3.5539122018955025, 'gradnorm/compressor.layer1': 126.2743556333787, 'gradnorm/compressor.layer2': 25.93075142815777, 'gradnorm/compressor.layer3': 10.895997220636986, 'gradnorm/compressor.layer4': 4.101162787442903, 'gradnorm/compressor.layer5': 2.348576065615126, 'gradnorm/compressor.layer6': 1.3306343607217155, 'gradnorm/compressor.layer7': 1.5021589954100913, 'gradnorm/decoder.layer0': 1.9105043932602903, 'gradnorm/decoder.layer1': 1.0630080747360486, 'gradnorm/decoder.layer2': 0.5867892622958589, 'gradnorm/max_param': 89.08445739746094, 'epoch': 5.63}

 56%|█████▋    | 10740/19090 [5:13:44<3:47:02,  1.63s/it]
 56%|█████▋    | 10741/19090 [5:13:47<4:41:55,  2.03s/it]
 56%|█████▋    | 10742/19090 [5:13:48<4:11:12,  1.81s/it]
 56%|█████▋    | 10743/19090 [5:13:50<3:49:58,  1.65s/it]
 56%|█████▋    | 10744/19090 [5:13:52<4:31:42,  1.95s/it]
 56%|█████▋    | 10745/19090 [5:13:54<4:04:07,  1.76s/it]
 56%|█████▋    | 10746/19090 [5:13:55<3:44:58,  1.62s/it]
 56%|█████▋    | 10747/19090 [5:13:58<4:36:49,  1.99s/it]
 56%|█████▋    | 10748/19090 [5:13:59<4:08:07,  1.78s/it]
 56%|█████▋    | 10749/19090 [5:14:00<3:47:37,  1.64s/it]
 56%|█████▋    | 10750/19090 [5:14:03<4:20:22,  1.87s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 132.15542602539062, 'learning_rate': 2.3859851736071537e-05, 'recon_loss': 0.0014475632779067382, 'recon_static': 0.0004803766198165249, 'recon_motion': 0.0009671866806456819, 'recon_norm': 1.2513796895742417, 'recon_norm_std': 0.04485138198360801, 'target_norm': 1.6199562102556229, 'recon_cos': 0.6599736377596855, 'gradnorm/compressor.query': 0.11758525025205742, 'gradnorm/compressor.layer0': 3.9150664665650403, 'gradnorm/compressor.layer1': 129.54797464404837, 'gradnorm/compressor.layer2': 23.175438553529844, 'gradnorm/compressor.layer3': 10.389670421377367, 'gradnorm/compressor.layer4': 3.127830139348241, 'gradnorm/compressor.layer5': 2.264689916812349, 'gradnorm/compressor.layer6': 1.0772324336512662, 'gradnorm/compressor.layer7': 1.3875246097341114, 'gradnorm/decoder.layer0': 1.5548866182799879, 'gradnorm/decoder.layer1': 1.127095125927581, 'gradnorm/decoder.layer2': 0.5545051167409389, 'gradnorm/max_param': 91.38465881347656, 'epoch': 5.63}

 56%|█████▋    | 10750/19090 [5:14:03<4:20:22,  1.87s/it]
 56%|█████▋    | 10751/19090 [5:14:04<3:56:12,  1.70s/it]
 56%|█████▋    | 10752/19090 [5:14:05<3:39:24,  1.58s/it]
 56%|█████▋    | 10753/19090 [5:14:08<4:37:54,  2.00s/it]
 56%|█████▋    | 10754/19090 [5:14:10<4:08:44,  1.79s/it]
 56%|█████▋    | 10755/19090 [5:14:11<3:49:24,  1.65s/it]
 56%|█████▋    | 10756/19090 [5:14:14<4:35:18,  1.98s/it]
 56%|█████▋    | 10757/19090 [5:14:15<4:07:03,  1.78s/it]
 56%|█████▋    | 10758/19090 [5:14:16<3:48:00,  1.64s/it]
 56%|█████▋    | 10759/19090 [5:14:19<4:35:32,  1.98s/it]
 56%|█████▋    | 10760/19090 [5:14:20<4:07:08,  1.78s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 134.33750915527344, 'learning_rate': 2.3814188050719144e-05, 'recon_loss': 0.0014854833134450018, 'recon_static': 0.0005128753473400138, 'recon_motion': 0.0009726079864776693, 'recon_norm': 1.2727505028247834, 'recon_norm_std': 0.04292789846658707, 'target_norm': 1.6213416039943696, 'recon_cos': 0.6525079622864723, 'gradnorm/compressor.query': 0.11720673085412432, 'gradnorm/compressor.layer0': 3.536556476622607, 'gradnorm/compressor.layer1': 131.02687134944878, 'gradnorm/compressor.layer2': 26.68663091383493, 'gradnorm/compressor.layer3': 11.284383153793613, 'gradnorm/compressor.layer4': 3.538967914454175, 'gradnorm/compressor.layer5': 2.3151662765999474, 'gradnorm/compressor.layer6': 1.2498260376995347, 'gradnorm/compressor.layer7': 1.5022247822791137, 'gradnorm/decoder.layer0': 1.7151907393817567, 'gradnorm/decoder.layer1': 1.230400979734038, 'gradnorm/decoder.layer2': 0.5734946947392835, 'gradnorm/max_param': 92.42994689941406, 'epoch': 5.64}

 56%|█████▋    | 10760/19090 [5:14:20<4:07:08,  1.78s/it]
 56%|█████▋    | 10761/19090 [5:14:22<3:46:42,  1.63s/it]
 56%|█████▋    | 10762/19090 [5:14:24<4:31:25,  1.96s/it]
 56%|█████▋    | 10763/19090 [5:14:26<4:03:23,  1.75s/it]
 56%|█████▋    | 10764/19090 [5:14:27<3:43:56,  1.61s/it]
 56%|█████▋    | 10765/19090 [5:14:30<4:41:23,  2.03s/it]
 56%|█████▋    | 10766/19090 [5:14:31<4:10:36,  1.81s/it]
 56%|█████▋    | 10767/19090 [5:14:33<3:48:52,  1.65s/it]
 56%|█████▋    | 10768/19090 [5:14:35<4:30:50,  1.95s/it]
 56%|█████▋    | 10769/19090 [5:14:37<4:03:00,  1.75s/it]
 56%|█████▋    | 10770/19090 [5:14:38<3:43:37,  1.61s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 132.68203735351562, 'learning_rate': 2.3768528330141985e-05, 'recon_loss': 0.001436590967932716, 'recon_static': 0.0004726209292130079, 'recon_motion': 0.0009639700583647937, 'recon_norm': 1.246719202399254, 'recon_norm_std': 0.044477208983153105, 'target_norm': 1.6205769598484039, 'recon_cos': 0.6630110457539559, 'gradnorm/compressor.query': 0.10892521387300952, 'gradnorm/compressor.layer0': 3.2246976548049875, 'gradnorm/compressor.layer1': 129.4933561881651, 'gradnorm/compressor.layer2': 26.293749582169596, 'gradnorm/compressor.layer3': 10.306380612185938, 'gradnorm/compressor.layer4': 3.9168116495339427, 'gradnorm/compressor.layer5': 1.9920263487019265, 'gradnorm/compressor.layer6': 1.3194279613915338, 'gradnorm/compressor.layer7': 1.366576227048944, 'gradnorm/decoder.layer0': 1.8221022030737037, 'gradnorm/decoder.layer1': 1.2035590045444569, 'gradnorm/decoder.layer2': 0.5596170722105076, 'gradnorm/max_param': 91.3365249633789, 'epoch': 5.64}

 56%|█████▋    | 10770/19090 [5:14:38<3:43:37,  1.61s/it]
 56%|█████▋    | 10771/19090 [5:14:41<4:38:42,  2.01s/it]
 56%|█████▋    | 10772/19090 [5:14:42<4:08:56,  1.80s/it]
 56%|█████▋    | 10773/19090 [5:14:43<3:47:45,  1.64s/it]
 56%|█████▋    | 10774/19090 [5:14:46<4:33:16,  1.97s/it]
 56%|█████▋    | 10775/19090 [5:14:47<4:04:49,  1.77s/it]
 56%|█████▋    | 10776/19090 [5:14:49<3:45:44,  1.63s/it]
 56%|█████▋    | 10777/19090 [5:14:51<4:31:32,  1.96s/it]
 56%|█████▋    | 10778/19090 [5:14:53<4:03:38,  1.76s/it]
 56%|█████▋    | 10779/19090 [5:14:54<3:44:18,  1.62s/it]
 56%|█████▋    | 10780/19090 [5:14:57<4:33:48,  1.98s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 116.9403076171875, 'learning_rate': 2.372287272700383e-05, 'recon_loss': 0.001435644883895293, 'recon_static': 0.0004773938017024193, 'recon_motion': 0.000958251090196427, 'recon_norm': 1.2515806585550309, 'recon_norm_std': 0.043725743517279626, 'target_norm': 1.6180303633213042, 'recon_cos': 0.662473452091217, 'gradnorm/compressor.query': 0.11809267693637991, 'gradnorm/compressor.layer0': 3.418487962796158, 'gradnorm/compressor.layer1': 114.02963567643282, 'gradnorm/compressor.layer2': 23.348864953759662, 'gradnorm/compressor.layer3': 9.476453813525543, 'gradnorm/compressor.layer4': 3.8605853819825238, 'gradnorm/compressor.layer5': 1.9092952605483822, 'gradnorm/compressor.layer6': 1.2190506393141174, 'gradnorm/compressor.layer7': 1.3925030916820016, 'gradnorm/decoder.layer0': 1.4830376473224995, 'gradnorm/decoder.layer1': 1.074699932180665, 'gradnorm/decoder.layer2': 0.5212739384137672, 'gradnorm/max_param': 80.41891479492188, 'epoch': 5.65}

 56%|█████▋    | 10780/19090 [5:14:57<4:33:48,  1.98s/it]
 56%|█████▋    | 10781/19090 [5:14:58<4:05:05,  1.77s/it]
 56%|█████▋    | 10782/19090 [5:14:59<3:45:02,  1.63s/it]
 56%|█████▋    | 10783/19090 [5:15:02<4:33:55,  1.98s/it]
 56%|█████▋    | 10784/19090 [5:15:03<4:05:02,  1.77s/it]
 56%|█████▋    | 10785/19090 [5:15:05<3:44:57,  1.63s/it]
 57%|█████▋    | 10786/19090 [5:15:08<4:34:01,  1.98s/it]
 57%|█████▋    | 10787/19090 [5:15:09<4:05:03,  1.77s/it]
 57%|█████▋    | 10788/19090 [5:15:10<3:44:48,  1.62s/it]
 57%|█████▋    | 10789/19090 [5:15:13<4:47:49,  2.08s/it][gradnorm] step 10789: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 103.4

 57%|█████▋    | 10790/19090 [5:15:15<4:15:02,  1.84s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 149.663330078125, 'learning_rate': 2.367722139395468e-05, 'recon_loss': 0.0014317805675091223, 'recon_static': 0.0004735069036541972, 'recon_motion': 0.0009582736922311597, 'recon_norm': 1.2504267066717147, 'recon_norm_std': 0.044713839422911406, 'target_norm': 1.6198803901672363, 'recon_cos': 0.6641129463911056, 'gradnorm/compressor.query': 0.12943330579673565, 'gradnorm/compressor.layer0': 3.747678674022839, 'gradnorm/compressor.layer1': 146.6137903323729, 'gradnorm/compressor.layer2': 27.349881009081876, 'gradnorm/compressor.layer3': 10.62330344623742, 'gradnorm/compressor.layer4': 4.112211807495697, 'gradnorm/compressor.layer5': 2.0610556891782528, 'gradnorm/compressor.layer6': 1.1130215525978107, 'gradnorm/compressor.layer7': 1.397802368509412, 'gradnorm/decoder.layer0': 1.713008376351012, 'gradnorm/decoder.layer1': 1.0007253043076814, 'gradnorm/decoder.layer2': 0.5366827640380016, 'gradnorm/max_param': 103.3953857421875, 'epoch': 5.65}

 57%|█████▋    | 10790/19090 [5:15:15<4:15:02,  1.84s/it]
 57%|█████▋    | 10791/19090 [5:15:16<3:51:53,  1.68s/it]
 57%|█████▋    | 10792/19090 [5:15:19<4:34:53,  1.99s/it]
 57%|█████▋    | 10793/19090 [5:15:20<4:05:40,  1.78s/it]
 57%|█████▋    | 10794/19090 [5:15:21<3:45:13,  1.63s/it]
 57%|█████▋    | 10795/19090 [5:15:24<4:34:02,  1.98s/it]
 57%|█████▋    | 10796/19090 [5:15:25<4:05:18,  1.77s/it]
 57%|█████▋    | 10797/19090 [5:15:27<3:44:57,  1.63s/it]
 57%|█████▋    | 10798/19090 [5:15:29<4:23:01,  1.90s/it]
 57%|█████▋    | 10799/19090 [5:15:30<3:57:34,  1.72s/it]
 57%|█████▋    | 10800/19090 [5:15:32<3:39:45,  1.59s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 137.51695251464844, 'learning_rate': 2.3631574483630275e-05, 'recon_loss': 0.0014625414100009949, 'recon_static': 0.0004956354336172808, 'recon_motion': 0.0009669060047599487, 'recon_norm': 1.2555421859025955, 'recon_norm_std': 0.044142411276698115, 'target_norm': 1.6219387203454971, 'recon_cos': 0.6571017622947692, 'gradnorm/compressor.query': 0.13003654989769842, 'gradnorm/compressor.layer0': 3.758078837324516, 'gradnorm/compressor.layer1': 134.39410474654386, 'gradnorm/compressor.layer2': 26.5477969417603, 'gradnorm/compressor.layer3': 10.018069155018175, 'gradnorm/compressor.layer4': 4.139308273835115, 'gradnorm/compressor.layer5': 2.2604567938786593, 'gradnorm/compressor.layer6': 1.2415971612784782, 'gradnorm/compressor.layer7': 1.3889057804775915, 'gradnorm/decoder.layer0': 1.6684886113304838, 'gradnorm/decoder.layer1': 1.0096429569545633, 'gradnorm/decoder.layer2': 0.5719710232941119, 'gradnorm/max_param': 94.75758361816406, 'epoch': 5.66}

 57%|█████▋    | 10800/19090 [5:15:32<3:39:45,  1.59s/it]
 57%|█████▋    | 10801/19090 [5:15:34<4:29:11,  1.95s/it]
 57%|█████▋    | 10802/19090 [5:15:36<4:01:48,  1.75s/it]
 57%|█████▋    | 10803/19090 [5:15:37<3:42:26,  1.61s/it]
 57%|█████▋    | 10804/19090 [5:15:40<4:21:15,  1.89s/it]
 57%|█████▋    | 10805/19090 [5:15:41<3:56:07,  1.71s/it]
 57%|█████▋    | 10806/19090 [5:15:42<3:38:27,  1.58s/it]
 57%|█████▋    | 10807/19090 [5:15:45<4:24:11,  1.91s/it]
 57%|█████▋    | 10808/19090 [5:15:46<3:58:16,  1.73s/it]
 57%|█████▋    | 10809/19090 [5:15:47<3:39:49,  1.59s/it]
 57%|█████▋    | 10810/19090 [5:15:50<4:25:28,  1.92s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 143.39419555664062, 'learning_rate': 2.3585932148651563e-05, 'recon_loss': 0.0014872771658701821, 'recon_static': 0.0005229136753769126, 'recon_motion': 0.00096436342864763, 'recon_norm': 1.2729074031114578, 'recon_norm_std': 0.043564860988408324, 'target_norm': 1.620682018995285, 'recon_cos': 0.6516486391425133, 'gradnorm/compressor.query': 0.15248779676029972, 'gradnorm/compressor.layer0': 4.846737635502613, 'gradnorm/compressor.layer1': 139.86055415797193, 'gradnorm/compressor.layer2': 28.27594585101498, 'gradnorm/compressor.layer3': 12.293111494245911, 'gradnorm/compressor.layer4': 3.63913889665944, 'gradnorm/compressor.layer5': 2.360108015368738, 'gradnorm/compressor.layer6': 1.2595285591803749, 'gradnorm/compressor.layer7': 1.434970122235646, 'gradnorm/decoder.layer0': 1.7264444302195152, 'gradnorm/decoder.layer1': 1.0098041106513815, 'gradnorm/decoder.layer2': 0.5437822080259235, 'gradnorm/max_param': 98.64556884765625, 'epoch': 5.66}

 57%|█████▋    | 10810/19090 [5:15:50<4:25:28,  1.92s/it]
 57%|█████▋    | 10811/19090 [5:15:51<3:59:11,  1.73s/it]
 57%|█████▋    | 10812/19090 [5:15:53<3:40:40,  1.60s/it]
 57%|█████▋    | 10813/19090 [5:15:56<4:36:09,  2.00s/it]
 57%|█████▋    | 10814/19090 [5:15:57<4:06:24,  1.79s/it]
 57%|█████▋    | 10815/19090 [5:15:58<3:45:37,  1.64s/it]
 57%|█████▋    | 10816/19090 [5:16:01<4:30:02,  1.96s/it]
 57%|█████▋    | 10817/19090 [5:16:02<4:02:07,  1.76s/it]
 57%|█████▋    | 10818/19090 [5:16:03<3:42:37,  1.61s/it]
 57%|█████▋    | 10819/19090 [5:16:06<4:37:14,  2.01s/it]
 57%|█████▋    | 10820/19090 [5:16:08<4:07:16,  1.79s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 134.6886749267578, 'learning_rate': 2.354029454162417e-05, 'recon_loss': 0.0014707034308230505, 'recon_static': 0.0005142152986081783, 'recon_motion': 0.0009564881140249782, 'recon_norm': 1.2639709413051605, 'recon_norm_std': 0.04474653704091906, 'target_norm': 1.619550034403801, 'recon_cos': 0.6542184248566627, 'gradnorm/compressor.query': 0.15692063891166988, 'gradnorm/compressor.layer0': 4.197912224072298, 'gradnorm/compressor.layer1': 131.88588521426394, 'gradnorm/compressor.layer2': 24.034273470806877, 'gradnorm/compressor.layer3': 11.22565182023131, 'gradnorm/compressor.layer4': 3.7028733823897815, 'gradnorm/compressor.layer5': 2.0860658527560707, 'gradnorm/compressor.layer6': 1.272028231001611, 'gradnorm/compressor.layer7': 1.406590568840425, 'gradnorm/decoder.layer0': 1.7020831639987422, 'gradnorm/decoder.layer1': 0.9965177942382155, 'gradnorm/decoder.layer2': 0.5414379138136921, 'gradnorm/max_param': 93.00608825683594, 'epoch': 5.67}

 57%|█████▋    | 10820/19090 [5:16:08<4:07:16,  1.79s/it]
 57%|█████▋    | 10821/19090 [5:16:09<3:46:23,  1.64s/it]
 57%|█████▋    | 10822/19090 [5:16:12<4:26:41,  1.94s/it]
 57%|█████▋    | 10823/19090 [5:16:13<3:59:42,  1.74s/it]
 57%|█████▋    | 10824/19090 [5:16:14<3:41:10,  1.61s/it]
 57%|█████▋    | 10825/19090 [5:16:17<4:43:53,  2.06s/it]
 57%|█████▋    | 10826/19090 [5:16:19<4:11:58,  1.83s/it]
 57%|█████▋    | 10827/19090 [5:16:20<3:49:33,  1.67s/it]
 57%|█████▋    | 10828/19090 [5:16:22<4:20:19,  1.89s/it]
 57%|█████▋    | 10829/19090 [5:16:24<3:55:23,  1.71s/it]
 57%|█████▋    | 10830/19090 [5:16:25<3:38:14,  1.59s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 139.64846801757812, 'learning_rate': 2.3494661815137956e-05, 'recon_loss': 0.0015383507736260072, 'recon_static': 0.0005737936131481547, 'recon_motion': 0.0009645570607972331, 'recon_norm': 1.2938013404607773, 'recon_norm_std': 0.04347534915432334, 'target_norm': 1.6210474908351897, 'recon_cos': 0.6399488359689712, 'gradnorm/compressor.query': 0.1752633731265813, 'gradnorm/compressor.layer0': 4.841880782224749, 'gradnorm/compressor.layer1': 136.6197820366802, 'gradnorm/compressor.layer2': 26.399438609309833, 'gradnorm/compressor.layer3': 9.607292234907506, 'gradnorm/compressor.layer4': 3.397064282320871, 'gradnorm/compressor.layer5': 2.2420424570628588, 'gradnorm/compressor.layer6': 1.2042414946664566, 'gradnorm/compressor.layer7': 1.4158256876924107, 'gradnorm/decoder.layer0': 1.6457905421750267, 'gradnorm/decoder.layer1': 1.0162173377837562, 'gradnorm/decoder.layer2': 0.509264264619999, 'gradnorm/max_param': 96.34745025634766, 'epoch': 5.67}

 57%|█████▋    | 10830/19090 [5:16:25<3:38:14,  1.59s/it]
 57%|█████▋    | 10831/19090 [5:16:27<4:16:15,  1.86s/it]
 57%|█████▋    | 10832/19090 [5:16:29<3:52:37,  1.69s/it]
 57%|█████▋    | 10833/19090 [5:16:30<3:35:51,  1.57s/it]
 57%|█████▋    | 10834/19090 [5:16:32<4:17:51,  1.87s/it]
 57%|█████▋    | 10835/19090 [5:16:34<3:53:41,  1.70s/it]
 57%|█████▋    | 10836/19090 [5:16:35<3:36:51,  1.58s/it]
 57%|█████▋    | 10837/19090 [5:16:38<4:28:58,  1.96s/it]
 57%|█████▋    | 10838/19090 [5:16:39<4:01:35,  1.76s/it]
 57%|█████▋    | 10839/19090 [5:16:40<3:42:22,  1.62s/it]
 57%|█████▋    | 10840/19090 [5:16:43<4:30:44,  1.97s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 139.28643798828125, 'learning_rate': 2.344903412176644e-05, 'recon_loss': 0.0014956366445403545, 'recon_static': 0.0005180879183171782, 'recon_motion': 0.0009775487138540483, 'recon_norm': 1.2729610949754715, 'recon_norm_std': 0.04290047101676464, 'target_norm': 1.621742057800293, 'recon_cos': 0.6497299283742904, 'gradnorm/compressor.query': 0.15411886376752212, 'gradnorm/compressor.layer0': 4.695583037391113, 'gradnorm/compressor.layer1': 136.04293325086755, 'gradnorm/compressor.layer2': 27.166947377008633, 'gradnorm/compressor.layer3': 10.240178683033282, 'gradnorm/compressor.layer4': 3.9440347849799493, 'gradnorm/compressor.layer5': 2.269108691904438, 'gradnorm/compressor.layer6': 1.1702262979090288, 'gradnorm/compressor.layer7': 1.3975873099607472, 'gradnorm/decoder.layer0': 1.602749566032459, 'gradnorm/decoder.layer1': 1.071724899503487, 'gradnorm/decoder.layer2': 0.5558554214261878, 'gradnorm/max_param': 95.94271087646484, 'epoch': 5.68}

 57%|█████▋    | 10840/19090 [5:16:43<4:30:44,  1.97s/it]
 57%|█████▋    | 10841/19090 [5:16:45<4:02:30,  1.76s/it]
 57%|█████▋    | 10842/19090 [5:16:46<3:42:53,  1.62s/it]
 57%|█████▋    | 10843/19090 [5:16:49<4:33:47,  1.99s/it]
 57%|█████▋    | 10844/19090 [5:16:50<4:04:48,  1.78s/it]
 57%|█████▋    | 10845/19090 [5:16:51<3:44:18,  1.63s/it]
 57%|█████▋    | 10846/19090 [5:16:54<4:36:02,  2.01s/it]
 57%|█████▋    | 10847/19090 [5:16:55<4:06:19,  1.79s/it]
 57%|█████▋    | 10848/19090 [5:16:57<3:45:39,  1.64s/it]
 57%|█████▋    | 10849/19090 [5:17:00<4:42:42,  2.06s/it]
 57%|█████▋    | 10850/19090 [5:17:01<4:10:56,  1.83s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 133.8334197998047, 'learning_rate': 2.3403411614066293e-05, 'recon_loss': 0.001459377896389924, 'recon_static': 0.0004937165620503947, 'recon_motion': 0.0009656613343395293, 'recon_norm': 1.256809264421463, 'recon_norm_std': 0.044443551916629076, 'target_norm': 1.6206926912069322, 'recon_cos': 0.6574931830167771, 'gradnorm/compressor.query': 0.12456946062578456, 'gradnorm/compressor.layer0': 3.948825593993402, 'gradnorm/compressor.layer1': 130.61196854854074, 'gradnorm/compressor.layer2': 26.292374187769493, 'gradnorm/compressor.layer3': 10.639173321840751, 'gradnorm/compressor.layer4': 4.323651918444087, 'gradnorm/compressor.layer5': 2.230986687005374, 'gradnorm/compressor.layer6': 1.1790773948782127, 'gradnorm/compressor.layer7': 1.5001079997283526, 'gradnorm/decoder.layer0': 1.720028280535787, 'gradnorm/decoder.layer1': 1.109006686253252, 'gradnorm/decoder.layer2': 0.5573697585060273, 'gradnorm/max_param': 92.13047790527344, 'epoch': 5.68}

 57%|█████▋    | 10850/19090 [5:17:01<4:10:56,  1.83s/it]
 57%|█████▋    | 10851/19090 [5:17:02<3:48:50,  1.67s/it]
 57%|█████▋    | 10852/19090 [5:17:05<4:26:58,  1.94s/it]
 57%|█████▋    | 10853/19090 [5:17:06<3:59:51,  1.75s/it]
 57%|█████▋    | 10854/19090 [5:17:08<3:40:52,  1.61s/it]
 57%|█████▋    | 10855/19090 [5:17:10<4:26:40,  1.94s/it]
 57%|█████▋    | 10856/19090 [5:17:12<3:59:22,  1.74s/it]
 57%|█████▋    | 10857/19090 [5:17:13<3:40:21,  1.61s/it]
 57%|█████▋    | 10858/19090 [5:17:15<4:17:31,  1.88s/it]
 57%|█████▋    | 10859/19090 [5:17:17<3:53:40,  1.70s/it]
 57%|█████▋    | 10860/19090 [5:17:18<3:36:53,  1.58s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 135.30169677734375, 'learning_rate': 2.335779444457689e-05, 'recon_loss': 0.0014538329472998158, 'recon_static': 0.0004976463336788583, 'recon_motion': 0.0009561866114381701, 'recon_norm': 1.2591759026050569, 'recon_norm_std': 0.04506606506183743, 'target_norm': 1.6182681143283844, 'recon_cos': 0.6580694541335106, 'gradnorm/compressor.query': 0.12178827892846443, 'gradnorm/compressor.layer0': 4.200778008109103, 'gradnorm/compressor.layer1': 132.4493859696979, 'gradnorm/compressor.layer2': 24.831029319982765, 'gradnorm/compressor.layer3': 10.049180972000046, 'gradnorm/compressor.layer4': 4.207366051629835, 'gradnorm/compressor.layer5': 1.9669025169268404, 'gradnorm/compressor.layer6': 1.1358961032213895, 'gradnorm/compressor.layer7': 1.2502103628531873, 'gradnorm/decoder.layer0': 1.6632601814136145, 'gradnorm/decoder.layer1': 1.0155712993803345, 'gradnorm/decoder.layer2': 0.5121046169359673, 'gradnorm/max_param': 93.4146957397461, 'epoch': 5.69}

 57%|█████▋    | 10860/19090 [5:17:18<3:36:53,  1.58s/it]
 57%|█████▋    | 10861/19090 [5:17:21<4:36:33,  2.02s/it]
 57%|█████▋    | 10862/19090 [5:17:22<4:06:46,  1.80s/it]
 57%|█████▋    | 10863/19090 [5:17:24<3:46:04,  1.65s/it]
 57%|█████▋    | 10864/19090 [5:17:26<4:17:03,  1.87s/it]
 57%|█████▋    | 10865/19090 [5:17:27<3:53:10,  1.70s/it]
 57%|█████▋    | 10866/19090 [5:17:29<3:36:19,  1.58s/it]
 57%|█████▋    | 10867/19090 [5:17:32<4:39:56,  2.04s/it]
 57%|█████▋    | 10868/19090 [5:17:33<4:08:55,  1.82s/it]
 57%|█████▋    | 10869/19090 [5:17:34<3:47:17,  1.66s/it]
 57%|█████▋    | 10870/19090 [5:17:37<4:35:34,  2.01s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 144.1773223876953, 'learning_rate': 2.3312182765819734e-05, 'recon_loss': 0.0014997197402408346, 'recon_static': 0.0005277913776808418, 'recon_motion': 0.0009719283698359504, 'recon_norm': 1.274465337395668, 'recon_norm_std': 0.04345466690137982, 'target_norm': 1.6210979282855988, 'recon_cos': 0.6484750241041184, 'gradnorm/compressor.query': 0.150488493397518, 'gradnorm/compressor.layer0': 4.441573731175995, 'gradnorm/compressor.layer1': 140.97455167068276, 'gradnorm/compressor.layer2': 27.24353734646875, 'gradnorm/compressor.layer3': 10.995524796559636, 'gradnorm/compressor.layer4': 4.143321645469068, 'gradnorm/compressor.layer5': 2.2231790098804787, 'gradnorm/compressor.layer6': 1.1663187166598916, 'gradnorm/compressor.layer7': 1.423941316016907, 'gradnorm/decoder.layer0': 1.8380860438623987, 'gradnorm/decoder.layer1': 1.1128100552487856, 'gradnorm/decoder.layer2': 0.545501346187871, 'gradnorm/max_param': 99.43592071533203, 'epoch': 5.69}

 57%|█████▋    | 10870/19090 [5:17:37<4:35:34,  2.01s/it]
 57%|█████▋    | 10871/19090 [5:17:38<4:05:43,  1.79s/it]
 57%|█████▋    | 10872/19090 [5:17:40<3:45:10,  1.64s/it]
 57%|█████▋    | 10873/19090 [5:17:42<4:24:24,  1.93s/it]
 57%|█████▋    | 10874/19090 [5:17:44<3:58:20,  1.74s/it]
 57%|█████▋    | 10875/19090 [5:17:45<3:39:53,  1.61s/it]
 57%|█████▋    | 10876/19090 [5:17:47<4:21:54,  1.91s/it]
 57%|█████▋    | 10877/19090 [5:17:49<3:56:32,  1.73s/it]
 57%|█████▋    | 10878/19090 [5:17:50<3:38:48,  1.60s/it]
 57%|█████▋    | 10879/19090 [5:17:53<4:31:21,  1.98s/it]
 57%|█████▋    | 10880/19090 [5:17:54<4:03:10,  1.78s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 132.12269592285156, 'learning_rate': 2.3266576730297956e-05, 'recon_loss': 0.0014689401374198495, 'recon_static': 0.0005050223393482156, 'recon_motion': 0.0009639178533689119, 'recon_norm': 1.2620992183685302, 'recon_norm_std': 0.044632353074848655, 'target_norm': 1.620529016852379, 'recon_cos': 0.6554133430123329, 'gradnorm/compressor.query': 0.2384511388900069, 'gradnorm/compressor.layer0': 6.478085770145478, 'gradnorm/compressor.layer1': 128.69008769860636, 'gradnorm/compressor.layer2': 26.559148455837562, 'gradnorm/compressor.layer3': 10.568285559539389, 'gradnorm/compressor.layer4': 4.512344594376999, 'gradnorm/compressor.layer5': 2.561557968785307, 'gradnorm/compressor.layer6': 1.240775259224697, 'gradnorm/compressor.layer7': 1.5913802087195328, 'gradnorm/decoder.layer0': 1.7500352855939718, 'gradnorm/decoder.layer1': 1.319581681637844, 'gradnorm/decoder.layer2': 0.6060222925643334, 'gradnorm/max_param': 90.78360748291016, 'epoch': 5.7}

 57%|█████▋    | 10880/19090 [5:17:54<4:03:10,  1.78s/it]
 57%|█████▋    | 10881/19090 [5:17:56<3:43:44,  1.64s/it]
 57%|█████▋    | 10882/19090 [5:17:58<4:27:02,  1.95s/it]
 57%|█████▋    | 10883/19090 [5:18:00<3:59:48,  1.75s/it]
 57%|█████▋    | 10884/19090 [5:18:01<3:40:25,  1.61s/it]
 57%|█████▋    | 10885/19090 [5:18:04<4:29:48,  1.97s/it]
 57%|█████▋    | 10886/19090 [5:18:05<4:01:41,  1.77s/it]
 57%|█████▋    | 10887/19090 [5:18:06<3:41:41,  1.62s/it]
 57%|█████▋    | 10888/19090 [5:18:09<4:22:35,  1.92s/it]
 57%|█████▋    | 10889/19090 [5:18:10<3:56:27,  1.73s/it]
 57%|█████▋    | 10890/19090 [5:18:11<3:38:16,  1.60s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 124.06419372558594, 'learning_rate': 2.3220976490495856e-05, 'recon_loss': 0.0014763457118533551, 'recon_static': 0.0005143615504493937, 'recon_motion': 0.0009619841686799191, 'recon_norm': 1.2698303282260894, 'recon_norm_std': 0.044306311197578906, 'target_norm': 1.6179295986890794, 'recon_cos': 0.6529931128025055, 'gradnorm/compressor.query': 0.12866769738702744, 'gradnorm/compressor.layer0': 3.9003989382422675, 'gradnorm/compressor.layer1': 120.72951269004194, 'gradnorm/compressor.layer2': 25.604775984621625, 'gradnorm/compressor.layer3': 10.818627534081733, 'gradnorm/compressor.layer4': 3.7995589703063497, 'gradnorm/compressor.layer5': 2.3242919910335105, 'gradnorm/compressor.layer6': 1.3415387167645167, 'gradnorm/compressor.layer7': 1.5126438513831733, 'gradnorm/decoder.layer0': 1.6707074375556261, 'gradnorm/decoder.layer1': 1.167859126531367, 'gradnorm/decoder.layer2': 0.5930244379443838, 'gradnorm/max_param': 85.1758041381836, 'epoch': 5.7}

 57%|█████▋    | 10890/19090 [5:18:11<3:38:16,  1.60s/it]
 57%|█████▋    | 10891/19090 [5:18:14<4:26:09,  1.95s/it]
 57%|█████▋    | 10892/19090 [5:18:15<3:58:53,  1.75s/it]
 57%|█████▋    | 10893/19090 [5:18:17<3:39:54,  1.61s/it]
 57%|█████▋    | 10894/19090 [5:18:20<4:44:13,  2.08s/it]
 57%|█████▋    | 10895/19090 [5:18:21<4:11:37,  1.84s/it]
 57%|█████▋    | 10896/19090 [5:18:22<3:48:43,  1.67s/it]
 57%|█████▋    | 10897/19090 [5:18:25<4:29:23,  1.97s/it]
 57%|█████▋    | 10898/19090 [5:18:26<4:01:13,  1.77s/it]
 57%|█████▋    | 10899/19090 [5:18:28<3:41:29,  1.62s/it]
 57%|█████▋    | 10900/19090 [5:18:31<4:29:26,  1.97s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 133.68711853027344, 'learning_rate': 2.3175382198878328e-05, 'recon_loss': 0.0014483967446722091, 'recon_static': 0.0004936556892062071, 'recon_motion': 0.0009547410663799383, 'recon_norm': 1.2560271143913269, 'recon_norm_std': 0.04435970615595579, 'target_norm': 1.6176798731088637, 'recon_cos': 0.6590565145015717, 'gradnorm/compressor.query': 0.14830783149502655, 'gradnorm/compressor.layer0': 4.527602628558936, 'gradnorm/compressor.layer1': 130.7752126742679, 'gradnorm/compressor.layer2': 25.361704683501166, 'gradnorm/compressor.layer3': 9.115261368009635, 'gradnorm/compressor.layer4': 3.221491562738283, 'gradnorm/compressor.layer5': 2.289152801086553, 'gradnorm/compressor.layer6': 1.2821655723784957, 'gradnorm/compressor.layer7': 1.4426805606009645, 'gradnorm/decoder.layer0': 1.5658591972814988, 'gradnorm/decoder.layer1': 1.1083258382405183, 'gradnorm/decoder.layer2': 0.4893040739867103, 'gradnorm/max_param': 92.23217010498047, 'epoch': 5.71}

 57%|█████▋    | 10900/19090 [5:18:31<4:29:26,  1.97s/it]
 57%|█████▋    | 10901/19090 [5:18:32<4:01:27,  1.77s/it]
 57%|█████▋    | 10902/19090 [5:18:33<3:41:49,  1.63s/it]
 57%|█████▋    | 10903/19090 [5:18:36<4:35:24,  2.02s/it]
 57%|█████▋    | 10904/19090 [5:18:37<4:05:36,  1.80s/it]
 57%|█████▋    | 10905/19090 [5:18:39<3:44:50,  1.65s/it]
 57%|█████▋    | 10906/19090 [5:18:41<4:16:37,  1.88s/it]
 57%|█████▋    | 10907/19090 [5:18:42<3:52:17,  1.70s/it]
 57%|█████▋    | 10908/19090 [5:18:44<3:35:50,  1.58s/it]
 57%|█████▋    | 10909/19090 [5:18:46<4:26:30,  1.95s/it][gradnorm] step 10909: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 101.8

 57%|█████▋    | 10910/19090 [5:18:48<3:59:25,  1.76s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 146.65234375, 'learning_rate': 2.3129794007890366e-05, 'recon_loss': 0.0014590992796001956, 'recon_static': 0.000496450402715709, 'recon_motion': 0.0009626489161746576, 'recon_norm': 1.2581275165081025, 'recon_norm_std': 0.04386498145759106, 'target_norm': 1.6212890237569808, 'recon_cos': 0.6578722670674324, 'gradnorm/compressor.query': 0.14328707839152036, 'gradnorm/compressor.layer0': 4.269205121064418, 'gradnorm/compressor.layer1': 144.29460017833307, 'gradnorm/compressor.layer2': 23.38007287450549, 'gradnorm/compressor.layer3': 9.840786926933669, 'gradnorm/compressor.layer4': 3.5296411140572026, 'gradnorm/compressor.layer5': 2.2627297679926794, 'gradnorm/compressor.layer6': 1.1474771441674638, 'gradnorm/compressor.layer7': 1.3839414912381047, 'gradnorm/decoder.layer0': 1.4616084062774068, 'gradnorm/decoder.layer1': 1.0205281378807245, 'gradnorm/decoder.layer2': 0.5224477891069673, 'gradnorm/max_param': 101.78031158447266, 'epoch': 5.72}

 57%|█████▋    | 10910/19090 [5:18:48<3:59:25,  1.76s/it]
 57%|█████▋    | 10911/19090 [5:18:49<3:40:32,  1.62s/it]
 57%|█████▋    | 10912/19090 [5:18:52<4:22:30,  1.93s/it]
 57%|█████▋    | 10913/19090 [5:18:53<3:56:33,  1.74s/it]
 57%|█████▋    | 10914/19090 [5:18:54<3:38:18,  1.60s/it]
 57%|█████▋    | 10915/19090 [5:18:57<4:20:18,  1.91s/it]
 57%|█████▋    | 10916/19090 [5:18:58<3:54:56,  1.72s/it]
 57%|█████▋    | 10917/19090 [5:18:59<3:37:11,  1.59s/it]
 57%|█████▋    | 10918/19090 [5:19:02<4:12:41,  1.86s/it]
 57%|█████▋    | 10919/19090 [5:19:03<3:49:12,  1.68s/it]
 57%|█████▋    | 10920/19090 [5:19:05<3:33:08,  1.57s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 140.2261505126953, 'learning_rate': 2.308421206995661e-05, 'recon_loss': 0.0014607845660066232, 'recon_static': 0.0005034275556681678, 'recon_motion': 0.0009573570292559453, 'recon_norm': 1.25849429666996, 'recon_norm_std': 0.04432952888309956, 'target_norm': 1.6191580712795257, 'recon_cos': 0.6564161032438278, 'gradnorm/compressor.query': 0.12389268139991871, 'gradnorm/compressor.layer0': 3.662567679655441, 'gradnorm/compressor.layer1': 137.26782244257026, 'gradnorm/compressor.layer2': 26.13830986647341, 'gradnorm/compressor.layer3': 9.939045719875592, 'gradnorm/compressor.layer4': 3.5826266427293043, 'gradnorm/compressor.layer5': 2.2110287273443756, 'gradnorm/compressor.layer6': 1.295963472575661, 'gradnorm/compressor.layer7': 1.3500008794993608, 'gradnorm/decoder.layer0': 1.6708345116716576, 'gradnorm/decoder.layer1': 1.0877566571848407, 'gradnorm/decoder.layer2': 0.5379066514762932, 'gradnorm/max_param': 96.81510162353516, 'epoch': 5.72}

 57%|█████▋    | 10920/19090 [5:19:05<3:33:08,  1.57s/it]
 57%|█████▋    | 10921/19090 [5:19:07<4:14:14,  1.87s/it]
 57%|█████▋    | 10922/19090 [5:19:08<3:55:03,  1.73s/it]
 57%|█████▋    | 10923/19090 [5:19:10<3:48:38,  1.68s/it]
 57%|█████▋    | 10924/19090 [5:19:13<4:27:18,  1.96s/it]
 57%|█████▋    | 10925/19090 [5:19:14<3:59:21,  1.76s/it]
 57%|█████▋    | 10926/19090 [5:19:15<3:42:32,  1.64s/it]
 57%|█████▋    | 10927/19090 [5:19:18<4:33:41,  2.01s/it]
 57%|█████▋    | 10928/19090 [5:19:19<4:04:33,  1.80s/it]
 57%|█████▋    | 10929/19090 [5:19:21<3:43:33,  1.64s/it][gradnorm] step 10929: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 121.0

 57%|█████▋    | 10930/19090 [5:19:24<4:43:25,  2.08s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 174.43946838378906, 'learning_rate': 2.3038636537480764e-05, 'recon_loss': 0.0014758551493287086, 'recon_static': 0.0005069378144980874, 'recon_motion': 0.0009689173137303441, 'recon_norm': 1.2651998102664948, 'recon_norm_std': 0.04351715315133333, 'target_norm': 1.6220011770725251, 'recon_cos': 0.6544433385133743, 'gradnorm/compressor.query': 0.1387293924673641, 'gradnorm/compressor.layer0': 4.392355146977828, 'gradnorm/compressor.layer1': 171.58338329184153, 'gradnorm/compressor.layer2': 28.43917298478621, 'gradnorm/compressor.layer3': 11.236668125885407, 'gradnorm/compressor.layer4': 4.319723078717737, 'gradnorm/compressor.layer5': 2.436058498507003, 'gradnorm/compressor.layer6': 1.3996123594677572, 'gradnorm/compressor.layer7': 1.4325014365357125, 'gradnorm/decoder.layer0': 1.9125489770938455, 'gradnorm/decoder.layer1': 1.120456900129002, 'gradnorm/decoder.layer2': 0.6262544440684754, 'gradnorm/max_param': 121.03781127929688, 'epoch': 5.73}

 57%|█████▋    | 10930/19090 [5:19:24<4:43:25,  2.08s/it]
 57%|█████▋    | 10931/19090 [5:19:25<4:15:23,  1.88s/it]
 57%|█████▋    | 10932/19090 [5:19:27<3:51:01,  1.70s/it]
 57%|█████▋    | 10933/19090 [5:19:29<4:36:04,  2.03s/it]
 57%|█████▋    | 10934/19090 [5:19:31<4:05:51,  1.81s/it]
 57%|█████▋    | 10935/19090 [5:19:32<3:44:49,  1.65s/it]
 57%|█████▋    | 10936/19090 [5:19:34<4:11:19,  1.85s/it]
 57%|█████▋    | 10937/19090 [5:19:36<4:00:40,  1.77s/it]
 57%|█████▋    | 10938/19090 [5:19:37<3:41:00,  1.63s/it]
 57%|█████▋    | 10939/19090 [5:19:40<4:12:14,  1.86s/it][gradnorm] step 10939: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 107.4

 57%|█████▋    | 10940/19090 [5:19:41<4:14:04,  1.87s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 154.863525390625, 'learning_rate': 2.299306756284511e-05, 'recon_loss': 0.0014570731262210756, 'recon_static': 0.0004959887919540051, 'recon_motion': 0.0009610843218979426, 'recon_norm': 1.2557187110185624, 'recon_norm_std': 0.04516741679981351, 'target_norm': 1.6204318672418594, 'recon_cos': 0.6578723132610321, 'gradnorm/compressor.query': 0.1675899914894976, 'gradnorm/compressor.layer0': 4.818002304868356, 'gradnorm/compressor.layer1': 152.29890044949437, 'gradnorm/compressor.layer2': 25.31822610773092, 'gradnorm/compressor.layer3': 9.928586020855898, 'gradnorm/compressor.layer4': 3.6407487709060624, 'gradnorm/compressor.layer5': 2.080220707407827, 'gradnorm/compressor.layer6': 1.2928422479893034, 'gradnorm/compressor.layer7': 1.3696122384374614, 'gradnorm/decoder.layer0': 1.6010445850604662, 'gradnorm/decoder.layer1': 0.9694258270613508, 'gradnorm/decoder.layer2': 0.5402722536434961, 'gradnorm/max_param': 107.43438720703125, 'epoch': 5.73}

 57%|█████▋    | 10940/19090 [5:19:41<4:14:04,  1.87s/it]
 57%|█████▋    | 10941/19090 [5:19:43<3:50:34,  1.70s/it]
 57%|█████▋    | 10942/19090 [5:19:45<4:10:13,  1.84s/it]
 57%|█████▋    | 10943/19090 [5:19:47<4:03:39,  1.79s/it]
 57%|█████▋    | 10944/19090 [5:19:48<3:42:57,  1.64s/it]
 57%|█████▋    | 10945/19090 [5:19:50<4:21:10,  1.92s/it]
 57%|█████▋    | 10946/19090 [5:19:52<4:09:25,  1.84s/it]
 57%|█████▋    | 10947/19090 [5:19:53<3:46:47,  1.67s/it]
 57%|█████▋    | 10948/19090 [5:19:56<4:08:16,  1.83s/it]
 57%|█████▋    | 10949/19090 [5:19:58<4:32:13,  2.01s/it]
 57%|█████▋    | 10950/19090 [5:19:59<4:02:47,  1.79s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 133.1605682373047, 'learning_rate': 2.2947505298410016e-05, 'recon_loss': 0.0014642135647591203, 'recon_static': 0.000510534144268604, 'recon_motion': 0.0009536794546875172, 'recon_norm': 1.2667513370513916, 'recon_norm_std': 0.04655009442940354, 'target_norm': 1.620010071992874, 'recon_cos': 0.6566769436001778, 'gradnorm/compressor.query': 0.13098806487151934, 'gradnorm/compressor.layer0': 3.821051840044393, 'gradnorm/compressor.layer1': 130.03738945491793, 'gradnorm/compressor.layer2': 25.50794290740019, 'gradnorm/compressor.layer3': 11.412126589763243, 'gradnorm/compressor.layer4': 3.7164421293119716, 'gradnorm/compressor.layer5': 2.1162341459416107, 'gradnorm/compressor.layer6': 1.2705637327960686, 'gradnorm/compressor.layer7': 1.4661957465213538, 'gradnorm/decoder.layer0': 1.7281130909293825, 'gradnorm/decoder.layer1': 1.0774820594444636, 'gradnorm/decoder.layer2': 0.5566872025126366, 'gradnorm/max_param': 91.73631286621094, 'epoch': 5.74}

 57%|█████▋    | 10950/19090 [5:19:59<4:02:47,  1.79s/it]
 57%|█████▋    | 10951/19090 [5:20:01<3:57:07,  1.75s/it]
 57%|█████▋    | 10952/19090 [5:20:03<4:22:00,  1.93s/it]
 57%|█████▋    | 10953/19090 [5:20:05<3:55:36,  1.74s/it]
 57%|█████▋    | 10954/19090 [5:20:07<4:07:55,  1.83s/it]
 57%|█████▋    | 10955/19090 [5:20:09<4:13:09,  1.87s/it]
 57%|█████▋    | 10956/19090 [5:20:10<3:49:28,  1.69s/it]
 57%|█████▋    | 10957/19090 [5:20:12<4:20:04,  1.92s/it]
 57%|█████▋    | 10958/19090 [5:20:14<4:10:58,  1.85s/it]
 57%|█████▋    | 10959/19090 [5:20:15<3:48:25,  1.69s/it]
 57%|█████▋    | 10960/19090 [5:20:17<4:06:28,  1.82s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 133.34910583496094, 'learning_rate': 2.290194989651342e-05, 'recon_loss': 0.0014442125306231902, 'recon_static': 0.0004794821878022049, 'recon_motion': 0.0009647304483223707, 'recon_norm': 1.2513748764991761, 'recon_norm_std': 0.04447174966335297, 'target_norm': 1.6219598680734635, 'recon_cos': 0.6619217485189438, 'gradnorm/compressor.query': 0.1053355488471857, 'gradnorm/compressor.layer0': 3.3845843108624765, 'gradnorm/compressor.layer1': 130.00895401855982, 'gradnorm/compressor.layer2': 26.730848640530827, 'gradnorm/compressor.layer3': 11.161138921888366, 'gradnorm/compressor.layer4': 4.051203824315368, 'gradnorm/compressor.layer5': 2.30464125764652, 'gradnorm/compressor.layer6': 1.2298504926176574, 'gradnorm/compressor.layer7': 1.3735053435181062, 'gradnorm/decoder.layer0': 1.6143466488855311, 'gradnorm/decoder.layer1': 1.0272032783115814, 'gradnorm/decoder.layer2': 0.5444155018627241, 'gradnorm/max_param': 91.73381042480469, 'epoch': 5.74}

 57%|█████▋    | 10960/19090 [5:20:17<4:06:28,  1.82s/it]
 57%|█████▋    | 10961/19090 [5:20:20<4:26:37,  1.97s/it]
 57%|█████▋    | 10962/19090 [5:20:21<3:59:10,  1.77s/it]
 57%|█████▋    | 10963/19090 [5:20:23<4:02:04,  1.79s/it]
 57%|█████▋    | 10964/19090 [5:20:25<4:13:02,  1.87s/it]
 57%|█████▋    | 10965/19090 [5:20:26<3:49:29,  1.69s/it]
 57%|█████▋    | 10966/19090 [5:20:28<4:06:47,  1.82s/it]
 57%|█████▋    | 10967/19090 [5:20:30<4:18:40,  1.91s/it]
 57%|█████▋    | 10968/19090 [5:20:32<3:53:24,  1.72s/it]
 57%|█████▋    | 10969/19090 [5:20:34<4:01:05,  1.78s/it][gradnorm] step 10969: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 106.7

 57%|█████▋    | 10970/19090 [5:20:35<4:01:49,  1.79s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 155.3056182861328, 'learning_rate': 2.2856401509470297e-05, 'recon_loss': 0.0014663472451502458, 'recon_static': 0.0005017275143472943, 'recon_motion': 0.0009646197402616962, 'recon_norm': 1.2564963340759276, 'recon_norm_std': 0.044416739512234925, 'target_norm': 1.620804014801979, 'recon_cos': 0.655395145714283, 'gradnorm/compressor.query': 0.1716113072595403, 'gradnorm/compressor.layer0': 5.014780324002044, 'gradnorm/compressor.layer1': 151.2119722339306, 'gradnorm/compressor.layer2': 32.38120001262789, 'gradnorm/compressor.layer3': 12.362702395611606, 'gradnorm/compressor.layer4': 3.745628988415679, 'gradnorm/compressor.layer5': 2.272244694367295, 'gradnorm/compressor.layer6': 1.3034987373389832, 'gradnorm/compressor.layer7': 1.4489392116371809, 'gradnorm/decoder.layer0': 1.8973673249151093, 'gradnorm/decoder.layer1': 1.1078154656675858, 'gradnorm/decoder.layer2': 0.6313994372405528, 'gradnorm/max_param': 106.67843627929688, 'epoch': 5.75}

 57%|█████▋    | 10970/19090 [5:20:35<4:01:49,  1.79s/it]
 57%|█████▋    | 10971/19090 [5:20:37<3:41:48,  1.64s/it]
 57%|█████▋    | 10972/19090 [5:20:39<4:24:58,  1.96s/it]
 57%|█████▋    | 10973/19090 [5:20:41<3:58:11,  1.76s/it]
 57%|█████▋    | 10974/19090 [5:20:42<3:39:00,  1.62s/it]
 57%|█████▋    | 10975/19090 [5:20:44<4:10:44,  1.85s/it]
 57%|█████▋    | 10976/19090 [5:20:46<4:02:36,  1.79s/it]
 58%|█████▊    | 10977/19090 [5:20:47<3:42:14,  1.64s/it]
 58%|█████▊    | 10978/19090 [5:20:50<4:09:51,  1.85s/it]
 58%|█████▊    | 10979/19090 [5:20:51<3:52:53,  1.72s/it][gradnorm] step 10979: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 107.0

 58%|█████▊    | 10980/19090 [5:20:52<3:35:57,  1.60s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 155.00936889648438, 'learning_rate': 2.281086028957217e-05, 'recon_loss': 0.001505197276128456, 'recon_static': 0.000537282413279172, 'recon_motion': 0.0009679149035946466, 'recon_norm': 1.278060832619667, 'recon_norm_std': 0.044023568369448186, 'target_norm': 1.6215319991111756, 'recon_cos': 0.6474899619817733, 'gradnorm/compressor.query': 0.13472779060792817, 'gradnorm/compressor.layer0': 4.131397949893685, 'gradnorm/compressor.layer1': 151.66358742295395, 'gradnorm/compressor.layer2': 29.495921742065672, 'gradnorm/compressor.layer3': 10.667749548622188, 'gradnorm/compressor.layer4': 3.5248749839311677, 'gradnorm/compressor.layer5': 2.134811749880532, 'gradnorm/compressor.layer6': 1.3454458934830282, 'gradnorm/compressor.layer7': 1.4133766998229513, 'gradnorm/decoder.layer0': 1.7854081858815962, 'gradnorm/decoder.layer1': 0.9554568049010899, 'gradnorm/decoder.layer2': 0.52479275291283, 'gradnorm/max_param': 106.99654388427734, 'epoch': 5.75}

 58%|█████▊    | 10980/19090 [5:20:52<3:35:57,  1.60s/it]
 58%|█████▊    | 10981/19090 [5:20:55<4:15:09,  1.89s/it]
 58%|█████▊    | 10982/19090 [5:20:57<3:58:43,  1.77s/it]
 58%|█████▊    | 10983/19090 [5:20:58<3:39:33,  1.62s/it]
 58%|█████▊    | 10984/19090 [5:21:00<4:18:09,  1.91s/it]
 58%|█████▊    | 10985/19090 [5:21:02<3:59:50,  1.78s/it]
 58%|█████▊    | 10986/19090 [5:21:03<3:40:11,  1.63s/it]
 58%|█████▊    | 10987/19090 [5:21:06<4:14:19,  1.88s/it]
 58%|█████▊    | 10988/19090 [5:21:07<3:57:02,  1.76s/it]
 58%|█████▊    | 10989/19090 [5:21:08<3:38:18,  1.62s/it]
 58%|█████▊    | 10990/19090 [5:21:11<4:19:12,  1.92s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 134.77191162109375, 'learning_rate': 2.2765326389086636e-05, 'recon_loss': 0.001470267612603493, 'recon_static': 0.0004957180637575221, 'recon_motion': 0.0009745495888637378, 'recon_norm': 1.2604508757591248, 'recon_norm_std': 0.04381692027673125, 'target_norm': 1.622200635075569, 'recon_cos': 0.6557058200240136, 'gradnorm/compressor.query': 0.12160384507186023, 'gradnorm/compressor.layer0': 3.6747882081330605, 'gradnorm/compressor.layer1': 131.80678034058036, 'gradnorm/compressor.layer2': 25.37680994353223, 'gradnorm/compressor.layer3': 10.202961615149537, 'gradnorm/compressor.layer4': 3.7748243733155906, 'gradnorm/compressor.layer5': 2.4743805427999868, 'gradnorm/compressor.layer6': 1.199538297047399, 'gradnorm/compressor.layer7': 1.4316666042578035, 'gradnorm/decoder.layer0': 1.9168197391033375, 'gradnorm/decoder.layer1': 1.0105083987619825, 'gradnorm/decoder.layer2': 0.5315736457961387, 'gradnorm/max_param': 92.96752166748047, 'epoch': 5.76}

 58%|█████▊    | 10990/19090 [5:21:11<4:19:12,  1.92s/it]
 58%|█████▊    | 10991/19090 [5:21:12<3:53:41,  1.73s/it]
 58%|█████▊    | 10992/19090 [5:21:14<3:35:52,  1.60s/it]
 58%|█████▊    | 10993/19090 [5:21:16<4:05:02,  1.82s/it]
 58%|█████▊    | 10994/19090 [5:21:17<3:56:37,  1.75s/it]
 58%|█████▊    | 10995/19090 [5:21:19<3:37:59,  1.62s/it]
 58%|█████▊    | 10996/19090 [5:21:21<4:13:08,  1.88s/it]
 58%|█████▊    | 10997/19090 [5:21:23<4:06:02,  1.82s/it]
 58%|█████▊    | 10998/19090 [5:21:24<3:44:20,  1.66s/it]
 58%|█████▊    | 10999/19090 [5:21:27<4:14:29,  1.89s/it]
 58%|█████▊    | 11000/19090 [5:21:28<3:56:11,  1.75s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 131.87193298339844, 'learning_rate': 2.271979996025676e-05, 'recon_loss': 0.001498390466440469, 'recon_static': 0.0005390294376411475, 'recon_motion': 0.0009593610011506826, 'recon_norm': 1.2787573605775833, 'recon_norm_std': 0.04371660491451621, 'target_norm': 1.6197875201702119, 'recon_cos': 0.648367366194725, 'gradnorm/compressor.query': 0.12578994027912985, 'gradnorm/compressor.layer0': 3.9437056791351175, 'gradnorm/compressor.layer1': 129.4679693133595, 'gradnorm/compressor.layer2': 22.743273065382706, 'gradnorm/compressor.layer3': 8.484200997438792, 'gradnorm/compressor.layer4': 3.2619102558862734, 'gradnorm/compressor.layer5': 2.431291841671555, 'gradnorm/compressor.layer6': 1.1914480733565886, 'gradnorm/compressor.layer7': 1.3283255930491185, 'gradnorm/decoder.layer0': 1.4925057436322953, 'gradnorm/decoder.layer1': 1.0957663342744308, 'gradnorm/decoder.layer2': 0.5264427965883451, 'gradnorm/max_param': 91.33968353271484, 'epoch': 5.76}

 58%|█████▊    | 11000/19090 [5:21:28<3:56:11,  1.75s/it]
 58%|█████▊    | 11001/19090 [5:21:32<5:05:43,  2.27s/it]
 58%|█████▊    | 11002/19090 [5:21:33<4:26:11,  1.97s/it]
 58%|█████▊    | 11003/19090 [5:21:34<3:58:07,  1.77s/it]
 58%|█████▊    | 11004/19090 [5:21:35<3:38:56,  1.62s/it]
 58%|█████▊    | 11005/19090 [5:21:37<3:37:37,  1.62s/it]
 58%|█████▊    | 11006/19090 [5:21:38<3:24:40,  1.52s/it]
 58%|█████▊    | 11007/19090 [5:21:40<3:15:32,  1.45s/it]
 58%|█████▊    | 11008/19090 [5:21:42<4:09:52,  1.86s/it]
 58%|█████▊    | 11009/19090 [5:21:44<3:47:10,  1.69s/it][gradnorm] step 11009: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 109.3

 58%|█████▊    | 11010/19090 [5:21:45<3:31:07,  1.57s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 157.85470581054688, 'learning_rate': 2.267428115530066e-05, 'recon_loss': 0.0014991697098594159, 'recon_static': 0.0005340381801943294, 'recon_motion': 0.0009651315122027881, 'recon_norm': 1.2739746928215028, 'recon_norm_std': 0.04462838266044855, 'target_norm': 1.6216404914855957, 'recon_cos': 0.648720794916153, 'gradnorm/compressor.query': 0.13921957922323547, 'gradnorm/compressor.layer0': 4.407199622016319, 'gradnorm/compressor.layer1': 154.9584621761587, 'gradnorm/compressor.layer2': 26.34556173337124, 'gradnorm/compressor.layer3': 12.614403233816196, 'gradnorm/compressor.layer4': 4.330095964738284, 'gradnorm/compressor.layer5': 2.359692078561828, 'gradnorm/compressor.layer6': 1.4503805023012715, 'gradnorm/compressor.layer7': 1.5083145333623633, 'gradnorm/decoder.layer0': 1.7506967247539504, 'gradnorm/decoder.layer1': 1.131330902410206, 'gradnorm/decoder.layer2': 0.5368529383841593, 'gradnorm/max_param': 109.32866668701172, 'epoch': 5.77}

 58%|█████▊    | 11010/19090 [5:21:45<3:31:07,  1.57s/it]
 58%|█████▊    | 11011/19090 [5:21:48<4:15:47,  1.90s/it]
 58%|█████▊    | 11012/19090 [5:21:49<3:51:30,  1.72s/it]
 58%|█████▊    | 11013/19090 [5:21:50<3:34:17,  1.59s/it]
 58%|█████▊    | 11014/19090 [5:21:53<4:20:01,  1.93s/it]
 58%|█████▊    | 11015/19090 [5:21:54<3:54:21,  1.74s/it]
 58%|█████▊    | 11016/19090 [5:21:56<3:36:17,  1.61s/it]
 58%|█████▊    | 11017/19090 [5:21:58<4:16:40,  1.91s/it]
 58%|█████▊    | 11018/19090 [5:22:00<3:51:58,  1.72s/it]
 58%|█████▊    | 11019/19090 [5:22:01<3:34:20,  1.59s/it]
 58%|█████▊    | 11020/19090 [5:22:04<4:24:45,  1.97s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 126.25228118896484, 'learning_rate': 2.2628770126410974e-05, 'recon_loss': 0.0014722706866450608, 'recon_static': 0.0004990329223801381, 'recon_motion': 0.000973237762809731, 'recon_norm': 1.2608541935682296, 'recon_norm_std': 0.04440675703808665, 'target_norm': 1.6229687631130219, 'recon_cos': 0.6555174931883812, 'gradnorm/compressor.query': 0.11091641373441896, 'gradnorm/compressor.layer0': 3.5216223530494517, 'gradnorm/compressor.layer1': 123.48398649976846, 'gradnorm/compressor.layer2': 22.992539232358887, 'gradnorm/compressor.layer3': 11.144263915804293, 'gradnorm/compressor.layer4': 3.689090838196232, 'gradnorm/compressor.layer5': 2.146431169900484, 'gradnorm/compressor.layer6': 1.3065312315532482, 'gradnorm/compressor.layer7': 1.397763436111119, 'gradnorm/decoder.layer0': 1.603737819206143, 'gradnorm/decoder.layer1': 1.1422125337416997, 'gradnorm/decoder.layer2': 0.570501452981568, 'gradnorm/max_param': 87.11322021484375, 'epoch': 5.77}

 58%|█████▊    | 11020/19090 [5:22:04<4:24:45,  1.97s/it]
 58%|█████▊    | 11021/19090 [5:22:05<3:57:27,  1.77s/it]
 58%|█████▊    | 11022/19090 [5:22:06<3:38:08,  1.62s/it]
 58%|█████▊    | 11023/19090 [5:22:09<4:28:06,  1.99s/it]
 58%|█████▊    | 11024/19090 [5:22:10<3:59:44,  1.78s/it]
 58%|█████▊    | 11025/19090 [5:22:12<3:39:59,  1.64s/it]
 58%|█████▊    | 11026/19090 [5:22:14<4:21:54,  1.95s/it]
 58%|█████▊    | 11027/19090 [5:22:16<3:55:28,  1.75s/it]
 58%|█████▊    | 11028/19090 [5:22:17<3:36:59,  1.61s/it]
 58%|█████▊    | 11029/19090 [5:22:19<4:10:36,  1.87s/it][gradnorm] step 11029: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 108.2

 58%|█████▊    | 11030/19090 [5:22:21<4:02:01,  1.80s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 156.7244110107422, 'learning_rate': 2.2583267025754312e-05, 'recon_loss': 0.0014710239163832739, 'recon_static': 0.0005043343153374735, 'recon_motion': 0.0009666895814007149, 'recon_norm': 1.2636532485485077, 'recon_norm_std': 0.04408248560503125, 'target_norm': 1.6196057051420212, 'recon_cos': 0.6545405805110931, 'gradnorm/compressor.query': 0.15007533426782943, 'gradnorm/compressor.layer0': 4.764478601867811, 'gradnorm/compressor.layer1': 153.3663913639491, 'gradnorm/compressor.layer2': 28.956942459478736, 'gradnorm/compressor.layer3': 12.314832747172018, 'gradnorm/compressor.layer4': 3.881679561512222, 'gradnorm/compressor.layer5': 2.3363295006122122, 'gradnorm/compressor.layer6': 1.2797419814737747, 'gradnorm/compressor.layer7': 1.2663780138655578, 'gradnorm/decoder.layer0': 1.7334062087193758, 'gradnorm/decoder.layer1': 1.1698297582161294, 'gradnorm/decoder.layer2': 0.5216644776189698, 'gradnorm/max_param': 108.16966247558594, 'epoch': 5.78}

 58%|█████▊    | 11030/19090 [5:22:21<4:02:01,  1.80s/it]
 58%|█████▊    | 11031/19090 [5:22:22<3:41:34,  1.65s/it]
 58%|█████▊    | 11032/19090 [5:22:25<4:08:53,  1.85s/it]
 58%|█████▊    | 11033/19090 [5:22:26<4:08:22,  1.85s/it]
 58%|█████▊    | 11034/19090 [5:22:28<3:45:30,  1.68s/it]
 58%|█████▊    | 11035/19090 [5:22:30<4:01:10,  1.80s/it]
 58%|█████▊    | 11036/19090 [5:22:32<3:55:57,  1.76s/it]
 58%|█████▊    | 11037/19090 [5:22:33<3:42:56,  1.66s/it]
 58%|█████▊    | 11038/19090 [5:22:35<4:12:25,  1.88s/it]
 58%|█████▊    | 11039/19090 [5:22:37<3:48:33,  1.70s/it]
 58%|█████▊    | 11040/19090 [5:22:38<3:35:24,  1.61s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 140.0323944091797, 'learning_rate': 2.2537772005470782e-05, 'recon_loss': 0.0014829079125775024, 'recon_static': 0.0005105739743157756, 'recon_motion': 0.0009723339942866005, 'recon_norm': 1.2619139313697816, 'recon_norm_std': 0.043710894975811244, 'target_norm': 1.6224887728691102, 'recon_cos': 0.6524625524878502, 'gradnorm/compressor.query': 0.14865784227554132, 'gradnorm/compressor.layer0': 4.404606938919403, 'gradnorm/compressor.layer1': 136.90420635886247, 'gradnorm/compressor.layer2': 26.35739517238829, 'gradnorm/compressor.layer3': 11.18560809116521, 'gradnorm/compressor.layer4': 3.8517655324909326, 'gradnorm/compressor.layer5': 2.1407576471147824, 'gradnorm/compressor.layer6': 1.1753816654349398, 'gradnorm/compressor.layer7': 1.4460142339425308, 'gradnorm/decoder.layer0': 1.7119833480123163, 'gradnorm/decoder.layer1': 0.9535045493426961, 'gradnorm/decoder.layer2': 0.5753242594249078, 'gradnorm/max_param': 96.56452178955078, 'epoch': 5.78}

 58%|█████▊    | 11040/19090 [5:22:38<3:35:24,  1.61s/it]
 58%|█████▊    | 11041/19090 [5:22:41<4:24:20,  1.97s/it]
 58%|█████▊    | 11042/19090 [5:22:42<3:56:48,  1.77s/it]
 58%|█████▊    | 11043/19090 [5:22:43<3:37:30,  1.62s/it]
 58%|█████▊    | 11044/19090 [5:22:46<4:08:09,  1.85s/it]
 58%|█████▊    | 11045/19090 [5:22:47<3:54:38,  1.75s/it]
 58%|█████▊    | 11046/19090 [5:22:49<3:39:07,  1.63s/it]
 58%|█████▊    | 11047/19090 [5:22:51<3:58:57,  1.78s/it]
 58%|█████▊    | 11048/19090 [5:22:52<3:53:20,  1.74s/it]
 58%|█████▊    | 11049/19090 [5:22:54<3:50:19,  1.72s/it][gradnorm] step 11049: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 126.0

 58%|█████▊    | 11050/19090 [5:22:56<4:03:41,  1.82s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 181.77037048339844, 'learning_rate': 2.2492285217673502e-05, 'recon_loss': 0.0014647011586930603, 'recon_static': 0.0004921725987514947, 'recon_motion': 0.0009725285548483953, 'recon_norm': 1.2580964714288712, 'recon_norm_std': 0.045404094457626346, 'target_norm': 1.624306657910347, 'recon_cos': 0.6580572918057441, 'gradnorm/compressor.query': 0.162669048731062, 'gradnorm/compressor.layer0': 4.591822420906714, 'gradnorm/compressor.layer1': 178.55405015883846, 'gradnorm/compressor.layer2': 30.859692442987523, 'gradnorm/compressor.layer3': 12.49935423136619, 'gradnorm/compressor.layer4': 3.6468209518441634, 'gradnorm/compressor.layer5': 2.3968237004764807, 'gradnorm/compressor.layer6': 1.3843315531024711, 'gradnorm/compressor.layer7': 1.7121956436456498, 'gradnorm/decoder.layer0': 1.901552697738672, 'gradnorm/decoder.layer1': 1.1860703344927093, 'gradnorm/decoder.layer2': 0.5816330539679452, 'gradnorm/max_param': 125.96009826660156, 'epoch': 5.79}

 58%|█████▊    | 11050/19090 [5:22:56<4:03:41,  1.82s/it]
 58%|█████▊    | 11051/19090 [5:22:58<4:14:54,  1.90s/it]
 58%|█████▊    | 11052/19090 [5:23:00<3:53:14,  1.74s/it]
 58%|█████▊    | 11053/19090 [5:23:02<4:07:37,  1.85s/it]
 58%|█████▊    | 11054/19090 [5:23:04<4:12:34,  1.89s/it]
 58%|█████▊    | 11055/19090 [5:23:05<3:51:09,  1.73s/it]
 58%|█████▊    | 11056/19090 [5:23:07<4:03:49,  1.82s/it]
 58%|█████▊    | 11057/19090 [5:23:09<4:03:20,  1.82s/it]
 58%|█████▊    | 11058/19090 [5:23:10<3:51:12,  1.73s/it]
 58%|█████▊    | 11059/19090 [5:23:12<4:05:03,  1.83s/it][gradnorm] step 11059: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 106.4

 58%|█████▊    | 11060/19090 [5:23:14<3:49:06,  1.71s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 153.57383728027344, 'learning_rate': 2.2446806814448033e-05, 'recon_loss': 0.0014600076392525807, 'recon_static': 0.0005016211754991673, 'recon_motion': 0.0009583865088643506, 'recon_norm': 1.2605345785617827, 'recon_norm_std': 0.04412935925647617, 'target_norm': 1.6207609683275224, 'recon_cos': 0.6577584177255631, 'gradnorm/compressor.query': 0.18553102325186577, 'gradnorm/compressor.layer0': 6.008269333553672, 'gradnorm/compressor.layer1': 150.8370201264431, 'gradnorm/compressor.layer2': 25.825186283588184, 'gradnorm/compressor.layer3': 10.176935815018576, 'gradnorm/compressor.layer4': 3.519414597080051, 'gradnorm/compressor.layer5': 2.173607773079169, 'gradnorm/compressor.layer6': 1.322685186404183, 'gradnorm/compressor.layer7': 1.517728623791511, 'gradnorm/decoder.layer0': 1.917823988041091, 'gradnorm/decoder.layer1': 1.1510590714509588, 'gradnorm/decoder.layer2': 0.5530890954212598, 'gradnorm/max_param': 106.38431549072266, 'epoch': 5.79}

 58%|█████▊    | 11060/19090 [5:23:14<3:49:06,  1.71s/it]
 58%|█████▊    | 11061/19090 [5:23:16<4:06:00,  1.84s/it]
 58%|█████▊    | 11062/19090 [5:23:18<4:15:02,  1.91s/it]
 58%|█████▊    | 11063/19090 [5:23:20<4:00:03,  1.79s/it]
 58%|█████▊    | 11064/19090 [5:23:22<4:08:18,  1.86s/it]
 58%|█████▊    | 11065/19090 [5:23:24<4:27:16,  2.00s/it]
 58%|█████▊    | 11066/19090 [5:23:25<3:58:46,  1.79s/it]
 58%|█████▊    | 11067/19090 [5:23:27<3:58:35,  1.78s/it]
 58%|█████▊    | 11068/19090 [5:23:30<4:27:56,  2.00s/it]
 58%|█████▊    | 11069/19090 [5:23:31<3:59:09,  1.79s/it][gradnorm] step 11069: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 104.4

 58%|█████▊    | 11070/19090 [5:23:33<3:54:42,  1.76s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 151.21731567382812, 'learning_rate': 2.2401336947851908e-05, 'recon_loss': 0.001467789095477201, 'recon_static': 0.0005096263143059332, 'recon_motion': 0.0009581627484294586, 'recon_norm': 1.2666858285665512, 'recon_norm_std': 0.04402933586388826, 'target_norm': 1.6175713151693345, 'recon_cos': 0.654617503285408, 'gradnorm/compressor.query': 0.11937742074164687, 'gradnorm/compressor.layer0': 4.367774227902882, 'gradnorm/compressor.layer1': 148.0200379699485, 'gradnorm/compressor.layer2': 27.408110639162885, 'gradnorm/compressor.layer3': 12.278273360049512, 'gradnorm/compressor.layer4': 4.265479452144277, 'gradnorm/compressor.layer5': 2.7275598663836935, 'gradnorm/compressor.layer6': 1.2745921978741486, 'gradnorm/compressor.layer7': 1.5767585201538268, 'gradnorm/decoder.layer0': 2.0951107784289498, 'gradnorm/decoder.layer1': 1.1030266412645782, 'gradnorm/decoder.layer2': 0.5827953520187605, 'gradnorm/max_param': 104.42617797851562, 'epoch': 5.8}

 58%|█████▊    | 11070/19090 [5:23:33<3:54:42,  1.76s/it]
 58%|█████▊    | 11071/19090 [5:23:35<4:17:26,  1.93s/it]
 58%|█████▊    | 11072/19090 [5:23:36<3:52:15,  1.74s/it]
 58%|█████▊    | 11073/19090 [5:23:37<3:35:27,  1.61s/it]
 58%|█████▊    | 11074/19090 [5:23:40<4:21:20,  1.96s/it]
 58%|█████▊    | 11075/19090 [5:23:42<3:54:41,  1.76s/it]
 58%|█████▊    | 11076/19090 [5:23:43<3:37:36,  1.63s/it]
 58%|█████▊    | 11077/19090 [5:23:46<4:22:27,  1.97s/it]
 58%|█████▊    | 11078/19090 [5:23:47<3:55:47,  1.77s/it]
 58%|█████▊    | 11079/19090 [5:23:48<3:43:54,  1.68s/it][gradnorm] step 11079: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 100.4

 58%|█████▊    | 11080/19090 [5:23:51<4:30:17,  2.02s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 145.83079528808594, 'learning_rate': 2.2355875769914137e-05, 'recon_loss': 0.0014517002215143292, 'recon_static': 0.0004841009729716461, 'recon_motion': 0.0009675992987467907, 'recon_norm': 1.2530299842357635, 'recon_norm_std': 0.043838109076023105, 'target_norm': 1.6192680656909944, 'recon_cos': 0.6587389528751373, 'gradnorm/compressor.query': 0.10906869709476374, 'gradnorm/compressor.layer0': 3.585356222674288, 'gradnorm/compressor.layer1': 142.27042911300998, 'gradnorm/compressor.layer2': 29.290711851732965, 'gradnorm/compressor.layer3': 11.21001432120151, 'gradnorm/compressor.layer4': 3.94424478416584, 'gradnorm/compressor.layer5': 2.4005035547644513, 'gradnorm/compressor.layer6': 1.2828430413758565, 'gradnorm/compressor.layer7': 1.3830053766081278, 'gradnorm/decoder.layer0': 1.7290906391511893, 'gradnorm/decoder.layer1': 1.0321088162539707, 'gradnorm/decoder.layer2': 0.5763566060453834, 'gradnorm/max_param': 100.3699722290039, 'epoch': 5.8}

 58%|█████▊    | 11080/19090 [5:23:51<4:30:17,  2.02s/it]
 58%|█████▊    | 11081/19090 [5:23:53<4:01:06,  1.81s/it]
 58%|█████▊    | 11082/19090 [5:23:54<3:40:38,  1.65s/it]
 58%|█████▊    | 11083/19090 [5:23:57<4:35:10,  2.06s/it]
 58%|█████▊    | 11084/19090 [5:23:58<4:04:54,  1.84s/it]
 58%|█████▊    | 11085/19090 [5:23:59<3:43:40,  1.68s/it]
 58%|█████▊    | 11086/19090 [5:24:02<4:32:43,  2.04s/it]
 58%|█████▊    | 11087/19090 [5:24:04<4:02:50,  1.82s/it]
 58%|█████▊    | 11088/19090 [5:24:05<3:42:03,  1.66s/it]
 58%|█████▊    | 11089/19090 [5:24:08<4:22:36,  1.97s/it][gradnorm] step 11089: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 104.4

 58%|█████▊    | 11090/19090 [5:24:09<3:55:42,  1.77s/it]
                                                         
{'loss': 0.0014, 'grad_norm': 150.80441284179688, 'learning_rate': 2.2310423432634663e-05, 'recon_loss': 0.0014368506439495833, 'recon_static': 0.0004801984716323204, 'recon_motion': 0.0009566522101522423, 'recon_norm': 1.2517831534147263, 'recon_norm_std': 0.04423598013818264, 'target_norm': 1.6185620933771134, 'recon_cos': 0.6624216422438621, 'gradnorm/compressor.query': 0.1233173304509302, 'gradnorm/compressor.layer0': 3.73070955947239, 'gradnorm/compressor.layer1': 147.99002289005836, 'gradnorm/compressor.layer2': 25.81952060056439, 'gradnorm/compressor.layer3': 11.554673633992694, 'gradnorm/compressor.layer4': 3.8584788899879214, 'gradnorm/compressor.layer5': 2.1708965704768737, 'gradnorm/compressor.layer6': 1.2813908569319354, 'gradnorm/compressor.layer7': 1.288118745213944, 'gradnorm/decoder.layer0': 1.658149996963426, 'gradnorm/decoder.layer1': 0.9692213388494823, 'gradnorm/decoder.layer2': 0.4920945003990963, 'gradnorm/max_param': 104.38034057617188, 'epoch': 5.81}

 58%|█████▊    | 11090/19090 [5:24:09<3:55:42,  1.77s/it]
 58%|█████▊    | 11091/19090 [5:24:10<3:37:50,  1.63s/it]
 58%|█████▊    | 11092/19090 [5:24:13<4:12:47,  1.90s/it]
 58%|█████▊    | 11093/19090 [5:24:14<3:48:40,  1.72s/it]
 58%|█████▊    | 11094/19090 [5:24:15<3:31:51,  1.59s/it]
 58%|█████▊    | 11095/19090 [5:24:18<4:06:47,  1.85s/it]
 58%|█████▊    | 11096/19090 [5:24:19<3:44:36,  1.69s/it]
 58%|█████▊    | 11097/19090 [5:24:21<3:34:02,  1.61s/it]
 58%|█████▊    | 11098/19090 [5:24:23<4:08:10,  1.86s/it]
 58%|█████▊    | 11099/19090 [5:24:24<3:45:15,  1.69s/it]
 58%|█████▊    | 11100/19090 [5:24:26<3:48:14,  1.71s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 125.72156524658203, 'learning_rate': 2.226498008798386e-05, 'recon_loss': 0.001470962687744759, 'recon_static': 0.0005129217192006763, 'recon_motion': 0.0009580409692716785, 'recon_norm': 1.2615960866212845, 'recon_norm_std': 0.04472419237717986, 'target_norm': 1.6196285963058472, 'recon_cos': 0.6541122987866401, 'gradnorm/compressor.query': 0.1378809374359124, 'gradnorm/compressor.layer0': 3.833983753144053, 'gradnorm/compressor.layer1': 122.87379990512623, 'gradnorm/compressor.layer2': 24.13369262315459, 'gradnorm/compressor.layer3': 9.373386498522484, 'gradnorm/compressor.layer4': 3.5723294293611674, 'gradnorm/compressor.layer5': 1.9885480960231119, 'gradnorm/compressor.layer6': 1.145406811254201, 'gradnorm/compressor.layer7': 1.2547603087244028, 'gradnorm/decoder.layer0': 1.522178090206814, 'gradnorm/decoder.layer1': 0.883987589925937, 'gradnorm/decoder.layer2': 0.47370807796888237, 'gradnorm/max_param': 86.68244934082031, 'epoch': 5.81}

 58%|█████▊    | 11100/19090 [5:24:26<3:48:14,  1.71s/it]
 58%|█████▊    | 11101/19090 [5:24:28<4:15:58,  1.92s/it]
 58%|█████▊    | 11102/19090 [5:24:30<3:50:52,  1.73s/it]
 58%|█████▊    | 11103/19090 [5:24:32<3:52:34,  1.75s/it]
 58%|█████▊    | 11104/19090 [5:24:34<4:10:19,  1.88s/it]
 58%|█████▊    | 11105/19090 [5:24:35<3:47:06,  1.71s/it]
 58%|█████▊    | 11106/19090 [5:24:37<4:03:53,  1.83s/it]
 58%|█████▊    | 11107/19090 [5:24:39<4:03:20,  1.83s/it]
 58%|█████▊    | 11108/19090 [5:24:40<3:41:43,  1.67s/it]
 58%|█████▊    | 11109/19090 [5:24:42<4:03:43,  1.83s/it]
 58%|█████▊    | 11110/19090 [5:24:44<4:08:47,  1.87s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 134.79168701171875, 'learning_rate': 2.221954588790206e-05, 'recon_loss': 0.0014810636290349065, 'recon_static': 0.0005136714571563061, 'recon_motion': 0.0009673921696958133, 'recon_norm': 1.2661883115768433, 'recon_norm_std': 0.044198749400675294, 'target_norm': 1.6223903805017472, 'recon_cos': 0.6534199312329292, 'gradnorm/compressor.query': 0.12665246426910418, 'gradnorm/compressor.layer0': 3.8194389358634915, 'gradnorm/compressor.layer1': 132.18630318607143, 'gradnorm/compressor.layer2': 23.805554115337902, 'gradnorm/compressor.layer3': 9.425012676523927, 'gradnorm/compressor.layer4': 3.7528833112945783, 'gradnorm/compressor.layer5': 2.045158190403176, 'gradnorm/compressor.layer6': 1.197670350113076, 'gradnorm/compressor.layer7': 1.3950753739889514, 'gradnorm/decoder.layer0': 1.5963223211981712, 'gradnorm/decoder.layer1': 0.9953133611510667, 'gradnorm/decoder.layer2': 0.5056838095453348, 'gradnorm/max_param': 93.24773406982422, 'epoch': 5.82}

 58%|█████▊    | 11110/19090 [5:24:44<4:08:47,  1.87s/it]
 58%|█████▊    | 11111/19090 [5:24:46<3:45:37,  1.70s/it]
 58%|█████▊    | 11112/19090 [5:24:48<4:09:35,  1.88s/it]
 58%|█████▊    | 11113/19090 [5:24:50<3:58:30,  1.79s/it]
 58%|█████▊    | 11114/19090 [5:24:51<3:38:28,  1.64s/it]
 58%|█████▊    | 11115/19090 [5:24:53<4:04:13,  1.84s/it]
 58%|█████▊    | 11116/19090 [5:24:55<3:58:45,  1.80s/it]
 58%|█████▊    | 11117/19090 [5:24:56<3:38:18,  1.64s/it]
 58%|█████▊    | 11118/19090 [5:24:59<4:12:59,  1.90s/it]
 58%|█████▊    | 11119/19090 [5:25:00<4:05:13,  1.85s/it]
 58%|█████▊    | 11120/19090 [5:25:02<3:42:59,  1.68s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 128.2480010986328, 'learning_rate': 2.2174120984299014e-05, 'recon_loss': 0.00148668511537835, 'recon_static': 0.0005253558221738786, 'recon_motion': 0.0009613292597350665, 'recon_norm': 1.2691933989524842, 'recon_norm_std': 0.04415872953832149, 'target_norm': 1.6202668309211732, 'recon_cos': 0.6510421335697174, 'gradnorm/compressor.query': 0.1080872950315269, 'gradnorm/compressor.layer0': 3.335694636989824, 'gradnorm/compressor.layer1': 125.80423314867032, 'gradnorm/compressor.layer2': 21.968999348469353, 'gradnorm/compressor.layer3': 10.257535815956238, 'gradnorm/compressor.layer4': 3.3764247006493826, 'gradnorm/compressor.layer5': 1.9921350210420925, 'gradnorm/compressor.layer6': 1.2072843020952067, 'gradnorm/compressor.layer7': 1.2021263238281374, 'gradnorm/decoder.layer0': 1.5372962692625405, 'gradnorm/decoder.layer1': 0.9669445503041446, 'gradnorm/decoder.layer2': 0.5300888105424426, 'gradnorm/max_param': 88.74308013916016, 'epoch': 5.83}

 58%|█████▊    | 11120/19090 [5:25:02<3:42:59,  1.68s/it]
 58%|█████▊    | 11121/19090 [5:25:04<4:12:02,  1.90s/it]
 58%|█████▊    | 11122/19090 [5:25:06<4:00:56,  1.81s/it]
 58%|█████▊    | 11123/19090 [5:25:07<3:39:50,  1.66s/it]
 58%|█████▊    | 11124/19090 [5:25:09<4:03:24,  1.83s/it]
 58%|█████▊    | 11125/19090 [5:25:11<3:55:11,  1.77s/it]
 58%|█████▊    | 11126/19090 [5:25:12<3:35:49,  1.63s/it]
 58%|█████▊    | 11127/19090 [5:25:15<4:08:16,  1.87s/it]
 58%|█████▊    | 11128/19090 [5:25:16<3:49:11,  1.73s/it]
 58%|█████▊    | 11129/19090 [5:25:17<3:31:50,  1.60s/it][gradnorm] step 11129: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 113.9

 58%|█████▊    | 11130/19090 [5:25:20<4:16:10,  1.93s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 164.74998474121094, 'learning_rate': 2.2128705529053364e-05, 'recon_loss': 0.0014529420237522571, 'recon_static': 0.0004914043958706316, 'recon_motion': 0.0009615375660359859, 'recon_norm': 1.2568311393260956, 'recon_norm_std': 0.044396286364644766, 'target_norm': 1.6199466019868851, 'recon_cos': 0.6589578419923783, 'gradnorm/compressor.query': 0.11930649576820784, 'gradnorm/compressor.layer0': 3.7566177467426205, 'gradnorm/compressor.layer1': 161.47331919701162, 'gradnorm/compressor.layer2': 29.993001503019936, 'gradnorm/compressor.layer3': 11.257126310086, 'gradnorm/compressor.layer4': 3.984909440642544, 'gradnorm/compressor.layer5': 2.1614926500162377, 'gradnorm/compressor.layer6': 1.2789007374359804, 'gradnorm/compressor.layer7': 1.4355296023620734, 'gradnorm/decoder.layer0': 1.6232940816082961, 'gradnorm/decoder.layer1': 1.1162320299400128, 'gradnorm/decoder.layer2': 0.6040770995955591, 'gradnorm/max_param': 113.90167999267578, 'epoch': 5.83}

 58%|█████▊    | 11130/19090 [5:25:20<4:16:10,  1.93s/it]
 58%|█████▊    | 11131/19090 [5:25:21<3:50:40,  1.74s/it]
 58%|█████▊    | 11132/19090 [5:25:23<3:33:01,  1.61s/it]
 58%|█████▊    | 11133/19090 [5:25:25<4:18:11,  1.95s/it]
 58%|█████▊    | 11134/19090 [5:25:27<3:57:09,  1.79s/it]
 58%|█████▊    | 11135/19090 [5:25:28<3:37:11,  1.64s/it]
 58%|█████▊    | 11136/19090 [5:25:31<4:21:21,  1.97s/it]
 58%|█████▊    | 11137/19090 [5:25:32<4:04:44,  1.85s/it]
 58%|█████▊    | 11138/19090 [5:25:34<3:42:33,  1.68s/it]
 58%|█████▊    | 11139/19090 [5:25:36<4:19:12,  1.96s/it]
 58%|█████▊    | 11140/19090 [5:25:38<4:09:30,  1.88s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 144.52296447753906, 'learning_rate': 2.2083299674012197e-05, 'recon_loss': 0.0015164864249527454, 'recon_static': 0.0005459223095385824, 'recon_motion': 0.000970564161252696, 'recon_norm': 1.2800039142370223, 'recon_norm_std': 0.04335567336529493, 'target_norm': 1.6227192670106887, 'recon_cos': 0.6456590071320534, 'gradnorm/compressor.query': 0.1109643186805961, 'gradnorm/compressor.layer0': 3.6547074935907657, 'gradnorm/compressor.layer1': 141.5417548265352, 'gradnorm/compressor.layer2': 26.43980278645661, 'gradnorm/compressor.layer3': 10.556085749466876, 'gradnorm/compressor.layer4': 4.129283415025717, 'gradnorm/compressor.layer5': 2.101740039005499, 'gradnorm/compressor.layer6': 1.2638013440275464, 'gradnorm/compressor.layer7': 1.408363957420119, 'gradnorm/decoder.layer0': 1.6346975150555407, 'gradnorm/decoder.layer1': 0.9768688179254931, 'gradnorm/decoder.layer2': 0.5306957382840407, 'gradnorm/max_param': 99.8486328125, 'epoch': 5.84}

 58%|█████▊    | 11140/19090 [5:25:38<4:09:30,  1.88s/it]
 58%|█████▊    | 11141/19090 [5:25:39<3:45:49,  1.70s/it]
 58%|█████▊    | 11142/19090 [5:25:42<4:21:05,  1.97s/it]
 58%|█████▊    | 11143/19090 [5:25:44<4:10:19,  1.89s/it]
 58%|█████▊    | 11144/19090 [5:25:45<3:46:15,  1.71s/it]
 58%|█████▊    | 11145/19090 [5:25:47<4:17:26,  1.94s/it]
 58%|█████▊    | 11146/19090 [5:25:49<3:58:54,  1.80s/it]
 58%|█████▊    | 11147/19090 [5:25:50<3:38:15,  1.65s/it]
 58%|█████▊    | 11148/19090 [5:25:53<4:13:11,  1.91s/it]
 58%|█████▊    | 11149/19090 [5:25:54<4:03:22,  1.84s/it][gradnorm] step 11149: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 101.4

 58%|█████▊    | 11150/19090 [5:25:56<3:41:34,  1.67s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 147.1746368408203, 'learning_rate': 2.203790357099049e-05, 'recon_loss': 0.0015226460847770796, 'recon_static': 0.0005594253321760334, 'recon_motion': 0.000963220804987941, 'recon_norm': 1.2828659296035767, 'recon_norm_std': 0.04403751557692885, 'target_norm': 1.6214177846908568, 'recon_cos': 0.6429830610752105, 'gradnorm/compressor.query': 0.13056437848156371, 'gradnorm/compressor.layer0': 4.246092851598341, 'gradnorm/compressor.layer1': 143.76999181635574, 'gradnorm/compressor.layer2': 28.440753591347576, 'gradnorm/compressor.layer3': 11.670915393116397, 'gradnorm/compressor.layer4': 3.6204971760292457, 'gradnorm/compressor.layer5': 2.4319474972654254, 'gradnorm/compressor.layer6': 1.2962800005362825, 'gradnorm/compressor.layer7': 1.5166969834256514, 'gradnorm/decoder.layer0': 1.7222560149469992, 'gradnorm/decoder.layer1': 1.0669742510798217, 'gradnorm/decoder.layer2': 0.563798993978046, 'gradnorm/max_param': 101.42937469482422, 'epoch': 5.84}

 58%|█████▊    | 11150/19090 [5:25:56<3:41:34,  1.67s/it]
 58%|█████▊    | 11151/19090 [5:25:58<4:13:10,  1.91s/it]
 58%|█████▊    | 11152/19090 [5:26:00<4:00:09,  1.82s/it]
 58%|█████▊    | 11153/19090 [5:26:01<3:39:12,  1.66s/it]
 58%|█████▊    | 11154/19090 [5:26:03<4:07:30,  1.87s/it]
 58%|█████▊    | 11155/19090 [5:26:05<3:57:46,  1.80s/it]
 58%|█████▊    | 11156/19090 [5:26:06<3:37:42,  1.65s/it]
 58%|█████▊    | 11157/19090 [5:26:09<4:06:44,  1.87s/it]
 58%|█████▊    | 11158/19090 [5:26:10<3:51:54,  1.75s/it]
 58%|█████▊    | 11159/19090 [5:26:11<3:33:27,  1.61s/it][gradnorm] step 11159: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 124.2

 58%|█████▊    | 11160/19090 [5:26:14<4:10:56,  1.90s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 179.42591857910156, 'learning_rate': 2.199251737177059e-05, 'recon_loss': 0.0015195522777503357, 'recon_static': 0.0005617105838609859, 'recon_motion': 0.0009578417884767987, 'recon_norm': 1.2881717562675477, 'recon_norm_std': 0.044099060166627166, 'target_norm': 1.6191413521766662, 'recon_cos': 0.643471221625805, 'gradnorm/compressor.query': 0.15469248672115335, 'gradnorm/compressor.layer0': 4.736447727227009, 'gradnorm/compressor.layer1': 176.130777495729, 'gradnorm/compressor.layer2': 31.116271484132202, 'gradnorm/compressor.layer3': 12.031697054397151, 'gradnorm/compressor.layer4': 4.280360888302112, 'gradnorm/compressor.layer5': 2.8668194360465122, 'gradnorm/compressor.layer6': 1.273324692299035, 'gradnorm/compressor.layer7': 1.6122126330616229, 'gradnorm/decoder.layer0': 1.9134550415211882, 'gradnorm/decoder.layer1': 1.1566077014137, 'gradnorm/decoder.layer2': 0.6404703930631075, 'gradnorm/max_param': 124.24285888671875, 'epoch': 5.85}

 58%|█████▊    | 11160/19090 [5:26:14<4:10:56,  1.90s/it]
 58%|█████▊    | 11161/19090 [5:26:15<3:48:28,  1.73s/it]
 58%|█████▊    | 11162/19090 [5:26:17<3:36:14,  1.64s/it]
 58%|█████▊    | 11163/19090 [5:26:19<4:18:05,  1.95s/it]
 58%|█████▊    | 11164/19090 [5:26:21<3:51:42,  1.75s/it]
 58%|█████▊    | 11165/19090 [5:26:22<3:35:06,  1.63s/it]
 58%|█████▊    | 11166/19090 [5:26:25<4:19:56,  1.97s/it]
 58%|█████▊    | 11167/19090 [5:26:26<3:52:54,  1.76s/it]
 59%|█████▊    | 11168/19090 [5:26:27<3:34:48,  1.63s/it]
 59%|█████▊    | 11169/19090 [5:26:30<4:20:49,  1.98s/it][gradnorm] step 11169: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 123.4

 59%|█████▊    | 11170/19090 [5:26:31<3:54:19,  1.78s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 177.88717651367188, 'learning_rate': 2.194714122810177e-05, 'recon_loss': 0.001506014820188284, 'recon_static': 0.0005380513277486898, 'recon_motion': 0.0009679635317297652, 'recon_norm': 1.2761658996343612, 'recon_norm_std': 0.043511633202433585, 'target_norm': 1.6214064419269563, 'recon_cos': 0.6469288140535354, 'gradnorm/compressor.query': 0.18061224735674525, 'gradnorm/compressor.layer0': 5.096853609583985, 'gradnorm/compressor.layer1': 174.98873289622105, 'gradnorm/compressor.layer2': 28.66886692326822, 'gradnorm/compressor.layer3': 11.907298284724636, 'gradnorm/compressor.layer4': 4.165172665097826, 'gradnorm/compressor.layer5': 2.5917627466020248, 'gradnorm/compressor.layer6': 1.3303420467379694, 'gradnorm/compressor.layer7': 1.4285322848134279, 'gradnorm/decoder.layer0': 1.9254026214851214, 'gradnorm/decoder.layer1': 1.0952189798740446, 'gradnorm/decoder.layer2': 0.5574950236367819, 'gradnorm/max_param': 123.4363784790039, 'epoch': 5.85}

 59%|█████▊    | 11170/19090 [5:26:31<3:54:19,  1.78s/it]
 59%|█████▊    | 11171/19090 [5:26:33<3:43:55,  1.70s/it]
 59%|█████▊    | 11172/19090 [5:26:35<4:01:47,  1.83s/it]
 59%|█████▊    | 11173/19090 [5:26:36<3:40:09,  1.67s/it]
 59%|█████▊    | 11174/19090 [5:26:38<3:46:22,  1.72s/it]
 59%|█████▊    | 11175/19090 [5:26:40<4:06:33,  1.87s/it]
 59%|█████▊    | 11176/19090 [5:26:42<3:43:13,  1.69s/it]
 59%|█████▊    | 11177/19090 [5:26:44<3:49:39,  1.74s/it]
 59%|█████▊    | 11178/19090 [5:26:46<4:04:53,  1.86s/it]
 59%|█████▊    | 11179/19090 [5:26:47<3:44:55,  1.71s/it][gradnorm] step 11179: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 118.0

 59%|█████▊    | 11180/19090 [5:26:49<3:53:32,  1.77s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 170.33984375, 'learning_rate': 2.1901775291699657e-05, 'recon_loss': 0.0014862470154184849, 'recon_static': 0.0005238351710431744, 'recon_motion': 0.0009624118887586519, 'recon_norm': 1.2682694494724274, 'recon_norm_std': 0.04384546047076583, 'target_norm': 1.6193132013082505, 'recon_cos': 0.65045166015625, 'gradnorm/compressor.query': 0.16554209550598284, 'gradnorm/compressor.layer0': 4.973114016574583, 'gradnorm/compressor.layer1': 167.30042331274896, 'gradnorm/compressor.layer2': 28.698626186971584, 'gradnorm/compressor.layer3': 11.97292006234636, 'gradnorm/compressor.layer4': 4.312106349504102, 'gradnorm/compressor.layer5': 2.435023590380937, 'gradnorm/compressor.layer6': 1.3851558420661545, 'gradnorm/compressor.layer7': 1.5677516420112763, 'gradnorm/decoder.layer0': 1.9542071587028245, 'gradnorm/decoder.layer1': 1.1866959309094869, 'gradnorm/decoder.layer2': 0.6183002195419892, 'gradnorm/max_param': 117.99385070800781, 'epoch': 5.86}

 59%|█████▊    | 11180/19090 [5:26:49<3:53:32,  1.77s/it]
 59%|█████▊    | 11181/19090 [5:26:52<4:22:59,  2.00s/it]
 59%|█████▊    | 11182/19090 [5:26:53<3:54:36,  1.78s/it]
 59%|█████▊    | 11183/19090 [5:26:55<3:58:45,  1.81s/it]
 59%|█████▊    | 11184/19090 [5:26:57<4:11:36,  1.91s/it]
 59%|█████▊    | 11185/19090 [5:26:58<3:47:39,  1.73s/it]
 59%|█████▊    | 11186/19090 [5:27:00<3:58:56,  1.81s/it]
 59%|█████▊    | 11187/19090 [5:27:02<4:03:04,  1.85s/it]
 59%|█████▊    | 11188/19090 [5:27:03<3:41:38,  1.68s/it]
 59%|█████▊    | 11189/19090 [5:27:06<4:03:57,  1.85s/it][gradnorm] step 11189: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 106.3

 59%|█████▊    | 11190/19090 [5:27:07<3:55:24,  1.79s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 153.7064666748047, 'learning_rate': 2.185641971424575e-05, 'recon_loss': 0.0015229630429530517, 'recon_static': 0.000557340272644069, 'recon_motion': 0.0009656228095991537, 'recon_norm': 1.2893285751342773, 'recon_norm_std': 0.04323044968768954, 'target_norm': 1.6204541712999343, 'recon_cos': 0.643316687643528, 'gradnorm/compressor.query': 0.13006761496988392, 'gradnorm/compressor.layer0': 4.699894299231653, 'gradnorm/compressor.layer1': 150.68387075679334, 'gradnorm/compressor.layer2': 27.523922034835216, 'gradnorm/compressor.layer3': 10.665409212698611, 'gradnorm/compressor.layer4': 3.7229548171439855, 'gradnorm/compressor.layer5': 2.264030158296974, 'gradnorm/compressor.layer6': 1.199361140343527, 'gradnorm/compressor.layer7': 1.3499088203580027, 'gradnorm/decoder.layer0': 1.728168137962168, 'gradnorm/decoder.layer1': 1.0433853974519516, 'gradnorm/decoder.layer2': 0.5477762425507333, 'gradnorm/max_param': 106.2756576538086, 'epoch': 5.86}

 59%|█████▊    | 11190/19090 [5:27:07<3:55:24,  1.79s/it]
 59%|█████▊    | 11191/19090 [5:27:09<3:37:21,  1.65s/it]
 59%|█████▊    | 11192/19090 [5:27:11<4:12:21,  1.92s/it]
 59%|█████▊    | 11193/19090 [5:27:13<3:56:09,  1.79s/it]
 59%|█████▊    | 11194/19090 [5:27:14<3:37:27,  1.65s/it]
 59%|█████▊    | 11195/19090 [5:27:17<4:22:26,  1.99s/it]
 59%|█████▊    | 11196/19090 [5:27:18<4:01:51,  1.84s/it]
 59%|█████▊    | 11197/19090 [5:27:20<3:40:43,  1.68s/it]
 59%|█████▊    | 11198/19090 [5:27:22<4:08:17,  1.89s/it]
 59%|█████▊    | 11199/19090 [5:27:23<3:53:23,  1.77s/it]
 59%|█████▊    | 11200/19090 [5:27:25<3:34:55,  1.63s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 126.0007553100586, 'learning_rate': 2.1811074647386927e-05, 'recon_loss': 0.001500052071060054, 'recon_static': 0.000543502441723831, 'recon_motion': 0.0009565496409777551, 'recon_norm': 1.2795467615127563, 'recon_norm_std': 0.04377415329217911, 'target_norm': 1.6176887035369873, 'recon_cos': 0.6470576301217079, 'gradnorm/compressor.query': 0.1184526104585668, 'gradnorm/compressor.layer0': 4.186816031551382, 'gradnorm/compressor.layer1': 122.90825227201996, 'gradnorm/compressor.layer2': 24.736576023358104, 'gradnorm/compressor.layer3': 10.55168437799817, 'gradnorm/compressor.layer4': 4.013616987535287, 'gradnorm/compressor.layer5': 2.2043857078627465, 'gradnorm/compressor.layer6': 1.252688568295969, 'gradnorm/compressor.layer7': 1.3602032823664085, 'gradnorm/decoder.layer0': 1.7011571592270376, 'gradnorm/decoder.layer1': 1.1572443835982749, 'gradnorm/decoder.layer2': 0.5931075033053165, 'gradnorm/max_param': 86.70291137695312, 'epoch': 5.87}

 59%|█████▊    | 11200/19090 [5:27:25<3:34:55,  1.63s/it]
 59%|█████▊    | 11201/19090 [5:27:27<4:15:42,  1.94s/it]
 59%|█████▊    | 11202/19090 [5:27:29<3:50:36,  1.75s/it]
 59%|█████▊    | 11203/19090 [5:27:30<3:33:07,  1.62s/it]
 59%|█████▊    | 11204/19090 [5:27:33<4:16:55,  1.95s/it]
 59%|█████▊    | 11205/19090 [5:27:34<3:51:17,  1.76s/it]
 59%|█████▊    | 11206/19090 [5:27:35<3:33:27,  1.62s/it]
 59%|█████▊    | 11207/19090 [5:27:38<4:22:11,  2.00s/it]
 59%|█████▊    | 11208/19090 [5:27:39<3:54:24,  1.78s/it]
 59%|█████▊    | 11209/19090 [5:27:41<3:35:42,  1.64s/it]
 59%|█████▊    | 11210/19090 [5:27:44<4:20:31,  1.98s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 138.21360778808594, 'learning_rate': 2.176574024273491e-05, 'recon_loss': 0.0015869847207795828, 'recon_static': 0.0006278339642449282, 'recon_motion': 0.0009591507958248257, 'recon_norm': 1.3164264976978302, 'recon_norm_std': 0.04260210245847702, 'target_norm': 1.6184092074632646, 'recon_cos': 0.6287212818861008, 'gradnorm/compressor.query': 0.12134499020597225, 'gradnorm/compressor.layer0': 4.186777536499488, 'gradnorm/compressor.layer1': 135.07716255440815, 'gradnorm/compressor.layer2': 26.115997953765692, 'gradnorm/compressor.layer3': 11.21568524250678, 'gradnorm/compressor.layer4': 4.114128583412834, 'gradnorm/compressor.layer5': 2.408054418373502, 'gradnorm/compressor.layer6': 1.4438409685812739, 'gradnorm/compressor.layer7': 1.4817978019025628, 'gradnorm/decoder.layer0': 1.7589994365090513, 'gradnorm/decoder.layer1': 1.155060388134106, 'gradnorm/decoder.layer2': 0.5857712064641615, 'gradnorm/max_param': 95.29528045654297, 'epoch': 5.87}

 59%|█████▊    | 11210/19090 [5:27:44<4:20:31,  1.98s/it]
 59%|█████▊    | 11211/19090 [5:27:45<3:53:01,  1.77s/it]
 59%|█████▊    | 11212/19090 [5:27:46<3:33:44,  1.63s/it]
 59%|█████▊    | 11213/19090 [5:27:49<4:16:11,  1.95s/it]
 59%|█████▊    | 11214/19090 [5:27:50<3:50:03,  1.75s/it]
 59%|█████▊    | 11215/19090 [5:27:51<3:31:40,  1.61s/it]
 59%|█████▉    | 11216/19090 [5:27:54<4:16:05,  1.95s/it]
 59%|█████▉    | 11217/19090 [5:27:55<3:49:50,  1.75s/it]
 59%|█████▉    | 11218/19090 [5:27:57<3:31:36,  1.61s/it]
 59%|█████▉    | 11219/19090 [5:27:59<4:04:41,  1.87s/it]
 59%|█████▉    | 11220/19090 [5:28:00<3:41:57,  1.69s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 144.29234313964844, 'learning_rate': 2.1720416651865775e-05, 'recon_loss': 0.0015789739176398144, 'recon_static': 0.0006160128832561895, 'recon_motion': 0.0009629610140109435, 'recon_norm': 1.3120305061340332, 'recon_norm_std': 0.043072646204382184, 'target_norm': 1.6201217830181123, 'recon_cos': 0.6306823626160621, 'gradnorm/compressor.query': 0.13002775465666302, 'gradnorm/compressor.layer0': 4.242256065643522, 'gradnorm/compressor.layer1': 141.60444328966165, 'gradnorm/compressor.layer2': 24.464269624929535, 'gradnorm/compressor.layer3': 11.14593730176879, 'gradnorm/compressor.layer4': 3.9189686740052507, 'gradnorm/compressor.layer5': 2.1910132397417486, 'gradnorm/compressor.layer6': 1.2374574827346303, 'gradnorm/compressor.layer7': 1.3706476078004084, 'gradnorm/decoder.layer0': 1.640741616600389, 'gradnorm/decoder.layer1': 1.0686177110502615, 'gradnorm/decoder.layer2': 0.5563806830351418, 'gradnorm/max_param': 99.89862823486328, 'epoch': 5.88}

 59%|█████▉    | 11220/19090 [5:28:00<3:41:57,  1.69s/it]
 59%|█████▉    | 11221/19090 [5:28:02<3:26:00,  1.57s/it]
 59%|█████▉    | 11222/19090 [5:28:04<4:01:59,  1.85s/it]
 59%|█████▉    | 11223/19090 [5:28:06<3:39:52,  1.68s/it]
 59%|█████▉    | 11224/19090 [5:28:07<3:24:42,  1.56s/it]
 59%|█████▉    | 11225/19090 [5:28:10<4:22:52,  2.01s/it]
 59%|█████▉    | 11226/19090 [5:28:11<3:55:06,  1.79s/it]
 59%|█████▉    | 11227/19090 [5:28:12<3:35:32,  1.64s/it]
 59%|█████▉    | 11228/19090 [5:28:15<4:18:44,  1.97s/it]
 59%|█████▉    | 11229/19090 [5:28:17<3:51:43,  1.77s/it]
 59%|█████▉    | 11230/19090 [5:28:18<3:33:08,  1.63s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 139.94442749023438, 'learning_rate': 2.167510402631945e-05, 'recon_loss': 0.0014940112334443257, 'recon_static': 0.0005296361821820028, 'recon_motion': 0.0009643750134273432, 'recon_norm': 1.2709743201732635, 'recon_norm_std': 0.04288582289591432, 'target_norm': 1.6191232979297638, 'recon_cos': 0.648347644507885, 'gradnorm/compressor.query': 0.11076555409229413, 'gradnorm/compressor.layer0': 3.7168403659026072, 'gradnorm/compressor.layer1': 136.80885245251858, 'gradnorm/compressor.layer2': 25.962632543219087, 'gradnorm/compressor.layer3': 12.373472793823575, 'gradnorm/compressor.layer4': 3.6861761270586713, 'gradnorm/compressor.layer5': 2.439577537766459, 'gradnorm/compressor.layer6': 1.1781006302071182, 'gradnorm/compressor.layer7': 1.4670780486373098, 'gradnorm/decoder.layer0': 1.5601405067012473, 'gradnorm/decoder.layer1': 0.989046968696708, 'gradnorm/decoder.layer2': 0.54746535997241, 'gradnorm/max_param': 96.51593017578125, 'epoch': 5.88}

 59%|█████▉    | 11230/19090 [5:28:18<3:33:08,  1.63s/it]
 59%|█████▉    | 11231/19090 [5:28:21<4:15:42,  1.95s/it]
 59%|█████▉    | 11232/19090 [5:28:22<3:49:52,  1.76s/it]
 59%|█████▉    | 11233/19090 [5:28:23<3:32:04,  1.62s/it]
 59%|█████▉    | 11234/19090 [5:28:26<4:18:50,  1.98s/it]
 59%|█████▉    | 11235/19090 [5:28:28<4:07:05,  1.89s/it]
 59%|█████▉    | 11236/19090 [5:28:29<3:43:50,  1.71s/it]
 59%|█████▉    | 11237/19090 [5:28:31<3:59:26,  1.83s/it]
 59%|█████▉    | 11238/19090 [5:28:33<3:53:08,  1.78s/it]
 59%|█████▉    | 11239/19090 [5:28:34<3:33:54,  1.63s/it]
 59%|█████▉    | 11240/19090 [5:28:36<4:02:26,  1.85s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 136.64169311523438, 'learning_rate': 2.1629802517599198e-05, 'recon_loss': 0.0015003206470282748, 'recon_static': 0.0005299240569001995, 'recon_motion': 0.0009703966075903736, 'recon_norm': 1.2746908098459244, 'recon_norm_std': 0.04333877861499787, 'target_norm': 1.6210169285535811, 'recon_cos': 0.6481131836771965, 'gradnorm/compressor.query': 0.14079463184912144, 'gradnorm/compressor.layer0': 4.066403904748363, 'gradnorm/compressor.layer1': 133.75452679255383, 'gradnorm/compressor.layer2': 25.505250876351422, 'gradnorm/compressor.layer3': 9.401317228186697, 'gradnorm/compressor.layer4': 3.740107904193376, 'gradnorm/compressor.layer5': 2.151078477102716, 'gradnorm/compressor.layer6': 1.175968953608372, 'gradnorm/compressor.layer7': 1.3755997736834988, 'gradnorm/decoder.layer0': 1.4967320129017165, 'gradnorm/decoder.layer1': 0.9122359455517802, 'gradnorm/decoder.layer2': 0.5056677790535041, 'gradnorm/max_param': 94.34745025634766, 'epoch': 5.89}

 59%|█████▉    | 11240/19090 [5:28:36<4:02:26,  1.85s/it]
 59%|█████▉    | 11241/19090 [5:28:38<4:06:06,  1.88s/it]
 59%|█████▉    | 11242/19090 [5:28:40<3:43:18,  1.71s/it]
 59%|█████▉    | 11243/19090 [5:28:41<3:46:13,  1.73s/it]
 59%|█████▉    | 11244/19090 [5:28:44<4:03:15,  1.86s/it]
 59%|█████▉    | 11245/19090 [5:28:45<3:40:55,  1.69s/it]
 59%|█████▉    | 11246/19090 [5:28:46<3:33:14,  1.63s/it]
 59%|█████▉    | 11247/19090 [5:28:49<4:18:42,  1.98s/it]
 59%|█████▉    | 11248/19090 [5:28:50<3:51:50,  1.77s/it]
 59%|█████▉    | 11249/19090 [5:28:52<3:32:38,  1.63s/it][gradnorm] step 11249: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 102.6

 59%|█████▉    | 11250/19090 [5:28:55<4:21:00,  2.00s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 148.8502960205078, 'learning_rate': 2.158451227717109e-05, 'recon_loss': 0.0015144336386583746, 'recon_static': 0.0005496572885022033, 'recon_motion': 0.0009647763232351281, 'recon_norm': 1.280249160528183, 'recon_norm_std': 0.04285059031099081, 'target_norm': 1.618442216515541, 'recon_cos': 0.643604801595211, 'gradnorm/compressor.query': 0.12434127211324812, 'gradnorm/compressor.layer0': 4.3509369816636285, 'gradnorm/compressor.layer1': 145.46134849591488, 'gradnorm/compressor.layer2': 28.803715953577502, 'gradnorm/compressor.layer3': 10.916928897260968, 'gradnorm/compressor.layer4': 3.8904101454150486, 'gradnorm/compressor.layer5': 2.435702614902108, 'gradnorm/compressor.layer6': 1.2221569467703657, 'gradnorm/compressor.layer7': 1.5562846352273503, 'gradnorm/decoder.layer0': 1.7887824226602536, 'gradnorm/decoder.layer1': 1.041838243976892, 'gradnorm/decoder.layer2': 0.600049183737002, 'gradnorm/max_param': 102.62793731689453, 'epoch': 5.89}

 59%|█████▉    | 11250/19090 [5:28:55<4:21:00,  2.00s/it]
 59%|█████▉    | 11251/19090 [5:28:56<3:53:36,  1.79s/it]
 59%|█████▉    | 11252/19090 [5:28:57<3:34:04,  1.64s/it]
 59%|█████▉    | 11253/19090 [5:29:00<4:19:37,  1.99s/it]
 59%|█████▉    | 11254/19090 [5:29:01<3:52:10,  1.78s/it]
 59%|█████▉    | 11255/19090 [5:29:03<3:32:53,  1.63s/it]
 59%|█████▉    | 11256/19090 [5:29:05<4:25:46,  2.04s/it]
 59%|█████▉    | 11257/19090 [5:29:07<3:56:54,  1.81s/it]
 59%|█████▉    | 11258/19090 [5:29:08<3:36:16,  1.66s/it]
 59%|█████▉    | 11259/19090 [5:29:11<4:21:13,  2.00s/it][gradnorm] step 11259: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 104.9

 59%|█████▉    | 11260/19090 [5:29:12<3:53:34,  1.79s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 151.4096221923828, 'learning_rate': 2.153923345646356e-05, 'recon_loss': 0.00159925670013763, 'recon_static': 0.0006380358681781217, 'recon_motion': 0.0009612208203179761, 'recon_norm': 1.3223756462335587, 'recon_norm_std': 0.04297196716070175, 'target_norm': 1.6214102774858474, 'recon_cos': 0.6285955280065536, 'gradnorm/compressor.query': 0.17098832082810048, 'gradnorm/compressor.layer0': 5.00316881855177, 'gradnorm/compressor.layer1': 148.6709611440647, 'gradnorm/compressor.layer2': 25.24007530430897, 'gradnorm/compressor.layer3': 11.503141554592816, 'gradnorm/compressor.layer4': 3.6430949072086616, 'gradnorm/compressor.layer5': 2.338016669142979, 'gradnorm/compressor.layer6': 1.2714161187996456, 'gradnorm/compressor.layer7': 1.561330662912115, 'gradnorm/decoder.layer0': 1.8103222753971047, 'gradnorm/decoder.layer1': 1.012541800683964, 'gradnorm/decoder.layer2': 0.5226564415901415, 'gradnorm/max_param': 104.88188934326172, 'epoch': 5.9}

 59%|█████▉    | 11260/19090 [5:29:12<3:53:34,  1.79s/it]
 59%|█████▉    | 11261/19090 [5:29:13<3:34:13,  1.64s/it]
 59%|█████▉    | 11262/19090 [5:29:16<4:17:27,  1.97s/it]
 59%|█████▉    | 11263/19090 [5:29:18<3:50:33,  1.77s/it]
 59%|█████▉    | 11264/19090 [5:29:19<3:31:46,  1.62s/it]
 59%|█████▉    | 11265/19090 [5:29:22<4:15:04,  1.96s/it]
 59%|█████▉    | 11266/19090 [5:29:23<3:48:59,  1.76s/it]
 59%|█████▉    | 11267/19090 [5:29:24<3:30:45,  1.62s/it]
 59%|█████▉    | 11268/19090 [5:29:27<4:15:45,  1.96s/it]
 59%|█████▉    | 11269/19090 [5:29:28<3:49:46,  1.76s/it]
 59%|█████▉    | 11270/19090 [5:29:29<3:31:24,  1.62s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 135.52015686035156, 'learning_rate': 2.149396620686684e-05, 'recon_loss': 0.0016604112664936111, 'recon_static': 0.0006879464432131499, 'recon_motion': 0.0009724647563416511, 'recon_norm': 1.3404987514019013, 'recon_norm_std': 0.04192238999530673, 'target_norm': 1.624504840373993, 'recon_cos': 0.6161655902862548, 'gradnorm/compressor.query': 0.14149327577159052, 'gradnorm/compressor.layer0': 4.355508162658, 'gradnorm/compressor.layer1': 132.86738571786532, 'gradnorm/compressor.layer2': 23.166381102341088, 'gradnorm/compressor.layer3': 11.35944014875575, 'gradnorm/compressor.layer4': 3.7173751725953252, 'gradnorm/compressor.layer5': 2.1562337460112584, 'gradnorm/compressor.layer6': 1.2256634083613691, 'gradnorm/compressor.layer7': 1.440784433938664, 'gradnorm/decoder.layer0': 1.9325004800694747, 'gradnorm/decoder.layer1': 1.0826334465357148, 'gradnorm/decoder.layer2': 0.5556984406414193, 'gradnorm/max_param': 93.68061065673828, 'epoch': 5.9}

 59%|█████▉    | 11270/19090 [5:29:29<3:31:24,  1.62s/it]
 59%|█████▉    | 11271/19090 [5:29:32<4:09:41,  1.92s/it]
 59%|█████▉    | 11272/19090 [5:29:33<3:45:40,  1.73s/it]
 59%|█████▉    | 11273/19090 [5:29:35<3:28:34,  1.60s/it]
 59%|█████▉    | 11274/19090 [5:29:37<4:12:51,  1.94s/it]
 59%|█████▉    | 11275/19090 [5:29:39<3:47:22,  1.75s/it]
 59%|█████▉    | 11276/19090 [5:29:40<3:29:31,  1.61s/it]
 59%|█████▉    | 11277/19090 [5:29:43<4:11:49,  1.93s/it]
 59%|█████▉    | 11278/19090 [5:29:44<3:46:51,  1.74s/it]
 59%|█████▉    | 11279/19090 [5:29:46<3:39:21,  1.68s/it][gradnorm] step 11279: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 118.7

 59%|█████▉    | 11280/19090 [5:29:48<4:19:36,  1.99s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 171.7643280029297, 'learning_rate': 2.1448710679732452e-05, 'recon_loss': 0.0015956364950397983, 'recon_static': 0.0006296771898632869, 'recon_motion': 0.0009659593168180435, 'recon_norm': 1.3205675423145293, 'recon_norm_std': 0.04246992515400052, 'target_norm': 1.6192407310009003, 'recon_cos': 0.627339905500412, 'gradnorm/compressor.query': 0.159446241547009, 'gradnorm/compressor.layer0': 5.214590276494308, 'gradnorm/compressor.layer1': 168.28752019423484, 'gradnorm/compressor.layer2': 31.03658768921972, 'gradnorm/compressor.layer3': 12.595998558570098, 'gradnorm/compressor.layer4': 4.306240902297288, 'gradnorm/compressor.layer5': 2.1094525358290306, 'gradnorm/compressor.layer6': 1.4301057631055134, 'gradnorm/compressor.layer7': 1.5768724513023846, 'gradnorm/decoder.layer0': 2.029986766339843, 'gradnorm/decoder.layer1': 1.0890886979530052, 'gradnorm/decoder.layer2': 0.5636442784396215, 'gradnorm/max_param': 118.72896575927734, 'epoch': 5.91}

 59%|█████▉    | 11280/19090 [5:29:48<4:19:36,  1.99s/it]
 59%|█████▉    | 11281/19090 [5:29:50<3:52:05,  1.78s/it]
 59%|█████▉    | 11282/19090 [5:29:51<3:36:37,  1.66s/it]
 59%|█████▉    | 11283/19090 [5:29:53<4:11:23,  1.93s/it]
 59%|█████▉    | 11284/19090 [5:29:55<3:46:11,  1.74s/it]
 59%|█████▉    | 11285/19090 [5:29:56<3:28:35,  1.60s/it]
 59%|█████▉    | 11286/19090 [5:29:59<4:11:49,  1.94s/it]
 59%|█████▉    | 11287/19090 [5:30:00<3:46:34,  1.74s/it]
 59%|█████▉    | 11288/19090 [5:30:01<3:32:04,  1.63s/it]
 59%|█████▉    | 11289/19090 [5:30:04<4:11:48,  1.94s/it][gradnorm] step 11289: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 110.6

 59%|█████▉    | 11290/19090 [5:30:05<3:46:28,  1.74s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 160.00460815429688, 'learning_rate': 2.1403467026372773e-05, 'recon_loss': 0.0015581409417791293, 'recon_static': 0.0005833103467011824, 'recon_motion': 0.0009748305586981587, 'recon_norm': 1.29525545835495, 'recon_norm_std': 0.04226529579609632, 'target_norm': 1.6224786430597304, 'recon_cos': 0.6354045927524566, 'gradnorm/compressor.query': 0.14231536115250537, 'gradnorm/compressor.layer0': 4.456800583563324, 'gradnorm/compressor.layer1': 156.71834184892015, 'gradnorm/compressor.layer2': 29.3180871124566, 'gradnorm/compressor.layer3': 11.389030774067486, 'gradnorm/compressor.layer4': 4.051729449956467, 'gradnorm/compressor.layer5': 2.4536054590204914, 'gradnorm/compressor.layer6': 1.3587066498186926, 'gradnorm/compressor.layer7': 1.3809735845792062, 'gradnorm/decoder.layer0': 1.99002664097728, 'gradnorm/decoder.layer1': 1.1030097814802353, 'gradnorm/decoder.layer2': 0.5747001342878156, 'gradnorm/max_param': 110.55731964111328, 'epoch': 5.91}

 59%|█████▉    | 11290/19090 [5:30:05<3:46:28,  1.74s/it]
 59%|█████▉    | 11291/19090 [5:30:07<3:28:41,  1.61s/it]
 59%|█████▉    | 11292/19090 [5:30:10<4:19:41,  2.00s/it]
 59%|█████▉    | 11293/19090 [5:30:11<3:51:53,  1.78s/it]
 59%|█████▉    | 11294/19090 [5:30:12<3:32:24,  1.63s/it]
 59%|█████▉    | 11295/19090 [5:30:15<4:27:20,  2.06s/it]
 59%|█████▉    | 11296/19090 [5:30:16<3:57:20,  1.83s/it]
 59%|█████▉    | 11297/19090 [5:30:18<3:36:46,  1.67s/it]
 59%|█████▉    | 11298/19090 [5:30:21<4:40:03,  2.16s/it]
 59%|█████▉    | 11299/19090 [5:30:22<4:06:13,  1.90s/it][gradnorm] step 11299: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 103.1

 59%|█████▉    | 11300/19090 [5:30:24<3:42:37,  1.71s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 149.6881866455078, 'learning_rate': 2.1358235398060444e-05, 'recon_loss': 0.0015379135584225878, 'recon_static': 0.0005609215644653886, 'recon_motion': 0.0009769920201506467, 'recon_norm': 1.2861561566591262, 'recon_norm_std': 0.04144152067601681, 'target_norm': 1.6223759949207306, 'recon_cos': 0.6397836223244667, 'gradnorm/compressor.query': 0.19447040219028838, 'gradnorm/compressor.layer0': 6.77866268461177, 'gradnorm/compressor.layer1': 146.1692685831225, 'gradnorm/compressor.layer2': 29.165542086764713, 'gradnorm/compressor.layer3': 10.502740411547062, 'gradnorm/compressor.layer4': 4.341156768632498, 'gradnorm/compressor.layer5': 2.46832088169389, 'gradnorm/compressor.layer6': 1.4292817921141097, 'gradnorm/compressor.layer7': 1.4195862679114808, 'gradnorm/decoder.layer0': 1.9064897480270582, 'gradnorm/decoder.layer1': 1.1205986072604477, 'gradnorm/decoder.layer2': 0.5341892834347195, 'gradnorm/max_param': 103.11872100830078, 'epoch': 5.92}

 59%|█████▉    | 11300/19090 [5:30:24<3:42:37,  1.71s/it]
 59%|█████▉    | 11301/19090 [5:30:26<4:17:17,  1.98s/it]
 59%|█████▉    | 11302/19090 [5:30:28<3:50:24,  1.78s/it]
 59%|█████▉    | 11303/19090 [5:30:29<3:31:31,  1.63s/it]
 59%|█████▉    | 11304/19090 [5:30:32<4:17:55,  1.99s/it]
 59%|█████▉    | 11305/19090 [5:30:33<3:50:44,  1.78s/it]
 59%|█████▉    | 11306/19090 [5:30:34<3:31:52,  1.63s/it]
 59%|█████▉    | 11307/19090 [5:30:37<4:29:29,  2.08s/it]
 59%|█████▉    | 11308/19090 [5:30:39<3:58:41,  1.84s/it]
 59%|█████▉    | 11309/19090 [5:30:40<3:37:13,  1.68s/it]
 59%|█████▉    | 11310/19090 [5:30:43<4:29:25,  2.08s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 136.67701721191406, 'learning_rate': 2.1313015946027892e-05, 'recon_loss': 0.001595923534478061, 'recon_static': 0.0006350466239382513, 'recon_motion': 0.0009608769410988316, 'recon_norm': 1.3200046986341476, 'recon_norm_std': 0.04181049345061183, 'target_norm': 1.6200865089893342, 'recon_cos': 0.627606375515461, 'gradnorm/compressor.query': 0.16085779822912089, 'gradnorm/compressor.layer0': 5.219983897275679, 'gradnorm/compressor.layer1': 133.6598703332455, 'gradnorm/compressor.layer2': 25.047848497535046, 'gradnorm/compressor.layer3': 11.45479955570276, 'gradnorm/compressor.layer4': 3.5724139216672763, 'gradnorm/compressor.layer5': 2.6440393167824503, 'gradnorm/compressor.layer6': 1.5332584055271006, 'gradnorm/compressor.layer7': 1.5365181431740331, 'gradnorm/decoder.layer0': 1.9024493992977651, 'gradnorm/decoder.layer1': 1.1544775393826006, 'gradnorm/decoder.layer2': 0.5900236954214233, 'gradnorm/max_param': 94.29653930664062, 'epoch': 5.92}

 59%|█████▉    | 11310/19090 [5:30:43<4:29:25,  2.08s/it]
 59%|█████▉    | 11311/19090 [5:30:44<3:58:47,  1.84s/it]
 59%|█████▉    | 11312/19090 [5:30:46<3:37:24,  1.68s/it]
 59%|█████▉    | 11313/19090 [5:30:49<4:28:42,  2.07s/it]
 59%|█████▉    | 11314/19090 [5:30:50<3:57:59,  1.84s/it]
 59%|█████▉    | 11315/19090 [5:30:51<3:36:42,  1.67s/it]
 59%|█████▉    | 11316/19090 [5:30:54<4:10:25,  1.93s/it]
 59%|█████▉    | 11317/19090 [5:30:55<3:46:15,  1.75s/it]
 59%|█████▉    | 11318/19090 [5:30:56<3:28:28,  1.61s/it]
 59%|█████▉    | 11319/19090 [5:30:59<4:21:24,  2.02s/it]
 59%|█████▉    | 11320/19090 [5:31:01<3:53:11,  1.80s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 136.49024963378906, 'learning_rate': 2.1267808821466882e-05, 'recon_loss': 0.0016435222671134398, 'recon_static': 0.0006776471971534192, 'recon_motion': 0.0009658750597736799, 'recon_norm': 1.3361588448286057, 'recon_norm_std': 0.04249313343316317, 'target_norm': 1.6220956891775131, 'recon_cos': 0.6180667832493782, 'gradnorm/compressor.query': 0.14363935789206894, 'gradnorm/compressor.layer0': 4.980114111148318, 'gradnorm/compressor.layer1': 133.2031094208202, 'gradnorm/compressor.layer2': 26.90164694583656, 'gradnorm/compressor.layer3': 10.335326669049074, 'gradnorm/compressor.layer4': 4.152112694212364, 'gradnorm/compressor.layer5': 2.336934977085945, 'gradnorm/compressor.layer6': 1.2672919139812695, 'gradnorm/compressor.layer7': 1.5537073313960745, 'gradnorm/decoder.layer0': 1.692097934314537, 'gradnorm/decoder.layer1': 1.1174510298203815, 'gradnorm/decoder.layer2': 0.5857588177667293, 'gradnorm/max_param': 93.96939086914062, 'epoch': 5.93}

 59%|█████▉    | 11320/19090 [5:31:01<3:53:11,  1.80s/it]
 59%|█████▉    | 11321/19090 [5:31:02<3:33:14,  1.65s/it]
 59%|█████▉    | 11322/19090 [5:31:04<4:08:46,  1.92s/it]
 59%|█████▉    | 11323/19090 [5:31:06<3:44:00,  1.73s/it]
 59%|█████▉    | 11324/19090 [5:31:07<3:26:44,  1.60s/it]
 59%|█████▉    | 11325/19090 [5:31:10<4:21:37,  2.02s/it]
 59%|█████▉    | 11326/19090 [5:31:11<3:53:25,  1.80s/it]
 59%|█████▉    | 11327/19090 [5:31:13<3:33:30,  1.65s/it]
 59%|█████▉    | 11328/19090 [5:31:15<4:21:51,  2.02s/it]
 59%|█████▉    | 11329/19090 [5:31:17<3:53:14,  1.80s/it][gradnorm] step 11329: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 108.0

 59%|█████▉    | 11330/19090 [5:31:18<3:33:15,  1.65s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 155.91024780273438, 'learning_rate': 2.122261417552791e-05, 'recon_loss': 0.0015795745479408652, 'recon_static': 0.0006159707118058577, 'recon_motion': 0.0009636038346798159, 'recon_norm': 1.3072255730628968, 'recon_norm_std': 0.043434140644967555, 'target_norm': 1.6211288332939149, 'recon_cos': 0.6306687101721764, 'gradnorm/compressor.query': 0.14124949523742755, 'gradnorm/compressor.layer0': 5.167957124242876, 'gradnorm/compressor.layer1': 153.21050546347988, 'gradnorm/compressor.layer2': 25.544556230777307, 'gradnorm/compressor.layer3': 11.31697322859079, 'gradnorm/compressor.layer4': 3.6880790126553507, 'gradnorm/compressor.layer5': 2.250395422203488, 'gradnorm/compressor.layer6': 1.2335094845905439, 'gradnorm/compressor.layer7': 1.5998853135729147, 'gradnorm/decoder.layer0': 1.7642392822562796, 'gradnorm/decoder.layer1': 1.0385754813476693, 'gradnorm/decoder.layer2': 0.5292821790618679, 'gradnorm/max_param': 108.04708862304688, 'epoch': 5.94}

 59%|█████▉    | 11330/19090 [5:31:18<3:33:15,  1.65s/it]
 59%|█████▉    | 11331/19090 [5:31:21<4:17:49,  1.99s/it]
 59%|█████▉    | 11332/19090 [5:31:22<3:50:32,  1.78s/it]
 59%|█████▉    | 11333/19090 [5:31:23<3:31:24,  1.64s/it]
 59%|█████▉    | 11334/19090 [5:31:26<4:18:28,  2.00s/it]
 59%|█████▉    | 11335/19090 [5:31:28<3:50:55,  1.79s/it]
 59%|█████▉    | 11336/19090 [5:31:29<3:31:48,  1.64s/it]
 59%|█████▉    | 11337/19090 [5:31:32<4:12:53,  1.96s/it]
 59%|█████▉    | 11338/19090 [5:31:33<3:46:58,  1.76s/it]
 59%|█████▉    | 11339/19090 [5:31:34<3:28:52,  1.62s/it][gradnorm] step 11339: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 105.9

 59%|█████▉    | 11340/19090 [5:31:37<4:24:56,  2.05s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 153.08624267578125, 'learning_rate': 2.1177432159319754e-05, 'recon_loss': 0.0015440091956406832, 'recon_static': 0.0005828132707392797, 'recon_motion': 0.0009611959278117866, 'recon_norm': 1.2975565850734712, 'recon_norm_std': 0.04358960445970297, 'target_norm': 1.6191419661045074, 'recon_cos': 0.6379529744386673, 'gradnorm/compressor.query': 0.1250280706358468, 'gradnorm/compressor.layer0': 4.360990928943125, 'gradnorm/compressor.layer1': 150.09204857977986, 'gradnorm/compressor.layer2': 25.98335789876498, 'gradnorm/compressor.layer3': 13.302492777179548, 'gradnorm/compressor.layer4': 4.391537174484899, 'gradnorm/compressor.layer5': 2.615098535202052, 'gradnorm/compressor.layer6': 1.3344977132578266, 'gradnorm/compressor.layer7': 1.8174050487078344, 'gradnorm/decoder.layer0': 1.917804594454074, 'gradnorm/decoder.layer1': 1.1734367644895642, 'gradnorm/decoder.layer2': 0.6223396424874601, 'gradnorm/max_param': 105.88307189941406, 'epoch': 5.94}

 59%|█████▉    | 11340/19090 [5:31:37<4:24:56,  2.05s/it]
 59%|█████▉    | 11341/19090 [5:31:38<3:55:33,  1.82s/it]
 59%|█████▉    | 11342/19090 [5:31:40<3:34:38,  1.66s/it]
 59%|█████▉    | 11343/19090 [5:31:43<4:29:25,  2.09s/it]
 59%|█████▉    | 11344/19090 [5:31:44<3:58:19,  1.85s/it]
 59%|█████▉    | 11345/19090 [5:31:45<3:36:48,  1.68s/it]
 59%|█████▉    | 11346/19090 [5:31:48<4:23:20,  2.04s/it]
 59%|█████▉    | 11347/19090 [5:31:50<3:54:09,  1.81s/it]
 59%|█████▉    | 11348/19090 [5:31:51<3:33:46,  1.66s/it]
 59%|█████▉    | 11349/19090 [5:31:54<4:22:18,  2.03s/it]
 59%|█████▉    | 11350/19090 [5:31:55<3:53:29,  1.81s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 128.8455047607422, 'learning_rate': 2.1132262923909018e-05, 'recon_loss': 0.0015694331610575317, 'recon_static': 0.000607640715315938, 'recon_motion': 0.0009617924253689126, 'recon_norm': 1.3037850618362428, 'recon_norm_std': 0.042004691623151305, 'target_norm': 1.620132002234459, 'recon_cos': 0.6321114361286163, 'gradnorm/compressor.query': 0.15541801079939258, 'gradnorm/compressor.layer0': 4.074780258151558, 'gradnorm/compressor.layer1': 125.80394205245517, 'gradnorm/compressor.layer2': 24.650254984498257, 'gradnorm/compressor.layer3': 10.789545466878387, 'gradnorm/compressor.layer4': 4.524655087640314, 'gradnorm/compressor.layer5': 2.1019940128259296, 'gradnorm/compressor.layer6': 1.333487869287618, 'gradnorm/compressor.layer7': 1.5447576654987627, 'gradnorm/decoder.layer0': 1.8419105923478607, 'gradnorm/decoder.layer1': 1.0421290070963578, 'gradnorm/decoder.layer2': 0.5657618573101334, 'gradnorm/max_param': 88.73461151123047, 'epoch': 5.95}

 59%|█████▉    | 11350/19090 [5:31:55<3:53:29,  1.81s/it]
 59%|█████▉    | 11351/19090 [5:31:56<3:33:20,  1.65s/it]
 59%|█████▉    | 11352/19090 [5:31:59<4:28:21,  2.08s/it]
 59%|█████▉    | 11353/19090 [5:32:01<3:57:47,  1.84s/it]
 59%|█████▉    | 11354/19090 [5:32:02<3:36:16,  1.68s/it]
 59%|█████▉    | 11355/19090 [5:32:05<4:11:44,  1.95s/it]
 59%|█████▉    | 11356/19090 [5:32:06<3:46:09,  1.75s/it]
 59%|█████▉    | 11357/19090 [5:32:07<3:28:07,  1.61s/it]
 59%|█████▉    | 11358/19090 [5:32:10<4:23:09,  2.04s/it]
 60%|█████▉    | 11359/19090 [5:32:12<3:53:49,  1.81s/it][gradnorm] step 11359: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 110.6

 60%|█████▉    | 11360/19090 [5:32:13<3:33:30,  1.66s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 159.6621551513672, 'learning_rate': 2.1087106620319503e-05, 'recon_loss': 0.00155935451912228, 'recon_static': 0.0005871528876014054, 'recon_motion': 0.0009722016373416408, 'recon_norm': 1.29973024725914, 'recon_norm_std': 0.04350974885746837, 'target_norm': 1.6226764231920243, 'recon_cos': 0.6356924057006836, 'gradnorm/compressor.query': 0.1531779484776832, 'gradnorm/compressor.layer0': 4.344322331547896, 'gradnorm/compressor.layer1': 156.82811176770892, 'gradnorm/compressor.layer2': 26.480364740631405, 'gradnorm/compressor.layer3': 12.103180119048904, 'gradnorm/compressor.layer4': 4.0285166855545, 'gradnorm/compressor.layer5': 2.286213426500783, 'gradnorm/compressor.layer6': 1.3268912924486365, 'gradnorm/compressor.layer7': 1.4907250552715712, 'gradnorm/decoder.layer0': 1.8474867503061456, 'gradnorm/decoder.layer1': 1.0622043759340087, 'gradnorm/decoder.layer2': 0.6131351078145199, 'gradnorm/max_param': 110.61422729492188, 'epoch': 5.95}

 60%|█████▉    | 11360/19090 [5:32:13<3:33:30,  1.66s/it]
 60%|█████▉    | 11361/19090 [5:32:16<4:30:20,  2.10s/it]
 60%|█████▉    | 11362/19090 [5:32:17<3:58:51,  1.85s/it]
 60%|█████▉    | 11363/19090 [5:32:18<3:37:02,  1.69s/it]
 60%|█████▉    | 11364/19090 [5:32:21<4:26:06,  2.07s/it]
 60%|█████▉    | 11365/19090 [5:32:23<3:55:57,  1.83s/it]
 60%|█████▉    | 11366/19090 [5:32:24<3:34:58,  1.67s/it]
 60%|█████▉    | 11367/19090 [5:32:27<4:15:02,  1.98s/it]
 60%|█████▉    | 11368/19090 [5:32:28<3:48:25,  1.77s/it]
 60%|█████▉    | 11369/19090 [5:32:29<3:29:42,  1.63s/it][gradnorm] step 11369: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 101.0

 60%|█████▉    | 11370/19090 [5:32:32<4:05:00,  1.90s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 146.21530151367188, 'learning_rate': 2.1041963399531812e-05, 'recon_loss': 0.0015409359126351773, 'recon_static': 0.0005682599774445407, 'recon_motion': 0.000972675927914679, 'recon_norm': 1.2917907059192657, 'recon_norm_std': 0.04329215437173843, 'target_norm': 1.6215195804834366, 'recon_cos': 0.6396211475133896, 'gradnorm/compressor.query': 0.1181551874784424, 'gradnorm/compressor.layer0': 3.7301838231748436, 'gradnorm/compressor.layer1': 143.25282910600754, 'gradnorm/compressor.layer2': 26.221916944088132, 'gradnorm/compressor.layer3': 11.222043714792719, 'gradnorm/compressor.layer4': 4.005390588047236, 'gradnorm/compressor.layer5': 2.289307044184705, 'gradnorm/compressor.layer6': 1.3090910464826184, 'gradnorm/compressor.layer7': 1.4630320828219083, 'gradnorm/decoder.layer0': 1.8986735256081002, 'gradnorm/decoder.layer1': 1.0256618758597904, 'gradnorm/decoder.layer2': 0.5422601810517855, 'gradnorm/max_param': 101.04804992675781, 'epoch': 5.96}

 60%|█████▉    | 11370/19090 [5:32:32<4:05:00,  1.90s/it]
 60%|█████▉    | 11371/19090 [5:32:33<3:41:17,  1.72s/it]
 60%|█████▉    | 11372/19090 [5:32:34<3:24:41,  1.59s/it]
 60%|█████▉    | 11373/19090 [5:32:37<4:12:22,  1.96s/it]
 60%|█████▉    | 11374/19090 [5:32:39<3:46:31,  1.76s/it]
 60%|█████▉    | 11375/19090 [5:32:40<3:28:27,  1.62s/it]
 60%|█████▉    | 11376/19090 [5:32:43<4:22:26,  2.04s/it]
 60%|█████▉    | 11377/19090 [5:32:44<3:53:49,  1.82s/it]
 60%|█████▉    | 11378/19090 [5:32:45<3:33:28,  1.66s/it]
 60%|█████▉    | 11379/19090 [5:32:48<4:16:21,  1.99s/it][gradnorm] step 11379: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 121.5

 60%|█████▉    | 11380/19090 [5:32:50<3:49:12,  1.78s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 175.78292846679688, 'learning_rate': 2.0996833412482815e-05, 'recon_loss': 0.0015504624228924514, 'recon_static': 0.0005773139957454987, 'recon_motion': 0.0009731483878567815, 'recon_norm': 1.2891461968421936, 'recon_norm_std': 0.044064464140683414, 'target_norm': 1.62268023788929, 'recon_cos': 0.636831933259964, 'gradnorm/compressor.query': 0.14662748339892254, 'gradnorm/compressor.layer0': 4.958647239856937, 'gradnorm/compressor.layer1': 172.18764178760043, 'gradnorm/compressor.layer2': 31.79437183021745, 'gradnorm/compressor.layer3': 13.260584795960522, 'gradnorm/compressor.layer4': 4.747086484717104, 'gradnorm/compressor.layer5': 2.7018025421232275, 'gradnorm/compressor.layer6': 1.3173783785078736, 'gradnorm/compressor.layer7': 1.5941376308197117, 'gradnorm/decoder.layer0': 2.022116329170988, 'gradnorm/decoder.layer1': 1.0745142859622665, 'gradnorm/decoder.layer2': 0.6011207494381179, 'gradnorm/max_param': 121.48126983642578, 'epoch': 5.96}

 60%|█████▉    | 11380/19090 [5:32:50<3:49:12,  1.78s/it]
 60%|█████▉    | 11381/19090 [5:32:51<3:30:16,  1.64s/it]
 60%|█████▉    | 11382/19090 [5:32:54<4:18:00,  2.01s/it]
 60%|█████▉    | 11383/19090 [5:32:55<3:50:08,  1.79s/it]
 60%|█████▉    | 11384/19090 [5:32:56<3:31:15,  1.64s/it]
 60%|█████▉    | 11385/19090 [5:32:59<4:03:42,  1.90s/it]
 60%|█████▉    | 11386/19090 [5:33:00<3:40:19,  1.72s/it]
 60%|█████▉    | 11387/19090 [5:33:01<3:24:03,  1.59s/it]
 60%|█████▉    | 11388/19090 [5:33:04<4:13:31,  1.97s/it]
 60%|█████▉    | 11389/19090 [5:33:06<3:46:48,  1.77s/it][gradnorm] step 11389: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 110.0

 60%|█████▉    | 11390/19090 [5:33:07<3:28:24,  1.62s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 159.44430541992188, 'learning_rate': 2.095171681006509e-05, 'recon_loss': 0.0015062115737237037, 'recon_static': 0.0005321635995642282, 'recon_motion': 0.0009740479523316026, 'recon_norm': 1.2758957535028457, 'recon_norm_std': 0.043532269168645145, 'target_norm': 1.6205189287662507, 'recon_cos': 0.6463067039847374, 'gradnorm/compressor.query': 0.13604509737670084, 'gradnorm/compressor.layer0': 4.322080722858186, 'gradnorm/compressor.layer1': 155.94274164537123, 'gradnorm/compressor.layer2': 29.75894363989589, 'gradnorm/compressor.layer3': 12.857597960637236, 'gradnorm/compressor.layer4': 4.201870982917024, 'gradnorm/compressor.layer5': 2.4689594614525654, 'gradnorm/compressor.layer6': 1.46349197068014, 'gradnorm/compressor.layer7': 1.6669401977512655, 'gradnorm/decoder.layer0': 2.122286129478011, 'gradnorm/decoder.layer1': 1.0797902942972346, 'gradnorm/decoder.layer2': 0.6268007087233675, 'gradnorm/max_param': 110.00909423828125, 'epoch': 5.97}

 60%|█████▉    | 11390/19090 [5:33:07<3:28:24,  1.62s/it]
 60%|█████▉    | 11391/19090 [5:33:10<4:23:43,  2.06s/it]
 60%|█████▉    | 11392/19090 [5:33:11<3:53:54,  1.82s/it]
 60%|█████▉    | 11393/19090 [5:33:12<3:33:10,  1.66s/it]
 60%|█████▉    | 11394/19090 [5:33:15<4:20:26,  2.03s/it]
 60%|█████▉    | 11395/19090 [5:33:17<3:51:34,  1.81s/it]
 60%|█████▉    | 11396/19090 [5:33:18<3:31:34,  1.65s/it]
 60%|█████▉    | 11397/19090 [5:33:21<4:19:24,  2.02s/it]
 60%|█████▉    | 11398/19090 [5:33:22<3:51:00,  1.80s/it]
 60%|█████▉    | 11399/19090 [5:33:23<3:30:54,  1.65s/it][gradnorm] step 11399: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 113.4

 60%|█████▉    | 11400/19090 [5:33:26<4:04:13,  1.91s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 164.28131103515625, 'learning_rate': 2.0906613743126517e-05, 'recon_loss': 0.0014716353383846582, 'recon_static': 0.0005167427960259374, 'recon_motion': 0.0009548925561830401, 'recon_norm': 1.2646631836891173, 'recon_norm_std': 0.04410769734531641, 'target_norm': 1.6181331306695939, 'recon_cos': 0.6531893759965897, 'gradnorm/compressor.query': 0.13804185155462168, 'gradnorm/compressor.layer0': 4.223949026277059, 'gradnorm/compressor.layer1': 160.74629418356741, 'gradnorm/compressor.layer2': 30.14126590978772, 'gradnorm/compressor.layer3': 13.698611134764151, 'gradnorm/compressor.layer4': 4.104265989596007, 'gradnorm/compressor.layer5': 2.604948938495334, 'gradnorm/compressor.layer6': 1.5704224201656027, 'gradnorm/compressor.layer7': 1.7712583218892546, 'gradnorm/decoder.layer0': 2.0290496646223093, 'gradnorm/decoder.layer1': 1.1229388641163995, 'gradnorm/decoder.layer2': 0.5980491593290489, 'gradnorm/max_param': 113.38227081298828, 'epoch': 5.97}

 60%|█████▉    | 11400/19090 [5:33:26<4:04:13,  1.91s/it]
 60%|█████▉    | 11401/19090 [5:33:27<3:40:40,  1.72s/it]
 60%|█████▉    | 11402/19090 [5:33:28<3:23:51,  1.59s/it]
 60%|█████▉    | 11403/19090 [5:33:31<4:18:10,  2.02s/it]
 60%|█████▉    | 11404/19090 [5:33:33<3:50:51,  1.80s/it]
 60%|█████▉    | 11405/19090 [5:33:34<3:30:50,  1.65s/it]
 60%|█████▉    | 11406/19090 [5:33:37<4:07:05,  1.93s/it]
 60%|█████▉    | 11407/19090 [5:33:38<3:42:27,  1.74s/it]
 60%|█████▉    | 11408/19090 [5:33:39<3:25:13,  1.60s/it]
 60%|█████▉    | 11409/19090 [5:33:42<4:09:45,  1.95s/it][gradnorm] step 11409: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 116.6

 60%|█████▉    | 11410/19090 [5:33:43<3:44:17,  1.75s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 168.4441680908203, 'learning_rate': 2.0861524362469663e-05, 'recon_loss': 0.001500167723861523, 'recon_static': 0.0005324134508555289, 'recon_motion': 0.0009677542824647389, 'recon_norm': 1.2755942821502686, 'recon_norm_std': 0.04440945442765951, 'target_norm': 1.6205414265394211, 'recon_cos': 0.6482927158474923, 'gradnorm/compressor.query': 0.12688529329873552, 'gradnorm/compressor.layer0': 4.181241725120852, 'gradnorm/compressor.layer1': 165.2366319290308, 'gradnorm/compressor.layer2': 28.900312946037563, 'gradnorm/compressor.layer3': 13.49603721949087, 'gradnorm/compressor.layer4': 4.155832140575416, 'gradnorm/compressor.layer5': 2.5505188694554852, 'gradnorm/compressor.layer6': 1.6321146080455797, 'gradnorm/compressor.layer7': 1.8565782943847378, 'gradnorm/decoder.layer0': 1.9448908459941991, 'gradnorm/decoder.layer1': 1.1546110961579106, 'gradnorm/decoder.layer2': 0.6457279309021305, 'gradnorm/max_param': 116.55653381347656, 'epoch': 5.98}

 60%|█████▉    | 11410/19090 [5:33:43<3:44:17,  1.75s/it]
 60%|█████▉    | 11411/19090 [5:33:45<3:26:28,  1.61s/it]
 60%|█████▉    | 11412/19090 [5:33:47<4:09:39,  1.95s/it]
 60%|█████▉    | 11413/19090 [5:33:49<3:44:13,  1.75s/it]
 60%|█████▉    | 11414/19090 [5:33:50<3:26:32,  1.61s/it]
 60%|█████▉    | 11415/19090 [5:33:52<3:56:05,  1.85s/it]
 60%|█████▉    | 11416/19090 [5:33:54<3:34:37,  1.68s/it]
 60%|█████▉    | 11417/19090 [5:33:55<3:19:29,  1.56s/it]
 60%|█████▉    | 11418/19090 [5:33:58<4:16:13,  2.00s/it]
 60%|█████▉    | 11419/19090 [5:33:59<3:48:44,  1.79s/it][gradnorm] step 11419: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 102.1

 60%|█████▉    | 11420/19090 [5:34:00<3:29:43,  1.64s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 148.10614013671875, 'learning_rate': 2.0816448818851393e-05, 'recon_loss': 0.0015074805618496613, 'recon_static': 0.0005516718796570785, 'recon_motion': 0.0009558086676406674, 'recon_norm': 1.2832325756549836, 'recon_norm_std': 0.04441317832097411, 'target_norm': 1.6204445689916611, 'recon_cos': 0.6468692436814308, 'gradnorm/compressor.query': 0.15783098391126513, 'gradnorm/compressor.layer0': 5.856078825477427, 'gradnorm/compressor.layer1': 144.76275715588937, 'gradnorm/compressor.layer2': 27.758877820466058, 'gradnorm/compressor.layer3': 11.87259432869455, 'gradnorm/compressor.layer4': 4.250167394595191, 'gradnorm/compressor.layer5': 2.477502782764277, 'gradnorm/compressor.layer6': 1.3653377117292291, 'gradnorm/compressor.layer7': 1.5808096230993922, 'gradnorm/decoder.layer0': 1.8088457166842526, 'gradnorm/decoder.layer1': 1.0965399445034651, 'gradnorm/decoder.layer2': 0.5487799843957706, 'gradnorm/max_param': 102.09564208984375, 'epoch': 5.98}

 60%|█████▉    | 11420/19090 [5:34:00<3:29:43,  1.64s/it]
 60%|█████▉    | 11421/19090 [5:34:03<4:16:36,  2.01s/it]
 60%|█████▉    | 11422/19090 [5:34:05<3:52:06,  1.82s/it]
 60%|█████▉    | 11423/19090 [5:34:06<3:31:54,  1.66s/it]
 60%|█████▉    | 11424/19090 [5:34:09<4:10:41,  1.96s/it]
 60%|█████▉    | 11425/19090 [5:34:10<4:03:10,  1.90s/it]
 60%|█████▉    | 11426/19090 [5:34:12<3:39:27,  1.72s/it]
 60%|█████▉    | 11427/19090 [5:34:14<4:00:52,  1.89s/it]
 60%|█████▉    | 11428/19090 [5:34:16<3:56:30,  1.85s/it]
 60%|█████▉    | 11429/19090 [5:34:17<3:34:43,  1.68s/it][gradnorm] step 11429: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 118.3

 60%|█████▉    | 11430/19090 [5:34:19<4:02:43,  1.90s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 170.8267822265625, 'learning_rate': 2.0771387262982277e-05, 'recon_loss': 0.0015949233609717339, 'recon_static': 0.0006378990030498244, 'recon_motion': 0.0009570243797497824, 'recon_norm': 1.3186481982469558, 'recon_norm_std': 0.04230711469426751, 'target_norm': 1.6177929043769836, 'recon_cos': 0.6261286854743957, 'gradnorm/compressor.query': 0.18558885199896516, 'gradnorm/compressor.layer0': 5.57366423064151, 'gradnorm/compressor.layer1': 167.72391167294842, 'gradnorm/compressor.layer2': 28.97921534263825, 'gradnorm/compressor.layer3': 12.182001213448189, 'gradnorm/compressor.layer4': 4.0494640419566155, 'gradnorm/compressor.layer5': 2.232328745050012, 'gradnorm/compressor.layer6': 1.542468721514701, 'gradnorm/compressor.layer7': 1.5734288053452283, 'gradnorm/decoder.layer0': 1.7597832377205458, 'gradnorm/decoder.layer1': 1.1917011833971862, 'gradnorm/decoder.layer2': 0.6558841412101523, 'gradnorm/max_param': 118.27236938476562, 'epoch': 5.99}

 60%|█████▉    | 11430/19090 [5:34:19<4:02:43,  1.90s/it]
 60%|█████▉    | 11431/19090 [5:34:21<3:50:39,  1.81s/it]
 60%|█████▉    | 11432/19090 [5:34:22<3:30:50,  1.65s/it]
 60%|█████▉    | 11433/19090 [5:34:25<4:15:02,  2.00s/it]
 60%|█████▉    | 11434/19090 [5:34:26<3:50:31,  1.81s/it]
 60%|█████▉    | 11435/19090 [5:34:28<3:30:40,  1.65s/it]
 60%|█████▉    | 11436/19090 [5:34:31<4:16:16,  2.01s/it]
 60%|█████▉    | 11437/19090 [5:34:32<3:53:23,  1.83s/it]
 60%|█████▉    | 11438/19090 [5:34:33<3:34:07,  1.68s/it]
 60%|█████▉    | 11439/19090 [5:34:36<4:09:05,  1.95s/it][gradnorm] step 11439: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 108.8

 60%|█████▉    | 11440/19090 [5:34:38<4:07:41,  1.94s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 157.6542205810547, 'learning_rate': 2.0726339845526098e-05, 'recon_loss': 0.001615719514666125, 'recon_static': 0.0006487226302851923, 'recon_motion': 0.0009669968945672736, 'recon_norm': 1.327012050151825, 'recon_norm_std': 0.04241368239745498, 'target_norm': 1.6201646119356155, 'recon_cos': 0.6239869311451912, 'gradnorm/compressor.query': 0.1470459853337826, 'gradnorm/compressor.layer0': 5.0234375, 'gradnorm/compressor.layer1': 154.20378283500375, 'gradnorm/compressor.layer2': 28.923820673343798, 'gradnorm/compressor.layer3': 13.35068076466076, 'gradnorm/compressor.layer4': 4.438890266232574, 'gradnorm/compressor.layer5': 2.399770062081073, 'gradnorm/compressor.layer6': 1.6110003513688615, 'gradnorm/compressor.layer7': 1.4616051438598652, 'gradnorm/decoder.layer0': 1.976085741043903, 'gradnorm/decoder.layer1': 1.2054731071032476, 'gradnorm/decoder.layer2': 0.6403681542564713, 'gradnorm/max_param': 108.77962493896484, 'epoch': 5.99}

 60%|█████▉    | 11440/19090 [5:34:38<4:07:41,  1.94s/it]
 60%|█████▉    | 11441/19090 [5:34:39<3:42:45,  1.75s/it]
 60%|█████▉    | 11442/19090 [5:34:42<4:09:42,  1.96s/it]
 60%|█████▉    | 11443/19090 [5:34:44<4:10:25,  1.96s/it]
 60%|█████▉    | 11444/19090 [5:34:45<3:44:27,  1.76s/it]
 60%|█████▉    | 11445/19090 [5:34:47<4:03:55,  1.91s/it]
 60%|█████▉    | 11446/19090 [5:34:50<4:23:59,  2.07s/it]
 60%|█████▉    | 11447/19090 [5:34:51<3:54:00,  1.84s/it]
 60%|█████▉    | 11448/19090 [5:34:52<3:44:11,  1.76s/it]
 60%|█████▉    | 11449/19090 [5:34:55<4:25:30,  2.08s/it][gradnorm] step 11449: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 105.8

 60%|█████▉    | 11450/19090 [5:34:57<3:55:14,  1.85s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 153.71539306640625, 'learning_rate': 2.0681306717099412e-05, 'recon_loss': 0.0016512429836438968, 'recon_static': 0.0006823731135227717, 'recon_motion': 0.0009688698599347844, 'recon_norm': 1.3390118449926376, 'recon_norm_std': 0.040890207141637804, 'target_norm': 1.6196193099021912, 'recon_cos': 0.6158030614256859, 'gradnorm/compressor.query': 0.17572388242688278, 'gradnorm/compressor.layer0': 5.129840332877849, 'gradnorm/compressor.layer1': 149.97984239556993, 'gradnorm/compressor.layer2': 29.848332845115696, 'gradnorm/compressor.layer3': 13.670295667708805, 'gradnorm/compressor.layer4': 3.9526868735719267, 'gradnorm/compressor.layer5': 2.338353059521881, 'gradnorm/compressor.layer6': 1.3454310525546616, 'gradnorm/compressor.layer7': 1.513800476517856, 'gradnorm/decoder.layer0': 1.8184120834365318, 'gradnorm/decoder.layer1': 1.1962617084723415, 'gradnorm/decoder.layer2': 0.6233133445771509, 'gradnorm/max_param': 105.79467010498047, 'epoch': 6.0}

 60%|█████▉    | 11450/19090 [5:34:57<3:55:14,  1.85s/it]
 60%|█████▉    | 11451/19090 [5:34:58<3:33:49,  1.68s/it]
 60%|█████▉    | 11452/19090 [5:34:59<3:18:10,  1.56s/it]
 60%|█████▉    | 11453/19090 [5:35:00<3:07:14,  1.47s/it]
 60%|██████    | 11454/19090 [5:35:02<3:17:13,  1.55s/it]
 60%|██████    | 11455/19090 [5:35:10<7:29:25,  3.53s/it]
 60%|██████    | 11456/19090 [5:35:12<6:04:12,  2.86s/it]
 60%|██████    | 11457/19090 [5:35:15<6:19:16,  2.98s/it]
 60%|██████    | 11458/19090 [5:35:16<5:15:28,  2.48s/it]
 60%|██████    | 11459/19090 [5:35:17<4:30:17,  2.13s/it][gradnorm] step 11459: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 119.1

 60%|██████    | 11460/19090 [5:35:20<4:45:20,  2.24s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 172.26675415039062, 'learning_rate': 2.0636288028270976e-05, 'recon_loss': 0.0015671573113650084, 'recon_static': 0.0005993874379782938, 'recon_motion': 0.0009677699112216942, 'recon_norm': 1.298700252175331, 'recon_norm_std': 0.04186438219621778, 'target_norm': 1.6212554544210434, 'recon_cos': 0.6326572701334954, 'gradnorm/compressor.query': 0.169710444309484, 'gradnorm/compressor.layer0': 5.64105087743685, 'gradnorm/compressor.layer1': 168.88334430717256, 'gradnorm/compressor.layer2': 29.924014691772854, 'gradnorm/compressor.layer3': 13.86294774535451, 'gradnorm/compressor.layer4': 4.348609456518516, 'gradnorm/compressor.layer5': 2.502233270688883, 'gradnorm/compressor.layer6': 1.4275829897672252, 'gradnorm/compressor.layer7': 1.500890944054912, 'gradnorm/decoder.layer0': 1.8843473770875212, 'gradnorm/decoder.layer1': 1.1884472231350576, 'gradnorm/decoder.layer2': 0.5855478644283147, 'gradnorm/max_param': 119.09210968017578, 'epoch': 6.0}

 60%|██████    | 11460/19090 [5:35:20<4:45:20,  2.24s/it]
 60%|██████    | 11461/19090 [5:35:21<4:09:46,  1.96s/it]
 60%|██████    | 11462/19090 [5:35:23<3:44:07,  1.76s/it]
 60%|██████    | 11463/19090 [5:35:25<4:14:35,  2.00s/it]
 60%|██████    | 11464/19090 [5:35:26<3:47:37,  1.79s/it]
 60%|██████    | 11465/19090 [5:35:28<3:29:43,  1.65s/it]
 60%|██████    | 11466/19090 [5:35:30<4:09:10,  1.96s/it]
 60%|██████    | 11467/19090 [5:35:32<3:44:01,  1.76s/it]
 60%|██████    | 11468/19090 [5:35:33<3:26:22,  1.62s/it]
 60%|██████    | 11469/19090 [5:35:35<3:53:57,  1.84s/it][gradnorm] step 11469: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 105.1

 60%|██████    | 11470/19090 [5:35:37<3:33:39,  1.68s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 152.56134033203125, 'learning_rate': 2.0591283929561252e-05, 'recon_loss': 0.0015077284799190239, 'recon_static': 0.0005410513113019988, 'recon_motion': 0.0009666771584306844, 'recon_norm': 1.2827026396989822, 'recon_norm_std': 0.0423142240382731, 'target_norm': 1.6199101567268372, 'recon_cos': 0.6463056921958923, 'gradnorm/compressor.query': 0.1257356143286869, 'gradnorm/compressor.layer0': 3.8831907276170172, 'gradnorm/compressor.layer1': 148.9966704745445, 'gradnorm/compressor.layer2': 29.61312744197295, 'gradnorm/compressor.layer3': 12.294652405952888, 'gradnorm/compressor.layer4': 3.9056690241786005, 'gradnorm/compressor.layer5': 2.739637183372078, 'gradnorm/compressor.layer6': 1.3323283381581208, 'gradnorm/compressor.layer7': 1.4743133854170218, 'gradnorm/decoder.layer0': 1.8312063088266297, 'gradnorm/decoder.layer1': 1.1775710971693174, 'gradnorm/decoder.layer2': 0.5606658490288783, 'gradnorm/max_param': 105.0927963256836, 'epoch': 6.01}

 60%|██████    | 11470/19090 [5:35:37<3:33:39,  1.68s/it]
 60%|██████    | 11471/19090 [5:35:38<3:19:07,  1.57s/it]
 60%|██████    | 11472/19090 [5:35:40<3:52:16,  1.83s/it]
 60%|██████    | 11473/19090 [5:35:42<3:31:28,  1.67s/it]
 60%|██████    | 11474/19090 [5:35:43<3:17:09,  1.55s/it]
 60%|██████    | 11475/19090 [5:35:45<3:37:45,  1.72s/it]
 60%|██████    | 11476/19090 [5:35:46<3:21:12,  1.59s/it]
 60%|██████    | 11477/19090 [5:35:48<3:10:21,  1.50s/it]
 60%|██████    | 11478/19090 [5:35:51<4:01:27,  1.90s/it]
 60%|██████    | 11479/19090 [5:35:52<3:38:01,  1.72s/it][gradnorm] step 11479: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 117.4

 60%|██████    | 11480/19090 [5:35:53<3:21:32,  1.59s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 169.58590698242188, 'learning_rate': 2.054629457144195e-05, 'recon_loss': 0.001497110683703795, 'recon_static': 0.0005356040688639041, 'recon_motion': 0.0009615066112019122, 'recon_norm': 1.2759142458438872, 'recon_norm_std': 0.042710973601788285, 'target_norm': 1.6181039482355117, 'recon_cos': 0.6475193783640861, 'gradnorm/compressor.query': 0.11940086583825175, 'gradnorm/compressor.layer0': 4.165612481592804, 'gradnorm/compressor.layer1': 166.36351857453604, 'gradnorm/compressor.layer2': 29.54986599086576, 'gradnorm/compressor.layer3': 12.655785220053824, 'gradnorm/compressor.layer4': 4.0986638753067, 'gradnorm/compressor.layer5': 2.342778729412531, 'gradnorm/compressor.layer6': 1.4778198862808793, 'gradnorm/compressor.layer7': 1.5990314114880724, 'gradnorm/decoder.layer0': 1.8086843117765876, 'gradnorm/decoder.layer1': 1.1154312521192624, 'gradnorm/decoder.layer2': 0.5338890219374107, 'gradnorm/max_param': 117.3623275756836, 'epoch': 6.01}

 60%|██████    | 11480/19090 [5:35:53<3:21:32,  1.59s/it]
 60%|██████    | 11481/19090 [5:35:55<3:44:41,  1.77s/it]
 60%|██████    | 11482/19090 [5:35:57<3:26:11,  1.63s/it]
 60%|██████    | 11483/19090 [5:35:58<3:13:29,  1.53s/it]
 60%|██████    | 11484/19090 [5:36:00<3:21:16,  1.59s/it]
 60%|██████    | 11485/19090 [5:36:01<3:09:54,  1.50s/it]
 60%|██████    | 11486/19090 [5:36:02<3:01:57,  1.44s/it]
 60%|██████    | 11487/19090 [5:36:05<3:40:58,  1.74s/it]
 60%|██████    | 11488/19090 [5:36:06<3:23:55,  1.61s/it]
 60%|██████    | 11489/19090 [5:36:07<3:11:30,  1.51s/it][gradnorm] step 11489: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 116.2

 60%|██████    | 11490/19090 [5:36:10<3:46:40,  1.79s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 168.50701904296875, 'learning_rate': 2.050132010433548e-05, 'recon_loss': 0.0014833995868684725, 'recon_static': 0.0005300816083035897, 'recon_motion': 0.0009533179763820954, 'recon_norm': 1.2710002690553666, 'recon_norm_std': 0.04408654337748885, 'target_norm': 1.6172084182500839, 'recon_cos': 0.6504182577133178, 'gradnorm/compressor.query': 0.13555770797754482, 'gradnorm/compressor.layer0': 4.672774314441173, 'gradnorm/compressor.layer1': 164.72782950218217, 'gradnorm/compressor.layer2': 31.816419438686648, 'gradnorm/compressor.layer3': 13.510256791572052, 'gradnorm/compressor.layer4': 4.928527598847605, 'gradnorm/compressor.layer5': 2.776023153752666, 'gradnorm/compressor.layer6': 1.5471077946130534, 'gradnorm/compressor.layer7': 1.759746995969133, 'gradnorm/decoder.layer0': 1.8522101448622375, 'gradnorm/decoder.layer1': 1.1926466691291888, 'gradnorm/decoder.layer2': 0.5915240171785228, 'gradnorm/max_param': 116.21163177490234, 'epoch': 6.02}

 60%|██████    | 11490/19090 [5:36:10<3:46:40,  1.79s/it]
 60%|██████    | 11491/19090 [5:36:11<3:27:32,  1.64s/it]
 60%|██████    | 11492/19090 [5:36:12<3:14:26,  1.54s/it]
 60%|██████    | 11493/19090 [5:36:15<3:51:58,  1.83s/it]
 60%|██████    | 11494/19090 [5:36:16<3:32:18,  1.68s/it]
 60%|██████    | 11495/19090 [5:36:17<3:18:19,  1.57s/it]
 60%|██████    | 11496/19090 [5:36:20<3:47:27,  1.80s/it]
 60%|██████    | 11497/19090 [5:36:21<3:28:44,  1.65s/it]
 60%|██████    | 11498/19090 [5:36:22<3:15:53,  1.55s/it]
 60%|██████    | 11499/19090 [5:36:25<3:57:43,  1.88s/it][gradnorm] step 11499: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 118.2

 60%|██████    | 11500/19090 [5:36:26<3:36:13,  1.71s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 171.2158660888672, 'learning_rate': 2.0456360678614456e-05, 'recon_loss': 0.0014922640140866861, 'recon_static': 0.0005322638819052373, 'recon_motion': 0.0009600001460057684, 'recon_norm': 1.2728784680366516, 'recon_norm_std': 0.04370158342644572, 'target_norm': 1.6199478209018707, 'recon_cos': 0.6494475454092026, 'gradnorm/compressor.query': 0.17224712618505317, 'gradnorm/compressor.layer0': 4.979947123419161, 'gradnorm/compressor.layer1': 167.52203096719845, 'gradnorm/compressor.layer2': 31.406766942263015, 'gradnorm/compressor.layer3': 14.226105583683998, 'gradnorm/compressor.layer4': 4.576305874251842, 'gradnorm/compressor.layer5': 2.339437357106864, 'gradnorm/compressor.layer6': 1.442434879466052, 'gradnorm/compressor.layer7': 1.8554415088460814, 'gradnorm/decoder.layer0': 1.9485406403212746, 'gradnorm/decoder.layer1': 1.2351748132728286, 'gradnorm/decoder.layer2': 0.6200766482581344, 'gradnorm/max_param': 118.1803970336914, 'epoch': 6.02}

 60%|██████    | 11500/19090 [5:36:26<3:36:13,  1.71s/it]
 60%|██████    | 11501/19090 [5:36:31<5:14:07,  2.48s/it]
 60%|██████    | 11502/19090 [5:36:32<4:28:43,  2.12s/it]
 60%|██████    | 11503/19090 [5:36:33<3:57:02,  1.87s/it]
 60%|██████    | 11504/19090 [5:36:35<3:34:37,  1.70s/it]
 60%|██████    | 11505/19090 [5:36:36<3:19:08,  1.58s/it]
 60%|██████    | 11506/19090 [5:36:37<3:08:06,  1.49s/it]
 60%|██████    | 11507/19090 [5:36:38<3:00:26,  1.43s/it]
 60%|██████    | 11508/19090 [5:36:40<2:55:00,  1.38s/it]
 60%|██████    | 11509/19090 [5:36:41<2:51:33,  1.36s/it][gradnorm] step 11509: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 140.4

 60%|██████    | 11510/19090 [5:36:42<2:49:23,  1.34s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 201.8330841064453, 'learning_rate': 2.041141644460122e-05, 'recon_loss': 0.0015508708747802302, 'recon_static': 0.0005898309405893088, 'recon_motion': 0.0009610399487428367, 'recon_norm': 1.3006627768278123, 'recon_norm_std': 0.042832032218575476, 'target_norm': 1.6198245763778687, 'recon_cos': 0.6366176202893257, 'gradnorm/compressor.query': 0.19804697807956398, 'gradnorm/compressor.layer0': 6.993957636446631, 'gradnorm/compressor.layer1': 199.1044991016024, 'gradnorm/compressor.layer2': 29.114320819105714, 'gradnorm/compressor.layer3': 12.755057136051775, 'gradnorm/compressor.layer4': 4.523646638084147, 'gradnorm/compressor.layer5': 2.01865886092766, 'gradnorm/compressor.layer6': 1.348685581906837, 'gradnorm/compressor.layer7': 1.6768564896160847, 'gradnorm/decoder.layer0': 1.9958322730755096, 'gradnorm/decoder.layer1': 1.1176810608185428, 'gradnorm/decoder.layer2': 0.5497174371631363, 'gradnorm/max_param': 140.42630004882812, 'epoch': 6.03}

 60%|██████    | 11510/19090 [5:36:42<2:49:23,  1.34s/it]
 60%|██████    | 11511/19090 [5:36:44<2:47:25,  1.33s/it]
 60%|██████    | 11512/19090 [5:36:45<2:45:53,  1.31s/it]
 60%|██████    | 11513/19090 [5:36:46<2:45:01,  1.31s/it]
 60%|██████    | 11514/19090 [5:36:47<2:44:29,  1.30s/it]
 60%|██████    | 11515/19090 [5:36:49<2:44:06,  1.30s/it]
 60%|██████    | 11516/19090 [5:36:50<3:00:40,  1.43s/it]
 60%|██████    | 11517/19090 [5:36:52<3:05:03,  1.47s/it]
 60%|██████    | 11518/19090 [5:36:53<2:58:24,  1.41s/it]
 60%|██████    | 11519/19090 [5:36:56<3:31:37,  1.68s/it][gradnorm] step 11519: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 125.3

 60%|██████    | 11520/19090 [5:36:57<3:34:14,  1.70s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 180.45057678222656, 'learning_rate': 2.036648755256731e-05, 'recon_loss': 0.0015065048966789619, 'recon_static': 0.0005352474909159355, 'recon_motion': 0.0009712574043078348, 'recon_norm': 1.2725845575332642, 'recon_norm_std': 0.0435790304094553, 'target_norm': 1.6224047869443894, 'recon_cos': 0.6468932285904885, 'gradnorm/compressor.query': 0.16213138270075084, 'gradnorm/compressor.layer0': 4.5481353795713035, 'gradnorm/compressor.layer1': 177.6167529753317, 'gradnorm/compressor.layer2': 28.541602037639723, 'gradnorm/compressor.layer3': 12.015154488308907, 'gradnorm/compressor.layer4': 4.62126462464086, 'gradnorm/compressor.layer5': 2.086027450622296, 'gradnorm/compressor.layer6': 1.3672941329788888, 'gradnorm/compressor.layer7': 1.4928441225710887, 'gradnorm/decoder.layer0': 1.9483914808086678, 'gradnorm/decoder.layer1': 1.0487672585668886, 'gradnorm/decoder.layer2': 0.5441564838319322, 'gradnorm/max_param': 125.25464630126953, 'epoch': 6.03}

 60%|██████    | 11520/19090 [5:36:57<3:34:14,  1.70s/it]
 60%|██████    | 11521/19090 [5:36:59<3:18:46,  1.58s/it]
 60%|██████    | 11522/19090 [5:37:01<3:43:20,  1.77s/it]
 60%|██████    | 11523/19090 [5:37:02<3:31:35,  1.68s/it]
 60%|██████    | 11524/19090 [5:37:04<3:16:51,  1.56s/it]
 60%|██████    | 11525/19090 [5:37:06<3:41:49,  1.76s/it]
 60%|██████    | 11526/19090 [5:37:07<3:37:42,  1.73s/it]
 60%|██████    | 11527/19090 [5:37:09<3:21:22,  1.60s/it]
 60%|██████    | 11528/19090 [5:37:11<3:39:51,  1.74s/it]
 60%|██████    | 11529/19090 [5:37:13<3:41:33,  1.76s/it][gradnorm] step 11529: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 135.1

 60%|██████    | 11530/19090 [5:37:14<3:24:02,  1.62s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 195.3773651123047, 'learning_rate': 2.0321574152732957e-05, 'recon_loss': 0.0014806204388150945, 'recon_static': 0.0005244368316198233, 'recon_motion': 0.000956183641392272, 'recon_norm': 1.2708914786577226, 'recon_norm_std': 0.043442994076758626, 'target_norm': 1.619715052843094, 'recon_cos': 0.6523325920104981, 'gradnorm/compressor.query': 0.1434160003776393, 'gradnorm/compressor.layer0': 5.348038229943345, 'gradnorm/compressor.layer1': 191.58950666072502, 'gradnorm/compressor.layer2': 34.77858080222923, 'gradnorm/compressor.layer3': 13.949515735602168, 'gradnorm/compressor.layer4': 4.046748214561977, 'gradnorm/compressor.layer5': 2.4704717113092785, 'gradnorm/compressor.layer6': 1.3786556026898984, 'gradnorm/compressor.layer7': 1.6180832450727711, 'gradnorm/decoder.layer0': 2.067763000329343, 'gradnorm/decoder.layer1': 1.168950207539876, 'gradnorm/decoder.layer2': 0.606886772135871, 'gradnorm/max_param': 135.14337158203125, 'epoch': 6.04}

 60%|██████    | 11530/19090 [5:37:14<3:24:02,  1.62s/it]
 60%|██████    | 11531/19090 [5:37:16<3:31:12,  1.68s/it]
 60%|██████    | 11532/19090 [5:37:17<3:32:19,  1.69s/it]
 60%|██████    | 11533/19090 [5:37:19<3:21:38,  1.60s/it]
 60%|██████    | 11534/19090 [5:37:20<3:22:45,  1.61s/it]
 60%|██████    | 11535/19090 [5:37:23<3:39:16,  1.74s/it]
 60%|██████    | 11536/19090 [5:37:24<3:22:10,  1.61s/it]
 60%|██████    | 11537/19090 [5:37:26<3:30:32,  1.67s/it]
 60%|██████    | 11538/19090 [5:37:28<3:41:40,  1.76s/it]
 60%|██████    | 11539/19090 [5:37:29<3:23:49,  1.62s/it][gradnorm] step 11539: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 124.1

 60%|██████    | 11540/19090 [5:37:31<3:31:15,  1.68s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 178.93099975585938, 'learning_rate': 2.0276676395266613e-05, 'recon_loss': 0.0015056929230922833, 'recon_static': 0.0005445498449262231, 'recon_motion': 0.0009611430446966551, 'recon_norm': 1.2781246781349183, 'recon_norm_std': 0.044664569292217494, 'target_norm': 1.6200086325407028, 'recon_cos': 0.646402756869793, 'gradnorm/compressor.query': 0.12376727284674617, 'gradnorm/compressor.layer0': 4.242332048570355, 'gradnorm/compressor.layer1': 175.87703223772567, 'gradnorm/compressor.layer2': 29.91716775146891, 'gradnorm/compressor.layer3': 11.684342717760968, 'gradnorm/compressor.layer4': 4.05372634697294, 'gradnorm/compressor.layer5': 2.6387688246958594, 'gradnorm/compressor.layer6': 1.5525266112474825, 'gradnorm/compressor.layer7': 1.6171559501187245, 'gradnorm/decoder.layer0': 1.9464296250946673, 'gradnorm/decoder.layer1': 1.2001452517021036, 'gradnorm/decoder.layer2': 0.5893659108776808, 'gradnorm/max_param': 124.08936309814453, 'epoch': 6.05}

 60%|██████    | 11540/19090 [5:37:31<3:31:15,  1.68s/it]
 60%|██████    | 11541/19090 [5:37:33<3:46:18,  1.80s/it]
 60%|██████    | 11542/19090 [5:37:34<3:27:06,  1.65s/it]
 60%|██████    | 11543/19090 [5:37:36<3:35:05,  1.71s/it]
 60%|██████    | 11544/19090 [5:37:38<3:52:12,  1.85s/it]
 60%|██████    | 11545/19090 [5:37:39<3:31:12,  1.68s/it]
 60%|██████    | 11546/19090 [5:37:41<3:17:18,  1.57s/it]
 60%|██████    | 11547/19090 [5:37:43<3:41:27,  1.76s/it]
 60%|██████    | 11548/19090 [5:37:44<3:28:54,  1.66s/it]
 60%|██████    | 11549/19090 [5:37:46<3:15:13,  1.55s/it][gradnorm] step 11549: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 106.5

 61%|██████    | 11550/19090 [5:37:48<3:51:11,  1.84s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 154.0299530029297, 'learning_rate': 2.023179443028443e-05, 'recon_loss': 0.0015283058251952752, 'recon_static': 0.00055090998794185, 'recon_motion': 0.000977395813970361, 'recon_norm': 1.2811986714601518, 'recon_norm_std': 0.042497513722628354, 'target_norm': 1.6229195922613144, 'recon_cos': 0.6422478437423706, 'gradnorm/compressor.query': 0.15690304067665234, 'gradnorm/compressor.layer0': 4.638870061555728, 'gradnorm/compressor.layer1': 151.04717159508482, 'gradnorm/compressor.layer2': 27.152280528357558, 'gradnorm/compressor.layer3': 11.142892914319871, 'gradnorm/compressor.layer4': 3.466645160632794, 'gradnorm/compressor.layer5': 2.4341642589703016, 'gradnorm/compressor.layer6': 1.4449243797657985, 'gradnorm/compressor.layer7': 1.4120239822928395, 'gradnorm/decoder.layer0': 1.889392167094455, 'gradnorm/decoder.layer1': 1.0442502724517255, 'gradnorm/decoder.layer2': 0.5823648763866912, 'gradnorm/max_param': 106.54051971435547, 'epoch': 6.05}

 61%|██████    | 11550/19090 [5:37:48<3:51:11,  1.84s/it]
 61%|██████    | 11551/19090 [5:37:49<3:30:36,  1.68s/it]
 61%|██████    | 11552/19090 [5:37:51<3:16:04,  1.56s/it]
 61%|██████    | 11553/19090 [5:37:53<3:24:11,  1.63s/it]
 61%|██████    | 11554/19090 [5:37:54<3:21:03,  1.60s/it]
 61%|██████    | 11555/19090 [5:37:56<3:35:35,  1.72s/it]
 61%|██████    | 11556/19090 [5:37:58<3:30:25,  1.68s/it]
 61%|██████    | 11557/19090 [5:38:00<3:37:10,  1.73s/it]
 61%|██████    | 11558/19090 [5:38:01<3:40:41,  1.76s/it]
 61%|██████    | 11559/19090 [5:38:03<3:37:11,  1.73s/it][gradnorm] step 11559: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 112.7

 61%|██████    | 11560/19090 [5:38:05<3:43:44,  1.78s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 162.8057098388672, 'learning_rate': 2.018692840784973e-05, 'recon_loss': 0.0016016563138691708, 'recon_static': 0.0006390195980202406, 'recon_motion': 0.0009626367886085063, 'recon_norm': 1.3193682610988617, 'recon_norm_std': 0.04304173970595002, 'target_norm': 1.619766280055046, 'recon_cos': 0.6253057077527047, 'gradnorm/compressor.query': 0.13920318875111704, 'gradnorm/compressor.layer0': 5.061463049665087, 'gradnorm/compressor.layer1': 159.8012009595829, 'gradnorm/compressor.layer2': 27.349527291461662, 'gradnorm/compressor.layer3': 12.710285641321535, 'gradnorm/compressor.layer4': 4.139662835954849, 'gradnorm/compressor.layer5': 2.6786213970064856, 'gradnorm/compressor.layer6': 1.4520297280938537, 'gradnorm/compressor.layer7': 1.6624549658849652, 'gradnorm/decoder.layer0': 1.8127668118584406, 'gradnorm/decoder.layer1': 1.1128784521795547, 'gradnorm/decoder.layer2': 0.6155024111148144, 'gradnorm/max_param': 112.72832489013672, 'epoch': 6.06}

 61%|██████    | 11560/19090 [5:38:05<3:43:44,  1.78s/it]
 61%|██████    | 11561/19090 [5:38:07<3:41:19,  1.76s/it]
 61%|██████    | 11562/19090 [5:38:08<3:37:49,  1.74s/it]
 61%|██████    | 11563/19090 [5:38:10<3:42:56,  1.78s/it]
 61%|██████    | 11564/19090 [5:38:12<3:34:29,  1.71s/it]
 61%|██████    | 11565/19090 [5:38:14<3:38:35,  1.74s/it]
 61%|██████    | 11566/19090 [5:38:15<3:38:07,  1.74s/it]
 61%|██████    | 11567/19090 [5:38:17<3:39:37,  1.75s/it]
 61%|██████    | 11568/19090 [5:38:19<3:32:39,  1.70s/it]
 61%|██████    | 11569/19090 [5:38:21<3:41:07,  1.76s/it][gradnorm] step 11569: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 101.4

 61%|██████    | 11570/19090 [5:38:22<3:32:50,  1.70s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 147.13442993164062, 'learning_rate': 2.014207847797256e-05, 'recon_loss': 0.0015658728196285665, 'recon_static': 0.0006066860194550827, 'recon_motion': 0.0009591867405106313, 'recon_norm': 1.3036222726106643, 'recon_norm_std': 0.04349236413836479, 'target_norm': 1.6184569269418716, 'recon_cos': 0.632179993391037, 'gradnorm/compressor.query': 0.14906275344323258, 'gradnorm/compressor.layer0': 4.731092668585674, 'gradnorm/compressor.layer1': 143.7263159837126, 'gradnorm/compressor.layer2': 27.670701556147343, 'gradnorm/compressor.layer3': 12.947484093844317, 'gradnorm/compressor.layer4': 4.255070522633895, 'gradnorm/compressor.layer5': 2.638102393036521, 'gradnorm/compressor.layer6': 1.5758031250660394, 'gradnorm/compressor.layer7': 1.652164963593239, 'gradnorm/decoder.layer0': 1.9227174602686934, 'gradnorm/decoder.layer1': 1.1219426573007452, 'gradnorm/decoder.layer2': 0.5900426112563574, 'gradnorm/max_param': 101.38497161865234, 'epoch': 6.06}

 61%|██████    | 11570/19090 [5:38:22<3:32:50,  1.70s/it]
 61%|██████    | 11571/19090 [5:38:24<3:43:39,  1.78s/it]
 61%|██████    | 11572/19090 [5:38:25<3:24:55,  1.64s/it]
 61%|██████    | 11573/19090 [5:38:27<3:37:10,  1.73s/it]
 61%|██████    | 11574/19090 [5:38:29<3:29:07,  1.67s/it]
 61%|██████    | 11575/19090 [5:38:30<3:17:07,  1.57s/it]
 61%|██████    | 11576/19090 [5:38:32<3:37:47,  1.74s/it]
 61%|██████    | 11577/19090 [5:38:34<3:46:14,  1.81s/it]
 61%|██████    | 11578/19090 [5:38:36<3:26:45,  1.65s/it]
 61%|██████    | 11579/19090 [5:38:37<3:36:57,  1.73s/it][gradnorm] step 11579: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 106.3

 61%|██████    | 11580/19090 [5:38:39<3:42:32,  1.78s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 153.57864379882812, 'learning_rate': 2.009724479060916e-05, 'recon_loss': 0.0016630612051812931, 'recon_static': 0.000696280506963376, 'recon_motion': 0.0009667806662037037, 'recon_norm': 1.3486420452594756, 'recon_norm_std': 0.0411708720959723, 'target_norm': 1.6206068247556686, 'recon_cos': 0.6142987489700318, 'gradnorm/compressor.query': 0.1245187981780777, 'gradnorm/compressor.layer0': 4.453340652330366, 'gradnorm/compressor.layer1': 150.6911356363572, 'gradnorm/compressor.layer2': 25.845945105914947, 'gradnorm/compressor.layer3': 12.432586772960304, 'gradnorm/compressor.layer4': 4.6245426261126, 'gradnorm/compressor.layer5': 2.342307498802263, 'gradnorm/compressor.layer6': 1.5241974329081003, 'gradnorm/compressor.layer7': 1.5842508451734338, 'gradnorm/decoder.layer0': 1.7714441330617339, 'gradnorm/decoder.layer1': 1.0131158792319328, 'gradnorm/decoder.layer2': 0.5311709794010153, 'gradnorm/max_param': 106.30011749267578, 'epoch': 6.07}

 61%|██████    | 11580/19090 [5:38:39<3:42:32,  1.78s/it]
 61%|██████    | 11581/19090 [5:38:41<3:24:20,  1.63s/it]
 61%|██████    | 11582/19090 [5:38:42<3:31:41,  1.69s/it]
 61%|██████    | 11583/19090 [5:38:45<3:54:08,  1.87s/it]
 61%|██████    | 11584/19090 [5:38:46<3:32:08,  1.70s/it]
 61%|██████    | 11585/19090 [5:38:48<3:28:05,  1.66s/it]
 61%|██████    | 11586/19090 [5:38:50<3:51:44,  1.85s/it]
 61%|██████    | 11587/19090 [5:38:51<3:30:27,  1.68s/it]
 61%|██████    | 11588/19090 [5:38:53<3:32:59,  1.70s/it]
 61%|██████    | 11589/19090 [5:38:55<3:52:33,  1.86s/it][gradnorm] step 11589: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 137.4

 61%|██████    | 11590/19090 [5:38:57<3:31:07,  1.69s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 198.457275390625, 'learning_rate': 2.005242749566143e-05, 'recon_loss': 0.0015486184740439058, 'recon_static': 0.0005842722253873944, 'recon_motion': 0.0009643462530220859, 'recon_norm': 1.2962578326463698, 'recon_norm_std': 0.04374632341787219, 'target_norm': 1.6205121725797653, 'recon_cos': 0.6367482274770737, 'gradnorm/compressor.query': 0.15930409630706957, 'gradnorm/compressor.layer0': 5.364521191678414, 'gradnorm/compressor.layer1': 194.77516084257252, 'gradnorm/compressor.layer2': 34.16171938426201, 'gradnorm/compressor.layer3': 14.340966846001876, 'gradnorm/compressor.layer4': 5.3777710624701935, 'gradnorm/compressor.layer5': 2.5757128904583846, 'gradnorm/compressor.layer6': 1.4788826772803143, 'gradnorm/compressor.layer7': 1.8203421659876429, 'gradnorm/decoder.layer0': 1.9169743401526878, 'gradnorm/decoder.layer1': 1.141124929320498, 'gradnorm/decoder.layer2': 0.6289342080293295, 'gradnorm/max_param': 137.3702850341797, 'epoch': 6.07}

 61%|██████    | 11590/19090 [5:38:57<3:31:07,  1.69s/it]
 61%|██████    | 11591/19090 [5:38:59<3:47:25,  1.82s/it]
 61%|██████    | 11592/19090 [5:39:01<3:52:13,  1.86s/it]
 61%|██████    | 11593/19090 [5:39:02<3:30:42,  1.69s/it]
 61%|██████    | 11594/19090 [5:39:04<3:37:11,  1.74s/it]
 61%|██████    | 11595/19090 [5:39:06<3:44:22,  1.80s/it]
 61%|██████    | 11596/19090 [5:39:07<3:25:16,  1.64s/it]
 61%|██████    | 11597/19090 [5:39:08<3:17:16,  1.58s/it]
 61%|██████    | 11598/19090 [5:39:11<3:55:54,  1.89s/it]
 61%|██████    | 11599/19090 [5:39:12<3:33:16,  1.71s/it][gradnorm] step 11599: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 147.6

 61%|██████    | 11600/19090 [5:39:14<3:23:08,  1.63s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 212.25657653808594, 'learning_rate': 2.000762674297651e-05, 'recon_loss': 0.0015853502380196005, 'recon_static': 0.0006265734293265269, 'recon_motion': 0.000958776829065755, 'recon_norm': 1.313717457652092, 'recon_norm_std': 0.04293477339670062, 'target_norm': 1.619607549905777, 'recon_cos': 0.6294332683086395, 'gradnorm/compressor.query': 0.14615392924438655, 'gradnorm/compressor.layer0': 4.456278865340857, 'gradnorm/compressor.layer1': 209.26822837270353, 'gradnorm/compressor.layer2': 31.867235382410705, 'gradnorm/compressor.layer3': 13.907533187812898, 'gradnorm/compressor.layer4': 3.842219598549964, 'gradnorm/compressor.layer5': 2.25221217866421, 'gradnorm/compressor.layer6': 1.4642650223974656, 'gradnorm/compressor.layer7': 1.8001472863231678, 'gradnorm/decoder.layer0': 1.9926032735751558, 'gradnorm/decoder.layer1': 1.1312384884074707, 'gradnorm/decoder.layer2': 0.6191771824389991, 'gradnorm/max_param': 147.59034729003906, 'epoch': 6.08}

 61%|██████    | 11600/19090 [5:39:14<3:23:08,  1.63s/it]
 61%|██████    | 11601/19090 [5:39:16<3:58:28,  1.91s/it]
 61%|██████    | 11602/19090 [5:39:18<3:34:49,  1.72s/it]
 61%|██████    | 11603/19090 [5:39:19<3:18:36,  1.59s/it]
 61%|██████    | 11604/19090 [5:39:22<4:09:15,  2.00s/it]
 61%|██████    | 11605/19090 [5:39:23<3:42:47,  1.79s/it]
 61%|██████    | 11606/19090 [5:39:24<3:24:05,  1.64s/it]
 61%|██████    | 11607/19090 [5:39:27<3:58:28,  1.91s/it]
 61%|██████    | 11608/19090 [5:39:28<3:35:04,  1.72s/it]
 61%|██████    | 11609/19090 [5:39:30<3:20:17,  1.61s/it][gradnorm] step 11609: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 106.1

 61%|██████    | 11610/19090 [5:39:32<3:58:15,  1.91s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 153.3333282470703, 'learning_rate': 1.99628426823462e-05, 'recon_loss': 0.0015682847413700074, 'recon_static': 0.0006039904124918394, 'recon_motion': 0.0009642943245125934, 'recon_norm': 1.3027281492948533, 'recon_norm_std': 0.0428960332646966, 'target_norm': 1.6213415443897248, 'recon_cos': 0.6330719724297523, 'gradnorm/compressor.query': 0.12792238549512347, 'gradnorm/compressor.layer0': 4.074593019347507, 'gradnorm/compressor.layer1': 150.30997918884162, 'gradnorm/compressor.layer2': 27.116916278174035, 'gradnorm/compressor.layer3': 11.527677980371443, 'gradnorm/compressor.layer4': 4.017480799896754, 'gradnorm/compressor.layer5': 2.40299698546971, 'gradnorm/compressor.layer6': 1.5483828815023408, 'gradnorm/compressor.layer7': 1.6925642529325662, 'gradnorm/decoder.layer0': 2.05869126496089, 'gradnorm/decoder.layer1': 1.2024251215059816, 'gradnorm/decoder.layer2': 0.5707223804105461, 'gradnorm/max_param': 106.05146026611328, 'epoch': 6.08}

 61%|██████    | 11610/19090 [5:39:32<3:58:15,  1.91s/it]
 61%|██████    | 11611/19090 [5:39:33<3:34:59,  1.72s/it]
 61%|██████    | 11612/19090 [5:39:35<3:29:42,  1.68s/it]
 61%|██████    | 11613/19090 [5:39:37<3:51:38,  1.86s/it]
 61%|██████    | 11614/19090 [5:39:39<3:30:15,  1.69s/it]
 61%|██████    | 11615/19090 [5:39:40<3:36:05,  1.73s/it]
 61%|██████    | 11616/19090 [5:39:43<4:05:19,  1.97s/it]
 61%|██████    | 11617/19090 [5:39:44<3:39:59,  1.77s/it]
 61%|██████    | 11618/19090 [5:39:46<3:23:39,  1.64s/it]
 61%|██████    | 11619/19090 [5:39:49<4:14:08,  2.04s/it][gradnorm] step 11619: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 100.2

 61%|██████    | 11620/19090 [5:39:50<3:46:09,  1.82s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 144.82614135742188, 'learning_rate': 1.991807546350648e-05, 'recon_loss': 0.0015554278972558678, 'recon_static': 0.00059856678271899, 'recon_motion': 0.0009568611450958997, 'recon_norm': 1.305675318837166, 'recon_norm_std': 0.0431548117659986, 'target_norm': 1.6183081179857255, 'recon_cos': 0.6352265357971192, 'gradnorm/compressor.query': 0.15543995930912277, 'gradnorm/compressor.layer0': 5.542400091679156, 'gradnorm/compressor.layer1': 142.08195945694513, 'gradnorm/compressor.layer2': 24.65639852616862, 'gradnorm/compressor.layer3': 10.991779029701833, 'gradnorm/compressor.layer4': 3.8029754985428306, 'gradnorm/compressor.layer5': 2.284198162367836, 'gradnorm/compressor.layer6': 1.1971679448517414, 'gradnorm/compressor.layer7': 1.4335219861604138, 'gradnorm/decoder.layer0': 1.8060380250300974, 'gradnorm/decoder.layer1': 1.0581965528293225, 'gradnorm/decoder.layer2': 0.5262007866090544, 'gradnorm/max_param': 100.21988677978516, 'epoch': 6.09}

 61%|██████    | 11620/19090 [5:39:50<3:46:09,  1.82s/it]
 61%|██████    | 11621/19090 [5:39:51<3:30:10,  1.69s/it]
 61%|██████    | 11622/19090 [5:39:53<3:42:10,  1.79s/it]
 61%|██████    | 11623/19090 [5:39:55<3:24:13,  1.64s/it]
 61%|██████    | 11624/19090 [5:39:56<3:32:16,  1.71s/it]
 61%|██████    | 11625/19090 [5:39:59<3:51:59,  1.86s/it]
 61%|██████    | 11626/19090 [5:40:00<3:31:08,  1.70s/it]
 61%|██████    | 11627/19090 [5:40:02<3:41:28,  1.78s/it]
 61%|██████    | 11628/19090 [5:40:04<3:49:03,  1.84s/it]
 61%|██████    | 11629/19090 [5:40:05<3:28:42,  1.68s/it][gradnorm] step 11629: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 124.8

 61%|██████    | 11630/19090 [5:40:07<3:35:55,  1.74s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 179.78111267089844, 'learning_rate': 1.9873325236137068e-05, 'recon_loss': 0.0015566386748105288, 'recon_static': 0.0005848064000019803, 'recon_motion': 0.0009718323213746771, 'recon_norm': 1.291768842935562, 'recon_norm_std': 0.043438476882874964, 'target_norm': 1.621392846107483, 'recon_cos': 0.6345738276839257, 'gradnorm/compressor.query': 0.13430707428989144, 'gradnorm/compressor.layer0': 4.266376359321794, 'gradnorm/compressor.layer1': 176.92125057847065, 'gradnorm/compressor.layer2': 28.94053182517935, 'gradnorm/compressor.layer3': 11.512822838977401, 'gradnorm/compressor.layer4': 3.9660855457969624, 'gradnorm/compressor.layer5': 2.5171155600801987, 'gradnorm/compressor.layer6': 1.5187820117229796, 'gradnorm/compressor.layer7': 1.5589200780078782, 'gradnorm/decoder.layer0': 1.889275628901797, 'gradnorm/decoder.layer1': 1.0535374906121595, 'gradnorm/decoder.layer2': 0.5654432067836102, 'gradnorm/max_param': 124.79823303222656, 'epoch': 6.09}

 61%|██████    | 11630/19090 [5:40:07<3:35:55,  1.74s/it]
 61%|██████    | 11631/19090 [5:40:09<3:51:37,  1.86s/it]
 61%|██████    | 11632/19090 [5:40:11<3:30:40,  1.69s/it]
 61%|██████    | 11633/19090 [5:40:12<3:37:20,  1.75s/it]
 61%|██████    | 11634/19090 [5:40:15<3:53:33,  1.88s/it]
 61%|██████    | 11635/19090 [5:40:16<3:32:21,  1.71s/it]
 61%|██████    | 11636/19090 [5:40:18<3:28:00,  1.67s/it]
 61%|██████    | 11637/19090 [5:40:20<4:00:01,  1.93s/it]
 61%|██████    | 11638/19090 [5:40:21<3:36:21,  1.74s/it]
 61%|██████    | 11639/19090 [5:40:23<3:31:36,  1.70s/it][gradnorm] step 11639: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 140.4

 61%|██████    | 11640/19090 [5:40:25<3:50:42,  1.86s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 202.83084106445312, 'learning_rate': 1.9828592149860813e-05, 'recon_loss': 0.0015221865207422525, 'recon_static': 0.0005537688062759117, 'recon_motion': 0.0009684177493909374, 'recon_norm': 1.2864223778247834, 'recon_norm_std': 0.0432303705252707, 'target_norm': 1.6220233500003816, 'recon_cos': 0.644038838148117, 'gradnorm/compressor.query': 0.17441286362153738, 'gradnorm/compressor.layer0': 5.336669692912883, 'gradnorm/compressor.layer1': 199.0237697801446, 'gradnorm/compressor.layer2': 35.42482308805949, 'gradnorm/compressor.layer3': 14.305170172939496, 'gradnorm/compressor.layer4': 4.884740049225828, 'gradnorm/compressor.layer5': 2.588219779332727, 'gradnorm/compressor.layer6': 1.4871204885834561, 'gradnorm/compressor.layer7': 1.705034847518681, 'gradnorm/decoder.layer0': 2.0547096628364474, 'gradnorm/decoder.layer1': 1.2427484936978426, 'gradnorm/decoder.layer2': 0.6165884466652252, 'gradnorm/max_param': 140.39613342285156, 'epoch': 6.1}

 61%|██████    | 11640/19090 [5:40:25<3:50:42,  1.86s/it]
 61%|██████    | 11641/19090 [5:40:26<3:30:01,  1.69s/it]
 61%|██████    | 11642/19090 [5:40:28<3:35:30,  1.74s/it]
 61%|██████    | 11643/19090 [5:40:30<3:46:50,  1.83s/it]
 61%|██████    | 11644/19090 [5:40:32<3:26:58,  1.67s/it]
 61%|██████    | 11645/19090 [5:40:34<3:37:15,  1.75s/it]
 61%|██████    | 11646/19090 [5:40:36<3:50:03,  1.85s/it]
 61%|██████    | 11647/19090 [5:40:37<3:28:52,  1.68s/it]
 61%|██████    | 11648/19090 [5:40:39<3:35:28,  1.74s/it]
 61%|██████    | 11649/19090 [5:40:41<3:50:05,  1.86s/it][gradnorm] step 11649: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 146.4

 61%|██████    | 11650/19090 [5:40:42<3:29:05,  1.69s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 211.49281311035156, 'learning_rate': 1.9783876354243283e-05, 'recon_loss': 0.0015163511532591655, 'recon_static': 0.0005455258215079084, 'recon_motion': 0.0009708253550343215, 'recon_norm': 1.2771344512701035, 'recon_norm_std': 0.044268982205539945, 'target_norm': 1.6219632089138032, 'recon_cos': 0.6444053784012794, 'gradnorm/compressor.query': 0.16764964242661545, 'gradnorm/compressor.layer0': 5.282247595834209, 'gradnorm/compressor.layer1': 207.52875365536698, 'gradnorm/compressor.layer2': 36.510707173119094, 'gradnorm/compressor.layer3': 16.00782012566924, 'gradnorm/compressor.layer4': 4.910852690888959, 'gradnorm/compressor.layer5': 2.7679199352261374, 'gradnorm/compressor.layer6': 1.4406278829483676, 'gradnorm/compressor.layer7': 1.723197246554804, 'gradnorm/decoder.layer0': 2.2467179202375305, 'gradnorm/decoder.layer1': 1.2369228577764189, 'gradnorm/decoder.layer2': 0.650933277008379, 'gradnorm/max_param': 146.37831115722656, 'epoch': 6.1}

 61%|██████    | 11650/19090 [5:40:42<3:29:05,  1.69s/it]
 61%|██████    | 11651/19090 [5:40:44<3:48:43,  1.84s/it]
 61%|██████    | 11652/19090 [5:40:46<3:50:38,  1.86s/it]
 61%|██████    | 11653/19090 [5:40:48<3:29:20,  1.69s/it]
 61%|██████    | 11654/19090 [5:40:50<3:56:26,  1.91s/it]
 61%|██████    | 11655/19090 [5:40:51<3:35:48,  1.74s/it]
 61%|██████    | 11656/19090 [5:40:53<3:18:47,  1.60s/it]
 61%|██████    | 11657/19090 [5:40:55<3:58:29,  1.93s/it]
 61%|██████    | 11658/19090 [5:40:57<3:34:42,  1.73s/it]
 61%|██████    | 11659/19090 [5:40:58<3:18:15,  1.60s/it][gradnorm] step 11659: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 138.6

 61%|██████    | 11660/19090 [5:41:01<4:03:28,  1.97s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 200.28903198242188, 'learning_rate': 1.9739177998792245e-05, 'recon_loss': 0.0015042380226077511, 'recon_static': 0.0005299522505083587, 'recon_motion': 0.0009742857597302646, 'recon_norm': 1.2736413836479188, 'recon_norm_std': 0.043241219501942395, 'target_norm': 1.6222835928201675, 'recon_cos': 0.6476405188441277, 'gradnorm/compressor.query': 0.1745743912762159, 'gradnorm/compressor.layer0': 5.956632607050804, 'gradnorm/compressor.layer1': 196.5470284054175, 'gradnorm/compressor.layer2': 34.97370489638877, 'gradnorm/compressor.layer3': 13.524320947096683, 'gradnorm/compressor.layer4': 4.765151303622072, 'gradnorm/compressor.layer5': 2.879609766059484, 'gradnorm/compressor.layer6': 1.5653206352377376, 'gradnorm/compressor.layer7': 1.8578991843261767, 'gradnorm/decoder.layer0': 2.1425769009582307, 'gradnorm/decoder.layer1': 1.2050773336200804, 'gradnorm/decoder.layer2': 0.6624128608256756, 'gradnorm/max_param': 138.6193389892578, 'epoch': 6.11}

 61%|██████    | 11660/19090 [5:41:01<4:03:28,  1.97s/it]
 61%|██████    | 11661/19090 [5:41:02<3:38:28,  1.76s/it]
 61%|██████    | 11662/19090 [5:41:03<3:20:44,  1.62s/it]
 61%|██████    | 11663/19090 [5:41:06<3:56:48,  1.91s/it]
 61%|██████    | 11664/19090 [5:41:07<3:36:01,  1.75s/it]
 61%|██████    | 11665/19090 [5:41:09<3:19:48,  1.61s/it]
 61%|██████    | 11666/19090 [5:41:11<3:58:33,  1.93s/it]
 61%|██████    | 11667/19090 [5:41:13<3:34:54,  1.74s/it]
 61%|██████    | 11668/19090 [5:41:14<3:18:37,  1.61s/it]
 61%|██████    | 11669/19090 [5:41:16<3:55:26,  1.90s/it][gradnorm] step 11669: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 139.8

 61%|██████    | 11670/19090 [5:41:18<3:49:23,  1.85s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 201.9024200439453, 'learning_rate': 1.969449723295712e-05, 'recon_loss': 0.0015326728578656911, 'recon_static': 0.0005610286083538085, 'recon_motion': 0.0009716442626086063, 'recon_norm': 1.287957713007927, 'recon_norm_std': 0.04299033721908927, 'target_norm': 1.6228869795799254, 'recon_cos': 0.6417750269174576, 'gradnorm/compressor.query': 0.14411929419836747, 'gradnorm/compressor.layer0': 4.9940224202011425, 'gradnorm/compressor.layer1': 198.15183177932016, 'gradnorm/compressor.layer2': 34.095334303817324, 'gradnorm/compressor.layer3': 16.371332820820612, 'gradnorm/compressor.layer4': 4.925933617724216, 'gradnorm/compressor.layer5': 2.8443884499654533, 'gradnorm/compressor.layer6': 1.6780341239205685, 'gradnorm/compressor.layer7': 1.942536054102324, 'gradnorm/decoder.layer0': 2.040989809009878, 'gradnorm/decoder.layer1': 1.2069661638310059, 'gradnorm/decoder.layer2': 0.6339694855187862, 'gradnorm/max_param': 139.7808074951172, 'epoch': 6.11}

 61%|██████    | 11670/19090 [5:41:18<3:49:23,  1.85s/it]
 61%|██████    | 11671/19090 [5:41:20<3:29:10,  1.69s/it]
 61%|██████    | 11672/19090 [5:41:22<3:49:43,  1.86s/it]
 61%|██████    | 11673/19090 [5:41:23<3:43:59,  1.81s/it]
 61%|██████    | 11674/19090 [5:41:25<3:24:34,  1.66s/it]
 61%|██████    | 11675/19090 [5:41:27<3:57:43,  1.92s/it]
 61%|██████    | 11676/19090 [5:41:29<3:38:46,  1.77s/it]
 61%|██████    | 11677/19090 [5:41:30<3:21:09,  1.63s/it]
 61%|██████    | 11678/19090 [5:41:33<4:03:31,  1.97s/it]
 61%|██████    | 11679/19090 [5:41:34<3:38:37,  1.77s/it][gradnorm] step 11679: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 132.3

 61%|██████    | 11680/19090 [5:41:35<3:21:25,  1.63s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 190.92941284179688, 'learning_rate': 1.9649834206128543e-05, 'recon_loss': 0.0014971119293477387, 'recon_static': 0.0005341024057997857, 'recon_motion': 0.0009630095286411233, 'recon_norm': 1.2740178883075715, 'recon_norm_std': 0.042825981695204976, 'target_norm': 1.620265594124794, 'recon_cos': 0.6483322441577911, 'gradnorm/compressor.query': 0.1564508994315728, 'gradnorm/compressor.layer0': 5.1137047949942005, 'gradnorm/compressor.layer1': 187.5271959443483, 'gradnorm/compressor.layer2': 32.103882970099, 'gradnorm/compressor.layer3': 14.018457507119185, 'gradnorm/compressor.layer4': 4.124457583744488, 'gradnorm/compressor.layer5': 2.4660198712565413, 'gradnorm/compressor.layer6': 1.4888432760669261, 'gradnorm/compressor.layer7': 1.8533279436476757, 'gradnorm/decoder.layer0': 1.8821385867245406, 'gradnorm/decoder.layer1': 1.2356624890488286, 'gradnorm/decoder.layer2': 0.6568488159187599, 'gradnorm/max_param': 132.2567901611328, 'epoch': 6.12}

 61%|██████    | 11680/19090 [5:41:35<3:21:25,  1.63s/it]
 61%|██████    | 11681/19090 [5:41:38<3:46:54,  1.84s/it]
 61%|██████    | 11682/19090 [5:41:39<3:38:37,  1.77s/it]
 61%|██████    | 11683/19090 [5:41:41<3:21:12,  1.63s/it]
 61%|██████    | 11684/19090 [5:41:43<3:45:30,  1.83s/it]
 61%|██████    | 11685/19090 [5:41:45<3:50:01,  1.86s/it]
 61%|██████    | 11686/19090 [5:41:46<3:30:33,  1.71s/it]
 61%|██████    | 11687/19090 [5:41:48<3:39:14,  1.78s/it]
 61%|██████    | 11688/19090 [5:41:50<3:43:34,  1.81s/it]
 61%|██████    | 11689/19090 [5:41:51<3:24:40,  1.66s/it][gradnorm] step 11689: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 123.6

 61%|██████    | 11690/19090 [5:41:54<3:49:00,  1.86s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 179.05303955078125, 'learning_rate': 1.9605189067637844e-05, 'recon_loss': 0.0015642007638234645, 'recon_static': 0.0006027051640558057, 'recon_motion': 0.0009614956099539996, 'recon_norm': 1.306514436006546, 'recon_norm_std': 0.042240670416504145, 'target_norm': 1.6192755788564681, 'recon_cos': 0.6340161472558975, 'gradnorm/compressor.query': 0.13455377734599103, 'gradnorm/compressor.layer0': 4.7567947376513375, 'gradnorm/compressor.layer1': 175.22010154196064, 'gradnorm/compressor.layer2': 33.68948749065222, 'gradnorm/compressor.layer3': 12.973446062414373, 'gradnorm/compressor.layer4': 3.9040677498113667, 'gradnorm/compressor.layer5': 2.483087938033571, 'gradnorm/compressor.layer6': 1.6359446294969855, 'gradnorm/compressor.layer7': 1.652490126887619, 'gradnorm/decoder.layer0': 1.8479084383133735, 'gradnorm/decoder.layer1': 1.147251944481471, 'gradnorm/decoder.layer2': 0.6528284933012888, 'gradnorm/max_param': 123.61875915527344, 'epoch': 6.12}

 61%|██████    | 11690/19090 [5:41:54<3:49:00,  1.86s/it]
 61%|██████    | 11691/19090 [5:41:55<3:38:27,  1.77s/it]
 61%|██████    | 11692/19090 [5:41:57<3:20:54,  1.63s/it]
 61%|██████▏   | 11693/19090 [5:41:59<3:47:42,  1.85s/it]
 61%|██████▏   | 11694/19090 [5:42:01<3:44:25,  1.82s/it]
 61%|██████▏   | 11695/19090 [5:42:02<3:24:59,  1.66s/it]
 61%|██████▏   | 11696/19090 [5:42:04<3:43:30,  1.81s/it]
 61%|██████▏   | 11697/19090 [5:42:06<3:50:57,  1.87s/it]
 61%|██████▏   | 11698/19090 [5:42:07<3:30:13,  1.71s/it]
 61%|██████▏   | 11699/19090 [5:42:10<3:45:33,  1.83s/it][gradnorm] step 11699: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 120.6

 61%|██████▏   | 11700/19090 [5:42:11<3:35:48,  1.75s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 174.73570251464844, 'learning_rate': 1.9560561966756518e-05, 'recon_loss': 0.0016086807561805471, 'recon_static': 0.000651960053073708, 'recon_motion': 0.0009567207103827968, 'recon_norm': 1.3262901782989502, 'recon_norm_std': 0.04266857299953699, 'target_norm': 1.619326862692833, 'recon_cos': 0.6243631586432457, 'gradnorm/compressor.query': 0.14203194949583195, 'gradnorm/compressor.layer0': 5.082496052978304, 'gradnorm/compressor.layer1': 170.9897429306799, 'gradnorm/compressor.layer2': 32.74885755104195, 'gradnorm/compressor.layer3': 12.639986415807451, 'gradnorm/compressor.layer4': 4.672103008715438, 'gradnorm/compressor.layer5': 2.305259814565939, 'gradnorm/compressor.layer6': 1.5225418979536063, 'gradnorm/compressor.layer7': 1.537696890901046, 'gradnorm/decoder.layer0': 1.8983222706678649, 'gradnorm/decoder.layer1': 1.1137017777575258, 'gradnorm/decoder.layer2': 0.5415596825756961, 'gradnorm/max_param': 120.62680053710938, 'epoch': 6.13}

 61%|██████▏   | 11700/19090 [5:42:11<3:35:48,  1.75s/it]
 61%|██████▏   | 11701/19090 [5:42:12<3:19:07,  1.62s/it]
 61%|██████▏   | 11702/19090 [5:42:15<3:45:20,  1.83s/it]
 61%|██████▏   | 11703/19090 [5:42:16<3:26:01,  1.67s/it]
 61%|██████▏   | 11704/19090 [5:42:17<3:12:13,  1.56s/it]
 61%|██████▏   | 11705/19090 [5:42:20<4:05:11,  1.99s/it]
 61%|██████▏   | 11706/19090 [5:42:22<3:39:40,  1.78s/it]
 61%|██████▏   | 11707/19090 [5:42:23<3:21:47,  1.64s/it]
 61%|██████▏   | 11708/19090 [5:42:26<4:01:15,  1.96s/it]
 61%|██████▏   | 11709/19090 [5:42:27<3:36:44,  1.76s/it][gradnorm] step 11709: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 105.0

 61%|██████▏   | 11710/19090 [5:42:28<3:19:55,  1.63s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 152.51211547851562, 'learning_rate': 1.951595305269577e-05, 'recon_loss': 0.0015753585728816688, 'recon_static': 0.0006110439368057996, 'recon_motion': 0.0009643147030146792, 'recon_norm': 1.307057249546051, 'recon_norm_std': 0.04175389632582664, 'target_norm': 1.619762685894966, 'recon_cos': 0.6306402653455734, 'gradnorm/compressor.query': 0.15057223983787585, 'gradnorm/compressor.layer0': 5.140727218956641, 'gradnorm/compressor.layer1': 148.8578718631467, 'gradnorm/compressor.layer2': 30.147241003725675, 'gradnorm/compressor.layer3': 11.719828075410812, 'gradnorm/compressor.layer4': 3.6610447671294963, 'gradnorm/compressor.layer5': 2.5287599914323224, 'gradnorm/compressor.layer6': 1.4263774304712322, 'gradnorm/compressor.layer7': 1.3589483007964913, 'gradnorm/decoder.layer0': 1.8783942335873725, 'gradnorm/decoder.layer1': 1.0437749254368507, 'gradnorm/decoder.layer2': 0.5685388078637503, 'gradnorm/max_param': 105.00310516357422, 'epoch': 6.13}

 61%|██████▏   | 11710/19090 [5:42:28<3:19:55,  1.63s/it]
 61%|██████▏   | 11711/19090 [5:42:31<3:51:06,  1.88s/it]
 61%|██████▏   | 11712/19090 [5:42:32<3:29:35,  1.70s/it]
 61%|██████▏   | 11713/19090 [5:42:33<3:14:43,  1.58s/it]
 61%|██████▏   | 11714/19090 [5:42:36<3:57:15,  1.93s/it]
 61%|██████▏   | 11715/19090 [5:42:37<3:34:05,  1.74s/it]
 61%|██████▏   | 11716/19090 [5:42:39<3:17:52,  1.61s/it]
 61%|██████▏   | 11717/19090 [5:42:41<3:52:22,  1.89s/it]
 61%|██████▏   | 11718/19090 [5:42:43<3:30:36,  1.71s/it]
 61%|██████▏   | 11719/19090 [5:42:44<3:15:14,  1.59s/it][gradnorm] step 11719: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 153.2

 61%|██████▏   | 11720/19090 [5:42:46<3:49:08,  1.87s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 220.98007202148438, 'learning_rate': 1.9471362474606e-05, 'recon_loss': 0.0016354636056348682, 'recon_static': 0.0006762540026102215, 'recon_motion': 0.0009592096554115415, 'recon_norm': 1.3345075488090514, 'recon_norm_std': 0.0424440611153841, 'target_norm': 1.6214220851659775, 'recon_cos': 0.6189906552433968, 'gradnorm/compressor.query': 0.17471016827192942, 'gradnorm/compressor.layer0': 5.917340235807689, 'gradnorm/compressor.layer1': 217.2584506411661, 'gradnorm/compressor.layer2': 37.19560048328969, 'gradnorm/compressor.layer3': 13.14553476388525, 'gradnorm/compressor.layer4': 4.829349300868576, 'gradnorm/compressor.layer5': 2.537148843154326, 'gradnorm/compressor.layer6': 1.3173652574233405, 'gradnorm/compressor.layer7': 1.722598881533942, 'gradnorm/decoder.layer0': 1.8955858579091585, 'gradnorm/decoder.layer1': 1.0609113372588173, 'gradnorm/decoder.layer2': 0.6411659468949299, 'gradnorm/max_param': 153.240966796875, 'epoch': 6.14}

 61%|██████▏   | 11720/19090 [5:42:46<3:49:08,  1.87s/it]
 61%|██████▏   | 11721/19090 [5:42:48<3:28:30,  1.70s/it]
 61%|██████▏   | 11722/19090 [5:42:49<3:13:38,  1.58s/it]
 61%|██████▏   | 11723/19090 [5:42:52<3:52:31,  1.89s/it]
 61%|██████▏   | 11724/19090 [5:42:53<3:32:07,  1.73s/it]
 61%|██████▏   | 11725/19090 [5:42:54<3:17:22,  1.61s/it]
 61%|██████▏   | 11726/19090 [5:42:57<3:54:13,  1.91s/it]
 61%|██████▏   | 11727/19090 [5:42:58<3:31:45,  1.73s/it]
 61%|██████▏   | 11728/19090 [5:42:59<3:16:10,  1.60s/it]
 61%|██████▏   | 11729/19090 [5:43:02<4:03:48,  1.99s/it][gradnorm] step 11729: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 125.7

 61%|██████▏   | 11730/19090 [5:43:04<3:38:30,  1.78s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 181.3079376220703, 'learning_rate': 1.9426790381576275e-05, 'recon_loss': 0.0015631745074642823, 'recon_static': 0.000592186753056012, 'recon_motion': 0.0009709877514978871, 'recon_norm': 1.299743613600731, 'recon_norm_std': 0.04235806167125702, 'target_norm': 1.6203237265348434, 'recon_cos': 0.6335775628685951, 'gradnorm/compressor.query': 0.16168799845091877, 'gradnorm/compressor.layer0': 5.347388384110283, 'gradnorm/compressor.layer1': 178.19886078304486, 'gradnorm/compressor.layer2': 29.68858859023899, 'gradnorm/compressor.layer3': 12.863867233464255, 'gradnorm/compressor.layer4': 4.699732979839079, 'gradnorm/compressor.layer5': 2.9732740765242043, 'gradnorm/compressor.layer6': 1.403511559947386, 'gradnorm/compressor.layer7': 1.6936583997442973, 'gradnorm/decoder.layer0': 2.2199448068304615, 'gradnorm/decoder.layer1': 1.0871128335802478, 'gradnorm/decoder.layer2': 0.5906887443899904, 'gradnorm/max_param': 125.69833374023438, 'epoch': 6.14}

 61%|██████▏   | 11730/19090 [5:43:04<3:38:30,  1.78s/it]
 61%|██████▏   | 11731/19090 [5:43:05<3:20:50,  1.64s/it]
 61%|██████▏   | 11732/19090 [5:43:08<3:58:41,  1.95s/it]
 61%|██████▏   | 11733/19090 [5:43:09<3:35:20,  1.76s/it]
 61%|██████▏   | 11734/19090 [5:43:10<3:18:33,  1.62s/it]
 61%|██████▏   | 11735/19090 [5:43:13<4:05:45,  2.00s/it]
 61%|██████▏   | 11736/19090 [5:43:14<3:39:31,  1.79s/it]
 61%|██████▏   | 11737/19090 [5:43:16<3:20:58,  1.64s/it]
 61%|██████▏   | 11738/19090 [5:43:19<4:05:59,  2.01s/it]
 61%|██████▏   | 11739/19090 [5:43:20<3:40:08,  1.80s/it][gradnorm] step 11739: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 139.5

 61%|██████▏   | 11740/19090 [5:43:21<3:21:58,  1.65s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 201.369384765625, 'learning_rate': 1.9382236922633888e-05, 'recon_loss': 0.0015710183361079544, 'recon_static': 0.000608507971628569, 'recon_motion': 0.0009625103892176412, 'recon_norm': 1.3043397814035416, 'recon_norm_std': 0.04234383851289749, 'target_norm': 1.619971713423729, 'recon_cos': 0.6314130008220673, 'gradnorm/compressor.query': 0.15216198494845257, 'gradnorm/compressor.layer0': 4.943225679432191, 'gradnorm/compressor.layer1': 197.80230028996124, 'gradnorm/compressor.layer2': 33.22335229466335, 'gradnorm/compressor.layer3': 15.907398234894217, 'gradnorm/compressor.layer4': 5.090022683383483, 'gradnorm/compressor.layer5': 2.606158712787188, 'gradnorm/compressor.layer6': 1.4401373511613262, 'gradnorm/compressor.layer7': 1.4985384178188277, 'gradnorm/decoder.layer0': 1.9758403197569498, 'gradnorm/decoder.layer1': 1.134690037637582, 'gradnorm/decoder.layer2': 0.6099057698166569, 'gradnorm/max_param': 139.51544189453125, 'epoch': 6.15}

 61%|██████▏   | 11740/19090 [5:43:21<3:21:58,  1.65s/it]
 62%|██████▏   | 11741/19090 [5:43:24<3:52:16,  1.90s/it]
 62%|██████▏   | 11742/19090 [5:43:25<3:30:30,  1.72s/it]
 62%|██████▏   | 11743/19090 [5:43:26<3:15:09,  1.59s/it]
 62%|██████▏   | 11744/19090 [5:43:29<4:04:57,  2.00s/it]
 62%|██████▏   | 11745/19090 [5:43:31<3:39:20,  1.79s/it]
 62%|██████▏   | 11746/19090 [5:43:32<3:21:12,  1.64s/it]
 62%|██████▏   | 11747/19090 [5:43:35<4:07:18,  2.02s/it]
 62%|██████▏   | 11748/19090 [5:43:36<3:40:15,  1.80s/it]
 62%|██████▏   | 11749/19090 [5:43:37<3:21:52,  1.65s/it][gradnorm] step 11749: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 153.1

 62%|██████▏   | 11750/19090 [5:43:40<4:00:53,  1.97s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 220.7080841064453, 'learning_rate': 1.9337702246743823e-05, 'recon_loss': 0.001588922945666127, 'recon_static': 0.0006234215834410861, 'recon_motion': 0.0009655013476731255, 'recon_norm': 1.3152276933193208, 'recon_norm_std': 0.04197775050997734, 'target_norm': 1.620706680417061, 'recon_cos': 0.6289107948541641, 'gradnorm/compressor.query': 0.21478053810293593, 'gradnorm/compressor.layer0': 5.840815332535287, 'gradnorm/compressor.layer1': 217.1935790516147, 'gradnorm/compressor.layer2': 35.44934077974864, 'gradnorm/compressor.layer3': 14.165395043840002, 'gradnorm/compressor.layer4': 5.313422986801439, 'gradnorm/compressor.layer5': 2.9023097409788234, 'gradnorm/compressor.layer6': 1.4870986044283017, 'gradnorm/compressor.layer7': 1.7238940854120086, 'gradnorm/decoder.layer0': 2.041309857492403, 'gradnorm/decoder.layer1': 1.0850589409213849, 'gradnorm/decoder.layer2': 0.5938415205841941, 'gradnorm/max_param': 153.1171417236328, 'epoch': 6.16}

 62%|██████▏   | 11750/19090 [5:43:40<4:00:53,  1.97s/it]
 62%|██████▏   | 11751/19090 [5:43:41<3:36:33,  1.77s/it]
 62%|██████▏   | 11752/19090 [5:43:43<3:19:16,  1.63s/it]
 62%|██████▏   | 11753/19090 [5:43:45<3:58:25,  1.95s/it]
 62%|██████▏   | 11754/19090 [5:43:47<3:34:51,  1.76s/it]
 62%|██████▏   | 11755/19090 [5:43:48<3:18:13,  1.62s/it]
 62%|██████▏   | 11756/19090 [5:43:51<3:54:55,  1.92s/it]
 62%|██████▏   | 11757/19090 [5:43:52<3:32:33,  1.74s/it]
 62%|██████▏   | 11758/19090 [5:43:53<3:16:59,  1.61s/it]
 62%|██████▏   | 11759/19090 [5:43:55<3:35:56,  1.77s/it][gradnorm] step 11759: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 116.3

 62%|██████▏   | 11760/19090 [5:43:57<3:18:32,  1.63s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 168.56503295898438, 'learning_rate': 1.9293186502808246e-05, 'recon_loss': 0.0016878704045666382, 'recon_static': 0.0007257109115016647, 'recon_motion': 0.0009621595105272718, 'recon_norm': 1.3539420157670974, 'recon_norm_std': 0.040990426857024434, 'target_norm': 1.6192987740039826, 'recon_cos': 0.6093155279755592, 'gradnorm/compressor.query': 0.16861721660361192, 'gradnorm/compressor.layer0': 5.899921558957836, 'gradnorm/compressor.layer1': 164.82810056047785, 'gradnorm/compressor.layer2': 32.33702186965333, 'gradnorm/compressor.layer3': 11.52774813442683, 'gradnorm/compressor.layer4': 4.0538220961255975, 'gradnorm/compressor.layer5': 2.5039657609506043, 'gradnorm/compressor.layer6': 1.4752489332496808, 'gradnorm/compressor.layer7': 1.6907042705122528, 'gradnorm/decoder.layer0': 1.7876318662861133, 'gradnorm/decoder.layer1': 1.091603325081189, 'gradnorm/decoder.layer2': 0.5787175981831589, 'gradnorm/max_param': 116.2655258178711, 'epoch': 6.16}

 62%|██████▏   | 11760/19090 [5:43:57<3:18:32,  1.63s/it]
 62%|██████▏   | 11761/19090 [5:43:58<3:06:24,  1.53s/it]
 62%|██████▏   | 11762/19090 [5:44:01<3:58:06,  1.95s/it]
 62%|██████▏   | 11763/19090 [5:44:02<3:33:52,  1.75s/it]
 62%|██████▏   | 11764/19090 [5:44:03<3:16:56,  1.61s/it]
 62%|██████▏   | 11765/19090 [5:44:06<4:08:40,  2.04s/it]
 62%|██████▏   | 11766/19090 [5:44:08<3:41:17,  1.81s/it]
 62%|██████▏   | 11767/19090 [5:44:09<3:21:52,  1.65s/it]
 62%|██████▏   | 11768/19090 [5:44:12<4:04:17,  2.00s/it]
 62%|██████▏   | 11769/19090 [5:44:13<3:38:30,  1.79s/it][gradnorm] step 11769: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 137.6

 62%|██████▏   | 11770/19090 [5:44:14<3:20:38,  1.64s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 199.7722930908203, 'learning_rate': 1.924868983966603e-05, 'recon_loss': 0.0017113440815592184, 'recon_static': 0.000743542549025733, 'recon_motion': 0.0009678015878307634, 'recon_norm': 1.3648196935653687, 'recon_norm_std': 0.041597832646220925, 'target_norm': 1.6205485552549361, 'recon_cos': 0.6042888298630714, 'gradnorm/compressor.query': 0.18211262402608486, 'gradnorm/compressor.layer0': 6.130284734837019, 'gradnorm/compressor.layer1': 195.09233591110134, 'gradnorm/compressor.layer2': 39.715333805191825, 'gradnorm/compressor.layer3': 13.972670992618726, 'gradnorm/compressor.layer4': 4.5443091215107385, 'gradnorm/compressor.layer5': 2.4800181062098936, 'gradnorm/compressor.layer6': 1.479056457210587, 'gradnorm/compressor.layer7': 1.7267152882631813, 'gradnorm/decoder.layer0': 2.0041139729235176, 'gradnorm/decoder.layer1': 1.166318103400484, 'gradnorm/decoder.layer2': 0.6637925889743087, 'gradnorm/max_param': 137.62527465820312, 'epoch': 6.17}

 62%|██████▏   | 11770/19090 [5:44:14<3:20:38,  1.64s/it]
 62%|██████▏   | 11771/19090 [5:44:17<3:47:57,  1.87s/it]
 62%|██████▏   | 11772/19090 [5:44:18<3:26:32,  1.69s/it]
 62%|██████▏   | 11773/19090 [5:44:19<3:11:39,  1.57s/it]
 62%|██████▏   | 11774/19090 [5:44:21<3:29:03,  1.71s/it]
 62%|██████▏   | 11775/19090 [5:44:23<3:13:17,  1.59s/it]
 62%|██████▏   | 11776/19090 [5:44:24<3:02:19,  1.50s/it]
 62%|██████▏   | 11777/19090 [5:44:27<3:55:09,  1.93s/it]
 62%|██████▏   | 11778/19090 [5:44:28<3:31:39,  1.74s/it]
 62%|██████▏   | 11779/19090 [5:44:30<3:15:14,  1.60s/it][gradnorm] step 11779: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 162.5

 62%|██████▏   | 11780/19090 [5:44:32<3:58:43,  1.96s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 234.669921875, 'learning_rate': 1.9204212406092275e-05, 'recon_loss': 0.0016450347495265305, 'recon_static': 0.0006867074029287323, 'recon_motion': 0.0009583273465977981, 'recon_norm': 1.3408795833587646, 'recon_norm_std': 0.04222578974440694, 'target_norm': 1.618374028801918, 'recon_cos': 0.616419106721878, 'gradnorm/compressor.query': 0.18544310629495322, 'gradnorm/compressor.layer0': 6.34878205160788, 'gradnorm/compressor.layer1': 230.4239045352066, 'gradnorm/compressor.layer2': 41.27062398724879, 'gradnorm/compressor.layer3': 13.95271982456932, 'gradnorm/compressor.layer4': 4.3502493337448, 'gradnorm/compressor.layer5': 2.554477041365708, 'gradnorm/compressor.layer6': 1.526030148754551, 'gradnorm/compressor.layer7': 1.7159518265842537, 'gradnorm/decoder.layer0': 1.9824492955308053, 'gradnorm/decoder.layer1': 1.2082392502473662, 'gradnorm/decoder.layer2': 0.6276644892741742, 'gradnorm/max_param': 162.53846740722656, 'epoch': 6.17}

 62%|██████▏   | 11780/19090 [5:44:32<3:58:43,  1.96s/it]
 62%|██████▏   | 11781/19090 [5:44:34<3:34:18,  1.76s/it]
 62%|██████▏   | 11782/19090 [5:44:35<3:17:39,  1.62s/it]
 62%|██████▏   | 11783/19090 [5:44:38<4:02:07,  1.99s/it]
 62%|██████▏   | 11784/19090 [5:44:39<3:36:20,  1.78s/it]
 62%|██████▏   | 11785/19090 [5:44:40<3:18:24,  1.63s/it]
 62%|██████▏   | 11786/19090 [5:44:43<4:00:35,  1.98s/it]
 62%|██████▏   | 11787/19090 [5:44:44<3:35:36,  1.77s/it]
 62%|██████▏   | 11788/19090 [5:44:46<3:17:53,  1.63s/it]
 62%|██████▏   | 11789/19090 [5:44:49<4:12:57,  2.08s/it][gradnorm] step 11789: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 137.8

 62%|██████▏   | 11790/19090 [5:44:50<3:44:06,  1.84s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 199.3979949951172, 'learning_rate': 1.9159754350797746e-05, 'recon_loss': 0.0016146297712111846, 'recon_static': 0.0006499813272967004, 'recon_motion': 0.000964648426452186, 'recon_norm': 1.3223249226808549, 'recon_norm_std': 0.04170049177482724, 'target_norm': 1.6217427581548691, 'recon_cos': 0.6232243165373802, 'gradnorm/compressor.query': 0.14573386254173984, 'gradnorm/compressor.layer0': 5.037551629722433, 'gradnorm/compressor.layer1': 195.39146403758278, 'gradnorm/compressor.layer2': 36.38174973619586, 'gradnorm/compressor.layer3': 14.064545207178426, 'gradnorm/compressor.layer4': 4.24674077442928, 'gradnorm/compressor.layer5': 2.5358180517997533, 'gradnorm/compressor.layer6': 1.5306688101041277, 'gradnorm/compressor.layer7': 1.6380625013460384, 'gradnorm/decoder.layer0': 1.911585485679696, 'gradnorm/decoder.layer1': 1.176527223906761, 'gradnorm/decoder.layer2': 0.6072091959926653, 'gradnorm/max_param': 137.83242797851562, 'epoch': 6.18}

 62%|██████▏   | 11790/19090 [5:44:50<3:44:06,  1.84s/it]
 62%|██████▏   | 11791/19090 [5:44:51<3:24:00,  1.68s/it]
 62%|██████▏   | 11792/19090 [5:44:54<4:03:36,  2.00s/it]
 62%|██████▏   | 11793/19090 [5:44:56<3:37:47,  1.79s/it]
 62%|██████▏   | 11794/19090 [5:44:57<3:19:30,  1.64s/it]
 62%|██████▏   | 11795/19090 [5:44:59<3:56:34,  1.95s/it]
 62%|██████▏   | 11796/19090 [5:45:01<3:32:38,  1.75s/it]
 62%|██████▏   | 11797/19090 [5:45:02<3:15:49,  1.61s/it]
 62%|██████▏   | 11798/19090 [5:45:05<3:48:47,  1.88s/it]
 62%|██████▏   | 11799/19090 [5:45:06<3:26:59,  1.70s/it][gradnorm] step 11799: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 114.4

 62%|██████▏   | 11800/19090 [5:45:07<3:11:42,  1.58s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 166.12158203125, 'learning_rate': 1.9115315822428437e-05, 'recon_loss': 0.0016179318394279108, 'recon_static': 0.0006597773535759188, 'recon_motion': 0.0009581544785760343, 'recon_norm': 1.3279823392629624, 'recon_norm_std': 0.04227084470912814, 'target_norm': 1.6208517879247666, 'recon_cos': 0.6230137795209885, 'gradnorm/compressor.query': 0.17468102305795857, 'gradnorm/compressor.layer0': 5.895639081614822, 'gradnorm/compressor.layer1': 162.0896140064193, 'gradnorm/compressor.layer2': 32.867147573959585, 'gradnorm/compressor.layer3': 13.057314194714477, 'gradnorm/compressor.layer4': 4.711987025827574, 'gradnorm/compressor.layer5': 2.4706163715315133, 'gradnorm/compressor.layer6': 1.4111010485179358, 'gradnorm/compressor.layer7': 1.5027443893650265, 'gradnorm/decoder.layer0': 1.8807762816901399, 'gradnorm/decoder.layer1': 1.182635129986206, 'gradnorm/decoder.layer2': 0.5564408328571274, 'gradnorm/max_param': 114.35856628417969, 'epoch': 6.18}

 62%|██████▏   | 11800/19090 [5:45:07<3:11:42,  1.58s/it]
 62%|██████▏   | 11801/19090 [5:45:10<4:00:40,  1.98s/it]
 62%|██████▏   | 11802/19090 [5:45:11<3:35:28,  1.77s/it]
 62%|██████▏   | 11803/19090 [5:45:13<3:17:26,  1.63s/it]
 62%|██████▏   | 11804/19090 [5:45:15<3:49:02,  1.89s/it]
 62%|██████▏   | 11805/19090 [5:45:16<3:27:09,  1.71s/it]
 62%|██████▏   | 11806/19090 [5:45:18<3:11:46,  1.58s/it]
 62%|██████▏   | 11807/19090 [5:45:21<3:58:52,  1.97s/it]
 62%|██████▏   | 11808/19090 [5:45:22<3:34:16,  1.77s/it]
 62%|██████▏   | 11809/19090 [5:45:23<3:16:55,  1.62s/it][gradnorm] step 11809: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 120.3

 62%|██████▏   | 11810/19090 [5:45:26<4:03:27,  2.01s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 173.86045837402344, 'learning_rate': 1.9070896969565075e-05, 'recon_loss': 0.0016311009007040412, 'recon_static': 0.0006604571535717696, 'recon_motion': 0.0009706436714623124, 'recon_norm': 1.3275940626859666, 'recon_norm_std': 0.04111582571640611, 'target_norm': 1.620844441652298, 'recon_cos': 0.6194737896323204, 'gradnorm/compressor.query': 0.1576025950165835, 'gradnorm/compressor.layer0': 5.5470383311120335, 'gradnorm/compressor.layer1': 170.4858578720681, 'gradnorm/compressor.layer2': 30.662670151510167, 'gradnorm/compressor.layer3': 12.518757898306296, 'gradnorm/compressor.layer4': 4.243790072671857, 'gradnorm/compressor.layer5': 2.620718779938616, 'gradnorm/compressor.layer6': 1.4310713777314643, 'gradnorm/compressor.layer7': 1.5895093711487984, 'gradnorm/decoder.layer0': 1.741749387659672, 'gradnorm/decoder.layer1': 1.2313392103378855, 'gradnorm/decoder.layer2': 0.5615360424451404, 'gradnorm/max_param': 120.25997161865234, 'epoch': 6.19}

 62%|██████▏   | 11810/19090 [5:45:26<4:03:27,  2.01s/it]
 62%|██████▏   | 11811/19090 [5:45:27<3:37:37,  1.79s/it]
 62%|██████▏   | 11812/19090 [5:45:29<3:19:12,  1.64s/it]
 62%|██████▏   | 11813/19090 [5:45:31<4:01:39,  1.99s/it]
 62%|██████▏   | 11814/19090 [5:45:33<3:36:01,  1.78s/it]
 62%|██████▏   | 11815/19090 [5:45:34<3:18:17,  1.64s/it]
 62%|██████▏   | 11816/19090 [5:45:37<3:59:24,  1.97s/it]
 62%|██████▏   | 11817/19090 [5:45:38<3:34:33,  1.77s/it]
 62%|██████▏   | 11818/19090 [5:45:39<3:17:24,  1.63s/it]
 62%|██████▏   | 11819/19090 [5:45:42<3:59:03,  1.97s/it][gradnorm] step 11819: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 122.8

 62%|██████▏   | 11820/19090 [5:45:43<3:34:37,  1.77s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 177.67823791503906, 'learning_rate': 1.9026497940722564e-05, 'recon_loss': 0.001604513407801278, 'recon_static': 0.0006350076029775665, 'recon_motion': 0.0009695058484794572, 'recon_norm': 1.3174052685499191, 'recon_norm_std': 0.04279039576649666, 'target_norm': 1.6230615943670272, 'recon_cos': 0.6260880663990974, 'gradnorm/compressor.query': 0.12637005939787282, 'gradnorm/compressor.layer0': 4.690177661320622, 'gradnorm/compressor.layer1': 174.09890777227753, 'gradnorm/compressor.layer2': 32.12899339843804, 'gradnorm/compressor.layer3': 12.941005950457082, 'gradnorm/compressor.layer4': 4.567492118540511, 'gradnorm/compressor.layer5': 2.56803454531551, 'gradnorm/compressor.layer6': 1.376741433611033, 'gradnorm/compressor.layer7': 1.537938593512882, 'gradnorm/decoder.layer0': 2.008753218490975, 'gradnorm/decoder.layer1': 1.120773216441065, 'gradnorm/decoder.layer2': 0.6008726846592011, 'gradnorm/max_param': 122.83570861816406, 'epoch': 6.19}

 62%|██████▏   | 11820/19090 [5:45:43<3:34:37,  1.77s/it]
 62%|██████▏   | 11821/19090 [5:45:45<3:17:25,  1.63s/it]
 62%|██████▏   | 11822/19090 [5:45:47<3:54:07,  1.93s/it]
 62%|██████▏   | 11823/19090 [5:45:49<3:30:59,  1.74s/it]
 62%|██████▏   | 11824/19090 [5:45:50<3:14:43,  1.61s/it]
 62%|██████▏   | 11825/19090 [5:45:52<3:41:52,  1.83s/it]
 62%|██████▏   | 11826/19090 [5:45:54<3:22:28,  1.67s/it]
 62%|██████▏   | 11827/19090 [5:45:55<3:08:39,  1.56s/it]
 62%|██████▏   | 11828/19090 [5:45:58<3:45:15,  1.86s/it]
 62%|██████▏   | 11829/19090 [5:45:59<3:24:28,  1.69s/it][gradnorm] step 11829: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 152.0

 62%|██████▏   | 11830/19090 [5:46:00<3:09:48,  1.57s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 219.39479064941406, 'learning_rate': 1.8982118884349538e-05, 'recon_loss': 0.0016211541456868871, 'recon_static': 0.0006519146307255141, 'recon_motion': 0.0009692394916783087, 'recon_norm': 1.3287227869033813, 'recon_norm_std': 0.04134215442463755, 'target_norm': 1.6196378082036973, 'recon_cos': 0.6221282541751861, 'gradnorm/compressor.query': 0.1707220093256947, 'gradnorm/compressor.layer0': 5.574197706304444, 'gradnorm/compressor.layer1': 215.42662976637777, 'gradnorm/compressor.layer2': 37.93294189823256, 'gradnorm/compressor.layer3': 14.623343610294008, 'gradnorm/compressor.layer4': 4.586851168699376, 'gradnorm/compressor.layer5': 3.056503653580441, 'gradnorm/compressor.layer6': 1.4862082153555265, 'gradnorm/compressor.layer7': 1.7699129817068213, 'gradnorm/decoder.layer0': 2.03177682940405, 'gradnorm/decoder.layer1': 1.1512642151889998, 'gradnorm/decoder.layer2': 0.6361270331546842, 'gradnorm/max_param': 151.979248046875, 'epoch': 6.2}

 62%|██████▏   | 11830/19090 [5:46:00<3:09:48,  1.57s/it]
 62%|██████▏   | 11831/19090 [5:46:03<3:50:50,  1.91s/it]
 62%|██████▏   | 11832/19090 [5:46:04<3:28:44,  1.73s/it]
 62%|██████▏   | 11833/19090 [5:46:05<3:13:16,  1.60s/it]
 62%|██████▏   | 11834/19090 [5:46:08<3:58:23,  1.97s/it]
 62%|██████▏   | 11835/19090 [5:46:10<3:33:38,  1.77s/it]
 62%|██████▏   | 11836/19090 [5:46:11<3:16:25,  1.62s/it]
 62%|██████▏   | 11837/19090 [5:46:13<3:54:12,  1.94s/it]
 62%|██████▏   | 11838/19090 [5:46:15<3:30:46,  1.74s/it]
 62%|██████▏   | 11839/19090 [5:46:16<3:14:07,  1.61s/it][gradnorm] step 11839: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 135.5

 62%|██████▏   | 11840/19090 [5:46:19<4:10:58,  2.08s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 195.6497802734375, 'learning_rate': 1.893775994882787e-05, 'recon_loss': 0.0016002392396330834, 'recon_static': 0.0006382466221111826, 'recon_motion': 0.000961992634984199, 'recon_norm': 1.3193111896514893, 'recon_norm_std': 0.04203275153413415, 'target_norm': 1.620208775997162, 'recon_cos': 0.6257592305541039, 'gradnorm/compressor.query': 0.13532980018700264, 'gradnorm/compressor.layer0': 4.747793387798447, 'gradnorm/compressor.layer1': 192.05650006899012, 'gradnorm/compressor.layer2': 34.2055589586236, 'gradnorm/compressor.layer3': 12.833773782941584, 'gradnorm/compressor.layer4': 4.052098971273782, 'gradnorm/compressor.layer5': 2.687523597791112, 'gradnorm/compressor.layer6': 1.466094518094358, 'gradnorm/compressor.layer7': 1.8994773246212993, 'gradnorm/decoder.layer0': 2.204390250429852, 'gradnorm/decoder.layer1': 1.1255374790121704, 'gradnorm/decoder.layer2': 0.5717092941015139, 'gradnorm/max_param': 135.48779296875, 'epoch': 6.2}

 62%|██████▏   | 11840/19090 [5:46:19<4:10:58,  2.08s/it]
 62%|██████▏   | 11841/19090 [5:46:21<3:42:29,  1.84s/it]
 62%|██████▏   | 11842/19090 [5:46:22<3:22:55,  1.68s/it]
 62%|██████▏   | 11843/19090 [5:46:24<3:55:44,  1.95s/it]
 62%|██████▏   | 11844/19090 [5:46:26<3:32:02,  1.76s/it]
 62%|██████▏   | 11845/19090 [5:46:27<3:15:40,  1.62s/it]
 62%|██████▏   | 11846/19090 [5:46:30<3:53:14,  1.93s/it]
 62%|██████▏   | 11847/19090 [5:46:31<3:29:46,  1.74s/it]
 62%|██████▏   | 11848/19090 [5:46:32<3:13:24,  1.60s/it]
 62%|██████▏   | 11849/19090 [5:46:35<4:00:31,  1.99s/it][gradnorm] step 11849: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 172.8

 62%|██████▏   | 11850/19090 [5:46:36<3:35:11,  1.78s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 248.74549865722656, 'learning_rate': 1.8893421282472147e-05, 'recon_loss': 0.0016087275318568573, 'recon_static': 0.0006456440198235214, 'recon_motion': 0.0009630834916606546, 'recon_norm': 1.3197080850601197, 'recon_norm_std': 0.0425250587053597, 'target_norm': 1.621052587032318, 'recon_cos': 0.6245284900069237, 'gradnorm/compressor.query': 0.17210890286346658, 'gradnorm/compressor.layer0': 6.355655716652929, 'gradnorm/compressor.layer1': 245.08761188603557, 'gradnorm/compressor.layer2': 39.09019174703947, 'gradnorm/compressor.layer3': 14.00648729701231, 'gradnorm/compressor.layer4': 4.3687280657085275, 'gradnorm/compressor.layer5': 2.8338886352604096, 'gradnorm/compressor.layer6': 1.7522695674492172, 'gradnorm/compressor.layer7': 2.061948991634857, 'gradnorm/decoder.layer0': 2.3189603931158653, 'gradnorm/decoder.layer1': 1.205481809395778, 'gradnorm/decoder.layer2': 0.6788986714843301, 'gradnorm/max_param': 172.8446807861328, 'epoch': 6.21}

 62%|██████▏   | 11850/19090 [5:46:36<3:35:11,  1.78s/it]
 62%|██████▏   | 11851/19090 [5:46:38<3:17:24,  1.64s/it]
 62%|██████▏   | 11852/19090 [5:46:40<3:55:34,  1.95s/it]
 62%|██████▏   | 11853/19090 [5:46:42<3:31:27,  1.75s/it]
 62%|██████▏   | 11854/19090 [5:46:43<3:14:21,  1.61s/it]
 62%|██████▏   | 11855/19090 [5:46:46<4:03:36,  2.02s/it]
 62%|██████▏   | 11856/19090 [5:46:47<3:37:16,  1.80s/it]
 62%|██████▏   | 11857/19090 [5:46:49<3:18:48,  1.65s/it]
 62%|██████▏   | 11858/19090 [5:46:51<4:02:48,  2.01s/it]
 62%|██████▏   | 11859/19090 [5:46:53<3:36:23,  1.80s/it][gradnorm] step 11859: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 136.1

 62%|██████▏   | 11860/19090 [5:46:54<3:18:00,  1.64s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 196.73924255371094, 'learning_rate': 1.884910303352918e-05, 'recon_loss': 0.0016209277382586151, 'recon_static': 0.0006557821878232061, 'recon_motion': 0.0009651455620769411, 'recon_norm': 1.334008240699768, 'recon_norm_std': 0.0419027310796082, 'target_norm': 1.6184248864650725, 'recon_cos': 0.6223289966583252, 'gradnorm/compressor.query': 0.15738363298442282, 'gradnorm/compressor.layer0': 4.952843498630239, 'gradnorm/compressor.layer1': 192.94022309383806, 'gradnorm/compressor.layer2': 35.25893164493484, 'gradnorm/compressor.layer3': 13.14198932981312, 'gradnorm/compressor.layer4': 4.4544840562513475, 'gradnorm/compressor.layer5': 2.7912674423882287, 'gradnorm/compressor.layer6': 1.740889951309617, 'gradnorm/compressor.layer7': 1.8708934318331354, 'gradnorm/decoder.layer0': 2.00173981809498, 'gradnorm/decoder.layer1': 1.1602384493186493, 'gradnorm/decoder.layer2': 0.651373981857895, 'gradnorm/max_param': 136.12733459472656, 'epoch': 6.21}

 62%|██████▏   | 11860/19090 [5:46:54<3:18:00,  1.64s/it]
 62%|██████▏   | 11861/19090 [5:46:56<3:45:55,  1.88s/it]
 62%|██████▏   | 11862/19090 [5:46:58<3:25:11,  1.70s/it]
 62%|██████▏   | 11863/19090 [5:46:59<3:10:15,  1.58s/it]
 62%|██████▏   | 11864/19090 [5:47:02<3:51:59,  1.93s/it]
 62%|██████▏   | 11865/19090 [5:47:03<3:29:20,  1.74s/it]
 62%|██████▏   | 11866/19090 [5:47:04<3:12:58,  1.60s/it]
 62%|██████▏   | 11867/19090 [5:47:07<3:38:33,  1.82s/it]
 62%|██████▏   | 11868/19090 [5:47:08<3:19:18,  1.66s/it]
 62%|██████▏   | 11869/19090 [5:47:09<3:05:53,  1.54s/it][gradnorm] step 11869: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 135.7

 62%|██████▏   | 11870/19090 [5:47:12<3:43:13,  1.85s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 196.64678955078125, 'learning_rate': 1.8804805350177505e-05, 'recon_loss': 0.0015517876017838717, 'recon_static': 0.0005916367241297849, 'recon_motion': 0.0009601507932529785, 'recon_norm': 1.2974329978227614, 'recon_norm_std': 0.04191800346598029, 'target_norm': 1.6179736077785491, 'recon_cos': 0.6347564160823822, 'gradnorm/compressor.query': 0.14195369476808653, 'gradnorm/compressor.layer0': 4.97872135400253, 'gradnorm/compressor.layer1': 192.31183985912048, 'gradnorm/compressor.layer2': 37.661634080860395, 'gradnorm/compressor.layer3': 14.142356269191021, 'gradnorm/compressor.layer4': 4.667425071306946, 'gradnorm/compressor.layer5': 2.789428601426255, 'gradnorm/compressor.layer6': 1.6946103096471767, 'gradnorm/compressor.layer7': 1.9650944877795289, 'gradnorm/decoder.layer0': 2.153233726646363, 'gradnorm/decoder.layer1': 1.2084096796756123, 'gradnorm/decoder.layer2': 0.7154812497681797, 'gradnorm/max_param': 135.65565490722656, 'epoch': 6.22}

 62%|██████▏   | 11870/19090 [5:47:12<3:43:13,  1.85s/it]
 62%|██████▏   | 11871/19090 [5:47:13<3:22:29,  1.68s/it]
 62%|██████▏   | 11872/19090 [5:47:14<3:07:55,  1.56s/it]
 62%|██████▏   | 11873/19090 [5:47:17<3:56:23,  1.97s/it]
 62%|██████▏   | 11874/19090 [5:47:19<3:32:11,  1.76s/it]
 62%|██████▏   | 11875/19090 [5:47:20<3:14:46,  1.62s/it]
 62%|██████▏   | 11876/19090 [5:47:22<3:46:05,  1.88s/it]
 62%|██████▏   | 11877/19090 [5:47:24<3:24:35,  1.70s/it]
 62%|██████▏   | 11878/19090 [5:47:25<3:09:30,  1.58s/it]
 62%|██████▏   | 11879/19090 [5:47:28<3:50:39,  1.92s/it][gradnorm] step 11879: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 138.0

 62%|██████▏   | 11880/19090 [5:47:29<3:27:42,  1.73s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 199.95249938964844, 'learning_rate': 1.8760528380526934e-05, 'recon_loss': 0.0015994471759768203, 'recon_static': 0.0006182434124639257, 'recon_motion': 0.000981203828996513, 'recon_norm': 1.3130094975233078, 'recon_norm_std': 0.04116650875657797, 'target_norm': 1.623493430018425, 'recon_cos': 0.6274368911981583, 'gradnorm/compressor.query': 0.13964202839476014, 'gradnorm/compressor.layer0': 4.788604664659572, 'gradnorm/compressor.layer1': 195.58267313657925, 'gradnorm/compressor.layer2': 38.112922331314515, 'gradnorm/compressor.layer3': 14.624894622683932, 'gradnorm/compressor.layer4': 4.487253999475687, 'gradnorm/compressor.layer5': 2.6227456357994345, 'gradnorm/compressor.layer6': 1.6380239303713229, 'gradnorm/compressor.layer7': 1.781533268941486, 'gradnorm/decoder.layer0': 2.107346640331939, 'gradnorm/decoder.layer1': 1.1322065113734165, 'gradnorm/decoder.layer2': 0.6244638766171281, 'gradnorm/max_param': 137.97291564941406, 'epoch': 6.22}

 62%|██████▏   | 11880/19090 [5:47:29<3:27:42,  1.73s/it]
 62%|██████▏   | 11881/19090 [5:47:30<3:11:51,  1.60s/it]
 62%|██████▏   | 11882/19090 [5:47:33<4:05:10,  2.04s/it]
 62%|██████▏   | 11883/19090 [5:47:35<3:37:48,  1.81s/it]
 62%|██████▏   | 11884/19090 [5:47:36<3:18:39,  1.65s/it]
 62%|██████▏   | 11885/19090 [5:47:38<3:48:30,  1.90s/it]
 62%|██████▏   | 11886/19090 [5:47:40<3:26:16,  1.72s/it]
 62%|██████▏   | 11887/19090 [5:47:41<3:10:48,  1.59s/it]
 62%|██████▏   | 11888/19090 [5:47:43<3:45:58,  1.88s/it]
 62%|██████▏   | 11889/19090 [5:47:45<3:24:37,  1.70s/it][gradnorm] step 11889: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 148.2

 62%|██████▏   | 11890/19090 [5:47:46<3:09:32,  1.58s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 214.52578735351562, 'learning_rate': 1.8716272272617997e-05, 'recon_loss': 0.001607620230061002, 'recon_static': 0.0006423865532269701, 'recon_motion': 0.0009652336870203726, 'recon_norm': 1.319476020336151, 'recon_norm_std': 0.04160541789606213, 'target_norm': 1.622155812382698, 'recon_cos': 0.6251942172646523, 'gradnorm/compressor.query': 0.17906933684886495, 'gradnorm/compressor.layer0': 6.198246110981309, 'gradnorm/compressor.layer1': 210.02956451116592, 'gradnorm/compressor.layer2': 40.02781320629601, 'gradnorm/compressor.layer3': 15.158054519670031, 'gradnorm/compressor.layer4': 4.523250490318765, 'gradnorm/compressor.layer5': 2.620565028524495, 'gradnorm/compressor.layer6': 1.5331530520286696, 'gradnorm/compressor.layer7': 1.6149263653340513, 'gradnorm/decoder.layer0': 2.0924325470438045, 'gradnorm/decoder.layer1': 1.181876772148462, 'gradnorm/decoder.layer2': 0.627236394378108, 'gradnorm/max_param': 148.1594696044922, 'epoch': 6.23}

 62%|██████▏   | 11890/19090 [5:47:46<3:09:32,  1.58s/it]
 62%|██████▏   | 11891/19090 [5:47:49<3:51:04,  1.93s/it]
 62%|██████▏   | 11892/19090 [5:47:50<3:27:52,  1.73s/it]
 62%|██████▏   | 11893/19090 [5:47:51<3:11:56,  1.60s/it]
 62%|██████▏   | 11894/19090 [5:47:54<3:54:49,  1.96s/it]
 62%|██████▏   | 11895/19090 [5:47:55<3:30:29,  1.76s/it]
 62%|██████▏   | 11896/19090 [5:47:57<3:13:39,  1.62s/it]
 62%|██████▏   | 11897/19090 [5:47:59<3:51:07,  1.93s/it]
 62%|██████▏   | 11898/19090 [5:48:01<3:27:59,  1.74s/it]
 62%|██████▏   | 11899/19090 [5:48:02<3:11:52,  1.60s/it][gradnorm] step 11899: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 139.3

 62%|██████▏   | 11900/19090 [5:48:05<3:50:45,  1.93s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 200.65992736816406, 'learning_rate': 1.8672037174421453e-05, 'recon_loss': 0.0015545588248642161, 'recon_static': 0.0005793151227408089, 'recon_motion': 0.0009752436890266836, 'recon_norm': 1.2888285517692566, 'recon_norm_std': 0.042075532488524914, 'target_norm': 1.623472973704338, 'recon_cos': 0.6361069664359092, 'gradnorm/compressor.query': 0.14123842438806272, 'gradnorm/compressor.layer0': 5.0029377889781665, 'gradnorm/compressor.layer1': 197.5008504728524, 'gradnorm/compressor.layer2': 31.977898594735024, 'gradnorm/compressor.layer3': 13.04977629605018, 'gradnorm/compressor.layer4': 4.457106353100258, 'gradnorm/compressor.layer5': 2.925130570586631, 'gradnorm/compressor.layer6': 1.5879537609728802, 'gradnorm/compressor.layer7': 1.7707547637862449, 'gradnorm/decoder.layer0': 2.0026295779781687, 'gradnorm/decoder.layer1': 1.1601953467771755, 'gradnorm/decoder.layer2': 0.6638272710316325, 'gradnorm/max_param': 139.33755493164062, 'epoch': 6.23}

 62%|██████▏   | 11900/19090 [5:48:05<3:50:45,  1.93s/it]
 62%|██████▏   | 11901/19090 [5:48:06<3:27:50,  1.73s/it]
 62%|██████▏   | 11902/19090 [5:48:07<3:11:39,  1.60s/it]
 62%|██████▏   | 11903/19090 [5:48:10<3:50:19,  1.92s/it]
 62%|██████▏   | 11904/19090 [5:48:11<3:27:23,  1.73s/it]
 62%|██████▏   | 11905/19090 [5:48:12<3:11:19,  1.60s/it]
 62%|██████▏   | 11906/19090 [5:48:15<3:50:40,  1.93s/it]
 62%|██████▏   | 11907/19090 [5:48:16<3:27:43,  1.74s/it]
 62%|██████▏   | 11908/19090 [5:48:18<3:11:44,  1.60s/it]
 62%|██████▏   | 11909/19090 [5:48:20<3:49:13,  1.92s/it][gradnorm] step 11909: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 146.0

 62%|██████▏   | 11910/19090 [5:48:22<3:26:35,  1.73s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 211.00875854492188, 'learning_rate': 1.862782323383786e-05, 'recon_loss': 0.0016198277124203741, 'recon_static': 0.0006522915122332051, 'recon_motion': 0.0009675362525740638, 'recon_norm': 1.3231584548950195, 'recon_norm_std': 0.04200124917551875, 'target_norm': 1.6219855487346648, 'recon_cos': 0.6224811628460885, 'gradnorm/compressor.query': 0.17600905703715408, 'gradnorm/compressor.layer0': 6.432218292270858, 'gradnorm/compressor.layer1': 206.98901691877276, 'gradnorm/compressor.layer2': 37.43467943778035, 'gradnorm/compressor.layer3': 14.007405638437353, 'gradnorm/compressor.layer4': 5.029959756072118, 'gradnorm/compressor.layer5': 2.3441209626672355, 'gradnorm/compressor.layer6': 1.545416404386066, 'gradnorm/compressor.layer7': 1.6878195566030911, 'gradnorm/decoder.layer0': 1.894619717449594, 'gradnorm/decoder.layer1': 1.1380062611546766, 'gradnorm/decoder.layer2': 0.6011194851995756, 'gradnorm/max_param': 146.0283660888672, 'epoch': 6.24}

 62%|██████▏   | 11910/19090 [5:48:22<3:26:35,  1.73s/it]
 62%|██████▏   | 11911/19090 [5:48:23<3:11:07,  1.60s/it]
 62%|██████▏   | 11912/19090 [5:48:26<3:51:00,  1.93s/it]
 62%|██████▏   | 11913/19090 [5:48:27<3:27:52,  1.74s/it]
 62%|██████▏   | 11914/19090 [5:48:28<3:12:04,  1.61s/it]
 62%|██████▏   | 11915/19090 [5:48:31<3:41:34,  1.85s/it]
 62%|██████▏   | 11916/19090 [5:48:32<3:21:21,  1.68s/it]
 62%|██████▏   | 11917/19090 [5:48:33<3:07:06,  1.57s/it]
 62%|██████▏   | 11918/19090 [5:48:36<3:55:12,  1.97s/it]
 62%|██████▏   | 11919/19090 [5:48:37<3:30:53,  1.76s/it][gradnorm] step 11919: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 125.2

 62%|██████▏   | 11920/19090 [5:48:39<3:13:42,  1.62s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 181.50564575195312, 'learning_rate': 1.858363059869701e-05, 'recon_loss': 0.0016256226430414245, 'recon_static': 0.0006565862233401277, 'recon_motion': 0.0009690364153357222, 'recon_norm': 1.329103448987007, 'recon_norm_std': 0.04169961772859097, 'target_norm': 1.6215400606393815, 'recon_cos': 0.6214810818433761, 'gradnorm/compressor.query': 0.1575846533304819, 'gradnorm/compressor.layer0': 5.3537954753446915, 'gradnorm/compressor.layer1': 177.41601776439182, 'gradnorm/compressor.layer2': 34.99967564023808, 'gradnorm/compressor.layer3': 13.302126285699709, 'gradnorm/compressor.layer4': 4.555523489759116, 'gradnorm/compressor.layer5': 2.7413713158908912, 'gradnorm/compressor.layer6': 1.5048474029791366, 'gradnorm/compressor.layer7': 1.4954910857286634, 'gradnorm/decoder.layer0': 1.7361489923371034, 'gradnorm/decoder.layer1': 1.0812586701326656, 'gradnorm/decoder.layer2': 0.5419634287309594, 'gradnorm/max_param': 125.1795654296875, 'epoch': 6.24}

 62%|██████▏   | 11920/19090 [5:48:39<3:13:42,  1.62s/it]
 62%|██████▏   | 11921/19090 [5:48:41<3:47:29,  1.90s/it]
 62%|██████▏   | 11922/19090 [5:48:43<3:25:23,  1.72s/it]
 62%|██████▏   | 11923/19090 [5:48:44<3:09:50,  1.59s/it]
 62%|██████▏   | 11924/19090 [5:48:47<3:48:00,  1.91s/it]
 62%|██████▏   | 11925/19090 [5:48:48<3:25:40,  1.72s/it]
 62%|██████▏   | 11926/19090 [5:48:49<3:10:03,  1.59s/it]
 62%|██████▏   | 11927/19090 [5:48:52<3:48:02,  1.91s/it]
 62%|██████▏   | 11928/19090 [5:48:53<3:25:51,  1.72s/it]
 62%|██████▏   | 11929/19090 [5:48:54<3:10:35,  1.60s/it][gradnorm] step 11929: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 146.2

 62%|██████▏   | 11930/19090 [5:48:57<3:45:54,  1.89s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 211.91253662109375, 'learning_rate': 1.853945941675744e-05, 'recon_loss': 0.001652857835870236, 'recon_static': 0.0006939365252037532, 'recon_motion': 0.0009589212684659287, 'recon_norm': 1.3396527528762818, 'recon_norm_std': 0.041143900342285634, 'target_norm': 1.6194028466939927, 'recon_cos': 0.6154544487595558, 'gradnorm/compressor.query': 0.14865432763996816, 'gradnorm/compressor.layer0': 5.631273628210695, 'gradnorm/compressor.layer1': 207.19953773476428, 'gradnorm/compressor.layer2': 41.09086402651218, 'gradnorm/compressor.layer3': 14.567568441334632, 'gradnorm/compressor.layer4': 4.653643192005172, 'gradnorm/compressor.layer5': 2.4669152697303094, 'gradnorm/compressor.layer6': 1.7145909707450488, 'gradnorm/compressor.layer7': 2.0572185519460957, 'gradnorm/decoder.layer0': 2.3724901087788886, 'gradnorm/decoder.layer1': 1.3486400606815543, 'gradnorm/decoder.layer2': 0.7060020501023264, 'gradnorm/max_param': 146.1821746826172, 'epoch': 6.25}

 62%|██████▏   | 11930/19090 [5:48:57<3:45:54,  1.89s/it]
 62%|██████▏   | 11931/19090 [5:48:58<3:24:41,  1.72s/it]
 63%|██████▎   | 11932/19090 [5:49:00<3:09:21,  1.59s/it]
 63%|██████▎   | 11933/19090 [5:49:02<3:45:30,  1.89s/it]
 63%|██████▎   | 11934/19090 [5:49:03<3:23:39,  1.71s/it]
 63%|██████▎   | 11935/19090 [5:49:05<3:08:49,  1.58s/it]
 63%|██████▎   | 11936/19090 [5:49:07<3:44:57,  1.89s/it]
 63%|██████▎   | 11937/19090 [5:49:09<3:23:21,  1.71s/it]
 63%|██████▎   | 11938/19090 [5:49:10<3:08:19,  1.58s/it]
 63%|██████▎   | 11939/19090 [5:49:13<3:55:11,  1.97s/it][gradnorm] step 11939: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 141.9

 63%|██████▎   | 11940/19090 [5:49:14<3:30:37,  1.77s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 205.623291015625, 'learning_rate': 1.8495309835706013e-05, 'recon_loss': 0.0016040730784879997, 'recon_static': 0.0006337256738333963, 'recon_motion': 0.000970347406109795, 'recon_norm': 1.3174849778413773, 'recon_norm_std': 0.04079907545819879, 'target_norm': 1.6203555643558503, 'recon_cos': 0.6248906522989273, 'gradnorm/compressor.query': 0.16559088171119576, 'gradnorm/compressor.layer0': 5.739238121696058, 'gradnorm/compressor.layer1': 201.0804507063031, 'gradnorm/compressor.layer2': 39.281037128997596, 'gradnorm/compressor.layer3': 14.895161276997756, 'gradnorm/compressor.layer4': 5.304389417135873, 'gradnorm/compressor.layer5': 2.833571947559234, 'gradnorm/compressor.layer6': 1.7962023595599266, 'gradnorm/compressor.layer7': 2.0044026791666, 'gradnorm/decoder.layer0': 2.126766704590133, 'gradnorm/decoder.layer1': 1.219677254479606, 'gradnorm/decoder.layer2': 0.5851738085487039, 'gradnorm/max_param': 141.87966918945312, 'epoch': 6.25}

 63%|██████▎   | 11940/19090 [5:49:14<3:30:37,  1.77s/it]
 63%|██████▎   | 11941/19090 [5:49:15<3:13:28,  1.62s/it]
 63%|██████▎   | 11942/19090 [5:49:18<4:01:56,  2.03s/it]
 63%|██████▎   | 11943/19090 [5:49:20<3:35:25,  1.81s/it]
 63%|██████▎   | 11944/19090 [5:49:21<3:17:12,  1.66s/it]
 63%|██████▎   | 11945/19090 [5:49:24<3:57:29,  1.99s/it]
 63%|██████▎   | 11946/19090 [5:49:25<3:32:39,  1.79s/it]
 63%|██████▎   | 11947/19090 [5:49:26<3:15:16,  1.64s/it]
 63%|██████▎   | 11948/19090 [5:49:29<3:57:10,  1.99s/it]
 63%|██████▎   | 11949/19090 [5:49:30<3:32:37,  1.79s/it][gradnorm] step 11949: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 142.0

 63%|██████▎   | 11950/19090 [5:49:32<3:15:08,  1.64s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 205.6710662841797, 'learning_rate': 1.8451182003157313e-05, 'recon_loss': 0.0015845826274016872, 'recon_static': 0.0006155702838441357, 'recon_motion': 0.0009690123042673804, 'recon_norm': 1.3113274186849595, 'recon_norm_std': 0.04167212387546897, 'target_norm': 1.6215900599956512, 'recon_cos': 0.629899674654007, 'gradnorm/compressor.query': 0.16359177110927045, 'gradnorm/compressor.layer0': 5.619605805225837, 'gradnorm/compressor.layer1': 201.33297978721717, 'gradnorm/compressor.layer2': 38.07321822868488, 'gradnorm/compressor.layer3': 15.514351876109494, 'gradnorm/compressor.layer4': 4.8926056090171155, 'gradnorm/compressor.layer5': 2.629451610119208, 'gradnorm/compressor.layer6': 1.6572607303259919, 'gradnorm/compressor.layer7': 1.968708764507242, 'gradnorm/decoder.layer0': 2.054707110061185, 'gradnorm/decoder.layer1': 1.278245520097629, 'gradnorm/decoder.layer2': 0.6460704342603169, 'gradnorm/max_param': 142.03428649902344, 'epoch': 6.26}

 63%|██████▎   | 11950/19090 [5:49:32<3:15:08,  1.64s/it]
 63%|██████▎   | 11951/19090 [5:49:34<3:54:34,  1.97s/it]
 63%|██████▎   | 11952/19090 [5:49:36<3:30:04,  1.77s/it]
 63%|██████▎   | 11953/19090 [5:49:37<3:13:03,  1.62s/it]
 63%|██████▎   | 11954/19090 [5:49:40<3:59:30,  2.01s/it]
 63%|██████▎   | 11955/19090 [5:49:41<3:33:22,  1.79s/it]
 63%|██████▎   | 11956/19090 [5:49:43<3:15:27,  1.64s/it]
 63%|██████▎   | 11957/19090 [5:49:45<4:02:28,  2.04s/it]
 63%|██████▎   | 11958/19090 [5:49:47<3:35:32,  1.81s/it]
 63%|██████▎   | 11959/19090 [5:49:48<3:16:44,  1.66s/it][gradnorm] step 11959: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 160.1

 63%|██████▎   | 11960/19090 [5:49:51<3:59:10,  2.01s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 232.8195343017578, 'learning_rate': 1.8407076066653232e-05, 'recon_loss': 0.0015479450434213505, 'recon_static': 0.0005759229956311174, 'recon_motion': 0.0009720220710732974, 'recon_norm': 1.2909633994102478, 'recon_norm_std': 0.043386678025126454, 'target_norm': 1.6234747409820556, 'recon_cos': 0.6381413593888283, 'gradnorm/compressor.query': 0.1762416637931499, 'gradnorm/compressor.layer0': 5.834403493900527, 'gradnorm/compressor.layer1': 226.94141589846487, 'gradnorm/compressor.layer2': 48.33472968528944, 'gradnorm/compressor.layer3': 16.626727251086766, 'gradnorm/compressor.layer4': 5.598886188367555, 'gradnorm/compressor.layer5': 2.926201702307599, 'gradnorm/compressor.layer6': 1.8966164979964748, 'gradnorm/compressor.layer7': 1.9728943935964494, 'gradnorm/decoder.layer0': 2.5300698060368196, 'gradnorm/decoder.layer1': 1.2556455913733198, 'gradnorm/decoder.layer2': 0.6266821397441994, 'gradnorm/max_param': 160.13491821289062, 'epoch': 6.27}

 63%|██████▎   | 11960/19090 [5:49:51<3:59:10,  2.01s/it]
 63%|██████▎   | 11961/19090 [5:49:52<3:33:09,  1.79s/it]
 63%|██████▎   | 11962/19090 [5:49:53<3:15:17,  1.64s/it]
 63%|██████▎   | 11963/19090 [5:49:56<3:59:09,  2.01s/it]
 63%|██████▎   | 11964/19090 [5:49:58<3:33:09,  1.79s/it]
 63%|██████▎   | 11965/19090 [5:49:59<3:15:01,  1.64s/it]
 63%|██████▎   | 11966/19090 [5:50:01<3:44:12,  1.89s/it]
 63%|██████▎   | 11967/19090 [5:50:03<3:22:39,  1.71s/it]
 63%|██████▎   | 11968/19090 [5:50:04<3:07:39,  1.58s/it]
 63%|██████▎   | 11969/19090 [5:50:07<3:59:53,  2.02s/it][gradnorm] step 11969: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 158.7

 63%|██████▎   | 11970/19090 [5:50:08<3:33:40,  1.80s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 230.04043579101562, 'learning_rate': 1.836299217366247e-05, 'recon_loss': 0.0016195770469494164, 'recon_static': 0.0006522840703837574, 'recon_motion': 0.0009672929591033608, 'recon_norm': 1.3287732154130936, 'recon_norm_std': 0.04171201381832361, 'target_norm': 1.6186810463666916, 'recon_cos': 0.6220588520169258, 'gradnorm/compressor.query': 0.1718114930647578, 'gradnorm/compressor.layer0': 6.452182950400543, 'gradnorm/compressor.layer1': 224.97485970659034, 'gradnorm/compressor.layer2': 43.797403169754965, 'gradnorm/compressor.layer3': 17.191441309467773, 'gradnorm/compressor.layer4': 4.957621846589279, 'gradnorm/compressor.layer5': 3.1485356393668598, 'gradnorm/compressor.layer6': 1.8802133561874745, 'gradnorm/compressor.layer7': 1.9217147760360227, 'gradnorm/decoder.layer0': 2.3753621678528933, 'gradnorm/decoder.layer1': 1.3186681315243887, 'gradnorm/decoder.layer2': 0.6319718133731138, 'gradnorm/max_param': 158.738037109375, 'epoch': 6.27}

 63%|██████▎   | 11970/19090 [5:50:08<3:33:40,  1.80s/it]
 63%|██████▎   | 11971/19090 [5:50:10<3:15:31,  1.65s/it]
 63%|██████▎   | 11972/19090 [5:50:12<3:57:12,  2.00s/it]
 63%|██████▎   | 11973/19090 [5:50:14<3:31:42,  1.78s/it]
 63%|██████▎   | 11974/19090 [5:50:15<3:13:55,  1.64s/it]
 63%|██████▎   | 11975/19090 [5:50:18<3:50:28,  1.94s/it]
 63%|██████▎   | 11976/19090 [5:50:19<3:27:15,  1.75s/it]
 63%|██████▎   | 11977/19090 [5:50:20<3:10:51,  1.61s/it]
 63%|██████▎   | 11978/19090 [5:50:23<3:49:50,  1.94s/it]
 63%|██████▎   | 11979/19090 [5:50:24<3:26:35,  1.74s/it][gradnorm] step 11979: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 141.6

 63%|██████▎   | 11980/19090 [5:50:25<3:10:36,  1.61s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 204.48321533203125, 'learning_rate': 1.8318930471579986e-05, 'recon_loss': 0.001622895454056561, 'recon_static': 0.0006436165553168393, 'recon_motion': 0.000979278857994359, 'recon_norm': 1.3206814646720886, 'recon_norm_std': 0.04181240079924464, 'target_norm': 1.6242399275302888, 'recon_cos': 0.6219830393791199, 'gradnorm/compressor.query': 0.19569751937056187, 'gradnorm/compressor.layer0': 5.385219174372597, 'gradnorm/compressor.layer1': 200.83493688848063, 'gradnorm/compressor.layer2': 34.31564477847531, 'gradnorm/compressor.layer3': 15.25226963851533, 'gradnorm/compressor.layer4': 4.518355972606677, 'gradnorm/compressor.layer5': 2.8103247813895265, 'gradnorm/compressor.layer6': 1.5786944579973117, 'gradnorm/compressor.layer7': 1.689431392110938, 'gradnorm/decoder.layer0': 2.0264238979263234, 'gradnorm/decoder.layer1': 1.146800165047582, 'gradnorm/decoder.layer2': 0.6307247478814377, 'gradnorm/max_param': 141.6409149169922, 'epoch': 6.28}

 63%|██████▎   | 11980/19090 [5:50:25<3:10:36,  1.61s/it]
 63%|██████▎   | 11981/19090 [5:50:28<3:43:44,  1.89s/it]
 63%|██████▎   | 11982/19090 [5:50:29<3:22:15,  1.71s/it]
 63%|██████▎   | 11983/19090 [5:50:31<3:07:35,  1.58s/it]
 63%|██████▎   | 11984/19090 [5:50:33<3:50:09,  1.94s/it]
 63%|██████▎   | 11985/19090 [5:50:35<3:26:43,  1.75s/it]
 63%|██████▎   | 11986/19090 [5:50:36<3:10:19,  1.61s/it]
 63%|██████▎   | 11987/19090 [5:50:39<3:50:16,  1.95s/it]
 63%|██████▎   | 11988/19090 [5:50:40<3:26:46,  1.75s/it]
 63%|██████▎   | 11989/19090 [5:50:41<3:10:45,  1.61s/it][gradnorm] step 11989: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 105.9

 63%|██████▎   | 11990/19090 [5:50:44<3:50:10,  1.95s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 154.7158966064453, 'learning_rate': 1.8274891107726567e-05, 'recon_loss': 0.001611613979912363, 'recon_static': 0.0006553999934112653, 'recon_motion': 0.0009562140287016519, 'recon_norm': 1.323022612929344, 'recon_norm_std': 0.0419471500441432, 'target_norm': 1.6195791900157928, 'recon_cos': 0.6229666516184806, 'gradnorm/compressor.query': 0.18201382557721787, 'gradnorm/compressor.layer0': 5.912267455352731, 'gradnorm/compressor.layer1': 150.07863824092354, 'gradnorm/compressor.layer2': 33.76133526371643, 'gradnorm/compressor.layer3': 14.020971666912924, 'gradnorm/compressor.layer4': 4.810300249971206, 'gradnorm/compressor.layer5': 2.9966901958872816, 'gradnorm/compressor.layer6': 1.4273106563345905, 'gradnorm/compressor.layer7': 1.555999585575745, 'gradnorm/decoder.layer0': 1.9507150048315804, 'gradnorm/decoder.layer1': 1.1031122331177663, 'gradnorm/decoder.layer2': 0.6419567593929737, 'gradnorm/max_param': 105.88903045654297, 'epoch': 6.28}

 63%|██████▎   | 11990/19090 [5:50:44<3:50:10,  1.95s/it]
 63%|██████▎   | 11991/19090 [5:50:45<3:27:27,  1.75s/it]
 63%|██████▎   | 11992/19090 [5:50:47<3:11:25,  1.62s/it]
 63%|██████▎   | 11993/19090 [5:50:49<3:50:20,  1.95s/it]
 63%|██████▎   | 11994/19090 [5:50:51<3:27:16,  1.75s/it]
 63%|██████▎   | 11995/19090 [5:50:52<3:11:29,  1.62s/it]
 63%|██████▎   | 11996/19090 [5:50:55<3:53:21,  1.97s/it]
 63%|██████▎   | 11997/19090 [5:50:56<3:29:31,  1.77s/it]
 63%|██████▎   | 11998/19090 [5:50:57<3:13:07,  1.63s/it]
 63%|██████▎   | 11999/19090 [5:51:00<3:49:43,  1.94s/it][gradnorm] step 11999: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 120.6

 63%|██████▎   | 12000/19090 [5:51:01<3:26:24,  1.75s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 174.34742736816406, 'learning_rate': 1.8230874229348317e-05, 'recon_loss': 0.0016929876495851205, 'recon_static': 0.0007311905283131637, 'recon_motion': 0.0009617971256375313, 'recon_norm': 1.3607090145349503, 'recon_norm_std': 0.0397342930547893, 'target_norm': 1.618497258424759, 'recon_cos': 0.6076178342103958, 'gradnorm/compressor.query': 0.1583419414478714, 'gradnorm/compressor.layer0': 5.4688560257579235, 'gradnorm/compressor.layer1': 170.8654212750491, 'gradnorm/compressor.layer2': 31.236057436553793, 'gradnorm/compressor.layer3': 12.753614324760122, 'gradnorm/compressor.layer4': 4.197022272566694, 'gradnorm/compressor.layer5': 2.449033690644908, 'gradnorm/compressor.layer6': 1.4602226915749799, 'gradnorm/compressor.layer7': 1.521456403560878, 'gradnorm/decoder.layer0': 2.0116457435175086, 'gradnorm/decoder.layer1': 1.1425775511650536, 'gradnorm/decoder.layer2': 0.5888052381127015, 'gradnorm/max_param': 120.5530014038086, 'epoch': 6.29}

 63%|██████▎   | 12000/19090 [5:51:01<3:26:24,  1.75s/it]
 63%|██████▎   | 12001/19090 [5:51:05<4:26:24,  2.25s/it]
 63%|██████▎   | 12002/19090 [5:51:06<3:51:55,  1.96s/it]
 63%|██████▎   | 12003/19090 [5:51:07<3:28:00,  1.76s/it]
 63%|██████▎   | 12004/19090 [5:51:09<3:11:10,  1.62s/it]
 63%|██████▎   | 12005/19090 [5:51:10<2:59:21,  1.52s/it]
 63%|██████▎   | 12006/19090 [5:51:11<2:51:01,  1.45s/it]
 63%|██████▎   | 12007/19090 [5:51:12<2:45:44,  1.40s/it]
 63%|██████▎   | 12008/19090 [5:51:15<3:09:23,  1.60s/it]
 63%|██████▎   | 12009/19090 [5:51:16<2:59:06,  1.52s/it][gradnorm] step 12009: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 116.4

 63%|██████▎   | 12010/19090 [5:51:17<2:51:35,  1.45s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 168.8183135986328, 'learning_rate': 1.818687998361614e-05, 'recon_loss': 0.0017284810164710507, 'recon_static': 0.0007634947818587534, 'recon_motion': 0.0009649861953221261, 'recon_norm': 1.366986373066902, 'recon_norm_std': 0.04023155849426985, 'target_norm': 1.6206832081079483, 'recon_cos': 0.6001829579472542, 'gradnorm/compressor.query': 0.14626951354418014, 'gradnorm/compressor.layer0': 5.209680408479593, 'gradnorm/compressor.layer1': 164.9306554093265, 'gradnorm/compressor.layer2': 32.657545204578106, 'gradnorm/compressor.layer3': 12.967463932843977, 'gradnorm/compressor.layer4': 4.334693084024569, 'gradnorm/compressor.layer5': 2.4137298453549896, 'gradnorm/compressor.layer6': 1.48385209359796, 'gradnorm/compressor.layer7': 1.837866561034844, 'gradnorm/decoder.layer0': 1.932949505994559, 'gradnorm/decoder.layer1': 1.149926602053041, 'gradnorm/decoder.layer2': 0.6453229492302105, 'gradnorm/max_param': 116.36872100830078, 'epoch': 6.29}

 63%|██████▎   | 12010/19090 [5:51:17<2:51:35,  1.45s/it]
 63%|██████▎   | 12011/19090 [5:51:20<3:33:26,  1.81s/it]
 63%|██████▎   | 12012/19090 [5:51:21<3:19:42,  1.69s/it]
 63%|██████▎   | 12013/19090 [5:51:23<3:05:22,  1.57s/it]
 63%|██████▎   | 12014/19090 [5:51:25<3:35:27,  1.83s/it]
 63%|██████▎   | 12015/19090 [5:51:26<3:22:36,  1.72s/it]
 63%|██████▎   | 12016/19090 [5:51:28<3:07:36,  1.59s/it]
 63%|██████▎   | 12017/19090 [5:51:30<3:45:43,  1.91s/it]
 63%|██████▎   | 12018/19090 [5:51:32<3:23:46,  1.73s/it]
 63%|██████▎   | 12019/19090 [5:51:33<3:08:14,  1.60s/it][gradnorm] step 12019: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 126.6

 63%|██████▎   | 12020/19090 [5:51:36<3:42:55,  1.89s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 183.47412109375, 'learning_rate': 1.8142908517625277e-05, 'recon_loss': 0.001649425711366348, 'recon_static': 0.0006890629432746209, 'recon_motion': 0.000960362765181344, 'recon_norm': 1.3467524915933609, 'recon_norm_std': 0.040905925538390876, 'target_norm': 1.6179767817258834, 'recon_cos': 0.6166422829031944, 'gradnorm/compressor.query': 0.17307180177771667, 'gradnorm/compressor.layer0': 6.3216471353328565, 'gradnorm/compressor.layer1': 179.51737598697792, 'gradnorm/compressor.layer2': 34.3125462488304, 'gradnorm/compressor.layer3': 13.372693014386865, 'gradnorm/compressor.layer4': 4.80189514894738, 'gradnorm/compressor.layer5': 2.408263812882165, 'gradnorm/compressor.layer6': 1.5742426035093764, 'gradnorm/compressor.layer7': 1.6530997348289747, 'gradnorm/decoder.layer0': 2.06625417571618, 'gradnorm/decoder.layer1': 1.1576955368124666, 'gradnorm/decoder.layer2': 0.6484210919981437, 'gradnorm/max_param': 126.64531707763672, 'epoch': 6.3}

 63%|██████▎   | 12020/19090 [5:51:36<3:42:55,  1.89s/it]
 63%|██████▎   | 12021/19090 [5:51:37<3:21:39,  1.71s/it]
 63%|██████▎   | 12022/19090 [5:51:38<3:06:38,  1.58s/it]
 63%|██████▎   | 12023/19090 [5:51:41<3:49:24,  1.95s/it]
 63%|██████▎   | 12024/19090 [5:51:42<3:26:02,  1.75s/it]
 63%|██████▎   | 12025/19090 [5:51:43<3:09:32,  1.61s/it]
 63%|██████▎   | 12026/19090 [5:51:46<3:40:39,  1.87s/it]
 63%|██████▎   | 12027/19090 [5:51:47<3:19:58,  1.70s/it]
 63%|██████▎   | 12028/19090 [5:51:49<3:05:20,  1.57s/it]
 63%|██████▎   | 12029/19090 [5:51:52<3:55:10,  2.00s/it][gradnorm] step 12029: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 145.3

 63%|██████▎   | 12030/19090 [5:51:53<3:30:06,  1.79s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 210.1899871826172, 'learning_rate': 1.809895997839482e-05, 'recon_loss': 0.0016845045378431679, 'recon_static': 0.0007266366606927477, 'recon_motion': 0.000957867932447698, 'recon_norm': 1.3537718683481217, 'recon_norm_std': 0.04126190850511193, 'target_norm': 1.6205190241336822, 'recon_cos': 0.6089243218302727, 'gradnorm/compressor.query': 0.20774355296748181, 'gradnorm/compressor.layer0': 8.477611902207668, 'gradnorm/compressor.layer1': 205.88183997441348, 'gradnorm/compressor.layer2': 38.285811271625974, 'gradnorm/compressor.layer3': 14.57684484096535, 'gradnorm/compressor.layer4': 4.662989711867111, 'gradnorm/compressor.layer5': 2.9379642200927027, 'gradnorm/compressor.layer6': 1.5701166287006096, 'gradnorm/compressor.layer7': 1.9592389657285558, 'gradnorm/decoder.layer0': 2.00152684581248, 'gradnorm/decoder.layer1': 1.095909330442739, 'gradnorm/decoder.layer2': 0.6417048122990641, 'gradnorm/max_param': 145.25244140625, 'epoch': 6.3}

 63%|██████▎   | 12030/19090 [5:51:53<3:30:06,  1.79s/it]
 63%|██████▎   | 12031/19090 [5:51:54<3:12:31,  1.64s/it]
 63%|██████▎   | 12032/19090 [5:51:57<3:50:06,  1.96s/it]
 63%|██████▎   | 12033/19090 [5:51:58<3:26:24,  1.75s/it]
 63%|██████▎   | 12034/19090 [5:51:59<3:09:49,  1.61s/it]
 63%|██████▎   | 12035/19090 [5:52:02<3:46:29,  1.93s/it]
 63%|██████▎   | 12036/19090 [5:52:03<3:23:49,  1.73s/it]
 63%|██████▎   | 12037/19090 [5:52:05<3:08:17,  1.60s/it]
 63%|██████▎   | 12038/19090 [5:52:07<3:42:58,  1.90s/it]
 63%|██████▎   | 12039/19090 [5:52:08<3:21:19,  1.71s/it][gradnorm] step 12039: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 156.8

 63%|██████▎   | 12040/19090 [5:52:10<3:06:13,  1.58s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 226.3523406982422, 'learning_rate': 1.8055034512867176e-05, 'recon_loss': 0.0017271562363021075, 'recon_static': 0.0007606006489368156, 'recon_motion': 0.0009665556426625699, 'recon_norm': 1.3719773590564728, 'recon_norm_std': 0.04069722769781947, 'target_norm': 1.6204716235399246, 'recon_cos': 0.6015677362680435, 'gradnorm/compressor.query': 0.18004072756751524, 'gradnorm/compressor.layer0': 6.2823955952553, 'gradnorm/compressor.layer1': 222.2260075632463, 'gradnorm/compressor.layer2': 39.5107144660431, 'gradnorm/compressor.layer3': 14.195710751327372, 'gradnorm/compressor.layer4': 5.319601002982519, 'gradnorm/compressor.layer5': 2.830311135755768, 'gradnorm/compressor.layer6': 1.6561590925495637, 'gradnorm/compressor.layer7': 1.8032705185480562, 'gradnorm/decoder.layer0': 2.176370373178982, 'gradnorm/decoder.layer1': 1.200262752218106, 'gradnorm/decoder.layer2': 0.6202570239418547, 'gradnorm/max_param': 156.778076171875, 'epoch': 6.31}

 63%|██████▎   | 12040/19090 [5:52:10<3:06:13,  1.58s/it]
 63%|██████▎   | 12041/19090 [5:52:13<3:57:52,  2.02s/it]
 63%|██████▎   | 12042/19090 [5:52:14<3:35:46,  1.84s/it]
 63%|██████▎   | 12043/19090 [5:52:15<3:16:14,  1.67s/it]
 63%|██████▎   | 12044/19090 [5:52:18<3:55:51,  2.01s/it]
 63%|██████▎   | 12045/19090 [5:52:20<3:30:33,  1.79s/it]
 63%|██████▎   | 12046/19090 [5:52:21<3:12:49,  1.64s/it]
 63%|██████▎   | 12047/19090 [5:52:24<4:03:00,  2.07s/it]
 63%|██████▎   | 12048/19090 [5:52:25<3:35:25,  1.84s/it]
 63%|██████▎   | 12049/19090 [5:52:27<3:16:14,  1.67s/it][gradnorm] step 12049: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 148.9

 63%|██████▎   | 12050/19090 [5:52:29<3:59:18,  2.04s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 214.79345703125, 'learning_rate': 1.8011132267907623e-05, 'recon_loss': 0.0016242634359514342, 'recon_static': 0.0006578541098861024, 'recon_motion': 0.0009664093013270758, 'recon_norm': 1.3289769440889359, 'recon_norm_std': 0.04138678070157766, 'target_norm': 1.6201663821935655, 'recon_cos': 0.6206816658377647, 'gradnorm/compressor.query': 0.1716151549098714, 'gradnorm/compressor.layer0': 5.719737150115051, 'gradnorm/compressor.layer1': 211.06017728955882, 'gradnorm/compressor.layer2': 36.231801062093574, 'gradnorm/compressor.layer3': 14.080360836260443, 'gradnorm/compressor.layer4': 5.2416255098232165, 'gradnorm/compressor.layer5': 2.6509754544866175, 'gradnorm/compressor.layer6': 1.619597256718953, 'gradnorm/compressor.layer7': 1.7128798376807202, 'gradnorm/decoder.layer0': 2.0670589577175043, 'gradnorm/decoder.layer1': 1.1718390395686975, 'gradnorm/decoder.layer2': 0.6205719249192889, 'gradnorm/max_param': 148.9285430908203, 'epoch': 6.31}

 63%|██████▎   | 12050/19090 [5:52:29<3:59:18,  2.04s/it]
 63%|██████▎   | 12051/19090 [5:52:31<3:33:04,  1.82s/it]
 63%|██████▎   | 12052/19090 [5:52:32<3:14:29,  1.66s/it]
 63%|██████▎   | 12053/19090 [5:52:35<3:55:50,  2.01s/it]
 63%|██████▎   | 12054/19090 [5:52:36<3:30:35,  1.80s/it]
 63%|██████▎   | 12055/19090 [5:52:37<3:12:34,  1.64s/it]
 63%|██████▎   | 12056/19090 [5:52:40<4:00:05,  2.05s/it]
 63%|██████▎   | 12057/19090 [5:52:42<3:33:22,  1.82s/it]
 63%|██████▎   | 12058/19090 [5:52:43<3:14:46,  1.66s/it]
 63%|██████▎   | 12059/19090 [5:52:46<3:51:24,  1.97s/it][gradnorm] step 12059: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 132.6

 63%|██████▎   | 12060/19090 [5:52:47<3:27:18,  1.77s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 192.49795532226562, 'learning_rate': 1.7967253390303805e-05, 'recon_loss': 0.0016296541405608877, 'recon_static': 0.0006645031709922478, 'recon_motion': 0.0009651509972172789, 'recon_norm': 1.3272176772356032, 'recon_norm_std': 0.041656048595905305, 'target_norm': 1.6201044380664826, 'recon_cos': 0.6188417539000511, 'gradnorm/compressor.query': 0.1598489592981423, 'gradnorm/compressor.layer0': 6.275328620725428, 'gradnorm/compressor.layer1': 187.9435690739377, 'gradnorm/compressor.layer2': 38.41829582505662, 'gradnorm/compressor.layer3': 13.227729939618088, 'gradnorm/compressor.layer4': 4.841662935571771, 'gradnorm/compressor.layer5': 2.7157103109183733, 'gradnorm/compressor.layer6': 1.5229341115443795, 'gradnorm/compressor.layer7': 1.7395877343158292, 'gradnorm/decoder.layer0': 2.0608516232730034, 'gradnorm/decoder.layer1': 1.2381590292508442, 'gradnorm/decoder.layer2': 0.6325694665343335, 'gradnorm/max_param': 132.60345458984375, 'epoch': 6.32}

 63%|██████▎   | 12060/19090 [5:52:47<3:27:18,  1.77s/it]
 63%|██████▎   | 12061/19090 [5:52:48<3:10:17,  1.62s/it]
 63%|██████▎   | 12062/19090 [5:52:51<3:55:01,  2.01s/it]
 63%|██████▎   | 12063/19090 [5:52:52<3:29:36,  1.79s/it]
 63%|██████▎   | 12064/19090 [5:52:54<3:11:46,  1.64s/it]
 63%|██████▎   | 12065/19090 [5:52:56<3:49:46,  1.96s/it]
 63%|██████▎   | 12066/19090 [5:52:58<3:25:53,  1.76s/it]
 63%|██████▎   | 12067/19090 [5:52:59<3:09:12,  1.62s/it]
 63%|██████▎   | 12068/19090 [5:53:02<3:51:35,  1.98s/it]
 63%|██████▎   | 12069/19090 [5:53:03<3:27:16,  1.77s/it][gradnorm] step 12069: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 130.5

 63%|██████▎   | 12070/19090 [5:53:04<3:10:13,  1.63s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 188.83505249023438, 'learning_rate': 1.792339802676522e-05, 'recon_loss': 0.0016673656296916305, 'recon_static': 0.0007025287952274085, 'recon_motion': 0.0009648368490161374, 'recon_norm': 1.3486163586378097, 'recon_norm_std': 0.0401817811653018, 'target_norm': 1.619781967997551, 'recon_cos': 0.6129113033413887, 'gradnorm/compressor.query': 0.15021535642935938, 'gradnorm/compressor.layer0': 5.151073284714271, 'gradnorm/compressor.layer1': 184.96639263998745, 'gradnorm/compressor.layer2': 34.57427855074763, 'gradnorm/compressor.layer3': 13.473223657955979, 'gradnorm/compressor.layer4': 4.689198300747725, 'gradnorm/compressor.layer5': 2.8310978054883096, 'gradnorm/compressor.layer6': 1.606003646661174, 'gradnorm/compressor.layer7': 1.7719936458955943, 'gradnorm/decoder.layer0': 2.1483102101851563, 'gradnorm/decoder.layer1': 1.3560215155461526, 'gradnorm/decoder.layer2': 0.6792580354894094, 'gradnorm/max_param': 130.50180053710938, 'epoch': 6.32}

 63%|██████▎   | 12070/19090 [5:53:04<3:10:13,  1.63s/it]
 63%|██████▎   | 12071/19090 [5:53:07<3:57:35,  2.03s/it]
 63%|██████▎   | 12072/19090 [5:53:09<3:31:45,  1.81s/it]
 63%|██████▎   | 12073/19090 [5:53:10<3:13:28,  1.65s/it]
 63%|██████▎   | 12074/19090 [5:53:13<3:51:15,  1.98s/it]
 63%|██████▎   | 12075/19090 [5:53:14<3:27:38,  1.78s/it]
 63%|██████▎   | 12076/19090 [5:53:15<3:10:25,  1.63s/it]
 63%|██████▎   | 12077/19090 [5:53:18<3:58:14,  2.04s/it]
 63%|██████▎   | 12078/19090 [5:53:20<3:31:52,  1.81s/it]
 63%|██████▎   | 12079/19090 [5:53:21<3:13:50,  1.66s/it][gradnorm] step 12079: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 114.8

 63%|██████▎   | 12080/19090 [5:53:24<4:00:25,  2.06s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 167.7396240234375, 'learning_rate': 1.7879566323922754e-05, 'recon_loss': 0.001595835149055347, 'recon_static': 0.0006261273680138402, 'recon_motion': 0.0009697078392491675, 'recon_norm': 1.31081320643425, 'recon_norm_std': 0.04085584739223123, 'target_norm': 1.6212606072425841, 'recon_cos': 0.6264903590083122, 'gradnorm/compressor.query': 0.16527941655789968, 'gradnorm/compressor.layer0': 4.90971436860378, 'gradnorm/compressor.layer1': 162.78443856063146, 'gradnorm/compressor.layer2': 37.39476118472346, 'gradnorm/compressor.layer3': 12.890394174069721, 'gradnorm/compressor.layer4': 5.307778807480642, 'gradnorm/compressor.layer5': 2.782068539299641, 'gradnorm/compressor.layer6': 1.4811049732403176, 'gradnorm/compressor.layer7': 1.9035503469612085, 'gradnorm/decoder.layer0': 2.243666423633052, 'gradnorm/decoder.layer1': 1.3863637550399435, 'gradnorm/decoder.layer2': 0.6796260016670584, 'gradnorm/max_param': 114.82144165039062, 'epoch': 6.33}

 63%|██████▎   | 12080/19090 [5:53:24<4:00:25,  2.06s/it]
 63%|██████▎   | 12081/19090 [5:53:25<3:33:30,  1.83s/it]
 63%|██████▎   | 12082/19090 [5:53:26<3:14:42,  1.67s/it]
 63%|██████▎   | 12083/19090 [5:53:29<3:57:39,  2.04s/it]
 63%|██████▎   | 12084/19090 [5:53:31<3:31:37,  1.81s/it]
 63%|██████▎   | 12085/19090 [5:53:32<3:13:04,  1.65s/it]
 63%|██████▎   | 12086/19090 [5:53:34<3:43:08,  1.91s/it]
 63%|██████▎   | 12087/19090 [5:53:36<3:21:07,  1.72s/it]
 63%|██████▎   | 12088/19090 [5:53:37<3:05:40,  1.59s/it]
 63%|██████▎   | 12089/19090 [5:53:40<3:42:10,  1.90s/it][gradnorm] step 12089: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 184.6

 63%|██████▎   | 12090/19090 [5:53:41<3:20:29,  1.72s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 266.6330261230469, 'learning_rate': 1.78357584283282e-05, 'recon_loss': 0.0016037468303693458, 'recon_static': 0.0006392433759174309, 'recon_motion': 0.000964503396244254, 'recon_norm': 1.3274413883686065, 'recon_norm_std': 0.04104497507214546, 'target_norm': 1.6186983168125153, 'recon_cos': 0.6257080927491188, 'gradnorm/compressor.query': 0.22372790984110272, 'gradnorm/compressor.layer0': 7.327006500711912, 'gradnorm/compressor.layer1': 261.6718782655102, 'gradnorm/compressor.layer2': 47.11012126274592, 'gradnorm/compressor.layer3': 17.0966731889821, 'gradnorm/compressor.layer4': 5.587764158554473, 'gradnorm/compressor.layer5': 3.107793818092102, 'gradnorm/compressor.layer6': 1.579131909929864, 'gradnorm/compressor.layer7': 2.16080689935469, 'gradnorm/decoder.layer0': 2.3078631747395884, 'gradnorm/decoder.layer1': 1.3253283507683913, 'gradnorm/decoder.layer2': 0.6920019266991998, 'gradnorm/max_param': 184.56332397460938, 'epoch': 6.33}

 63%|██████▎   | 12090/19090 [5:53:41<3:20:29,  1.72s/it]
 63%|██████▎   | 12091/19090 [5:53:42<3:05:14,  1.59s/it]
 63%|██████▎   | 12092/19090 [5:53:45<3:44:54,  1.93s/it]
 63%|██████▎   | 12093/19090 [5:53:46<3:22:19,  1.73s/it]
 63%|██████▎   | 12094/19090 [5:53:47<3:06:29,  1.60s/it]
 63%|██████▎   | 12095/19090 [5:53:50<3:42:57,  1.91s/it]
 63%|██████▎   | 12096/19090 [5:53:51<3:20:59,  1.72s/it]
 63%|██████▎   | 12097/19090 [5:53:53<3:05:33,  1.59s/it]
 63%|██████▎   | 12098/19090 [5:53:56<3:57:14,  2.04s/it]
 63%|██████▎   | 12099/19090 [5:53:57<3:31:07,  1.81s/it][gradnorm] step 12099: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 142.7

 63%|██████▎   | 12100/19090 [5:53:58<3:12:54,  1.66s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 206.45413208007812, 'learning_rate': 1.7791974486453726e-05, 'recon_loss': 0.001593000630964525, 'recon_static': 0.000626810958783608, 'recon_motion': 0.0009661896590841934, 'recon_norm': 1.3104702889919282, 'recon_norm_std': 0.042466895654797554, 'target_norm': 1.6227414458990097, 'recon_cos': 0.627744123339653, 'gradnorm/compressor.query': 0.20883702047620278, 'gradnorm/compressor.layer0': 7.75621226729591, 'gradnorm/compressor.layer1': 202.25505053953535, 'gradnorm/compressor.layer2': 37.748617589029834, 'gradnorm/compressor.layer3': 13.198693898635684, 'gradnorm/compressor.layer4': 5.500267369100506, 'gradnorm/compressor.layer5': 3.3344717624856943, 'gradnorm/compressor.layer6': 1.6674061167995198, 'gradnorm/compressor.layer7': 2.238111346388787, 'gradnorm/decoder.layer0': 2.2388040887293337, 'gradnorm/decoder.layer1': 1.416505252767802, 'gradnorm/decoder.layer2': 0.8093324686901537, 'gradnorm/max_param': 142.69602966308594, 'epoch': 6.34}

 63%|██████▎   | 12100/19090 [5:53:58<3:12:54,  1.66s/it]
 63%|██████▎   | 12101/19090 [5:54:01<3:47:49,  1.96s/it]
 63%|██████▎   | 12102/19090 [5:54:02<3:24:33,  1.76s/it]
 63%|██████▎   | 12103/19090 [5:54:04<3:08:06,  1.62s/it]
 63%|██████▎   | 12104/19090 [5:54:06<3:44:50,  1.93s/it]
 63%|██████▎   | 12105/19090 [5:54:08<3:22:16,  1.74s/it]
 63%|██████▎   | 12106/19090 [5:54:09<3:06:39,  1.60s/it]
 63%|██████▎   | 12107/19090 [5:54:12<3:43:33,  1.92s/it]
 63%|██████▎   | 12108/19090 [5:54:13<3:21:55,  1.74s/it]
 63%|██████▎   | 12109/19090 [5:54:14<3:06:46,  1.61s/it][gradnorm] step 12109: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 131.3

 63%|██████▎   | 12110/19090 [5:54:17<3:36:13,  1.86s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 190.7080841064453, 'learning_rate': 1.7748214644691412e-05, 'recon_loss': 0.0016718985774787144, 'recon_static': 0.0007047635779599659, 'recon_motion': 0.0009671349602285772, 'recon_norm': 1.346320351958275, 'recon_norm_std': 0.04133232850581407, 'target_norm': 1.621558490395546, 'recon_cos': 0.6126825407147407, 'gradnorm/compressor.query': 0.1595287766204153, 'gradnorm/compressor.layer0': 5.211418804491781, 'gradnorm/compressor.layer1': 186.15386570992825, 'gradnorm/compressor.layer2': 37.9457620739787, 'gradnorm/compressor.layer3': 14.500337267932393, 'gradnorm/compressor.layer4': 4.546053572977216, 'gradnorm/compressor.layer5': 2.656763363765189, 'gradnorm/compressor.layer6': 1.6146018324581821, 'gradnorm/compressor.layer7': 1.791912158145699, 'gradnorm/decoder.layer0': 1.936206570477484, 'gradnorm/decoder.layer1': 1.159494281812273, 'gradnorm/decoder.layer2': 0.5943597373246274, 'gradnorm/max_param': 131.34066772460938, 'epoch': 6.34}

 63%|██████▎   | 12110/19090 [5:54:17<3:36:13,  1.86s/it]
 63%|██████▎   | 12111/19090 [5:54:18<3:16:29,  1.69s/it]
 63%|██████▎   | 12112/19090 [5:54:19<3:02:53,  1.57s/it]
 63%|██████▎   | 12113/19090 [5:54:22<3:44:21,  1.93s/it]
 63%|██████▎   | 12114/19090 [5:54:23<3:22:25,  1.74s/it]
 63%|██████▎   | 12115/19090 [5:54:25<3:07:33,  1.61s/it]
 63%|██████▎   | 12116/19090 [5:54:27<3:51:01,  1.99s/it]
 63%|██████▎   | 12117/19090 [5:54:29<3:27:08,  1.78s/it]
 63%|██████▎   | 12118/19090 [5:54:30<3:10:19,  1.64s/it]
 63%|██████▎   | 12119/19090 [5:54:33<3:56:34,  2.04s/it][gradnorm] step 12119: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 125.0

 63%|██████▎   | 12120/19090 [5:54:34<3:31:27,  1.82s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 180.57627868652344, 'learning_rate': 1.7704479049352787e-05, 'recon_loss': 0.0016474550066050142, 'recon_static': 0.0006760685821063817, 'recon_motion': 0.0009713863764773123, 'recon_norm': 1.3304379463195801, 'recon_norm_std': 0.04132164111360907, 'target_norm': 1.6230337411165237, 'recon_cos': 0.6165407299995422, 'gradnorm/compressor.query': 0.14256312017000694, 'gradnorm/compressor.layer0': 4.855021105483722, 'gradnorm/compressor.layer1': 177.20061128739087, 'gradnorm/compressor.layer2': 31.13768194352955, 'gradnorm/compressor.layer3': 13.134995496362118, 'gradnorm/compressor.layer4': 4.845012475658058, 'gradnorm/compressor.layer5': 2.6151577038740115, 'gradnorm/compressor.layer6': 1.6790906688571539, 'gradnorm/compressor.layer7': 1.6949141777164467, 'gradnorm/decoder.layer0': 2.0335504270050357, 'gradnorm/decoder.layer1': 1.2308217828906418, 'gradnorm/decoder.layer2': 0.6392863639331997, 'gradnorm/max_param': 125.0189437866211, 'epoch': 6.35}

 63%|██████▎   | 12120/19090 [5:54:34<3:31:27,  1.82s/it]
 63%|██████▎   | 12121/19090 [5:54:36<3:13:23,  1.67s/it]
 63%|██████▎   | 12122/19090 [5:54:38<3:54:38,  2.02s/it]
 64%|██████▎   | 12123/19090 [5:54:40<3:29:01,  1.80s/it]
 64%|██████▎   | 12124/19090 [5:54:41<3:11:01,  1.65s/it]
 64%|██████▎   | 12125/19090 [5:54:44<3:43:25,  1.92s/it]
 64%|██████▎   | 12126/19090 [5:54:45<3:21:14,  1.73s/it]
 64%|██████▎   | 12127/19090 [5:54:46<3:05:35,  1.60s/it]
 64%|██████▎   | 12128/19090 [5:54:49<3:47:36,  1.96s/it]
 64%|██████▎   | 12129/19090 [5:54:50<3:24:17,  1.76s/it][gradnorm] step 12129: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 141.0

 64%|██████▎   | 12130/19090 [5:54:52<3:08:21,  1.62s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 204.25534057617188, 'learning_rate': 1.766076784666828e-05, 'recon_loss': 0.0018466203269781544, 'recon_static': 0.0008809312785160728, 'recon_motion': 0.0009656890892074443, 'recon_norm': 1.4249631643295289, 'recon_norm_std': 0.03893172452226281, 'target_norm': 1.6196245551109314, 'recon_cos': 0.5819122970104218, 'gradnorm/compressor.query': 0.1607849006405205, 'gradnorm/compressor.layer0': 5.035678975484897, 'gradnorm/compressor.layer1': 199.90169459011597, 'gradnorm/compressor.layer2': 38.84276635419876, 'gradnorm/compressor.layer3': 13.616348179454748, 'gradnorm/compressor.layer4': 4.511758911482505, 'gradnorm/compressor.layer5': 2.823526764616006, 'gradnorm/compressor.layer6': 1.6722560697842594, 'gradnorm/compressor.layer7': 1.7759748280347987, 'gradnorm/decoder.layer0': 1.9715430414047086, 'gradnorm/decoder.layer1': 1.2037504972969841, 'gradnorm/decoder.layer2': 0.6245245794264995, 'gradnorm/max_param': 141.0135040283203, 'epoch': 6.35}

 64%|██████▎   | 12130/19090 [5:54:52<3:08:21,  1.62s/it]
 64%|██████▎   | 12131/19090 [5:54:54<3:45:10,  1.94s/it]
 64%|██████▎   | 12132/19090 [5:54:56<3:22:37,  1.75s/it]
 64%|██████▎   | 12133/19090 [5:54:57<3:07:03,  1.61s/it]
 64%|██████▎   | 12134/19090 [5:55:00<3:46:22,  1.95s/it]
 64%|██████▎   | 12135/19090 [5:55:01<3:23:30,  1.76s/it]
 64%|██████▎   | 12136/19090 [5:55:02<3:07:57,  1.62s/it]
 64%|██████▎   | 12137/19090 [5:55:05<4:00:32,  2.08s/it]
 64%|██████▎   | 12138/19090 [5:55:07<3:33:44,  1.84s/it]
 64%|██████▎   | 12139/19090 [5:55:08<3:14:42,  1.68s/it][gradnorm] step 12139: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 121.4

 64%|██████▎   | 12140/19090 [5:55:11<4:01:47,  2.09s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 176.8892364501953, 'learning_rate': 1.761708118278677e-05, 'recon_loss': 0.0017581816180609166, 'recon_static': 0.0007950769228045829, 'recon_motion': 0.0009631047214497812, 'recon_norm': 1.384049853682518, 'recon_norm_std': 0.04006349546834827, 'target_norm': 1.6207875102758407, 'recon_cos': 0.596670101583004, 'gradnorm/compressor.query': 0.14181912250279066, 'gradnorm/compressor.layer0': 4.883525924443518, 'gradnorm/compressor.layer1': 172.05622264248393, 'gradnorm/compressor.layer2': 38.290676454201126, 'gradnorm/compressor.layer3': 12.348657796102149, 'gradnorm/compressor.layer4': 4.789269995122601, 'gradnorm/compressor.layer5': 2.9481406154870027, 'gradnorm/compressor.layer6': 1.7319477727076775, 'gradnorm/compressor.layer7': 1.8166942081055248, 'gradnorm/decoder.layer0': 2.03764826229318, 'gradnorm/decoder.layer1': 1.193575775454585, 'gradnorm/decoder.layer2': 0.6496097355954562, 'gradnorm/max_param': 121.4048843383789, 'epoch': 6.36}

 64%|██████▎   | 12140/19090 [5:55:11<4:01:47,  2.09s/it]
 64%|██████▎   | 12141/19090 [5:55:12<3:34:02,  1.85s/it]
 64%|██████▎   | 12142/19090 [5:55:14<3:14:28,  1.68s/it]
 64%|██████▎   | 12143/19090 [5:55:16<3:41:38,  1.91s/it]
 64%|██████▎   | 12144/19090 [5:55:17<3:20:15,  1.73s/it]
 64%|██████▎   | 12145/19090 [5:55:19<3:04:56,  1.60s/it]
 64%|██████▎   | 12146/19090 [5:55:22<3:52:11,  2.01s/it]
 64%|██████▎   | 12147/19090 [5:55:23<3:27:05,  1.79s/it]
 64%|██████▎   | 12148/19090 [5:55:24<3:09:48,  1.64s/it]
 64%|██████▎   | 12149/19090 [5:55:27<3:50:43,  1.99s/it][gradnorm] step 12149: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 135.1

 64%|██████▎   | 12150/19090 [5:55:28<3:26:10,  1.78s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 195.74363708496094, 'learning_rate': 1.7573419203775116e-05, 'recon_loss': 0.001631282173912041, 'recon_static': 0.0006710561210638843, 'recon_motion': 0.000960226081951987, 'recon_norm': 1.3293487071990966, 'recon_norm_std': 0.04095111694186926, 'target_norm': 1.620352205634117, 'recon_cos': 0.6190753057599068, 'gradnorm/compressor.query': 0.14955541369639436, 'gradnorm/compressor.layer0': 5.500874709885014, 'gradnorm/compressor.layer1': 191.4127983573721, 'gradnorm/compressor.layer2': 37.73561236325293, 'gradnorm/compressor.layer3': 13.417722545958107, 'gradnorm/compressor.layer4': 4.827733885098891, 'gradnorm/compressor.layer5': 2.522907402174866, 'gradnorm/compressor.layer6': 1.8000125858078806, 'gradnorm/compressor.layer7': 1.8347740220252748, 'gradnorm/decoder.layer0': 2.048015131350588, 'gradnorm/decoder.layer1': 1.1807078817990222, 'gradnorm/decoder.layer2': 0.7426899865351279, 'gradnorm/max_param': 135.0686798095703, 'epoch': 6.36}

 64%|██████▎   | 12150/19090 [5:55:28<3:26:10,  1.78s/it]
 64%|██████▎   | 12151/19090 [5:55:30<3:09:09,  1.64s/it]
 64%|██████▎   | 12152/19090 [5:55:32<3:49:02,  1.98s/it]
 64%|██████▎   | 12153/19090 [5:55:34<3:25:04,  1.77s/it]
 64%|██████▎   | 12154/19090 [5:55:35<3:08:04,  1.63s/it]
 64%|██████▎   | 12155/19090 [5:55:38<3:48:24,  1.98s/it]
 64%|██████▎   | 12156/19090 [5:55:39<3:24:32,  1.77s/it]
 64%|██████▎   | 12157/19090 [5:55:40<3:07:39,  1.62s/it]
 64%|██████▎   | 12158/19090 [5:55:43<3:35:57,  1.87s/it]
 64%|██████▎   | 12159/19090 [5:55:44<3:15:41,  1.69s/it][gradnorm] step 12159: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 160.1

 64%|██████▎   | 12160/19090 [5:55:45<3:01:33,  1.57s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 233.3560028076172, 'learning_rate': 1.7529782055617607e-05, 'recon_loss': 0.0016996547143207862, 'recon_static': 0.0007457221639924682, 'recon_motion': 0.0009539325343212113, 'recon_norm': 1.364142230153084, 'recon_norm_std': 0.04136216873303056, 'target_norm': 1.617125204205513, 'recon_cos': 0.6054995685815812, 'gradnorm/compressor.query': 0.18849725560257208, 'gradnorm/compressor.layer0': 6.697953085160112, 'gradnorm/compressor.layer1': 226.93808523747177, 'gradnorm/compressor.layer2': 50.66169821488049, 'gradnorm/compressor.layer3': 17.10919217225376, 'gradnorm/compressor.layer4': 5.234877468303751, 'gradnorm/compressor.layer5': 3.169166949828265, 'gradnorm/compressor.layer6': 1.7267875697300237, 'gradnorm/compressor.layer7': 1.915641040330351, 'gradnorm/decoder.layer0': 1.9938517481751676, 'gradnorm/decoder.layer1': 1.1807544758825135, 'gradnorm/decoder.layer2': 0.6593196239992405, 'gradnorm/max_param': 160.12950134277344, 'epoch': 6.37}

 64%|██████▎   | 12160/19090 [5:55:45<3:01:33,  1.57s/it]
 64%|██████▎   | 12161/19090 [5:55:48<3:40:18,  1.91s/it]
 64%|██████▎   | 12162/19090 [5:55:49<3:18:34,  1.72s/it]
 64%|██████▎   | 12163/19090 [5:55:51<3:03:42,  1.59s/it]
 64%|██████▎   | 12164/19090 [5:55:54<3:55:24,  2.04s/it]
 64%|██████▎   | 12165/19090 [5:55:55<3:29:21,  1.81s/it]
 64%|██████▎   | 12166/19090 [5:55:56<3:11:04,  1.66s/it]
 64%|██████▎   | 12167/19090 [5:55:59<3:57:02,  2.05s/it]
 64%|██████▎   | 12168/19090 [5:56:00<3:30:36,  1.83s/it]
 64%|██████▎   | 12169/19090 [5:56:02<3:12:09,  1.67s/it][gradnorm] step 12169: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 155.8

 64%|██████▍   | 12170/19090 [5:56:05<3:56:32,  2.05s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 226.2707061767578, 'learning_rate': 1.748616988421552e-05, 'recon_loss': 0.001605421499698423, 'recon_static': 0.0006429608474718407, 'recon_motion': 0.0009624607162550091, 'recon_norm': 1.3222571820020677, 'recon_norm_std': 0.041613736748695375, 'target_norm': 1.6202796965837478, 'recon_cos': 0.6251661866903305, 'gradnorm/compressor.query': 0.16250616369112003, 'gradnorm/compressor.layer0': 5.926657323382937, 'gradnorm/compressor.layer1': 220.827421304511, 'gradnorm/compressor.layer2': 46.151073994646914, 'gradnorm/compressor.layer3': 14.613356952323297, 'gradnorm/compressor.layer4': 5.6632842272417685, 'gradnorm/compressor.layer5': 2.977889118279766, 'gradnorm/compressor.layer6': 1.7740574899924924, 'gradnorm/compressor.layer7': 1.948238760961899, 'gradnorm/decoder.layer0': 2.2925383441676086, 'gradnorm/decoder.layer1': 1.1989406300465948, 'gradnorm/decoder.layer2': 0.6912439231191554, 'gradnorm/max_param': 155.7969970703125, 'epoch': 6.38}

 64%|██████▍   | 12170/19090 [5:56:05<3:56:32,  2.05s/it]
 64%|██████▍   | 12171/19090 [5:56:06<3:30:24,  1.82s/it]
 64%|██████▍   | 12172/19090 [5:56:07<3:11:57,  1.66s/it]
 64%|██████▍   | 12173/19090 [5:56:10<3:38:37,  1.90s/it]
 64%|██████▍   | 12174/19090 [5:56:11<3:17:32,  1.71s/it]
 64%|██████▍   | 12175/19090 [5:56:12<3:03:03,  1.59s/it]
 64%|██████▍   | 12176/19090 [5:56:15<3:43:19,  1.94s/it]
 64%|██████▍   | 12177/19090 [5:56:16<3:20:46,  1.74s/it]
 64%|██████▍   | 12178/19090 [5:56:18<3:05:15,  1.61s/it]
 64%|██████▍   | 12179/19090 [5:56:21<3:52:01,  2.01s/it][gradnorm] step 12179: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 142.5

 64%|██████▍   | 12180/19090 [5:56:22<3:27:10,  1.80s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 206.19503784179688, 'learning_rate': 1.7442582835386646e-05, 'recon_loss': 0.0016143679153174163, 'recon_static': 0.0006543632611283101, 'recon_motion': 0.0009600046352716163, 'recon_norm': 1.3219066977500915, 'recon_norm_std': 0.04178093057125807, 'target_norm': 1.6207011699676515, 'recon_cos': 0.6228328987956047, 'gradnorm/compressor.query': 0.16711656500573036, 'gradnorm/compressor.layer0': 5.484594617831681, 'gradnorm/compressor.layer1': 201.9346274168945, 'gradnorm/compressor.layer2': 37.82276946031897, 'gradnorm/compressor.layer3': 15.182419417701515, 'gradnorm/compressor.layer4': 5.010326022506583, 'gradnorm/compressor.layer5': 3.009036917229579, 'gradnorm/compressor.layer6': 1.7428241473118764, 'gradnorm/compressor.layer7': 2.042846792749334, 'gradnorm/decoder.layer0': 2.10098326688092, 'gradnorm/decoder.layer1': 1.1859151905961927, 'gradnorm/decoder.layer2': 0.6621437176375339, 'gradnorm/max_param': 142.46054077148438, 'epoch': 6.38}

 64%|██████▍   | 12180/19090 [5:56:22<3:27:10,  1.80s/it]
 64%|██████▍   | 12181/19090 [5:56:23<3:09:44,  1.65s/it]
 64%|██████▍   | 12182/19090 [5:56:26<3:46:01,  1.96s/it]
 64%|██████▍   | 12183/19090 [5:56:27<3:22:45,  1.76s/it]
 64%|██████▍   | 12184/19090 [5:56:28<3:06:20,  1.62s/it]
 64%|██████▍   | 12185/19090 [5:56:31<3:43:45,  1.94s/it]
 64%|██████▍   | 12186/19090 [5:56:33<3:22:44,  1.76s/it]
 64%|██████▍   | 12187/19090 [5:56:34<3:07:07,  1.63s/it]
 64%|██████▍   | 12188/19090 [5:56:36<3:43:10,  1.94s/it]
 64%|██████▍   | 12189/19090 [5:56:38<3:21:19,  1.75s/it][gradnorm] step 12189: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 135.1

 64%|██████▍   | 12190/19090 [5:56:39<3:05:46,  1.62s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 195.68231201171875, 'learning_rate': 1.7399021054864748e-05, 'recon_loss': 0.0016253594367299229, 'recon_static': 0.0006689488829579205, 'recon_motion': 0.0009564105232129805, 'recon_norm': 1.331920748949051, 'recon_norm_std': 0.04107310576364398, 'target_norm': 1.6193545818328858, 'recon_cos': 0.6205983623862267, 'gradnorm/compressor.query': 0.2184963970160057, 'gradnorm/compressor.layer0': 7.460476814147578, 'gradnorm/compressor.layer1': 191.54839874689634, 'gradnorm/compressor.layer2': 35.98654601570738, 'gradnorm/compressor.layer3': 14.428373548561735, 'gradnorm/compressor.layer4': 4.693737508500668, 'gradnorm/compressor.layer5': 2.75199748234301, 'gradnorm/compressor.layer6': 1.6995711038007175, 'gradnorm/compressor.layer7': 1.9242395880775423, 'gradnorm/decoder.layer0': 1.9887475326973543, 'gradnorm/decoder.layer1': 1.1422827183894009, 'gradnorm/decoder.layer2': 0.6299249442743815, 'gradnorm/max_param': 135.1034698486328, 'epoch': 6.39}

 64%|██████▍   | 12190/19090 [5:56:39<3:05:46,  1.62s/it]
 64%|██████▍   | 12191/19090 [5:56:42<3:42:33,  1.94s/it]
 64%|██████▍   | 12192/19090 [5:56:43<3:20:27,  1.74s/it]
 64%|██████▍   | 12193/19090 [5:56:44<3:05:10,  1.61s/it]
 64%|██████▍   | 12194/19090 [5:56:47<3:54:42,  2.04s/it]
 64%|██████▍   | 12195/19090 [5:56:49<3:29:11,  1.82s/it]
 64%|██████▍   | 12196/19090 [5:56:50<3:11:20,  1.67s/it]
 64%|██████▍   | 12197/19090 [5:56:53<3:48:01,  1.98s/it]
 64%|██████▍   | 12198/19090 [5:56:54<3:24:03,  1.78s/it]
 64%|██████▍   | 12199/19090 [5:56:55<3:07:31,  1.63s/it][gradnorm] step 12199: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 123.9

 64%|██████▍   | 12200/19090 [5:56:58<3:52:51,  2.03s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 180.09243774414062, 'learning_rate': 1.7355484688299108e-05, 'recon_loss': 0.001833836772129871, 'recon_static': 0.0008692021016031504, 'recon_motion': 0.0009646346428780816, 'recon_norm': 1.415477028489113, 'recon_norm_std': 0.03966403938829899, 'target_norm': 1.620588904619217, 'recon_cos': 0.5842500641942024, 'gradnorm/compressor.query': 0.17279085612695286, 'gradnorm/compressor.layer0': 6.094859799782428, 'gradnorm/compressor.layer1': 175.55642566921327, 'gradnorm/compressor.layer2': 36.907311717370106, 'gradnorm/compressor.layer3': 13.11894682309027, 'gradnorm/compressor.layer4': 4.506175995323538, 'gradnorm/compressor.layer5': 3.021771270013064, 'gradnorm/compressor.layer6': 1.6409378480084527, 'gradnorm/compressor.layer7': 1.7890643657024934, 'gradnorm/decoder.layer0': 2.143641441866298, 'gradnorm/decoder.layer1': 1.2135784760269486, 'gradnorm/decoder.layer2': 0.629368748221841, 'gradnorm/max_param': 123.8639907836914, 'epoch': 6.39}

 64%|██████▍   | 12200/19090 [5:56:58<3:52:51,  2.03s/it]
 64%|██████▍   | 12201/19090 [5:57:00<3:27:56,  1.81s/it]
 64%|██████▍   | 12202/19090 [5:57:01<3:09:45,  1.65s/it]
 64%|██████▍   | 12203/19090 [5:57:04<3:48:13,  1.99s/it]
 64%|██████▍   | 12204/19090 [5:57:05<3:24:04,  1.78s/it]
 64%|██████▍   | 12205/19090 [5:57:06<3:07:02,  1.63s/it]
 64%|██████▍   | 12206/19090 [5:57:09<3:37:21,  1.89s/it]
 64%|██████▍   | 12207/19090 [5:57:10<3:16:18,  1.71s/it]
 64%|██████▍   | 12208/19090 [5:57:11<3:01:42,  1.58s/it]
 64%|██████▍   | 12209/19090 [5:57:14<3:45:31,  1.97s/it][gradnorm] step 12209: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 147.3

 64%|██████▍   | 12210/19090 [5:57:15<3:22:08,  1.76s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 213.7101593017578, 'learning_rate': 1.7311973881254063e-05, 'recon_loss': 0.001892222632886842, 'recon_static': 0.0009180789842503145, 'recon_motion': 0.00097414368938189, 'recon_norm': 1.4297877252101898, 'recon_norm_std': 0.038681084755808115, 'target_norm': 1.6223849684000016, 'recon_cos': 0.5732427060604095, 'gradnorm/compressor.query': 0.1847749010140762, 'gradnorm/compressor.layer0': 6.465307821806545, 'gradnorm/compressor.layer1': 208.7839605609109, 'gradnorm/compressor.layer2': 42.58689989854568, 'gradnorm/compressor.layer3': 13.467983002740588, 'gradnorm/compressor.layer4': 4.919865467434183, 'gradnorm/compressor.layer5': 2.5575961199137085, 'gradnorm/compressor.layer6': 1.768922949648863, 'gradnorm/compressor.layer7': 1.8202432774819248, 'gradnorm/decoder.layer0': 2.260138768518474, 'gradnorm/decoder.layer1': 1.2901390897448834, 'gradnorm/decoder.layer2': 0.6780105599731612, 'gradnorm/max_param': 147.3106231689453, 'epoch': 6.4}

 64%|██████▍   | 12210/19090 [5:57:15<3:22:08,  1.76s/it]
 64%|██████▍   | 12211/19090 [5:57:17<3:05:49,  1.62s/it]
 64%|██████▍   | 12212/19090 [5:57:19<3:39:19,  1.91s/it]
 64%|██████▍   | 12213/19090 [5:57:21<3:17:37,  1.72s/it]
 64%|██████▍   | 12214/19090 [5:57:22<3:02:35,  1.59s/it]
 64%|██████▍   | 12215/19090 [5:57:25<3:40:52,  1.93s/it]
 64%|██████▍   | 12216/19090 [5:57:26<3:18:58,  1.74s/it]
 64%|██████▍   | 12217/19090 [5:57:27<3:03:31,  1.60s/it]
 64%|██████▍   | 12218/19090 [5:57:30<3:54:33,  2.05s/it]
 64%|██████▍   | 12219/19090 [5:57:32<3:28:32,  1.82s/it][gradnorm] step 12219: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 142.3

 64%|██████▍   | 12220/19090 [5:57:33<3:10:15,  1.66s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 206.3902130126953, 'learning_rate': 1.7268488779208464e-05, 'recon_loss': 0.001675545531907119, 'recon_static': 0.0007077530637616292, 'recon_motion': 0.0009677924972493201, 'recon_norm': 1.3542836010456085, 'recon_norm_std': 0.03992992341518402, 'target_norm': 1.6204033493995667, 'recon_cos': 0.6115990281105042, 'gradnorm/compressor.query': 0.14248809192702516, 'gradnorm/compressor.layer0': 5.292066704154503, 'gradnorm/compressor.layer1': 201.63332089153818, 'gradnorm/compressor.layer2': 40.68639591435939, 'gradnorm/compressor.layer3': 14.668596025202477, 'gradnorm/compressor.layer4': 4.780614673564948, 'gradnorm/compressor.layer5': 2.8131677470659877, 'gradnorm/compressor.layer6': 1.7364710608437173, 'gradnorm/compressor.layer7': 1.7794949099968198, 'gradnorm/decoder.layer0': 1.8273770359068933, 'gradnorm/decoder.layer1': 1.2743210331077635, 'gradnorm/decoder.layer2': 0.6589011501000007, 'gradnorm/max_param': 142.27320861816406, 'epoch': 6.4}

 64%|██████▍   | 12220/19090 [5:57:33<3:10:15,  1.66s/it]
 64%|██████▍   | 12221/19090 [5:57:36<3:43:15,  1.95s/it]
 64%|██████▍   | 12222/19090 [5:57:37<3:20:22,  1.75s/it]
 64%|██████▍   | 12223/19090 [5:57:38<3:04:20,  1.61s/it]
 64%|██████▍   | 12224/19090 [5:57:41<3:49:07,  2.00s/it]
 64%|██████▍   | 12225/19090 [5:57:42<3:24:31,  1.79s/it]
 64%|██████▍   | 12226/19090 [5:57:44<3:07:45,  1.64s/it]
 64%|██████▍   | 12227/19090 [5:57:46<3:39:05,  1.92s/it]
 64%|██████▍   | 12228/19090 [5:57:47<3:17:32,  1.73s/it]
 64%|██████▍   | 12229/19090 [5:57:49<3:02:36,  1.60s/it][gradnorm] step 12229: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 150.2

 64%|██████▍   | 12230/19090 [5:57:52<3:51:52,  2.03s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 219.0401153564453, 'learning_rate': 1.7225029527555214e-05, 'recon_loss': 0.0015759602247271687, 'recon_static': 0.0006144471175502985, 'recon_motion': 0.0009615131100872531, 'recon_norm': 1.305237165093422, 'recon_norm_std': 0.04309590132907033, 'target_norm': 1.6202932566404342, 'recon_cos': 0.6304644450545311, 'gradnorm/compressor.query': 0.15019588123109684, 'gradnorm/compressor.layer0': 5.3166259899596335, 'gradnorm/compressor.layer1': 212.79865043040098, 'gradnorm/compressor.layer2': 48.93247848997509, 'gradnorm/compressor.layer3': 14.848265143888128, 'gradnorm/compressor.layer4': 5.577171094524763, 'gradnorm/compressor.layer5': 2.847844791319277, 'gradnorm/compressor.layer6': 1.7821955931598648, 'gradnorm/compressor.layer7': 1.897252658912001, 'gradnorm/decoder.layer0': 1.9263966410849998, 'gradnorm/decoder.layer1': 1.4130890166717704, 'gradnorm/decoder.layer2': 0.6717987017224626, 'gradnorm/max_param': 150.16151428222656, 'epoch': 6.41}

 64%|██████▍   | 12230/19090 [5:57:52<3:51:52,  2.03s/it]
 64%|██████▍   | 12231/19090 [5:57:53<3:26:30,  1.81s/it]
 64%|██████▍   | 12232/19090 [5:57:54<3:08:57,  1.65s/it]
 64%|██████▍   | 12233/19090 [5:57:57<3:51:06,  2.02s/it]
 64%|██████▍   | 12234/19090 [5:57:59<3:26:08,  1.80s/it]
 64%|██████▍   | 12235/19090 [5:58:00<3:08:52,  1.65s/it]
 64%|██████▍   | 12236/19090 [5:58:02<3:37:25,  1.90s/it]
 64%|██████▍   | 12237/19090 [5:58:04<3:16:30,  1.72s/it]
 64%|██████▍   | 12238/19090 [5:58:05<3:02:41,  1.60s/it]
 64%|██████▍   | 12239/19090 [5:58:07<3:32:52,  1.86s/it][gradnorm] step 12239: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 123.6

 64%|██████▍   | 12240/19090 [5:58:09<3:13:19,  1.69s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 179.6707763671875, 'learning_rate': 1.7181596271600814e-05, 'recon_loss': 0.0016168140718946233, 'recon_static': 0.0006443590231356211, 'recon_motion': 0.0009724549992824905, 'recon_norm': 1.319482770562172, 'recon_norm_std': 0.04075832031667233, 'target_norm': 1.622117030620575, 'recon_cos': 0.6224885687232018, 'gradnorm/compressor.query': 0.11774892235549299, 'gradnorm/compressor.layer0': 4.378231707844303, 'gradnorm/compressor.layer1': 175.2242034776446, 'gradnorm/compressor.layer2': 36.40717206450365, 'gradnorm/compressor.layer3': 13.73334507123822, 'gradnorm/compressor.layer4': 4.9700689904319555, 'gradnorm/compressor.layer5': 2.8446814719757207, 'gradnorm/compressor.layer6': 1.5729790058371935, 'gradnorm/compressor.layer7': 1.6794089152833733, 'gradnorm/decoder.layer0': 2.12262739232805, 'gradnorm/decoder.layer1': 1.2819144689174127, 'gradnorm/decoder.layer2': 0.748571187226528, 'gradnorm/max_param': 123.6399154663086, 'epoch': 6.41}

 64%|██████▍   | 12240/19090 [5:58:09<3:13:19,  1.69s/it]
 64%|██████▍   | 12241/19090 [5:58:10<2:59:30,  1.57s/it]
 64%|██████▍   | 12242/19090 [5:58:13<3:49:24,  2.01s/it]
 64%|██████▍   | 12243/19090 [5:58:14<3:24:45,  1.79s/it]
 64%|██████▍   | 12244/19090 [5:58:16<3:08:04,  1.65s/it]
 64%|██████▍   | 12245/19090 [5:58:18<3:33:35,  1.87s/it]
 64%|██████▍   | 12246/19090 [5:58:19<3:14:12,  1.70s/it]
 64%|██████▍   | 12247/19090 [5:58:21<3:00:31,  1.58s/it]
 64%|██████▍   | 12248/19090 [5:58:24<3:48:05,  2.00s/it]
 64%|██████▍   | 12249/19090 [5:58:25<3:24:13,  1.79s/it][gradnorm] step 12249: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 161.6

 64%|██████▍   | 12250/19090 [5:58:26<3:07:32,  1.65s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 234.20518493652344, 'learning_rate': 1.7138189156564826e-05, 'recon_loss': 0.00158162891457323, 'recon_static': 0.0006077399972127751, 'recon_motion': 0.0009738889057189226, 'recon_norm': 1.3081006109714508, 'recon_norm_std': 0.04153263727203012, 'target_norm': 1.6206901371479034, 'recon_cos': 0.629948103427887, 'gradnorm/compressor.query': 0.16277728883394907, 'gradnorm/compressor.layer0': 5.848773286450657, 'gradnorm/compressor.layer1': 228.9787534494194, 'gradnorm/compressor.layer2': 46.31459804626939, 'gradnorm/compressor.layer3': 14.08928492965848, 'gradnorm/compressor.layer4': 4.755901886360309, 'gradnorm/compressor.layer5': 2.8971859268185853, 'gradnorm/compressor.layer6': 1.6195694340622315, 'gradnorm/compressor.layer7': 1.7836867947731587, 'gradnorm/decoder.layer0': 2.0406969323558464, 'gradnorm/decoder.layer1': 1.2582437001234903, 'gradnorm/decoder.layer2': 0.660319426645109, 'gradnorm/max_param': 161.56613159179688, 'epoch': 6.42}

 64%|██████▍   | 12250/19090 [5:58:26<3:07:32,  1.65s/it]
 64%|██████▍   | 12251/19090 [5:58:29<3:55:26,  2.07s/it]
 64%|██████▍   | 12252/19090 [5:58:31<3:29:18,  1.84s/it]
 64%|██████▍   | 12253/19090 [5:58:32<3:10:52,  1.68s/it]
 64%|██████▍   | 12254/19090 [5:58:34<3:39:00,  1.92s/it]
 64%|██████▍   | 12255/19090 [5:58:36<3:17:45,  1.74s/it]
 64%|██████▍   | 12256/19090 [5:58:37<3:02:55,  1.61s/it]
 64%|██████▍   | 12257/19090 [5:58:40<3:40:20,  1.93s/it]
 64%|██████▍   | 12258/19090 [5:58:41<3:18:19,  1.74s/it]
 64%|██████▍   | 12259/19090 [5:58:42<3:03:12,  1.61s/it][gradnorm] step 12259: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 181.4

 64%|██████▍   | 12260/19090 [5:58:45<3:34:53,  1.89s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 263.06591796875, 'learning_rate': 1.70948083275794e-05, 'recon_loss': 0.001544340731925331, 'recon_static': 0.000576799051486887, 'recon_motion': 0.0009675417197286152, 'recon_norm': 1.287823924422264, 'recon_norm_std': 0.04264450548216701, 'target_norm': 1.6208895146846771, 'recon_cos': 0.6374342799186706, 'gradnorm/compressor.query': 0.2109524138794437, 'gradnorm/compressor.layer0': 7.295771055829005, 'gradnorm/compressor.layer1': 257.1182850362844, 'gradnorm/compressor.layer2': 51.96233236080729, 'gradnorm/compressor.layer3': 16.829181322754255, 'gradnorm/compressor.layer4': 5.573290185502751, 'gradnorm/compressor.layer5': 3.5315659643280615, 'gradnorm/compressor.layer6': 1.700771911062763, 'gradnorm/compressor.layer7': 1.8591230846901208, 'gradnorm/decoder.layer0': 2.323855280304703, 'gradnorm/decoder.layer1': 1.230598660831848, 'gradnorm/decoder.layer2': 0.6983008726840826, 'gradnorm/max_param': 181.42059326171875, 'epoch': 6.42}

 64%|██████▍   | 12260/19090 [5:58:45<3:34:53,  1.89s/it]
 64%|██████▍   | 12261/19090 [5:58:46<3:14:34,  1.71s/it]
 64%|██████▍   | 12262/19090 [5:58:47<3:00:25,  1.59s/it]
 64%|██████▍   | 12263/19090 [5:58:50<3:50:23,  2.02s/it]
 64%|██████▍   | 12264/19090 [5:58:52<3:25:18,  1.80s/it]
 64%|██████▍   | 12265/19090 [5:58:53<3:07:52,  1.65s/it]
 64%|██████▍   | 12266/19090 [5:58:56<3:49:47,  2.02s/it]
 64%|██████▍   | 12267/19090 [5:58:57<3:25:11,  1.80s/it]
 64%|██████▍   | 12268/19090 [5:58:58<3:08:01,  1.65s/it]
 64%|██████▍   | 12269/19090 [5:59:02<3:57:09,  2.09s/it][gradnorm] step 12269: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 160.5

 64%|██████▍   | 12270/19090 [5:59:03<3:30:19,  1.85s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 232.2455596923828, 'learning_rate': 1.7051453929688834e-05, 'recon_loss': 0.0016030885191867128, 'recon_static': 0.0006332706325338222, 'recon_motion': 0.0009698179375845939, 'recon_norm': 1.312979409098625, 'recon_norm_std': 0.041145762987434865, 'target_norm': 1.621633529663086, 'recon_cos': 0.6248259678483009, 'gradnorm/compressor.query': 0.1461722545527103, 'gradnorm/compressor.layer0': 5.287427731283881, 'gradnorm/compressor.layer1': 227.52988299616823, 'gradnorm/compressor.layer2': 43.67411527125149, 'gradnorm/compressor.layer3': 13.82980712314805, 'gradnorm/compressor.layer4': 4.753967635824912, 'gradnorm/compressor.layer5': 2.6983255987152717, 'gradnorm/compressor.layer6': 1.5941450340024945, 'gradnorm/compressor.layer7': 1.7857743076046328, 'gradnorm/decoder.layer0': 2.0700231961779663, 'gradnorm/decoder.layer1': 1.1073162156281862, 'gradnorm/decoder.layer2': 0.5928671195648358, 'gradnorm/max_param': 160.53976440429688, 'epoch': 6.43}

 64%|██████▍   | 12270/19090 [5:59:03<3:30:19,  1.85s/it]
 64%|██████▍   | 12271/19090 [5:59:04<3:11:59,  1.69s/it]
 64%|██████▍   | 12272/19090 [5:59:07<3:39:23,  1.93s/it]
 64%|██████▍   | 12273/19090 [5:59:08<3:17:18,  1.74s/it]
 64%|██████▍   | 12274/19090 [5:59:09<3:01:55,  1.60s/it]
 64%|██████▍   | 12275/19090 [5:59:12<3:38:12,  1.92s/it]
 64%|██████▍   | 12276/19090 [5:59:13<3:16:29,  1.73s/it]
 64%|██████▍   | 12277/19090 [5:59:14<3:01:17,  1.60s/it]
 64%|██████▍   | 12278/19090 [5:59:17<3:45:06,  1.98s/it]
 64%|██████▍   | 12279/19090 [5:59:19<3:21:24,  1.77s/it][gradnorm] step 12279: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 154.0

 64%|██████▍   | 12280/19090 [5:59:20<3:04:47,  1.63s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 224.63536071777344, 'learning_rate': 1.700812610784902e-05, 'recon_loss': 0.0015912863134872168, 'recon_static': 0.0006264600902795792, 'recon_motion': 0.000964826246490702, 'recon_norm': 1.315934669971466, 'recon_norm_std': 0.04204115169122815, 'target_norm': 1.6214575737714767, 'recon_cos': 0.6290530681610107, 'gradnorm/compressor.query': 0.15297557561334066, 'gradnorm/compressor.layer0': 5.255078720243592, 'gradnorm/compressor.layer1': 218.27521689944555, 'gradnorm/compressor.layer2': 50.06464570932266, 'gradnorm/compressor.layer3': 15.354265824320818, 'gradnorm/compressor.layer4': 4.880243855621564, 'gradnorm/compressor.layer5': 3.146833972227815, 'gradnorm/compressor.layer6': 1.5454157101499728, 'gradnorm/compressor.layer7': 1.9542119778041263, 'gradnorm/decoder.layer0': 2.2744247547650134, 'gradnorm/decoder.layer1': 1.2681119526529843, 'gradnorm/decoder.layer2': 0.6403019253261109, 'gradnorm/max_param': 154.01715087890625, 'epoch': 6.43}

 64%|██████▍   | 12280/19090 [5:59:20<3:04:47,  1.63s/it]
 64%|██████▍   | 12281/19090 [5:59:23<3:42:42,  1.96s/it]
 64%|██████▍   | 12282/19090 [5:59:24<3:20:06,  1.76s/it]
 64%|██████▍   | 12283/19090 [5:59:25<3:04:02,  1.62s/it]
 64%|██████▍   | 12284/19090 [5:59:28<3:38:29,  1.93s/it]
 64%|██████▍   | 12285/19090 [5:59:29<3:17:06,  1.74s/it]
 64%|██████▍   | 12286/19090 [5:59:31<3:02:22,  1.61s/it]
 64%|██████▍   | 12287/19090 [5:59:34<3:52:22,  2.05s/it]
 64%|██████▍   | 12288/19090 [5:59:35<3:26:44,  1.82s/it]
 64%|██████▍   | 12289/19090 [5:59:36<3:08:58,  1.67s/it][gradnorm] step 12289: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 147.8

 64%|██████▍   | 12290/19090 [5:59:39<3:51:30,  2.04s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 214.0692901611328, 'learning_rate': 1.6964825006926997e-05, 'recon_loss': 0.0016056068940088152, 'recon_static': 0.0006298383334069513, 'recon_motion': 0.0009757686173543334, 'recon_norm': 1.3154904335737228, 'recon_norm_std': 0.041345651354640724, 'target_norm': 1.62402805685997, 'recon_cos': 0.626805117726326, 'gradnorm/compressor.query': 0.15934137896658826, 'gradnorm/compressor.layer0': 6.120715141150692, 'gradnorm/compressor.layer1': 209.44749677723533, 'gradnorm/compressor.layer2': 40.65243965138347, 'gradnorm/compressor.layer3': 15.053278589365922, 'gradnorm/compressor.layer4': 4.452798610571625, 'gradnorm/compressor.layer5': 2.827912886447346, 'gradnorm/compressor.layer6': 1.4530343468394313, 'gradnorm/compressor.layer7': 1.8677708819741943, 'gradnorm/decoder.layer0': 2.3283913479210248, 'gradnorm/decoder.layer1': 1.176111878201773, 'gradnorm/decoder.layer2': 0.683654281116072, 'gradnorm/max_param': 147.78253173828125, 'epoch': 6.44}

 64%|██████▍   | 12290/19090 [5:59:39<3:51:30,  2.04s/it]
 64%|██████▍   | 12291/19090 [5:59:40<3:26:21,  1.82s/it]
 64%|██████▍   | 12292/19090 [5:59:42<3:08:43,  1.67s/it]
 64%|██████▍   | 12293/19090 [5:59:44<3:22:44,  1.79s/it]
 64%|██████▍   | 12294/19090 [5:59:45<3:06:15,  1.64s/it]
 64%|██████▍   | 12295/19090 [5:59:46<2:54:39,  1.54s/it]
 64%|██████▍   | 12296/19090 [5:59:49<3:25:59,  1.82s/it]
 64%|██████▍   | 12297/19090 [5:59:50<3:08:27,  1.66s/it]
 64%|██████▍   | 12298/19090 [5:59:51<2:56:09,  1.56s/it]
 64%|██████▍   | 12299/19090 [5:59:54<3:42:41,  1.97s/it][gradnorm] step 12299: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 156.8

 64%|██████▍   | 12300/19090 [5:59:56<3:20:10,  1.77s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 226.46409606933594, 'learning_rate': 1.6921550771700484e-05, 'recon_loss': 0.001653852709569037, 'recon_static': 0.000695913634262979, 'recon_motion': 0.0009579390869475901, 'recon_norm': 1.3385105937719346, 'recon_norm_std': 0.04131994703784585, 'target_norm': 1.6199868142604827, 'recon_cos': 0.6142010986804962, 'gradnorm/compressor.query': 0.1506643158107188, 'gradnorm/compressor.layer0': 5.546808462684114, 'gradnorm/compressor.layer1': 222.24213463303937, 'gradnorm/compressor.layer2': 40.67947219431211, 'gradnorm/compressor.layer3': 12.892481518773415, 'gradnorm/compressor.layer4': 4.755408971870463, 'gradnorm/compressor.layer5': 3.021015784355277, 'gradnorm/compressor.layer6': 1.530567406302511, 'gradnorm/compressor.layer7': 1.8097061460335087, 'gradnorm/decoder.layer0': 1.8857785364141433, 'gradnorm/decoder.layer1': 1.1033833922432947, 'gradnorm/decoder.layer2': 0.6775371473411221, 'gradnorm/max_param': 156.80015563964844, 'epoch': 6.44}

 64%|██████▍   | 12300/19090 [5:59:56<3:20:10,  1.77s/it]
 64%|██████▍   | 12301/19090 [5:59:57<3:04:31,  1.63s/it]
 64%|██████▍   | 12302/19090 [6:00:00<3:45:10,  1.99s/it]
 64%|██████▍   | 12303/19090 [6:00:01<3:21:10,  1.78s/it]
 64%|██████▍   | 12304/19090 [6:00:02<3:04:31,  1.63s/it]
 64%|██████▍   | 12305/19090 [6:00:06<3:57:56,  2.10s/it]
 64%|██████▍   | 12306/19090 [6:00:07<3:30:16,  1.86s/it]
 64%|██████▍   | 12307/19090 [6:00:08<3:11:04,  1.69s/it]
 64%|██████▍   | 12308/19090 [6:00:11<3:50:35,  2.04s/it]
 64%|██████▍   | 12309/19090 [6:00:12<3:25:32,  1.82s/it][gradnorm] step 12309: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 153.2

 64%|██████▍   | 12310/19090 [6:00:14<3:08:21,  1.67s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 222.25250244140625, 'learning_rate': 1.687830354685736e-05, 'recon_loss': 0.0016491477377712726, 'recon_static': 0.0006800409086281433, 'recon_motion': 0.0009691068044048734, 'recon_norm': 1.3351068377494812, 'recon_norm_std': 0.04079077541828156, 'target_norm': 1.621751543879509, 'recon_cos': 0.6162516996264458, 'gradnorm/compressor.query': 0.17822370136947419, 'gradnorm/compressor.layer0': 6.745669070484861, 'gradnorm/compressor.layer1': 217.08333833173378, 'gradnorm/compressor.layer2': 44.41093945154448, 'gradnorm/compressor.layer3': 14.702565619340579, 'gradnorm/compressor.layer4': 4.282443137874218, 'gradnorm/compressor.layer5': 2.7744516157316363, 'gradnorm/compressor.layer6': 1.6612594708703017, 'gradnorm/compressor.layer7': 1.7161641870037232, 'gradnorm/decoder.layer0': 1.881134044837639, 'gradnorm/decoder.layer1': 1.128626277221944, 'gradnorm/decoder.layer2': 0.6450404381089069, 'gradnorm/max_param': 153.1670684814453, 'epoch': 6.45}

 64%|██████▍   | 12310/19090 [6:00:14<3:08:21,  1.67s/it]
 64%|██████▍   | 12311/19090 [6:00:17<3:47:51,  2.02s/it]
 64%|██████▍   | 12312/19090 [6:00:18<3:23:32,  1.80s/it]
 64%|██████▍   | 12313/19090 [6:00:19<3:06:10,  1.65s/it]
 65%|██████▍   | 12314/19090 [6:00:22<3:37:14,  1.92s/it]
 65%|██████▍   | 12315/19090 [6:00:23<3:15:41,  1.73s/it]
 65%|██████▍   | 12316/19090 [6:00:24<3:00:56,  1.60s/it]
 65%|██████▍   | 12317/19090 [6:00:27<3:33:07,  1.89s/it]
 65%|██████▍   | 12318/19090 [6:00:28<3:12:46,  1.71s/it]
 65%|██████▍   | 12319/19090 [6:00:29<2:58:32,  1.58s/it][gradnorm] step 12319: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 156.9

 65%|██████▍   | 12320/19090 [6:00:32<3:34:26,  1.90s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 226.84837341308594, 'learning_rate': 1.6835083476995184e-05, 'recon_loss': 0.0017135854286607355, 'recon_static': 0.0007449348835507408, 'recon_motion': 0.0009686505683930591, 'recon_norm': 1.3642482072114945, 'recon_norm_std': 0.04057648936286569, 'target_norm': 1.6199004799127579, 'recon_cos': 0.6033367410302162, 'gradnorm/compressor.query': 0.18508264792935308, 'gradnorm/compressor.layer0': 6.855170766761035, 'gradnorm/compressor.layer1': 222.36473240770442, 'gradnorm/compressor.layer2': 41.99832004638697, 'gradnorm/compressor.layer3': 12.546260832960227, 'gradnorm/compressor.layer4': 4.886441424922061, 'gradnorm/compressor.layer5': 2.8565359560938544, 'gradnorm/compressor.layer6': 1.8650021073889396, 'gradnorm/compressor.layer7': 1.940967963192596, 'gradnorm/decoder.layer0': 2.0971148198845815, 'gradnorm/decoder.layer1': 1.2835599034607208, 'gradnorm/decoder.layer2': 0.6965541468753094, 'gradnorm/max_param': 156.90634155273438, 'epoch': 6.45}

 65%|██████▍   | 12320/19090 [6:00:32<3:34:26,  1.90s/it]
 65%|██████▍   | 12321/19090 [6:00:33<3:13:59,  1.72s/it]
 65%|██████▍   | 12322/19090 [6:00:35<2:59:22,  1.59s/it]
 65%|██████▍   | 12323/19090 [6:00:37<3:35:41,  1.91s/it]
 65%|██████▍   | 12324/19090 [6:00:39<3:14:39,  1.73s/it]
 65%|██████▍   | 12325/19090 [6:00:40<2:59:51,  1.60s/it]
 65%|██████▍   | 12326/19090 [6:00:42<3:34:29,  1.90s/it]
 65%|██████▍   | 12327/19090 [6:00:44<3:13:53,  1.72s/it]
 65%|██████▍   | 12328/19090 [6:00:45<2:59:29,  1.59s/it]
 65%|██████▍   | 12329/19090 [6:00:48<3:30:24,  1.87s/it][gradnorm] step 12329: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 176.3

 65%|██████▍   | 12330/19090 [6:00:49<3:20:17,  1.78s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 254.08673095703125, 'learning_rate': 1.6791890706620726e-05, 'recon_loss': 0.0016395570361055433, 'recon_static': 0.0006638878767262213, 'recon_motion': 0.0009756692044902593, 'recon_norm': 1.3302105784416198, 'recon_norm_std': 0.04084534859284759, 'target_norm': 1.6223645567893983, 'recon_cos': 0.6186202436685562, 'gradnorm/compressor.query': 0.24021288922999862, 'gradnorm/compressor.layer0': 7.3909865567117725, 'gradnorm/compressor.layer1': 249.91840856067407, 'gradnorm/compressor.layer2': 42.646255697305946, 'gradnorm/compressor.layer3': 13.55402612309375, 'gradnorm/compressor.layer4': 4.672666144603719, 'gradnorm/compressor.layer5': 3.0888573076233077, 'gradnorm/compressor.layer6': 1.7868207187687069, 'gradnorm/compressor.layer7': 1.71270075825554, 'gradnorm/decoder.layer0': 2.0700192801641886, 'gradnorm/decoder.layer1': 1.2554097891103349, 'gradnorm/decoder.layer2': 0.6259303797499242, 'gradnorm/max_param': 176.34385681152344, 'epoch': 6.46}

 65%|██████▍   | 12330/19090 [6:00:49<3:20:17,  1.78s/it]
 65%|██████▍   | 12331/19090 [6:00:50<3:03:52,  1.63s/it]
 65%|██████▍   | 12332/19090 [6:00:53<3:41:07,  1.96s/it]
 65%|██████▍   | 12333/19090 [6:00:55<3:22:51,  1.80s/it]
 65%|██████▍   | 12334/19090 [6:00:56<3:05:31,  1.65s/it]
 65%|██████▍   | 12335/19090 [6:00:59<3:48:19,  2.03s/it]
 65%|██████▍   | 12336/19090 [6:01:00<3:23:16,  1.81s/it]
 65%|██████▍   | 12337/19090 [6:01:01<3:05:38,  1.65s/it]
 65%|██████▍   | 12338/19090 [6:01:04<3:44:59,  2.00s/it]
 65%|██████▍   | 12339/19090 [6:01:05<3:20:49,  1.78s/it][gradnorm] step 12339: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 210.6

 65%|██████▍   | 12340/19090 [6:01:07<3:04:07,  1.64s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 303.7735595703125, 'learning_rate': 1.6748725380149514e-05, 'recon_loss': 0.0015917323005851358, 'recon_static': 0.000626341305905953, 'recon_motion': 0.0009653909612097778, 'recon_norm': 1.3108404159545899, 'recon_norm_std': 0.04142040042206645, 'target_norm': 1.6206042557954787, 'recon_cos': 0.6275743454694748, 'gradnorm/compressor.query': 0.1980470251048947, 'gradnorm/compressor.layer0': 7.159552865585073, 'gradnorm/compressor.layer1': 298.5183987210839, 'gradnorm/compressor.layer2': 52.57827126651275, 'gradnorm/compressor.layer3': 17.244087422226514, 'gradnorm/compressor.layer4': 5.421491889720277, 'gradnorm/compressor.layer5': 3.1315860773517437, 'gradnorm/compressor.layer6': 1.6843814998352673, 'gradnorm/compressor.layer7': 1.8028805095919294, 'gradnorm/decoder.layer0': 2.163633228238532, 'gradnorm/decoder.layer1': 1.268128732494353, 'gradnorm/decoder.layer2': 0.6473863653675777, 'gradnorm/max_param': 210.6140594482422, 'epoch': 6.46}

 65%|██████▍   | 12340/19090 [6:01:07<3:04:07,  1.64s/it]
 65%|██████▍   | 12341/19090 [6:01:09<3:38:18,  1.94s/it]
 65%|██████▍   | 12342/19090 [6:01:11<3:18:40,  1.77s/it]
 65%|██████▍   | 12343/19090 [6:01:12<3:02:25,  1.62s/it]
 65%|██████▍   | 12344/19090 [6:01:15<3:43:12,  1.99s/it]
 65%|██████▍   | 12345/19090 [6:01:16<3:19:56,  1.78s/it]
 65%|██████▍   | 12346/19090 [6:01:17<3:03:27,  1.63s/it]
 65%|██████▍   | 12347/19090 [6:01:20<3:43:26,  1.99s/it]
 65%|██████▍   | 12348/19090 [6:01:22<3:19:50,  1.78s/it]
 65%|██████▍   | 12349/19090 [6:01:23<3:03:22,  1.63s/it][gradnorm] step 12349: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 158.8

 65%|██████▍   | 12350/19090 [6:01:26<3:39:16,  1.95s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 229.9600067138672, 'learning_rate': 1.6705587641905267e-05, 'recon_loss': 0.0015604940359480678, 'recon_static': 0.0005989143435726873, 'recon_motion': 0.0009615796792786568, 'recon_norm': 1.3007634103298187, 'recon_norm_std': 0.04135163873434067, 'target_norm': 1.6190058678388595, 'recon_cos': 0.6338303595781326, 'gradnorm/compressor.query': 0.1932656513756583, 'gradnorm/compressor.layer0': 7.310239523996007, 'gradnorm/compressor.layer1': 225.10521845350453, 'gradnorm/compressor.layer2': 43.76073668478757, 'gradnorm/compressor.layer3': 13.73301452217287, 'gradnorm/compressor.layer4': 5.354356735160937, 'gradnorm/compressor.layer5': 3.2045594515762263, 'gradnorm/compressor.layer6': 1.6093559634369803, 'gradnorm/compressor.layer7': 2.021516219178529, 'gradnorm/decoder.layer0': 2.0987335791521957, 'gradnorm/decoder.layer1': 1.325439655488737, 'gradnorm/decoder.layer2': 0.6254412047443316, 'gradnorm/max_param': 158.8176727294922, 'epoch': 6.47}

 65%|██████▍   | 12350/19090 [6:01:26<3:39:16,  1.95s/it]
 65%|██████▍   | 12351/19090 [6:01:27<3:17:00,  1.75s/it]
 65%|██████▍   | 12352/19090 [6:01:28<3:01:21,  1.61s/it]
 65%|██████▍   | 12353/19090 [6:01:31<3:43:01,  1.99s/it]
 65%|██████▍   | 12354/19090 [6:01:32<3:19:35,  1.78s/it]
 65%|██████▍   | 12355/19090 [6:01:34<3:03:03,  1.63s/it]
 65%|██████▍   | 12356/19090 [6:01:36<3:45:13,  2.01s/it]
 65%|██████▍   | 12357/19090 [6:01:38<3:21:21,  1.79s/it]
 65%|██████▍   | 12358/19090 [6:01:39<3:04:37,  1.65s/it]
 65%|██████▍   | 12359/19090 [6:01:42<3:36:03,  1.93s/it][gradnorm] step 12359: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 143.8

 65%|██████▍   | 12360/19090 [6:01:43<3:18:05,  1.77s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 209.95945739746094, 'learning_rate': 1.6662477636119482e-05, 'recon_loss': 0.0015929638466332108, 'recon_static': 0.0006383499945513904, 'recon_motion': 0.0009546139015583321, 'recon_norm': 1.3199335783720016, 'recon_norm_std': 0.0416244201362133, 'target_norm': 1.6165914982557297, 'recon_cos': 0.6262358605861664, 'gradnorm/compressor.query': 0.15044048646982427, 'gradnorm/compressor.layer0': 5.632187722697209, 'gradnorm/compressor.layer1': 203.79067845463393, 'gradnorm/compressor.layer2': 48.30087839279556, 'gradnorm/compressor.layer3': 12.079764073575369, 'gradnorm/compressor.layer4': 4.704174444272708, 'gradnorm/compressor.layer5': 2.798327021119309, 'gradnorm/compressor.layer6': 1.5402754443171698, 'gradnorm/compressor.layer7': 1.932571294932624, 'gradnorm/decoder.layer0': 1.948843329288693, 'gradnorm/decoder.layer1': 1.2176687873225243, 'gradnorm/decoder.layer2': 0.592034874323818, 'gradnorm/max_param': 143.79457092285156, 'epoch': 6.47}

 65%|██████▍   | 12360/19090 [6:01:43<3:18:05,  1.77s/it]
 65%|██████▍   | 12361/19090 [6:01:44<3:03:11,  1.63s/it]
 65%|██████▍   | 12362/19090 [6:01:48<3:53:45,  2.08s/it]
 65%|██████▍   | 12363/19090 [6:01:49<3:27:23,  1.85s/it]
 65%|██████▍   | 12364/19090 [6:01:50<3:09:01,  1.69s/it]
 65%|██████▍   | 12365/19090 [6:01:53<3:47:21,  2.03s/it]
 65%|██████▍   | 12366/19090 [6:01:54<3:22:48,  1.81s/it]
 65%|██████▍   | 12367/19090 [6:01:56<3:05:35,  1.66s/it]
 65%|██████▍   | 12368/19090 [6:01:58<3:34:59,  1.92s/it]
 65%|██████▍   | 12369/19090 [6:01:59<3:14:06,  1.73s/it][gradnorm] step 12369: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 161.5

 65%|██████▍   | 12370/19090 [6:02:01<2:59:50,  1.61s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 234.07333374023438, 'learning_rate': 1.6619395506930955e-05, 'recon_loss': 0.001591593405464664, 'recon_static': 0.0006261588132474571, 'recon_motion': 0.0009654346198658459, 'recon_norm': 1.3124647676944732, 'recon_norm_std': 0.041818319167941806, 'target_norm': 1.6208136558532715, 'recon_cos': 0.6275818228721619, 'gradnorm/compressor.query': 0.17995170160518786, 'gradnorm/compressor.layer0': 6.098992386442437, 'gradnorm/compressor.layer1': 228.95187471933923, 'gradnorm/compressor.layer2': 45.47465865818154, 'gradnorm/compressor.layer3': 14.85250769611924, 'gradnorm/compressor.layer4': 4.9121685896654546, 'gradnorm/compressor.layer5': 2.7478858450509986, 'gradnorm/compressor.layer6': 1.6270812717999412, 'gradnorm/compressor.layer7': 1.9602115700837899, 'gradnorm/decoder.layer0': 2.36701385566192, 'gradnorm/decoder.layer1': 1.207982253449548, 'gradnorm/decoder.layer2': 0.6051617490093583, 'gradnorm/max_param': 161.54966735839844, 'epoch': 6.48}

 65%|██████▍   | 12370/19090 [6:02:01<2:59:50,  1.61s/it]
 65%|██████▍   | 12371/19090 [6:02:03<3:31:09,  1.89s/it]
 65%|██████▍   | 12372/19090 [6:02:05<3:11:41,  1.71s/it]
 65%|██████▍   | 12373/19090 [6:02:06<2:57:50,  1.59s/it]
 65%|██████▍   | 12374/19090 [6:02:09<3:38:48,  1.95s/it]
 65%|██████▍   | 12375/19090 [6:02:10<3:16:42,  1.76s/it]
 65%|██████▍   | 12376/19090 [6:02:11<3:01:31,  1.62s/it]
 65%|██████▍   | 12377/19090 [6:02:14<3:27:58,  1.86s/it]
 65%|██████▍   | 12378/19090 [6:02:15<3:08:59,  1.69s/it]
 65%|██████▍   | 12379/19090 [6:02:16<2:55:40,  1.57s/it][gradnorm] step 12379: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 185.8

 65%|██████▍   | 12380/19090 [6:02:19<3:34:01,  1.91s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 268.584228515625, 'learning_rate': 1.657634139838524e-05, 'recon_loss': 0.0015468209545360877, 'recon_static': 0.0005836988362716511, 'recon_motion': 0.0009631221575546079, 'recon_norm': 1.2941139757633209, 'recon_norm_std': 0.04240991929545999, 'target_norm': 1.6179512470960618, 'recon_cos': 0.6359715580940246, 'gradnorm/compressor.query': 0.15179158673226303, 'gradnorm/compressor.layer0': 5.98652757620428, 'gradnorm/compressor.layer1': 263.3171062378971, 'gradnorm/compressor.layer2': 49.8225023285789, 'gradnorm/compressor.layer3': 15.158115924927962, 'gradnorm/compressor.layer4': 5.46088528369394, 'gradnorm/compressor.layer5': 3.059663977530132, 'gradnorm/compressor.layer6': 1.7246929420624486, 'gradnorm/compressor.layer7': 2.1013045028525723, 'gradnorm/decoder.layer0': 2.3062355144428315, 'gradnorm/decoder.layer1': 1.1968418883681184, 'gradnorm/decoder.layer2': 0.6029778711202929, 'gradnorm/max_param': 185.79049682617188, 'epoch': 6.49}

 65%|██████▍   | 12380/19090 [6:02:19<3:34:01,  1.91s/it]
 65%|██████▍   | 12381/19090 [6:02:20<3:16:01,  1.75s/it]
 65%|██████▍   | 12382/19090 [6:02:22<3:00:28,  1.61s/it]
 65%|██████▍   | 12383/19090 [6:02:24<3:33:54,  1.91s/it]
 65%|██████▍   | 12384/19090 [6:02:26<3:32:41,  1.90s/it]
 65%|██████▍   | 12385/19090 [6:02:27<3:12:10,  1.72s/it]
 65%|██████▍   | 12386/19090 [6:02:29<3:16:37,  1.76s/it]
 65%|██████▍   | 12387/19090 [6:02:32<3:38:31,  1.96s/it]
 65%|██████▍   | 12388/19090 [6:02:33<3:16:09,  1.76s/it]
 65%|██████▍   | 12389/19090 [6:02:35<3:19:17,  1.78s/it][gradnorm] step 12389: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 177.3

 65%|██████▍   | 12390/19090 [6:02:37<3:38:44,  1.96s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 255.898681640625, 'learning_rate': 1.6533315454434224e-05, 'recon_loss': 0.0015566733694868163, 'recon_static': 0.0005942631280049682, 'recon_motion': 0.0009624102516681887, 'recon_norm': 1.2987755239009857, 'recon_norm_std': 0.04251321433112025, 'target_norm': 1.620547553896904, 'recon_cos': 0.6352001845836639, 'gradnorm/compressor.query': 0.20743978426641907, 'gradnorm/compressor.layer0': 7.803776868327575, 'gradnorm/compressor.layer1': 251.3690094388129, 'gradnorm/compressor.layer2': 43.997978857609894, 'gradnorm/compressor.layer3': 15.666332051917497, 'gradnorm/compressor.layer4': 5.3601947302104955, 'gradnorm/compressor.layer5': 3.4196798172475495, 'gradnorm/compressor.layer6': 1.770383919927536, 'gradnorm/compressor.layer7': 2.0837008597123803, 'gradnorm/decoder.layer0': 2.328548520884057, 'gradnorm/decoder.layer1': 1.343782690670779, 'gradnorm/decoder.layer2': 0.6565988839411597, 'gradnorm/max_param': 177.34654235839844, 'epoch': 6.49}

 65%|██████▍   | 12390/19090 [6:02:37<3:38:44,  1.96s/it]
 65%|██████▍   | 12391/19090 [6:02:38<3:16:55,  1.76s/it]
 65%|██████▍   | 12392/19090 [6:02:40<3:17:24,  1.77s/it]
 65%|██████▍   | 12393/19090 [6:02:42<3:19:29,  1.79s/it]
 65%|██████▍   | 12394/19090 [6:02:43<3:02:54,  1.64s/it]
 65%|██████▍   | 12395/19090 [6:02:46<3:23:14,  1.82s/it]
 65%|██████▍   | 12396/19090 [6:02:48<3:29:14,  1.88s/it]
 65%|██████▍   | 12397/19090 [6:02:49<3:09:47,  1.70s/it]
 65%|██████▍   | 12398/19090 [6:02:51<3:24:18,  1.83s/it]
 65%|██████▍   | 12399/19090 [6:02:53<3:30:30,  1.89s/it][gradnorm] step 12399: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 155.5

 65%|██████▍   | 12400/19090 [6:02:54<3:10:29,  1.71s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 224.29891967773438, 'learning_rate': 1.649031781893563e-05, 'recon_loss': 0.001566798021667637, 'recon_static': 0.0005995498024276458, 'recon_motion': 0.0009672481945017352, 'recon_norm': 1.3085618644952774, 'recon_norm_std': 0.041417747642844914, 'target_norm': 1.6196778059005736, 'recon_cos': 0.6334029540419579, 'gradnorm/compressor.query': 0.1537538158420068, 'gradnorm/compressor.layer0': 5.4390186512973235, 'gradnorm/compressor.layer1': 220.4232700255352, 'gradnorm/compressor.layer2': 38.48853160916492, 'gradnorm/compressor.layer3': 12.924509796940498, 'gradnorm/compressor.layer4': 4.911235633027114, 'gradnorm/compressor.layer5': 2.9116687095851823, 'gradnorm/compressor.layer6': 1.724309149905797, 'gradnorm/compressor.layer7': 1.878820849597319, 'gradnorm/decoder.layer0': 2.18027928566945, 'gradnorm/decoder.layer1': 1.1491578626030339, 'gradnorm/decoder.layer2': 0.6242855995868043, 'gradnorm/max_param': 155.51565551757812, 'epoch': 6.5}

 65%|██████▍   | 12400/19090 [6:02:54<3:10:29,  1.71s/it]
 65%|██████▍   | 12401/19090 [6:02:56<3:21:39,  1.81s/it]
 65%|██████▍   | 12402/19090 [6:02:59<3:36:22,  1.94s/it]
 65%|██████▍   | 12403/19090 [6:03:00<3:14:34,  1.75s/it]
 65%|██████▍   | 12404/19090 [6:03:02<3:26:16,  1.85s/it]
 65%|██████▍   | 12405/19090 [6:03:04<3:33:28,  1.92s/it]
 65%|██████▍   | 12406/19090 [6:03:05<3:12:28,  1.73s/it]
 65%|██████▍   | 12407/19090 [6:03:07<3:19:37,  1.79s/it]
 65%|██████▍   | 12408/19090 [6:03:09<3:22:58,  1.82s/it]
 65%|██████▌   | 12409/19090 [6:03:11<3:05:08,  1.66s/it][gradnorm] step 12409: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 171.2

 65%|██████▌   | 12410/19090 [6:03:13<3:27:40,  1.87s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 248.0114288330078, 'learning_rate': 1.644734863565251e-05, 'recon_loss': 0.0015198896173387766, 'recon_static': 0.0005620593423373066, 'recon_motion': 0.0009578302022418938, 'recon_norm': 1.2797356754541398, 'recon_norm_std': 0.04272623173892498, 'target_norm': 1.6187530368566514, 'recon_cos': 0.6422105595469475, 'gradnorm/compressor.query': 0.1462632863212609, 'gradnorm/compressor.layer0': 5.514429930121847, 'gradnorm/compressor.layer1': 242.6101231915107, 'gradnorm/compressor.layer2': 48.090188809835475, 'gradnorm/compressor.layer3': 15.594424674627057, 'gradnorm/compressor.layer4': 5.925700781218636, 'gradnorm/compressor.layer5': 3.405944670545298, 'gradnorm/compressor.layer6': 2.0974569954174824, 'gradnorm/compressor.layer7': 1.9599201253263618, 'gradnorm/decoder.layer0': 2.4794065108339085, 'gradnorm/decoder.layer1': 1.4532498541897028, 'gradnorm/decoder.layer2': 0.7060025988684956, 'gradnorm/max_param': 171.1841278076172, 'epoch': 6.5}

 65%|██████▌   | 12410/19090 [6:03:13<3:27:40,  1.87s/it]
 65%|██████▌   | 12411/19090 [6:03:15<3:26:10,  1.85s/it]
 65%|██████▌   | 12412/19090 [6:03:16<3:07:33,  1.69s/it]
 65%|██████▌   | 12413/19090 [6:03:18<3:27:17,  1.86s/it]
 65%|██████▌   | 12414/19090 [6:03:20<3:20:28,  1.80s/it]
 65%|██████▌   | 12415/19090 [6:03:21<3:03:05,  1.65s/it]
 65%|██████▌   | 12416/19090 [6:03:24<3:31:38,  1.90s/it]
 65%|██████▌   | 12417/19090 [6:03:25<3:19:57,  1.80s/it]
 65%|██████▌   | 12418/19090 [6:03:27<3:02:55,  1.65s/it]
 65%|██████▌   | 12419/19090 [6:03:29<3:20:36,  1.80s/it][gradnorm] step 12419: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 183.4

 65%|██████▌   | 12420/19090 [6:03:30<3:15:51,  1.76s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 265.7901611328125, 'learning_rate': 1.6404408048252794e-05, 'recon_loss': 0.0015326687833294272, 'recon_static': 0.0005621547577902674, 'recon_motion': 0.0009705140255391598, 'recon_norm': 1.2819440215826035, 'recon_norm_std': 0.042098873760551214, 'target_norm': 1.6237132906913758, 'recon_cos': 0.6409624472260476, 'gradnorm/compressor.query': 0.1578585498493358, 'gradnorm/compressor.layer0': 5.872915019720396, 'gradnorm/compressor.layer1': 259.87481962475704, 'gradnorm/compressor.layer2': 51.60283223740607, 'gradnorm/compressor.layer3': 18.828011540114574, 'gradnorm/compressor.layer4': 5.5324791873311145, 'gradnorm/compressor.layer5': 3.3890226594826447, 'gradnorm/compressor.layer6': 2.0304309073807993, 'gradnorm/compressor.layer7': 1.9768848985538148, 'gradnorm/decoder.layer0': 2.3150913050794544, 'gradnorm/decoder.layer1': 1.2539665231449415, 'gradnorm/decoder.layer2': 0.7542322669468701, 'gradnorm/max_param': 183.37054443359375, 'epoch': 6.51}

 65%|██████▌   | 12420/19090 [6:03:30<3:15:51,  1.76s/it]
 65%|██████▌   | 12421/19090 [6:03:32<3:00:02,  1.62s/it]
 65%|██████▌   | 12422/19090 [6:03:34<3:34:49,  1.93s/it]
 65%|██████▌   | 12423/19090 [6:03:36<3:13:13,  1.74s/it]
 65%|██████▌   | 12424/19090 [6:03:37<2:57:56,  1.60s/it]
 65%|██████▌   | 12425/19090 [6:03:40<3:39:11,  1.97s/it]
 65%|██████▌   | 12426/19090 [6:03:41<3:16:06,  1.77s/it]
 65%|██████▌   | 12427/19090 [6:03:42<3:00:17,  1.62s/it]
 65%|██████▌   | 12428/19090 [6:03:45<3:22:45,  1.83s/it]
 65%|██████▌   | 12429/19090 [6:03:46<3:04:30,  1.66s/it][gradnorm] step 12429: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 204.8

 65%|██████▌   | 12430/19090 [6:03:47<2:52:00,  1.55s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 295.268798828125, 'learning_rate': 1.636149620030883e-05, 'recon_loss': 0.0015130930405575782, 'recon_static': 0.0005533619740162976, 'recon_motion': 0.00095973112329375, 'recon_norm': 1.2819634228944778, 'recon_norm_std': 0.042898022662848236, 'target_norm': 1.6196366280317307, 'recon_cos': 0.6446704342961311, 'gradnorm/compressor.query': 0.16106411596582226, 'gradnorm/compressor.layer0': 5.543414516684115, 'gradnorm/compressor.layer1': 290.2953237652994, 'gradnorm/compressor.layer2': 50.71973678183622, 'gradnorm/compressor.layer3': 15.947595129212223, 'gradnorm/compressor.layer4': 5.456334784374471, 'gradnorm/compressor.layer5': 3.196492693015115, 'gradnorm/compressor.layer6': 1.8968687129662787, 'gradnorm/compressor.layer7': 1.9632814540172248, 'gradnorm/decoder.layer0': 2.3019109002575684, 'gradnorm/decoder.layer1': 1.2389378296138163, 'gradnorm/decoder.layer2': 0.6684106625398768, 'gradnorm/max_param': 204.83924865722656, 'epoch': 6.51}

 65%|██████▌   | 12430/19090 [6:03:47<2:52:00,  1.55s/it]
 65%|██████▌   | 12431/19090 [6:03:50<3:21:46,  1.82s/it]
 65%|██████▌   | 12432/19090 [6:03:51<3:03:52,  1.66s/it]
 65%|██████▌   | 12433/19090 [6:03:52<2:51:32,  1.55s/it]
 65%|██████▌   | 12434/19090 [6:03:55<3:28:19,  1.88s/it]
 65%|██████▌   | 12435/19090 [6:03:56<3:09:03,  1.70s/it]
 65%|██████▌   | 12436/19090 [6:03:57<2:55:12,  1.58s/it]
 65%|██████▌   | 12437/19090 [6:04:00<3:37:57,  1.97s/it]
 65%|██████▌   | 12438/19090 [6:04:02<3:15:49,  1.77s/it]
 65%|██████▌   | 12439/19090 [6:04:03<2:59:38,  1.62s/it][gradnorm] step 12439: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 183.3

 65%|██████▌   | 12440/19090 [6:04:06<3:40:36,  1.99s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 265.7008361816406, 'learning_rate': 1.631861323529683e-05, 'recon_loss': 0.0015284734399756417, 'recon_static': 0.0005576191557338461, 'recon_motion': 0.0009708543017040938, 'recon_norm': 1.2798606395721435, 'recon_norm_std': 0.042308190185576675, 'target_norm': 1.620806559920311, 'recon_cos': 0.6407565325498581, 'gradnorm/compressor.query': 0.15260475980971053, 'gradnorm/compressor.layer0': 5.025469948727821, 'gradnorm/compressor.layer1': 259.763353843455, 'gradnorm/compressor.layer2': 52.750312406509735, 'gradnorm/compressor.layer3': 16.356346163323064, 'gradnorm/compressor.layer4': 4.8699451551541095, 'gradnorm/compressor.layer5': 2.7821773740277163, 'gradnorm/compressor.layer6': 1.8911432233824108, 'gradnorm/compressor.layer7': 1.8258883495854772, 'gradnorm/decoder.layer0': 2.118825243848805, 'gradnorm/decoder.layer1': 1.1080806326980388, 'gradnorm/decoder.layer2': 0.6546288402562006, 'gradnorm/max_param': 183.30140686035156, 'epoch': 6.52}

 65%|██████▌   | 12440/19090 [6:04:06<3:40:36,  1.99s/it]
 65%|██████▌   | 12441/19090 [6:04:07<3:17:10,  1.78s/it]
 65%|██████▌   | 12442/19090 [6:04:08<3:01:09,  1.64s/it]
 65%|██████▌   | 12443/19090 [6:04:11<3:37:00,  1.96s/it]
 65%|██████▌   | 12444/19090 [6:04:12<3:14:31,  1.76s/it]
 65%|██████▌   | 12445/19090 [6:04:14<2:58:41,  1.61s/it]
 65%|██████▌   | 12446/19090 [6:04:16<3:36:32,  1.96s/it]
 65%|██████▌   | 12447/19090 [6:04:18<3:14:04,  1.75s/it]
 65%|██████▌   | 12448/19090 [6:04:19<2:58:31,  1.61s/it]
 65%|██████▌   | 12449/19090 [6:04:22<3:35:27,  1.95s/it][gradnorm] step 12449: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 198.6

 65%|██████▌   | 12450/19090 [6:04:23<3:13:29,  1.75s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 287.0721130371094, 'learning_rate': 1.627575929659646e-05, 'recon_loss': 0.0015446414385223762, 'recon_static': 0.0005679034846252761, 'recon_motion': 0.0009767379437107594, 'recon_norm': 1.2900382697582244, 'recon_norm_std': 0.04179319040849805, 'target_norm': 1.621958500146866, 'recon_cos': 0.6384870022535324, 'gradnorm/compressor.query': 0.17073523767134152, 'gradnorm/compressor.layer0': 5.948903907665405, 'gradnorm/compressor.layer1': 281.4785321520986, 'gradnorm/compressor.layer2': 52.93456907839739, 'gradnorm/compressor.layer3': 17.143941608822598, 'gradnorm/compressor.layer4': 4.997078041784866, 'gradnorm/compressor.layer5': 3.1087139711336085, 'gradnorm/compressor.layer6': 1.8372763439225406, 'gradnorm/compressor.layer7': 1.94438461786906, 'gradnorm/decoder.layer0': 2.268168107330308, 'gradnorm/decoder.layer1': 1.1930550594622393, 'gradnorm/decoder.layer2': 0.6817159529109896, 'gradnorm/max_param': 198.61605834960938, 'epoch': 6.52}

 65%|██████▌   | 12450/19090 [6:04:23<3:13:29,  1.75s/it]
 65%|██████▌   | 12451/19090 [6:04:24<2:58:12,  1.61s/it]
 65%|██████▌   | 12452/19090 [6:04:27<3:35:19,  1.95s/it]
 65%|██████▌   | 12453/19090 [6:04:28<3:13:26,  1.75s/it]
 65%|██████▌   | 12454/19090 [6:04:30<2:58:16,  1.61s/it]
 65%|██████▌   | 12455/19090 [6:04:33<3:45:06,  2.04s/it]
 65%|██████▌   | 12456/19090 [6:04:34<3:20:26,  1.81s/it]
 65%|██████▌   | 12457/19090 [6:04:35<3:03:05,  1.66s/it]
 65%|██████▌   | 12458/19090 [6:04:38<3:38:32,  1.98s/it]
 65%|██████▌   | 12459/19090 [6:04:39<3:15:50,  1.77s/it][gradnorm] step 12459: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 142.9

 65%|██████▌   | 12460/19090 [6:04:40<2:59:55,  1.63s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 206.8150177001953, 'learning_rate': 1.6232934527490342e-05, 'recon_loss': 0.0015733340580482036, 'recon_static': 0.0006146029307274147, 'recon_motion': 0.0009587311433278956, 'recon_norm': 1.307548463344574, 'recon_norm_std': 0.04201278565451503, 'target_norm': 1.6192911237478256, 'recon_cos': 0.630982854962349, 'gradnorm/compressor.query': 0.13728292409395815, 'gradnorm/compressor.layer0': 4.936470214073865, 'gradnorm/compressor.layer1': 202.4574415154948, 'gradnorm/compressor.layer2': 39.12228081854779, 'gradnorm/compressor.layer3': 13.384627540373742, 'gradnorm/compressor.layer4': 5.249732601077173, 'gradnorm/compressor.layer5': 3.0635382001888134, 'gradnorm/compressor.layer6': 1.6511891617060706, 'gradnorm/compressor.layer7': 1.8706320429910561, 'gradnorm/decoder.layer0': 2.1268943869586865, 'gradnorm/decoder.layer1': 1.1072578106809474, 'gradnorm/decoder.layer2': 0.6501012163951522, 'gradnorm/max_param': 142.85299682617188, 'epoch': 6.53}

 65%|██████▌   | 12460/19090 [6:04:40<2:59:55,  1.63s/it]
 65%|██████▌   | 12461/19090 [6:04:44<3:54:02,  2.12s/it]
 65%|██████▌   | 12462/19090 [6:04:45<3:26:48,  1.87s/it]
 65%|██████▌   | 12463/19090 [6:04:46<3:07:31,  1.70s/it]
 65%|██████▌   | 12464/19090 [6:04:49<3:35:59,  1.96s/it]
 65%|██████▌   | 12465/19090 [6:04:50<3:13:59,  1.76s/it]
 65%|██████▌   | 12466/19090 [6:04:51<2:58:49,  1.62s/it]
 65%|██████▌   | 12467/19090 [6:04:55<3:47:25,  2.06s/it]
 65%|██████▌   | 12468/19090 [6:04:56<3:22:03,  1.83s/it]
 65%|██████▌   | 12469/19090 [6:04:57<3:03:58,  1.67s/it][gradnorm] step 12469: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 178.4

 65%|██████▌   | 12470/19090 [6:05:00<3:45:10,  2.04s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 257.4918518066406, 'learning_rate': 1.619013907116355e-05, 'recon_loss': 0.0016088930191472174, 'recon_static': 0.0006339790037600323, 'recon_motion': 0.0009749140997882932, 'recon_norm': 1.3174200266599656, 'recon_norm_std': 0.04095433773472905, 'target_norm': 1.6233456760644913, 'recon_cos': 0.6250096932053566, 'gradnorm/compressor.query': 0.1849466553987642, 'gradnorm/compressor.layer0': 6.399327254699836, 'gradnorm/compressor.layer1': 252.81803145988223, 'gradnorm/compressor.layer2': 44.71069655819231, 'gradnorm/compressor.layer3': 17.11465386600398, 'gradnorm/compressor.layer4': 5.23683040590349, 'gradnorm/compressor.layer5': 3.2410677895193376, 'gradnorm/compressor.layer6': 1.6920925096107875, 'gradnorm/compressor.layer7': 2.058961549508275, 'gradnorm/decoder.layer0': 2.2682123603300366, 'gradnorm/decoder.layer1': 1.2066784685745002, 'gradnorm/decoder.layer2': 0.7140432993771929, 'gradnorm/max_param': 178.37240600585938, 'epoch': 6.53}

 65%|██████▌   | 12470/19090 [6:05:00<3:45:10,  2.04s/it]
 65%|██████▌   | 12471/19090 [6:05:01<3:20:26,  1.82s/it]
 65%|██████▌   | 12472/19090 [6:05:03<3:02:49,  1.66s/it]
 65%|██████▌   | 12473/19090 [6:05:05<3:36:03,  1.96s/it]
 65%|██████▌   | 12474/19090 [6:05:07<3:14:06,  1.76s/it]
 65%|██████▌   | 12475/19090 [6:05:08<2:58:37,  1.62s/it]
 65%|██████▌   | 12476/19090 [6:05:11<3:38:32,  1.98s/it]
 65%|██████▌   | 12477/19090 [6:05:12<3:15:44,  1.78s/it]
 65%|██████▌   | 12478/19090 [6:05:13<3:00:12,  1.64s/it]
 65%|██████▌   | 12479/19090 [6:05:16<3:34:51,  1.95s/it][gradnorm] step 12479: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 206.2

 65%|██████▌   | 12480/19090 [6:05:17<3:12:53,  1.75s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 296.5600891113281, 'learning_rate': 1.614737307070316e-05, 'recon_loss': 0.001626153729739599, 'recon_static': 0.0006708251617965289, 'recon_motion': 0.0009553285912261344, 'recon_norm': 1.330409687757492, 'recon_norm_std': 0.041994479950517416, 'target_norm': 1.619647029042244, 'recon_cos': 0.6204042986035347, 'gradnorm/compressor.query': 0.19649752654394687, 'gradnorm/compressor.layer0': 7.012215718373605, 'gradnorm/compressor.layer1': 292.3421449047332, 'gradnorm/compressor.layer2': 46.36493320657326, 'gradnorm/compressor.layer3': 14.937646075057204, 'gradnorm/compressor.layer4': 6.202804288092358, 'gradnorm/compressor.layer5': 2.772709796696946, 'gradnorm/compressor.layer6': 1.7044159082833696, 'gradnorm/compressor.layer7': 2.317063236827392, 'gradnorm/decoder.layer0': 2.3303072130087767, 'gradnorm/decoder.layer1': 1.22070390624975, 'gradnorm/decoder.layer2': 0.7287624237676606, 'gradnorm/max_param': 206.20632934570312, 'epoch': 6.54}

 65%|██████▌   | 12480/19090 [6:05:17<3:12:53,  1.75s/it]
 65%|██████▌   | 12481/19090 [6:05:19<2:57:29,  1.61s/it]
 65%|██████▌   | 12482/19090 [6:05:21<3:31:42,  1.92s/it]
 65%|██████▌   | 12483/19090 [6:05:23<3:10:29,  1.73s/it]
 65%|██████▌   | 12484/19090 [6:05:24<2:55:39,  1.60s/it]
 65%|██████▌   | 12485/19090 [6:05:27<3:41:54,  2.02s/it]
 65%|██████▌   | 12486/19090 [6:05:28<3:17:36,  1.80s/it]
 65%|██████▌   | 12487/19090 [6:05:29<3:00:38,  1.64s/it]
 65%|██████▌   | 12488/19090 [6:05:32<3:33:34,  1.94s/it]
 65%|██████▌   | 12489/19090 [6:05:33<3:11:54,  1.74s/it][gradnorm] step 12489: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 193.8

 65%|██████▌   | 12490/19090 [6:05:35<2:56:48,  1.61s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 281.5662536621094, 'learning_rate': 1.6104636669097776e-05, 'recon_loss': 0.0016326084936736152, 'recon_static': 0.0006563104790984653, 'recon_motion': 0.0009762979578226804, 'recon_norm': 1.322321170568466, 'recon_norm_std': 0.04077375428751111, 'target_norm': 1.6216588407754897, 'recon_cos': 0.618389467895031, 'gradnorm/compressor.query': 0.18901208292027905, 'gradnorm/compressor.layer0': 6.819954809223703, 'gradnorm/compressor.layer1': 274.6622784211549, 'gradnorm/compressor.layer2': 58.155238270350594, 'gradnorm/compressor.layer3': 18.410301708130017, 'gradnorm/compressor.layer4': 6.60904629658769, 'gradnorm/compressor.layer5': 3.2909859222973408, 'gradnorm/compressor.layer6': 1.8975172913056242, 'gradnorm/compressor.layer7': 2.369917400224472, 'gradnorm/decoder.layer0': 2.575906064518397, 'gradnorm/decoder.layer1': 1.3291283351828007, 'gradnorm/decoder.layer2': 0.7303097393749263, 'gradnorm/max_param': 193.80613708496094, 'epoch': 6.54}

 65%|██████▌   | 12490/19090 [6:05:35<2:56:48,  1.61s/it]
 65%|██████▌   | 12491/19090 [6:05:37<3:30:59,  1.92s/it]
 65%|██████▌   | 12492/19090 [6:05:38<3:10:14,  1.73s/it]
 65%|██████▌   | 12493/19090 [6:05:40<2:55:33,  1.60s/it]
 65%|██████▌   | 12494/19090 [6:05:43<3:43:05,  2.03s/it]
 65%|██████▌   | 12495/19090 [6:05:44<3:18:27,  1.81s/it]
 65%|██████▌   | 12496/19090 [6:05:45<3:01:22,  1.65s/it]
 65%|██████▌   | 12497/19090 [6:05:48<3:31:59,  1.93s/it]
 65%|██████▌   | 12498/19090 [6:05:49<3:10:39,  1.74s/it]
 65%|██████▌   | 12499/19090 [6:05:51<2:56:12,  1.60s/it][gradnorm] step 12499: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 196.7

 65%|██████▌   | 12500/19090 [6:05:53<3:32:16,  1.93s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 283.108154296875, 'learning_rate': 1.6061930009237014e-05, 'recon_loss': 0.0016301974101224914, 'recon_static': 0.000672710096114315, 'recon_motion': 0.0009574873096426017, 'recon_norm': 1.332811203598976, 'recon_norm_std': 0.041348299384117125, 'target_norm': 1.6185044825077057, 'recon_cos': 0.6188823133707047, 'gradnorm/compressor.query': 0.19374179591990703, 'gradnorm/compressor.layer0': 6.314760313340595, 'gradnorm/compressor.layer1': 278.74854259708695, 'gradnorm/compressor.layer2': 45.96289954697838, 'gradnorm/compressor.layer3': 15.795968924929669, 'gradnorm/compressor.layer4': 4.830698259636922, 'gradnorm/compressor.layer5': 3.226416633024315, 'gradnorm/compressor.layer6': 1.8272363834226049, 'gradnorm/compressor.layer7': 2.0591781915794742, 'gradnorm/decoder.layer0': 2.0235315249480292, 'gradnorm/decoder.layer1': 1.2512136289815265, 'gradnorm/decoder.layer2': 0.6919645221551588, 'gradnorm/max_param': 196.65330505371094, 'epoch': 6.55}

 65%|██████▌   | 12500/19090 [6:05:53<3:32:16,  1.93s/it]
 65%|██████▌   | 12501/19090 [6:05:57<4:45:21,  2.60s/it]
 65%|██████▌   | 12502/19090 [6:05:59<4:01:50,  2.20s/it]
 65%|██████▌   | 12503/19090 [6:06:00<3:31:46,  1.93s/it]
 66%|██████▌   | 12504/19090 [6:06:01<3:10:39,  1.74s/it]
 66%|██████▌   | 12505/19090 [6:06:03<2:55:43,  1.60s/it]
 66%|██████▌   | 12506/19090 [6:06:04<2:45:26,  1.51s/it]
 66%|██████▌   | 12507/19090 [6:06:05<2:37:56,  1.44s/it]
 66%|██████▌   | 12508/19090 [6:06:06<2:33:02,  1.40s/it]
 66%|██████▌   | 12509/19090 [6:06:08<2:29:25,  1.36s/it][gradnorm] step 12509: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 177.2

 66%|██████▌   | 12510/19090 [6:06:09<2:27:01,  1.34s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 256.3769836425781, 'learning_rate': 1.601925323391105e-05, 'recon_loss': 0.0016740108811063692, 'recon_static': 0.0007132578903110698, 'recon_motion': 0.0009607530198991298, 'recon_norm': 1.3531923115253448, 'recon_norm_std': 0.0402006808668375, 'target_norm': 1.6183546483516693, 'recon_cos': 0.6112244695425033, 'gradnorm/compressor.query': 0.18507320779151396, 'gradnorm/compressor.layer0': 7.027215368042837, 'gradnorm/compressor.layer1': 251.1263843914056, 'gradnorm/compressor.layer2': 47.87724629602903, 'gradnorm/compressor.layer3': 16.734693223223143, 'gradnorm/compressor.layer4': 4.348763625607988, 'gradnorm/compressor.layer5': 2.9955002098586445, 'gradnorm/compressor.layer6': 2.2083714559851257, 'gradnorm/compressor.layer7': 1.9318742601139622, 'gradnorm/decoder.layer0': 2.1099310248070573, 'gradnorm/decoder.layer1': 1.2803009053534646, 'gradnorm/decoder.layer2': 0.6404621801251097, 'gradnorm/max_param': 177.18692016601562, 'epoch': 6.55}

 66%|██████▌   | 12510/19090 [6:06:09<2:27:01,  1.34s/it]
 66%|██████▌   | 12511/19090 [6:06:10<2:25:27,  1.33s/it]
 66%|██████▌   | 12512/19090 [6:06:12<2:24:07,  1.31s/it]
 66%|██████▌   | 12513/19090 [6:06:13<2:23:25,  1.31s/it]
 66%|██████▌   | 12514/19090 [6:06:15<2:34:52,  1.41s/it]
 66%|██████▌   | 12515/19090 [6:06:16<2:37:27,  1.44s/it]
 66%|██████▌   | 12516/19090 [6:06:18<3:03:03,  1.67s/it]
 66%|██████▌   | 12517/19090 [6:06:20<3:06:58,  1.71s/it]
 66%|██████▌   | 12518/19090 [6:06:21<2:53:05,  1.58s/it]
 66%|██████▌   | 12519/19090 [6:06:24<3:22:32,  1.85s/it][gradnorm] step 12519: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 168.9

 66%|██████▌   | 12520/19090 [6:06:25<3:15:08,  1.78s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 245.03768920898438, 'learning_rate': 1.597660648581017e-05, 'recon_loss': 0.0016608065459877252, 'recon_static': 0.0006889341995702126, 'recon_motion': 0.0009718723755213432, 'recon_norm': 1.3424457520246507, 'recon_norm_std': 0.04014696227386594, 'target_norm': 1.6198761969804765, 'recon_cos': 0.6133474797010422, 'gradnorm/compressor.query': 0.16534660960060416, 'gradnorm/compressor.layer0': 5.948225753779014, 'gradnorm/compressor.layer1': 239.4284046171214, 'gradnorm/compressor.layer2': 47.21948419349792, 'gradnorm/compressor.layer3': 20.102234278757347, 'gradnorm/compressor.layer4': 4.834537136956679, 'gradnorm/compressor.layer5': 3.019313789620561, 'gradnorm/compressor.layer6': 1.9979342879180237, 'gradnorm/compressor.layer7': 1.9878621737062712, 'gradnorm/decoder.layer0': 2.3491580875380356, 'gradnorm/decoder.layer1': 1.3886079196715626, 'gradnorm/decoder.layer2': 0.6695016148884607, 'gradnorm/max_param': 168.90248107910156, 'epoch': 6.56}

 66%|██████▌   | 12520/19090 [6:06:25<3:15:08,  1.78s/it]
 66%|██████▌   | 12521/19090 [6:06:27<3:00:21,  1.65s/it]
 66%|██████▌   | 12522/19090 [6:06:29<3:29:34,  1.91s/it]
 66%|██████▌   | 12523/19090 [6:06:31<3:28:03,  1.90s/it]
 66%|██████▌   | 12524/19090 [6:06:32<3:07:57,  1.72s/it]
 66%|██████▌   | 12525/19090 [6:06:35<3:21:25,  1.84s/it]
 66%|██████▌   | 12526/19090 [6:06:37<3:30:22,  1.92s/it]
 66%|██████▌   | 12527/19090 [6:06:38<3:09:27,  1.73s/it]
 66%|██████▌   | 12528/19090 [6:06:40<3:12:10,  1.76s/it]
 66%|██████▌   | 12529/19090 [6:06:42<3:22:55,  1.86s/it][gradnorm] step 12529: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 214.0

 66%|██████▌   | 12530/19090 [6:06:43<3:04:30,  1.69s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 309.2039794921875, 'learning_rate': 1.5933989907524232e-05, 'recon_loss': 0.001612915936857462, 'recon_static': 0.0006479253017459996, 'recon_motion': 0.0009649906613049097, 'recon_norm': 1.326339441537857, 'recon_norm_std': 0.041153051983565093, 'target_norm': 1.6191479712724686, 'recon_cos': 0.6229581475257874, 'gradnorm/compressor.query': 0.24228228744536637, 'gradnorm/compressor.layer0': 7.998609898910023, 'gradnorm/compressor.layer1': 303.3711500579777, 'gradnorm/compressor.layer2': 56.305673403502375, 'gradnorm/compressor.layer3': 16.301610125253923, 'gradnorm/compressor.layer4': 6.351624060375874, 'gradnorm/compressor.layer5': 3.805535631212732, 'gradnorm/compressor.layer6': 2.0390989804567137, 'gradnorm/compressor.layer7': 2.1562682441271317, 'gradnorm/decoder.layer0': 2.543806040753428, 'gradnorm/decoder.layer1': 1.4905966222209117, 'gradnorm/decoder.layer2': 0.7856275368151073, 'gradnorm/max_param': 214.0457763671875, 'epoch': 6.56}

 66%|██████▌   | 12530/19090 [6:06:43<3:04:30,  1.69s/it]
 66%|██████▌   | 12531/19090 [6:06:45<3:20:01,  1.83s/it]
 66%|██████▌   | 12532/19090 [6:06:47<3:23:09,  1.86s/it]
 66%|██████▌   | 12533/19090 [6:06:49<3:04:25,  1.69s/it]
 66%|██████▌   | 12534/19090 [6:06:51<3:23:38,  1.86s/it]
 66%|██████▌   | 12535/19090 [6:06:53<3:23:49,  1.87s/it]
 66%|██████▌   | 12536/19090 [6:06:54<3:05:04,  1.69s/it]
 66%|██████▌   | 12537/19090 [6:06:56<3:13:27,  1.77s/it]
 66%|██████▌   | 12538/19090 [6:06:58<3:29:13,  1.92s/it]
 66%|██████▌   | 12539/19090 [6:06:59<3:08:54,  1.73s/it][gradnorm] step 12539: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 140.3

 66%|██████▌   | 12540/19090 [6:07:01<3:13:31,  1.77s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 203.0888671875, 'learning_rate': 1.5891403641542226e-05, 'recon_loss': 0.0016899045731406659, 'recon_static': 0.0007328272695303895, 'recon_motion': 0.0009570773210725747, 'recon_norm': 1.3543233126401901, 'recon_norm_std': 0.04195472877472639, 'target_norm': 1.6200815320014954, 'recon_cos': 0.6077806740999222, 'gradnorm/compressor.query': 0.20085469954519256, 'gradnorm/compressor.layer0': 7.001995074839468, 'gradnorm/compressor.layer1': 198.78805649799486, 'gradnorm/compressor.layer2': 37.84030808563304, 'gradnorm/compressor.layer3': 14.283269760245561, 'gradnorm/compressor.layer4': 4.6381769884036155, 'gradnorm/compressor.layer5': 2.9551641296659246, 'gradnorm/compressor.layer6': 1.7528386934512856, 'gradnorm/compressor.layer7': 1.8197510471479859, 'gradnorm/decoder.layer0': 2.1644757148294986, 'gradnorm/decoder.layer1': 1.2886280136761425, 'gradnorm/decoder.layer2': 0.667567126795672, 'gradnorm/max_param': 140.25657653808594, 'epoch': 6.57}

 66%|██████▌   | 12540/19090 [6:07:01<3:13:31,  1.77s/it]
 66%|██████▌   | 12541/19090 [6:07:03<3:24:12,  1.87s/it]
 66%|██████▌   | 12542/19090 [6:07:05<3:05:09,  1.70s/it]
 66%|██████▌   | 12543/19090 [6:07:07<3:12:49,  1.77s/it]
 66%|██████▌   | 12544/19090 [6:07:09<3:28:39,  1.91s/it]
 66%|██████▌   | 12545/19090 [6:07:10<3:08:08,  1.72s/it]
 66%|██████▌   | 12546/19090 [6:07:12<3:03:05,  1.68s/it]
 66%|██████▌   | 12547/19090 [6:07:14<3:26:03,  1.89s/it]
 66%|██████▌   | 12548/19090 [6:07:15<3:06:20,  1.71s/it]
 66%|██████▌   | 12549/19090 [6:07:17<3:17:33,  1.81s/it][gradnorm] step 12549: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 152.2

 66%|██████▌   | 12550/19090 [6:07:19<3:16:20,  1.80s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 220.18228149414062, 'learning_rate': 1.58488478302518e-05, 'recon_loss': 0.001940401890897192, 'recon_static': 0.000971498696890194, 'recon_motion': 0.0009689031852758489, 'recon_norm': 1.4514541447162628, 'recon_norm_std': 0.03819980314001441, 'target_norm': 1.6210823446512221, 'recon_cos': 0.566706919670105, 'gradnorm/compressor.query': 0.19998573870774733, 'gradnorm/compressor.layer0': 7.7110112118120675, 'gradnorm/compressor.layer1': 215.72126474341837, 'gradnorm/compressor.layer2': 40.28240155452502, 'gradnorm/compressor.layer3': 14.556257358809594, 'gradnorm/compressor.layer4': 5.340213966512005, 'gradnorm/compressor.layer5': 2.960940076680937, 'gradnorm/compressor.layer6': 1.6321928317738668, 'gradnorm/compressor.layer7': 1.891861251312294, 'gradnorm/decoder.layer0': 2.1682210139691667, 'gradnorm/decoder.layer1': 1.362426052361627, 'gradnorm/decoder.layer2': 0.655302976576368, 'gradnorm/max_param': 152.21205139160156, 'epoch': 6.57}

 66%|██████▌   | 12550/19090 [6:07:19<3:16:20,  1.80s/it]
 66%|██████▌   | 12551/19090 [6:07:21<2:59:27,  1.65s/it]
 66%|██████▌   | 12552/19090 [6:07:22<3:08:41,  1.73s/it]
 66%|██████▌   | 12553/19090 [6:07:25<3:21:46,  1.85s/it]
 66%|██████▌   | 12554/19090 [6:07:26<3:03:28,  1.68s/it]
 66%|██████▌   | 12555/19090 [6:07:28<3:13:10,  1.77s/it]
 66%|██████▌   | 12556/19090 [6:07:30<3:16:06,  1.80s/it]
 66%|██████▌   | 12557/19090 [6:07:31<2:59:17,  1.65s/it]
 66%|██████▌   | 12558/19090 [6:07:33<3:21:13,  1.85s/it]
 66%|██████▌   | 12559/19090 [6:07:35<3:20:31,  1.84s/it][gradnorm] step 12559: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 166.5

 66%|██████▌   | 12560/19090 [6:07:36<3:02:32,  1.68s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 242.2913818359375, 'learning_rate': 1.5806322615938793e-05, 'recon_loss': 0.0018450228759320452, 'recon_static': 0.0008821274997899309, 'recon_motion': 0.0009628953950596042, 'recon_norm': 1.42444007396698, 'recon_norm_std': 0.039039991423487665, 'target_norm': 1.6204987585544586, 'recon_cos': 0.5821843937039375, 'gradnorm/compressor.query': 0.19553317001787698, 'gradnorm/compressor.layer0': 7.439897679681547, 'gradnorm/compressor.layer1': 235.97795183025045, 'gradnorm/compressor.layer2': 51.09959351140917, 'gradnorm/compressor.layer3': 17.34994698552104, 'gradnorm/compressor.layer4': 5.292828569984128, 'gradnorm/compressor.layer5': 3.4238431661463684, 'gradnorm/compressor.layer6': 1.6174371499021738, 'gradnorm/compressor.layer7': 1.9699775864928464, 'gradnorm/decoder.layer0': 1.9543583752175442, 'gradnorm/decoder.layer1': 1.210356443726794, 'gradnorm/decoder.layer2': 0.6449792865266065, 'gradnorm/max_param': 166.48031616210938, 'epoch': 6.58}

 66%|██████▌   | 12560/19090 [6:07:36<3:02:32,  1.68s/it]
 66%|██████▌   | 12561/19090 [6:07:39<3:28:45,  1.92s/it]
 66%|██████▌   | 12562/19090 [6:07:41<3:23:49,  1.87s/it]
 66%|██████▌   | 12563/19090 [6:07:42<3:04:54,  1.70s/it]
 66%|██████▌   | 12564/19090 [6:07:44<3:24:22,  1.88s/it]
 66%|██████▌   | 12565/19090 [6:07:46<3:14:46,  1.79s/it]
 66%|██████▌   | 12566/19090 [6:07:47<2:58:18,  1.64s/it]
 66%|██████▌   | 12567/19090 [6:07:50<3:29:24,  1.93s/it]
 66%|██████▌   | 12568/19090 [6:07:51<3:15:35,  1.80s/it]
 66%|██████▌   | 12569/19090 [6:07:53<3:00:39,  1.66s/it][gradnorm] step 12569: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 165.6

 66%|██████▌   | 12570/19090 [6:07:55<3:14:29,  1.79s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 240.07632446289062, 'learning_rate': 1.5763828140786698e-05, 'recon_loss': 0.0017473534971941262, 'recon_static': 0.0007667857280466706, 'recon_motion': 0.0009805678186239676, 'recon_norm': 1.3712306827306748, 'recon_norm_std': 0.04007718311622739, 'target_norm': 1.6241487324237824, 'recon_cos': 0.5981478899717331, 'gradnorm/compressor.query': 0.1662243148494696, 'gradnorm/compressor.layer0': 6.158387098821862, 'gradnorm/compressor.layer1': 234.6701225257489, 'gradnorm/compressor.layer2': 47.08927587406447, 'gradnorm/compressor.layer3': 16.235780098902605, 'gradnorm/compressor.layer4': 5.249447021426326, 'gradnorm/compressor.layer5': 2.823859354158612, 'gradnorm/compressor.layer6': 1.4332815552755556, 'gradnorm/compressor.layer7': 2.0380275628394506, 'gradnorm/decoder.layer0': 2.0072268808306526, 'gradnorm/decoder.layer1': 1.2028317279819938, 'gradnorm/decoder.layer2': 0.6336627005466847, 'gradnorm/max_param': 165.57479858398438, 'epoch': 6.58}

 66%|██████▌   | 12570/19090 [6:07:55<3:14:29,  1.79s/it]
 66%|██████▌   | 12571/19090 [6:07:57<3:17:23,  1.82s/it]
 66%|██████▌   | 12572/19090 [6:07:58<3:04:45,  1.70s/it]
 66%|██████▌   | 12573/19090 [6:08:00<3:07:32,  1.73s/it]
 66%|██████▌   | 12574/19090 [6:08:02<3:28:16,  1.92s/it]
 66%|██████▌   | 12575/19090 [6:08:04<3:12:26,  1.77s/it]
 66%|██████▌   | 12576/19090 [6:08:06<3:16:41,  1.81s/it]
 66%|██████▌   | 12577/19090 [6:08:08<3:22:57,  1.87s/it]
 66%|██████▌   | 12578/19090 [6:08:09<3:15:54,  1.81s/it]
 66%|██████▌   | 12579/19090 [6:08:11<3:15:18,  1.80s/it][gradnorm] step 12579: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 160.6

 66%|██████▌   | 12580/19090 [6:08:13<3:13:03,  1.78s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 232.50967407226562, 'learning_rate': 1.572136454687627e-05, 'recon_loss': 0.0016883609525393695, 'recon_static': 0.0007298672324395738, 'recon_motion': 0.0009584937346517108, 'recon_norm': 1.353434643149376, 'recon_norm_std': 0.039846825785934924, 'target_norm': 1.6191697984933853, 'recon_cos': 0.60760118663311, 'gradnorm/compressor.query': 0.1946246145247046, 'gradnorm/compressor.layer0': 6.75472602700888, 'gradnorm/compressor.layer1': 227.56268023887398, 'gradnorm/compressor.layer2': 43.56460999642514, 'gradnorm/compressor.layer3': 16.825060835525697, 'gradnorm/compressor.layer4': 4.820174204420615, 'gradnorm/compressor.layer5': 3.410866505001538, 'gradnorm/compressor.layer6': 1.9266536199178559, 'gradnorm/compressor.layer7': 2.033954639045288, 'gradnorm/decoder.layer0': 2.176635464926103, 'gradnorm/decoder.layer1': 1.2031975761420863, 'gradnorm/decoder.layer2': 0.6692236773629924, 'gradnorm/max_param': 160.561279296875, 'epoch': 6.59}

 66%|██████▌   | 12580/19090 [6:08:13<3:13:03,  1.78s/it]
 66%|██████▌   | 12581/19090 [6:08:15<3:15:49,  1.81s/it]
 66%|██████▌   | 12582/19090 [6:08:16<3:16:30,  1.81s/it]
 66%|██████▌   | 12583/19090 [6:08:18<3:10:56,  1.76s/it]
 66%|██████▌   | 12584/19090 [6:08:20<3:18:25,  1.83s/it]
 66%|██████▌   | 12585/19090 [6:08:22<3:07:42,  1.73s/it]
 66%|██████▌   | 12586/19090 [6:08:23<3:08:38,  1.74s/it]
 66%|██████▌   | 12587/19090 [6:08:26<3:25:37,  1.90s/it]
 66%|██████▌   | 12588/19090 [6:08:27<3:08:51,  1.74s/it]
 66%|██████▌   | 12589/19090 [6:08:29<3:12:22,  1.78s/it][gradnorm] step 12589: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 166.3

 66%|██████▌   | 12590/19090 [6:08:31<3:19:06,  1.84s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 240.6368865966797, 'learning_rate': 1.567893197618501e-05, 'recon_loss': 0.0016827477316837757, 'recon_static': 0.0007114177293260582, 'recon_motion': 0.000971330056199804, 'recon_norm': 1.346593400835991, 'recon_norm_std': 0.040229312144219875, 'target_norm': 1.6223564505577088, 'recon_cos': 0.6094782501459122, 'gradnorm/compressor.query': 0.2026626018861079, 'gradnorm/compressor.layer0': 7.607639749367234, 'gradnorm/compressor.layer1': 235.63470636506415, 'gradnorm/compressor.layer2': 45.02503101574445, 'gradnorm/compressor.layer3': 15.699097742286815, 'gradnorm/compressor.layer4': 5.436021450323207, 'gradnorm/compressor.layer5': 2.756896907120039, 'gradnorm/compressor.layer6': 1.9333386185453236, 'gradnorm/compressor.layer7': 1.900884863949211, 'gradnorm/decoder.layer0': 2.089938551989634, 'gradnorm/decoder.layer1': 1.2350175850886538, 'gradnorm/decoder.layer2': 0.6004311711079147, 'gradnorm/max_param': 166.2584228515625, 'epoch': 6.6}

 66%|██████▌   | 12590/19090 [6:08:31<3:19:06,  1.84s/it]
 66%|██████▌   | 12591/19090 [6:08:33<3:19:06,  1.84s/it]
 66%|██████▌   | 12592/19090 [6:08:34<3:11:06,  1.76s/it]
 66%|██████▌   | 12593/19090 [6:08:36<3:21:32,  1.86s/it]
 66%|██████▌   | 12594/19090 [6:08:38<3:18:01,  1.83s/it]
 66%|██████▌   | 12595/19090 [6:08:40<3:09:42,  1.75s/it]
 66%|██████▌   | 12596/19090 [6:08:42<3:14:56,  1.80s/it]
 66%|██████▌   | 12597/19090 [6:08:43<3:17:41,  1.83s/it]
 66%|██████▌   | 12598/19090 [6:08:45<3:11:41,  1.77s/it]
 66%|██████▌   | 12599/19090 [6:08:47<3:11:04,  1.77s/it][gradnorm] step 12599: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 155.8

 66%|██████▌   | 12600/19090 [6:08:49<3:09:03,  1.75s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 226.03053283691406, 'learning_rate': 1.563653057058666e-05, 'recon_loss': 0.0018738533573923633, 'recon_static': 0.0009170873498078435, 'recon_motion': 0.0009567660468746908, 'recon_norm': 1.4321051239967346, 'recon_norm_std': 0.03866869136691094, 'target_norm': 1.6178250074386598, 'recon_cos': 0.5773358687758445, 'gradnorm/compressor.query': 0.1801341042219789, 'gradnorm/compressor.layer0': 6.249555953464095, 'gradnorm/compressor.layer1': 220.84438455052464, 'gradnorm/compressor.layer2': 44.80242815249387, 'gradnorm/compressor.layer3': 14.784412650554469, 'gradnorm/compressor.layer4': 5.390816593911755, 'gradnorm/compressor.layer5': 3.0508061016013244, 'gradnorm/compressor.layer6': 1.6886198778373398, 'gradnorm/compressor.layer7': 2.086738689731518, 'gradnorm/decoder.layer0': 2.3256767764603263, 'gradnorm/decoder.layer1': 1.1222366727605726, 'gradnorm/decoder.layer2': 0.628356314052519, 'gradnorm/max_param': 155.81773376464844, 'epoch': 6.6}

 66%|██████▌   | 12600/19090 [6:08:49<3:09:03,  1.75s/it]
 66%|██████▌   | 12601/19090 [6:08:50<3:02:17,  1.69s/it]
 66%|██████▌   | 12602/19090 [6:08:52<3:12:41,  1.78s/it]
 66%|██████▌   | 12603/19090 [6:08:54<3:14:57,  1.80s/it]
 66%|██████▌   | 12604/19090 [6:08:56<3:08:43,  1.75s/it]
 66%|██████▌   | 12605/19090 [6:08:58<3:23:57,  1.89s/it]
 66%|██████▌   | 12606/19090 [6:08:59<3:07:39,  1.74s/it]
 66%|██████▌   | 12607/19090 [6:09:01<3:12:49,  1.78s/it]
 66%|██████▌   | 12608/19090 [6:09:03<3:20:41,  1.86s/it]
 66%|██████▌   | 12609/19090 [6:09:05<3:13:59,  1.80s/it][gradnorm] step 12609: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 150.9

 66%|██████▌   | 12610/19090 [6:09:07<3:22:21,  1.87s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 218.14614868164062, 'learning_rate': 1.5594160471850797e-05, 'recon_loss': 0.0017023812572006137, 'recon_static': 0.000730104828835465, 'recon_motion': 0.0009722764327307232, 'recon_norm': 1.3602899968624116, 'recon_norm_std': 0.03990319138392806, 'target_norm': 1.6227076649665833, 'recon_cos': 0.6068488880991936, 'gradnorm/compressor.query': 0.17008818785442195, 'gradnorm/compressor.layer0': 5.910290178898097, 'gradnorm/compressor.layer1': 213.89398476523363, 'gradnorm/compressor.layer2': 38.23171859386686, 'gradnorm/compressor.layer3': 17.14665601509519, 'gradnorm/compressor.layer4': 4.816448226956169, 'gradnorm/compressor.layer5': 2.974669808408183, 'gradnorm/compressor.layer6': 1.848085252826581, 'gradnorm/compressor.layer7': 1.877579313652498, 'gradnorm/decoder.layer0': 2.342815467168062, 'gradnorm/decoder.layer1': 1.25699095818508, 'gradnorm/decoder.layer2': 0.6492667583246098, 'gradnorm/max_param': 150.9232635498047, 'epoch': 6.61}

 66%|██████▌   | 12610/19090 [6:09:07<3:22:21,  1.87s/it]
 66%|██████▌   | 12611/19090 [6:09:09<3:21:56,  1.87s/it]
 66%|██████▌   | 12612/19090 [6:09:10<3:03:44,  1.70s/it]
 66%|██████▌   | 12613/19090 [6:09:12<3:14:49,  1.80s/it]
 66%|██████▌   | 12614/19090 [6:09:14<3:24:47,  1.90s/it]
 66%|██████▌   | 12615/19090 [6:09:15<3:05:11,  1.72s/it]
 66%|██████▌   | 12616/19090 [6:09:17<3:06:53,  1.73s/it]
 66%|██████▌   | 12617/19090 [6:09:20<3:28:08,  1.93s/it]
 66%|██████▌   | 12618/19090 [6:09:21<3:07:25,  1.74s/it]
 66%|██████▌   | 12619/19090 [6:09:22<3:02:50,  1.70s/it][gradnorm] step 12619: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 204.2

 66%|██████▌   | 12620/19090 [6:09:25<3:28:45,  1.94s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 295.13726806640625, 'learning_rate': 1.5551821821642315e-05, 'recon_loss': 0.0017410814441973343, 'recon_static': 0.0007742911198874936, 'recon_motion': 0.0009667903548688627, 'recon_norm': 1.3759513288736342, 'recon_norm_std': 0.040237517468631266, 'target_norm': 1.6215723246335982, 'recon_cos': 0.5992657721042634, 'gradnorm/compressor.query': 0.34628248429007424, 'gradnorm/compressor.layer0': 11.210418582829371, 'gradnorm/compressor.layer1': 289.5608744288496, 'gradnorm/compressor.layer2': 52.78472827497078, 'gradnorm/compressor.layer3': 17.189428158891367, 'gradnorm/compressor.layer4': 5.494247549395022, 'gradnorm/compressor.layer5': 3.0185305181922235, 'gradnorm/compressor.layer6': 1.9122066921138174, 'gradnorm/compressor.layer7': 1.8720238908196878, 'gradnorm/decoder.layer0': 2.103161625309399, 'gradnorm/decoder.layer1': 1.276951114143009, 'gradnorm/decoder.layer2': 0.7014840368900319, 'gradnorm/max_param': 204.22235107421875, 'epoch': 6.61}

 66%|██████▌   | 12620/19090 [6:09:25<3:28:45,  1.94s/it]
 66%|██████▌   | 12621/19090 [6:09:26<3:08:02,  1.74s/it]
 66%|██████▌   | 12622/19090 [6:09:28<3:07:29,  1.74s/it]
 66%|██████▌   | 12623/19090 [6:09:30<3:25:14,  1.90s/it]
 66%|██████▌   | 12624/19090 [6:09:32<3:05:43,  1.72s/it]
 66%|██████▌   | 12625/19090 [6:09:33<2:58:05,  1.65s/it]
 66%|██████▌   | 12626/19090 [6:09:36<3:32:58,  1.98s/it]
 66%|██████▌   | 12627/19090 [6:09:37<3:10:39,  1.77s/it]
 66%|██████▌   | 12628/19090 [6:09:39<3:00:32,  1.68s/it]
 66%|██████▌   | 12629/19090 [6:09:41<3:28:03,  1.93s/it][gradnorm] step 12629: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 178.0

 66%|██████▌   | 12630/19090 [6:09:42<3:07:31,  1.74s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 257.4164123535156, 'learning_rate': 1.550951476152094e-05, 'recon_loss': 0.0017520863300887867, 'recon_static': 0.0007946155121317134, 'recon_motion': 0.0009574707873980515, 'recon_norm': 1.3849462479352952, 'recon_norm_std': 0.03972218008711934, 'target_norm': 1.6187768459320069, 'recon_cos': 0.5965486034750939, 'gradnorm/compressor.query': 0.1854356231446616, 'gradnorm/compressor.layer0': 7.013591106207869, 'gradnorm/compressor.layer1': 252.29234964124458, 'gradnorm/compressor.layer2': 47.480014812978155, 'gradnorm/compressor.layer3': 16.396842540943666, 'gradnorm/compressor.layer4': 4.508070807766545, 'gradnorm/compressor.layer5': 2.6794759149680827, 'gradnorm/compressor.layer6': 1.6653288558297072, 'gradnorm/compressor.layer7': 1.7648968165551628, 'gradnorm/decoder.layer0': 2.0379793644647464, 'gradnorm/decoder.layer1': 1.1965029902789461, 'gradnorm/decoder.layer2': 0.6070385916522459, 'gradnorm/max_param': 177.9709930419922, 'epoch': 6.62}

 66%|██████▌   | 12630/19090 [6:09:42<3:07:31,  1.74s/it]
 66%|██████▌   | 12631/19090 [6:09:44<3:12:49,  1.79s/it]
 66%|██████▌   | 12632/19090 [6:09:46<3:18:23,  1.84s/it]
 66%|██████▌   | 12633/19090 [6:09:48<3:00:44,  1.68s/it]
 66%|██████▌   | 12634/19090 [6:09:50<3:14:58,  1.81s/it]
 66%|██████▌   | 12635/19090 [6:09:52<3:30:25,  1.96s/it]
 66%|██████▌   | 12636/19090 [6:09:53<3:09:17,  1.76s/it]
 66%|██████▌   | 12637/19090 [6:09:55<3:05:31,  1.73s/it]
 66%|██████▌   | 12638/19090 [6:09:57<3:27:38,  1.93s/it]
 66%|██████▌   | 12639/19090 [6:09:59<3:06:58,  1.74s/it][gradnorm] step 12639: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 167.5

 66%|██████▌   | 12640/19090 [6:10:00<2:56:07,  1.64s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 243.038330078125, 'learning_rate': 1.54672394329408e-05, 'recon_loss': 0.0018060573027469217, 'recon_static': 0.0008374793265829794, 'recon_motion': 0.0009685779688879848, 'recon_norm': 1.4006646871566772, 'recon_norm_std': 0.03953950917348266, 'target_norm': 1.6220438718795775, 'recon_cos': 0.5872609734535217, 'gradnorm/compressor.query': 0.30160942256128526, 'gradnorm/compressor.layer0': 10.261409828929263, 'gradnorm/compressor.layer1': 237.476873874068, 'gradnorm/compressor.layer2': 47.134003201643075, 'gradnorm/compressor.layer3': 16.959518703613117, 'gradnorm/compressor.layer4': 5.723924307567297, 'gradnorm/compressor.layer5': 3.296108789583711, 'gradnorm/compressor.layer6': 1.72447278355606, 'gradnorm/compressor.layer7': 2.093719026706537, 'gradnorm/decoder.layer0': 2.1540220609294947, 'gradnorm/decoder.layer1': 1.2651593328859811, 'gradnorm/decoder.layer2': 0.7112950274891768, 'gradnorm/max_param': 167.5449676513672, 'epoch': 6.62}

 66%|██████▌   | 12640/19090 [6:10:00<2:56:07,  1.64s/it]
 66%|██████▌   | 12641/19090 [6:10:02<3:22:59,  1.89s/it]
 66%|██████▌   | 12642/19090 [6:10:04<3:03:51,  1.71s/it]
 66%|██████▌   | 12643/19090 [6:10:05<2:51:16,  1.59s/it]
 66%|██████▌   | 12644/19090 [6:10:08<3:23:05,  1.89s/it]
 66%|██████▌   | 12645/19090 [6:10:09<3:03:27,  1.71s/it]
 66%|██████▌   | 12646/19090 [6:10:10<2:49:52,  1.58s/it]
 66%|██████▌   | 12647/19090 [6:10:13<3:24:16,  1.90s/it]
 66%|██████▋   | 12648/19090 [6:10:14<3:04:27,  1.72s/it]
 66%|██████▋   | 12649/19090 [6:10:15<2:50:44,  1.59s/it][gradnorm] step 12649: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 165.5

 66%|██████▋   | 12650/19090 [6:10:18<3:27:26,  1.93s/it]
                                                         
{'loss': 0.0022, 'grad_norm': 240.1595458984375, 'learning_rate': 1.542499597724992e-05, 'recon_loss': 0.002192776827723719, 'recon_static': 0.0012314754640101455, 'recon_motion': 0.0009613013666239567, 'recon_norm': 1.5530562162399293, 'recon_norm_std': 0.03638545460999012, 'target_norm': 1.61939577460289, 'recon_cos': 0.5314657784998417, 'gradnorm/compressor.query': 0.23667511018418258, 'gradnorm/compressor.layer0': 8.602133694704035, 'gradnorm/compressor.layer1': 234.69093705871984, 'gradnorm/compressor.layer2': 45.705432634049366, 'gradnorm/compressor.layer3': 19.594001537211188, 'gradnorm/compressor.layer4': 5.222389642362982, 'gradnorm/compressor.layer5': 3.0559132645965765, 'gradnorm/compressor.layer6': 1.7538968704074782, 'gradnorm/compressor.layer7': 1.9617264213092986, 'gradnorm/decoder.layer0': 2.079421592582535, 'gradnorm/decoder.layer1': 1.2227067053380984, 'gradnorm/decoder.layer2': 0.672795618663157, 'gradnorm/max_param': 165.54464721679688, 'epoch': 6.63}

 66%|██████▋   | 12650/19090 [6:10:18<3:27:26,  1.93s/it]
 66%|██████▋   | 12651/19090 [6:10:19<3:06:51,  1.74s/it]
 66%|██████▋   | 12652/19090 [6:10:21<2:52:36,  1.61s/it]
 66%|██████▋   | 12653/19090 [6:10:23<3:23:59,  1.90s/it]
 66%|██████▋   | 12654/19090 [6:10:25<3:04:22,  1.72s/it]
 66%|██████▋   | 12655/19090 [6:10:26<2:50:40,  1.59s/it]
 66%|██████▋   | 12656/19090 [6:10:29<3:30:18,  1.96s/it]
 66%|██████▋   | 12657/19090 [6:10:30<3:08:28,  1.76s/it]
 66%|██████▋   | 12658/19090 [6:10:31<2:56:52,  1.65s/it]
 66%|██████▋   | 12659/19090 [6:10:35<3:48:33,  2.13s/it][gradnorm] step 12659: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 161.5

 66%|██████▋   | 12660/19090 [6:10:36<3:21:35,  1.88s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 233.65591430664062, 'learning_rate': 1.5382784535689742e-05, 'recon_loss': 0.0019381430465728045, 'recon_static': 0.000978267281607259, 'recon_motion': 0.0009598757897038013, 'recon_norm': 1.4575931757688523, 'recon_norm_std': 0.037754641845822334, 'target_norm': 1.6214616537094115, 'recon_cos': 0.5663581937551498, 'gradnorm/compressor.query': 0.3071627332923154, 'gradnorm/compressor.layer0': 11.384928970617938, 'gradnorm/compressor.layer1': 228.8473068308867, 'gradnorm/compressor.layer2': 42.508765180695384, 'gradnorm/compressor.layer3': 15.407198801285837, 'gradnorm/compressor.layer4': 5.291034009865623, 'gradnorm/compressor.layer5': 3.0051641521588963, 'gradnorm/compressor.layer6': 1.6858090299600945, 'gradnorm/compressor.layer7': 2.002121991734669, 'gradnorm/decoder.layer0': 1.9396542294087074, 'gradnorm/decoder.layer1': 1.3525905337679154, 'gradnorm/decoder.layer2': 0.6601290951171824, 'gradnorm/max_param': 161.47354125976562, 'epoch': 6.63}

 66%|██████▋   | 12660/19090 [6:10:36<3:21:35,  1.88s/it]
 66%|██████▋   | 12661/19090 [6:10:37<3:02:28,  1.70s/it]
 66%|██████▋   | 12662/19090 [6:10:40<3:36:05,  2.02s/it]
 66%|██████▋   | 12663/19090 [6:10:41<3:12:38,  1.80s/it]
 66%|██████▋   | 12664/19090 [6:10:43<2:56:06,  1.64s/it]
 66%|██████▋   | 12665/19090 [6:10:45<3:30:43,  1.97s/it]
 66%|██████▋   | 12666/19090 [6:10:47<3:08:53,  1.76s/it]
 66%|██████▋   | 12667/19090 [6:10:48<2:53:35,  1.62s/it]
 66%|██████▋   | 12668/19090 [6:10:50<3:10:21,  1.78s/it]
 66%|██████▋   | 12669/19090 [6:10:51<2:54:32,  1.63s/it][gradnorm] step 12669: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 177.3

 66%|██████▋   | 12670/19090 [6:10:53<2:43:38,  1.53s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 256.7326965332031, 'learning_rate': 1.5340605249394703e-05, 'recon_loss': 0.0019392067100852729, 'recon_static': 0.0009701585731818341, 'recon_motion': 0.0009690481165307574, 'recon_norm': 1.4542054206132888, 'recon_norm_std': 0.03801063671708107, 'target_norm': 1.6208584576845169, 'recon_cos': 0.5665875524282455, 'gradnorm/compressor.query': 0.25472188670642304, 'gradnorm/compressor.layer0': 8.979783558435452, 'gradnorm/compressor.layer1': 251.2289325744549, 'gradnorm/compressor.layer2': 48.36881470873692, 'gradnorm/compressor.layer3': 17.98116503852234, 'gradnorm/compressor.layer4': 5.2160153105224385, 'gradnorm/compressor.layer5': 2.8966386256234333, 'gradnorm/compressor.layer6': 2.156100613491696, 'gradnorm/compressor.layer7': 2.1688461590207413, 'gradnorm/decoder.layer0': 2.2199634940780135, 'gradnorm/decoder.layer1': 1.2984146722102081, 'gradnorm/decoder.layer2': 0.7635234483641841, 'gradnorm/max_param': 177.25721740722656, 'epoch': 6.64}

 66%|██████▋   | 12670/19090 [6:10:53<2:43:38,  1.53s/it]
 66%|██████▋   | 12671/19090 [6:10:55<3:16:46,  1.84s/it]
 66%|██████▋   | 12672/19090 [6:10:56<2:59:05,  1.67s/it]
 66%|██████▋   | 12673/19090 [6:10:58<2:47:53,  1.57s/it]
 66%|██████▋   | 12674/19090 [6:11:01<3:24:57,  1.92s/it]
 66%|██████▋   | 12675/19090 [6:11:02<3:04:44,  1.73s/it]
 66%|██████▋   | 12676/19090 [6:11:03<3:00:33,  1.69s/it]
 66%|██████▋   | 12677/19090 [6:11:06<3:33:01,  1.99s/it]
 66%|██████▋   | 12678/19090 [6:11:07<3:10:28,  1.78s/it]
 66%|██████▋   | 12679/19090 [6:11:09<2:54:30,  1.63s/it][gradnorm] step 12679: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 164.8

 66%|██████▋   | 12680/19090 [6:11:12<3:32:53,  1.99s/it]
                                                         
{'loss': 0.002, 'grad_norm': 238.2424774169922, 'learning_rate': 1.52984582593917e-05, 'recon_loss': 0.001962300075683743, 'recon_static': 0.0009972142084734513, 'recon_motion': 0.0009650858861277811, 'recon_norm': 1.4656953006982802, 'recon_norm_std': 0.03851100271567702, 'target_norm': 1.6207410991191864, 'recon_cos': 0.5627362832427025, 'gradnorm/compressor.query': 0.17137585575639963, 'gradnorm/compressor.layer0': 7.002123374412099, 'gradnorm/compressor.layer1': 233.55381974771896, 'gradnorm/compressor.layer2': 43.252336913021246, 'gradnorm/compressor.layer3': 15.374207018740124, 'gradnorm/compressor.layer4': 5.423379387045505, 'gradnorm/compressor.layer5': 3.2571098612784772, 'gradnorm/compressor.layer6': 2.0657558751738865, 'gradnorm/compressor.layer7': 2.0053668731503858, 'gradnorm/decoder.layer0': 2.233183322916787, 'gradnorm/decoder.layer1': 1.4560004057254854, 'gradnorm/decoder.layer2': 0.6920775479031452, 'gradnorm/max_param': 164.7869873046875, 'epoch': 6.64}

 66%|██████▋   | 12680/19090 [6:11:12<3:32:53,  1.99s/it]
 66%|██████▋   | 12681/19090 [6:11:13<3:10:22,  1.78s/it]
 66%|██████▋   | 12682/19090 [6:11:14<2:54:22,  1.63s/it]
 66%|██████▋   | 12683/19090 [6:11:17<3:25:07,  1.92s/it]
 66%|██████▋   | 12684/19090 [6:11:18<3:04:51,  1.73s/it]
 66%|██████▋   | 12685/19090 [6:11:19<2:50:37,  1.60s/it]
 66%|██████▋   | 12686/19090 [6:11:22<3:15:06,  1.83s/it]
 66%|██████▋   | 12687/19090 [6:11:23<2:57:36,  1.66s/it]
 66%|██████▋   | 12688/19090 [6:11:24<2:49:46,  1.59s/it]
 66%|██████▋   | 12689/19090 [6:11:27<3:26:27,  1.94s/it][gradnorm] step 12689: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 158.7

 66%|██████▋   | 12690/19090 [6:11:29<3:34:23,  2.01s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 229.62054443359375, 'learning_rate': 1.5256343706599646e-05, 'recon_loss': 0.0017306580732110888, 'recon_static': 0.0007642376047442667, 'recon_motion': 0.000966420529584866, 'recon_norm': 1.3746841073036193, 'recon_norm_std': 0.0394324503839016, 'target_norm': 1.6202451020479203, 'recon_cos': 0.6008374273777009, 'gradnorm/compressor.query': 0.15323935637142733, 'gradnorm/compressor.layer0': 5.634893236748216, 'gradnorm/compressor.layer1': 224.94633040461008, 'gradnorm/compressor.layer2': 42.34304946376235, 'gradnorm/compressor.layer3': 15.668679182503507, 'gradnorm/compressor.layer4': 5.481018604366003, 'gradnorm/compressor.layer5': 3.0666682623430606, 'gradnorm/compressor.layer6': 2.0905032819545615, 'gradnorm/compressor.layer7': 1.7165061869500053, 'gradnorm/decoder.layer0': 2.322759089805941, 'gradnorm/decoder.layer1': 1.368158050458552, 'gradnorm/decoder.layer2': 0.7053355542202342, 'gradnorm/max_param': 158.73306274414062, 'epoch': 6.65}

 66%|██████▋   | 12690/19090 [6:11:29<3:34:23,  2.01s/it]
 66%|██████▋   | 12691/19090 [6:11:31<3:11:17,  1.79s/it]
 66%|██████▋   | 12692/19090 [6:11:32<3:03:05,  1.72s/it]
 66%|██████▋   | 12693/19090 [6:11:33<2:49:18,  1.59s/it]
 66%|██████▋   | 12694/19090 [6:11:35<2:50:19,  1.60s/it]
 67%|██████▋   | 12695/19090 [6:11:37<3:15:09,  1.83s/it]
 67%|██████▋   | 12696/19090 [6:11:39<2:57:50,  1.67s/it]
 67%|██████▋   | 12697/19090 [6:11:40<2:56:25,  1.66s/it]
 67%|██████▋   | 12698/19090 [6:11:43<3:24:26,  1.92s/it]
 67%|██████▋   | 12699/19090 [6:11:44<3:04:18,  1.73s/it][gradnorm] step 12699: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 185.0

 67%|██████▋   | 12700/19090 [6:11:46<3:05:55,  1.75s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 266.79461669921875, 'learning_rate': 1.5214261731829022e-05, 'recon_loss': 0.001734611569554545, 'recon_static': 0.0007696055195992813, 'recon_motion': 0.0009650060543208383, 'recon_norm': 1.372962561249733, 'recon_norm_std': 0.03936906512826681, 'target_norm': 1.6204938769340516, 'recon_cos': 0.5999280229210854, 'gradnorm/compressor.query': 0.20506860620115122, 'gradnorm/compressor.layer0': 7.53355226901885, 'gradnorm/compressor.layer1': 262.1748134594549, 'gradnorm/compressor.layer2': 46.02921329506404, 'gradnorm/compressor.layer3': 14.750586546534239, 'gradnorm/compressor.layer4': 5.164272102107372, 'gradnorm/compressor.layer5': 3.1681425938256016, 'gradnorm/compressor.layer6': 1.860914987064377, 'gradnorm/compressor.layer7': 1.781317525554137, 'gradnorm/decoder.layer0': 2.239708037933589, 'gradnorm/decoder.layer1': 1.3224722899362509, 'gradnorm/decoder.layer2': 0.7508401139094651, 'gradnorm/max_param': 184.97982788085938, 'epoch': 6.65}

 67%|██████▋   | 12700/19090 [6:11:46<3:05:55,  1.75s/it]
 67%|██████▋   | 12701/19090 [6:11:48<3:30:55,  1.98s/it]
 67%|██████▋   | 12702/19090 [6:11:50<3:08:45,  1.77s/it]
 67%|██████▋   | 12703/19090 [6:11:51<3:01:17,  1.70s/it]
 67%|██████▋   | 12704/19090 [6:11:54<3:32:23,  2.00s/it]
 67%|██████▋   | 12705/19090 [6:11:55<3:09:42,  1.78s/it]
 67%|██████▋   | 12706/19090 [6:11:57<2:53:58,  1.64s/it]
 67%|██████▋   | 12707/19090 [6:11:59<3:29:49,  1.97s/it]
 67%|██████▋   | 12708/19090 [6:12:01<3:08:03,  1.77s/it]
 67%|██████▋   | 12709/19090 [6:12:02<2:53:07,  1.63s/it][gradnorm] step 12709: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 188.2

 67%|██████▋   | 12710/19090 [6:12:04<3:24:33,  1.92s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 271.73638916015625, 'learning_rate': 1.5172212475781367e-05, 'recon_loss': 0.0016302346310112625, 'recon_static': 0.0006641319341724738, 'recon_motion': 0.0009661026517278515, 'recon_norm': 1.3265716135501862, 'recon_norm_std': 0.040659682173281905, 'target_norm': 1.6214767694473267, 'recon_cos': 0.619897823035717, 'gradnorm/compressor.query': 0.16997183690053233, 'gradnorm/compressor.layer0': 6.688422308807738, 'gradnorm/compressor.layer1': 266.71394437674235, 'gradnorm/compressor.layer2': 47.931681014980846, 'gradnorm/compressor.layer3': 17.67476818574778, 'gradnorm/compressor.layer4': 5.109770313333986, 'gradnorm/compressor.layer5': 3.0625473330206976, 'gradnorm/compressor.layer6': 1.7786205358833271, 'gradnorm/compressor.layer7': 1.906609548211126, 'gradnorm/decoder.layer0': 2.340084006276934, 'gradnorm/decoder.layer1': 1.1971159650101786, 'gradnorm/decoder.layer2': 0.7322450958535751, 'gradnorm/max_param': 188.19631958007812, 'epoch': 6.66}

 67%|██████▋   | 12710/19090 [6:12:04<3:24:33,  1.92s/it]
 67%|██████▋   | 12711/19090 [6:12:06<3:04:09,  1.73s/it]
 67%|██████▋   | 12712/19090 [6:12:08<3:07:23,  1.76s/it]
 67%|██████▋   | 12713/19090 [6:12:10<3:15:19,  1.84s/it]
 67%|██████▋   | 12714/19090 [6:12:11<2:57:39,  1.67s/it]
 67%|██████▋   | 12715/19090 [6:12:13<3:10:50,  1.80s/it]
 67%|██████▋   | 12716/19090 [6:12:15<3:08:47,  1.78s/it]
 67%|██████▋   | 12717/19090 [6:12:16<2:53:05,  1.63s/it]
 67%|██████▋   | 12718/19090 [6:12:18<3:10:38,  1.80s/it]
 67%|██████▋   | 12719/19090 [6:12:20<3:13:56,  1.83s/it][gradnorm] step 12719: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 179.5

 67%|██████▋   | 12720/19090 [6:12:21<2:58:28,  1.68s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 260.82989501953125, 'learning_rate': 1.513019607904882e-05, 'recon_loss': 0.0017158651375211776, 'recon_static': 0.0007417798580718227, 'recon_motion': 0.0009740852867253125, 'recon_norm': 1.3590073704719543, 'recon_norm_std': 0.04046033238992095, 'target_norm': 1.6206665068864823, 'recon_cos': 0.6042106866836547, 'gradnorm/compressor.query': 0.26590090335102146, 'gradnorm/compressor.layer0': 9.444728483030312, 'gradnorm/compressor.layer1': 254.47702959943163, 'gradnorm/compressor.layer2': 52.09121216448725, 'gradnorm/compressor.layer3': 20.19142023057651, 'gradnorm/compressor.layer4': 6.05481728691541, 'gradnorm/compressor.layer5': 3.1418987367217874, 'gradnorm/compressor.layer6': 1.932043698861152, 'gradnorm/compressor.layer7': 2.21750713514967, 'gradnorm/decoder.layer0': 2.3774407544059746, 'gradnorm/decoder.layer1': 1.287766145771218, 'gradnorm/decoder.layer2': 0.8516119146446184, 'gradnorm/max_param': 179.53990173339844, 'epoch': 6.66}

 67%|██████▋   | 12720/19090 [6:12:21<2:58:28,  1.68s/it]
 67%|██████▋   | 12721/19090 [6:12:23<3:10:23,  1.79s/it]
 67%|██████▋   | 12722/19090 [6:12:26<3:18:53,  1.87s/it]
 67%|██████▋   | 12723/19090 [6:12:27<3:02:46,  1.72s/it]
 67%|██████▋   | 12724/19090 [6:12:28<2:55:37,  1.66s/it]
 67%|██████▋   | 12725/19090 [6:12:31<3:18:09,  1.87s/it]
 67%|██████▋   | 12726/19090 [6:12:32<3:10:50,  1.80s/it]
 67%|██████▋   | 12727/19090 [6:12:34<2:54:33,  1.65s/it]
 67%|██████▋   | 12728/19090 [6:12:36<3:18:54,  1.88s/it]
 67%|██████▋   | 12729/19090 [6:12:38<3:23:50,  1.92s/it][gradnorm] step 12729: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 157.8

 67%|██████▋   | 12730/19090 [6:12:39<3:04:13,  1.74s/it]
                                                         
{'loss': 0.002, 'grad_norm': 228.24501037597656, 'learning_rate': 1.5088212682113678e-05, 'recon_loss': 0.001990769515396096, 'recon_static': 0.0010270045328070409, 'recon_motion': 0.0009637648763600737, 'recon_norm': 1.4796946942806244, 'recon_norm_std': 0.0369876092299819, 'target_norm': 1.6195575207471848, 'recon_cos': 0.55916668176651, 'gradnorm/compressor.query': 0.19858496960934177, 'gradnorm/compressor.layer0': 7.047125902805358, 'gradnorm/compressor.layer1': 223.64748000480577, 'gradnorm/compressor.layer2': 41.53598631523859, 'gradnorm/compressor.layer3': 15.706384242767184, 'gradnorm/compressor.layer4': 4.979558931910744, 'gradnorm/compressor.layer5': 3.546912777065135, 'gradnorm/compressor.layer6': 2.0423458163502164, 'gradnorm/compressor.layer7': 2.297499176124865, 'gradnorm/decoder.layer0': 2.5787646800523403, 'gradnorm/decoder.layer1': 1.3520967607267376, 'gradnorm/decoder.layer2': 0.8034082633077139, 'gradnorm/max_param': 157.79177856445312, 'epoch': 6.67}

 67%|██████▋   | 12730/19090 [6:12:39<3:04:13,  1.74s/it]
 67%|██████▋   | 12731/19090 [6:12:41<3:13:59,  1.83s/it]
 67%|██████▋   | 12732/19090 [6:12:44<3:19:56,  1.89s/it]
 67%|██████▋   | 12733/19090 [6:12:45<3:01:08,  1.71s/it]
 67%|██████▋   | 12734/19090 [6:12:47<3:07:34,  1.77s/it]
 67%|██████▋   | 12735/19090 [6:12:49<3:24:49,  1.93s/it]
 67%|██████▋   | 12736/19090 [6:12:50<3:04:21,  1.74s/it]
 67%|██████▋   | 12737/19090 [6:12:52<3:06:04,  1.76s/it]
 67%|██████▋   | 12738/19090 [6:12:54<3:00:21,  1.70s/it]
 67%|██████▋   | 12739/19090 [6:12:55<2:50:07,  1.61s/it][gradnorm] step 12739: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 202.3

 67%|██████▋   | 12740/19090 [6:12:57<3:11:49,  1.81s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 291.2348937988281, 'learning_rate': 1.5046262425347882e-05, 'recon_loss': 0.0018414624326396733, 'recon_static': 0.0008790686406427994, 'recon_motion': 0.0009623937905416824, 'recon_norm': 1.414188766479492, 'recon_norm_std': 0.03872607639059424, 'target_norm': 1.6204226583242416, 'recon_cos': 0.5805087417364121, 'gradnorm/compressor.query': 0.19357466993913441, 'gradnorm/compressor.layer0': 7.852077486220407, 'gradnorm/compressor.layer1': 286.72554658156986, 'gradnorm/compressor.layer2': 46.93177291125651, 'gradnorm/compressor.layer3': 17.080166456496862, 'gradnorm/compressor.layer4': 4.961502451529989, 'gradnorm/compressor.layer5': 3.2688821857048556, 'gradnorm/compressor.layer6': 2.0262905906080375, 'gradnorm/compressor.layer7': 2.0303775967975497, 'gradnorm/decoder.layer0': 2.204815155378057, 'gradnorm/decoder.layer1': 1.164915655463726, 'gradnorm/decoder.layer2': 0.6774043398182273, 'gradnorm/max_param': 202.2837371826172, 'epoch': 6.67}

 67%|██████▋   | 12740/19090 [6:12:57<3:11:49,  1.81s/it]
 67%|██████▋   | 12741/19090 [6:12:59<3:19:46,  1.89s/it]
 67%|██████▋   | 12742/19090 [6:13:01<3:00:47,  1.71s/it]
 67%|██████▋   | 12743/19090 [6:13:03<3:18:22,  1.88s/it]
 67%|██████▋   | 12744/19090 [6:13:05<3:14:26,  1.84s/it]
 67%|██████▋   | 12745/19090 [6:13:06<2:56:56,  1.67s/it]
 67%|██████▋   | 12746/19090 [6:13:08<3:20:27,  1.90s/it]
 67%|██████▋   | 12747/19090 [6:13:10<3:16:04,  1.85s/it]
 67%|██████▋   | 12748/19090 [6:13:12<2:58:13,  1.69s/it]
 67%|██████▋   | 12749/19090 [6:13:14<3:19:47,  1.89s/it][gradnorm] step 12749: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 170.5

 67%|██████▋   | 12750/19090 [6:13:15<3:04:00,  1.74s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 246.22573852539062, 'learning_rate': 1.5004345449012563e-05, 'recon_loss': 0.0016536372160771862, 'recon_static': 0.0007088120124535635, 'recon_motion': 0.0009448251978028566, 'recon_norm': 1.3469216734170915, 'recon_norm_std': 0.041714598331600426, 'target_norm': 1.61715047955513, 'recon_cos': 0.6143443331122398, 'gradnorm/compressor.query': 0.15988746045419078, 'gradnorm/compressor.layer0': 5.8799544201595895, 'gradnorm/compressor.layer1': 241.6118804637926, 'gradnorm/compressor.layer2': 43.46189035334218, 'gradnorm/compressor.layer3': 16.5887294149828, 'gradnorm/compressor.layer4': 5.220357316163603, 'gradnorm/compressor.layer5': 3.045380680736086, 'gradnorm/compressor.layer6': 1.9420922523767874, 'gradnorm/compressor.layer7': 2.2039412481158576, 'gradnorm/decoder.layer0': 2.2196671645360664, 'gradnorm/decoder.layer1': 1.265485873640503, 'gradnorm/decoder.layer2': 0.6384906803412835, 'gradnorm/max_param': 170.47317504882812, 'epoch': 6.68}

 67%|██████▋   | 12750/19090 [6:13:15<3:04:00,  1.74s/it]
 67%|██████▋   | 12751/19090 [6:13:17<2:49:37,  1.61s/it]
 67%|██████▋   | 12752/19090 [6:13:19<3:15:14,  1.85s/it]
 67%|██████▋   | 12753/19090 [6:13:21<3:12:23,  1.82s/it]
 67%|██████▋   | 12754/19090 [6:13:22<2:55:22,  1.66s/it]
 67%|██████▋   | 12755/19090 [6:13:24<3:00:39,  1.71s/it]
 67%|██████▋   | 12756/19090 [6:13:26<3:18:45,  1.88s/it]
 67%|██████▋   | 12757/19090 [6:13:27<3:00:03,  1.71s/it]
 67%|██████▋   | 12758/19090 [6:13:29<3:06:19,  1.77s/it]
 67%|██████▋   | 12759/19090 [6:13:31<3:15:09,  1.85s/it][gradnorm] step 12759: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 195.8

 67%|██████▋   | 12760/19090 [6:13:33<2:58:19,  1.69s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 282.9200134277344, 'learning_rate': 1.4962461893257606e-05, 'recon_loss': 0.0016212373448070139, 'recon_static': 0.0006624547182582319, 'recon_motion': 0.0009587825756170787, 'recon_norm': 1.326297327876091, 'recon_norm_std': 0.041389019694179296, 'target_norm': 1.6201752364635467, 'recon_cos': 0.6213353708386421, 'gradnorm/compressor.query': 0.16173335783399678, 'gradnorm/compressor.layer0': 6.314343475985041, 'gradnorm/compressor.layer1': 277.44868613403094, 'gradnorm/compressor.layer2': 51.86610192887909, 'gradnorm/compressor.layer3': 16.756600517070037, 'gradnorm/compressor.layer4': 5.419590185515014, 'gradnorm/compressor.layer5': 3.2774936064331173, 'gradnorm/compressor.layer6': 2.011210375502779, 'gradnorm/compressor.layer7': 1.9900515007179196, 'gradnorm/decoder.layer0': 2.2522368438468567, 'gradnorm/decoder.layer1': 1.2506691095522167, 'gradnorm/decoder.layer2': 0.6682789177237362, 'gradnorm/max_param': 195.796142578125, 'epoch': 6.68}

 67%|██████▋   | 12760/19090 [6:13:33<2:58:19,  1.69s/it]
 67%|██████▋   | 12761/19090 [6:13:35<3:08:54,  1.79s/it]
 67%|██████▋   | 12762/19090 [6:13:37<3:17:29,  1.87s/it]
 67%|██████▋   | 12763/19090 [6:13:38<2:59:18,  1.70s/it]
 67%|██████▋   | 12764/19090 [6:13:40<3:11:06,  1.81s/it]
 67%|██████▋   | 12765/19090 [6:13:42<3:12:23,  1.82s/it]
 67%|██████▋   | 12766/19090 [6:13:43<2:55:25,  1.66s/it]
 67%|██████▋   | 12767/19090 [6:13:45<2:59:25,  1.70s/it]
 67%|██████▋   | 12768/19090 [6:13:47<3:17:10,  1.87s/it]
 67%|██████▋   | 12769/19090 [6:13:49<2:58:52,  1.70s/it][gradnorm] step 12769: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 178.6

 67%|██████▋   | 12770/19090 [6:13:50<2:50:04,  1.61s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 257.8955383300781, 'learning_rate': 1.4920611898121139e-05, 'recon_loss': 0.001686549390433356, 'recon_static': 0.0007187433700892143, 'recon_motion': 0.0009678060101578012, 'recon_norm': 1.3453450471162796, 'recon_norm_std': 0.0414044258184731, 'target_norm': 1.6204269587993623, 'recon_cos': 0.6076580494642257, 'gradnorm/compressor.query': 0.18028829519557615, 'gradnorm/compressor.layer0': 6.570942416482026, 'gradnorm/compressor.layer1': 253.18399775410373, 'gradnorm/compressor.layer2': 45.00415833651514, 'gradnorm/compressor.layer3': 16.76755162072017, 'gradnorm/compressor.layer4': 5.4246646619248216, 'gradnorm/compressor.layer5': 3.334208484842834, 'gradnorm/compressor.layer6': 2.1347361402927234, 'gradnorm/compressor.layer7': 2.2539944689326554, 'gradnorm/decoder.layer0': 2.420011430981334, 'gradnorm/decoder.layer1': 1.3177233708947218, 'gradnorm/decoder.layer2': 0.6788694347588261, 'gradnorm/max_param': 178.60728454589844, 'epoch': 6.69}

 67%|██████▋   | 12770/19090 [6:13:50<2:50:04,  1.61s/it]
 67%|██████▋   | 12771/19090 [6:13:53<3:20:48,  1.91s/it]
 67%|██████▋   | 12772/19090 [6:13:54<3:01:40,  1.73s/it]
 67%|██████▋   | 12773/19090 [6:13:55<2:48:15,  1.60s/it]
 67%|██████▋   | 12774/19090 [6:13:58<3:23:21,  1.93s/it]
 67%|██████▋   | 12775/19090 [6:13:59<3:02:59,  1.74s/it]
 67%|██████▋   | 12776/19090 [6:14:01<2:56:14,  1.67s/it]
 67%|██████▋   | 12777/19090 [6:14:03<3:23:57,  1.94s/it]
 67%|██████▋   | 12778/19090 [6:14:05<3:03:38,  1.75s/it]
 67%|██████▋   | 12779/19090 [6:14:06<2:54:39,  1.66s/it][gradnorm] step 12779: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 217.6

 67%|██████▋   | 12780/19090 [6:14:09<3:23:28,  1.93s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 314.4946594238281, 'learning_rate': 1.4878795603529066e-05, 'recon_loss': 0.0017061599937733263, 'recon_static': 0.0007418582492391579, 'recon_motion': 0.0009643017736379988, 'recon_norm': 1.3676822900772094, 'recon_norm_std': 0.040003233961760996, 'target_norm': 1.619515061378479, 'recon_cos': 0.6056641414761543, 'gradnorm/compressor.query': 0.17489977333655046, 'gradnorm/compressor.layer0': 6.226423427124729, 'gradnorm/compressor.layer1': 308.44333201173276, 'gradnorm/compressor.layer2': 58.10495286010479, 'gradnorm/compressor.layer3': 17.090670069308832, 'gradnorm/compressor.layer4': 5.915146681707334, 'gradnorm/compressor.layer5': 3.219016943667455, 'gradnorm/compressor.layer6': 2.2834551878479394, 'gradnorm/compressor.layer7': 1.9004724843663532, 'gradnorm/decoder.layer0': 2.4727002180628777, 'gradnorm/decoder.layer1': 1.2787304272090643, 'gradnorm/decoder.layer2': 0.7621073978462949, 'gradnorm/max_param': 217.6265106201172, 'epoch': 6.69}

 67%|██████▋   | 12780/19090 [6:14:09<3:23:28,  1.93s/it]
 67%|██████▋   | 12781/19090 [6:14:10<3:03:05,  1.74s/it]
 67%|██████▋   | 12782/19090 [6:14:11<2:56:22,  1.68s/it]
 67%|██████▋   | 12783/19090 [6:14:14<3:27:19,  1.97s/it]
 67%|██████▋   | 12784/19090 [6:14:15<3:05:42,  1.77s/it]
 67%|██████▋   | 12785/19090 [6:14:17<2:50:48,  1.63s/it]
 67%|██████▋   | 12786/19090 [6:14:19<3:24:13,  1.94s/it]
 67%|██████▋   | 12787/19090 [6:14:21<3:04:04,  1.75s/it]
 67%|██████▋   | 12788/19090 [6:14:22<2:50:03,  1.62s/it]
 67%|██████▋   | 12789/19090 [6:14:25<3:29:01,  1.99s/it][gradnorm] step 12789: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 177.3

 67%|██████▋   | 12790/19090 [6:14:26<3:06:56,  1.78s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 257.248779296875, 'learning_rate': 1.4837013149294627e-05, 'recon_loss': 0.0016119766980409623, 'recon_static': 0.0006520171285956166, 'recon_motion': 0.000959959601459559, 'recon_norm': 1.324264895915985, 'recon_norm_std': 0.040984494518488644, 'target_norm': 1.6182830691337586, 'recon_cos': 0.6226632922887803, 'gradnorm/compressor.query': 0.14923417725628332, 'gradnorm/compressor.layer0': 5.648932513103485, 'gradnorm/compressor.layer1': 251.31541437007002, 'gradnorm/compressor.layer2': 50.522103184212845, 'gradnorm/compressor.layer3': 19.313580825338292, 'gradnorm/compressor.layer4': 5.731205950685242, 'gradnorm/compressor.layer5': 3.1460869962640134, 'gradnorm/compressor.layer6': 2.2907859381413154, 'gradnorm/compressor.layer7': 2.080797928750937, 'gradnorm/decoder.layer0': 2.3856936187110027, 'gradnorm/decoder.layer1': 1.2605517869303364, 'gradnorm/decoder.layer2': 0.6798598849171106, 'gradnorm/max_param': 177.33670043945312, 'epoch': 6.7}

 67%|██████▋   | 12790/19090 [6:14:26<3:06:56,  1.78s/it]
 67%|██████▋   | 12791/19090 [6:14:27<2:51:26,  1.63s/it]
 67%|██████▋   | 12792/19090 [6:14:30<3:27:55,  1.98s/it]
 67%|██████▋   | 12793/19090 [6:14:32<3:06:04,  1.77s/it]
 67%|██████▋   | 12794/19090 [6:14:33<2:50:41,  1.63s/it]
 67%|██████▋   | 12795/19090 [6:14:36<3:24:31,  1.95s/it]
 67%|██████▋   | 12796/19090 [6:14:37<3:03:42,  1.75s/it]
 67%|██████▋   | 12797/19090 [6:14:38<2:52:34,  1.65s/it]
 67%|██████▋   | 12798/19090 [6:14:41<3:23:38,  1.94s/it]
 67%|██████▋   | 12799/19090 [6:14:42<3:03:14,  1.75s/it][gradnorm] step 12799: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 181.3

 67%|██████▋   | 12800/19090 [6:14:43<2:49:37,  1.62s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 262.95233154296875, 'learning_rate': 1.479526467511793e-05, 'recon_loss': 0.0017490991653176025, 'recon_static': 0.0007864052007789723, 'recon_motion': 0.0009626939558074809, 'recon_norm': 1.3801551580429077, 'recon_norm_std': 0.04018948096781969, 'target_norm': 1.6196418941020965, 'recon_cos': 0.5972913280129433, 'gradnorm/compressor.query': 0.25319754655878735, 'gradnorm/compressor.layer0': 9.426797823859069, 'gradnorm/compressor.layer1': 257.0214150853971, 'gradnorm/compressor.layer2': 51.46730322040635, 'gradnorm/compressor.layer3': 17.170659363249566, 'gradnorm/compressor.layer4': 5.114703371875576, 'gradnorm/compressor.layer5': 3.1895338563343585, 'gradnorm/compressor.layer6': 1.80429270996353, 'gradnorm/compressor.layer7': 2.0005299342937537, 'gradnorm/decoder.layer0': 2.2265956943112264, 'gradnorm/decoder.layer1': 1.4937887338380729, 'gradnorm/decoder.layer2': 0.7052079395610512, 'gradnorm/max_param': 181.31370544433594, 'epoch': 6.71}

 67%|██████▋   | 12800/19090 [6:14:43<2:49:37,  1.62s/it]
 67%|██████▋   | 12801/19090 [6:14:46<3:25:16,  1.96s/it]
 67%|██████▋   | 12802/19090 [6:14:48<3:04:28,  1.76s/it]
 67%|██████▋   | 12803/19090 [6:14:49<2:53:20,  1.65s/it]
 67%|██████▋   | 12804/19090 [6:14:51<3:20:23,  1.91s/it]
 67%|██████▋   | 12805/19090 [6:14:53<3:00:45,  1.73s/it]
 67%|██████▋   | 12806/19090 [6:14:54<2:57:26,  1.69s/it]
 67%|██████▋   | 12807/19090 [6:14:57<3:29:27,  2.00s/it]
 67%|██████▋   | 12808/19090 [6:14:58<3:07:03,  1.79s/it]
 67%|██████▋   | 12809/19090 [6:15:00<2:51:33,  1.64s/it][gradnorm] step 12809: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 234.9

 67%|██████▋   | 12810/19090 [6:15:02<3:15:51,  1.87s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 338.6583557128906, 'learning_rate': 1.4753550320585435e-05, 'recon_loss': 0.0018601388728711754, 'recon_static': 0.0008823220880003646, 'recon_motion': 0.0009778168037883005, 'recon_norm': 1.4208087295293808, 'recon_norm_std': 0.03745107408612967, 'target_norm': 1.620112991333008, 'recon_cos': 0.5771563172340393, 'gradnorm/compressor.query': 0.21486259724570808, 'gradnorm/compressor.layer0': 7.93181929472796, 'gradnorm/compressor.layer1': 332.98301383103615, 'gradnorm/compressor.layer2': 56.81338094063977, 'gradnorm/compressor.layer3': 21.278155176032357, 'gradnorm/compressor.layer4': 6.141641236679177, 'gradnorm/compressor.layer5': 3.683463004290866, 'gradnorm/compressor.layer6': 2.0129665615831174, 'gradnorm/compressor.layer7': 2.1354630876163085, 'gradnorm/decoder.layer0': 2.4009969269948517, 'gradnorm/decoder.layer1': 1.4133009572339446, 'gradnorm/decoder.layer2': 0.7963165121280564, 'gradnorm/max_param': 234.93748474121094, 'epoch': 6.71}

 67%|██████▋   | 12810/19090 [6:15:02<3:15:51,  1.87s/it]
 67%|██████▋   | 12811/19090 [6:15:04<3:28:37,  1.99s/it]
 67%|██████▋   | 12812/19090 [6:15:06<3:06:24,  1.78s/it]
 67%|██████▋   | 12813/19090 [6:15:09<3:42:28,  2.13s/it]
 67%|██████▋   | 12814/19090 [6:15:10<3:16:01,  1.87s/it]
 67%|██████▋   | 12815/19090 [6:15:11<2:57:37,  1.70s/it]
 67%|██████▋   | 12816/19090 [6:15:12<2:44:40,  1.57s/it]
 67%|██████▋   | 12817/19090 [6:15:14<2:35:37,  1.49s/it]
 67%|██████▋   | 12818/19090 [6:15:15<2:31:18,  1.45s/it]
 67%|██████▋   | 12819/19090 [6:15:18<3:07:55,  1.80s/it][gradnorm] step 12819: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 191.1

 67%|██████▋   | 12820/19090 [6:15:19<2:52:04,  1.65s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 276.7154541015625, 'learning_rate': 1.471187022516955e-05, 'recon_loss': 0.0017131204338511452, 'recon_static': 0.000738051743246615, 'recon_motion': 0.0009750686527695507, 'recon_norm': 1.3557041972875594, 'recon_norm_std': 0.03973017819225788, 'target_norm': 1.623045802116394, 'recon_cos': 0.6035021498799324, 'gradnorm/compressor.query': 0.2128105625478914, 'gradnorm/compressor.layer0': 7.558644975197868, 'gradnorm/compressor.layer1': 270.79248570169005, 'gradnorm/compressor.layer2': 52.415214740837655, 'gradnorm/compressor.layer3': 19.371901528958283, 'gradnorm/compressor.layer4': 5.8601597781743475, 'gradnorm/compressor.layer5': 3.444727287848268, 'gradnorm/compressor.layer6': 1.8101563922485242, 'gradnorm/compressor.layer7': 2.2092323872294855, 'gradnorm/decoder.layer0': 2.4941886114584055, 'gradnorm/decoder.layer1': 1.3442543768759174, 'gradnorm/decoder.layer2': 0.813091539578516, 'gradnorm/max_param': 191.05343627929688, 'epoch': 6.72}

 67%|██████▋   | 12820/19090 [6:15:19<2:52:04,  1.65s/it]
 67%|██████▋   | 12821/19090 [6:15:21<2:49:30,  1.62s/it]
 67%|██████▋   | 12822/19090 [6:15:23<3:12:11,  1.84s/it]
 67%|██████▋   | 12823/19090 [6:15:24<2:54:53,  1.67s/it]
 67%|██████▋   | 12824/19090 [6:15:26<2:56:29,  1.69s/it]
 67%|██████▋   | 12825/19090 [6:15:28<3:20:44,  1.92s/it]
 67%|██████▋   | 12826/19090 [6:15:30<3:00:44,  1.73s/it]
 67%|██████▋   | 12827/19090 [6:15:32<3:09:22,  1.81s/it]
 67%|██████▋   | 12828/19090 [6:15:34<3:23:10,  1.95s/it]
 67%|██████▋   | 12829/19090 [6:15:35<3:02:31,  1.75s/it][gradnorm] step 12829: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 139.4

 67%|██████▋   | 12830/19090 [6:15:37<3:00:44,  1.73s/it]
                                                         
{'loss': 0.002, 'grad_norm': 202.63987731933594, 'learning_rate': 1.4670224528228138e-05, 'recon_loss': 0.001957141497405246, 'recon_static': 0.000994403836375568, 'recon_motion': 0.0009627377017750405, 'recon_norm': 1.468586590886116, 'recon_norm_std': 0.03800569484010339, 'target_norm': 1.6192308783531189, 'recon_cos': 0.5646099656820297, 'gradnorm/compressor.query': 0.18279055153901008, 'gradnorm/compressor.layer0': 5.626115815436186, 'gradnorm/compressor.layer1': 197.65893525843956, 'gradnorm/compressor.layer2': 41.932810491345855, 'gradnorm/compressor.layer3': 12.637576043176567, 'gradnorm/compressor.layer4': 4.458260342484731, 'gradnorm/compressor.layer5': 3.082379648994768, 'gradnorm/compressor.layer6': 1.7476635049013685, 'gradnorm/compressor.layer7': 2.01690751311082, 'gradnorm/decoder.layer0': 2.27934436372712, 'gradnorm/decoder.layer1': 1.3197340290557849, 'gradnorm/decoder.layer2': 0.8217358910208482, 'gradnorm/max_param': 139.4297637939453, 'epoch': 6.72}

 67%|██████▋   | 12830/19090 [6:15:37<3:00:44,  1.73s/it]
 67%|██████▋   | 12831/19090 [6:15:39<3:12:52,  1.85s/it]
 67%|██████▋   | 12832/19090 [6:15:40<2:55:38,  1.68s/it]
 67%|██████▋   | 12833/19090 [6:15:42<3:09:46,  1.82s/it]
 67%|██████▋   | 12834/19090 [6:15:44<3:11:21,  1.84s/it]
 67%|██████▋   | 12835/19090 [6:15:46<2:54:28,  1.67s/it]
 67%|██████▋   | 12836/19090 [6:15:48<3:14:23,  1.86s/it]
 67%|██████▋   | 12837/19090 [6:15:50<3:07:27,  1.80s/it]
 67%|██████▋   | 12838/19090 [6:15:51<2:51:35,  1.65s/it]
 67%|██████▋   | 12839/19090 [6:15:53<3:21:08,  1.93s/it][gradnorm] step 12839: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 177.0

 67%|██████▋   | 12840/19090 [6:15:55<3:08:04,  1.81s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 256.98834228515625, 'learning_rate': 1.462861336900404e-05, 'recon_loss': 0.0018515321804443373, 'recon_static': 0.0008841306713293307, 'recon_motion': 0.0009674015134805813, 'recon_norm': 1.4196594536304474, 'recon_norm_std': 0.03897255742922425, 'target_norm': 1.6198409616947174, 'recon_cos': 0.5792303055524826, 'gradnorm/compressor.query': 0.17545080384446454, 'gradnorm/compressor.layer0': 6.852441850907482, 'gradnorm/compressor.layer1': 250.88022384944972, 'gradnorm/compressor.layer2': 52.39889352809609, 'gradnorm/compressor.layer3': 15.715573909522492, 'gradnorm/compressor.layer4': 5.169221779488861, 'gradnorm/compressor.layer5': 3.679317123660308, 'gradnorm/compressor.layer6': 2.2733083996379597, 'gradnorm/compressor.layer7': 2.600280262440436, 'gradnorm/decoder.layer0': 2.7793832579194753, 'gradnorm/decoder.layer1': 1.3968974322879624, 'gradnorm/decoder.layer2': 0.8030138109475514, 'gradnorm/max_param': 176.99179077148438, 'epoch': 6.73}

 67%|██████▋   | 12840/19090 [6:15:55<3:08:04,  1.81s/it]
 67%|██████▋   | 12841/19090 [6:15:56<2:52:01,  1.65s/it]
 67%|██████▋   | 12842/19090 [6:15:59<3:23:14,  1.95s/it]
 67%|██████▋   | 12843/19090 [6:16:00<3:07:28,  1.80s/it]
 67%|██████▋   | 12844/19090 [6:16:02<2:51:24,  1.65s/it]
 67%|██████▋   | 12845/19090 [6:16:04<3:13:39,  1.86s/it]
 67%|██████▋   | 12846/19090 [6:16:06<3:04:47,  1.78s/it]
 67%|██████▋   | 12847/19090 [6:16:07<2:49:39,  1.63s/it]
 67%|██████▋   | 12848/19090 [6:16:09<3:14:29,  1.87s/it]
 67%|██████▋   | 12849/19090 [6:16:11<3:01:49,  1.75s/it][gradnorm] step 12849: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 184.1

 67%|██████▋   | 12850/19090 [6:16:12<2:47:24,  1.61s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 267.52490234375, 'learning_rate': 1.4587036886624603e-05, 'recon_loss': 0.0017922973638633266, 'recon_static': 0.0008311415032949299, 'recon_motion': 0.0009611558998585679, 'recon_norm': 1.403368318080902, 'recon_norm_std': 0.03827114999294281, 'target_norm': 1.618554052710533, 'recon_cos': 0.5898359373211861, 'gradnorm/compressor.query': 0.1759101289256076, 'gradnorm/compressor.layer0': 7.4130243794580535, 'gradnorm/compressor.layer1': 261.00113744962874, 'gradnorm/compressor.layer2': 54.26282430506635, 'gradnorm/compressor.layer3': 19.967487611701117, 'gradnorm/compressor.layer4': 4.874052689284341, 'gradnorm/compressor.layer5': 2.9472813185570415, 'gradnorm/compressor.layer6': 2.1357153954938184, 'gradnorm/compressor.layer7': 2.3137552102399677, 'gradnorm/decoder.layer0': 2.226351543940128, 'gradnorm/decoder.layer1': 1.3871985236836533, 'gradnorm/decoder.layer2': 0.7784631997624547, 'gradnorm/max_param': 184.14793395996094, 'epoch': 6.73}

 67%|██████▋   | 12850/19090 [6:16:12<2:47:24,  1.61s/it]
 67%|██████▋   | 12851/19090 [6:16:15<3:21:14,  1.94s/it]
 67%|██████▋   | 12852/19090 [6:16:16<3:07:14,  1.80s/it]
 67%|██████▋   | 12853/19090 [6:16:18<2:51:24,  1.65s/it]
 67%|██████▋   | 12854/19090 [6:16:20<3:24:20,  1.97s/it]
 67%|██████▋   | 12855/19090 [6:16:22<3:03:02,  1.76s/it]
 67%|██████▋   | 12856/19090 [6:16:23<2:48:13,  1.62s/it]
 67%|██████▋   | 12857/19090 [6:16:25<3:20:48,  1.93s/it]
 67%|██████▋   | 12858/19090 [6:16:27<3:03:58,  1.77s/it]
 67%|██████▋   | 12859/19090 [6:16:28<2:48:56,  1.63s/it][gradnorm] step 12859: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 164.4

 67%|██████▋   | 12860/19090 [6:16:31<3:23:57,  1.96s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 237.63453674316406, 'learning_rate': 1.4545495220101287e-05, 'recon_loss': 0.0018677576415939256, 'recon_static': 0.0008995027747005224, 'recon_motion': 0.0009682548610726371, 'recon_norm': 1.4256376206874848, 'recon_norm_std': 0.03871197300031781, 'target_norm': 1.6203839153051376, 'recon_cos': 0.5760698407888413, 'gradnorm/compressor.query': 0.17399014733600754, 'gradnorm/compressor.layer0': 6.45619110346442, 'gradnorm/compressor.layer1': 233.03818758950217, 'gradnorm/compressor.layer2': 41.90537173995791, 'gradnorm/compressor.layer3': 17.28181862197415, 'gradnorm/compressor.layer4': 6.275604291900153, 'gradnorm/compressor.layer5': 3.230829626958094, 'gradnorm/compressor.layer6': 1.9844201751132455, 'gradnorm/compressor.layer7': 2.174676090544383, 'gradnorm/decoder.layer0': 2.440756847224682, 'gradnorm/decoder.layer1': 1.3755624661060617, 'gradnorm/decoder.layer2': 0.7736550997957303, 'gradnorm/max_param': 164.43092346191406, 'epoch': 6.74}

 67%|██████▋   | 12860/19090 [6:16:31<3:23:57,  1.96s/it]
 67%|██████▋   | 12861/19090 [6:16:32<3:02:49,  1.76s/it]
 67%|██████▋   | 12862/19090 [6:16:33<2:48:13,  1.62s/it]
 67%|██████▋   | 12863/19090 [6:16:36<3:20:37,  1.93s/it]
 67%|██████▋   | 12864/19090 [6:16:37<3:00:34,  1.74s/it]
 67%|██████▋   | 12865/19090 [6:16:39<2:46:35,  1.61s/it]
 67%|██████▋   | 12866/19090 [6:16:42<3:24:24,  1.97s/it]
 67%|██████▋   | 12867/19090 [6:16:43<3:07:44,  1.81s/it]
 67%|██████▋   | 12868/19090 [6:16:44<2:51:36,  1.65s/it]
 67%|██████▋   | 12869/19090 [6:16:47<3:11:02,  1.84s/it][gradnorm] step 12869: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 179.3

 67%|██████▋   | 12870/19090 [6:16:48<3:01:41,  1.75s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 258.70550537109375, 'learning_rate': 1.450398850832908e-05, 'recon_loss': 0.001705719856545329, 'recon_static': 0.0007313302252441644, 'recon_motion': 0.000974389657494612, 'recon_norm': 1.3597340345382691, 'recon_norm_std': 0.03934201914817095, 'target_norm': 1.6206766545772553, 'recon_cos': 0.6050742849707603, 'gradnorm/compressor.query': 0.1733692305463709, 'gradnorm/compressor.layer0': 5.864914048549646, 'gradnorm/compressor.layer1': 254.01993032043765, 'gradnorm/compressor.layer2': 44.24764818741429, 'gradnorm/compressor.layer3': 18.549092906561494, 'gradnorm/compressor.layer4': 5.855614027342225, 'gradnorm/compressor.layer5': 3.7829379263012033, 'gradnorm/compressor.layer6': 2.133641232186031, 'gradnorm/compressor.layer7': 2.0423858569595748, 'gradnorm/decoder.layer0': 2.527036008420538, 'gradnorm/decoder.layer1': 1.2865822252341719, 'gradnorm/decoder.layer2': 0.8358093591321691, 'gradnorm/max_param': 179.25003051757812, 'epoch': 6.74}

 67%|██████▋   | 12870/19090 [6:16:48<3:01:41,  1.75s/it]
 67%|██████▋   | 12871/19090 [6:16:49<2:47:19,  1.61s/it]
 67%|██████▋   | 12872/19090 [6:16:52<3:24:09,  1.97s/it]
 67%|██████▋   | 12873/19090 [6:16:53<3:02:53,  1.77s/it]
 67%|██████▋   | 12874/19090 [6:16:55<2:48:13,  1.62s/it]
 67%|██████▋   | 12875/19090 [6:16:58<3:23:41,  1.97s/it]
 67%|██████▋   | 12876/19090 [6:16:59<3:02:25,  1.76s/it]
 67%|██████▋   | 12877/19090 [6:17:00<2:47:46,  1.62s/it]
 67%|██████▋   | 12878/19090 [6:17:03<3:17:07,  1.90s/it]
 67%|██████▋   | 12879/19090 [6:17:04<2:57:58,  1.72s/it][gradnorm] step 12879: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 184.6

 67%|██████▋   | 12880/19090 [6:17:05<2:44:40,  1.59s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 268.20074462890625, 'learning_rate': 1.446251689008612e-05, 'recon_loss': 0.0016499352001119404, 'recon_static': 0.0006940692270291038, 'recon_motion': 0.0009558660036418587, 'recon_norm': 1.3381815910339356, 'recon_norm_std': 0.04048047037795186, 'target_norm': 1.620950809121132, 'recon_cos': 0.6159970954060554, 'gradnorm/compressor.query': 0.18339941209365543, 'gradnorm/compressor.layer0': 6.610559235662526, 'gradnorm/compressor.layer1': 261.6046934307563, 'gradnorm/compressor.layer2': 55.44244534734872, 'gradnorm/compressor.layer3': 17.67092517893977, 'gradnorm/compressor.layer4': 6.133334969782957, 'gradnorm/compressor.layer5': 3.101421670094607, 'gradnorm/compressor.layer6': 1.8949052893691867, 'gradnorm/compressor.layer7': 2.1905811953008194, 'gradnorm/decoder.layer0': 2.5604924968005425, 'gradnorm/decoder.layer1': 1.3999585026312962, 'gradnorm/decoder.layer2': 0.7912578070739028, 'gradnorm/max_param': 184.58560180664062, 'epoch': 6.75}

 67%|██████▋   | 12880/19090 [6:17:05<2:44:40,  1.59s/it]
 67%|██████▋   | 12881/19090 [6:17:08<3:05:27,  1.79s/it]
 67%|██████▋   | 12882/19090 [6:17:09<2:49:57,  1.64s/it]
 67%|██████▋   | 12883/19090 [6:17:10<2:38:57,  1.54s/it]
 67%|██████▋   | 12884/19090 [6:17:12<3:05:21,  1.79s/it]
 67%|██████▋   | 12885/19090 [6:17:14<2:49:47,  1.64s/it]
 68%|██████▊   | 12886/19090 [6:17:15<2:38:54,  1.54s/it]
 68%|██████▊   | 12887/19090 [6:17:17<2:57:31,  1.72s/it]
 68%|██████▊   | 12888/19090 [6:17:19<2:44:39,  1.59s/it]
 68%|██████▊   | 12889/19090 [6:17:20<2:35:18,  1.50s/it][gradnorm] step 12889: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 179.3

 68%|██████▊   | 12890/19090 [6:17:22<3:03:50,  1.78s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 260.7484436035156, 'learning_rate': 1.4421080504033257e-05, 'recon_loss': 0.0016991622804198414, 'recon_static': 0.0007353100649197586, 'recon_motion': 0.000963852196582593, 'recon_norm': 1.3574594914913178, 'recon_norm_std': 0.0394072875380516, 'target_norm': 1.6202848196029662, 'recon_cos': 0.6060356706380844, 'gradnorm/compressor.query': 0.22842396422073366, 'gradnorm/compressor.layer0': 8.202517648647103, 'gradnorm/compressor.layer1': 254.12944228778372, 'gradnorm/compressor.layer2': 54.02987409685382, 'gradnorm/compressor.layer3': 19.177447043276022, 'gradnorm/compressor.layer4': 5.2355339930733855, 'gradnorm/compressor.layer5': 3.3171597221314726, 'gradnorm/compressor.layer6': 1.8061223126854191, 'gradnorm/compressor.layer7': 2.134326886239364, 'gradnorm/decoder.layer0': 2.276658650333409, 'gradnorm/decoder.layer1': 1.2982951741534132, 'gradnorm/decoder.layer2': 0.7352393319251959, 'gradnorm/max_param': 179.32472229003906, 'epoch': 6.75}

 68%|██████▊   | 12890/19090 [6:17:22<3:03:50,  1.78s/it]
 68%|██████▊   | 12891/19090 [6:17:24<2:48:37,  1.63s/it]
 68%|██████▊   | 12892/19090 [6:17:25<2:38:01,  1.53s/it]
 68%|██████▊   | 12893/19090 [6:17:27<2:55:54,  1.70s/it]
 68%|██████▊   | 12894/19090 [6:17:28<2:47:57,  1.63s/it]
 68%|██████▊   | 12895/19090 [6:17:30<2:37:22,  1.52s/it]
 68%|██████▊   | 12896/19090 [6:17:32<2:53:53,  1.68s/it]
 68%|██████▊   | 12897/19090 [6:17:33<2:52:14,  1.67s/it]
 68%|██████▊   | 12898/19090 [6:17:35<2:40:27,  1.55s/it]
 68%|██████▊   | 12899/19090 [6:17:37<3:13:28,  1.88s/it][gradnorm] step 12899: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 195.8

 68%|██████▊   | 12900/19090 [6:17:39<2:55:33,  1.70s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 283.99639892578125, 'learning_rate': 1.437967948871346e-05, 'recon_loss': 0.001808799387072213, 'recon_static': 0.0008352055330760777, 'recon_motion': 0.0009735938554513268, 'recon_norm': 1.4006174743175506, 'recon_norm_std': 0.039064882416278125, 'target_norm': 1.6210063576698304, 'recon_cos': 0.5866069674491883, 'gradnorm/compressor.query': 0.19154627213187564, 'gradnorm/compressor.layer0': 6.869873632520953, 'gradnorm/compressor.layer1': 277.5508540114766, 'gradnorm/compressor.layer2': 55.6007452386544, 'gradnorm/compressor.layer3': 20.621894741381293, 'gradnorm/compressor.layer4': 5.483498353803618, 'gradnorm/compressor.layer5': 3.166299062439873, 'gradnorm/compressor.layer6': 1.87487131312949, 'gradnorm/compressor.layer7': 2.0756220838738195, 'gradnorm/decoder.layer0': 2.308124113643093, 'gradnorm/decoder.layer1': 1.3318772212940384, 'gradnorm/decoder.layer2': 0.7274296006161277, 'gradnorm/max_param': 195.8310546875, 'epoch': 6.76}

 68%|██████▊   | 12900/19090 [6:17:39<2:55:33,  1.70s/it]
 68%|██████▊   | 12901/19090 [6:17:40<2:42:45,  1.58s/it]
 68%|██████▊   | 12902/19090 [6:17:43<3:16:55,  1.91s/it]
 68%|██████▊   | 12903/19090 [6:17:44<2:57:42,  1.72s/it]
 68%|██████▊   | 12904/19090 [6:17:45<2:44:03,  1.59s/it]
 68%|██████▊   | 12905/19090 [6:17:48<3:11:14,  1.86s/it]
 68%|██████▊   | 12906/19090 [6:17:49<2:53:42,  1.69s/it]
 68%|██████▊   | 12907/19090 [6:17:50<2:43:42,  1.59s/it]
 68%|██████▊   | 12908/19090 [6:17:53<3:14:17,  1.89s/it]
 68%|██████▊   | 12909/19090 [6:17:54<2:55:58,  1.71s/it][gradnorm] step 12909: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 169.2

 68%|██████▊   | 12910/19090 [6:17:56<2:47:38,  1.63s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 245.78961181640625, 'learning_rate': 1.4338313982551488e-05, 'recon_loss': 0.0017436454829294235, 'recon_static': 0.000775262204115279, 'recon_motion': 0.0009683833050075918, 'recon_norm': 1.3782110124826432, 'recon_norm_std': 0.038557304069399834, 'target_norm': 1.6217703998088837, 'recon_cos': 0.5991744965314865, 'gradnorm/compressor.query': 0.21221180404442908, 'gradnorm/compressor.layer0': 8.709914895834274, 'gradnorm/compressor.layer1': 239.8405720473498, 'gradnorm/compressor.layer2': 49.45077996019754, 'gradnorm/compressor.layer3': 17.59885756512781, 'gradnorm/compressor.layer4': 5.202244039652091, 'gradnorm/compressor.layer5': 3.401497561562542, 'gradnorm/compressor.layer6': 2.0911599836112216, 'gradnorm/compressor.layer7': 2.354531881534958, 'gradnorm/decoder.layer0': 2.689225374882671, 'gradnorm/decoder.layer1': 1.2657741411661905, 'gradnorm/decoder.layer2': 0.7971335346972783, 'gradnorm/max_param': 169.1839599609375, 'epoch': 6.76}

 68%|██████▊   | 12910/19090 [6:17:56<2:47:38,  1.63s/it]
 68%|██████▊   | 12911/19090 [6:17:58<3:20:53,  1.95s/it]
 68%|██████▊   | 12912/19090 [6:18:00<3:00:43,  1.76s/it]
 68%|██████▊   | 12913/19090 [6:18:01<2:46:25,  1.62s/it]
 68%|██████▊   | 12914/19090 [6:18:03<3:13:45,  1.88s/it]
 68%|██████▊   | 12915/19090 [6:18:05<3:02:53,  1.78s/it]
 68%|██████▊   | 12916/19090 [6:18:06<2:47:45,  1.63s/it]
 68%|██████▊   | 12917/19090 [6:18:09<3:12:00,  1.87s/it]
 68%|██████▊   | 12918/19090 [6:18:10<3:05:25,  1.80s/it]
 68%|██████▊   | 12919/19090 [6:18:12<2:49:42,  1.65s/it][gradnorm] step 12919: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 168.1

 68%|██████▊   | 12920/19090 [6:18:14<3:23:00,  1.97s/it]
                                                         
{'loss': 0.0021, 'grad_norm': 243.15333557128906, 'learning_rate': 1.4296984123853385e-05, 'recon_loss': 0.002057013305602595, 'recon_static': 0.00108957582997391, 'recon_motion': 0.0009674374989117496, 'recon_norm': 1.4959882706403733, 'recon_norm_std': 0.03747618924826383, 'target_norm': 1.6208239942789078, 'recon_cos': 0.5525787986814976, 'gradnorm/compressor.query': 0.2029664814905224, 'gradnorm/compressor.layer0': 8.22890016792237, 'gradnorm/compressor.layer1': 238.31631094513864, 'gradnorm/compressor.layer2': 43.88200370337979, 'gradnorm/compressor.layer3': 16.84663595696204, 'gradnorm/compressor.layer4': 5.396126325549566, 'gradnorm/compressor.layer5': 2.7643424470810207, 'gradnorm/compressor.layer6': 1.8408967262820415, 'gradnorm/compressor.layer7': 2.1858804429456, 'gradnorm/decoder.layer0': 2.183789403130843, 'gradnorm/decoder.layer1': 1.2155170108321958, 'gradnorm/decoder.layer2': 0.7030782895997127, 'gradnorm/max_param': 168.1497802734375, 'epoch': 6.77}

 68%|██████▊   | 12920/19090 [6:18:14<3:23:00,  1.97s/it]
 68%|██████▊   | 12921/19090 [6:18:16<3:01:47,  1.77s/it]
 68%|██████▊   | 12922/19090 [6:18:17<2:48:23,  1.64s/it]
 68%|██████▊   | 12923/19090 [6:18:20<3:32:05,  2.06s/it]
 68%|██████▊   | 12924/19090 [6:18:21<3:08:09,  1.83s/it]
 68%|██████▊   | 12925/19090 [6:18:23<2:51:16,  1.67s/it]
 68%|██████▊   | 12926/19090 [6:18:25<3:14:46,  1.90s/it]
 68%|██████▊   | 12927/19090 [6:18:26<2:55:56,  1.71s/it]
 68%|██████▊   | 12928/19090 [6:18:28<2:42:49,  1.59s/it]
 68%|██████▊   | 12929/19090 [6:18:30<3:07:56,  1.83s/it][gradnorm] step 12929: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 181.0

 68%|██████▊   | 12930/19090 [6:18:31<2:51:16,  1.67s/it]
                                                         
{'loss': 0.0021, 'grad_norm': 262.8875427246094, 'learning_rate': 1.4255690050805957e-05, 'recon_loss': 0.0021409984037745744, 'recon_static': 0.001166705867217388, 'recon_motion': 0.0009742926224134863, 'recon_norm': 1.5247293382883071, 'recon_norm_std': 0.03709131060168147, 'target_norm': 1.6229103058576584, 'recon_cos': 0.5391394749283791, 'gradnorm/compressor.query': 0.23346220127081904, 'gradnorm/compressor.layer0': 8.572941337787885, 'gradnorm/compressor.layer1': 256.5702876406385, 'gradnorm/compressor.layer2': 52.47269797907408, 'gradnorm/compressor.layer3': 19.93035358794783, 'gradnorm/compressor.layer4': 5.740674625321365, 'gradnorm/compressor.layer5': 2.978732381969477, 'gradnorm/compressor.layer6': 2.1480439536572016, 'gradnorm/compressor.layer7': 2.007098119088765, 'gradnorm/decoder.layer0': 2.2419566391181145, 'gradnorm/decoder.layer1': 1.2104074610208524, 'gradnorm/decoder.layer2': 0.662525282683272, 'gradnorm/max_param': 181.0157928466797, 'epoch': 6.77}

 68%|██████▊   | 12930/19090 [6:18:31<2:51:16,  1.67s/it]
 68%|██████▊   | 12931/19090 [6:18:32<2:39:37,  1.55s/it]
 68%|██████▊   | 12932/19090 [6:18:35<3:02:47,  1.78s/it]
 68%|██████▊   | 12933/19090 [6:18:36<2:47:33,  1.63s/it]
 68%|██████▊   | 12934/19090 [6:18:37<2:36:59,  1.53s/it]
 68%|██████▊   | 12935/19090 [6:18:40<3:06:52,  1.82s/it]
 68%|██████▊   | 12936/19090 [6:18:41<2:50:23,  1.66s/it]
 68%|██████▊   | 12937/19090 [6:18:42<2:38:48,  1.55s/it]
 68%|██████▊   | 12938/19090 [6:18:45<3:06:19,  1.82s/it]
 68%|██████▊   | 12939/19090 [6:18:46<2:50:06,  1.66s/it][gradnorm] step 12939: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 196.1

 68%|██████▊   | 12940/19090 [6:18:47<2:38:43,  1.55s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 282.8659362792969, 'learning_rate': 1.4214431901476388e-05, 'recon_loss': 0.0018378857377683744, 'recon_static': 0.0008619148604338989, 'recon_motion': 0.0009759709020727314, 'recon_norm': 1.411675125360489, 'recon_norm_std': 0.037727849930524825, 'target_norm': 1.6223869383335114, 'recon_cos': 0.5815699890255928, 'gradnorm/compressor.query': 0.1883364141544205, 'gradnorm/compressor.layer0': 6.753027060600392, 'gradnorm/compressor.layer1': 277.9778121766196, 'gradnorm/compressor.layer2': 48.24769009610071, 'gradnorm/compressor.layer3': 17.573758138134718, 'gradnorm/compressor.layer4': 5.88379287149773, 'gradnorm/compressor.layer5': 3.2545612312542835, 'gradnorm/compressor.layer6': 1.8453013472256379, 'gradnorm/compressor.layer7': 1.9183295817577402, 'gradnorm/decoder.layer0': 2.193412420050911, 'gradnorm/decoder.layer1': 1.274460551265345, 'gradnorm/decoder.layer2': 0.6514429051258164, 'gradnorm/max_param': 196.11956787109375, 'epoch': 6.78}

 68%|██████▊   | 12940/19090 [6:18:47<2:38:43,  1.55s/it]
 68%|██████▊   | 12941/19090 [6:18:50<3:08:04,  1.84s/it]
 68%|██████▊   | 12942/19090 [6:18:51<2:51:12,  1.67s/it]
 68%|██████▊   | 12943/19090 [6:18:53<2:39:40,  1.56s/it]
 68%|██████▊   | 12944/19090 [6:18:55<3:01:01,  1.77s/it]
 68%|██████▊   | 12945/19090 [6:18:56<2:46:34,  1.63s/it]
 68%|██████▊   | 12946/19090 [6:18:57<2:36:17,  1.53s/it]
 68%|██████▊   | 12947/19090 [6:19:00<2:58:45,  1.75s/it]
 68%|██████▊   | 12948/19090 [6:19:01<2:44:56,  1.61s/it]
 68%|██████▊   | 12949/19090 [6:19:02<2:35:15,  1.52s/it][gradnorm] step 12949: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 206.0

 68%|██████▊   | 12950/19090 [6:19:05<3:05:40,  1.81s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 299.390869140625, 'learning_rate': 1.4173209813811788e-05, 'recon_loss': 0.0017903090658364818, 'recon_static': 0.0008308378543006256, 'recon_motion': 0.0009594712129910477, 'recon_norm': 1.3967076599597932, 'recon_norm_std': 0.03900098977610469, 'target_norm': 1.6199443012475967, 'recon_cos': 0.589122374355793, 'gradnorm/compressor.query': 0.20138446907604912, 'gradnorm/compressor.layer0': 7.1367091286880955, 'gradnorm/compressor.layer1': 291.92116872539407, 'gradnorm/compressor.layer2': 62.35982131511543, 'gradnorm/compressor.layer3': 20.403071063883758, 'gradnorm/compressor.layer4': 5.758960542135959, 'gradnorm/compressor.layer5': 3.3736528604380296, 'gradnorm/compressor.layer6': 1.9129869202736494, 'gradnorm/compressor.layer7': 1.968706463529194, 'gradnorm/decoder.layer0': 2.5372612300590074, 'gradnorm/decoder.layer1': 1.3374468213499051, 'gradnorm/decoder.layer2': 0.7749613767659479, 'gradnorm/max_param': 205.97779846191406, 'epoch': 6.78}

 68%|██████▊   | 12950/19090 [6:19:05<3:05:40,  1.81s/it]
 68%|██████▊   | 12951/19090 [6:19:06<2:49:29,  1.66s/it]
 68%|██████▊   | 12952/19090 [6:19:07<2:42:30,  1.59s/it]
 68%|██████▊   | 12953/19090 [6:19:10<3:12:11,  1.88s/it]
 68%|██████▊   | 12954/19090 [6:19:11<2:54:01,  1.70s/it]
 68%|██████▊   | 12955/19090 [6:19:13<2:41:29,  1.58s/it]
 68%|██████▊   | 12956/19090 [6:19:16<3:26:21,  2.02s/it]
 68%|██████▊   | 12957/19090 [6:19:17<3:04:01,  1.80s/it]
 68%|██████▊   | 12958/19090 [6:19:18<2:48:28,  1.65s/it]
 68%|██████▊   | 12959/19090 [6:19:21<3:16:56,  1.93s/it][gradnorm] step 12959: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 150.7

 68%|██████▊   | 12960/19090 [6:19:22<2:57:32,  1.74s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 219.65972900390625, 'learning_rate': 1.4132023925638624e-05, 'recon_loss': 0.0016851840307936072, 'recon_static': 0.0007065804908052087, 'recon_motion': 0.0009786035632714629, 'recon_norm': 1.3468136221170426, 'recon_norm_std': 0.03941009445115924, 'target_norm': 1.6224718421697617, 'recon_cos': 0.609197610616684, 'gradnorm/compressor.query': 0.1783518492773104, 'gradnorm/compressor.layer0': 5.659779158957868, 'gradnorm/compressor.layer1': 213.5914799623337, 'gradnorm/compressor.layer2': 47.744145378975524, 'gradnorm/compressor.layer3': 16.06595654534136, 'gradnorm/compressor.layer4': 5.701768018677923, 'gradnorm/compressor.layer5': 3.3755409372362113, 'gradnorm/compressor.layer6': 1.891958412773889, 'gradnorm/compressor.layer7': 2.057776504236536, 'gradnorm/decoder.layer0': 2.313634748674617, 'gradnorm/decoder.layer1': 1.3350732249633723, 'gradnorm/decoder.layer2': 0.7309641254822321, 'gradnorm/max_param': 150.68528747558594, 'epoch': 6.79}

 68%|██████▊   | 12960/19090 [6:19:22<2:57:32,  1.74s/it]
 68%|██████▊   | 12961/19090 [6:19:23<2:43:49,  1.60s/it]
 68%|██████▊   | 12962/19090 [6:19:26<3:23:10,  1.99s/it]
 68%|██████▊   | 12963/19090 [6:19:28<3:01:46,  1.78s/it]
 68%|██████▊   | 12964/19090 [6:19:29<2:46:35,  1.63s/it]
 68%|██████▊   | 12965/19090 [6:19:32<3:24:10,  2.00s/it]
 68%|██████▊   | 12966/19090 [6:19:33<3:02:21,  1.79s/it]
 68%|██████▊   | 12967/19090 [6:19:34<2:47:11,  1.64s/it]
 68%|██████▊   | 12968/19090 [6:19:37<3:11:36,  1.88s/it]
 68%|██████▊   | 12969/19090 [6:19:38<2:53:31,  1.70s/it][gradnorm] step 12969: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 220.5

 68%|██████▊   | 12970/19090 [6:19:39<2:41:12,  1.58s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 317.7919006347656, 'learning_rate': 1.4090874374662364e-05, 'recon_loss': 0.0018097012623911723, 'recon_static': 0.0008402389794355258, 'recon_motion': 0.0009694623557152226, 'recon_norm': 1.398985466361046, 'recon_norm_std': 0.03970595700666309, 'target_norm': 1.6216510355472564, 'recon_cos': 0.5862397864460945, 'gradnorm/compressor.query': 0.1943358608226529, 'gradnorm/compressor.layer0': 7.495570336973276, 'gradnorm/compressor.layer1': 312.4234781310457, 'gradnorm/compressor.layer2': 55.591815505269075, 'gradnorm/compressor.layer3': 13.433950407812674, 'gradnorm/compressor.layer4': 5.085752984844188, 'gradnorm/compressor.layer5': 3.7109715510362107, 'gradnorm/compressor.layer6': 1.8627918168410593, 'gradnorm/compressor.layer7': 2.260076107499089, 'gradnorm/decoder.layer0': 2.292731251675003, 'gradnorm/decoder.layer1': 1.3741792049649468, 'gradnorm/decoder.layer2': 0.8727041506278117, 'gradnorm/max_param': 220.46533203125, 'epoch': 6.79}

 68%|██████▊   | 12970/19090 [6:19:39<2:41:12,  1.58s/it]
 68%|██████▊   | 12971/19090 [6:19:42<3:16:27,  1.93s/it]
 68%|██████▊   | 12972/19090 [6:19:43<2:57:17,  1.74s/it]
 68%|██████▊   | 12973/19090 [6:19:45<2:43:57,  1.61s/it]
 68%|██████▊   | 12974/19090 [6:19:48<3:24:51,  2.01s/it]
 68%|██████▊   | 12975/19090 [6:19:49<3:02:44,  1.79s/it]
 68%|██████▊   | 12976/19090 [6:19:50<2:47:37,  1.65s/it]
 68%|██████▊   | 12977/19090 [6:19:53<3:12:55,  1.89s/it]
 68%|██████▊   | 12978/19090 [6:19:54<2:54:20,  1.71s/it]
 68%|██████▊   | 12979/19090 [6:19:55<2:41:17,  1.58s/it][gradnorm] step 12979: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 189.0

 68%|██████▊   | 12980/19090 [6:19:58<3:23:28,  2.00s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 274.13470458984375, 'learning_rate': 1.404976129846701e-05, 'recon_loss': 0.0016720542131224646, 'recon_static': 0.0006849399855127558, 'recon_motion': 0.0009871142377960497, 'recon_norm': 1.3405587166547774, 'recon_norm_std': 0.03992539141327143, 'target_norm': 1.6240436732769012, 'recon_cos': 0.612284243106842, 'gradnorm/compressor.query': 0.1598666177353686, 'gradnorm/compressor.layer0': 6.278157398070027, 'gradnorm/compressor.layer1': 267.83915415786396, 'gradnorm/compressor.layer2': 55.03388737646674, 'gradnorm/compressor.layer3': 17.225531431728566, 'gradnorm/compressor.layer4': 4.820754663105205, 'gradnorm/compressor.layer5': 3.019701322476336, 'gradnorm/compressor.layer6': 1.8494254225199607, 'gradnorm/compressor.layer7': 2.092037696403927, 'gradnorm/decoder.layer0': 2.1381055944432448, 'gradnorm/decoder.layer1': 1.3827049084202787, 'gradnorm/decoder.layer2': 0.6883445883801358, 'gradnorm/max_param': 189.00723266601562, 'epoch': 6.8}

 68%|██████▊   | 12980/19090 [6:19:58<3:23:28,  2.00s/it]
 68%|██████▊   | 12981/19090 [6:20:00<3:01:40,  1.78s/it]
 68%|██████▊   | 12982/19090 [6:20:01<2:46:27,  1.64s/it]
 68%|██████▊   | 12983/19090 [6:20:04<3:20:40,  1.97s/it]
 68%|██████▊   | 12984/19090 [6:20:05<2:59:39,  1.77s/it]
 68%|██████▊   | 12985/19090 [6:20:06<2:44:57,  1.62s/it]
 68%|██████▊   | 12986/19090 [6:20:09<3:20:03,  1.97s/it]
 68%|██████▊   | 12987/19090 [6:20:10<2:59:04,  1.76s/it]
 68%|██████▊   | 12988/19090 [6:20:11<2:44:32,  1.62s/it]
 68%|██████▊   | 12989/19090 [6:20:14<3:16:39,  1.93s/it][gradnorm] step 12989: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 202.6

 68%|██████▊   | 12990/19090 [6:20:15<2:56:55,  1.74s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 291.6734619140625, 'learning_rate': 1.4008684834514554e-05, 'recon_loss': 0.0016438017541076989, 'recon_static': 0.0006737892064847984, 'recon_motion': 0.000970012546167709, 'recon_norm': 1.333427858352661, 'recon_norm_std': 0.04070243882015347, 'target_norm': 1.6208861589431762, 'recon_cos': 0.6168301150202751, 'gradnorm/compressor.query': 0.18353106057449403, 'gradnorm/compressor.layer0': 6.409004028833307, 'gradnorm/compressor.layer1': 287.16427408802093, 'gradnorm/compressor.layer2': 47.91505800062362, 'gradnorm/compressor.layer3': 15.034575223412117, 'gradnorm/compressor.layer4': 4.793329276630098, 'gradnorm/compressor.layer5': 3.16568215441911, 'gradnorm/compressor.layer6': 1.8366974252154906, 'gradnorm/compressor.layer7': 2.088987485213568, 'gradnorm/decoder.layer0': 2.089190058212684, 'gradnorm/decoder.layer1': 1.2228357347854162, 'gradnorm/decoder.layer2': 0.7012127197610186, 'gradnorm/max_param': 202.60858154296875, 'epoch': 6.8}

 68%|██████▊   | 12990/19090 [6:20:15<2:56:55,  1.74s/it]
 68%|██████▊   | 12991/19090 [6:20:17<2:43:03,  1.60s/it]
 68%|██████▊   | 12992/19090 [6:20:19<3:06:30,  1.84s/it]
 68%|██████▊   | 12993/19090 [6:20:20<2:49:42,  1.67s/it]
 68%|██████▊   | 12994/19090 [6:20:22<2:37:58,  1.55s/it]
 68%|██████▊   | 12995/19090 [6:20:25<3:20:11,  1.97s/it]
 68%|██████▊   | 12996/19090 [6:20:26<2:59:18,  1.77s/it]
 68%|██████▊   | 12997/19090 [6:20:27<2:44:33,  1.62s/it]
 68%|██████▊   | 12998/19090 [6:20:30<3:24:40,  2.02s/it]
 68%|██████▊   | 12999/19090 [6:20:31<3:02:24,  1.80s/it][gradnorm] step 12999: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 212.0

 68%|██████▊   | 13000/19090 [6:20:33<2:46:55,  1.64s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 307.49444580078125, 'learning_rate': 1.3967645120144605e-05, 'recon_loss': 0.0016038230067351833, 'recon_static': 0.0006424551087548025, 'recon_motion': 0.0009613679343601689, 'recon_norm': 1.3193100333213805, 'recon_norm_std': 0.04027644069865346, 'target_norm': 1.6182851642370224, 'recon_cos': 0.623992420732975, 'gradnorm/compressor.query': 0.1786463261107237, 'gradnorm/compressor.layer0': 6.798052385452142, 'gradnorm/compressor.layer1': 300.42131873420703, 'gradnorm/compressor.layer2': 62.08080514287408, 'gradnorm/compressor.layer3': 18.34620787808367, 'gradnorm/compressor.layer4': 5.585472076873467, 'gradnorm/compressor.layer5': 3.3887869779622966, 'gradnorm/compressor.layer6': 2.2114901980744115, 'gradnorm/compressor.layer7': 2.447634828183369, 'gradnorm/decoder.layer0': 2.5988790516555937, 'gradnorm/decoder.layer1': 1.3912701342583285, 'gradnorm/decoder.layer2': 0.8649574336013832, 'gradnorm/max_param': 211.9502410888672, 'epoch': 6.81}

 68%|██████▊   | 13000/19090 [6:20:33<2:46:55,  1.64s/it]
 68%|██████▊   | 13001/19090 [6:20:36<3:45:14,  2.22s/it]
 68%|██████▊   | 13002/19090 [6:20:38<3:16:42,  1.94s/it]
 68%|██████▊   | 13003/19090 [6:20:39<2:56:46,  1.74s/it]
 68%|██████▊   | 13004/19090 [6:20:40<2:42:56,  1.61s/it]
 68%|██████▊   | 13005/19090 [6:20:41<2:33:15,  1.51s/it]
 68%|██████▊   | 13006/19090 [6:20:43<2:26:28,  1.44s/it]
 68%|██████▊   | 13007/19090 [6:20:44<2:34:52,  1.53s/it]
 68%|██████▊   | 13008/19090 [6:20:46<2:27:25,  1.45s/it]
 68%|██████▊   | 13009/19090 [6:20:47<2:22:11,  1.40s/it][gradnorm] step 13009: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 191.8

 68%|██████▊   | 13010/19090 [6:20:50<3:01:57,  1.80s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 280.7192687988281, 'learning_rate': 1.3926642292573921e-05, 'recon_loss': 0.0016555447014980017, 'recon_static': 0.0006906740032718517, 'recon_motion': 0.0009648707477026619, 'recon_norm': 1.341590267419815, 'recon_norm_std': 0.04041572855785489, 'target_norm': 1.6196576744318008, 'recon_cos': 0.6143052145838738, 'gradnorm/compressor.query': 0.1885342188588266, 'gradnorm/compressor.layer0': 7.105332652713234, 'gradnorm/compressor.layer1': 271.78910999615124, 'gradnorm/compressor.layer2': 67.4483815942792, 'gradnorm/compressor.layer3': 16.325440867898482, 'gradnorm/compressor.layer4': 6.0159889804658, 'gradnorm/compressor.layer5': 3.2270686235250623, 'gradnorm/compressor.layer6': 2.474847241706343, 'gradnorm/compressor.layer7': 2.641171099833362, 'gradnorm/decoder.layer0': 2.3317589784590234, 'gradnorm/decoder.layer1': 1.3866874583837447, 'gradnorm/decoder.layer2': 0.7753343922093127, 'gradnorm/max_param': 191.79171752929688, 'epoch': 6.82}

 68%|██████▊   | 13010/19090 [6:20:50<3:01:57,  1.80s/it]
 68%|██████▊   | 13011/19090 [6:20:51<2:46:24,  1.64s/it]
 68%|██████▊   | 13012/19090 [6:20:52<2:35:23,  1.53s/it]
 68%|██████▊   | 13013/19090 [6:20:55<3:00:37,  1.78s/it]
 68%|██████▊   | 13014/19090 [6:20:56<2:45:25,  1.63s/it]
 68%|██████▊   | 13015/19090 [6:20:57<2:39:22,  1.57s/it]
 68%|██████▊   | 13016/19090 [6:21:00<3:13:01,  1.91s/it]
 68%|██████▊   | 13017/19090 [6:21:01<2:54:04,  1.72s/it]
 68%|██████▊   | 13018/19090 [6:21:03<2:48:23,  1.66s/it]
 68%|██████▊   | 13019/19090 [6:21:05<3:16:53,  1.95s/it][gradnorm] step 13019: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 236.5

 68%|██████▊   | 13020/19090 [6:21:07<2:56:53,  1.75s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 343.6374816894531, 'learning_rate': 1.3885676488895883e-05, 'recon_loss': 0.0016605609038379044, 'recon_static': 0.0006905447226017714, 'recon_motion': 0.0009700162278022617, 'recon_norm': 1.3433739960193634, 'recon_norm_std': 0.04016793761402369, 'target_norm': 1.6201731503009795, 'recon_cos': 0.6139090403914451, 'gradnorm/compressor.query': 0.2740617710308267, 'gradnorm/compressor.layer0': 9.254619785043719, 'gradnorm/compressor.layer1': 335.2174201171532, 'gradnorm/compressor.layer2': 70.44198147345267, 'gradnorm/compressor.layer3': 24.306961330874433, 'gradnorm/compressor.layer4': 6.733887267539723, 'gradnorm/compressor.layer5': 3.6377499733728835, 'gradnorm/compressor.layer6': 2.3164054265913374, 'gradnorm/compressor.layer7': 2.118799925809502, 'gradnorm/decoder.layer0': 2.536168911588483, 'gradnorm/decoder.layer1': 1.3370334970362412, 'gradnorm/decoder.layer2': 0.8516965987985565, 'gradnorm/max_param': 236.4700927734375, 'epoch': 6.82}

 68%|██████▊   | 13020/19090 [6:21:07<2:56:53,  1.75s/it]
 68%|██████▊   | 13021/19090 [6:21:08<2:52:34,  1.71s/it]
 68%|██████▊   | 13022/19090 [6:21:11<3:15:36,  1.93s/it]
 68%|██████▊   | 13023/19090 [6:21:12<2:55:48,  1.74s/it]
 68%|██████▊   | 13024/19090 [6:21:13<2:42:02,  1.60s/it]
 68%|██████▊   | 13025/19090 [6:21:16<3:12:28,  1.90s/it]
 68%|██████▊   | 13026/19090 [6:21:17<2:53:32,  1.72s/it]
 68%|██████▊   | 13027/19090 [6:21:19<2:50:44,  1.69s/it]
 68%|██████▊   | 13028/19090 [6:21:22<3:29:06,  2.07s/it]
 68%|██████▊   | 13029/19090 [6:21:23<3:05:15,  1.83s/it][gradnorm] step 13029: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 256.6

 68%|██████▊   | 13030/19090 [6:21:24<2:48:47,  1.67s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 371.5524597167969, 'learning_rate': 1.3844747846080094e-05, 'recon_loss': 0.0016552005137782544, 'recon_static': 0.0006813792497268878, 'recon_motion': 0.0009738212829688564, 'recon_norm': 1.3313746482133866, 'recon_norm_std': 0.04073912128806114, 'target_norm': 1.6227861434221267, 'recon_cos': 0.6147179737687111, 'gradnorm/compressor.query': 0.26979178124257214, 'gradnorm/compressor.layer0': 9.404228119881063, 'gradnorm/compressor.layer1': 363.7271685274555, 'gradnorm/compressor.layer2': 71.36797387703045, 'gradnorm/compressor.layer3': 22.10734517515297, 'gradnorm/compressor.layer4': 7.263100724973686, 'gradnorm/compressor.layer5': 3.5517216980309243, 'gradnorm/compressor.layer6': 2.127764866587502, 'gradnorm/compressor.layer7': 2.3139597435287484, 'gradnorm/decoder.layer0': 2.351818691758578, 'gradnorm/decoder.layer1': 1.305234965514181, 'gradnorm/decoder.layer2': 0.8102233240949879, 'gradnorm/max_param': 256.6167297363281, 'epoch': 6.83}

 68%|██████▊   | 13030/19090 [6:21:24<2:48:47,  1.67s/it]
 68%|██████▊   | 13031/19090 [6:21:27<3:28:28,  2.06s/it]
 68%|██████▊   | 13032/19090 [6:21:29<3:04:57,  1.83s/it]
 68%|██████▊   | 13033/19090 [6:21:30<2:48:18,  1.67s/it]
 68%|██████▊   | 13034/19090 [6:21:33<3:21:52,  2.00s/it]
 68%|██████▊   | 13035/19090 [6:21:34<3:00:19,  1.79s/it]
 68%|██████▊   | 13036/19090 [6:21:35<2:45:02,  1.64s/it]
 68%|██████▊   | 13037/19090 [6:21:38<3:26:36,  2.05s/it]
 68%|██████▊   | 13038/19090 [6:21:40<3:03:34,  1.82s/it]
 68%|██████▊   | 13039/19090 [6:21:41<2:47:33,  1.66s/it][gradnorm] step 13039: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 238.9

 68%|██████▊   | 13040/19090 [6:21:43<3:15:50,  1.94s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 346.19573974609375, 'learning_rate': 1.3803856500971956e-05, 'recon_loss': 0.0016988708666758612, 'recon_static': 0.0007363675234955735, 'recon_motion': 0.000962503338814713, 'recon_norm': 1.3516359686851502, 'recon_norm_std': 0.03991960436105728, 'target_norm': 1.6190856009721757, 'recon_cos': 0.6042770341038703, 'gradnorm/compressor.query': 0.20509705229109856, 'gradnorm/compressor.layer0': 7.637726466041341, 'gradnorm/compressor.layer1': 338.4985921314888, 'gradnorm/compressor.layer2': 67.90458177886453, 'gradnorm/compressor.layer3': 23.107310854822686, 'gradnorm/compressor.layer4': 5.720249333849552, 'gradnorm/compressor.layer5': 3.8634599696401306, 'gradnorm/compressor.layer6': 2.2622998270244237, 'gradnorm/compressor.layer7': 2.295197685387589, 'gradnorm/decoder.layer0': 2.5352013911620515, 'gradnorm/decoder.layer1': 1.3325265986953014, 'gradnorm/decoder.layer2': 0.776836785859666, 'gradnorm/max_param': 238.8661346435547, 'epoch': 6.83}

 68%|██████▊   | 13040/19090 [6:21:44<3:15:50,  1.94s/it]
 68%|██████▊   | 13041/19090 [6:21:45<2:56:02,  1.75s/it]
 68%|██████▊   | 13042/19090 [6:21:46<2:42:17,  1.61s/it]
 68%|██████▊   | 13043/19090 [6:21:49<3:14:56,  1.93s/it]
 68%|██████▊   | 13044/19090 [6:21:50<2:55:18,  1.74s/it]
 68%|██████▊   | 13045/19090 [6:21:51<2:41:56,  1.61s/it]
 68%|██████▊   | 13046/19090 [6:21:55<3:29:41,  2.08s/it]
 68%|██████▊   | 13047/19090 [6:21:56<3:05:33,  1.84s/it]
 68%|██████▊   | 13048/19090 [6:21:57<2:48:44,  1.68s/it]
 68%|██████▊   | 13049/19090 [6:22:00<3:21:57,  2.01s/it][gradnorm] step 13049: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 202.0

 68%|██████▊   | 13050/19090 [6:22:01<3:00:15,  1.79s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 293.85015869140625, 'learning_rate': 1.3763002590292096e-05, 'recon_loss': 0.0017626276647206395, 'recon_static': 0.0007946575977257452, 'recon_motion': 0.0009679700742708519, 'recon_norm': 1.386034232378006, 'recon_norm_std': 0.03917866870760918, 'target_norm': 1.6194152861833573, 'recon_cos': 0.5967827126383781, 'gradnorm/compressor.query': 0.2229854007991654, 'gradnorm/compressor.layer0': 7.202055415183164, 'gradnorm/compressor.layer1': 286.28539628314957, 'gradnorm/compressor.layer2': 62.405800104787936, 'gradnorm/compressor.layer3': 19.14780477585812, 'gradnorm/compressor.layer4': 6.340054643076609, 'gradnorm/compressor.layer5': 3.753749625821751, 'gradnorm/compressor.layer6': 2.166198777761494, 'gradnorm/compressor.layer7': 2.3645171661754083, 'gradnorm/decoder.layer0': 2.7806913693499227, 'gradnorm/decoder.layer1': 1.5185429121830432, 'gradnorm/decoder.layer2': 0.9671927210061398, 'gradnorm/max_param': 201.964599609375, 'epoch': 6.84}

 68%|██████▊   | 13050/19090 [6:22:01<3:00:15,  1.79s/it]
 68%|██████▊   | 13051/19090 [6:22:02<2:45:00,  1.64s/it]
 68%|██████▊   | 13052/19090 [6:22:05<3:21:16,  2.00s/it]
 68%|██████▊   | 13053/19090 [6:22:07<2:59:38,  1.79s/it]
 68%|██████▊   | 13054/19090 [6:22:08<2:44:31,  1.64s/it]
 68%|██████▊   | 13055/19090 [6:22:11<3:25:38,  2.04s/it]
 68%|██████▊   | 13056/19090 [6:22:12<3:02:49,  1.82s/it]
 68%|██████▊   | 13057/19090 [6:22:13<2:46:43,  1.66s/it]
 68%|██████▊   | 13058/19090 [6:22:16<3:25:37,  2.05s/it]
 68%|██████▊   | 13059/19090 [6:22:18<3:02:37,  1.82s/it][gradnorm] step 13059: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 242.8

 68%|██████▊   | 13060/19090 [6:22:19<2:46:39,  1.66s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 353.8168029785156, 'learning_rate': 1.3722186250636016e-05, 'recon_loss': 0.0019131654378725215, 'recon_static': 0.0009417364039109088, 'recon_motion': 0.000971429020864889, 'recon_norm': 1.446465927362442, 'recon_norm_std': 0.03730927370488644, 'target_norm': 1.6206314891576767, 'recon_cos': 0.5713307201862335, 'gradnorm/compressor.query': 0.23524196336327666, 'gradnorm/compressor.layer0': 8.96661032430917, 'gradnorm/compressor.layer1': 344.0181567701042, 'gradnorm/compressor.layer2': 79.6382077939038, 'gradnorm/compressor.layer3': 18.54962184140109, 'gradnorm/compressor.layer4': 6.092229252349553, 'gradnorm/compressor.layer5': 3.8041793828600587, 'gradnorm/compressor.layer6': 2.384853901896788, 'gradnorm/compressor.layer7': 2.1665484200513503, 'gradnorm/decoder.layer0': 2.5519074878965005, 'gradnorm/decoder.layer1': 1.3879160695383435, 'gradnorm/decoder.layer2': 0.809806982402136, 'gradnorm/max_param': 242.76011657714844, 'epoch': 6.84}

 68%|██████▊   | 13060/19090 [6:22:19<2:46:39,  1.66s/it]
 68%|██████▊   | 13061/19090 [6:22:21<3:11:51,  1.91s/it]
 68%|██████▊   | 13062/19090 [6:22:23<2:53:03,  1.72s/it]
 68%|██████▊   | 13063/19090 [6:22:24<2:40:08,  1.59s/it]
 68%|██████▊   | 13064/19090 [6:22:27<3:09:08,  1.88s/it]
 68%|██████▊   | 13065/19090 [6:22:28<2:51:27,  1.71s/it]
 68%|██████▊   | 13066/19090 [6:22:29<2:38:48,  1.58s/it]
 68%|██████▊   | 13067/19090 [6:22:32<3:17:22,  1.97s/it]
 68%|██████▊   | 13068/19090 [6:22:33<2:57:02,  1.76s/it]
 68%|██████▊   | 13069/19090 [6:22:35<2:42:40,  1.62s/it][gradnorm] step 13069: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 318.6

 68%|██████▊   | 13070/19090 [6:22:38<3:20:39,  2.00s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 459.5290832519531, 'learning_rate': 1.3681407618473615e-05, 'recon_loss': 0.0018039965914795175, 'recon_static': 0.0008371198520762846, 'recon_motion': 0.0009668767175753601, 'recon_norm': 1.403038340806961, 'recon_norm_std': 0.038010663632303474, 'target_norm': 1.6193968027830123, 'recon_cos': 0.587307807803154, 'gradnorm/compressor.query': 0.28016606045091375, 'gradnorm/compressor.layer0': 10.688796293827359, 'gradnorm/compressor.layer1': 451.5811804094586, 'gradnorm/compressor.layer2': 79.1051237624261, 'gradnorm/compressor.layer3': 28.382111452256, 'gradnorm/compressor.layer4': 5.954711382299261, 'gradnorm/compressor.layer5': 3.4339654957765786, 'gradnorm/compressor.layer6': 1.8378027348679749, 'gradnorm/compressor.layer7': 2.2106641189267546, 'gradnorm/decoder.layer0': 2.400084386772505, 'gradnorm/decoder.layer1': 1.413054723599191, 'gradnorm/decoder.layer2': 0.7963601861646813, 'gradnorm/max_param': 318.6478576660156, 'epoch': 6.85}

 68%|██████▊   | 13070/19090 [6:22:38<3:20:39,  2.00s/it]
 68%|██████▊   | 13071/19090 [6:22:39<2:59:16,  1.79s/it]
 68%|██████▊   | 13072/19090 [6:22:40<2:44:33,  1.64s/it]
 68%|██████▊   | 13073/19090 [6:22:43<3:19:58,  1.99s/it]
 68%|██████▊   | 13074/19090 [6:22:44<2:58:49,  1.78s/it]
 68%|██████▊   | 13075/19090 [6:22:46<2:44:02,  1.64s/it]
 68%|██████▊   | 13076/19090 [6:22:48<3:10:10,  1.90s/it]
 69%|██████▊   | 13077/19090 [6:22:49<2:51:56,  1.72s/it]
 69%|██████▊   | 13078/19090 [6:22:51<2:39:08,  1.59s/it]
 69%|██████▊   | 13079/19090 [6:22:53<3:08:03,  1.88s/it][gradnorm] step 13079: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 255.0

 69%|██████▊   | 13080/19090 [6:22:54<2:50:13,  1.70s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 368.24652099609375, 'learning_rate': 1.3640666830148674e-05, 'recon_loss': 0.0017679861921351403, 'recon_static': 0.0007886889681685716, 'recon_motion': 0.0009792972399736755, 'recon_norm': 1.3807148456573486, 'recon_norm_std': 0.03899400364607573, 'target_norm': 1.6215966194868088, 'recon_cos': 0.5939837485551834, 'gradnorm/compressor.query': 0.20644198534981248, 'gradnorm/compressor.layer0': 7.993072609911313, 'gradnorm/compressor.layer1': 361.43507369927454, 'gradnorm/compressor.layer2': 66.90299501657232, 'gradnorm/compressor.layer3': 19.072872754193487, 'gradnorm/compressor.layer4': 5.926262108662289, 'gradnorm/compressor.layer5': 3.508266903813904, 'gradnorm/compressor.layer6': 2.1579406302472446, 'gradnorm/compressor.layer7': 2.353879680413959, 'gradnorm/decoder.layer0': 2.4339412239080898, 'gradnorm/decoder.layer1': 1.459223838672288, 'gradnorm/decoder.layer2': 0.8829524849153173, 'gradnorm/max_param': 255.0482635498047, 'epoch': 6.85}

 69%|██████▊   | 13080/19090 [6:22:54<2:50:13,  1.70s/it]
 69%|██████▊   | 13081/19090 [6:22:56<2:37:51,  1.58s/it]
 69%|██████▊   | 13082/19090 [6:22:59<3:24:30,  2.04s/it]
 69%|██████▊   | 13083/19090 [6:23:00<3:01:44,  1.82s/it]
 69%|██████▊   | 13084/19090 [6:23:01<2:45:57,  1.66s/it]
 69%|██████▊   | 13085/19090 [6:23:04<3:23:05,  2.03s/it]
 69%|██████▊   | 13086/19090 [6:23:06<3:00:49,  1.81s/it]
 69%|██████▊   | 13087/19090 [6:23:07<2:45:41,  1.66s/it]
 69%|██████▊   | 13088/19090 [6:23:09<3:12:49,  1.93s/it]
 69%|██████▊   | 13089/19090 [6:23:11<2:53:25,  1.73s/it][gradnorm] step 13089: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 237.8

 69%|██████▊   | 13090/19090 [6:23:12<2:39:59,  1.60s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 342.8674011230469, 'learning_rate': 1.3599964021878465e-05, 'recon_loss': 0.0016759092774009331, 'recon_static': 0.000705285907315556, 'recon_motion': 0.0009706234020995907, 'recon_norm': 1.343207973241806, 'recon_norm_std': 0.039636795222759244, 'target_norm': 1.6212503433227539, 'recon_cos': 0.6100776016712188, 'gradnorm/compressor.query': 0.1941388822818881, 'gradnorm/compressor.layer0': 7.558465054621447, 'gradnorm/compressor.layer1': 336.99085438257816, 'gradnorm/compressor.layer2': 59.628508970384715, 'gradnorm/compressor.layer3': 17.804000164991013, 'gradnorm/compressor.layer4': 6.16982913332483, 'gradnorm/compressor.layer5': 3.2540041392463563, 'gradnorm/compressor.layer6': 2.0084792636810365, 'gradnorm/compressor.layer7': 2.2879333835987543, 'gradnorm/decoder.layer0': 2.2834558143157966, 'gradnorm/decoder.layer1': 1.3610592635437202, 'gradnorm/decoder.layer2': 0.7203473709503301, 'gradnorm/max_param': 237.80416870117188, 'epoch': 6.86}

 69%|██████▊   | 13090/19090 [6:23:12<2:39:59,  1.60s/it]
 69%|██████▊   | 13091/19090 [6:23:15<3:26:21,  2.06s/it]
 69%|██████▊   | 13092/19090 [6:23:16<3:02:55,  1.83s/it]
 69%|██████▊   | 13093/19090 [6:23:18<2:46:32,  1.67s/it]
 69%|██████▊   | 13094/19090 [6:23:21<3:21:19,  2.01s/it]
 69%|██████▊   | 13095/19090 [6:23:22<2:59:34,  1.80s/it]
 69%|██████▊   | 13096/19090 [6:23:23<2:44:15,  1.64s/it]
 69%|██████▊   | 13097/19090 [6:23:26<3:10:20,  1.91s/it]
 69%|██████▊   | 13098/19090 [6:23:27<2:51:53,  1.72s/it]
 69%|██████▊   | 13099/19090 [6:23:28<2:38:52,  1.59s/it][gradnorm] step 13099: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 254.0

 69%|██████▊   | 13100/19090 [6:23:31<3:19:29,  2.00s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 366.1058654785156, 'learning_rate': 1.3559299329753284e-05, 'recon_loss': 0.001627138801268302, 'recon_static': 0.0006597519997740165, 'recon_motion': 0.0009673867665696889, 'recon_norm': 1.3251311421394347, 'recon_norm_std': 0.04061134811490774, 'target_norm': 1.6203427165746689, 'recon_cos': 0.6197045594453812, 'gradnorm/compressor.query': 0.24351977179884948, 'gradnorm/compressor.layer0': 8.155782751971957, 'gradnorm/compressor.layer1': 359.97863434862353, 'gradnorm/compressor.layer2': 63.10918350738464, 'gradnorm/compressor.layer3': 18.07752402847248, 'gradnorm/compressor.layer4': 6.263157537056239, 'gradnorm/compressor.layer5': 3.7879030321674705, 'gradnorm/compressor.layer6': 2.1925561509287705, 'gradnorm/compressor.layer7': 2.3184761975621204, 'gradnorm/decoder.layer0': 2.5046260947579055, 'gradnorm/decoder.layer1': 1.4684784009083436, 'gradnorm/decoder.layer2': 0.7528674623180768, 'gradnorm/max_param': 254.0210723876953, 'epoch': 6.86}

 69%|██████▊   | 13100/19090 [6:23:31<3:19:29,  2.00s/it]
 69%|██████▊   | 13101/19090 [6:23:33<2:58:20,  1.79s/it]
 69%|██████▊   | 13102/19090 [6:23:34<2:43:21,  1.64s/it]
 69%|██████▊   | 13103/19090 [6:23:36<3:09:45,  1.90s/it]
 69%|██████▊   | 13104/19090 [6:23:38<2:51:15,  1.72s/it]
 69%|██████▊   | 13105/19090 [6:23:39<2:38:16,  1.59s/it]
 69%|██████▊   | 13106/19090 [6:23:42<3:11:52,  1.92s/it]
 69%|██████▊   | 13107/19090 [6:23:43<2:52:37,  1.73s/it]
 69%|██████▊   | 13108/19090 [6:23:44<2:39:14,  1.60s/it]
 69%|██████▊   | 13109/19090 [6:23:47<3:09:54,  1.91s/it][gradnorm] step 13109: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 252.9

 69%|██████▊   | 13110/19090 [6:23:48<2:51:40,  1.72s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 365.1562194824219, 'learning_rate': 1.3518672889735981e-05, 'recon_loss': 0.0016427668568212539, 'recon_static': 0.0006799205933930352, 'recon_motion': 0.0009628463056287728, 'recon_norm': 1.3383437126874924, 'recon_norm_std': 0.04127969061955809, 'target_norm': 1.6191642224788665, 'recon_cos': 0.617737203836441, 'gradnorm/compressor.query': 0.23239436908456132, 'gradnorm/compressor.layer0': 8.32900395147766, 'gradnorm/compressor.layer1': 358.377615965339, 'gradnorm/compressor.layer2': 66.74691662471196, 'gradnorm/compressor.layer3': 17.83479364898717, 'gradnorm/compressor.layer4': 5.895703137940601, 'gradnorm/compressor.layer5': 2.790136905032085, 'gradnorm/compressor.layer6': 2.336136927432418, 'gradnorm/compressor.layer7': 2.356001102648025, 'gradnorm/decoder.layer0': 2.381328883829406, 'gradnorm/decoder.layer1': 1.4135505213093535, 'gradnorm/decoder.layer2': 0.8264140322054301, 'gradnorm/max_param': 252.85861206054688, 'epoch': 6.87}

 69%|██████▊   | 13110/19090 [6:23:48<2:51:40,  1.72s/it]
 69%|██████▊   | 13111/19090 [6:23:49<2:38:41,  1.59s/it]
 69%|██████▊   | 13112/19090 [6:23:52<3:14:27,  1.95s/it]
 69%|██████▊   | 13113/19090 [6:23:53<2:54:26,  1.75s/it]
 69%|██████▊   | 13114/19090 [6:23:55<2:40:56,  1.62s/it]
 69%|██████▊   | 13115/19090 [6:23:58<3:30:45,  2.12s/it]
 69%|██████▊   | 13116/19090 [6:23:59<3:06:02,  1.87s/it]
 69%|██████▊   | 13117/19090 [6:24:01<2:48:35,  1.69s/it]
 69%|██████▊   | 13118/19090 [6:24:03<3:23:04,  2.04s/it]
 69%|██████▊   | 13119/19090 [6:24:05<3:00:24,  1.81s/it][gradnorm] step 13119: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 221.9

 69%|██████▊   | 13120/19090 [6:24:06<2:44:41,  1.66s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 319.5149841308594, 'learning_rate': 1.347808483766147e-05, 'recon_loss': 0.0017642775841522963, 'recon_static': 0.0007846931563108228, 'recon_motion': 0.0009795844365726226, 'recon_norm': 1.3885773181915284, 'recon_norm_std': 0.03920335816219449, 'target_norm': 1.6222061932086944, 'recon_cos': 0.5963123351335525, 'gradnorm/compressor.query': 0.21101241194401918, 'gradnorm/compressor.layer0': 8.488750418465147, 'gradnorm/compressor.layer1': 314.5229274234233, 'gradnorm/compressor.layer2': 52.1791090746874, 'gradnorm/compressor.layer3': 17.722578331463055, 'gradnorm/compressor.layer4': 5.432213679991601, 'gradnorm/compressor.layer5': 3.251528160577472, 'gradnorm/compressor.layer6': 2.0244854308603024, 'gradnorm/compressor.layer7': 2.1682542217736502, 'gradnorm/decoder.layer0': 2.260872321781022, 'gradnorm/decoder.layer1': 1.3354742574349472, 'gradnorm/decoder.layer2': 0.7303309999211322, 'gradnorm/max_param': 221.93702697753906, 'epoch': 6.87}

 69%|██████▊   | 13120/19090 [6:24:06<2:44:41,  1.66s/it]
 69%|██████▊   | 13121/19090 [6:24:09<3:14:15,  1.95s/it]
 69%|██████▊   | 13122/19090 [6:24:10<2:54:12,  1.75s/it]
 69%|██████▊   | 13123/19090 [6:24:11<2:40:18,  1.61s/it]
 69%|██████▊   | 13124/19090 [6:24:14<3:16:30,  1.98s/it]
 69%|██████▉   | 13125/19090 [6:24:15<2:55:50,  1.77s/it]
 69%|██████▉   | 13126/19090 [6:24:17<2:41:22,  1.62s/it]
 69%|██████▉   | 13127/19090 [6:24:19<3:18:03,  1.99s/it]
 69%|██████▉   | 13128/19090 [6:24:21<2:56:56,  1.78s/it]
 69%|██████▉   | 13129/19090 [6:24:22<2:42:16,  1.63s/it][gradnorm] step 13129: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 499.0

 69%|██████▉   | 13130/19090 [6:24:25<3:14:39,  1.96s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 713.8722534179688, 'learning_rate': 1.3437535309236381e-05, 'recon_loss': 0.0017015644203638657, 'recon_static': 0.0007272560585988685, 'recon_motion': 0.0009743083865032531, 'recon_norm': 1.3452365189790725, 'recon_norm_std': 0.040297511406242846, 'target_norm': 1.6219212114810944, 'recon_cos': 0.6038305044174195, 'gradnorm/compressor.query': 1.0553697428090618, 'gradnorm/compressor.layer0': 30.50739136233013, 'gradnorm/compressor.layer1': 708.0645936282932, 'gradnorm/compressor.layer2': 82.03546715350471, 'gradnorm/compressor.layer3': 22.819039355961202, 'gradnorm/compressor.layer4': 6.600025246311799, 'gradnorm/compressor.layer5': 3.425345833575034, 'gradnorm/compressor.layer6': 2.3596775290443097, 'gradnorm/compressor.layer7': 2.5249083398479617, 'gradnorm/decoder.layer0': 2.388821888016098, 'gradnorm/decoder.layer1': 1.4492255773666252, 'gradnorm/decoder.layer2': 0.8261384709235817, 'gradnorm/max_param': 498.99224853515625, 'epoch': 6.88}

 69%|██████▉   | 13130/19090 [6:24:25<3:14:39,  1.96s/it]
 69%|██████▉   | 13131/19090 [6:24:26<2:54:38,  1.76s/it]
 69%|██████▉   | 13132/19090 [6:24:27<2:40:38,  1.62s/it]
 69%|██████▉   | 13133/19090 [6:24:30<3:04:06,  1.85s/it]
 69%|██████▉   | 13134/19090 [6:24:31<2:47:22,  1.69s/it]
 69%|██████▉   | 13135/19090 [6:24:32<2:35:39,  1.57s/it]
 69%|██████▉   | 13136/19090 [6:24:35<3:03:56,  1.85s/it]
 69%|██████▉   | 13137/19090 [6:24:36<2:51:47,  1.73s/it]
 69%|██████▉   | 13138/19090 [6:24:38<2:38:43,  1.60s/it]
 69%|██████▉   | 13139/19090 [6:24:40<3:05:34,  1.87s/it][gradnorm] step 13139: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 235.6

 69%|██████▉   | 13140/19090 [6:24:42<2:52:13,  1.74s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 341.5334777832031, 'learning_rate': 1.3397024440038503e-05, 'recon_loss': 0.0017892974079586565, 'recon_static': 0.0008307017473271116, 'recon_motion': 0.0009585956577211618, 'recon_norm': 1.3981451839208603, 'recon_norm_std': 0.040222059935331345, 'target_norm': 1.621016162633896, 'recon_cos': 0.5933832779526711, 'gradnorm/compressor.query': 0.3054088343521983, 'gradnorm/compressor.layer0': 10.89829784016698, 'gradnorm/compressor.layer1': 333.8763922659103, 'gradnorm/compressor.layer2': 67.94078807015525, 'gradnorm/compressor.layer3': 18.955621389851366, 'gradnorm/compressor.layer4': 6.447492824874383, 'gradnorm/compressor.layer5': 3.9759180412364996, 'gradnorm/compressor.layer6': 2.4579537847595745, 'gradnorm/compressor.layer7': 2.295635381935296, 'gradnorm/decoder.layer0': 2.501213732775335, 'gradnorm/decoder.layer1': 1.379638563867287, 'gradnorm/decoder.layer2': 0.8320899548662689, 'gradnorm/max_param': 235.5719451904297, 'epoch': 6.88}

 69%|██████▉   | 13140/19090 [6:24:42<2:52:13,  1.74s/it]
 69%|██████▉   | 13141/19090 [6:24:43<2:38:58,  1.60s/it]
 69%|██████▉   | 13142/19090 [6:24:45<3:07:17,  1.89s/it]
 69%|██████▉   | 13143/19090 [6:24:47<2:51:59,  1.74s/it]
 69%|██████▉   | 13144/19090 [6:24:48<2:38:38,  1.60s/it]
 69%|██████▉   | 13145/19090 [6:24:51<3:05:17,  1.87s/it]
 69%|██████▉   | 13146/19090 [6:24:52<3:03:58,  1.86s/it]
 69%|██████▉   | 13147/19090 [6:24:54<2:46:59,  1.69s/it]
 69%|██████▉   | 13148/19090 [6:24:56<3:07:40,  1.90s/it]
 69%|██████▉   | 13149/19090 [6:24:58<3:04:32,  1.86s/it][gradnorm] step 13149: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 246.1

 69%|██████▉   | 13150/19090 [6:24:59<2:47:30,  1.69s/it]
                                                         
{'loss': 0.0021, 'grad_norm': 355.9693603515625, 'learning_rate': 1.3356552365516339e-05, 'recon_loss': 0.002063055039616302, 'recon_static': 0.0010988652487867513, 'recon_motion': 0.0009641898417612538, 'recon_norm': 1.4975631296634675, 'recon_norm_std': 0.03666120050475001, 'target_norm': 1.6188274592161178, 'recon_cos': 0.5472233414649963, 'gradnorm/compressor.query': 0.2501535391199023, 'gradnorm/compressor.layer0': 10.865323727555928, 'gradnorm/compressor.layer1': 348.84457722960235, 'gradnorm/compressor.layer2': 67.37317367654578, 'gradnorm/compressor.layer3': 17.71744097898311, 'gradnorm/compressor.layer4': 5.104656502192003, 'gradnorm/compressor.layer5': 2.9945295047730043, 'gradnorm/compressor.layer6': 1.866067845933647, 'gradnorm/compressor.layer7': 2.0694614028520286, 'gradnorm/decoder.layer0': 2.3585280357119474, 'gradnorm/decoder.layer1': 1.2945400037452268, 'gradnorm/decoder.layer2': 0.6731910900869876, 'gradnorm/max_param': 246.13662719726562, 'epoch': 6.89}

 69%|██████▉   | 13150/19090 [6:24:59<2:47:30,  1.69s/it]
 69%|██████▉   | 13151/19090 [6:25:01<3:01:37,  1.83s/it]
 69%|██████▉   | 13152/19090 [6:25:03<3:04:35,  1.87s/it]
 69%|██████▉   | 13153/19090 [6:25:05<2:47:29,  1.69s/it]
 69%|██████▉   | 13154/19090 [6:25:07<3:03:48,  1.86s/it]
 69%|██████▉   | 13155/19090 [6:25:08<2:54:03,  1.76s/it]
 69%|██████▉   | 13156/19090 [6:25:10<2:40:00,  1.62s/it]
 69%|██████▉   | 13157/19090 [6:25:12<3:12:12,  1.94s/it]
 69%|██████▉   | 13158/19090 [6:25:14<3:05:12,  1.87s/it]
 69%|██████▉   | 13159/19090 [6:25:15<2:48:00,  1.70s/it][gradnorm] step 13159: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 202.8

 69%|██████▉   | 13160/19090 [6:25:18<3:09:45,  1.92s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 292.99969482421875, 'learning_rate': 1.331611922098875e-05, 'recon_loss': 0.0016724773187888787, 'recon_static': 0.0007168842523242347, 'recon_motion': 0.00095559307228541, 'recon_norm': 1.3552871465682983, 'recon_norm_std': 0.04091010792180896, 'target_norm': 1.6190887093544006, 'recon_cos': 0.6121556594967842, 'gradnorm/compressor.query': 0.16139006453213747, 'gradnorm/compressor.layer0': 6.5284669047911335, 'gradnorm/compressor.layer1': 287.41830360991275, 'gradnorm/compressor.layer2': 54.19790440410842, 'gradnorm/compressor.layer3': 14.364183154305353, 'gradnorm/compressor.layer4': 5.605836565131979, 'gradnorm/compressor.layer5': 2.6077876945216394, 'gradnorm/compressor.layer6': 1.984107833058409, 'gradnorm/compressor.layer7': 2.0335322543330463, 'gradnorm/decoder.layer0': 2.219508833807243, 'gradnorm/decoder.layer1': 1.1983952879154784, 'gradnorm/decoder.layer2': 0.7111142955235203, 'gradnorm/max_param': 202.8106231689453, 'epoch': 6.89}

 69%|██████▉   | 13160/19090 [6:25:18<3:09:45,  1.92s/it]
 69%|██████▉   | 13161/19090 [6:25:19<3:04:35,  1.87s/it]
 69%|██████▉   | 13162/19090 [6:25:21<2:47:22,  1.69s/it]
 69%|██████▉   | 13163/19090 [6:25:23<2:56:40,  1.79s/it]
 69%|██████▉   | 13164/19090 [6:25:25<2:58:22,  1.81s/it]
 69%|██████▉   | 13165/19090 [6:25:26<2:43:23,  1.65s/it]
 69%|██████▉   | 13166/19090 [6:25:28<2:51:56,  1.74s/it]
 69%|██████▉   | 13167/19090 [6:25:30<3:00:59,  1.83s/it]
 69%|██████▉   | 13168/19090 [6:25:31<2:44:57,  1.67s/it]
 69%|██████▉   | 13169/19090 [6:25:33<3:00:44,  1.83s/it][gradnorm] step 13169: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 203.7

 69%|██████▉   | 13170/19090 [6:25:35<2:58:22,  1.81s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 294.9932556152344, 'learning_rate': 1.3275725141644386e-05, 'recon_loss': 0.0017119049851316958, 'recon_static': 0.0007427423741319216, 'recon_motion': 0.0009691626852145419, 'recon_norm': 1.3614292770624161, 'recon_norm_std': 0.03937413282692433, 'target_norm': 1.6203078776597977, 'recon_cos': 0.6030934259295464, 'gradnorm/compressor.query': 0.19029184879261965, 'gradnorm/compressor.layer0': 6.563113229347065, 'gradnorm/compressor.layer1': 288.7638524815736, 'gradnorm/compressor.layer2': 57.09370723862657, 'gradnorm/compressor.layer3': 16.51508450954526, 'gradnorm/compressor.layer4': 6.060705804293045, 'gradnorm/compressor.layer5': 2.8936054432563325, 'gradnorm/compressor.layer6': 2.064828194405556, 'gradnorm/compressor.layer7': 2.0121291251260245, 'gradnorm/decoder.layer0': 2.335528589902185, 'gradnorm/decoder.layer1': 1.2207385737040424, 'gradnorm/decoder.layer2': 0.725867400379053, 'gradnorm/max_param': 203.72805786132812, 'epoch': 6.9}

 69%|██████▉   | 13170/19090 [6:25:35<2:58:22,  1.81s/it]
 69%|██████▉   | 13171/19090 [6:25:36<2:43:15,  1.65s/it]
 69%|██████▉   | 13172/19090 [6:25:39<3:09:26,  1.92s/it]
 69%|██████▉   | 13173/19090 [6:25:40<2:53:36,  1.76s/it]
 69%|██████▉   | 13174/19090 [6:25:42<2:40:08,  1.62s/it]
 69%|██████▉   | 13175/19090 [6:25:45<3:16:58,  2.00s/it]
 69%|██████▉   | 13176/19090 [6:25:46<2:56:01,  1.79s/it]
 69%|██████▉   | 13177/19090 [6:25:47<2:41:15,  1.64s/it]
 69%|██████▉   | 13178/19090 [6:25:50<3:15:05,  1.98s/it]
 69%|██████▉   | 13179/19090 [6:25:51<2:58:19,  1.81s/it][gradnorm] step 13179: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 247.4

 69%|██████▉   | 13180/19090 [6:25:53<2:43:01,  1.66s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 357.0991516113281, 'learning_rate': 1.3235370262541272e-05, 'recon_loss': 0.0017091217479901389, 'recon_static': 0.0007556061857030727, 'recon_motion': 0.0009535155521007254, 'recon_norm': 1.3675668448209763, 'recon_norm_std': 0.03967254841700196, 'target_norm': 1.6170861214399337, 'recon_cos': 0.6036673158407211, 'gradnorm/compressor.query': 0.22275774298963064, 'gradnorm/compressor.layer0': 8.129289169861444, 'gradnorm/compressor.layer1': 350.6912927526146, 'gradnorm/compressor.layer2': 64.41845469030983, 'gradnorm/compressor.layer3': 16.294502239180435, 'gradnorm/compressor.layer4': 5.197471319269018, 'gradnorm/compressor.layer5': 3.269917856053781, 'gradnorm/compressor.layer6': 1.850428495528922, 'gradnorm/compressor.layer7': 1.8578945003857454, 'gradnorm/decoder.layer0': 2.611419698954133, 'gradnorm/decoder.layer1': 1.3989378391032483, 'gradnorm/decoder.layer2': 0.8182288454970991, 'gradnorm/max_param': 247.44639587402344, 'epoch': 6.9}

 69%|██████▉   | 13180/19090 [6:25:53<2:43:01,  1.66s/it]
 69%|██████▉   | 13181/19090 [6:25:55<3:13:50,  1.97s/it]
 69%|██████▉   | 13182/19090 [6:25:57<3:03:06,  1.86s/it]
 69%|██████▉   | 13183/19090 [6:25:58<2:46:13,  1.69s/it]
 69%|██████▉   | 13184/19090 [6:26:00<3:00:40,  1.84s/it]
 69%|██████▉   | 13185/19090 [6:26:02<3:06:00,  1.89s/it]
 69%|██████▉   | 13186/19090 [6:26:04<2:48:17,  1.71s/it]
 69%|██████▉   | 13187/19090 [6:26:06<3:06:21,  1.89s/it]
 69%|██████▉   | 13188/19090 [6:26:08<3:06:53,  1.90s/it]
 69%|██████▉   | 13189/19090 [6:26:09<2:48:50,  1.72s/it][gradnorm] step 13189: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 286.4

 69%|██████▉   | 13190/19090 [6:26:12<3:11:32,  1.95s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 412.47088623046875, 'learning_rate': 1.3195054718606414e-05, 'recon_loss': 0.0017498552973847837, 'recon_static': 0.0007788340342813171, 'recon_motion': 0.0009710213184007443, 'recon_norm': 1.376951465010643, 'recon_norm_std': 0.03938136268407107, 'target_norm': 1.6209832906723023, 'recon_cos': 0.5966319769620896, 'gradnorm/compressor.query': 0.28929427881026243, 'gradnorm/compressor.layer0': 11.247799806552447, 'gradnorm/compressor.layer1': 405.91289860018, 'gradnorm/compressor.layer2': 68.36121604663714, 'gradnorm/compressor.layer3': 22.467068456722664, 'gradnorm/compressor.layer4': 6.060634994726761, 'gradnorm/compressor.layer5': 2.962536543608669, 'gradnorm/compressor.layer6': 1.9827578706182294, 'gradnorm/compressor.layer7': 2.162635415308063, 'gradnorm/decoder.layer0': 2.4583583334820784, 'gradnorm/decoder.layer1': 1.306735048175322, 'gradnorm/decoder.layer2': 0.6753266188925399, 'gradnorm/max_param': 286.4396667480469, 'epoch': 6.91}

 69%|██████▉   | 13190/19090 [6:26:12<3:11:32,  1.95s/it]
 69%|██████▉   | 13191/19090 [6:26:13<2:56:08,  1.79s/it]
 69%|██████▉   | 13192/19090 [6:26:14<2:41:13,  1.64s/it]
 69%|██████▉   | 13193/19090 [6:26:17<3:07:18,  1.91s/it]
 69%|██████▉   | 13194/19090 [6:26:19<3:05:43,  1.89s/it]
 69%|██████▉   | 13195/19090 [6:26:20<2:48:01,  1.71s/it]
 69%|██████▉   | 13196/19090 [6:26:22<2:59:29,  1.83s/it]
 69%|██████▉   | 13197/19090 [6:26:24<3:11:10,  1.95s/it]
 69%|██████▉   | 13198/19090 [6:26:26<2:51:49,  1.75s/it]
 69%|██████▉   | 13199/19090 [6:26:27<2:51:41,  1.75s/it][gradnorm] step 13199: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 237.1

 69%|██████▉   | 13200/19090 [6:26:30<3:00:07,  1.83s/it]
                                                         
{'loss': 0.0021, 'grad_norm': 341.59033203125, 'learning_rate': 1.3154778644635275e-05, 'recon_loss': 0.0020622762414859606, 'recon_static': 0.0010940608422970398, 'recon_motion': 0.0009682153933681548, 'recon_norm': 1.5029650181531906, 'recon_norm_std': 0.0371168989688158, 'target_norm': 1.6207314282655716, 'recon_cos': 0.5494441464543343, 'gradnorm/compressor.query': 0.31695839307476986, 'gradnorm/compressor.layer0': 12.995058147422151, 'gradnorm/compressor.layer1': 336.0856243206484, 'gradnorm/compressor.layer2': 56.5795683846541, 'gradnorm/compressor.layer3': 17.254390434257957, 'gradnorm/compressor.layer4': 5.817606661826538, 'gradnorm/compressor.layer5': 3.3963521237370378, 'gradnorm/compressor.layer6': 1.9136950081686346, 'gradnorm/compressor.layer7': 2.3806601887748964, 'gradnorm/decoder.layer0': 2.277106192550558, 'gradnorm/decoder.layer1': 1.3608298135366588, 'gradnorm/decoder.layer2': 0.8400931764196359, 'gradnorm/max_param': 237.05670166015625, 'epoch': 6.91}

 69%|██████▉   | 13200/19090 [6:26:30<3:00:07,  1.83s/it]
 69%|██████▉   | 13201/19090 [6:26:31<2:44:04,  1.67s/it]
 69%|██████▉   | 13202/19090 [6:26:33<2:48:17,  1.71s/it]
 69%|██████▉   | 13203/19090 [6:26:35<3:06:02,  1.90s/it]
 69%|██████▉   | 13204/19090 [6:26:36<2:48:05,  1.71s/it]
 69%|██████▉   | 13205/19090 [6:26:38<2:44:29,  1.68s/it]
 69%|██████▉   | 13206/19090 [6:26:40<3:02:03,  1.86s/it]
 69%|██████▉   | 13207/19090 [6:26:41<2:45:20,  1.69s/it]
 69%|██████▉   | 13208/19090 [6:26:43<2:34:07,  1.57s/it]
 69%|██████▉   | 13209/19090 [6:26:46<3:12:02,  1.96s/it][gradnorm] step 13209: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 251.6

 69%|██████▉   | 13210/19090 [6:26:47<2:52:18,  1.76s/it]
                                                         
{'loss': 0.0021, 'grad_norm': 362.91259765625, 'learning_rate': 1.311454217529132e-05, 'recon_loss': 0.002059200088842772, 'recon_static': 0.0010884055503993296, 'recon_motion': 0.0009707945282571018, 'recon_norm': 1.495077547430992, 'recon_norm_std': 0.0363870857283473, 'target_norm': 1.6216408848762511, 'recon_cos': 0.5496613934636116, 'gradnorm/compressor.query': 0.2432032097860128, 'gradnorm/compressor.layer0': 9.395180509659303, 'gradnorm/compressor.layer1': 356.4840020954657, 'gradnorm/compressor.layer2': 64.2340614350167, 'gradnorm/compressor.layer3': 18.337545200076367, 'gradnorm/compressor.layer4': 5.964753575570843, 'gradnorm/compressor.layer5': 4.1783529557532075, 'gradnorm/compressor.layer6': 2.2135589479258018, 'gradnorm/compressor.layer7': 2.612911181056082, 'gradnorm/decoder.layer0': 2.5109578310434775, 'gradnorm/decoder.layer1': 1.466420863245897, 'gradnorm/decoder.layer2': 0.9421498381266085, 'gradnorm/max_param': 251.56333923339844, 'epoch': 6.92}

 69%|██████▉   | 13210/19090 [6:26:47<2:52:18,  1.76s/it]
 69%|██████▉   | 13211/19090 [6:26:48<2:39:12,  1.62s/it]
 69%|██████▉   | 13212/19090 [6:26:51<3:10:49,  1.95s/it]
 69%|██████▉   | 13213/19090 [6:26:52<2:51:20,  1.75s/it]
 69%|██████▉   | 13214/19090 [6:26:54<2:50:29,  1.74s/it]
 69%|██████▉   | 13215/19090 [6:26:56<3:10:43,  1.95s/it]
 69%|██████▉   | 13216/19090 [6:26:58<2:51:37,  1.75s/it]
 69%|██████▉   | 13217/19090 [6:26:59<2:43:08,  1.67s/it]
 69%|██████▉   | 13218/19090 [6:27:02<3:11:28,  1.96s/it]
 69%|██████▉   | 13219/19090 [6:27:03<2:52:14,  1.76s/it][gradnorm] step 13219: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 228.9

 69%|██████▉   | 13220/19090 [6:27:04<2:42:54,  1.67s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 330.29705810546875, 'learning_rate': 1.3074345445105654e-05, 'recon_loss': 0.0018230024172225967, 'recon_static': 0.0008455652539851144, 'recon_motion': 0.000977437228721101, 'recon_norm': 1.4052049785852432, 'recon_norm_std': 0.03843003800138831, 'target_norm': 1.6217058569192886, 'recon_cos': 0.5834882393479347, 'gradnorm/compressor.query': 0.22690566152610064, 'gradnorm/compressor.layer0': 7.4810299336660995, 'gradnorm/compressor.layer1': 324.42268435638096, 'gradnorm/compressor.layer2': 58.094098808796836, 'gradnorm/compressor.layer3': 18.86274048264528, 'gradnorm/compressor.layer4': 5.445975352242374, 'gradnorm/compressor.layer5': 3.2066195150492103, 'gradnorm/compressor.layer6': 2.2510517629114264, 'gradnorm/compressor.layer7': 2.5368804475961233, 'gradnorm/decoder.layer0': 2.364287056755834, 'gradnorm/decoder.layer1': 1.25499704041455, 'gradnorm/decoder.layer2': 0.8552298364634724, 'gradnorm/max_param': 228.88050842285156, 'epoch': 6.93}

 69%|██████▉   | 13220/19090 [6:27:04<2:42:54,  1.67s/it]
 69%|██████▉   | 13221/19090 [6:27:07<3:09:41,  1.94s/it]
 69%|██████▉   | 13222/19090 [6:27:08<2:50:53,  1.75s/it]
 69%|██████▉   | 13223/19090 [6:27:10<2:37:42,  1.61s/it]
 69%|██████▉   | 13224/19090 [6:27:12<3:13:39,  1.98s/it]
 69%|██████▉   | 13225/19090 [6:27:14<2:53:11,  1.77s/it]
 69%|██████▉   | 13226/19090 [6:27:15<2:39:26,  1.63s/it]
 69%|██████▉   | 13227/19090 [6:27:18<3:17:18,  2.02s/it]
 69%|██████▉   | 13228/19090 [6:27:19<2:55:42,  1.80s/it]
 69%|██████▉   | 13229/19090 [6:27:21<2:40:38,  1.64s/it][gradnorm] step 13229: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 251.6

 69%|██████▉   | 13230/19090 [6:27:23<3:17:07,  2.02s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 362.9461975097656, 'learning_rate': 1.303418858847648e-05, 'recon_loss': 0.0017270971555262804, 'recon_static': 0.0007545764281530864, 'recon_motion': 0.0009725206735311076, 'recon_norm': 1.364870262145996, 'recon_norm_std': 0.04015585146844387, 'target_norm': 1.6215452194213866, 'recon_cos': 0.6003024369478226, 'gradnorm/compressor.query': 0.21546728509871102, 'gradnorm/compressor.layer0': 8.518091582996076, 'gradnorm/compressor.layer1': 356.6149008384254, 'gradnorm/compressor.layer2': 63.523638793341135, 'gradnorm/compressor.layer3': 19.55188051043021, 'gradnorm/compressor.layer4': 5.943781691987177, 'gradnorm/compressor.layer5': 3.5340334294653486, 'gradnorm/compressor.layer6': 1.984416030099164, 'gradnorm/compressor.layer7': 2.358357494306868, 'gradnorm/decoder.layer0': 2.3551891660746924, 'gradnorm/decoder.layer1': 1.4358503578816568, 'gradnorm/decoder.layer2': 0.8247662574365996, 'gradnorm/max_param': 251.60302734375, 'epoch': 6.93}

 69%|██████▉   | 13230/19090 [6:27:23<3:17:07,  2.02s/it]
 69%|██████▉   | 13231/19090 [6:27:25<2:56:14,  1.80s/it]
 69%|██████▉   | 13232/19090 [6:27:26<2:41:15,  1.65s/it]
 69%|██████▉   | 13233/19090 [6:27:29<3:15:21,  2.00s/it]
 69%|██████▉   | 13234/19090 [6:27:30<2:54:36,  1.79s/it]
 69%|██████▉   | 13235/19090 [6:27:31<2:40:08,  1.64s/it]
 69%|██████▉   | 13236/19090 [6:27:34<3:11:37,  1.96s/it]
 69%|██████▉   | 13237/19090 [6:27:35<2:52:02,  1.76s/it]
 69%|██████▉   | 13238/19090 [6:27:37<2:38:05,  1.62s/it]
 69%|██████▉   | 13239/19090 [6:27:39<3:05:15,  1.90s/it][gradnorm] step 13239: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 313.8

 69%|██████▉   | 13240/19090 [6:27:41<2:47:50,  1.72s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 452.2278747558594, 'learning_rate': 1.2994071739668679e-05, 'recon_loss': 0.0016544910933589563, 'recon_static': 0.0006822649535024539, 'recon_motion': 0.0009722261092974804, 'recon_norm': 1.340604105591774, 'recon_norm_std': 0.04059314774349332, 'target_norm': 1.621859073638916, 'recon_cos': 0.616115739941597, 'gradnorm/compressor.query': 0.2811584191098205, 'gradnorm/compressor.layer0': 10.53352996867974, 'gradnorm/compressor.layer1': 444.71193274410797, 'gradnorm/compressor.layer2': 77.5032036888234, 'gradnorm/compressor.layer3': 23.239565240729547, 'gradnorm/compressor.layer4': 6.7006925424239325, 'gradnorm/compressor.layer5': 3.9211258495694348, 'gradnorm/compressor.layer6': 2.33591892400902, 'gradnorm/compressor.layer7': 2.5089864389623986, 'gradnorm/decoder.layer0': 2.7994984518170525, 'gradnorm/decoder.layer1': 1.6333644244420278, 'gradnorm/decoder.layer2': 0.9726383559944126, 'gradnorm/max_param': 313.7818603515625, 'epoch': 6.94}

 69%|██████▉   | 13240/19090 [6:27:41<2:47:50,  1.72s/it]
 69%|██████▉   | 13241/19090 [6:27:42<2:35:08,  1.59s/it]
 69%|██████▉   | 13242/19090 [6:27:44<3:01:59,  1.87s/it]
 69%|██████▉   | 13243/19090 [6:27:46<2:45:20,  1.70s/it]
 69%|██████▉   | 13244/19090 [6:27:47<2:33:25,  1.57s/it]
 69%|██████▉   | 13245/19090 [6:27:50<3:04:06,  1.89s/it]
 69%|██████▉   | 13246/19090 [6:27:51<2:46:27,  1.71s/it]
 69%|██████▉   | 13247/19090 [6:27:52<2:34:25,  1.59s/it]
 69%|██████▉   | 13248/19090 [6:27:55<3:11:38,  1.97s/it]
 69%|██████▉   | 13249/19090 [6:27:56<2:51:54,  1.77s/it][gradnorm] step 13249: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 273.0

 69%|██████▉   | 13250/19090 [6:27:58<2:38:01,  1.62s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 392.8645935058594, 'learning_rate': 1.295399503281336e-05, 'recon_loss': 0.001602616484160535, 'recon_static': 0.0006391967530362308, 'recon_motion': 0.000963419719482772, 'recon_norm': 1.3158679962158204, 'recon_norm_std': 0.04036427140235901, 'target_norm': 1.6188895791769027, 'recon_cos': 0.6241974905133247, 'gradnorm/compressor.query': 0.2408970477758596, 'gradnorm/compressor.layer0': 8.084083233947956, 'gradnorm/compressor.layer1': 387.0374860720858, 'gradnorm/compressor.layer2': 62.69641986950391, 'gradnorm/compressor.layer3': 22.23272720418138, 'gradnorm/compressor.layer4': 5.319284930824158, 'gradnorm/compressor.layer5': 2.9706976675451187, 'gradnorm/compressor.layer6': 1.9096810153407253, 'gradnorm/compressor.layer7': 2.2062652068654174, 'gradnorm/decoder.layer0': 2.4360375541022226, 'gradnorm/decoder.layer1': 1.3479064709112205, 'gradnorm/decoder.layer2': 0.788083214737397, 'gradnorm/max_param': 273.0470275878906, 'epoch': 6.94}

 69%|██████▉   | 13250/19090 [6:27:58<2:38:01,  1.62s/it]
 69%|██████▉   | 13251/19090 [6:28:01<3:21:39,  2.07s/it]
 69%|██████▉   | 13252/19090 [6:28:02<2:58:49,  1.84s/it]
 69%|██████▉   | 13253/19090 [6:28:03<2:42:38,  1.67s/it]
 69%|██████▉   | 13254/19090 [6:28:06<3:15:58,  2.01s/it]
 69%|██████▉   | 13255/19090 [6:28:07<2:54:37,  1.80s/it]
 69%|██████▉   | 13256/19090 [6:28:09<2:39:45,  1.64s/it]
 69%|██████▉   | 13257/19090 [6:28:12<3:14:52,  2.00s/it]
 69%|██████▉   | 13258/19090 [6:28:13<2:53:50,  1.79s/it]
 69%|██████▉   | 13259/19090 [6:28:14<2:39:11,  1.64s/it][gradnorm] step 13259: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 305.7

 69%|██████▉   | 13260/19090 [6:28:17<3:12:33,  1.98s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 439.9573059082031, 'learning_rate': 1.291395860190747e-05, 'recon_loss': 0.0016158850135980173, 'recon_static': 0.000645332460408099, 'recon_motion': 0.0009705525095341727, 'recon_norm': 1.3226341754198074, 'recon_norm_std': 0.03991403812542558, 'target_norm': 1.6210906475782394, 'recon_cos': 0.6229048848152161, 'gradnorm/compressor.query': 0.27282241032046933, 'gradnorm/compressor.layer0': 9.56865459172818, 'gradnorm/compressor.layer1': 433.3213479913031, 'gradnorm/compressor.layer2': 71.72774657624656, 'gradnorm/compressor.layer3': 22.076072257278778, 'gradnorm/compressor.layer4': 6.0853053341242855, 'gradnorm/compressor.layer5': 3.7594210221540716, 'gradnorm/compressor.layer6': 2.1652559921063697, 'gradnorm/compressor.layer7': 2.2216117563446938, 'gradnorm/decoder.layer0': 2.701364638593711, 'gradnorm/decoder.layer1': 1.5080268193650253, 'gradnorm/decoder.layer2': 0.8951079662563021, 'gradnorm/max_param': 305.73968505859375, 'epoch': 6.95}

 69%|██████▉   | 13260/19090 [6:28:17<3:12:33,  1.98s/it]
 69%|██████▉   | 13261/19090 [6:28:18<2:52:11,  1.77s/it]
 69%|██████▉   | 13262/19090 [6:28:19<2:37:57,  1.63s/it]
 69%|██████▉   | 13263/19090 [6:28:22<3:14:50,  2.01s/it]
 69%|██████▉   | 13264/19090 [6:28:24<2:53:46,  1.79s/it]
 69%|██████▉   | 13265/19090 [6:28:25<2:38:59,  1.64s/it]
 69%|██████▉   | 13266/19090 [6:28:28<3:12:08,  1.98s/it]
 69%|██████▉   | 13267/19090 [6:28:29<2:52:19,  1.78s/it]
 70%|██████▉   | 13268/19090 [6:28:30<2:38:09,  1.63s/it]
 70%|██████▉   | 13269/19090 [6:28:33<3:01:31,  1.87s/it][gradnorm] step 13269: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 232.0

 70%|██████▉   | 13270/19090 [6:28:34<2:44:27,  1.70s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 335.7764587402344, 'learning_rate': 1.287396258081322e-05, 'recon_loss': 0.0016153860866324975, 'recon_static': 0.000640927733911667, 'recon_motion': 0.0009744583629071712, 'recon_norm': 1.3189550310373306, 'recon_norm_std': 0.040654001478105786, 'target_norm': 1.6219162613153457, 'recon_cos': 0.6228192865848541, 'gradnorm/compressor.query': 0.2080708171004897, 'gradnorm/compressor.layer0': 7.774242508960906, 'gradnorm/compressor.layer1': 328.72356119922404, 'gradnorm/compressor.layer2': 64.62777284912038, 'gradnorm/compressor.layer3': 19.858459883764354, 'gradnorm/compressor.layer4': 5.447156202681462, 'gradnorm/compressor.layer5': 3.2762752059554905, 'gradnorm/compressor.layer6': 1.677383474973161, 'gradnorm/compressor.layer7': 1.9543910691418331, 'gradnorm/decoder.layer0': 2.327898360108082, 'gradnorm/decoder.layer1': 1.3960754649497045, 'gradnorm/decoder.layer2': 0.8076200658154196, 'gradnorm/max_param': 231.97877502441406, 'epoch': 6.95}

 70%|██████▉   | 13270/19090 [6:28:34<2:44:27,  1.70s/it]
 70%|██████▉   | 13271/19090 [6:28:35<2:32:32,  1.57s/it]
 70%|██████▉   | 13272/19090 [6:28:38<3:04:48,  1.91s/it]
 70%|██████▉   | 13273/19090 [6:28:39<2:46:48,  1.72s/it]
 70%|██████▉   | 13274/19090 [6:28:41<2:34:04,  1.59s/it]
 70%|██████▉   | 13275/19090 [6:28:44<3:16:05,  2.02s/it]
 70%|██████▉   | 13276/19090 [6:28:45<2:54:55,  1.81s/it]
 70%|██████▉   | 13277/19090 [6:28:46<2:39:49,  1.65s/it]
 70%|██████▉   | 13278/19090 [6:28:49<3:06:11,  1.92s/it]
 70%|██████▉   | 13279/19090 [6:28:50<2:47:37,  1.73s/it][gradnorm] step 13279: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 319.7

 70%|██████▉   | 13280/19090 [6:28:51<2:34:54,  1.60s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 461.5484313964844, 'learning_rate': 1.2834007103257745e-05, 'recon_loss': 0.001565770962042734, 'recon_static': 0.0005931411884375848, 'recon_motion': 0.0009726298128953203, 'recon_norm': 1.3007852911949158, 'recon_norm_std': 0.040229975245893, 'target_norm': 1.620092636346817, 'recon_cos': 0.6329089805483818, 'gradnorm/compressor.query': 0.29730781576585574, 'gradnorm/compressor.layer0': 9.506998796836301, 'gradnorm/compressor.layer1': 453.01740101788585, 'gradnorm/compressor.layer2': 84.3862810745547, 'gradnorm/compressor.layer3': 23.081905316228635, 'gradnorm/compressor.layer4': 5.280034754956619, 'gradnorm/compressor.layer5': 3.284938374119846, 'gradnorm/compressor.layer6': 2.117448853975094, 'gradnorm/compressor.layer7': 2.264786032197515, 'gradnorm/decoder.layer0': 2.574333315718122, 'gradnorm/decoder.layer1': 1.4614970721627558, 'gradnorm/decoder.layer2': 0.9336031031439332, 'gradnorm/max_param': 319.6727294921875, 'epoch': 6.96}

 70%|██████▉   | 13280/19090 [6:28:51<2:34:54,  1.60s/it]
 70%|██████▉   | 13281/19090 [6:28:54<2:55:05,  1.81s/it]
 70%|██████▉   | 13282/19090 [6:28:55<2:40:00,  1.65s/it]
 70%|██████▉   | 13283/19090 [6:28:56<2:30:10,  1.55s/it]
 70%|██████▉   | 13284/19090 [6:28:59<3:08:38,  1.95s/it]
 70%|██████▉   | 13285/19090 [6:29:00<2:49:24,  1.75s/it]
 70%|██████▉   | 13286/19090 [6:29:02<2:36:08,  1.61s/it]
 70%|██████▉   | 13287/19090 [6:29:04<3:09:51,  1.96s/it]
 70%|██████▉   | 13288/19090 [6:29:06<2:50:16,  1.76s/it]
 70%|██████▉   | 13289/19090 [6:29:07<2:36:28,  1.62s/it][gradnorm] step 13289: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 279.5

 70%|██████▉   | 13290/19090 [6:29:09<2:58:28,  1.85s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 402.4693603515625, 'learning_rate': 1.279409230283265e-05, 'recon_loss': 0.0015943820326356217, 'recon_static': 0.0006257728149648756, 'recon_motion': 0.0009686092467745766, 'recon_norm': 1.3091124355793, 'recon_norm_std': 0.04069842090830207, 'target_norm': 1.6198637783527374, 'recon_cos': 0.6262339979410172, 'gradnorm/compressor.query': 0.2005974057790482, 'gradnorm/compressor.layer0': 7.101118520826573, 'gradnorm/compressor.layer1': 396.11186472384287, 'gradnorm/compressor.layer2': 68.4295407757589, 'gradnorm/compressor.layer3': 16.930816795803057, 'gradnorm/compressor.layer4': 5.260629701234979, 'gradnorm/compressor.layer5': 3.414543065657395, 'gradnorm/compressor.layer6': 1.803854613888851, 'gradnorm/compressor.layer7': 2.3100906398875023, 'gradnorm/decoder.layer0': 2.575066619307561, 'gradnorm/decoder.layer1': 1.4448463306752546, 'gradnorm/decoder.layer2': 0.7845048242254095, 'gradnorm/max_param': 279.535400390625, 'epoch': 6.96}

 70%|██████▉   | 13290/19090 [6:29:09<2:58:28,  1.85s/it]
 70%|██████▉   | 13291/19090 [6:29:11<2:42:15,  1.68s/it]
 70%|██████▉   | 13292/19090 [6:29:13<2:46:49,  1.73s/it]
 70%|██████▉   | 13293/19090 [6:29:15<3:08:45,  1.95s/it]
 70%|██████▉   | 13294/19090 [6:29:16<2:49:20,  1.75s/it]
 70%|██████▉   | 13295/19090 [6:29:18<2:43:13,  1.69s/it]
 70%|██████▉   | 13296/19090 [6:29:21<3:12:31,  1.99s/it]
 70%|██████▉   | 13297/19090 [6:29:22<2:51:58,  1.78s/it]
 70%|██████▉   | 13298/19090 [6:29:23<2:40:46,  1.67s/it]
 70%|██████▉   | 13299/19090 [6:29:26<3:09:16,  1.96s/it][gradnorm] step 13299: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 290.5

 70%|██████▉   | 13300/19090 [6:29:27<2:49:46,  1.76s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 420.17822265625, 'learning_rate': 1.2754218312993482e-05, 'recon_loss': 0.0016465742199216038, 'recon_static': 0.0006700553887640126, 'recon_motion': 0.0009765188908204437, 'recon_norm': 1.3300939440727233, 'recon_norm_std': 0.039855904318392275, 'target_norm': 1.622423881292343, 'recon_cos': 0.6166256979107857, 'gradnorm/compressor.query': 0.3894059040670203, 'gradnorm/compressor.layer0': 12.46411413760928, 'gradnorm/compressor.layer1': 411.737173449277, 'gradnorm/compressor.layer2': 80.67601946791562, 'gradnorm/compressor.layer3': 16.850504863657832, 'gradnorm/compressor.layer4': 5.615816376088724, 'gradnorm/compressor.layer5': 4.517208679635146, 'gradnorm/compressor.layer6': 2.1170106933640818, 'gradnorm/compressor.layer7': 2.434451519959567, 'gradnorm/decoder.layer0': 2.9330696623235815, 'gradnorm/decoder.layer1': 1.5085764318466819, 'gradnorm/decoder.layer2': 0.9626810956243017, 'gradnorm/max_param': 290.52081298828125, 'epoch': 6.97}

 70%|██████▉   | 13300/19090 [6:29:27<2:49:46,  1.76s/it]
 70%|██████▉   | 13301/19090 [6:29:29<2:41:36,  1.67s/it]
 70%|██████▉   | 13302/19090 [6:29:31<3:06:27,  1.93s/it]
 70%|██████▉   | 13303/19090 [6:29:32<2:47:43,  1.74s/it]
 70%|██████▉   | 13304/19090 [6:29:34<2:34:35,  1.60s/it]
 70%|██████▉   | 13305/19090 [6:29:37<3:09:14,  1.96s/it]
 70%|██████▉   | 13306/19090 [6:29:38<2:49:41,  1.76s/it]
 70%|██████▉   | 13307/19090 [6:29:39<2:35:59,  1.62s/it]
 70%|██████▉   | 13308/19090 [6:29:42<3:14:27,  2.02s/it]
 70%|██████▉   | 13309/19090 [6:29:43<2:53:28,  1.80s/it][gradnorm] step 13309: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 260.1

 70%|██████▉   | 13310/19090 [6:29:45<2:38:41,  1.65s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 375.5431213378906, 'learning_rate': 1.2714385267059356e-05, 'recon_loss': 0.0017518990265671165, 'recon_static': 0.0007857112534111366, 'recon_motion': 0.0009661878197221085, 'recon_norm': 1.3824802607297897, 'recon_norm_std': 0.03973996806889772, 'target_norm': 1.6192575246095657, 'recon_cos': 0.5963417381048203, 'gradnorm/compressor.query': 0.22847896673144433, 'gradnorm/compressor.layer0': 8.674101892423026, 'gradnorm/compressor.layer1': 368.6282638037946, 'gradnorm/compressor.layer2': 67.83590465924, 'gradnorm/compressor.layer3': 20.198488628387366, 'gradnorm/compressor.layer4': 5.147454545652691, 'gradnorm/compressor.layer5': 3.245653914527201, 'gradnorm/compressor.layer6': 2.2487513998386675, 'gradnorm/compressor.layer7': 2.9452289599806982, 'gradnorm/decoder.layer0': 2.7077721136551585, 'gradnorm/decoder.layer1': 1.4025310508160285, 'gradnorm/decoder.layer2': 0.8870990626593686, 'gradnorm/max_param': 260.0936279296875, 'epoch': 6.97}

 70%|██████▉   | 13310/19090 [6:29:45<2:38:41,  1.65s/it]
 70%|██████▉   | 13311/19090 [6:29:47<3:01:44,  1.89s/it]
 70%|██████▉   | 13312/19090 [6:29:48<2:44:21,  1.71s/it]
 70%|██████▉   | 13313/19090 [6:29:50<2:35:34,  1.62s/it]
 70%|██████▉   | 13314/19090 [6:29:52<3:05:44,  1.93s/it]
 70%|██████▉   | 13315/19090 [6:29:54<2:47:02,  1.74s/it]
 70%|██████▉   | 13316/19090 [6:29:55<2:41:18,  1.68s/it]
 70%|██████▉   | 13317/19090 [6:29:58<3:08:12,  1.96s/it]
 70%|██████▉   | 13318/19090 [6:29:59<2:48:48,  1.75s/it]
 70%|██████▉   | 13319/19090 [6:30:01<2:41:05,  1.67s/it][gradnorm] step 13319: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 273.5

 70%|██████▉   | 13320/19090 [6:30:03<3:10:28,  1.98s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 396.1007080078125, 'learning_rate': 1.2674593298212528e-05, 'recon_loss': 0.0017014457640470936, 'recon_static': 0.0007362785094301216, 'recon_motion': 0.0009651672939071432, 'recon_norm': 1.359402397274971, 'recon_norm_std': 0.0397742360830307, 'target_norm': 1.6194476187229156, 'recon_cos': 0.6055873751640319, 'gradnorm/compressor.query': 0.22473822505241364, 'gradnorm/compressor.layer0': 8.003180348518828, 'gradnorm/compressor.layer1': 387.7150209883543, 'gradnorm/compressor.layer2': 78.05752711342129, 'gradnorm/compressor.layer3': 18.68609062672437, 'gradnorm/compressor.layer4': 5.577194007950533, 'gradnorm/compressor.layer5': 3.379393084960769, 'gradnorm/compressor.layer6': 2.1403053212055885, 'gradnorm/compressor.layer7': 2.7811909358578744, 'gradnorm/decoder.layer0': 2.9756325698368453, 'gradnorm/decoder.layer1': 1.483403018143232, 'gradnorm/decoder.layer2': 0.8425347795863516, 'gradnorm/max_param': 273.523193359375, 'epoch': 6.98}

 70%|██████▉   | 13320/19090 [6:30:03<3:10:28,  1.98s/it]
 70%|██████▉   | 13321/19090 [6:30:05<2:50:26,  1.77s/it]
 70%|██████▉   | 13322/19090 [6:30:06<2:36:23,  1.63s/it]
 70%|██████▉   | 13323/19090 [6:30:09<3:13:19,  2.01s/it]
 70%|██████▉   | 13324/19090 [6:30:10<2:52:21,  1.79s/it]
 70%|██████▉   | 13325/19090 [6:30:11<2:37:40,  1.64s/it]
 70%|██████▉   | 13326/19090 [6:30:14<3:07:28,  1.95s/it]
 70%|██████▉   | 13327/19090 [6:30:15<2:48:12,  1.75s/it]
 70%|██████▉   | 13328/19090 [6:30:17<2:34:44,  1.61s/it]
 70%|██████▉   | 13329/19090 [6:30:20<3:11:00,  1.99s/it][gradnorm] step 13329: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 351.8

 70%|██████▉   | 13330/19090 [6:30:21<2:50:45,  1.78s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 506.45416259765625, 'learning_rate': 1.2634842539497857e-05, 'recon_loss': 0.001605570237734355, 'recon_static': 0.0006370165836415253, 'recon_motion': 0.0009685537123004905, 'recon_norm': 1.3147342026233673, 'recon_norm_std': 0.04078161371871829, 'target_norm': 1.6209861934185028, 'recon_cos': 0.6244150027632713, 'gradnorm/compressor.query': 0.33753543994408985, 'gradnorm/compressor.layer0': 12.607410225836988, 'gradnorm/compressor.layer1': 498.6329749926292, 'gradnorm/compressor.layer2': 84.42668555838847, 'gradnorm/compressor.layer3': 22.42034783562272, 'gradnorm/compressor.layer4': 6.469456435988145, 'gradnorm/compressor.layer5': 3.305959037568532, 'gradnorm/compressor.layer6': 1.9685773016857637, 'gradnorm/compressor.layer7': 2.477552342482639, 'gradnorm/decoder.layer0': 2.472708895877645, 'gradnorm/decoder.layer1': 1.3809735845792062, 'gradnorm/decoder.layer2': 0.7433418933378387, 'gradnorm/max_param': 351.82623291015625, 'epoch': 6.98}

 70%|██████▉   | 13330/19090 [6:30:21<2:50:45,  1.78s/it]
 70%|██████▉   | 13331/19090 [6:30:22<2:36:36,  1.63s/it]
 70%|██████▉   | 13332/19090 [6:30:25<3:01:16,  1.89s/it]
 70%|██████▉   | 13333/19090 [6:30:26<2:43:47,  1.71s/it]
 70%|██████▉   | 13334/19090 [6:30:27<2:32:58,  1.59s/it]
 70%|██████▉   | 13335/19090 [6:30:30<3:12:54,  2.01s/it]
 70%|██████▉   | 13336/19090 [6:30:31<2:51:56,  1.79s/it]
 70%|██████▉   | 13337/19090 [6:30:33<2:37:21,  1.64s/it]
 70%|██████▉   | 13338/19090 [6:30:35<3:08:53,  1.97s/it]
 70%|██████▉   | 13339/19090 [6:30:37<2:49:09,  1.76s/it][gradnorm] step 13339: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 276.7

 70%|██████▉   | 13340/19090 [6:30:38<2:35:38,  1.62s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 399.5028381347656, 'learning_rate': 1.2595133123822433e-05, 'recon_loss': 0.0016372689016861842, 'recon_static': 0.0006716814546962269, 'recon_motion': 0.0009655874164309353, 'recon_norm': 1.3317697793245316, 'recon_norm_std': 0.04032058957964182, 'target_norm': 1.6194181174039841, 'recon_cos': 0.617645688354969, 'gradnorm/compressor.query': 0.2424176030546335, 'gradnorm/compressor.layer0': 8.586069899172529, 'gradnorm/compressor.layer1': 392.2084506356282, 'gradnorm/compressor.layer2': 72.47884591705879, 'gradnorm/compressor.layer3': 19.221346845643957, 'gradnorm/compressor.layer4': 6.533156997221313, 'gradnorm/compressor.layer5': 3.761287232680923, 'gradnorm/compressor.layer6': 2.401053924418927, 'gradnorm/compressor.layer7': 2.4839779518549503, 'gradnorm/decoder.layer0': 2.609171511088085, 'gradnorm/decoder.layer1': 1.539118499511633, 'gradnorm/decoder.layer2': 0.8193180777930017, 'gradnorm/max_param': 276.74871826171875, 'epoch': 6.99}

 70%|██████▉   | 13340/19090 [6:30:38<2:35:38,  1.62s/it]
 70%|██████▉   | 13341/19090 [6:30:41<3:03:08,  1.91s/it]
 70%|██████▉   | 13342/19090 [6:30:42<2:45:11,  1.72s/it]
 70%|██████▉   | 13343/19090 [6:30:43<2:33:33,  1.60s/it]
 70%|██████▉   | 13344/19090 [6:30:46<3:11:46,  2.00s/it]
 70%|██████▉   | 13345/19090 [6:30:47<2:50:58,  1.79s/it]
 70%|██████▉   | 13346/19090 [6:30:49<2:36:44,  1.64s/it]
 70%|██████▉   | 13347/19090 [6:30:52<3:12:36,  2.01s/it]
 70%|██████▉   | 13348/19090 [6:30:53<2:51:41,  1.79s/it]
 70%|██████▉   | 13349/19090 [6:30:54<2:37:13,  1.64s/it][gradnorm] step 13349: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 338.6

 70%|██████▉   | 13350/19090 [6:30:57<3:09:45,  1.98s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 488.71685791015625, 'learning_rate': 1.2555465183955156e-05, 'recon_loss': 0.0016048701072577387, 'recon_static': 0.0006350988609483465, 'recon_motion': 0.0009697712550405413, 'recon_norm': 1.3164355516433717, 'recon_norm_std': 0.040355578809976575, 'target_norm': 1.6196320921182632, 'recon_cos': 0.6242225497961045, 'gradnorm/compressor.query': 0.4043567263028071, 'gradnorm/compressor.layer0': 18.071055567298945, 'gradnorm/compressor.layer1': 479.87102759491535, 'gradnorm/compressor.layer2': 87.5074438797923, 'gradnorm/compressor.layer3': 22.538825551180846, 'gradnorm/compressor.layer4': 6.409582843507306, 'gradnorm/compressor.layer5': 3.69467628599264, 'gradnorm/compressor.layer6': 2.4352736453659003, 'gradnorm/compressor.layer7': 2.5213645711577883, 'gradnorm/decoder.layer0': 2.465304517526332, 'gradnorm/decoder.layer1': 1.36160126713863, 'gradnorm/decoder.layer2': 0.8645047454659612, 'gradnorm/max_param': 338.5739440917969, 'epoch': 6.99}

 70%|██████▉   | 13350/19090 [6:30:57<3:09:45,  1.98s/it]
 70%|██████▉   | 13351/19090 [6:30:58<2:49:47,  1.78s/it]
 70%|██████▉   | 13352/19090 [6:31:00<2:35:52,  1.63s/it]
 70%|██████▉   | 13353/19090 [6:31:03<3:14:17,  2.03s/it]
 70%|██████▉   | 13354/19090 [6:31:04<2:52:48,  1.81s/it]
 70%|██████▉   | 13355/19090 [6:31:05<2:37:57,  1.65s/it]
 70%|██████▉   | 13356/19090 [6:31:08<3:05:52,  1.94s/it]
 70%|██████▉   | 13357/19090 [6:31:09<2:48:13,  1.76s/it]
 70%|██████▉   | 13358/19090 [6:31:10<2:34:48,  1.62s/it]
 70%|██████▉   | 13359/19090 [6:31:12<2:38:22,  1.66s/it][gradnorm] step 13359: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 358.6

 70%|██████▉   | 13360/19090 [6:31:13<2:27:17,  1.54s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 517.5089721679688, 'learning_rate': 1.2515838852526185e-05, 'recon_loss': 0.0017365920793963596, 'recon_static': 0.0007754747974104248, 'recon_motion': 0.0009611172339646145, 'recon_norm': 1.3780615687370301, 'recon_norm_std': 0.03943872721865773, 'target_norm': 1.6184441536664962, 'recon_cos': 0.5996974110603333, 'gradnorm/compressor.query': 0.26459825046920354, 'gradnorm/compressor.layer0': 10.714880419983734, 'gradnorm/compressor.layer1': 508.22362204053445, 'gradnorm/compressor.layer2': 93.42304967692662, 'gradnorm/compressor.layer3': 24.4575013555881, 'gradnorm/compressor.layer4': 6.393002690319902, 'gradnorm/compressor.layer5': 4.195871440230782, 'gradnorm/compressor.layer6': 2.69295878816582, 'gradnorm/compressor.layer7': 2.63656294750912, 'gradnorm/decoder.layer0': 2.7079688976310474, 'gradnorm/decoder.layer1': 1.552402369674372, 'gradnorm/decoder.layer2': 0.97366198282727, 'gradnorm/max_param': 358.59063720703125, 'epoch': 7.0}

 70%|██████▉   | 13360/19090 [6:31:13<2:27:17,  1.54s/it]
 70%|██████▉   | 13361/19090 [6:31:15<2:19:24,  1.46s/it]
 70%|██████▉   | 13362/19090 [6:31:16<2:13:54,  1.40s/it]
 70%|███████   | 13363/19090 [6:31:18<2:22:30,  1.49s/it]
 70%|███████   | 13364/19090 [6:31:26<5:32:58,  3.49s/it]
 70%|███████   | 13365/19090 [6:31:27<4:30:13,  2.83s/it]
 70%|███████   | 13366/19090 [6:31:31<4:47:07,  3.01s/it]
 70%|███████   | 13367/19090 [6:31:32<3:58:05,  2.50s/it]
 70%|███████   | 13368/19090 [6:31:33<3:23:27,  2.13s/it]
 70%|███████   | 13369/19090 [6:31:36<3:31:22,  2.22s/it][gradnorm] step 13369: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 251.5

 70%|███████   | 13370/19090 [6:31:37<3:04:47,  1.94s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 364.6425476074219, 'learning_rate': 1.247625426202659e-05, 'recon_loss': 0.001748381950892508, 'recon_static': 0.0007780332642141729, 'recon_motion': 0.0009703486721264198, 'recon_norm': 1.3728771179914474, 'recon_norm_std': 0.03906475035473704, 'target_norm': 1.6214995294809342, 'recon_cos': 0.5968071281909942, 'gradnorm/compressor.query': 0.27695268273150125, 'gradnorm/compressor.layer0': 10.529223872578962, 'gradnorm/compressor.layer1': 356.39623095299424, 'gradnorm/compressor.layer2': 73.98260128549651, 'gradnorm/compressor.layer3': 17.187246979671713, 'gradnorm/compressor.layer4': 5.923790453329771, 'gradnorm/compressor.layer5': 3.4907790379700563, 'gradnorm/compressor.layer6': 2.2349245689344652, 'gradnorm/compressor.layer7': 2.155292242993966, 'gradnorm/decoder.layer0': 2.588235807610017, 'gradnorm/decoder.layer1': 1.3504259920882085, 'gradnorm/decoder.layer2': 0.8314087737614538, 'gradnorm/max_param': 251.4860076904297, 'epoch': 7.0}

 70%|███████   | 13370/19090 [6:31:37<3:04:47,  1.94s/it]
 70%|███████   | 13371/19090 [6:31:38<2:46:09,  1.74s/it]
 70%|███████   | 13372/19090 [6:31:41<3:17:50,  2.08s/it]
 70%|███████   | 13373/19090 [6:31:42<2:55:49,  1.85s/it]
 70%|███████   | 13374/19090 [6:31:44<2:40:00,  1.68s/it]
 70%|███████   | 13375/19090 [6:31:46<2:57:34,  1.86s/it]
 70%|███████   | 13376/19090 [6:31:47<2:41:37,  1.70s/it]
 70%|███████   | 13377/19090 [6:31:48<2:30:00,  1.58s/it]
 70%|███████   | 13378/19090 [6:31:51<2:45:11,  1.74s/it]
 70%|███████   | 13379/19090 [6:31:52<2:32:22,  1.60s/it][gradnorm] step 13379: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 245.6

 70%|███████   | 13380/19090 [6:31:53<2:23:34,  1.51s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 355.890869140625, 'learning_rate': 1.2436711544807916e-05, 'recon_loss': 0.001778695167740807, 'recon_static': 0.0008219486800953746, 'recon_motion': 0.0009567465662257746, 'recon_norm': 1.394116720557213, 'recon_norm_std': 0.0393257369287312, 'target_norm': 1.6170821487903595, 'recon_cos': 0.591910396516323, 'gradnorm/compressor.query': 0.3225803110386287, 'gradnorm/compressor.layer0': 9.002389802478103, 'gradnorm/compressor.layer1': 348.05094589370105, 'gradnorm/compressor.layer2': 71.49425648915792, 'gradnorm/compressor.layer3': 16.197022454048557, 'gradnorm/compressor.layer4': 5.940844106070025, 'gradnorm/compressor.layer5': 3.271275937856157, 'gradnorm/compressor.layer6': 2.311890289759344, 'gradnorm/compressor.layer7': 2.1382100760698743, 'gradnorm/decoder.layer0': 2.3229312183511652, 'gradnorm/decoder.layer1': 1.413961965454047, 'gradnorm/decoder.layer2': 0.7779552795506586, 'gradnorm/max_param': 245.56451416015625, 'epoch': 7.01}

 70%|███████   | 13380/19090 [6:31:53<2:23:34,  1.51s/it]
 70%|███████   | 13381/19090 [6:31:55<2:38:40,  1.67s/it]
 70%|███████   | 13382/19090 [6:31:56<2:27:45,  1.55s/it]
 70%|███████   | 13383/19090 [6:31:58<2:20:12,  1.47s/it]
 70%|███████   | 13384/19090 [6:32:00<2:42:46,  1.71s/it]
 70%|███████   | 13385/19090 [6:32:01<2:30:38,  1.58s/it]
 70%|███████   | 13386/19090 [6:32:03<2:22:09,  1.50s/it]
 70%|███████   | 13387/19090 [6:32:05<2:49:59,  1.79s/it]
 70%|███████   | 13388/19090 [6:32:06<2:35:33,  1.64s/it]
 70%|███████   | 13389/19090 [6:32:08<2:25:39,  1.53s/it][gradnorm] step 13389: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 290.3

 70%|███████   | 13390/19090 [6:32:09<2:34:54,  1.63s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 419.3336486816406, 'learning_rate': 1.2397210833081643e-05, 'recon_loss': 0.0017480494221672415, 'recon_static': 0.0007701510097831487, 'recon_motion': 0.0009778983774594962, 'recon_norm': 1.3742843300104142, 'recon_norm_std': 0.038546105939894916, 'target_norm': 1.62193643450737, 'recon_cos': 0.5977623611688614, 'gradnorm/compressor.query': 0.2670683215743461, 'gradnorm/compressor.layer0': 10.20617050822056, 'gradnorm/compressor.layer1': 411.50533486821286, 'gradnorm/compressor.layer2': 77.49099935133596, 'gradnorm/compressor.layer3': 18.004951643684198, 'gradnorm/compressor.layer4': 5.952238081897745, 'gradnorm/compressor.layer5': 3.894376608208937, 'gradnorm/compressor.layer6': 2.4429065702532156, 'gradnorm/compressor.layer7': 2.4441673839458353, 'gradnorm/decoder.layer0': 2.396120723128513, 'gradnorm/decoder.layer1': 1.3820082486288783, 'gradnorm/decoder.layer2': 0.7724215452853858, 'gradnorm/max_param': 290.3439636230469, 'epoch': 7.01}

 70%|███████   | 13390/19090 [6:32:09<2:34:54,  1.63s/it]
 70%|███████   | 13391/19090 [6:32:11<2:25:06,  1.53s/it]
 70%|███████   | 13392/19090 [6:32:12<2:18:09,  1.45s/it]
 70%|███████   | 13393/19090 [6:32:14<2:41:32,  1.70s/it]
 70%|███████   | 13394/19090 [6:32:16<2:29:47,  1.58s/it]
 70%|███████   | 13395/19090 [6:32:17<2:21:26,  1.49s/it]
 70%|███████   | 13396/19090 [6:32:19<2:48:18,  1.77s/it]
 70%|███████   | 13397/19090 [6:32:21<2:34:27,  1.63s/it]
 70%|███████   | 13398/19090 [6:32:22<2:24:50,  1.53s/it]
 70%|███████   | 13399/19090 [6:32:24<2:45:38,  1.75s/it][gradnorm] step 13399: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 238.4

 70%|███████   | 13400/19090 [6:32:25<2:32:32,  1.61s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 344.0155944824219, 'learning_rate': 1.2357752258918828e-05, 'recon_loss': 0.0018194204400060698, 'recon_static': 0.0008463910286081955, 'recon_motion': 0.0009730293706525118, 'recon_norm': 1.4071940243244172, 'recon_norm_std': 0.03886263780295849, 'target_norm': 1.6213530123233795, 'recon_cos': 0.5861711218953133, 'gradnorm/compressor.query': 0.25179720645116943, 'gradnorm/compressor.layer0': 8.062476816070108, 'gradnorm/compressor.layer1': 337.86749667954155, 'gradnorm/compressor.layer2': 61.123444270207195, 'gradnorm/compressor.layer3': 18.324938615016986, 'gradnorm/compressor.layer4': 5.472698421166034, 'gradnorm/compressor.layer5': 3.22334258198192, 'gradnorm/compressor.layer6': 2.076941475827485, 'gradnorm/compressor.layer7': 1.830091673842589, 'gradnorm/decoder.layer0': 2.240715686872878, 'gradnorm/decoder.layer1': 1.3138633640642534, 'gradnorm/decoder.layer2': 0.7523282787281088, 'gradnorm/max_param': 238.39535522460938, 'epoch': 7.02}

 70%|███████   | 13400/19090 [6:32:25<2:32:32,  1.61s/it]
 70%|███████   | 13401/19090 [6:32:27<2:23:27,  1.51s/it]
 70%|███████   | 13402/19090 [6:32:29<2:57:24,  1.87s/it]
 70%|███████   | 13403/19090 [6:32:31<2:40:46,  1.70s/it]
 70%|███████   | 13404/19090 [6:32:32<2:29:07,  1.57s/it]
 70%|███████   | 13405/19090 [6:32:35<2:55:18,  1.85s/it]
 70%|███████   | 13406/19090 [6:32:36<2:39:15,  1.68s/it]
 70%|███████   | 13407/19090 [6:32:37<2:27:55,  1.56s/it]
 70%|███████   | 13408/19090 [6:32:39<2:50:35,  1.80s/it]
 70%|███████   | 13409/19090 [6:32:41<2:35:53,  1.65s/it][gradnorm] step 13409: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 282.7

 70%|███████   | 13410/19090 [6:32:42<2:25:55,  1.54s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 407.26165771484375, 'learning_rate': 1.2318335954249669e-05, 'recon_loss': 0.0018278830480994656, 'recon_static': 0.0008542211959138513, 'recon_motion': 0.0009736618550959975, 'recon_norm': 1.4037445724010467, 'recon_norm_std': 0.038517984095960855, 'target_norm': 1.6219799399375916, 'recon_cos': 0.5840077832341194, 'gradnorm/compressor.query': 0.3152649747610488, 'gradnorm/compressor.layer0': 10.80626944368642, 'gradnorm/compressor.layer1': 400.6848629222222, 'gradnorm/compressor.layer2': 68.32814822489887, 'gradnorm/compressor.layer3': 21.504572892141745, 'gradnorm/compressor.layer4': 5.853812789629114, 'gradnorm/compressor.layer5': 3.6913800273952875, 'gradnorm/compressor.layer6': 2.000969055965971, 'gradnorm/compressor.layer7': 2.1832144037522894, 'gradnorm/decoder.layer0': 2.4740923286423593, 'gradnorm/decoder.layer1': 1.4463227736248854, 'gradnorm/decoder.layer2': 0.8884464187523271, 'gradnorm/max_param': 282.73980712890625, 'epoch': 7.02}

 70%|███████   | 13410/19090 [6:32:42<2:25:55,  1.54s/it]
 70%|███████   | 13411/19090 [6:32:44<2:42:56,  1.72s/it]
 70%|███████   | 13412/19090 [6:32:46<2:31:11,  1.60s/it]
 70%|███████   | 13413/19090 [6:32:47<2:22:29,  1.51s/it]
 70%|███████   | 13414/19090 [6:32:49<2:51:42,  1.82s/it]
 70%|███████   | 13415/19090 [6:32:51<2:36:52,  1.66s/it]
 70%|███████   | 13416/19090 [6:32:52<2:26:25,  1.55s/it]
 70%|███████   | 13417/19090 [6:32:55<2:59:51,  1.90s/it]
 70%|███████   | 13418/19090 [6:32:56<2:42:20,  1.72s/it]
 70%|███████   | 13419/19090 [6:32:57<2:30:12,  1.59s/it][gradnorm] step 13419: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 309.4

 70%|███████   | 13420/19090 [6:33:00<2:51:03,  1.81s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 445.67144775390625, 'learning_rate': 1.2278962050863007e-05, 'recon_loss': 0.0017685944330878556, 'recon_static': 0.0007980930473422632, 'recon_motion': 0.0009705014352221041, 'recon_norm': 1.385549196600914, 'recon_norm_std': 0.03872768515720963, 'target_norm': 1.6194970279932022, 'recon_cos': 0.5938719749450684, 'gradnorm/compressor.query': 0.2612320801872392, 'gradnorm/compressor.layer0': 10.355906002061586, 'gradnorm/compressor.layer1': 438.4674125576495, 'gradnorm/compressor.layer2': 76.04353129848555, 'gradnorm/compressor.layer3': 20.18529807916945, 'gradnorm/compressor.layer4': 6.443948127372048, 'gradnorm/compressor.layer5': 3.3935343131132982, 'gradnorm/compressor.layer6': 2.1277379741326614, 'gradnorm/compressor.layer7': 2.3268686271209296, 'gradnorm/decoder.layer0': 2.5506491751017077, 'gradnorm/decoder.layer1': 1.3904602724728443, 'gradnorm/decoder.layer2': 0.7356474383747565, 'gradnorm/max_param': 309.39837646484375, 'epoch': 7.03}

 70%|███████   | 13420/19090 [6:33:00<2:51:03,  1.81s/it]
 70%|███████   | 13421/19090 [6:33:01<2:36:17,  1.65s/it]
 70%|███████   | 13422/19090 [6:33:02<2:26:06,  1.55s/it]
 70%|███████   | 13423/19090 [6:33:05<2:58:41,  1.89s/it]
 70%|███████   | 13424/19090 [6:33:06<2:41:35,  1.71s/it]
 70%|███████   | 13425/19090 [6:33:07<2:29:43,  1.59s/it]
 70%|███████   | 13426/19090 [6:33:09<2:34:55,  1.64s/it]
 70%|███████   | 13427/19090 [6:33:10<2:24:55,  1.54s/it]
 70%|███████   | 13428/19090 [6:33:12<2:18:42,  1.47s/it]
 70%|███████   | 13429/19090 [6:33:14<2:39:18,  1.69s/it][gradnorm] step 13429: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 256.5

 70%|███████   | 13430/19090 [6:33:15<2:28:06,  1.57s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 371.090087890625, 'learning_rate': 1.2239630680405894e-05, 'recon_loss': 0.0016611716360785066, 'recon_static': 0.0006994892304646782, 'recon_motion': 0.0009616824216209352, 'recon_norm': 1.3475941479206086, 'recon_norm_std': 0.03969790516421199, 'target_norm': 1.6191641956567764, 'recon_cos': 0.613521172106266, 'gradnorm/compressor.query': 0.20107929173314168, 'gradnorm/compressor.layer0': 7.403689428722846, 'gradnorm/compressor.layer1': 363.43435617728824, 'gradnorm/compressor.layer2': 71.70656214522839, 'gradnorm/compressor.layer3': 19.030776144808918, 'gradnorm/compressor.layer4': 6.102392084088764, 'gradnorm/compressor.layer5': 3.307161888319681, 'gradnorm/compressor.layer6': 1.9736743075139076, 'gradnorm/compressor.layer7': 2.2271452559405165, 'gradnorm/decoder.layer0': 2.2739934601970697, 'gradnorm/decoder.layer1': 1.308479509064192, 'gradnorm/decoder.layer2': 0.7181282256529552, 'gradnorm/max_param': 256.4819030761719, 'epoch': 7.04}

 70%|███████   | 13430/19090 [6:33:15<2:28:06,  1.57s/it]
 70%|███████   | 13431/19090 [6:33:17<2:20:15,  1.49s/it]
 70%|███████   | 13432/19090 [6:33:19<2:51:35,  1.82s/it]
 70%|███████   | 13433/19090 [6:33:20<2:36:29,  1.66s/it]
 70%|███████   | 13434/19090 [6:33:22<2:25:53,  1.55s/it]
 70%|███████   | 13435/19090 [6:33:24<2:43:13,  1.73s/it]
 70%|███████   | 13436/19090 [6:33:25<2:30:43,  1.60s/it]
 70%|███████   | 13437/19090 [6:33:26<2:22:11,  1.51s/it]
 70%|███████   | 13438/19090 [6:33:29<2:38:59,  1.69s/it]
 70%|███████   | 13439/19090 [6:33:30<2:27:53,  1.57s/it][gradnorm] step 13439: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 389.2

 70%|███████   | 13440/19090 [6:33:31<2:19:55,  1.49s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 558.6715698242188, 'learning_rate': 1.2200341974383226e-05, 'recon_loss': 0.0016518020333023743, 'recon_static': 0.0006783620992791839, 'recon_motion': 0.0009734400096931495, 'recon_norm': 1.329833760857582, 'recon_norm_std': 0.03962419722229242, 'target_norm': 1.6229955166578294, 'recon_cos': 0.6151440724730491, 'gradnorm/compressor.query': 0.3438673577681448, 'gradnorm/compressor.layer0': 13.073245360762389, 'gradnorm/compressor.layer1': 551.6321237926595, 'gradnorm/compressor.layer2': 84.03079095263236, 'gradnorm/compressor.layer3': 22.803164560315828, 'gradnorm/compressor.layer4': 5.962269725391524, 'gradnorm/compressor.layer5': 3.156990493175268, 'gradnorm/compressor.layer6': 2.062675353023106, 'gradnorm/compressor.layer7': 2.29865170225611, 'gradnorm/decoder.layer0': 2.4659607014490943, 'gradnorm/decoder.layer1': 1.3980839031865815, 'gradnorm/decoder.layer2': 0.8111117681046297, 'gradnorm/max_param': 389.2161560058594, 'epoch': 7.04}

 70%|███████   | 13440/19090 [6:33:31<2:19:55,  1.49s/it]
 70%|███████   | 13441/19090 [6:33:33<2:34:43,  1.64s/it]
 70%|███████   | 13442/19090 [6:33:34<2:24:44,  1.54s/it]
 70%|███████   | 13443/19090 [6:33:36<2:17:46,  1.46s/it]
 70%|███████   | 13444/19090 [6:33:38<2:47:37,  1.78s/it]
 70%|███████   | 13445/19090 [6:33:40<2:33:46,  1.63s/it]
 70%|███████   | 13446/19090 [6:33:41<2:23:59,  1.53s/it]
 70%|███████   | 13447/19090 [6:33:43<2:45:28,  1.76s/it]
 70%|███████   | 13448/19090 [6:33:44<2:32:15,  1.62s/it]
 70%|███████   | 13449/19090 [6:33:46<2:23:09,  1.52s/it][gradnorm] step 13449: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 290.5

 70%|███████   | 13450/19090 [6:33:48<2:54:11,  1.85s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 419.4102478027344, 'learning_rate': 1.2161096064157218e-05, 'recon_loss': 0.0016640711372019723, 'recon_static': 0.0006974175295908936, 'recon_motion': 0.0009666536367149092, 'recon_norm': 1.3459057748317718, 'recon_norm_std': 0.03964829277247191, 'target_norm': 1.6205092787742614, 'recon_cos': 0.6136256098747254, 'gradnorm/compressor.query': 0.26777572018730667, 'gradnorm/compressor.layer0': 10.353959038620808, 'gradnorm/compressor.layer1': 411.6578524697422, 'gradnorm/compressor.layer2': 76.27264328238041, 'gradnorm/compressor.layer3': 21.42707985182681, 'gradnorm/compressor.layer4': 5.116122861578446, 'gradnorm/compressor.layer5': 3.2140103600648975, 'gradnorm/compressor.layer6': 2.4192831629728757, 'gradnorm/compressor.layer7': 2.588540141810704, 'gradnorm/decoder.layer0': 2.680080373896812, 'gradnorm/decoder.layer1': 1.3614207691866262, 'gradnorm/decoder.layer2': 0.8544731443190522, 'gradnorm/max_param': 290.4769592285156, 'epoch': 7.05}

 70%|███████   | 13450/19090 [6:33:48<2:54:11,  1.85s/it]
 70%|███████   | 13451/19090 [6:33:50<2:38:21,  1.68s/it]
 70%|███████   | 13452/19090 [6:33:51<2:27:19,  1.57s/it]
 70%|███████   | 13453/19090 [6:33:53<2:47:35,  1.78s/it]
 70%|███████   | 13454/19090 [6:33:55<2:33:47,  1.64s/it]
 70%|███████   | 13455/19090 [6:33:56<2:23:59,  1.53s/it]
 70%|███████   | 13456/19090 [6:33:59<2:56:31,  1.88s/it]
 70%|███████   | 13457/19090 [6:34:00<2:39:47,  1.70s/it]
 70%|███████   | 13458/19090 [6:34:01<2:28:22,  1.58s/it]
 71%|███████   | 13459/19090 [6:34:04<2:53:41,  1.85s/it][gradnorm] step 13459: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 302.0

 71%|███████   | 13460/19090 [6:34:05<2:37:49,  1.68s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 435.1683349609375, 'learning_rate': 1.2121893080946977e-05, 'recon_loss': 0.001875409862259403, 'recon_static': 0.0009057443414349109, 'recon_motion': 0.0009696655310108326, 'recon_norm': 1.4256651490926742, 'recon_norm_std': 0.03780193608254194, 'target_norm': 1.620202773809433, 'recon_cos': 0.5745807841420174, 'gradnorm/compressor.query': 0.2723760575190371, 'gradnorm/compressor.layer0': 12.329298304432719, 'gradnorm/compressor.layer1': 427.964824050996, 'gradnorm/compressor.layer2': 74.7281858324722, 'gradnorm/compressor.layer3': 20.20705881196168, 'gradnorm/compressor.layer4': 6.2669968470712645, 'gradnorm/compressor.layer5': 3.533470333490328, 'gradnorm/compressor.layer6': 2.481239787736531, 'gradnorm/compressor.layer7': 2.462884139098567, 'gradnorm/decoder.layer0': 2.5428321438382557, 'gradnorm/decoder.layer1': 1.3596389338894703, 'gradnorm/decoder.layer2': 0.7615843825548183, 'gradnorm/max_param': 301.9705505371094, 'epoch': 7.05}

 71%|███████   | 13460/19090 [6:34:05<2:37:49,  1.68s/it]
 71%|███████   | 13461/19090 [6:34:06<2:26:39,  1.56s/it]
 71%|███████   | 13462/19090 [6:34:09<2:57:38,  1.89s/it]
 71%|███████   | 13463/19090 [6:34:10<2:40:27,  1.71s/it]
 71%|███████   | 13464/19090 [6:34:11<2:28:33,  1.58s/it]
 71%|███████   | 13465/19090 [6:34:14<3:01:03,  1.93s/it]
 71%|███████   | 13466/19090 [6:34:15<2:42:51,  1.74s/it]
 71%|███████   | 13467/19090 [6:34:17<2:30:15,  1.60s/it]
 71%|███████   | 13468/19090 [6:34:19<2:54:55,  1.87s/it]
 71%|███████   | 13469/19090 [6:34:21<2:38:45,  1.69s/it][gradnorm] step 13469: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 365.1

 71%|███████   | 13470/19090 [6:34:22<2:27:22,  1.57s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 525.2266845703125, 'learning_rate': 1.208273315582813e-05, 'recon_loss': 0.0018024788674665614, 'recon_static': 0.0008496969880070537, 'recon_motion': 0.0009527818954666145, 'recon_norm': 1.405247578024864, 'recon_norm_std': 0.03855830784887075, 'target_norm': 1.6180595695972442, 'recon_cos': 0.5877909675240517, 'gradnorm/compressor.query': 0.3244123455371929, 'gradnorm/compressor.layer0': 11.61147777053485, 'gradnorm/compressor.layer1': 517.3930747023968, 'gradnorm/compressor.layer2': 86.78857267898954, 'gradnorm/compressor.layer3': 20.4182800070737, 'gradnorm/compressor.layer4': 6.655676015543196, 'gradnorm/compressor.layer5': 3.8836352207791314, 'gradnorm/compressor.layer6': 2.4521578686410135, 'gradnorm/compressor.layer7': 2.6467427370428105, 'gradnorm/decoder.layer0': 2.8626843255677987, 'gradnorm/decoder.layer1': 1.4224824551001496, 'gradnorm/decoder.layer2': 0.8602973150152644, 'gradnorm/max_param': 365.070068359375, 'epoch': 7.06}

 71%|███████   | 13470/19090 [6:34:22<2:27:22,  1.57s/it]
 71%|███████   | 13471/19090 [6:34:24<2:56:33,  1.89s/it]
 71%|███████   | 13472/19090 [6:34:26<2:39:49,  1.71s/it]
 71%|███████   | 13473/19090 [6:34:27<2:28:05,  1.58s/it]
 71%|███████   | 13474/19090 [6:34:29<2:45:13,  1.77s/it]
 71%|███████   | 13475/19090 [6:34:30<2:31:57,  1.62s/it]
 71%|███████   | 13476/19090 [6:34:32<2:22:31,  1.52s/it]
 71%|███████   | 13477/19090 [6:34:34<2:55:01,  1.87s/it]
 71%|███████   | 13478/19090 [6:34:36<2:38:29,  1.69s/it]
 71%|███████   | 13479/19090 [6:34:37<2:26:51,  1.57s/it][gradnorm] step 13479: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 264.8

 71%|███████   | 13480/19090 [6:34:40<2:52:38,  1.85s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 383.1924743652344, 'learning_rate': 1.204361641973232e-05, 'recon_loss': 0.0017652452661423013, 'recon_static': 0.0007894846668932587, 'recon_motion': 0.0009757605992490425, 'recon_norm': 1.376919189095497, 'recon_norm_std': 0.039961837138980626, 'target_norm': 1.6234603583812715, 'recon_cos': 0.5939426407217979, 'gradnorm/compressor.query': 0.3049451154628527, 'gradnorm/compressor.layer0': 12.552902579052372, 'gradnorm/compressor.layer1': 375.26759202601016, 'gradnorm/compressor.layer2': 73.07283796622039, 'gradnorm/compressor.layer3': 21.02776426815717, 'gradnorm/compressor.layer4': 6.070283906886488, 'gradnorm/compressor.layer5': 3.3760933694447766, 'gradnorm/compressor.layer6': 2.5267219073246556, 'gradnorm/compressor.layer7': 2.5075731965446297, 'gradnorm/decoder.layer0': 2.6131267872594175, 'gradnorm/decoder.layer1': 1.6409861575838212, 'gradnorm/decoder.layer2': 0.8776250699022464, 'gradnorm/max_param': 264.82421875, 'epoch': 7.06}

 71%|███████   | 13480/19090 [6:34:40<2:52:38,  1.85s/it]
 71%|███████   | 13481/19090 [6:34:41<2:37:09,  1.68s/it]
 71%|███████   | 13482/19090 [6:34:42<2:26:09,  1.56s/it]
 71%|███████   | 13483/19090 [6:34:45<2:53:39,  1.86s/it]
 71%|███████   | 13484/19090 [6:34:46<2:37:57,  1.69s/it]
 71%|███████   | 13485/19090 [6:34:47<2:26:52,  1.57s/it]
 71%|███████   | 13486/19090 [6:34:50<2:50:10,  1.82s/it]
 71%|███████   | 13487/19090 [6:34:51<2:35:34,  1.67s/it]
 71%|███████   | 13488/19090 [6:34:52<2:25:05,  1.55s/it]
 71%|███████   | 13489/19090 [6:34:55<2:45:24,  1.77s/it][gradnorm] step 13489: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 255.9

 71%|███████   | 13490/19090 [6:34:56<2:31:59,  1.63s/it]
                                                         
{'loss': 0.0021, 'grad_norm': 372.2926940917969, 'learning_rate': 1.2004543003446747e-05, 'recon_loss': 0.002061491782660596, 'recon_static': 0.0010972047937684692, 'recon_motion': 0.0009642869845265523, 'recon_norm': 1.505466839671135, 'recon_norm_std': 0.03658924456685782, 'target_norm': 1.6183282345533372, 'recon_cos': 0.5467919290065766, 'gradnorm/compressor.query': 0.3117939841514701, 'gradnorm/compressor.layer0': 10.854186035451368, 'gradnorm/compressor.layer1': 362.6938705713125, 'gradnorm/compressor.layer2': 80.13042042410454, 'gradnorm/compressor.layer3': 21.248397228158762, 'gradnorm/compressor.layer4': 5.686146334015326, 'gradnorm/compressor.layer5': 3.8019958625381647, 'gradnorm/compressor.layer6': 2.120068043930377, 'gradnorm/compressor.layer7': 2.371876117771373, 'gradnorm/decoder.layer0': 2.215923712088625, 'gradnorm/decoder.layer1': 1.5027942856739644, 'gradnorm/decoder.layer2': 0.8139047217176226, 'gradnorm/max_param': 255.946533203125, 'epoch': 7.07}

 71%|███████   | 13490/19090 [6:34:56<2:31:59,  1.63s/it]
 71%|███████   | 13491/19090 [6:34:57<2:22:45,  1.53s/it]
 71%|███████   | 13492/19090 [6:35:00<2:47:50,  1.80s/it]
 71%|███████   | 13493/19090 [6:35:01<2:33:40,  1.65s/it]
 71%|███████   | 13494/19090 [6:35:02<2:23:48,  1.54s/it]
 71%|███████   | 13495/19090 [6:35:05<2:49:37,  1.82s/it]
 71%|███████   | 13496/19090 [6:35:06<2:36:38,  1.68s/it]
 71%|███████   | 13497/19090 [6:35:07<2:25:53,  1.57s/it]
 71%|███████   | 13498/19090 [6:35:10<2:49:48,  1.82s/it]
 71%|███████   | 13499/19090 [6:35:11<2:38:15,  1.70s/it][gradnorm] step 13499: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 345.6

 71%|███████   | 13500/19090 [6:35:12<2:26:40,  1.57s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 498.8193054199219, 'learning_rate': 1.196551303761384e-05, 'recon_loss': 0.0018038080597762018, 'recon_static': 0.000838850493892096, 'recon_motion': 0.0009649574989452958, 'recon_norm': 1.398494827747345, 'recon_norm_std': 0.03889657277613878, 'target_norm': 1.620166113972664, 'recon_cos': 0.5864330336451531, 'gradnorm/compressor.query': 0.3123236039130403, 'gradnorm/compressor.layer0': 11.452753287368418, 'gradnorm/compressor.layer1': 489.7866882633704, 'gradnorm/compressor.layer2': 90.30616056902625, 'gradnorm/compressor.layer3': 23.955852594619884, 'gradnorm/compressor.layer4': 6.005870172754521, 'gradnorm/compressor.layer5': 3.7797335271410137, 'gradnorm/compressor.layer6': 2.2541391658926972, 'gradnorm/compressor.layer7': 2.3364274649778416, 'gradnorm/decoder.layer0': 2.285609221598255, 'gradnorm/decoder.layer1': 1.555857692433146, 'gradnorm/decoder.layer2': 0.8956614521645235, 'gradnorm/max_param': 345.6378479003906, 'epoch': 7.07}

 71%|███████   | 13500/19090 [6:35:12<2:26:40,  1.57s/it]
 71%|███████   | 13501/19090 [6:35:17<3:45:05,  2.42s/it]
 71%|███████   | 13502/19090 [6:35:18<3:13:36,  2.08s/it]
 71%|███████   | 13503/19090 [6:35:19<2:51:25,  1.84s/it]
 71%|███████   | 13504/19090 [6:35:21<2:35:49,  1.67s/it]
 71%|███████   | 13505/19090 [6:35:22<2:25:06,  1.56s/it]
 71%|███████   | 13506/19090 [6:35:23<2:17:25,  1.48s/it]
 71%|███████   | 13507/19090 [6:35:24<2:12:22,  1.42s/it]
 71%|███████   | 13508/19090 [6:35:26<2:14:56,  1.45s/it]
 71%|███████   | 13509/19090 [6:35:27<2:10:44,  1.41s/it][gradnorm] step 13509: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 280.9

 71%|███████   | 13510/19090 [6:35:29<2:21:46,  1.52s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 403.82757568359375, 'learning_rate': 1.1926526652730716e-05, 'recon_loss': 0.0016956269275397063, 'recon_static': 0.0007336636990658008, 'recon_motion': 0.0009619632561225444, 'recon_norm': 1.3582020998001099, 'recon_norm_std': 0.03910357290878892, 'target_norm': 1.6174024909734726, 'recon_cos': 0.6055313393473625, 'gradnorm/compressor.query': 0.3313457031372323, 'gradnorm/compressor.layer0': 13.221350447893574, 'gradnorm/compressor.layer1': 398.2117644859328, 'gradnorm/compressor.layer2': 62.54814942181543, 'gradnorm/compressor.layer3': 18.849629511341696, 'gradnorm/compressor.layer4': 5.984149659697541, 'gradnorm/compressor.layer5': 2.9497016561907863, 'gradnorm/compressor.layer6': 1.9599176923830597, 'gradnorm/compressor.layer7': 2.2114575317017198, 'gradnorm/decoder.layer0': 2.405514691450903, 'gradnorm/decoder.layer1': 1.3765949622146711, 'gradnorm/decoder.layer2': 0.8429669526004101, 'gradnorm/max_param': 280.89111328125, 'epoch': 7.08}

 71%|███████   | 13510/19090 [6:35:29<2:21:46,  1.52s/it]
 71%|███████   | 13511/19090 [6:35:31<2:35:02,  1.67s/it]
 71%|███████   | 13512/19090 [6:35:32<2:24:25,  1.55s/it]
 71%|███████   | 13513/19090 [6:35:34<2:38:53,  1.71s/it]
 71%|███████   | 13514/19090 [6:35:36<2:43:36,  1.76s/it]
 71%|███████   | 13515/19090 [6:35:38<2:30:22,  1.62s/it]
 71%|███████   | 13516/19090 [6:35:40<2:41:40,  1.74s/it]
 71%|███████   | 13517/19090 [6:35:41<2:41:51,  1.74s/it]
 71%|███████   | 13518/19090 [6:35:43<2:29:06,  1.61s/it]
 71%|███████   | 13519/19090 [6:35:45<2:51:19,  1.85s/it][gradnorm] step 13519: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 264.4

 71%|███████   | 13520/19090 [6:35:47<2:47:44,  1.81s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 381.4757080078125, 'learning_rate': 1.1887583979148748e-05, 'recon_loss': 0.0017530649638501927, 'recon_static': 0.0007796343212248758, 'recon_motion': 0.0009734306746395305, 'recon_norm': 1.380339938402176, 'recon_norm_std': 0.03843260705471039, 'target_norm': 1.6195890694856643, 'recon_cos': 0.5961686044931411, 'gradnorm/compressor.query': 0.34962254504752466, 'gradnorm/compressor.layer0': 12.941039554871246, 'gradnorm/compressor.layer1': 374.7396179349069, 'gradnorm/compressor.layer2': 66.21217965280255, 'gradnorm/compressor.layer3': 22.042670743267337, 'gradnorm/compressor.layer4': 5.191314949053713, 'gradnorm/compressor.layer5': 3.3286344469608165, 'gradnorm/compressor.layer6': 2.0982086761306866, 'gradnorm/compressor.layer7': 2.210776387197922, 'gradnorm/decoder.layer0': 2.4481120795852926, 'gradnorm/decoder.layer1': 1.4705141713621395, 'gradnorm/decoder.layer2': 0.9620813214652887, 'gradnorm/max_param': 264.4130859375, 'epoch': 7.08}

 71%|███████   | 13520/19090 [6:35:47<2:47:44,  1.81s/it]
 71%|███████   | 13521/19090 [6:35:48<2:33:13,  1.65s/it]
 71%|███████   | 13522/19090 [6:35:51<3:07:18,  2.02s/it]
 71%|███████   | 13523/19090 [6:35:52<2:46:52,  1.80s/it]
 71%|███████   | 13524/19090 [6:35:54<2:32:36,  1.65s/it]
 71%|███████   | 13525/19090 [6:35:56<2:53:19,  1.87s/it]
 71%|███████   | 13526/19090 [6:35:57<2:37:05,  1.69s/it]
 71%|███████   | 13527/19090 [6:35:58<2:25:46,  1.57s/it]
 71%|███████   | 13528/19090 [6:36:01<2:56:57,  1.91s/it]
 71%|███████   | 13529/19090 [6:36:02<2:39:35,  1.72s/it][gradnorm] step 13529: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 279.1

 71%|███████   | 13530/19090 [6:36:04<2:27:26,  1.59s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 400.1559753417969, 'learning_rate': 1.1848685147073222e-05, 'recon_loss': 0.0017961099481908604, 'recon_static': 0.0008301358262542635, 'recon_motion': 0.0009659741757786833, 'recon_norm': 1.397248339653015, 'recon_norm_std': 0.037882004678249356, 'target_norm': 1.6170467913150788, 'recon_cos': 0.5870586931705475, 'gradnorm/compressor.query': 0.33545755878503264, 'gradnorm/compressor.layer0': 11.409667456793356, 'gradnorm/compressor.layer1': 395.69426723797756, 'gradnorm/compressor.layer2': 54.57172566620466, 'gradnorm/compressor.layer3': 19.018515950899253, 'gradnorm/compressor.layer4': 6.845274872927858, 'gradnorm/compressor.layer5': 3.62172590579162, 'gradnorm/compressor.layer6': 2.2909660886854777, 'gradnorm/compressor.layer7': 2.302287983032884, 'gradnorm/decoder.layer0': 2.605457037080119, 'gradnorm/decoder.layer1': 1.5482713196363143, 'gradnorm/decoder.layer2': 1.0366170688143128, 'gradnorm/max_param': 279.0532531738281, 'epoch': 7.09}

 71%|███████   | 13530/19090 [6:36:04<2:27:26,  1.59s/it]
 71%|███████   | 13531/19090 [6:36:06<2:52:07,  1.86s/it]
 71%|███████   | 13532/19090 [6:36:08<2:36:08,  1.69s/it]
 71%|███████   | 13533/19090 [6:36:09<2:25:04,  1.57s/it]
 71%|███████   | 13534/19090 [6:36:11<2:49:28,  1.83s/it]
 71%|███████   | 13535/19090 [6:36:13<2:34:17,  1.67s/it]
 71%|███████   | 13536/19090 [6:36:14<2:23:42,  1.55s/it]
 71%|███████   | 13537/19090 [6:36:17<2:58:35,  1.93s/it]
 71%|███████   | 13538/19090 [6:36:18<2:45:07,  1.78s/it]
 71%|███████   | 13539/19090 [6:36:19<2:31:11,  1.63s/it][gradnorm] step 13539: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 283.0

 71%|███████   | 13540/19090 [6:36:22<2:59:02,  1.94s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 406.9466552734375, 'learning_rate': 1.1809830286562807e-05, 'recon_loss': 0.0017299747647484764, 'recon_static': 0.0007598315758514218, 'recon_motion': 0.0009701431947178208, 'recon_norm': 1.3721078246831895, 'recon_norm_std': 0.03895302694290877, 'target_norm': 1.6212055832147598, 'recon_cos': 0.6014555394649506, 'gradnorm/compressor.query': 0.3024278873603769, 'gradnorm/compressor.layer0': 9.877433380466146, 'gradnorm/compressor.layer1': 401.11078192314903, 'gradnorm/compressor.layer2': 64.08305310542328, 'gradnorm/compressor.layer3': 21.024608181127224, 'gradnorm/compressor.layer4': 5.922214146064435, 'gradnorm/compressor.layer5': 3.8466113845405228, 'gradnorm/compressor.layer6': 2.1278901361270983, 'gradnorm/compressor.layer7': 2.331982073752598, 'gradnorm/decoder.layer0': 2.6384379341787496, 'gradnorm/decoder.layer1': 1.37156230169483, 'gradnorm/decoder.layer2': 0.832276966186357, 'gradnorm/max_param': 283.0156555175781, 'epoch': 7.09}

 71%|███████   | 13540/19090 [6:36:22<2:59:02,  1.94s/it]
 71%|███████   | 13541/19090 [6:36:23<2:43:10,  1.76s/it]
 71%|███████   | 13542/19090 [6:36:25<2:29:47,  1.62s/it]
 71%|███████   | 13543/19090 [6:36:27<2:53:10,  1.87s/it]
 71%|███████   | 13544/19090 [6:36:29<2:41:42,  1.75s/it]
 71%|███████   | 13545/19090 [6:36:30<2:28:44,  1.61s/it]
 71%|███████   | 13546/19090 [6:36:33<3:03:14,  1.98s/it]
 71%|███████   | 13547/19090 [6:36:34<2:43:48,  1.77s/it]
 71%|███████   | 13548/19090 [6:36:35<2:30:15,  1.63s/it]
 71%|███████   | 13549/19090 [6:36:38<2:52:41,  1.87s/it][gradnorm] step 13549: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 375.1

 71%|███████   | 13550/19090 [6:36:39<2:36:26,  1.69s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 539.9653930664062, 'learning_rate': 1.1771019527529126e-05, 'recon_loss': 0.001744046524981968, 'recon_static': 0.0007755684695439413, 'recon_motion': 0.0009684780248790048, 'recon_norm': 1.379317507147789, 'recon_norm_std': 0.03953232057392597, 'target_norm': 1.61933414041996, 'recon_cos': 0.5979656457901001, 'gradnorm/compressor.query': 0.5538290661408183, 'gradnorm/compressor.layer0': 21.437691468031616, 'gradnorm/compressor.layer1': 531.7495592381812, 'gradnorm/compressor.layer2': 86.77688929447172, 'gradnorm/compressor.layer3': 27.009909612680236, 'gradnorm/compressor.layer4': 7.080159482291181, 'gradnorm/compressor.layer5': 3.8164314886096116, 'gradnorm/compressor.layer6': 2.3669653054882858, 'gradnorm/compressor.layer7': 2.4185867136168966, 'gradnorm/decoder.layer0': 2.5108143556090563, 'gradnorm/decoder.layer1': 1.491489746570031, 'gradnorm/decoder.layer2': 0.8424090219723703, 'gradnorm/max_param': 375.0926513671875, 'epoch': 7.1}

 71%|███████   | 13550/19090 [6:36:39<2:36:26,  1.69s/it]
 71%|███████   | 13551/19090 [6:36:40<2:25:09,  1.57s/it]
 71%|███████   | 13552/19090 [6:36:43<2:57:55,  1.93s/it]
 71%|███████   | 13553/19090 [6:36:44<2:40:05,  1.73s/it]
 71%|███████   | 13554/19090 [6:36:46<2:27:36,  1.60s/it]
 71%|███████   | 13555/19090 [6:36:48<3:00:36,  1.96s/it]
 71%|███████   | 13556/19090 [6:36:50<2:41:56,  1.76s/it]
 71%|███████   | 13557/19090 [6:36:51<2:28:57,  1.62s/it]
 71%|███████   | 13558/19090 [6:36:54<3:01:29,  1.97s/it]
 71%|███████   | 13559/19090 [6:36:55<2:42:36,  1.76s/it][gradnorm] step 13559: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 269.8

 71%|███████   | 13560/19090 [6:36:56<2:29:28,  1.62s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 390.2399597167969, 'learning_rate': 1.173225299973641e-05, 'recon_loss': 0.0018385290779406205, 'recon_static': 0.0008704733947524801, 'recon_motion': 0.0009680557355750352, 'recon_norm': 1.4088033616542817, 'recon_norm_std': 0.03834203518927097, 'target_norm': 1.61926212310791, 'recon_cos': 0.5801893308758735, 'gradnorm/compressor.query': 0.3220092678572591, 'gradnorm/compressor.layer0': 11.090781402989883, 'gradnorm/compressor.layer1': 382.43369523356597, 'gradnorm/compressor.layer2': 72.9002087081203, 'gradnorm/compressor.layer3': 22.713166654016455, 'gradnorm/compressor.layer4': 6.659545369253556, 'gradnorm/compressor.layer5': 3.460333846860166, 'gradnorm/compressor.layer6': 2.2597026371069697, 'gradnorm/compressor.layer7': 2.4840107776758176, 'gradnorm/decoder.layer0': 2.7465692274007116, 'gradnorm/decoder.layer1': 1.4633130837402952, 'gradnorm/decoder.layer2': 0.9778085617225072, 'gradnorm/max_param': 269.81878662109375, 'epoch': 7.1}

 71%|███████   | 13560/19090 [6:36:56<2:29:28,  1.62s/it]
 71%|███████   | 13561/19090 [6:36:59<2:59:51,  1.95s/it]
 71%|███████   | 13562/19090 [6:37:00<2:41:21,  1.75s/it]
 71%|███████   | 13563/19090 [6:37:02<2:28:25,  1.61s/it]
 71%|███████   | 13564/19090 [6:37:04<3:02:08,  1.98s/it]
 71%|███████   | 13565/19090 [6:37:06<2:42:52,  1.77s/it]
 71%|███████   | 13566/19090 [6:37:07<2:29:27,  1.62s/it]
 71%|███████   | 13567/19090 [6:37:09<2:45:48,  1.80s/it]
 71%|███████   | 13568/19090 [6:37:11<2:34:51,  1.68s/it]
 71%|███████   | 13569/19090 [6:37:12<2:30:25,  1.63s/it][gradnorm] step 13569: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 290.8

 71%|███████   | 13570/19090 [6:37:15<2:54:02,  1.89s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 418.8631286621094, 'learning_rate': 1.1693530832800962e-05, 'recon_loss': 0.0017671158711891621, 'recon_static': 0.0007857730786781758, 'recon_motion': 0.0009813427954213695, 'recon_norm': 1.3808847486972808, 'recon_norm_std': 0.03938292087987065, 'target_norm': 1.6231538861989976, 'recon_cos': 0.5941016927361489, 'gradnorm/compressor.query': 0.32509070652838223, 'gradnorm/compressor.layer0': 11.146956475732512, 'gradnorm/compressor.layer1': 412.2063727673312, 'gradnorm/compressor.layer2': 69.73030846030117, 'gradnorm/compressor.layer3': 21.65575255335967, 'gradnorm/compressor.layer4': 6.253066874495591, 'gradnorm/compressor.layer5': 4.066154127705785, 'gradnorm/compressor.layer6': 2.3180607102798167, 'gradnorm/compressor.layer7': 2.4954565724419746, 'gradnorm/decoder.layer0': 2.498228208210341, 'gradnorm/decoder.layer1': 1.4784258030324084, 'gradnorm/decoder.layer2': 0.9545482889872327, 'gradnorm/max_param': 290.8401794433594, 'epoch': 7.11}

 71%|███████   | 13570/19090 [6:37:15<2:54:02,  1.89s/it]
 71%|███████   | 13571/19090 [6:37:16<2:45:06,  1.79s/it]
 71%|███████   | 13572/19090 [6:37:18<2:30:58,  1.64s/it]
 71%|███████   | 13573/19090 [6:37:20<2:57:04,  1.93s/it]
 71%|███████   | 13574/19090 [6:37:22<2:47:22,  1.82s/it]
 71%|███████   | 13575/19090 [6:37:23<2:32:22,  1.66s/it]
 71%|███████   | 13576/19090 [6:37:26<3:05:57,  2.02s/it]
 71%|███████   | 13577/19090 [6:37:27<2:49:38,  1.85s/it]
 71%|███████   | 13578/19090 [6:37:29<2:34:05,  1.68s/it]
 71%|███████   | 13579/19090 [6:37:31<3:03:52,  2.00s/it][gradnorm] step 13579: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 422.2

 71%|███████   | 13580/19090 [6:37:33<2:44:52,  1.80s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 606.7113037109375, 'learning_rate': 1.1654853156190743e-05, 'recon_loss': 0.001744636512012221, 'recon_static': 0.0007663091193535366, 'recon_motion': 0.0009783273475477473, 'recon_norm': 1.3718710660934448, 'recon_norm_std': 0.03952690009027719, 'target_norm': 1.6210144191980362, 'recon_cos': 0.5975956857204437, 'gradnorm/compressor.query': 0.3655033039460315, 'gradnorm/compressor.layer0': 11.917084017472234, 'gradnorm/compressor.layer1': 598.3830033933116, 'gradnorm/compressor.layer2': 95.52164042279895, 'gradnorm/compressor.layer3': 26.06311941897317, 'gradnorm/compressor.layer4': 6.6947820357609595, 'gradnorm/compressor.layer5': 4.398072791690226, 'gradnorm/compressor.layer6': 2.4647815533270565, 'gradnorm/compressor.layer7': 2.8727844657845507, 'gradnorm/decoder.layer0': 2.9533195633348366, 'gradnorm/decoder.layer1': 1.5397321919016698, 'gradnorm/decoder.layer2': 0.9710087438989153, 'gradnorm/max_param': 422.2169494628906, 'epoch': 7.11}

 71%|███████   | 13580/19090 [6:37:33<2:44:52,  1.80s/it]
 71%|███████   | 13581/19090 [6:37:34<2:31:08,  1.65s/it]
 71%|███████   | 13582/19090 [6:37:36<2:54:21,  1.90s/it]
 71%|███████   | 13583/19090 [6:37:38<2:40:56,  1.75s/it]
 71%|███████   | 13584/19090 [6:37:39<2:31:54,  1.66s/it]
 71%|███████   | 13585/19090 [6:37:42<2:57:56,  1.94s/it]
 71%|███████   | 13586/19090 [6:37:43<2:44:11,  1.79s/it]
 71%|███████   | 13587/19090 [6:37:45<2:30:40,  1.64s/it]
 71%|███████   | 13588/19090 [6:37:47<2:59:59,  1.96s/it]
 71%|███████   | 13589/19090 [6:37:49<2:43:51,  1.79s/it][gradnorm] step 13589: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 345.9

 71%|███████   | 13590/19090 [6:37:50<2:30:22,  1.64s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 500.18255615234375, 'learning_rate': 1.161622009922502e-05, 'recon_loss': 0.0016403635323513298, 'recon_static': 0.0006696596712572501, 'recon_motion': 0.0009707038654596545, 'recon_norm': 1.3343349009752274, 'recon_norm_std': 0.039916053507477046, 'target_norm': 1.618761995434761, 'recon_cos': 0.6171335399150848, 'gradnorm/compressor.query': 0.26387396238313665, 'gradnorm/compressor.layer0': 10.3528647470968, 'gradnorm/compressor.layer1': 490.1279647959296, 'gradnorm/compressor.layer2': 95.0660954283913, 'gradnorm/compressor.layer3': 26.763078861310383, 'gradnorm/compressor.layer4': 7.157951319326204, 'gradnorm/compressor.layer5': 4.2613421167928465, 'gradnorm/compressor.layer6': 2.712273612001987, 'gradnorm/compressor.layer7': 2.8806355845281106, 'gradnorm/decoder.layer0': 2.7154749294381686, 'gradnorm/decoder.layer1': 1.61050893594086, 'gradnorm/decoder.layer2': 1.0316660648605078, 'gradnorm/max_param': 345.85833740234375, 'epoch': 7.12}

 71%|███████   | 13590/19090 [6:37:50<2:30:22,  1.64s/it]
 71%|███████   | 13591/19090 [6:37:52<2:53:48,  1.90s/it]
 71%|███████   | 13592/19090 [6:37:54<2:44:42,  1.80s/it]
 71%|███████   | 13593/19090 [6:37:55<2:30:36,  1.64s/it]
 71%|███████   | 13594/19090 [6:37:58<2:55:59,  1.92s/it]
 71%|███████   | 13595/19090 [6:38:00<2:50:30,  1.86s/it]
 71%|███████   | 13596/19090 [6:38:01<2:34:55,  1.69s/it]
 71%|███████   | 13597/19090 [6:38:03<2:56:25,  1.93s/it]
 71%|███████   | 13598/19090 [6:38:05<2:45:18,  1.81s/it]
 71%|███████   | 13599/19090 [6:38:06<2:30:57,  1.65s/it][gradnorm] step 13599: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 268.2

 71%|███████   | 13600/19090 [6:38:09<3:06:01,  2.03s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 390.978515625, 'learning_rate': 1.1577631791073843e-05, 'recon_loss': 0.001774422632297501, 'recon_static': 0.000796432928473223, 'recon_motion': 0.0009779897474800237, 'recon_norm': 1.3816220045089722, 'recon_norm_std': 0.04030970651656389, 'target_norm': 1.6230108618736268, 'recon_cos': 0.5947267800569535, 'gradnorm/compressor.query': 0.27812948544775024, 'gradnorm/compressor.layer0': 11.280658569049276, 'gradnorm/compressor.layer1': 380.1980158417453, 'gradnorm/compressor.layer2': 87.91799157275773, 'gradnorm/compressor.layer3': 19.431921646029295, 'gradnorm/compressor.layer4': 6.629324023738676, 'gradnorm/compressor.layer5': 3.8408954224079084, 'gradnorm/compressor.layer6': 2.1850765427859162, 'gradnorm/compressor.layer7': 2.4150402471691392, 'gradnorm/decoder.layer0': 2.5919135152964436, 'gradnorm/decoder.layer1': 1.5149381201919978, 'gradnorm/decoder.layer2': 0.9182829623476569, 'gradnorm/max_param': 268.2341003417969, 'epoch': 7.12}

 71%|███████   | 13600/19090 [6:38:09<3:06:01,  2.03s/it]
 71%|███████   | 13601/19090 [6:38:10<2:45:30,  1.81s/it]
 71%|███████▏  | 13602/19090 [6:38:12<2:30:59,  1.65s/it]
 71%|███████▏  | 13603/19090 [6:38:14<2:59:29,  1.96s/it]
 71%|███████▏  | 13604/19090 [6:38:16<2:41:39,  1.77s/it]
 71%|███████▏  | 13605/19090 [6:38:17<2:28:23,  1.62s/it]
 71%|███████▏  | 13606/19090 [6:38:20<3:08:21,  2.06s/it]
 71%|███████▏  | 13607/19090 [6:38:21<2:47:01,  1.83s/it]
 71%|███████▏  | 13608/19090 [6:38:23<2:32:14,  1.67s/it]
 71%|███████▏  | 13609/19090 [6:38:25<2:54:31,  1.91s/it][gradnorm] step 13609: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 379.1

 71%|███████▏  | 13610/19090 [6:38:26<2:37:25,  1.72s/it]
                                                         
{'loss': 0.0024, 'grad_norm': 544.7113037109375, 'learning_rate': 1.1539088360757609e-05, 'recon_loss': 0.002360154839698225, 'recon_static': 0.0013957181596197187, 'recon_motion': 0.0009644367542932742, 'recon_norm': 1.6097501248121262, 'recon_norm_std': 0.03489597588777542, 'target_norm': 1.6197396486997604, 'recon_cos': 0.5114952556788921, 'gradnorm/compressor.query': 0.4238618230827403, 'gradnorm/compressor.layer0': 16.942202882901345, 'gradnorm/compressor.layer1': 537.2870217583894, 'gradnorm/compressor.layer2': 84.00041271289982, 'gradnorm/compressor.layer3': 24.488429237170486, 'gradnorm/compressor.layer4': 7.424536940395466, 'gradnorm/compressor.layer5': 3.9395809427082966, 'gradnorm/compressor.layer6': 2.146680522700229, 'gradnorm/compressor.layer7': 2.1707218319696215, 'gradnorm/decoder.layer0': 2.577805747870635, 'gradnorm/decoder.layer1': 1.533444991745498, 'gradnorm/decoder.layer2': 1.0741831825610382, 'gradnorm/max_param': 379.063720703125, 'epoch': 7.13}

 71%|███████▏  | 13610/19090 [6:38:26<2:37:25,  1.72s/it]
 71%|███████▏  | 13611/19090 [6:38:28<2:25:32,  1.59s/it]
 71%|███████▏  | 13612/19090 [6:38:30<2:54:42,  1.91s/it]
 71%|███████▏  | 13613/19090 [6:38:32<2:37:28,  1.73s/it]
 71%|███████▏  | 13614/19090 [6:38:33<2:25:28,  1.59s/it]
 71%|███████▏  | 13615/19090 [6:38:36<2:54:08,  1.91s/it]
 71%|███████▏  | 13616/19090 [6:38:37<2:42:37,  1.78s/it]
 71%|███████▏  | 13617/19090 [6:38:38<2:28:59,  1.63s/it]
 71%|███████▏  | 13618/19090 [6:38:41<2:54:04,  1.91s/it]
 71%|███████▏  | 13619/19090 [6:38:42<2:38:30,  1.74s/it][gradnorm] step 13619: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 222.2

 71%|███████▏  | 13620/19090 [6:38:44<2:28:56,  1.63s/it]
                                                         
{'loss': 0.002, 'grad_norm': 321.1825866699219, 'learning_rate': 1.1500589937146727e-05, 'recon_loss': 0.001963709178380668, 'recon_static': 0.0009872395909042097, 'recon_motion': 0.0009764695394551381, 'recon_norm': 1.4597848117351533, 'recon_norm_std': 0.037673348747193816, 'target_norm': 1.623176211118698, 'recon_cos': 0.5622374549508095, 'gradnorm/compressor.query': 0.21776890429239024, 'gradnorm/compressor.layer0': 8.492757516661854, 'gradnorm/compressor.layer1': 314.93610463076476, 'gradnorm/compressor.layer2': 59.98659314731813, 'gradnorm/compressor.layer3': 15.549913824400734, 'gradnorm/compressor.layer4': 6.106522072047742, 'gradnorm/compressor.layer5': 2.685430927894749, 'gradnorm/compressor.layer6': 1.9489828513591307, 'gradnorm/compressor.layer7': 2.1659895376796365, 'gradnorm/decoder.layer0': 2.38066589720938, 'gradnorm/decoder.layer1': 1.348407480694738, 'gradnorm/decoder.layer2': 0.8121698515714962, 'gradnorm/max_param': 222.224853515625, 'epoch': 7.13}

 71%|███████▏  | 13620/19090 [6:38:44<2:28:56,  1.63s/it]
 71%|███████▏  | 13621/19090 [6:38:46<2:44:20,  1.80s/it]
 71%|███████▏  | 13622/19090 [6:38:47<2:36:40,  1.72s/it]
 71%|███████▏  | 13623/19090 [6:38:49<2:29:02,  1.64s/it]
 71%|███████▏  | 13624/19090 [6:38:52<3:01:31,  1.99s/it]
 71%|███████▏  | 13625/19090 [6:38:53<2:46:13,  1.82s/it]
 71%|███████▏  | 13626/19090 [6:38:54<2:31:34,  1.66s/it]
 71%|███████▏  | 13627/19090 [6:38:57<2:54:57,  1.92s/it]
 71%|███████▏  | 13628/19090 [6:38:59<2:48:45,  1.85s/it]
 71%|███████▏  | 13629/19090 [6:39:00<2:33:35,  1.69s/it][gradnorm] step 13629: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 285.1

 71%|███████▏  | 13630/19090 [6:39:02<2:49:02,  1.86s/it]
                                                         
{'loss': 0.002, 'grad_norm': 411.8660583496094, 'learning_rate': 1.1462136648961103e-05, 'recon_loss': 0.0019661088881548494, 'recon_static': 0.0009940413001459092, 'recon_motion': 0.0009720675545395352, 'recon_norm': 1.4648285627365112, 'recon_norm_std': 0.03709839172661304, 'target_norm': 1.620710274577141, 'recon_cos': 0.5631702110171318, 'gradnorm/compressor.query': 0.3333374584459323, 'gradnorm/compressor.layer0': 10.41446762269958, 'gradnorm/compressor.layer1': 403.997254012945, 'gradnorm/compressor.layer2': 76.9282145595002, 'gradnorm/compressor.layer3': 18.528106783122684, 'gradnorm/compressor.layer4': 4.945100366945059, 'gradnorm/compressor.layer5': 2.951164927244055, 'gradnorm/compressor.layer6': 2.0440401437105735, 'gradnorm/compressor.layer7': 2.082786437413128, 'gradnorm/decoder.layer0': 2.3410162686544878, 'gradnorm/decoder.layer1': 1.459433367785211, 'gradnorm/decoder.layer2': 0.8750935572607068, 'gradnorm/max_param': 285.1091613769531, 'epoch': 7.14}

 71%|███████▏  | 13630/19090 [6:39:02<2:49:02,  1.86s/it]
 71%|███████▏  | 13631/19090 [6:39:04<2:56:44,  1.94s/it]
 71%|███████▏  | 13632/19090 [6:39:06<2:38:40,  1.74s/it]
 71%|███████▏  | 13633/19090 [6:39:07<2:38:20,  1.74s/it]
 71%|███████▏  | 13634/19090 [6:39:10<2:57:34,  1.95s/it]
 71%|███████▏  | 13635/19090 [6:39:11<2:39:20,  1.75s/it]
 71%|███████▏  | 13636/19090 [6:39:13<2:35:50,  1.71s/it]
 71%|███████▏  | 13637/19090 [6:39:15<2:47:31,  1.84s/it]
 71%|███████▏  | 13638/19090 [6:39:16<2:32:11,  1.67s/it]
 71%|███████▏  | 13639/19090 [6:39:18<2:41:08,  1.77s/it][gradnorm] step 13639: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 259.2

 71%|███████▏  | 13640/19090 [6:39:20<2:37:48,  1.74s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 373.84234619140625, 'learning_rate': 1.1423728624769695e-05, 'recon_loss': 0.0017269743955694139, 'recon_static': 0.0007642898955964484, 'recon_motion': 0.000962684529076796, 'recon_norm': 1.3715960532426834, 'recon_norm_std': 0.03968776548281312, 'target_norm': 1.6189513564109803, 'recon_cos': 0.6011289224028588, 'gradnorm/compressor.query': 0.2532874710335635, 'gradnorm/compressor.layer0': 10.48978544846761, 'gradnorm/compressor.layer1': 367.3759144255377, 'gradnorm/compressor.layer2': 63.595828820445256, 'gradnorm/compressor.layer3': 24.020532090569155, 'gradnorm/compressor.layer4': 6.048625052088446, 'gradnorm/compressor.layer5': 2.9056307225161775, 'gradnorm/compressor.layer6': 1.9096110995551396, 'gradnorm/compressor.layer7': 2.1814903542206783, 'gradnorm/decoder.layer0': 2.3315265563067626, 'gradnorm/decoder.layer1': 1.449151461702281, 'gradnorm/decoder.layer2': 0.7492563614140866, 'gradnorm/max_param': 259.23175048828125, 'epoch': 7.15}

 71%|███████▏  | 13640/19090 [6:39:20<2:37:48,  1.74s/it]
 71%|███████▏  | 13641/19090 [6:39:21<2:25:35,  1.60s/it]
 71%|███████▏  | 13642/19090 [6:39:23<2:42:29,  1.79s/it]
 71%|███████▏  | 13643/19090 [6:39:25<2:42:21,  1.79s/it]
 71%|███████▏  | 13644/19090 [6:39:26<2:28:48,  1.64s/it]
 71%|███████▏  | 13645/19090 [6:39:29<2:52:06,  1.90s/it]
 71%|███████▏  | 13646/19090 [6:39:30<2:37:53,  1.74s/it]
 71%|███████▏  | 13647/19090 [6:39:31<2:25:35,  1.60s/it]
 71%|███████▏  | 13648/19090 [6:39:35<3:05:51,  2.05s/it]
 71%|███████▏  | 13649/19090 [6:39:36<2:45:05,  1.82s/it][gradnorm] step 13649: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 372.8

 72%|███████▏  | 13650/19090 [6:39:37<2:30:49,  1.66s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 535.921630859375, 'learning_rate': 1.138536599299018e-05, 'recon_loss': 0.0018587533937534317, 'recon_static': 0.0008900911212549545, 'recon_motion': 0.0009686622404842637, 'recon_norm': 1.4237423568964005, 'recon_norm_std': 0.038819039985537526, 'target_norm': 1.6215450465679169, 'recon_cos': 0.5791857540607452, 'gradnorm/compressor.query': 0.31926607540167734, 'gradnorm/compressor.layer0': 13.702560987359432, 'gradnorm/compressor.layer1': 528.2761765875875, 'gradnorm/compressor.layer2': 85.7788677947911, 'gradnorm/compressor.layer3': 23.051671905454466, 'gradnorm/compressor.layer4': 5.530701173612778, 'gradnorm/compressor.layer5': 3.1621550662136673, 'gradnorm/compressor.layer6': 2.090501457179177, 'gradnorm/compressor.layer7': 2.2411463642998863, 'gradnorm/decoder.layer0': 2.513281067862132, 'gradnorm/decoder.layer1': 1.4054722224516538, 'gradnorm/decoder.layer2': 0.8446194090162051, 'gradnorm/max_param': 372.77117919921875, 'epoch': 7.15}

 72%|███████▏  | 13650/19090 [6:39:37<2:30:49,  1.66s/it]
 72%|███████▏  | 13651/19090 [6:39:40<2:56:28,  1.95s/it]
 72%|███████▏  | 13652/19090 [6:39:41<2:40:29,  1.77s/it]
 72%|███████▏  | 13653/19090 [6:39:42<2:27:12,  1.62s/it]
 72%|███████▏  | 13654/19090 [6:39:45<2:54:42,  1.93s/it]
 72%|███████▏  | 13655/19090 [6:39:46<2:37:09,  1.73s/it]
 72%|███████▏  | 13656/19090 [6:39:48<2:26:50,  1.62s/it]
 72%|███████▏  | 13657/19090 [6:39:50<2:55:30,  1.94s/it]
 72%|███████▏  | 13658/19090 [6:39:52<2:37:38,  1.74s/it]
 72%|███████▏  | 13659/19090 [6:39:53<2:28:14,  1.64s/it][gradnorm] step 13659: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 326.3

 72%|███████▏  | 13660/19090 [6:39:56<2:56:33,  1.95s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 471.2474670410156, 'learning_rate': 1.1347048881888436e-05, 'recon_loss': 0.0017141771415481344, 'recon_static': 0.0007484222078346647, 'recon_motion': 0.0009657549657276831, 'recon_norm': 1.3601626813411714, 'recon_norm_std': 0.03968472182750702, 'target_norm': 1.6195935904979706, 'recon_cos': 0.6023479893803596, 'gradnorm/compressor.query': 0.29216217910474135, 'gradnorm/compressor.layer0': 11.585489708136802, 'gradnorm/compressor.layer1': 462.5176685814284, 'gradnorm/compressor.layer2': 86.25904334383584, 'gradnorm/compressor.layer3': 22.720821238974562, 'gradnorm/compressor.layer4': 5.82537040351382, 'gradnorm/compressor.layer5': 3.11035846128552, 'gradnorm/compressor.layer6': 2.1114203814906847, 'gradnorm/compressor.layer7': 2.0419973248381096, 'gradnorm/decoder.layer0': 2.431177988432354, 'gradnorm/decoder.layer1': 1.3323243565410074, 'gradnorm/decoder.layer2': 0.7256108686597332, 'gradnorm/max_param': 326.28729248046875, 'epoch': 7.16}

 72%|███████▏  | 13660/19090 [6:39:56<2:56:33,  1.95s/it]
 72%|███████▏  | 13661/19090 [6:39:57<2:38:23,  1.75s/it]
 72%|███████▏  | 13662/19090 [6:39:59<2:33:23,  1.70s/it]
 72%|███████▏  | 13663/19090 [6:40:01<2:56:41,  1.95s/it]
 72%|███████▏  | 13664/19090 [6:40:02<2:38:30,  1.75s/it]
 72%|███████▏  | 13665/19090 [6:40:04<2:34:48,  1.71s/it]
 72%|███████▏  | 13666/19090 [6:40:07<2:59:30,  1.99s/it]
 72%|███████▏  | 13667/19090 [6:40:08<2:40:27,  1.78s/it]
 72%|███████▏  | 13668/19090 [6:40:09<2:28:30,  1.64s/it]
 72%|███████▏  | 13669/19090 [6:40:12<2:55:55,  1.95s/it][gradnorm] step 13669: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 292.0

 72%|███████▏  | 13670/19090 [6:40:13<2:37:55,  1.75s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 421.9938049316406, 'learning_rate': 1.130877741957812e-05, 'recon_loss': 0.0017343447834718972, 'recon_static': 0.0007646175828995184, 'recon_motion': 0.0009697272194898688, 'recon_norm': 1.3705846160650252, 'recon_norm_std': 0.03910415153950453, 'target_norm': 1.6190514475107194, 'recon_cos': 0.5984467282891274, 'gradnorm/compressor.query': 0.2607584601764343, 'gradnorm/compressor.layer0': 11.435613424496792, 'gradnorm/compressor.layer1': 413.8318650128334, 'gradnorm/compressor.layer2': 78.55680045952418, 'gradnorm/compressor.layer3': 21.162869971478052, 'gradnorm/compressor.layer4': 6.262831067378046, 'gradnorm/compressor.layer5': 3.734700591765267, 'gradnorm/compressor.layer6': 2.2420803137489926, 'gradnorm/compressor.layer7': 2.1115908816537248, 'gradnorm/decoder.layer0': 2.6115553649315544, 'gradnorm/decoder.layer1': 1.4425125631912534, 'gradnorm/decoder.layer2': 0.7444398533701679, 'gradnorm/max_param': 291.9990539550781, 'epoch': 7.16}

 72%|███████▏  | 13670/19090 [6:40:13<2:37:55,  1.75s/it]
 72%|███████▏  | 13671/19090 [6:40:15<2:32:14,  1.69s/it]
 72%|███████▏  | 13672/19090 [6:40:17<2:48:16,  1.86s/it]
 72%|███████▏  | 13673/19090 [6:40:18<2:32:42,  1.69s/it]
 72%|███████▏  | 13674/19090 [6:40:20<2:34:54,  1.72s/it]
 72%|███████▏  | 13675/19090 [6:40:23<2:59:53,  1.99s/it]
 72%|███████▏  | 13676/19090 [6:40:24<2:40:44,  1.78s/it]
 72%|███████▏  | 13677/19090 [6:40:25<2:27:28,  1.63s/it]
 72%|███████▏  | 13678/19090 [6:40:28<2:51:47,  1.90s/it]
 72%|███████▏  | 13679/19090 [6:40:29<2:35:05,  1.72s/it][gradnorm] step 13679: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 281.6

 72%|███████▏  | 13680/19090 [6:40:31<2:33:01,  1.70s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 405.9906005859375, 'learning_rate': 1.127055173402031e-05, 'recon_loss': 0.001831741858040914, 'recon_static': 0.0008635064645204693, 'recon_motion': 0.0009682354328106157, 'recon_norm': 1.4170955777168275, 'recon_norm_std': 0.03791893031448126, 'target_norm': 1.6207482129335404, 'recon_cos': 0.5839515164494514, 'gradnorm/compressor.query': 0.2690345915166902, 'gradnorm/compressor.layer0': 11.90759301808619, 'gradnorm/compressor.layer1': 399.1696850964512, 'gradnorm/compressor.layer2': 69.66912594094316, 'gradnorm/compressor.layer3': 20.296125461285325, 'gradnorm/compressor.layer4': 6.661175691773915, 'gradnorm/compressor.layer5': 4.040952376861245, 'gradnorm/compressor.layer6': 2.6245278433370944, 'gradnorm/compressor.layer7': 2.4606760037469906, 'gradnorm/decoder.layer0': 2.8203447926540264, 'gradnorm/decoder.layer1': 1.4094184361296371, 'gradnorm/decoder.layer2': 0.8805801050398875, 'gradnorm/max_param': 281.64935302734375, 'epoch': 7.17}

 72%|███████▏  | 13680/19090 [6:40:31<2:33:01,  1.70s/it]
 72%|███████▏  | 13681/19090 [6:40:33<2:50:39,  1.89s/it]
 72%|███████▏  | 13682/19090 [6:40:34<2:34:19,  1.71s/it]
 72%|███████▏  | 13683/19090 [6:40:36<2:34:41,  1.72s/it]
 72%|███████▏  | 13684/19090 [6:40:38<2:43:41,  1.82s/it]
 72%|███████▏  | 13685/19090 [6:40:39<2:29:22,  1.66s/it]
 72%|███████▏  | 13686/19090 [6:40:41<2:34:39,  1.72s/it]
 72%|███████▏  | 13687/19090 [6:40:43<2:44:17,  1.82s/it]
 72%|███████▏  | 13688/19090 [6:40:45<2:29:49,  1.66s/it]
 72%|███████▏  | 13689/19090 [6:40:47<2:40:31,  1.78s/it][gradnorm] step 13689: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 306.4

 72%|███████▏  | 13690/19090 [6:40:49<2:45:28,  1.84s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 444.4706726074219, 'learning_rate': 1.1232371953022993e-05, 'recon_loss': 0.0018150785152101889, 'recon_static': 0.0008540166090824641, 'recon_motion': 0.0009610619017621502, 'recon_norm': 1.4062468528747558, 'recon_norm_std': 0.03856223598122597, 'target_norm': 1.6183870494365693, 'recon_cos': 0.5849722743034362, 'gradnorm/compressor.query': 0.25116169257755894, 'gradnorm/compressor.layer0': 10.522765049622214, 'gradnorm/compressor.layer1': 434.14605477304525, 'gradnorm/compressor.layer2': 91.26835752843095, 'gradnorm/compressor.layer3': 23.3441055863145, 'gradnorm/compressor.layer4': 6.6601771558506195, 'gradnorm/compressor.layer5': 4.37621486690031, 'gradnorm/compressor.layer6': 2.3899411488713187, 'gradnorm/compressor.layer7': 2.6570074460029716, 'gradnorm/decoder.layer0': 2.5649538154236518, 'gradnorm/decoder.layer1': 1.4796127219877069, 'gradnorm/decoder.layer2': 0.8072592369867182, 'gradnorm/max_param': 306.35235595703125, 'epoch': 7.17}

 72%|███████▏  | 13690/19090 [6:40:49<2:45:28,  1.84s/it]
 72%|███████▏  | 13691/19090 [6:40:50<2:31:26,  1.68s/it]
 72%|███████▏  | 13692/19090 [6:40:52<2:44:39,  1.83s/it]
 72%|███████▏  | 13693/19090 [6:40:54<2:51:11,  1.90s/it]
 72%|███████▏  | 13694/19090 [6:40:56<2:34:56,  1.72s/it]
 72%|███████▏  | 13695/19090 [6:40:57<2:38:49,  1.77s/it]
 72%|███████▏  | 13696/19090 [6:41:00<2:58:37,  1.99s/it]
 72%|███████▏  | 13697/19090 [6:41:01<2:40:05,  1.78s/it]
 72%|███████▏  | 13698/19090 [6:41:03<2:39:41,  1.78s/it]
 72%|███████▏  | 13699/19090 [6:41:05<2:45:04,  1.84s/it][gradnorm] step 13699: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 338.3

 72%|███████▏  | 13700/19090 [6:41:06<2:30:35,  1.68s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 487.4201354980469, 'learning_rate': 1.1194238204240698e-05, 'recon_loss': 0.00164527379674837, 'recon_static': 0.0006640142571995966, 'recon_motion': 0.0009812595759285613, 'recon_norm': 1.3274230539798737, 'recon_norm_std': 0.0391577803529799, 'target_norm': 1.6225877314805985, 'recon_cos': 0.6164579704403877, 'gradnorm/compressor.query': 0.24355362323690594, 'gradnorm/compressor.layer0': 9.083477287435008, 'gradnorm/compressor.layer1': 479.3560133136957, 'gradnorm/compressor.layer2': 84.39412977586178, 'gradnorm/compressor.layer3': 22.317979155907366, 'gradnorm/compressor.layer4': 7.378310155746676, 'gradnorm/compressor.layer5': 3.734128680544988, 'gradnorm/compressor.layer6': 2.691574733648309, 'gradnorm/compressor.layer7': 2.74405712937365, 'gradnorm/decoder.layer0': 2.585420124585475, 'gradnorm/decoder.layer1': 1.5501865428528954, 'gradnorm/decoder.layer2': 0.9533074313815373, 'gradnorm/max_param': 338.2985534667969, 'epoch': 7.18}

 72%|███████▏  | 13700/19090 [6:41:06<2:30:35,  1.68s/it]
 72%|███████▏  | 13701/19090 [6:41:09<2:44:24,  1.83s/it]
 72%|███████▏  | 13702/19090 [6:41:11<2:54:40,  1.95s/it]
 72%|███████▏  | 13703/19090 [6:41:12<2:37:12,  1.75s/it]
 72%|███████▏  | 13704/19090 [6:41:14<2:38:00,  1.76s/it]
 72%|███████▏  | 13705/19090 [6:41:16<2:46:44,  1.86s/it]
 72%|███████▏  | 13706/19090 [6:41:17<2:31:26,  1.69s/it]
 72%|███████▏  | 13707/19090 [6:41:19<2:42:33,  1.81s/it]
 72%|███████▏  | 13708/19090 [6:41:21<2:46:51,  1.86s/it]
 72%|███████▏  | 13709/19090 [6:41:23<2:31:21,  1.69s/it][gradnorm] step 13709: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 322.1

 72%|███████▏  | 13710/19090 [6:41:25<2:44:24,  1.83s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 464.1380310058594, 'learning_rate': 1.1156150615173998e-05, 'recon_loss': 0.001583241284242831, 'recon_static': 0.0006155124836368486, 'recon_motion': 0.0009677287729573436, 'recon_norm': 1.3080361008644104, 'recon_norm_std': 0.039687707740813497, 'target_norm': 1.6190832942724227, 'recon_cos': 0.6285990089178085, 'gradnorm/compressor.query': 0.22864839169338905, 'gradnorm/compressor.layer0': 7.7919713197058895, 'gradnorm/compressor.layer1': 456.4153125170101, 'gradnorm/compressor.layer2': 80.04649222775943, 'gradnorm/compressor.layer3': 23.659778004344076, 'gradnorm/compressor.layer4': 7.04952432096347, 'gradnorm/compressor.layer5': 3.280237532132683, 'gradnorm/compressor.layer6': 2.383668439376969, 'gradnorm/compressor.layer7': 2.4154214830028646, 'gradnorm/decoder.layer0': 2.569184864667284, 'gradnorm/decoder.layer1': 1.416920424155168, 'gradnorm/decoder.layer2': 0.8086918573248123, 'gradnorm/max_param': 322.0975036621094, 'epoch': 7.18}

 72%|███████▏  | 13710/19090 [6:41:25<2:44:24,  1.83s/it]
 72%|███████▏  | 13711/19090 [6:41:27<2:48:32,  1.88s/it]
 72%|███████▏  | 13712/19090 [6:41:28<2:32:48,  1.70s/it]
 72%|███████▏  | 13713/19090 [6:41:30<2:42:55,  1.82s/it]
 72%|███████▏  | 13714/19090 [6:41:32<2:41:12,  1.80s/it]
 72%|███████▏  | 13715/19090 [6:41:33<2:27:22,  1.65s/it]
 72%|███████▏  | 13716/19090 [6:41:36<2:48:56,  1.89s/it]
 72%|███████▏  | 13717/19090 [6:41:37<2:37:06,  1.75s/it]
 72%|███████▏  | 13718/19090 [6:41:38<2:24:43,  1.62s/it]
 72%|███████▏  | 13719/19090 [6:41:41<2:54:52,  1.95s/it][gradnorm] step 13719: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 354.2

 72%|███████▏  | 13720/19090 [6:41:43<2:41:49,  1.81s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 511.95489501953125, 'learning_rate': 1.1118109313169198e-05, 'recon_loss': 0.0015539219573838636, 'recon_static': 0.0005860842880792916, 'recon_motion': 0.0009678376707597635, 'recon_norm': 1.2936104744672776, 'recon_norm_std': 0.040868819784373046, 'target_norm': 1.6190212041139602, 'recon_cos': 0.6348337903618813, 'gradnorm/compressor.query': 0.3286067514210074, 'gradnorm/compressor.layer0': 15.35486506301881, 'gradnorm/compressor.layer1': 501.93320023684424, 'gradnorm/compressor.layer2': 96.32678045493113, 'gradnorm/compressor.layer3': 23.477692100527843, 'gradnorm/compressor.layer4': 7.5045950482606925, 'gradnorm/compressor.layer5': 3.6578398288388656, 'gradnorm/compressor.layer6': 2.673139574048188, 'gradnorm/compressor.layer7': 2.7173551291670095, 'gradnorm/decoder.layer0': 2.6715139401625643, 'gradnorm/decoder.layer1': 1.6453737832563144, 'gradnorm/decoder.layer2': 0.9158673196884222, 'gradnorm/max_param': 354.1713562011719, 'epoch': 7.19}

 72%|███████▏  | 13720/19090 [6:41:43<2:41:49,  1.81s/it]
 72%|███████▏  | 13721/19090 [6:41:44<2:27:54,  1.65s/it]
 72%|███████▏  | 13722/19090 [6:41:47<2:58:31,  2.00s/it]
 72%|███████▏  | 13723/19090 [6:41:48<2:39:21,  1.78s/it]
 72%|███████▏  | 13724/19090 [6:41:49<2:25:59,  1.63s/it]
 72%|███████▏  | 13725/19090 [6:41:52<2:58:25,  2.00s/it]
 72%|███████▏  | 13726/19090 [6:41:53<2:39:23,  1.78s/it]
 72%|███████▏  | 13727/19090 [6:41:55<2:26:03,  1.63s/it]
 72%|███████▏  | 13728/19090 [6:41:57<2:39:15,  1.78s/it]
 72%|███████▏  | 13729/19090 [6:41:58<2:27:16,  1.65s/it][gradnorm] step 13729: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 396.4

 72%|███████▏  | 13730/19090 [6:41:59<2:17:35,  1.54s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 571.5896606445312, 'learning_rate': 1.1080114425417804e-05, 'recon_loss': 0.0017391549481544644, 'recon_static': 0.0007709516416070983, 'recon_motion': 0.0009682033021817915, 'recon_norm': 1.373204019665718, 'recon_norm_std': 0.03959152800962329, 'target_norm': 1.6206813633441925, 'recon_cos': 0.599090752005577, 'gradnorm/compressor.query': 0.3211426096773067, 'gradnorm/compressor.layer0': 12.347582103937457, 'gradnorm/compressor.layer1': 561.8549913011364, 'gradnorm/compressor.layer2': 100.935415032646, 'gradnorm/compressor.layer3': 24.21374822342348, 'gradnorm/compressor.layer4': 8.123345778868853, 'gradnorm/compressor.layer5': 4.057766548281402, 'gradnorm/compressor.layer6': 2.6350701045169544, 'gradnorm/compressor.layer7': 2.342976149915894, 'gradnorm/decoder.layer0': 2.809195378452059, 'gradnorm/decoder.layer1': 1.7409248052766562, 'gradnorm/decoder.layer2': 1.0180890051073908, 'gradnorm/max_param': 396.4474792480469, 'epoch': 7.19}

 72%|███████▏  | 13730/19090 [6:41:59<2:17:35,  1.54s/it]
 72%|███████▏  | 13731/19090 [6:42:02<2:41:33,  1.81s/it]
 72%|███████▏  | 13732/19090 [6:42:03<2:30:16,  1.68s/it]
 72%|███████▏  | 13733/19090 [6:42:04<2:19:48,  1.57s/it]
 72%|███████▏  | 13734/19090 [6:42:07<2:47:02,  1.87s/it]
 72%|███████▏  | 13735/19090 [6:42:09<2:40:40,  1.80s/it]
 72%|███████▏  | 13736/19090 [6:42:10<2:26:54,  1.65s/it]
 72%|███████▏  | 13737/19090 [6:42:12<2:49:28,  1.90s/it]
 72%|███████▏  | 13738/19090 [6:42:14<2:48:14,  1.89s/it]
 72%|███████▏  | 13739/19090 [6:42:16<2:32:10,  1.71s/it][gradnorm] step 13739: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 365.6

 72%|███████▏  | 13740/19090 [6:42:18<2:43:49,  1.84s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 525.4800415039062, 'learning_rate': 1.1042166078956106e-05, 'recon_loss': 0.0016228167311055585, 'recon_static': 0.0006500880263047293, 'recon_motion': 0.0009727287120767869, 'recon_norm': 1.3221926748752595, 'recon_norm_std': 0.040470187272876504, 'target_norm': 1.6206489622592926, 'recon_cos': 0.6209329843521119, 'gradnorm/compressor.query': 0.2767122628277087, 'gradnorm/compressor.layer0': 10.122010719871007, 'gradnorm/compressor.layer1': 518.1134950471759, 'gradnorm/compressor.layer2': 83.08989507793652, 'gradnorm/compressor.layer3': 23.83966933462742, 'gradnorm/compressor.layer4': 8.54752482675906, 'gradnorm/compressor.layer5': 3.728757327894529, 'gradnorm/compressor.layer6': 2.797396222355248, 'gradnorm/compressor.layer7': 2.589806926114528, 'gradnorm/decoder.layer0': 2.716763878079632, 'gradnorm/decoder.layer1': 1.728768161063192, 'gradnorm/decoder.layer2': 1.0140096876972007, 'gradnorm/max_param': 365.57611083984375, 'epoch': 7.2}

 72%|███████▏  | 13740/19090 [6:42:18<2:43:49,  1.84s/it]
 72%|███████▏  | 13741/19090 [6:42:20<2:46:08,  1.86s/it]
 72%|███████▏  | 13742/19090 [6:42:21<2:30:37,  1.69s/it]
 72%|███████▏  | 13743/19090 [6:42:23<2:44:48,  1.85s/it]
 72%|███████▏  | 13744/19090 [6:42:25<2:43:42,  1.84s/it]
 72%|███████▏  | 13745/19090 [6:42:26<2:28:55,  1.67s/it]
 72%|███████▏  | 13746/19090 [6:42:28<2:37:16,  1.77s/it]
 72%|███████▏  | 13747/19090 [6:42:30<2:49:10,  1.90s/it]
 72%|███████▏  | 13748/19090 [6:42:32<2:33:02,  1.72s/it]
 72%|███████▏  | 13749/19090 [6:42:34<2:34:07,  1.73s/it][gradnorm] step 13749: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 317.9

 72%|███████▏  | 13750/19090 [6:42:35<2:40:54,  1.81s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 457.99639892578125, 'learning_rate': 1.1004264400664852e-05, 'recon_loss': 0.0016323772812029346, 'recon_static': 0.0006551125843543559, 'recon_motion': 0.0009772647259524091, 'recon_norm': 1.3221151053905487, 'recon_norm_std': 0.03950435137376189, 'target_norm': 1.6216616094112397, 'recon_cos': 0.6187536135315895, 'gradnorm/compressor.query': 0.3060167489945409, 'gradnorm/compressor.layer0': 9.998825004211433, 'gradnorm/compressor.layer1': 450.53357255591953, 'gradnorm/compressor.layer2': 77.86461921682562, 'gradnorm/compressor.layer3': 23.489290434934784, 'gradnorm/compressor.layer4': 5.726737692366219, 'gradnorm/compressor.layer5': 3.4742804077249527, 'gradnorm/compressor.layer6': 2.2668910205360153, 'gradnorm/compressor.layer7': 2.359534251920562, 'gradnorm/decoder.layer0': 2.56850539107868, 'gradnorm/decoder.layer1': 1.622364033672627, 'gradnorm/decoder.layer2': 0.83783389168407, 'gradnorm/max_param': 317.91668701171875, 'epoch': 7.2}

 72%|███████▏  | 13750/19090 [6:42:36<2:40:54,  1.81s/it]
 72%|███████▏  | 13751/19090 [6:42:37<2:26:58,  1.65s/it]
 72%|███████▏  | 13752/19090 [6:42:39<2:40:22,  1.80s/it]
 72%|███████▏  | 13753/19090 [6:42:41<2:39:03,  1.79s/it]
 72%|███████▏  | 13754/19090 [6:42:42<2:25:43,  1.64s/it]
 72%|███████▏  | 13755/19090 [6:42:44<2:44:42,  1.85s/it]
 72%|███████▏  | 13756/19090 [6:42:46<2:40:28,  1.81s/it]
 72%|███████▏  | 13757/19090 [6:42:47<2:26:37,  1.65s/it]
 72%|███████▏  | 13758/19090 [6:42:49<2:37:12,  1.77s/it]
 72%|███████▏  | 13759/19090 [6:42:51<2:34:47,  1.74s/it][gradnorm] step 13759: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 345.2

 72%|███████▏  | 13760/19090 [6:42:52<2:22:38,  1.61s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 495.8875427246094, 'learning_rate': 1.0966409517268713e-05, 'recon_loss': 0.001598072491469793, 'recon_static': 0.0006303312664385885, 'recon_motion': 0.0009677412002929486, 'recon_norm': 1.3120425939559937, 'recon_norm_std': 0.040486642811447385, 'target_norm': 1.6203275471925735, 'recon_cos': 0.6256812497973442, 'gradnorm/compressor.query': 0.23582266513977249, 'gradnorm/compressor.layer0': 8.918827638776035, 'gradnorm/compressor.layer1': 489.258175710943, 'gradnorm/compressor.layer2': 75.00941021694378, 'gradnorm/compressor.layer3': 27.53623544886469, 'gradnorm/compressor.layer4': 6.076937444237177, 'gradnorm/compressor.layer5': 3.00539818510586, 'gradnorm/compressor.layer6': 2.26283745119615, 'gradnorm/compressor.layer7': 2.3613425633799174, 'gradnorm/decoder.layer0': 2.678946968274229, 'gradnorm/decoder.layer1': 1.4505535055849188, 'gradnorm/decoder.layer2': 0.8492631228592246, 'gradnorm/max_param': 345.2266540527344, 'epoch': 7.21}

 72%|███████▏  | 13760/19090 [6:42:52<2:22:38,  1.61s/it]
 72%|███████▏  | 13761/19090 [6:42:55<2:45:56,  1.87s/it]
 72%|███████▏  | 13762/19090 [6:42:56<2:33:03,  1.72s/it]
 72%|███████▏  | 13763/19090 [6:42:57<2:21:26,  1.59s/it]
 72%|███████▏  | 13764/19090 [6:43:00<2:44:13,  1.85s/it]
 72%|███████▏  | 13765/19090 [6:43:02<2:46:33,  1.88s/it]
 72%|███████▏  | 13766/19090 [6:43:03<2:30:48,  1.70s/it]
 72%|███████▏  | 13767/19090 [6:43:05<2:36:52,  1.77s/it]
 72%|███████▏  | 13768/19090 [6:43:07<2:46:52,  1.88s/it]
 72%|███████▏  | 13769/19090 [6:43:09<2:30:58,  1.70s/it][gradnorm] step 13769: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 279.9

 72%|███████▏  | 13770/19090 [6:43:11<2:45:42,  1.87s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 404.37896728515625, 'learning_rate': 1.0928601555335883e-05, 'recon_loss': 0.0016723041539080442, 'recon_static': 0.0007010830828221515, 'recon_motion': 0.0009712211263831705, 'recon_norm': 1.3447341948747635, 'recon_norm_std': 0.0401708634570241, 'target_norm': 1.620073539018631, 'recon_cos': 0.6109316229820252, 'gradnorm/compressor.query': 0.19314229705657088, 'gradnorm/compressor.layer0': 7.506716962863784, 'gradnorm/compressor.layer1': 396.59888032494496, 'gradnorm/compressor.layer2': 75.52485621469266, 'gradnorm/compressor.layer3': 20.14451875181404, 'gradnorm/compressor.layer4': 6.0435980855632865, 'gradnorm/compressor.layer5': 3.1884200320409817, 'gradnorm/compressor.layer6': 2.1137822204335865, 'gradnorm/compressor.layer7': 2.2676292236248417, 'gradnorm/decoder.layer0': 2.531880053061041, 'gradnorm/decoder.layer1': 1.4533600975920267, 'gradnorm/decoder.layer2': 0.817901628977558, 'gradnorm/max_param': 279.8883361816406, 'epoch': 7.21}

 72%|███████▏  | 13770/19090 [6:43:11<2:45:42,  1.87s/it]
 72%|███████▏  | 13771/19090 [6:43:12<2:35:07,  1.75s/it]
 72%|███████▏  | 13772/19090 [6:43:14<2:23:02,  1.61s/it]
 72%|███████▏  | 13773/19090 [6:43:16<2:40:56,  1.82s/it]
 72%|███████▏  | 13774/19090 [6:43:18<2:37:24,  1.78s/it]
 72%|███████▏  | 13775/19090 [6:43:19<2:24:24,  1.63s/it]
 72%|███████▏  | 13776/19090 [6:43:21<2:41:16,  1.82s/it]
 72%|███████▏  | 13777/19090 [6:43:23<2:34:45,  1.75s/it]
 72%|███████▏  | 13778/19090 [6:43:24<2:22:32,  1.61s/it]
 72%|███████▏  | 13779/19090 [6:43:26<2:45:08,  1.87s/it][gradnorm] step 13779: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 311.8

 72%|███████▏  | 13780/19090 [6:43:28<2:39:50,  1.81s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 451.3915710449219, 'learning_rate': 1.0890840641277716e-05, 'recon_loss': 0.0016597344481851905, 'recon_static': 0.0006887676630867645, 'recon_motion': 0.0009709667661809363, 'recon_norm': 1.3416947811841964, 'recon_norm_std': 0.03969283746555448, 'target_norm': 1.6192719846963883, 'recon_cos': 0.6135159835219384, 'gradnorm/compressor.query': 0.319991765288853, 'gradnorm/compressor.layer0': 12.453977778204038, 'gradnorm/compressor.layer1': 441.8499992927464, 'gradnorm/compressor.layer2': 88.56014850000959, 'gradnorm/compressor.layer3': 21.38555683202158, 'gradnorm/compressor.layer4': 6.188405288591653, 'gradnorm/compressor.layer5': 3.0553850319028975, 'gradnorm/compressor.layer6': 2.2650079347248333, 'gradnorm/compressor.layer7': 2.278433432716022, 'gradnorm/decoder.layer0': 2.550033857083417, 'gradnorm/decoder.layer1': 1.4519407307216656, 'gradnorm/decoder.layer2': 0.9466007691660516, 'gradnorm/max_param': 311.8082580566406, 'epoch': 7.22}

 72%|███████▏  | 13780/19090 [6:43:28<2:39:50,  1.81s/it]
 72%|███████▏  | 13781/19090 [6:43:29<2:26:08,  1.65s/it]
 72%|███████▏  | 13782/19090 [6:43:32<2:42:41,  1.84s/it]
 72%|███████▏  | 13783/19090 [6:43:34<2:44:40,  1.86s/it]
 72%|███████▏  | 13784/19090 [6:43:35<2:29:35,  1.69s/it]
 72%|███████▏  | 13785/19090 [6:43:37<2:48:29,  1.91s/it]
 72%|███████▏  | 13786/19090 [6:43:39<2:34:14,  1.74s/it]
 72%|███████▏  | 13787/19090 [6:43:40<2:22:18,  1.61s/it]
 72%|███████▏  | 13788/19090 [6:43:43<2:49:39,  1.92s/it]
 72%|███████▏  | 13789/19090 [6:43:44<2:33:02,  1.73s/it][gradnorm] step 13789: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 405.2

 72%|███████▏  | 13790/19090 [6:43:45<2:21:18,  1.60s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 583.9305419921875, 'learning_rate': 1.0853126901348248e-05, 'recon_loss': 0.0016734101198380813, 'recon_static': 0.0006980003105127253, 'recon_motion': 0.0009754097744007594, 'recon_norm': 1.3449124068021774, 'recon_norm_std': 0.03987038787454367, 'target_norm': 1.6200991839170455, 'recon_cos': 0.6109052777290345, 'gradnorm/compressor.query': 0.38129865304652466, 'gradnorm/compressor.layer0': 14.72248316759497, 'gradnorm/compressor.layer1': 574.2869491813304, 'gradnorm/compressor.layer2': 100.97904040721768, 'gradnorm/compressor.layer3': 26.007767910745795, 'gradnorm/compressor.layer4': 6.385699101518699, 'gradnorm/compressor.layer5': 3.7981388202143993, 'gradnorm/compressor.layer6': 2.576330681823374, 'gradnorm/compressor.layer7': 2.3569781526597735, 'gradnorm/decoder.layer0': 2.8582282763452795, 'gradnorm/decoder.layer1': 1.6287414321029983, 'gradnorm/decoder.layer2': 1.0445562846137466, 'gradnorm/max_param': 405.2085876464844, 'epoch': 7.22}

 72%|███████▏  | 13790/19090 [6:43:45<2:21:18,  1.60s/it]
 72%|███████▏  | 13791/19090 [6:43:48<2:55:36,  1.99s/it]
 72%|███████▏  | 13792/19090 [6:43:49<2:40:26,  1.82s/it]
 72%|███████▏  | 13793/19090 [6:43:51<2:26:36,  1.66s/it]
 72%|███████▏  | 13794/19090 [6:43:53<2:45:07,  1.87s/it]
 72%|███████▏  | 13795/19090 [6:43:54<2:30:29,  1.71s/it]
 72%|███████▏  | 13796/19090 [6:43:56<2:19:16,  1.58s/it]
 72%|███████▏  | 13797/19090 [6:43:58<2:45:03,  1.87s/it]
 72%|███████▏  | 13798/19090 [6:44:00<2:31:19,  1.72s/it]
 72%|███████▏  | 13799/19090 [6:44:01<2:20:02,  1.59s/it][gradnorm] step 13799: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 312.9

 72%|███████▏  | 13800/19090 [6:44:03<2:45:44,  1.88s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 452.092529296875, 'learning_rate': 1.0815460461643753e-05, 'recon_loss': 0.0018350684491451829, 'recon_static': 0.0008655960235046222, 'recon_motion': 0.0009694725347799249, 'recon_norm': 1.4152712047100067, 'recon_norm_std': 0.037898655887693165, 'target_norm': 1.6205584049224853, 'recon_cos': 0.5829817995429039, 'gradnorm/compressor.query': 0.5589892181250655, 'gradnorm/compressor.layer0': 26.541321980280813, 'gradnorm/compressor.layer1': 443.46422135838645, 'gradnorm/compressor.layer2': 80.10417716460702, 'gradnorm/compressor.layer3': 22.914054509033726, 'gradnorm/compressor.layer4': 6.984444637079439, 'gradnorm/compressor.layer5': 3.731643124418521, 'gradnorm/compressor.layer6': 2.3059026068826016, 'gradnorm/compressor.layer7': 2.188697705038996, 'gradnorm/decoder.layer0': 2.27050445228246, 'gradnorm/decoder.layer1': 1.4347097434262026, 'gradnorm/decoder.layer2': 0.7842452044451342, 'gradnorm/max_param': 312.8712158203125, 'epoch': 7.23}

 72%|███████▏  | 13800/19090 [6:44:03<2:45:44,  1.88s/it]
 72%|███████▏  | 13801/19090 [6:44:05<2:40:54,  1.83s/it]
 72%|███████▏  | 13802/19090 [6:44:06<2:26:37,  1.66s/it]
 72%|███████▏  | 13803/19090 [6:44:09<2:37:28,  1.79s/it]
 72%|███████▏  | 13804/19090 [6:44:11<2:42:42,  1.85s/it]
 72%|███████▏  | 13805/19090 [6:44:12<2:27:55,  1.68s/it]
 72%|███████▏  | 13806/19090 [6:44:14<2:42:15,  1.84s/it]
 72%|███████▏  | 13807/19090 [6:44:16<2:41:22,  1.83s/it]
 72%|███████▏  | 13808/19090 [6:44:17<2:27:00,  1.67s/it]
 72%|███████▏  | 13809/19090 [6:44:20<2:45:55,  1.89s/it][gradnorm] step 13809: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 368.6

 72%|███████▏  | 13810/19090 [6:44:21<2:39:14,  1.81s/it]
                                                         
{'loss': 0.0027, 'grad_norm': 533.657958984375, 'learning_rate': 1.077784144810241e-05, 'recon_loss': 0.0027225811674725263, 'recon_static': 0.0017445616802433505, 'recon_motion': 0.0009780194362974727, 'recon_norm': 1.737289771437645, 'recon_norm_std': 0.0328729763161391, 'target_norm': 1.6206459015607835, 'recon_cos': 0.47886011451482774, 'gradnorm/compressor.query': 0.4567965206316452, 'gradnorm/compressor.layer0': 18.198185354460826, 'gradnorm/compressor.layer1': 522.3575463702999, 'gradnorm/compressor.layer2': 103.77412652457018, 'gradnorm/compressor.layer3': 27.117596015721507, 'gradnorm/compressor.layer4': 6.724376221270308, 'gradnorm/compressor.layer5': 4.722494698436178, 'gradnorm/compressor.layer6': 3.1712492527583063, 'gradnorm/compressor.layer7': 2.7912958003180255, 'gradnorm/decoder.layer0': 2.956388254945052, 'gradnorm/decoder.layer1': 1.5792183440651302, 'gradnorm/decoder.layer2': 0.9737419901311507, 'gradnorm/max_param': 368.591796875, 'epoch': 7.23}

 72%|███████▏  | 13810/19090 [6:44:21<2:39:14,  1.81s/it]
 72%|███████▏  | 13811/19090 [6:44:22<2:25:31,  1.65s/it]
 72%|███████▏  | 13812/19090 [6:44:25<2:52:38,  1.96s/it]
 72%|███████▏  | 13813/19090 [6:44:27<2:46:00,  1.89s/it]
 72%|███████▏  | 13814/19090 [6:44:28<2:30:19,  1.71s/it]
 72%|███████▏  | 13815/19090 [6:44:30<2:46:44,  1.90s/it]
 72%|███████▏  | 13816/19090 [6:44:32<2:43:21,  1.86s/it]
 72%|███████▏  | 13817/19090 [6:44:34<2:28:21,  1.69s/it]
 72%|███████▏  | 13818/19090 [6:44:36<2:46:36,  1.90s/it]
 72%|███████▏  | 13819/19090 [6:44:37<2:34:09,  1.75s/it][gradnorm] step 13819: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 280.3

 72%|███████▏  | 13820/19090 [6:44:39<2:21:53,  1.62s/it]
                                                         
{'loss': 0.0027, 'grad_norm': 403.8277282714844, 'learning_rate': 1.0740269986503801e-05, 'recon_loss': 0.0026483968773391097, 'recon_static': 0.0016783898317953572, 'recon_motion': 0.0009700070557300932, 'recon_norm': 1.7036119401454926, 'recon_norm_std': 0.03317825947888196, 'target_norm': 1.619224515557289, 'recon_cos': 0.4799986474215984, 'gradnorm/compressor.query': 0.3081506251692291, 'gradnorm/compressor.layer0': 11.35105438040339, 'gradnorm/compressor.layer1': 397.2101930968036, 'gradnorm/compressor.layer2': 68.16511505276178, 'gradnorm/compressor.layer3': 21.30010393144208, 'gradnorm/compressor.layer4': 6.103283745595321, 'gradnorm/compressor.layer5': 3.2923361825704403, 'gradnorm/compressor.layer6': 2.583383457148768, 'gradnorm/compressor.layer7': 2.4312082909779393, 'gradnorm/decoder.layer0': 2.717258965296621, 'gradnorm/decoder.layer1': 1.659300963048633, 'gradnorm/decoder.layer2': 0.9739669486632069, 'gradnorm/max_param': 280.2665710449219, 'epoch': 7.24}

 72%|███████▏  | 13820/19090 [6:44:39<2:21:53,  1.62s/it]
 72%|███████▏  | 13821/19090 [6:44:41<2:50:49,  1.95s/it]
 72%|███████▏  | 13822/19090 [6:44:43<2:33:26,  1.75s/it]
 72%|███████▏  | 13823/19090 [6:44:44<2:21:23,  1.61s/it]
 72%|███████▏  | 13824/19090 [6:44:47<2:58:49,  2.04s/it]
 72%|███████▏  | 13825/19090 [6:44:48<2:39:16,  1.82s/it]
 72%|███████▏  | 13826/19090 [6:44:50<2:25:29,  1.66s/it]
 72%|███████▏  | 13827/19090 [6:44:52<2:48:38,  1.92s/it]
 72%|███████▏  | 13828/19090 [6:44:53<2:31:51,  1.73s/it]
 72%|███████▏  | 13829/19090 [6:44:55<2:20:07,  1.60s/it][gradnorm] step 13829: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 322.6

 72%|███████▏  | 13830/19090 [6:44:58<2:54:22,  1.99s/it]
                                                         
{'loss': 0.0022, 'grad_norm': 464.772216796875, 'learning_rate': 1.0702746202468494e-05, 'recon_loss': 0.0021825358620844782, 'recon_static': 0.0012166286760475487, 'recon_motion': 0.0009659071904025041, 'recon_norm': 1.5496673345565797, 'recon_norm_std': 0.03648422062397003, 'target_norm': 1.6207174956798553, 'recon_cos': 0.5320043131709099, 'gradnorm/compressor.query': 0.3149375264719842, 'gradnorm/compressor.layer0': 12.684649368029207, 'gradnorm/compressor.layer1': 457.1410306185609, 'gradnorm/compressor.layer2': 77.86837853069756, 'gradnorm/compressor.layer3': 27.160315522059246, 'gradnorm/compressor.layer4': 6.300688920878855, 'gradnorm/compressor.layer5': 3.400802596113084, 'gradnorm/compressor.layer6': 2.420230627183559, 'gradnorm/compressor.layer7': 2.389928180142197, 'gradnorm/decoder.layer0': 2.6942382440833526, 'gradnorm/decoder.layer1': 1.4640734465524687, 'gradnorm/decoder.layer2': 0.9361402505332016, 'gradnorm/max_param': 322.57733154296875, 'epoch': 7.24}

 72%|███████▏  | 13830/19090 [6:44:58<2:54:22,  1.99s/it]
 72%|███████▏  | 13831/19090 [6:44:59<2:35:57,  1.78s/it]
 72%|███████▏  | 13832/19090 [6:45:00<2:23:08,  1.63s/it]
 72%|███████▏  | 13833/19090 [6:45:03<2:52:50,  1.97s/it]
 72%|███████▏  | 13834/19090 [6:45:04<2:34:50,  1.77s/it]
 72%|███████▏  | 13835/19090 [6:45:05<2:22:17,  1.62s/it]
 72%|███████▏  | 13836/19090 [6:45:08<2:54:53,  2.00s/it]
 72%|███████▏  | 13837/19090 [6:45:10<2:36:21,  1.79s/it]
 72%|███████▏  | 13838/19090 [6:45:11<2:23:29,  1.64s/it]
 72%|███████▏  | 13839/19090 [6:45:13<2:45:10,  1.89s/it][gradnorm] step 13839: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 261.6

 72%|███████▏  | 13840/19090 [6:45:15<2:29:26,  1.71s/it]
                                                         
{'loss': 0.002, 'grad_norm': 379.5070495605469, 'learning_rate': 1.0665270221457696e-05, 'recon_loss': 0.0020311843923991545, 'recon_static': 0.0010747364052804188, 'recon_motion': 0.0009564480962581002, 'recon_norm': 1.4989486902952194, 'recon_norm_std': 0.03753062319010496, 'target_norm': 1.6186111688613891, 'recon_cos': 0.5525050982832909, 'gradnorm/compressor.query': 0.2287689337425014, 'gradnorm/compressor.layer0': 8.81249350689588, 'gradnorm/compressor.layer1': 370.7041652382665, 'gradnorm/compressor.layer2': 77.43491155379303, 'gradnorm/compressor.layer3': 21.255339546300256, 'gradnorm/compressor.layer4': 6.5465248032457435, 'gradnorm/compressor.layer5': 3.4865919510207046, 'gradnorm/compressor.layer6': 2.2664912179288246, 'gradnorm/compressor.layer7': 2.6943093022367317, 'gradnorm/decoder.layer0': 2.728169926538904, 'gradnorm/decoder.layer1': 1.591648586363568, 'gradnorm/decoder.layer2': 0.9723847404763983, 'gradnorm/max_param': 261.5763244628906, 'epoch': 7.25}

 72%|███████▏  | 13840/19090 [6:45:15<2:29:26,  1.71s/it]
 73%|███████▎  | 13841/19090 [6:45:16<2:18:33,  1.58s/it]
 73%|███████▎  | 13842/19090 [6:45:18<2:38:30,  1.81s/it]
 73%|███████▎  | 13843/19090 [6:45:20<2:29:24,  1.71s/it]
 73%|███████▎  | 13844/19090 [6:45:21<2:18:28,  1.58s/it]
 73%|███████▎  | 13845/19090 [6:45:24<2:47:16,  1.91s/it]
 73%|███████▎  | 13846/19090 [6:45:25<2:30:56,  1.73s/it]
 73%|███████▎  | 13847/19090 [6:45:26<2:19:26,  1.60s/it]
 73%|███████▎  | 13848/19090 [6:45:29<2:55:27,  2.01s/it]
 73%|███████▎  | 13849/19090 [6:45:31<2:36:32,  1.79s/it][gradnorm] step 13849: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 379.3

 73%|███████▎  | 13850/19090 [6:45:32<2:23:16,  1.64s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 546.45703125, 'learning_rate': 1.0627842168772765e-05, 'recon_loss': 0.0018691736127948388, 'recon_static': 0.0009052460038219579, 'recon_motion': 0.0009639276104280725, 'recon_norm': 1.4274197578430177, 'recon_norm_std': 0.03872521677985787, 'target_norm': 1.6196352154016496, 'recon_cos': 0.5763096138834953, 'gradnorm/compressor.query': 0.2923859330433151, 'gradnorm/compressor.layer0': 10.90423442939535, 'gradnorm/compressor.layer1': 537.5422948475775, 'gradnorm/compressor.layer2': 93.94871128207402, 'gradnorm/compressor.layer3': 25.44422467547277, 'gradnorm/compressor.layer4': 6.397925648692799, 'gradnorm/compressor.layer5': 2.9144993677012616, 'gradnorm/compressor.layer6': 2.333840644273415, 'gradnorm/compressor.layer7': 2.346205074020413, 'gradnorm/decoder.layer0': 2.7357153522926194, 'gradnorm/decoder.layer1': 1.4632923913990807, 'gradnorm/decoder.layer2': 0.9416436493572704, 'gradnorm/max_param': 379.3063659667969, 'epoch': 7.26}

 73%|███████▎  | 13850/19090 [6:45:32<2:23:16,  1.64s/it]
 73%|███████▎  | 13851/19090 [6:45:35<2:55:04,  2.01s/it]
 73%|███████▎  | 13852/19090 [6:45:36<2:36:12,  1.79s/it]
 73%|███████▎  | 13853/19090 [6:45:37<2:23:00,  1.64s/it]
 73%|███████▎  | 13854/19090 [6:45:40<2:53:14,  1.99s/it]
 73%|███████▎  | 13855/19090 [6:45:41<2:35:05,  1.78s/it]
 73%|███████▎  | 13856/19090 [6:45:43<2:22:18,  1.63s/it]
 73%|███████▎  | 13857/19090 [6:45:46<2:53:14,  1.99s/it]
 73%|███████▎  | 13858/19090 [6:45:47<2:34:51,  1.78s/it]
 73%|███████▎  | 13859/19090 [6:45:48<2:22:03,  1.63s/it][gradnorm] step 13859: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 328.7

 73%|███████▎  | 13860/19090 [6:45:51<2:42:53,  1.87s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 473.1163330078125, 'learning_rate': 1.0590462169554772e-05, 'recon_loss': 0.0017792804312193767, 'recon_static': 0.0008091506999335252, 'recon_motion': 0.0009701297865831293, 'recon_norm': 1.3880375504493714, 'recon_norm_std': 0.03952089855447412, 'target_norm': 1.6208611607551575, 'recon_cos': 0.591159151494503, 'gradnorm/compressor.query': 0.27283340207982426, 'gradnorm/compressor.layer0': 9.935823886999838, 'gradnorm/compressor.layer1': 465.84794997938974, 'gradnorm/compressor.layer2': 79.08774917970861, 'gradnorm/compressor.layer3': 19.6316186714228, 'gradnorm/compressor.layer4': 7.107480108908864, 'gradnorm/compressor.layer5': 3.76363728719674, 'gradnorm/compressor.layer6': 2.3603202492621946, 'gradnorm/compressor.layer7': 2.358605366665921, 'gradnorm/decoder.layer0': 2.6630438593920864, 'gradnorm/decoder.layer1': 1.419173640385038, 'gradnorm/decoder.layer2': 0.8694325846092369, 'gradnorm/max_param': 328.7209167480469, 'epoch': 7.26}

 73%|███████▎  | 13860/19090 [6:45:51<2:42:53,  1.87s/it]
 73%|███████▎  | 13861/19090 [6:45:52<2:27:40,  1.69s/it]
 73%|███████▎  | 13862/19090 [6:45:53<2:16:57,  1.57s/it]
 73%|███████▎  | 13863/19090 [6:45:56<2:48:04,  1.93s/it]
 73%|███████▎  | 13864/19090 [6:45:57<2:31:16,  1.74s/it]
 73%|███████▎  | 13865/19090 [6:45:58<2:19:37,  1.60s/it]
 73%|███████▎  | 13866/19090 [6:46:01<2:51:58,  1.98s/it]
 73%|███████▎  | 13867/19090 [6:46:03<2:34:03,  1.77s/it]
 73%|███████▎  | 13868/19090 [6:46:04<2:21:32,  1.63s/it]
 73%|███████▎  | 13869/19090 [6:46:06<2:47:49,  1.93s/it][gradnorm] step 13869: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 390.3

 73%|███████▎  | 13870/19090 [6:46:08<2:31:10,  1.74s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 560.7617797851562, 'learning_rate': 1.0553130348784182e-05, 'recon_loss': 0.001829100350732915, 'recon_static': 0.0008676929923240095, 'recon_motion': 0.0009614073322154581, 'recon_norm': 1.4109907537698745, 'recon_norm_std': 0.03843131642788648, 'target_norm': 1.619478017091751, 'recon_cos': 0.5822807282209397, 'gradnorm/compressor.query': 0.2902672894016303, 'gradnorm/compressor.layer0': 11.72672580145289, 'gradnorm/compressor.layer1': 553.2278294247317, 'gradnorm/compressor.layer2': 87.95806440429439, 'gradnorm/compressor.layer3': 21.009469213448952, 'gradnorm/compressor.layer4': 6.4957311257214165, 'gradnorm/compressor.layer5': 3.5799682400536015, 'gradnorm/compressor.layer6': 2.277908909390628, 'gradnorm/compressor.layer7': 2.387616926595488, 'gradnorm/decoder.layer0': 2.901032882120106, 'gradnorm/decoder.layer1': 1.4477022776154762, 'gradnorm/decoder.layer2': 0.835503046385201, 'gradnorm/max_param': 390.3188171386719, 'epoch': 7.27}

 73%|███████▎  | 13870/19090 [6:46:08<2:31:10,  1.74s/it]
 73%|███████▎  | 13871/19090 [6:46:09<2:19:27,  1.60s/it]
 73%|███████▎  | 13872/19090 [6:46:12<2:47:38,  1.93s/it]
 73%|███████▎  | 13873/19090 [6:46:13<2:31:02,  1.74s/it]
 73%|███████▎  | 13874/19090 [6:46:14<2:19:25,  1.60s/it]
 73%|███████▎  | 13875/19090 [6:46:17<2:50:19,  1.96s/it]
 73%|███████▎  | 13876/19090 [6:46:18<2:33:10,  1.76s/it]
 73%|███████▎  | 13877/19090 [6:46:20<2:21:10,  1.62s/it]
 73%|███████▎  | 13878/19090 [6:46:22<2:45:23,  1.90s/it]
 73%|███████▎  | 13879/19090 [6:46:24<2:29:47,  1.72s/it][gradnorm] step 13879: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 330.7

 73%|███████▎  | 13880/19090 [6:46:25<2:18:47,  1.60s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 474.2239685058594, 'learning_rate': 1.0515846831280348e-05, 'recon_loss': 0.0017198630404891447, 'recon_static': 0.0007429668665281497, 'recon_motion': 0.0009768962016096338, 'recon_norm': 1.3669251829385758, 'recon_norm_std': 0.038448158837854865, 'target_norm': 1.6197004824876786, 'recon_cos': 0.6024382680654525, 'gradnorm/compressor.query': 0.255550955748856, 'gradnorm/compressor.layer0': 10.21222966372133, 'gradnorm/compressor.layer1': 468.60209333185867, 'gradnorm/compressor.layer2': 69.98138005703731, 'gradnorm/compressor.layer3': 15.791256109988527, 'gradnorm/compressor.layer4': 4.844340429929528, 'gradnorm/compressor.layer5': 3.3546078946565556, 'gradnorm/compressor.layer6': 1.8772991071926208, 'gradnorm/compressor.layer7': 1.8203847977193113, 'gradnorm/decoder.layer0': 2.308394525132324, 'gradnorm/decoder.layer1': 1.392492133825469, 'gradnorm/decoder.layer2': 0.7081682115587187, 'gradnorm/max_param': 330.67010498046875, 'epoch': 7.27}

 73%|███████▎  | 13880/19090 [6:46:25<2:18:47,  1.60s/it]
 73%|███████▎  | 13881/19090 [6:46:28<2:58:19,  2.05s/it]
 73%|███████▎  | 13882/19090 [6:46:29<2:38:54,  1.83s/it]
 73%|███████▎  | 13883/19090 [6:46:31<2:25:31,  1.68s/it]
 73%|███████▎  | 13884/19090 [6:46:33<2:47:24,  1.93s/it]
 73%|███████▎  | 13885/19090 [6:46:34<2:30:55,  1.74s/it]
 73%|███████▎  | 13886/19090 [6:46:36<2:19:34,  1.61s/it]
 73%|███████▎  | 13887/19090 [6:46:39<2:51:51,  1.98s/it]
 73%|███████▎  | 13888/19090 [6:46:40<2:34:16,  1.78s/it]
 73%|███████▎  | 13889/19090 [6:46:41<2:21:38,  1.63s/it][gradnorm] step 13889: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 375.4

 73%|███████▎  | 13890/19090 [6:46:44<2:53:58,  2.01s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 539.2470703125, 'learning_rate': 1.0478611741701088e-05, 'recon_loss': 0.0018005352554609998, 'recon_static': 0.0008417477089096792, 'recon_motion': 0.0009587875159922987, 'recon_norm': 1.3963889449834823, 'recon_norm_std': 0.03915941817685962, 'target_norm': 1.6188310712575913, 'recon_cos': 0.5867386192083359, 'gradnorm/compressor.query': 0.4717891874032878, 'gradnorm/compressor.layer0': 16.907027982908282, 'gradnorm/compressor.layer1': 532.0447584555269, 'gradnorm/compressor.layer2': 82.48012603898135, 'gradnorm/compressor.layer3': 23.28432105373901, 'gradnorm/compressor.layer4': 6.85965417107012, 'gradnorm/compressor.layer5': 4.2117180888140595, 'gradnorm/compressor.layer6': 2.384077693057255, 'gradnorm/compressor.layer7': 2.2539818815503327, 'gradnorm/decoder.layer0': 2.3859358532093613, 'gradnorm/decoder.layer1': 1.548302117365992, 'gradnorm/decoder.layer2': 0.9665803917389142, 'gradnorm/max_param': 375.3940124511719, 'epoch': 7.28}

 73%|███████▎  | 13890/19090 [6:46:44<2:53:58,  2.01s/it]
 73%|███████▎  | 13891/19090 [6:46:45<2:35:22,  1.79s/it]
 73%|███████▎  | 13892/19090 [6:46:47<2:22:12,  1.64s/it]
 73%|███████▎  | 13893/19090 [6:46:49<2:48:06,  1.94s/it]
 73%|███████▎  | 13894/19090 [6:46:51<2:31:15,  1.75s/it]
 73%|███████▎  | 13895/19090 [6:46:52<2:19:21,  1.61s/it]
 73%|███████▎  | 13896/19090 [6:46:55<2:56:30,  2.04s/it]
 73%|███████▎  | 13897/19090 [6:46:56<2:37:07,  1.82s/it]
 73%|███████▎  | 13898/19090 [6:46:58<2:23:22,  1.66s/it]
 73%|███████▎  | 13899/19090 [6:47:00<2:55:38,  2.03s/it][gradnorm] step 13899: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 248.7

 73%|███████▎  | 13900/19090 [6:47:02<2:36:33,  1.81s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 361.2591247558594, 'learning_rate': 1.0441425204542366e-05, 'recon_loss': 0.0019276489008916542, 'recon_static': 0.0009604172286344692, 'recon_motion': 0.0009672316882642917, 'recon_norm': 1.4475035935640335, 'recon_norm_std': 0.03673345427960158, 'target_norm': 1.6209422111511231, 'recon_cos': 0.5682836666703224, 'gradnorm/compressor.query': 0.499398197165242, 'gradnorm/compressor.layer0': 14.244690558465821, 'gradnorm/compressor.layer1': 352.87301850311536, 'gradnorm/compressor.layer2': 72.60400675496497, 'gradnorm/compressor.layer3': 20.96937362266582, 'gradnorm/compressor.layer4': 6.397093542080873, 'gradnorm/compressor.layer5': 3.9720206413308947, 'gradnorm/compressor.layer6': 2.161689972625641, 'gradnorm/compressor.layer7': 2.2435058545928257, 'gradnorm/decoder.layer0': 2.30406496077592, 'gradnorm/decoder.layer1': 1.4751861454807522, 'gradnorm/decoder.layer2': 0.8247669078531105, 'gradnorm/max_param': 248.70094299316406, 'epoch': 7.28}

 73%|███████▎  | 13900/19090 [6:47:02<2:36:33,  1.81s/it]
 73%|███████▎  | 13901/19090 [6:47:03<2:23:02,  1.65s/it]
 73%|███████▎  | 13902/19090 [6:47:06<2:46:20,  1.92s/it]
 73%|███████▎  | 13903/19090 [6:47:07<2:30:07,  1.74s/it]
 73%|███████▎  | 13904/19090 [6:47:08<2:18:27,  1.60s/it]
 73%|███████▎  | 13905/19090 [6:47:11<2:54:49,  2.02s/it]
 73%|███████▎  | 13906/19090 [6:47:12<2:35:44,  1.80s/it]
 73%|███████▎  | 13907/19090 [6:47:14<2:22:32,  1.65s/it]
 73%|███████▎  | 13908/19090 [6:47:16<2:49:51,  1.97s/it]
 73%|███████▎  | 13909/19090 [6:47:18<2:32:25,  1.77s/it][gradnorm] step 13909: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 340.6

 73%|███████▎  | 13910/19090 [6:47:19<2:20:14,  1.62s/it]
                                                         
{'loss': 0.002, 'grad_norm': 489.7396545410156, 'learning_rate': 1.0404287344137768e-05, 'recon_loss': 0.0019513269537128509, 'recon_static': 0.000981100402714219, 'recon_motion': 0.0009702265801024623, 'recon_norm': 1.4602770119905473, 'recon_norm_std': 0.036795619037002326, 'target_norm': 1.6206881433725357, 'recon_cos': 0.56481963545084, 'gradnorm/compressor.query': 0.3481624967800301, 'gradnorm/compressor.layer0': 13.132807271526325, 'gradnorm/compressor.layer1': 482.83291623500565, 'gradnorm/compressor.layer2': 77.18682628754566, 'gradnorm/compressor.layer3': 22.522965981016604, 'gradnorm/compressor.layer4': 6.272395179210227, 'gradnorm/compressor.layer5': 4.058779847153113, 'gradnorm/compressor.layer6': 2.5593475861996713, 'gradnorm/compressor.layer7': 2.8073754991082267, 'gradnorm/decoder.layer0': 2.512010619675609, 'gradnorm/decoder.layer1': 1.531551175763495, 'gradnorm/decoder.layer2': 0.8446453430291281, 'gradnorm/max_param': 340.6263732910156, 'epoch': 7.29}

 73%|███████▎  | 13910/19090 [6:47:19<2:20:14,  1.62s/it]
 73%|███████▎  | 13911/19090 [6:47:22<2:48:52,  1.96s/it]
 73%|███████▎  | 13912/19090 [6:47:23<2:31:34,  1.76s/it]
 73%|███████▎  | 13913/19090 [6:47:24<2:19:25,  1.62s/it]
 73%|███████▎  | 13914/19090 [6:47:27<2:57:25,  2.06s/it]
 73%|███████▎  | 13915/19090 [6:47:29<2:37:41,  1.83s/it]
 73%|███████▎  | 13916/19090 [6:47:30<2:23:42,  1.67s/it]
 73%|███████▎  | 13917/19090 [6:47:32<2:40:17,  1.86s/it]
 73%|███████▎  | 13918/19090 [6:47:34<2:25:59,  1.69s/it]
 73%|███████▎  | 13919/19090 [6:47:35<2:16:02,  1.58s/it][gradnorm] step 13919: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 309.0

 73%|███████▎  | 13920/19090 [6:47:37<2:35:51,  1.81s/it]
                                                         
{'loss': 0.002, 'grad_norm': 448.0540771484375, 'learning_rate': 1.0367198284658108e-05, 'recon_loss': 0.0019626072113169356, 'recon_static': 0.000993776753603015, 'recon_motion': 0.0009688304766314104, 'recon_norm': 1.458516263961792, 'recon_norm_std': 0.036745004914700986, 'target_norm': 1.6202958792448043, 'recon_cos': 0.5627770513296128, 'gradnorm/compressor.query': 0.2866109297085851, 'gradnorm/compressor.layer0': 11.689428812892428, 'gradnorm/compressor.layer1': 437.8240585554887, 'gradnorm/compressor.layer2': 91.13618222713468, 'gradnorm/compressor.layer3': 23.45924642173702, 'gradnorm/compressor.layer4': 6.423803183860064, 'gradnorm/compressor.layer5': 3.2236839921338656, 'gradnorm/compressor.layer6': 2.255655175417054, 'gradnorm/compressor.layer7': 2.2720474240699513, 'gradnorm/decoder.layer0': 2.2512373171231443, 'gradnorm/decoder.layer1': 1.427070849658726, 'gradnorm/decoder.layer2': 0.8247120542582195, 'gradnorm/max_param': 308.9722595214844, 'epoch': 7.29}

 73%|███████▎  | 13920/19090 [6:47:37<2:35:51,  1.81s/it]
 73%|███████▎  | 13921/19090 [6:47:39<2:22:42,  1.66s/it]
 73%|███████▎  | 13922/19090 [6:47:40<2:13:22,  1.55s/it]
 73%|███████▎  | 13923/19090 [6:47:43<2:44:32,  1.91s/it]
 73%|███████▎  | 13924/19090 [6:47:44<2:28:29,  1.72s/it]
 73%|███████▎  | 13925/19090 [6:47:45<2:17:28,  1.60s/it]
 73%|███████▎  | 13926/19090 [6:47:48<2:45:52,  1.93s/it]
 73%|███████▎  | 13927/19090 [6:47:49<2:29:22,  1.74s/it]
 73%|███████▎  | 13928/19090 [6:47:51<2:17:41,  1.60s/it]
 73%|███████▎  | 13929/19090 [6:47:54<2:54:43,  2.03s/it][gradnorm] step 13929: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 290.5

 73%|███████▎  | 13930/19090 [6:47:55<2:35:45,  1.81s/it]
                                                         
{'loss': 0.0021, 'grad_norm': 420.70562744140625, 'learning_rate': 1.0330158150111097e-05, 'recon_loss': 0.0020522347069345413, 'recon_static': 0.0010877876222366468, 'recon_motion': 0.0009644470235798507, 'recon_norm': 1.5001067787408828, 'recon_norm_std': 0.03618124881759286, 'target_norm': 1.6196032792329789, 'recon_cos': 0.5490523532032967, 'gradnorm/compressor.query': 0.26216065406896993, 'gradnorm/compressor.layer0': 9.483966749476496, 'gradnorm/compressor.layer1': 411.75650799349853, 'gradnorm/compressor.layer2': 82.86724900122636, 'gradnorm/compressor.layer3': 20.49102363379016, 'gradnorm/compressor.layer4': 6.185684400113114, 'gradnorm/compressor.layer5': 3.4202895301824783, 'gradnorm/compressor.layer6': 2.4918868023985397, 'gradnorm/compressor.layer7': 2.469448039201605, 'gradnorm/decoder.layer0': 2.7571405126983817, 'gradnorm/decoder.layer1': 1.5276736320294086, 'gradnorm/decoder.layer2': 0.835963061646, 'gradnorm/max_param': 290.521484375, 'epoch': 7.3}

 73%|███████▎  | 13930/19090 [6:47:55<2:35:45,  1.81s/it]
 73%|███████▎  | 13931/19090 [6:47:56<2:22:18,  1.66s/it]
 73%|███████▎  | 13932/19090 [6:47:59<2:45:22,  1.92s/it]
 73%|███████▎  | 13933/19090 [6:48:00<2:28:56,  1.73s/it]
 73%|███████▎  | 13934/19090 [6:48:01<2:17:37,  1.60s/it]
 73%|███████▎  | 13935/19090 [6:48:04<2:41:30,  1.88s/it]
 73%|███████▎  | 13936/19090 [6:48:05<2:26:18,  1.70s/it]
 73%|███████▎  | 13937/19090 [6:48:06<2:15:46,  1.58s/it]
 73%|███████▎  | 13938/19090 [6:48:09<2:50:10,  1.98s/it]
 73%|███████▎  | 13939/19090 [6:48:11<2:32:24,  1.78s/it][gradnorm] step 13939: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 339.4

 73%|███████▎  | 13940/19090 [6:48:12<2:19:54,  1.63s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 490.1949157714844, 'learning_rate': 1.0293167064340827e-05, 'recon_loss': 0.0018202265113359318, 'recon_static': 0.000847785230143927, 'recon_motion': 0.0009724413233925588, 'recon_norm': 1.4005950063467025, 'recon_norm_std': 0.0376118041574955, 'target_norm': 1.620883622765541, 'recon_cos': 0.5841392800211906, 'gradnorm/compressor.query': 0.28812599550186924, 'gradnorm/compressor.layer0': 10.556944339699054, 'gradnorm/compressor.layer1': 481.0327269271811, 'gradnorm/compressor.layer2': 90.11315586181077, 'gradnorm/compressor.layer3': 24.51119676984059, 'gradnorm/compressor.layer4': 6.349388886128721, 'gradnorm/compressor.layer5': 3.115116664850985, 'gradnorm/compressor.layer6': 1.8980234581569937, 'gradnorm/compressor.layer7': 2.2283363102978853, 'gradnorm/decoder.layer0': 2.2053875499105122, 'gradnorm/decoder.layer1': 1.2987223417184295, 'gradnorm/decoder.layer2': 0.8295602417099402, 'gradnorm/max_param': 339.4287109375, 'epoch': 7.3}

 73%|███████▎  | 13940/19090 [6:48:12<2:19:54,  1.63s/it]
 73%|███████▎  | 13941/19090 [6:48:14<2:45:13,  1.93s/it]
 73%|███████▎  | 13942/19090 [6:48:16<2:28:50,  1.73s/it]
 73%|███████▎  | 13943/19090 [6:48:17<2:17:30,  1.60s/it]
 73%|███████▎  | 13944/19090 [6:48:20<2:49:10,  1.97s/it]
 73%|███████▎  | 13945/19090 [6:48:21<2:31:32,  1.77s/it]
 73%|███████▎  | 13946/19090 [6:48:22<2:19:16,  1.62s/it]
 73%|███████▎  | 13947/19090 [6:48:25<2:44:43,  1.92s/it]
 73%|███████▎  | 13948/19090 [6:48:26<2:28:31,  1.73s/it]
 73%|███████▎  | 13949/19090 [6:48:28<2:17:11,  1.60s/it][gradnorm] step 13949: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 341.1

 73%|███████▎  | 13950/19090 [6:48:31<2:55:30,  2.05s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 491.6468505859375, 'learning_rate': 1.025622515102737e-05, 'recon_loss': 0.0016315354238031431, 'recon_static': 0.0006764828765881247, 'recon_motion': 0.000955052554490976, 'recon_norm': 1.3390336841344834, 'recon_norm_std': 0.03971142638474703, 'target_norm': 1.6157118141651154, 'recon_cos': 0.6185126081109047, 'gradnorm/compressor.query': 0.2650409054674519, 'gradnorm/compressor.layer0': 9.003543156317727, 'gradnorm/compressor.layer1': 483.4508926716342, 'gradnorm/compressor.layer2': 86.32815752261831, 'gradnorm/compressor.layer3': 19.309638048964857, 'gradnorm/compressor.layer4': 7.128976348114269, 'gradnorm/compressor.layer5': 3.6341115946286453, 'gradnorm/compressor.layer6': 2.5228284921733444, 'gradnorm/compressor.layer7': 2.5658491624486572, 'gradnorm/decoder.layer0': 2.4177830737284243, 'gradnorm/decoder.layer1': 1.3695396897468843, 'gradnorm/decoder.layer2': 0.7983516773898381, 'gradnorm/max_param': 341.1493225097656, 'epoch': 7.31}

 73%|███████▎  | 13950/19090 [6:48:31<2:55:30,  2.05s/it]
 73%|███████▎  | 13951/19090 [6:48:32<2:36:07,  1.82s/it]
 73%|███████▎  | 13952/19090 [6:48:33<2:22:24,  1.66s/it]
 73%|███████▎  | 13953/19090 [6:48:36<2:54:46,  2.04s/it]
 73%|███████▎  | 13954/19090 [6:48:38<2:35:26,  1.82s/it]
 73%|███████▎  | 13955/19090 [6:48:39<2:22:00,  1.66s/it]
 73%|███████▎  | 13956/19090 [6:48:42<2:57:15,  2.07s/it]
 73%|███████▎  | 13957/19090 [6:48:43<2:37:10,  1.84s/it]
 73%|███████▎  | 13958/19090 [6:48:44<2:23:01,  1.67s/it]
 73%|███████▎  | 13959/19090 [6:48:47<2:51:36,  2.01s/it][gradnorm] step 13959: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 341.0

 73%|███████▎  | 13960/19090 [6:48:49<2:33:15,  1.79s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 491.4914245605469, 'learning_rate': 1.0219332533686446e-05, 'recon_loss': 0.001645398820983246, 'recon_static': 0.0006854196210042574, 'recon_motion': 0.0009599791679647751, 'recon_norm': 1.3355909287929535, 'recon_norm_std': 0.04009850276634097, 'target_norm': 1.6168152570724488, 'recon_cos': 0.6148505717515945, 'gradnorm/compressor.query': 0.2559408829713471, 'gradnorm/compressor.layer0': 9.863342357483972, 'gradnorm/compressor.layer1': 483.28150763814665, 'gradnorm/compressor.layer2': 84.84038897578499, 'gradnorm/compressor.layer3': 25.09675247643171, 'gradnorm/compressor.layer4': 6.362968010917466, 'gradnorm/compressor.layer5': 3.195727183847448, 'gradnorm/compressor.layer6': 2.536951214307369, 'gradnorm/compressor.layer7': 2.922935578948267, 'gradnorm/decoder.layer0': 2.949432486563788, 'gradnorm/decoder.layer1': 1.5323955377070162, 'gradnorm/decoder.layer2': 1.0024527034875472, 'gradnorm/max_param': 341.0220947265625, 'epoch': 7.31}

 73%|███████▎  | 13960/19090 [6:48:49<2:33:15,  1.79s/it]
 73%|███████▎  | 13961/19090 [6:48:50<2:20:28,  1.64s/it]
 73%|███████▎  | 13962/19090 [6:48:52<2:41:16,  1.89s/it]
 73%|███████▎  | 13963/19090 [6:48:54<2:25:45,  1.71s/it]
 73%|███████▎  | 13964/19090 [6:48:55<2:15:02,  1.58s/it]
 73%|███████▎  | 13965/19090 [6:48:58<2:49:40,  1.99s/it]
 73%|███████▎  | 13966/19090 [6:48:59<2:32:03,  1.78s/it]
 73%|███████▎  | 13967/19090 [6:49:00<2:19:49,  1.64s/it]
 73%|███████▎  | 13968/19090 [6:49:03<2:44:24,  1.93s/it]
 73%|███████▎  | 13969/19090 [6:49:04<2:28:01,  1.73s/it][gradnorm] step 13969: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 315.6

 73%|███████▎  | 13970/19090 [6:49:06<2:16:33,  1.60s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 454.1336364746094, 'learning_rate': 1.0182489335668921e-05, 'recon_loss': 0.0017411507171345875, 'recon_static': 0.0007691493214224465, 'recon_motion': 0.0009720014175400137, 'recon_norm': 1.374056062102318, 'recon_norm_std': 0.03830725355073809, 'target_norm': 1.6200536996126176, 'recon_cos': 0.5979399919509888, 'gradnorm/compressor.query': 0.2128246100028825, 'gradnorm/compressor.layer0': 7.9314032741973985, 'gradnorm/compressor.layer1': 447.30968369240566, 'gradnorm/compressor.layer2': 74.60745252758768, 'gradnorm/compressor.layer3': 21.059621551967865, 'gradnorm/compressor.layer4': 6.475217083565375, 'gradnorm/compressor.layer5': 3.6572685453197784, 'gradnorm/compressor.layer6': 2.558378021230995, 'gradnorm/compressor.layer7': 2.638294342614001, 'gradnorm/decoder.layer0': 2.6365709955707732, 'gradnorm/decoder.layer1': 1.4826666387402465, 'gradnorm/decoder.layer2': 0.8588452874058748, 'gradnorm/max_param': 315.63653564453125, 'epoch': 7.32}

 73%|███████▎  | 13970/19090 [6:49:06<2:16:33,  1.60s/it]
 73%|███████▎  | 13971/19090 [6:49:09<2:59:12,  2.10s/it]
 73%|███████▎  | 13972/19090 [6:49:10<2:38:20,  1.86s/it]
 73%|███████▎  | 13973/19090 [6:49:11<2:23:48,  1.69s/it]
 73%|███████▎  | 13974/19090 [6:49:14<2:53:37,  2.04s/it]
 73%|███████▎  | 13975/19090 [6:49:16<2:34:23,  1.81s/it]
 73%|███████▎  | 13976/19090 [6:49:17<2:20:59,  1.65s/it]
 73%|███████▎  | 13977/19090 [6:49:20<2:51:59,  2.02s/it]
 73%|███████▎  | 13978/19090 [6:49:21<2:33:17,  1.80s/it]
 73%|███████▎  | 13979/19090 [6:49:22<2:20:17,  1.65s/it][gradnorm] step 13979: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 382.0

 73%|███████▎  | 13980/19090 [6:49:25<2:47:52,  1.97s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 549.8507080078125, 'learning_rate': 1.0145695680160407e-05, 'recon_loss': 0.0017175380809931085, 'recon_static': 0.0007420874535455368, 'recon_motion': 0.0009754506041645073, 'recon_norm': 1.3657123178243638, 'recon_norm_std': 0.03968390552327037, 'target_norm': 1.6208306014537812, 'recon_cos': 0.6032632276415825, 'gradnorm/compressor.query': 0.2993164933719477, 'gradnorm/compressor.layer0': 12.643444524856546, 'gradnorm/compressor.layer1': 541.2644051200855, 'gradnorm/compressor.layer2': 93.34926854292968, 'gradnorm/compressor.layer3': 20.34661423169367, 'gradnorm/compressor.layer4': 6.4205101029193505, 'gradnorm/compressor.layer5': 4.20692107925036, 'gradnorm/compressor.layer6': 2.148073366691693, 'gradnorm/compressor.layer7': 2.5972649530532386, 'gradnorm/decoder.layer0': 2.6774068753158846, 'gradnorm/decoder.layer1': 1.5038295974022693, 'gradnorm/decoder.layer2': 0.9525049725600478, 'gradnorm/max_param': 381.97064208984375, 'epoch': 7.32}

 73%|███████▎  | 13980/19090 [6:49:25<2:47:52,  1.97s/it]
 73%|███████▎  | 13981/19090 [6:49:26<2:30:42,  1.77s/it]
 73%|███████▎  | 13982/19090 [6:49:28<2:18:33,  1.63s/it]
 73%|███████▎  | 13983/19090 [6:49:30<2:47:34,  1.97s/it]
 73%|███████▎  | 13984/19090 [6:49:32<2:30:06,  1.76s/it]
 73%|███████▎  | 13985/19090 [6:49:33<2:18:00,  1.62s/it]
 73%|███████▎  | 13986/19090 [6:49:36<2:48:22,  1.98s/it]
 73%|███████▎  | 13987/19090 [6:49:37<2:30:55,  1.77s/it]
 73%|███████▎  | 13988/19090 [6:49:38<2:18:31,  1.63s/it]
 73%|███████▎  | 13989/19090 [6:49:41<2:36:33,  1.84s/it][gradnorm] step 13989: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 361.1

 73%|███████▎  | 13990/19090 [6:49:42<2:22:38,  1.68s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 518.0868530273438, 'learning_rate': 1.0108951690180923e-05, 'recon_loss': 0.0016716591926524416, 'recon_static': 0.0006944074586499483, 'recon_motion': 0.0009772517514647916, 'recon_norm': 1.3373471289873122, 'recon_norm_std': 0.04040643032640219, 'target_norm': 1.6229681223630905, 'recon_cos': 0.6111576482653618, 'gradnorm/compressor.query': 0.23826470864836646, 'gradnorm/compressor.layer0': 8.905863649788987, 'gradnorm/compressor.layer1': 511.67106743492934, 'gradnorm/compressor.layer2': 76.68975075649777, 'gradnorm/compressor.layer3': 23.974901427065348, 'gradnorm/compressor.layer4': 5.796604859671731, 'gradnorm/compressor.layer5': 3.647685919057028, 'gradnorm/compressor.layer6': 2.1046615497034313, 'gradnorm/compressor.layer7': 2.846016954908982, 'gradnorm/decoder.layer0': 2.938032872991853, 'gradnorm/decoder.layer1': 1.4594055956751668, 'gradnorm/decoder.layer2': 0.9923248195989876, 'gradnorm/max_param': 361.0779113769531, 'epoch': 7.33}

 73%|███████▎  | 13990/19090 [6:49:42<2:22:38,  1.68s/it]
 73%|███████▎  | 13991/19090 [6:49:43<2:13:01,  1.57s/it]
 73%|███████▎  | 13992/19090 [6:49:46<2:41:10,  1.90s/it]
 73%|███████▎  | 13993/19090 [6:49:47<2:25:47,  1.72s/it]
 73%|███████▎  | 13994/19090 [6:49:49<2:15:09,  1.59s/it]
 73%|███████▎  | 13995/19090 [6:49:51<2:45:02,  1.94s/it]
 73%|███████▎  | 13996/19090 [6:49:53<2:28:38,  1.75s/it]
 73%|███████▎  | 13997/19090 [6:49:54<2:16:49,  1.61s/it]
 73%|███████▎  | 13998/19090 [6:49:57<2:44:28,  1.94s/it]
 73%|███████▎  | 13999/19090 [6:49:58<2:27:45,  1.74s/it][gradnorm] step 13999: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 407.5

 73%|███████▎  | 14000/19090 [6:49:59<2:16:08,  1.60s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 586.2880249023438, 'learning_rate': 1.0072257488584385e-05, 'recon_loss': 0.0016603764932369813, 'recon_static': 0.0006931120282388292, 'recon_motion': 0.0009672644693637267, 'recon_norm': 1.3437204569578172, 'recon_norm_std': 0.03934174720197916, 'target_norm': 1.620671346783638, 'recon_cos': 0.6141070008277894, 'gradnorm/compressor.query': 0.2513028706487205, 'gradnorm/compressor.layer0': 9.631801431503765, 'gradnorm/compressor.layer1': 577.5348474334687, 'gradnorm/compressor.layer2': 96.87886585028491, 'gradnorm/compressor.layer3': 24.919997725142228, 'gradnorm/compressor.layer4': 7.234232363520285, 'gradnorm/compressor.layer5': 3.7716843854752344, 'gradnorm/compressor.layer6': 2.144681775867639, 'gradnorm/compressor.layer7': 2.584173609348957, 'gradnorm/decoder.layer0': 2.66181148169338, 'gradnorm/decoder.layer1': 1.4495692895945655, 'gradnorm/decoder.layer2': 0.8929508023865526, 'gradnorm/max_param': 407.5001220703125, 'epoch': 7.33}

 73%|███████▎  | 14000/19090 [6:49:59<2:16:08,  1.60s/it]
 73%|███████▎  | 14001/19090 [6:50:03<3:17:40,  2.33s/it]
 73%|███████▎  | 14002/19090 [6:50:05<2:51:10,  2.02s/it]
 73%|███████▎  | 14003/19090 [6:50:06<2:32:26,  1.80s/it]
 73%|███████▎  | 14004/19090 [6:50:07<2:19:26,  1.64s/it]
 73%|███████▎  | 14005/19090 [6:50:08<2:10:16,  1.54s/it]
 73%|███████▎  | 14006/19090 [6:50:10<2:03:51,  1.46s/it]
 73%|███████▎  | 14007/19090 [6:50:12<2:17:20,  1.62s/it]
 73%|███████▎  | 14008/19090 [6:50:13<2:08:55,  1.52s/it]
 73%|███████▎  | 14009/19090 [6:50:14<2:02:58,  1.45s/it][gradnorm] step 14009: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 442.2

 73%|███████▎  | 14010/19090 [6:50:17<2:45:53,  1.96s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 635.4774780273438, 'learning_rate': 1.0035613198058266e-05, 'recon_loss': 0.0016338204557541757, 'recon_static': 0.0006492979038739577, 'recon_motion': 0.000984522588260006, 'recon_norm': 1.3231703281402587, 'recon_norm_std': 0.03907775850966573, 'target_norm': 1.62130469083786, 'recon_cos': 0.6187452301383018, 'gradnorm/compressor.query': 0.3512150955335909, 'gradnorm/compressor.layer0': 13.696770302652842, 'gradnorm/compressor.layer1': 626.6443119904624, 'gradnorm/compressor.layer2': 100.79559006505443, 'gradnorm/compressor.layer3': 26.742198911770863, 'gradnorm/compressor.layer4': 6.828631978014491, 'gradnorm/compressor.layer5': 3.9595116450960677, 'gradnorm/compressor.layer6': 2.3944483084572488, 'gradnorm/compressor.layer7': 2.463386019950705, 'gradnorm/decoder.layer0': 2.9379411731333205, 'gradnorm/decoder.layer1': 1.5116122107211216, 'gradnorm/decoder.layer2': 0.8718526639094004, 'gradnorm/max_param': 442.1628723144531, 'epoch': 7.34}

 73%|███████▎  | 14010/19090 [6:50:17<2:45:53,  1.96s/it]
 73%|███████▎  | 14011/19090 [6:50:19<2:28:45,  1.76s/it]
 73%|███████▎  | 14012/19090 [6:50:20<2:17:04,  1.62s/it]
 73%|███████▎  | 14013/19090 [6:50:23<2:42:15,  1.92s/it]
 73%|███████▎  | 14014/19090 [6:50:24<2:26:26,  1.73s/it]
 73%|███████▎  | 14015/19090 [6:50:25<2:15:29,  1.60s/it]
 73%|███████▎  | 14016/19090 [6:50:28<2:36:16,  1.85s/it]
 73%|███████▎  | 14017/19090 [6:50:29<2:22:29,  1.69s/it]
 73%|███████▎  | 14018/19090 [6:50:30<2:12:54,  1.57s/it]
 73%|███████▎  | 14019/19090 [6:50:33<2:40:20,  1.90s/it][gradnorm] step 14019: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 439.2

 73%|███████▎  | 14020/19090 [6:50:34<2:25:14,  1.72s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 630.5458984375, 'learning_rate': 9.999018941123147e-06, 'recon_loss': 0.0016030333703383804, 'recon_static': 0.000629550869052764, 'recon_motion': 0.00097348251438234, 'recon_norm': 1.314658087491989, 'recon_norm_std': 0.040991259180009364, 'target_norm': 1.6212750911712646, 'recon_cos': 0.6253996163606643, 'gradnorm/compressor.query': 0.30258616849696396, 'gradnorm/compressor.layer0': 10.617076477413455, 'gradnorm/compressor.layer1': 622.5237445270662, 'gradnorm/compressor.layer2': 96.31904483765139, 'gradnorm/compressor.layer3': 23.667170272985082, 'gradnorm/compressor.layer4': 7.646253139463401, 'gradnorm/compressor.layer5': 4.112193718262026, 'gradnorm/compressor.layer6': 2.694653947513592, 'gradnorm/compressor.layer7': 2.6677143204599623, 'gradnorm/decoder.layer0': 2.954324952813899, 'gradnorm/decoder.layer1': 1.6333879980683974, 'gradnorm/decoder.layer2': 0.9351370596769035, 'gradnorm/max_param': 439.15863037109375, 'epoch': 7.34}

 73%|███████▎  | 14020/19090 [6:50:34<2:25:14,  1.72s/it]
 73%|███████▎  | 14021/19090 [6:50:35<2:14:31,  1.59s/it]
 73%|███████▎  | 14022/19090 [6:50:38<2:40:16,  1.90s/it]
 73%|███████▎  | 14023/19090 [6:50:39<2:24:52,  1.72s/it]
 73%|███████▎  | 14024/19090 [6:50:41<2:14:03,  1.59s/it]
 73%|███████▎  | 14025/19090 [6:50:43<2:45:13,  1.96s/it]
 73%|███████▎  | 14026/19090 [6:50:45<2:28:09,  1.76s/it]
 73%|███████▎  | 14027/19090 [6:50:46<2:16:14,  1.61s/it]
 73%|███████▎  | 14028/19090 [6:50:49<2:46:50,  1.98s/it]
 73%|███████▎  | 14029/19090 [6:50:50<2:29:13,  1.77s/it][gradnorm] step 14029: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 328.5

 73%|███████▎  | 14030/19090 [6:50:51<2:17:02,  1.63s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 475.1332702636719, 'learning_rate': 9.962474840132328e-06, 'recon_loss': 0.0016380963934352621, 'recon_static': 0.0006603200657991692, 'recon_motion': 0.0009777762898011133, 'recon_norm': 1.328616228699684, 'recon_norm_std': 0.039718132745474574, 'target_norm': 1.6199785977602006, 'recon_cos': 0.6172367721796036, 'gradnorm/compressor.query': 0.23878042122125795, 'gradnorm/compressor.layer0': 8.820949102358899, 'gradnorm/compressor.layer1': 465.6290603849807, 'gradnorm/compressor.layer2': 90.58135802622414, 'gradnorm/compressor.layer3': 23.935262470898063, 'gradnorm/compressor.layer4': 7.126492795663965, 'gradnorm/compressor.layer5': 3.9339141260654538, 'gradnorm/compressor.layer6': 2.208569987830089, 'gradnorm/compressor.layer7': 2.32183859631019, 'gradnorm/decoder.layer0': 2.4518464273523155, 'gradnorm/decoder.layer1': 1.5036777867761737, 'gradnorm/decoder.layer2': 0.81367305292199, 'gradnorm/max_param': 328.5466003417969, 'epoch': 7.35}

 73%|███████▎  | 14030/19090 [6:50:51<2:17:02,  1.63s/it]
 73%|███████▎  | 14031/19090 [6:50:54<2:42:53,  1.93s/it]
 74%|███████▎  | 14032/19090 [6:50:55<2:26:29,  1.74s/it]
 74%|███████▎  | 14033/19090 [6:50:57<2:15:02,  1.60s/it]
 74%|███████▎  | 14034/19090 [6:51:00<2:51:07,  2.03s/it]
 74%|███████▎  | 14035/19090 [6:51:01<2:32:15,  1.81s/it]
 74%|███████▎  | 14036/19090 [6:51:02<2:19:04,  1.65s/it]
 74%|███████▎  | 14037/19090 [6:51:05<2:49:55,  2.02s/it]
 74%|███████▎  | 14038/19090 [6:51:06<2:31:28,  1.80s/it]
 74%|███████▎  | 14039/19090 [6:51:08<2:18:33,  1.65s/it][gradnorm] step 14039: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 419.3

 74%|███████▎  | 14040/19090 [6:51:11<2:51:52,  2.04s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 607.071044921875, 'learning_rate': 9.925981017271411e-06, 'recon_loss': 0.001601899033994414, 'recon_static': 0.0006292566700722091, 'recon_motion': 0.0009726423944812268, 'recon_norm': 1.3117600500583648, 'recon_norm_std': 0.040644025057554246, 'target_norm': 1.6202369660139084, 'recon_cos': 0.6245289593935013, 'gradnorm/compressor.query': 0.3040594205278923, 'gradnorm/compressor.layer0': 12.21596548055332, 'gradnorm/compressor.layer1': 594.350328089419, 'gradnorm/compressor.layer2': 118.87137711356338, 'gradnorm/compressor.layer3': 29.945304347116032, 'gradnorm/compressor.layer4': 7.961512491502086, 'gradnorm/compressor.layer5': 4.498960586880561, 'gradnorm/compressor.layer6': 2.521185847663891, 'gradnorm/compressor.layer7': 2.4620068361263363, 'gradnorm/decoder.layer0': 2.6882669108683803, 'gradnorm/decoder.layer1': 1.5473690592596838, 'gradnorm/decoder.layer2': 0.9689611696999952, 'gradnorm/max_param': 419.3348083496094, 'epoch': 7.35}

 74%|███████▎  | 14040/19090 [6:51:11<2:51:52,  2.04s/it]
 74%|███████▎  | 14041/19090 [6:51:12<2:33:02,  1.82s/it]
 74%|███████▎  | 14042/19090 [6:51:13<2:19:39,  1.66s/it]
 74%|███████▎  | 14043/19090 [6:51:16<2:49:14,  2.01s/it]
 74%|███████▎  | 14044/19090 [6:51:17<2:30:48,  1.79s/it]
 74%|███████▎  | 14045/19090 [6:51:19<2:17:57,  1.64s/it]
 74%|███████▎  | 14046/19090 [6:51:21<2:44:51,  1.96s/it]
 74%|███████▎  | 14047/19090 [6:51:23<2:27:50,  1.76s/it]
 74%|███████▎  | 14048/19090 [6:51:24<2:15:54,  1.62s/it]
 74%|███████▎  | 14049/19090 [6:51:27<2:53:32,  2.07s/it][gradnorm] step 14049: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 348.8

 74%|███████▎  | 14050/19090 [6:51:28<2:33:54,  1.83s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 502.2137145996094, 'learning_rate': 9.889537594557891e-06, 'recon_loss': 0.001664880133466795, 'recon_static': 0.0006893011537613347, 'recon_motion': 0.0009755790058989077, 'recon_norm': 1.3467252999544144, 'recon_norm_std': 0.03906140858307481, 'target_norm': 1.61853948533535, 'recon_cos': 0.6124926194548607, 'gradnorm/compressor.query': 0.2504341855297477, 'gradnorm/compressor.layer0': 9.303462821508527, 'gradnorm/compressor.layer1': 494.2467279861345, 'gradnorm/compressor.layer2': 84.08030422514985, 'gradnorm/compressor.layer3': 26.524830826805005, 'gradnorm/compressor.layer4': 6.490388072303656, 'gradnorm/compressor.layer5': 3.6920569759613837, 'gradnorm/compressor.layer6': 2.4379657398223142, 'gradnorm/compressor.layer7': 2.525100679562114, 'gradnorm/decoder.layer0': 2.7367804709533075, 'gradnorm/decoder.layer1': 1.5773425286917944, 'gradnorm/decoder.layer2': 1.0250706927414774, 'gradnorm/max_param': 348.7624206542969, 'epoch': 7.36}

 74%|███████▎  | 14050/19090 [6:51:28<2:33:54,  1.83s/it]
 74%|███████▎  | 14051/19090 [6:51:30<2:20:10,  1.67s/it]
 74%|███████▎  | 14052/19090 [6:51:33<2:51:54,  2.05s/it]
 74%|███████▎  | 14053/19090 [6:51:34<2:32:47,  1.82s/it]
 74%|███████▎  | 14054/19090 [6:51:35<2:19:23,  1.66s/it]
 74%|███████▎  | 14055/19090 [6:51:38<2:48:48,  2.01s/it]
 74%|███████▎  | 14056/19090 [6:51:39<2:30:38,  1.80s/it]
 74%|███████▎  | 14057/19090 [6:51:41<2:17:48,  1.64s/it]
 74%|███████▎  | 14058/19090 [6:51:43<2:43:52,  1.95s/it]
 74%|███████▎  | 14059/19090 [6:51:45<2:27:08,  1.75s/it][gradnorm] step 14059: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 340.1

 74%|███████▎  | 14060/19090 [6:51:46<2:15:21,  1.61s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 489.343994140625, 'learning_rate': 9.85314469384075e-06, 'recon_loss': 0.001751791822607629, 'recon_static': 0.0007634579087607563, 'recon_motion': 0.000988333931309171, 'recon_norm': 1.3737423509359359, 'recon_norm_std': 0.038062725216150284, 'target_norm': 1.6239996522665023, 'recon_cos': 0.5982495486736298, 'gradnorm/compressor.query': 0.2856070078143302, 'gradnorm/compressor.layer0': 10.422180994851947, 'gradnorm/compressor.layer1': 481.9867412595496, 'gradnorm/compressor.layer2': 79.07451736732258, 'gradnorm/compressor.layer3': 26.65874223585307, 'gradnorm/compressor.layer4': 6.504435273154484, 'gradnorm/compressor.layer5': 3.4407248888752053, 'gradnorm/compressor.layer6': 2.2808938989575998, 'gradnorm/compressor.layer7': 2.358659850612961, 'gradnorm/decoder.layer0': 2.4479290792813346, 'gradnorm/decoder.layer1': 1.5367166679147133, 'gradnorm/decoder.layer2': 0.9261482573626839, 'gradnorm/max_param': 340.1109619140625, 'epoch': 7.37}

 74%|███████▎  | 14060/19090 [6:51:46<2:15:21,  1.61s/it]
 74%|███████▎  | 14061/19090 [6:51:49<2:46:51,  1.99s/it]
 74%|███████▎  | 14062/19090 [6:51:50<2:29:11,  1.78s/it]
 74%|███████▎  | 14063/19090 [6:51:51<2:16:45,  1.63s/it]
 74%|███████▎  | 14064/19090 [6:51:54<2:46:03,  1.98s/it]
 74%|███████▎  | 14065/19090 [6:51:55<2:28:32,  1.77s/it]
 74%|███████▎  | 14066/19090 [6:51:57<2:16:22,  1.63s/it]
 74%|███████▎  | 14067/19090 [6:51:59<2:45:22,  1.98s/it]
 74%|███████▎  | 14068/19090 [6:52:01<2:28:07,  1.77s/it]
 74%|███████▎  | 14069/19090 [6:52:02<2:16:03,  1.63s/it][gradnorm] step 14069: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 424.8

 74%|███████▎  | 14070/19090 [6:52:05<2:41:58,  1.94s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 609.7677612304688, 'learning_rate': 9.816802436800052e-06, 'recon_loss': 0.001916291654924862, 'recon_static': 0.0009363294666400179, 'recon_motion': 0.000979962151905056, 'recon_norm': 1.438936895132065, 'recon_norm_std': 0.037593257054686545, 'target_norm': 1.622615212202072, 'recon_cos': 0.5696955412626267, 'gradnorm/compressor.query': 0.4871664367145166, 'gradnorm/compressor.layer0': 18.722472468715154, 'gradnorm/compressor.layer1': 602.0726700324471, 'gradnorm/compressor.layer2': 90.56519578458382, 'gradnorm/compressor.layer3': 26.172094215499836, 'gradnorm/compressor.layer4': 6.967343312429204, 'gradnorm/compressor.layer5': 3.4959621298111525, 'gradnorm/compressor.layer6': 2.9067469346340413, 'gradnorm/compressor.layer7': 2.6712037074175, 'gradnorm/decoder.layer0': 2.797914194734172, 'gradnorm/decoder.layer1': 1.4699354055057157, 'gradnorm/decoder.layer2': 1.0551323976649436, 'gradnorm/max_param': 424.7726745605469, 'epoch': 7.37}

 74%|███████▎  | 14070/19090 [6:52:05<2:41:58,  1.94s/it]
 74%|███████▎  | 14071/19090 [6:52:06<2:25:44,  1.74s/it]
 74%|███████▎  | 14072/19090 [6:52:07<2:14:46,  1.61s/it]
 74%|███████▎  | 14073/19090 [6:52:10<2:42:53,  1.95s/it]
 74%|███████▎  | 14074/19090 [6:52:11<2:26:19,  1.75s/it]
 74%|███████▎  | 14075/19090 [6:52:13<2:15:05,  1.62s/it]
 74%|███████▎  | 14076/19090 [6:52:15<2:38:07,  1.89s/it]
 74%|███████▎  | 14077/19090 [6:52:16<2:23:08,  1.71s/it]
 74%|███████▎  | 14078/19090 [6:52:18<2:12:27,  1.59s/it]
 74%|███████▍  | 14079/19090 [6:52:20<2:41:49,  1.94s/it][gradnorm] step 14079: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 307.4

 74%|███████▍  | 14080/19090 [6:52:22<2:26:02,  1.75s/it]
                                                         
{'loss': 0.002, 'grad_norm': 441.856689453125, 'learning_rate': 9.780510944946527e-06, 'recon_loss': 0.0019785120093729347, 'recon_static': 0.0010027332522440703, 'recon_motion': 0.0009757787338458002, 'recon_norm': 1.4652915209531785, 'recon_norm_std': 0.03646675040945411, 'target_norm': 1.620790383219719, 'recon_cos': 0.5609849505126476, 'gradnorm/compressor.query': 0.2868208830616857, 'gradnorm/compressor.layer0': 10.961961346292787, 'gradnorm/compressor.layer1': 435.54843301290845, 'gradnorm/compressor.layer2': 69.69392136894903, 'gradnorm/compressor.layer3': 21.98505310010538, 'gradnorm/compressor.layer4': 5.9746017131775, 'gradnorm/compressor.layer5': 3.6997982485641194, 'gradnorm/compressor.layer6': 2.606296390696896, 'gradnorm/compressor.layer7': 2.6935367650294713, 'gradnorm/decoder.layer0': 2.664058562245074, 'gradnorm/decoder.layer1': 1.6081161019605033, 'gradnorm/decoder.layer2': 0.9846230451722575, 'gradnorm/max_param': 307.3761291503906, 'epoch': 7.38}

 74%|███████▍  | 14080/19090 [6:52:22<2:26:02,  1.75s/it]
 74%|███████▍  | 14081/19090 [6:52:23<2:14:43,  1.61s/it]
 74%|███████▍  | 14082/19090 [6:52:26<2:40:48,  1.93s/it]
 74%|███████▍  | 14083/19090 [6:52:27<2:24:53,  1.74s/it]
 74%|███████▍  | 14084/19090 [6:52:28<2:13:40,  1.60s/it]
 74%|███████▍  | 14085/19090 [6:52:31<2:51:50,  2.06s/it]
 74%|███████▍  | 14086/19090 [6:52:33<2:32:28,  1.83s/it]
 74%|███████▍  | 14087/19090 [6:52:34<2:19:00,  1.67s/it]
 74%|███████▍  | 14088/19090 [6:52:37<2:51:34,  2.06s/it]
 74%|███████▍  | 14089/19090 [6:52:38<2:32:10,  1.83s/it][gradnorm] step 14089: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 378.7

 74%|███████▍  | 14090/19090 [6:52:40<2:18:49,  1.67s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 547.3158569335938, 'learning_rate': 9.744270339621181e-06, 'recon_loss': 0.0017968283063964918, 'recon_static': 0.0008215121459215879, 'recon_motion': 0.0009753161735716276, 'recon_norm': 1.3948306769132615, 'recon_norm_std': 0.03758348487317562, 'target_norm': 1.621983414888382, 'recon_cos': 0.588164322078228, 'gradnorm/compressor.query': 0.39808588782021853, 'gradnorm/compressor.layer0': 11.673063286650194, 'gradnorm/compressor.layer1': 536.8969000189887, 'gradnorm/compressor.layer2': 101.4163226892619, 'gradnorm/compressor.layer3': 27.91625441820925, 'gradnorm/compressor.layer4': 6.826437595418318, 'gradnorm/compressor.layer5': 4.057900980035064, 'gradnorm/compressor.layer6': 3.138969174673944, 'gradnorm/compressor.layer7': 2.9655937683583335, 'gradnorm/decoder.layer0': 3.0737049880422203, 'gradnorm/decoder.layer1': 1.748005684308689, 'gradnorm/decoder.layer2': 1.054432138682267, 'gradnorm/max_param': 378.7344055175781, 'epoch': 7.38}

 74%|███████▍  | 14090/19090 [6:52:40<2:18:49,  1.67s/it]
 74%|███████▍  | 14091/19090 [6:52:42<2:47:37,  2.01s/it]
 74%|███████▍  | 14092/19090 [6:52:44<2:29:40,  1.80s/it]
 74%|███████▍  | 14093/19090 [6:52:45<2:16:53,  1.64s/it]
 74%|███████▍  | 14094/19090 [6:52:48<2:50:52,  2.05s/it]
 74%|███████▍  | 14095/19090 [6:52:49<2:31:53,  1.82s/it]
 74%|███████▍  | 14096/19090 [6:52:51<2:18:42,  1.67s/it]
 74%|███████▍  | 14097/19090 [6:52:53<2:43:19,  1.96s/it]
 74%|███████▍  | 14098/19090 [6:52:55<2:26:39,  1.76s/it]
 74%|███████▍  | 14099/19090 [6:52:56<2:15:03,  1.62s/it][gradnorm] step 14099: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 443.2

 74%|███████▍  | 14100/19090 [6:52:58<2:40:57,  1.94s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 634.9821166992188, 'learning_rate': 9.708080741994868e-06, 'recon_loss': 0.0019247969554271548, 'recon_static': 0.0009409160047653131, 'recon_motion': 0.0009838809521170333, 'recon_norm': 1.4411170870065688, 'recon_norm_std': 0.036694733798503874, 'target_norm': 1.6228024661540985, 'recon_cos': 0.5672825634479522, 'gradnorm/compressor.query': 0.37488623721359154, 'gradnorm/compressor.layer0': 15.131262121309943, 'gradnorm/compressor.layer1': 628.1973714526351, 'gradnorm/compressor.layer2': 86.80897614788317, 'gradnorm/compressor.layer3': 26.692721925809547, 'gradnorm/compressor.layer4': 7.252915946975169, 'gradnorm/compressor.layer5': 4.007749441740159, 'gradnorm/compressor.layer6': 2.5102901400474087, 'gradnorm/compressor.layer7': 2.623951657175554, 'gradnorm/decoder.layer0': 2.8027175317982573, 'gradnorm/decoder.layer1': 1.539281916589955, 'gradnorm/decoder.layer2': 0.9140153856443279, 'gradnorm/max_param': 443.2352600097656, 'epoch': 7.39}

 74%|███████▍  | 14100/19090 [6:52:58<2:40:57,  1.94s/it]
 74%|███████▍  | 14101/19090 [6:53:00<2:25:23,  1.75s/it]
 74%|███████▍  | 14102/19090 [6:53:01<2:13:55,  1.61s/it]
 74%|███████▍  | 14103/19090 [6:53:04<2:35:04,  1.87s/it]
 74%|███████▍  | 14104/19090 [6:53:05<2:20:36,  1.69s/it]
 74%|███████▍  | 14105/19090 [6:53:06<2:10:28,  1.57s/it]
 74%|███████▍  | 14106/19090 [6:53:09<2:41:42,  1.95s/it]
 74%|███████▍  | 14107/19090 [6:53:10<2:25:22,  1.75s/it]
 74%|███████▍  | 14108/19090 [6:53:12<2:13:53,  1.61s/it]
 74%|███████▍  | 14109/19090 [6:53:14<2:42:43,  1.96s/it][gradnorm] step 14109: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 457.3

 74%|███████▍  | 14110/19090 [6:53:16<2:25:57,  1.76s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 657.163330078125, 'learning_rate': 9.671942273067911e-06, 'recon_loss': 0.001899000399862416, 'recon_static': 0.0009349940635729581, 'recon_motion': 0.0009640063988626934, 'recon_norm': 1.4363990485668183, 'recon_norm_std': 0.03712926795706153, 'target_norm': 1.6176029413938522, 'recon_cos': 0.5704792365431786, 'gradnorm/compressor.query': 0.3512042127238467, 'gradnorm/compressor.layer0': 14.735962993529798, 'gradnorm/compressor.layer1': 648.1010482555942, 'gradnorm/compressor.layer2': 104.96175363703914, 'gradnorm/compressor.layer3': 22.625929333900437, 'gradnorm/compressor.layer4': 6.6430317904107765, 'gradnorm/compressor.layer5': 3.779038468232993, 'gradnorm/compressor.layer6': 2.6775503283132953, 'gradnorm/compressor.layer7': 2.6453652355798165, 'gradnorm/decoder.layer0': 2.625987820727075, 'gradnorm/decoder.layer1': 1.4639026925105536, 'gradnorm/decoder.layer2': 0.9318219214404528, 'gradnorm/max_param': 457.3191833496094, 'epoch': 7.39}

 74%|███████▍  | 14110/19090 [6:53:16<2:25:57,  1.76s/it]
 74%|███████▍  | 14111/19090 [6:53:17<2:14:16,  1.62s/it]
 74%|███████▍  | 14112/19090 [6:53:20<2:49:37,  2.04s/it]
 74%|███████▍  | 14113/19090 [6:53:21<2:30:39,  1.82s/it]
 74%|███████▍  | 14114/19090 [6:53:22<2:17:45,  1.66s/it]
 74%|███████▍  | 14115/19090 [6:53:26<2:53:58,  2.10s/it]
 74%|███████▍  | 14116/19090 [6:53:27<2:33:41,  1.85s/it]
 74%|███████▍  | 14117/19090 [6:53:28<2:19:32,  1.68s/it]
 74%|███████▍  | 14118/19090 [6:53:31<2:48:58,  2.04s/it]
 74%|███████▍  | 14119/19090 [6:53:32<2:30:16,  1.81s/it][gradnorm] step 14119: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 420.9

 74%|███████▍  | 14120/19090 [6:53:34<2:17:12,  1.66s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 606.3931884765625, 'learning_rate': 9.63585505366967e-06, 'recon_loss': 0.0017746735247783363, 'recon_static': 0.0007975620595971122, 'recon_motion': 0.0009771114477189259, 'recon_norm': 1.3877054870128631, 'recon_norm_std': 0.03685244023799896, 'target_norm': 1.6187872469425202, 'recon_cos': 0.5916240885853767, 'gradnorm/compressor.query': 0.33014085839388907, 'gradnorm/compressor.layer0': 12.440520720055604, 'gradnorm/compressor.layer1': 596.545838557273, 'gradnorm/compressor.layer2': 104.76225409379325, 'gradnorm/compressor.layer3': 24.958429793827925, 'gradnorm/compressor.layer4': 7.432096939013152, 'gradnorm/compressor.layer5': 3.788296777493183, 'gradnorm/compressor.layer6': 2.08924426455713, 'gradnorm/compressor.layer7': 2.6932051671884936, 'gradnorm/decoder.layer0': 2.9409688393849738, 'gradnorm/decoder.layer1': 1.4844242288557503, 'gradnorm/decoder.layer2': 0.8907683407728508, 'gradnorm/max_param': 420.86492919921875, 'epoch': 7.4}

 74%|███████▍  | 14120/19090 [6:53:34<2:17:12,  1.66s/it]
 74%|███████▍  | 14121/19090 [6:53:36<2:43:10,  1.97s/it]
 74%|███████▍  | 14122/19090 [6:53:38<2:26:13,  1.77s/it]
 74%|███████▍  | 14123/19090 [6:53:39<2:14:20,  1.62s/it]
 74%|███████▍  | 14124/19090 [6:53:42<2:40:27,  1.94s/it]
 74%|███████▍  | 14125/19090 [6:53:43<2:24:23,  1.75s/it]
 74%|███████▍  | 14126/19090 [6:53:44<2:13:08,  1.61s/it]
 74%|███████▍  | 14127/19090 [6:53:47<2:40:58,  1.95s/it]
 74%|███████▍  | 14128/19090 [6:53:48<2:24:29,  1.75s/it]
 74%|███████▍  | 14129/19090 [6:53:49<2:13:05,  1.61s/it][gradnorm] step 14129: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 493.3

 74%|███████▍  | 14130/19090 [6:53:52<2:40:57,  1.95s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 708.6570434570312, 'learning_rate': 9.59981920445816e-06, 'recon_loss': 0.0017608377209398895, 'recon_static': 0.0007842302467906848, 'recon_motion': 0.0009766074756043964, 'recon_norm': 1.3783545970916748, 'recon_norm_std': 0.03777504367753863, 'target_norm': 1.6224854439496994, 'recon_cos': 0.5948174327611924, 'gradnorm/compressor.query': 0.387539945358595, 'gradnorm/compressor.layer0': 16.681215167161085, 'gradnorm/compressor.layer1': 699.1068632548246, 'gradnorm/compressor.layer2': 110.7948202196294, 'gradnorm/compressor.layer3': 28.408160087733815, 'gradnorm/compressor.layer4': 6.828727503541707, 'gradnorm/compressor.layer5': 3.621531175120292, 'gradnorm/compressor.layer6': 2.252803327690181, 'gradnorm/compressor.layer7': 2.928210728061102, 'gradnorm/decoder.layer0': 2.658085626363929, 'gradnorm/decoder.layer1': 1.5070864653965672, 'gradnorm/decoder.layer2': 0.8538131369930522, 'gradnorm/max_param': 493.3052978515625, 'epoch': 7.4}

 74%|███████▍  | 14130/19090 [6:53:52<2:40:57,  1.95s/it]
 74%|███████▍  | 14131/19090 [6:53:53<2:24:31,  1.75s/it]
 74%|███████▍  | 14132/19090 [6:53:55<2:13:00,  1.61s/it]
 74%|███████▍  | 14133/19090 [6:53:57<2:38:13,  1.92s/it]
 74%|███████▍  | 14134/19090 [6:53:59<2:22:32,  1.73s/it]
 74%|███████▍  | 14135/19090 [6:54:00<2:11:30,  1.59s/it]
 74%|███████▍  | 14136/19090 [6:54:03<2:42:23,  1.97s/it]
 74%|███████▍  | 14137/19090 [6:54:04<2:25:29,  1.76s/it]
 74%|███████▍  | 14138/19090 [6:54:05<2:13:42,  1.62s/it]
 74%|███████▍  | 14139/19090 [6:54:08<2:35:02,  1.88s/it][gradnorm] step 14139: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 369.3

 74%|███████▍  | 14140/19090 [6:54:09<2:20:19,  1.70s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 531.2496948242188, 'learning_rate': 9.563834845919637e-06, 'recon_loss': 0.0017178408423205837, 'recon_static': 0.0007520461091189645, 'recon_motion': 0.0009657947361120023, 'recon_norm': 1.3659897118806839, 'recon_norm_std': 0.03937422949820757, 'target_norm': 1.6205518245697021, 'recon_cos': 0.6026243433356285, 'gradnorm/compressor.query': 0.29598706872585556, 'gradnorm/compressor.layer0': 11.821234163505835, 'gradnorm/compressor.layer1': 523.4960004622767, 'gradnorm/compressor.layer2': 85.78040755506179, 'gradnorm/compressor.layer3': 24.438154363042127, 'gradnorm/compressor.layer4': 7.174651495680655, 'gradnorm/compressor.layer5': 3.3804386258353443, 'gradnorm/compressor.layer6': 2.101326514399223, 'gradnorm/compressor.layer7': 2.5626358368388935, 'gradnorm/decoder.layer0': 2.559993355265577, 'gradnorm/decoder.layer1': 1.328201291754091, 'gradnorm/decoder.layer2': 0.7828527413568681, 'gradnorm/max_param': 369.34967041015625, 'epoch': 7.41}

 74%|███████▍  | 14140/19090 [6:54:09<2:20:19,  1.70s/it]
 74%|███████▍  | 14141/19090 [6:54:10<2:10:04,  1.58s/it]
 74%|███████▍  | 14142/19090 [6:54:13<2:38:10,  1.92s/it]
 74%|███████▍  | 14143/19090 [6:54:14<2:22:24,  1.73s/it]
 74%|███████▍  | 14144/19090 [6:54:16<2:11:25,  1.59s/it]
 74%|███████▍  | 14145/19090 [6:54:18<2:34:47,  1.88s/it]
 74%|███████▍  | 14146/19090 [6:54:20<2:20:02,  1.70s/it]
 74%|███████▍  | 14147/19090 [6:54:21<2:09:49,  1.58s/it]
 74%|███████▍  | 14148/19090 [6:54:24<2:40:41,  1.95s/it]
 74%|███████▍  | 14149/19090 [6:54:25<2:24:14,  1.75s/it][gradnorm] step 14149: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 415.2

 74%|███████▍  | 14150/19090 [6:54:26<2:12:46,  1.61s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 599.5120239257812, 'learning_rate': 9.527902098368197e-06, 'recon_loss': 0.0017256051331060007, 'recon_static': 0.0007575100040412508, 'recon_motion': 0.0009680951276095584, 'recon_norm': 1.3643540024757386, 'recon_norm_std': 0.03863654062151909, 'target_norm': 1.6205815941095352, 'recon_cos': 0.600638173520565, 'gradnorm/compressor.query': 0.3577774447131203, 'gradnorm/compressor.layer0': 15.736803611757981, 'gradnorm/compressor.layer1': 588.4623501125624, 'gradnorm/compressor.layer2': 109.89469391194463, 'gradnorm/compressor.layer3': 26.542537305636284, 'gradnorm/compressor.layer4': 8.04086547444521, 'gradnorm/compressor.layer5': 3.537886603160323, 'gradnorm/compressor.layer6': 2.164847992280612, 'gradnorm/compressor.layer7': 2.50195607908076, 'gradnorm/decoder.layer0': 2.611463430547032, 'gradnorm/decoder.layer1': 1.4219607652222408, 'gradnorm/decoder.layer2': 0.8158960893072754, 'gradnorm/max_param': 415.182861328125, 'epoch': 7.41}

 74%|███████▍  | 14150/19090 [6:54:26<2:12:46,  1.61s/it]
 74%|███████▍  | 14151/19090 [6:54:29<2:42:57,  1.98s/it]
 74%|███████▍  | 14152/19090 [6:54:30<2:25:46,  1.77s/it]
 74%|███████▍  | 14153/19090 [6:54:32<2:13:42,  1.63s/it]
 74%|███████▍  | 14154/19090 [6:54:35<2:47:47,  2.04s/it]
 74%|███████▍  | 14155/19090 [6:54:36<2:29:21,  1.82s/it]
 74%|███████▍  | 14156/19090 [6:54:37<2:16:40,  1.66s/it]
 74%|███████▍  | 14157/19090 [6:54:40<2:38:48,  1.93s/it]
 74%|███████▍  | 14158/19090 [6:54:41<2:23:03,  1.74s/it]
 74%|███████▍  | 14159/19090 [6:54:42<2:11:52,  1.60s/it][gradnorm] step 14159: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 266.5

 74%|███████▍  | 14160/19090 [6:54:45<2:43:07,  1.99s/it]
                                                         
{'loss': 0.0021, 'grad_norm': 385.1982116699219, 'learning_rate': 9.492021081945385e-06, 'recon_loss': 0.0020662172406446188, 'recon_static': 0.0010948833645670676, 'recon_motion': 0.0009713338207802735, 'recon_norm': 1.4965663224458694, 'recon_norm_std': 0.036038242559880015, 'target_norm': 1.6211012721061706, 'recon_cos': 0.5488902151584625, 'gradnorm/compressor.query': 0.2623555989682368, 'gradnorm/compressor.layer0': 9.068620390856431, 'gradnorm/compressor.layer1': 377.77401065319464, 'gradnorm/compressor.layer2': 71.50232185915364, 'gradnorm/compressor.layer3': 19.81424048447385, 'gradnorm/compressor.layer4': 6.86305030603549, 'gradnorm/compressor.layer5': 3.4420258292293555, 'gradnorm/compressor.layer6': 2.040217280877735, 'gradnorm/compressor.layer7': 2.426243932953235, 'gradnorm/decoder.layer0': 2.334979112682818, 'gradnorm/decoder.layer1': 1.3432076712647718, 'gradnorm/decoder.layer2': 0.8612255978276228, 'gradnorm/max_param': 266.5406799316406, 'epoch': 7.42}

 74%|███████▍  | 14160/19090 [6:54:45<2:43:07,  1.99s/it]
 74%|███████▍  | 14161/19090 [6:54:47<2:25:53,  1.78s/it]
 74%|███████▍  | 14162/19090 [6:54:48<2:13:50,  1.63s/it]
 74%|███████▍  | 14163/19090 [6:54:51<2:42:48,  1.98s/it]
 74%|███████▍  | 14164/19090 [6:54:52<2:25:47,  1.78s/it]
 74%|███████▍  | 14165/19090 [6:54:53<2:13:41,  1.63s/it]
 74%|███████▍  | 14166/19090 [6:54:56<2:30:47,  1.84s/it]
 74%|███████▍  | 14167/19090 [6:54:57<2:17:12,  1.67s/it]
 74%|███████▍  | 14168/19090 [6:54:58<2:07:52,  1.56s/it]
 74%|███████▍  | 14169/19090 [6:55:01<2:46:51,  2.03s/it][gradnorm] step 14169: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 244.1

 74%|███████▍  | 14170/19090 [6:55:03<2:28:31,  1.81s/it]
                                                         
{'loss': 0.0025, 'grad_norm': 350.57818603515625, 'learning_rate': 9.456191916619744e-06, 'recon_loss': 0.0025049774441868065, 'recon_static': 0.0015410033462103455, 'recon_motion': 0.0009639741067076102, 'recon_norm': 1.6646416276693343, 'recon_norm_std': 0.033786640129983427, 'target_norm': 1.6216134428977966, 'recon_cos': 0.4989872857928276, 'gradnorm/compressor.query': 0.45483591731464945, 'gradnorm/compressor.layer0': 17.922624221479815, 'gradnorm/compressor.layer1': 345.9944792731526, 'gradnorm/compressor.layer2': 49.446037703048056, 'gradnorm/compressor.layer3': 18.905286326163345, 'gradnorm/compressor.layer4': 5.911785781014521, 'gradnorm/compressor.layer5': 3.207122541236433, 'gradnorm/compressor.layer6': 2.2088834569191786, 'gradnorm/compressor.layer7': 2.654568487758636, 'gradnorm/decoder.layer0': 2.734141486959907, 'gradnorm/decoder.layer1': 1.8215015768064742, 'gradnorm/decoder.layer2': 1.0855838556569288, 'gradnorm/max_param': 244.05889892578125, 'epoch': 7.42}

 74%|███████▍  | 14170/19090 [6:55:03<2:28:31,  1.81s/it]
 74%|███████▍  | 14171/19090 [6:55:04<2:16:06,  1.66s/it]
 74%|███████▍  | 14172/19090 [6:55:06<2:35:57,  1.90s/it]
 74%|███████▍  | 14173/19090 [6:55:08<2:20:47,  1.72s/it]
 74%|███████▍  | 14174/19090 [6:55:09<2:10:13,  1.59s/it]
 74%|███████▍  | 14175/19090 [6:55:12<2:38:25,  1.93s/it]
 74%|███████▍  | 14176/19090 [6:55:13<2:22:25,  1.74s/it]
 74%|███████▍  | 14177/19090 [6:55:14<2:11:24,  1.60s/it]
 74%|███████▍  | 14178/19090 [6:55:17<2:43:09,  1.99s/it]
 74%|███████▍  | 14179/19090 [6:55:18<2:25:46,  1.78s/it][gradnorm] step 14179: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 378.6

 74%|███████▍  | 14180/19090 [6:55:20<2:13:36,  1.63s/it]
                                                         
{'loss': 0.0031, 'grad_norm': 546.2994995117188, 'learning_rate': 9.42041472218651e-06, 'recon_loss': 0.0030891456815879794, 'recon_static': 0.0021222420444246383, 'recon_motion': 0.0009669036473496817, 'recon_norm': 1.853659749031067, 'recon_norm_std': 0.03185633234679699, 'target_norm': 1.6215634793043137, 'recon_cos': 0.4464788131415844, 'gradnorm/compressor.query': 0.5226647950801053, 'gradnorm/compressor.layer0': 20.86776589777768, 'gradnorm/compressor.layer1': 536.8443966830612, 'gradnorm/compressor.layer2': 95.61032362864901, 'gradnorm/compressor.layer3': 24.46899694559296, 'gradnorm/compressor.layer4': 5.976377397051845, 'gradnorm/compressor.layer5': 3.1420589987923666, 'gradnorm/compressor.layer6': 2.45760698956797, 'gradnorm/compressor.layer7': 2.293270370687353, 'gradnorm/decoder.layer0': 2.3846932415048903, 'gradnorm/decoder.layer1': 1.4735321758836004, 'gradnorm/decoder.layer2': 0.8588809587329951, 'gradnorm/max_param': 378.6440124511719, 'epoch': 7.43}

 74%|███████▍  | 14180/19090 [6:55:20<2:13:36,  1.63s/it]
 74%|███████▍  | 14181/19090 [6:55:22<2:42:03,  1.98s/it]
 74%|███████▍  | 14182/19090 [6:55:24<2:24:54,  1.77s/it]
 74%|███████▍  | 14183/19090 [6:55:25<2:13:02,  1.63s/it]
 74%|███████▍  | 14184/19090 [6:55:28<2:44:57,  2.02s/it]
 74%|███████▍  | 14185/19090 [6:55:29<2:27:02,  1.80s/it]
 74%|███████▍  | 14186/19090 [6:55:31<2:14:58,  1.65s/it]
 74%|███████▍  | 14187/19090 [6:55:33<2:42:19,  1.99s/it]
 74%|███████▍  | 14188/19090 [6:55:35<2:25:11,  1.78s/it]
 74%|███████▍  | 14189/19090 [6:55:36<2:13:10,  1.63s/it][gradnorm] step 14189: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 421.4

 74%|███████▍  | 14190/19090 [6:55:39<2:40:01,  1.96s/it]
                                                         
{'loss': 0.0021, 'grad_norm': 607.15185546875, 'learning_rate': 9.384689618267117e-06, 'recon_loss': 0.0020811501919524742, 'recon_static': 0.001112635782919824, 'recon_motion': 0.0009685143755632453, 'recon_norm': 1.5053605049848557, 'recon_norm_std': 0.03647484201937914, 'target_norm': 1.619761437177658, 'recon_cos': 0.5445103421807289, 'gradnorm/compressor.query': 0.4587561831460392, 'gradnorm/compressor.layer0': 19.023204888208245, 'gradnorm/compressor.layer1': 597.2626620675362, 'gradnorm/compressor.layer2': 103.7576239668681, 'gradnorm/compressor.layer3': 26.658765130830446, 'gradnorm/compressor.layer4': 6.150921524601549, 'gradnorm/compressor.layer5': 3.387968226656499, 'gradnorm/compressor.layer6': 2.3486932124185125, 'gradnorm/compressor.layer7': 2.6076342778363344, 'gradnorm/decoder.layer0': 2.7195054681789257, 'gradnorm/decoder.layer1': 1.5584149174405137, 'gradnorm/decoder.layer2': 0.9396926352509977, 'gradnorm/max_param': 421.43280029296875, 'epoch': 7.43}

 74%|███████▍  | 14190/19090 [6:55:39<2:40:01,  1.96s/it]
 74%|███████▍  | 14191/19090 [6:55:40<2:23:32,  1.76s/it]
 74%|███████▍  | 14192/19090 [6:55:41<2:11:56,  1.62s/it]
 74%|███████▍  | 14193/19090 [6:55:44<2:37:03,  1.92s/it]
 74%|███████▍  | 14194/19090 [6:55:45<2:21:21,  1.73s/it]
 74%|███████▍  | 14195/19090 [6:55:46<2:10:38,  1.60s/it]
 74%|███████▍  | 14196/19090 [6:55:49<2:44:45,  2.02s/it]
 74%|███████▍  | 14197/19090 [6:55:51<2:26:59,  1.80s/it]
 74%|███████▍  | 14198/19090 [6:55:52<2:14:30,  1.65s/it]
 74%|███████▍  | 14199/19090 [6:55:55<2:41:11,  1.98s/it][gradnorm] step 14199: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 482.4

 74%|███████▍  | 14200/19090 [6:55:56<2:24:23,  1.77s/it]
                                                         
{'loss': 0.002, 'grad_norm': 693.4287109375, 'learning_rate': 9.349016724308824e-06, 'recon_loss': 0.001965696891420521, 'recon_static': 0.0010020500791142694, 'recon_motion': 0.000963646822492592, 'recon_norm': 1.4597240149974824, 'recon_norm_std': 0.03730105310678482, 'target_norm': 1.621217805147171, 'recon_cos': 0.5599763259291649, 'gradnorm/compressor.query': 0.45660189488014, 'gradnorm/compressor.layer0': 17.841446121518043, 'gradnorm/compressor.layer1': 683.8230765336893, 'gradnorm/compressor.layer2': 110.05207396671359, 'gradnorm/compressor.layer3': 26.921886335604114, 'gradnorm/compressor.layer4': 6.080281088505233, 'gradnorm/compressor.layer5': 3.730102904590302, 'gradnorm/compressor.layer6': 2.2515766130120234, 'gradnorm/compressor.layer7': 2.8104723933081717, 'gradnorm/decoder.layer0': 2.673230725156026, 'gradnorm/decoder.layer1': 1.6645583725809587, 'gradnorm/decoder.layer2': 0.9050548170148357, 'gradnorm/max_param': 482.43756103515625, 'epoch': 7.44}

 74%|███████▍  | 14200/19090 [6:55:56<2:24:23,  1.77s/it]
 74%|███████▍  | 14201/19090 [6:55:57<2:12:36,  1.63s/it]
 74%|███████▍  | 14202/19090 [6:56:00<2:40:58,  1.98s/it]
 74%|███████▍  | 14203/19090 [6:56:01<2:24:06,  1.77s/it]
 74%|███████▍  | 14204/19090 [6:56:03<2:12:15,  1.62s/it]
 74%|███████▍  | 14205/19090 [6:56:06<2:41:24,  1.98s/it]
 74%|███████▍  | 14206/19090 [6:56:07<2:24:21,  1.77s/it]
 74%|███████▍  | 14207/19090 [6:56:08<2:12:25,  1.63s/it]
 74%|███████▍  | 14208/19090 [6:56:11<2:39:21,  1.96s/it]
 74%|███████▍  | 14209/19090 [6:56:12<2:22:47,  1.76s/it][gradnorm] step 14209: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 358.8

 74%|███████▍  | 14210/19090 [6:56:13<2:11:25,  1.62s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 515.902099609375, 'learning_rate': 9.313396159584358e-06, 'recon_loss': 0.0019234217674238608, 'recon_static': 0.0009440745008760132, 'recon_motion': 0.0009793472505407407, 'recon_norm': 1.444054228067398, 'recon_norm_std': 0.03625772427767515, 'target_norm': 1.6223063915967941, 'recon_cos': 0.5680343464016915, 'gradnorm/compressor.query': 0.2526347970907872, 'gradnorm/compressor.layer0': 9.32061004803229, 'gradnorm/compressor.layer1': 508.46361718416, 'gradnorm/compressor.layer2': 83.43143119315106, 'gradnorm/compressor.layer3': 22.37529323961127, 'gradnorm/compressor.layer4': 6.330332731207181, 'gradnorm/compressor.layer5': 3.526804057896181, 'gradnorm/compressor.layer6': 2.081128581403072, 'gradnorm/compressor.layer7': 2.3564349924012515, 'gradnorm/decoder.layer0': 2.4455783687272565, 'gradnorm/decoder.layer1': 1.4761257433944743, 'gradnorm/decoder.layer2': 0.8830951137034755, 'gradnorm/max_param': 358.7609558105469, 'epoch': 7.44}

 74%|███████▍  | 14210/19090 [6:56:13<2:11:25,  1.62s/it]
 74%|███████▍  | 14211/19090 [6:56:16<2:38:09,  1.94s/it]
 74%|███████▍  | 14212/19090 [6:56:17<2:21:57,  1.75s/it]
 74%|███████▍  | 14213/19090 [6:56:19<2:10:39,  1.61s/it]
 74%|███████▍  | 14214/19090 [6:56:21<2:38:02,  1.94s/it]
 74%|███████▍  | 14215/19090 [6:56:23<2:22:02,  1.75s/it]
 74%|███████▍  | 14216/19090 [6:56:24<2:10:48,  1.61s/it]
 74%|███████▍  | 14217/19090 [6:56:27<2:34:21,  1.90s/it]
 74%|███████▍  | 14218/19090 [6:56:28<2:19:15,  1.71s/it]
 74%|███████▍  | 14219/19090 [6:56:29<2:09:12,  1.59s/it][gradnorm] step 14219: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 437.8

 74%|███████▍  | 14220/19090 [6:56:32<2:36:25,  1.93s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 629.811279296875, 'learning_rate': 9.277828043191475e-06, 'recon_loss': 0.0016864470555447042, 'recon_static': 0.0007177550593041815, 'recon_motion': 0.0009686920311651193, 'recon_norm': 1.3474143087863921, 'recon_norm_std': 0.03917409619316459, 'target_norm': 1.620691967010498, 'recon_cos': 0.6081450566649437, 'gradnorm/compressor.query': 0.32016238905316047, 'gradnorm/compressor.layer0': 10.621152652466312, 'gradnorm/compressor.layer1': 620.547968532651, 'gradnorm/compressor.layer2': 103.24796173078914, 'gradnorm/compressor.layer3': 27.073959215048784, 'gradnorm/compressor.layer4': 6.401707969690528, 'gradnorm/compressor.layer5': 3.7034577167909712, 'gradnorm/compressor.layer6': 2.2265326246967754, 'gradnorm/compressor.layer7': 2.7615859716025946, 'gradnorm/decoder.layer0': 2.5202581732062765, 'gradnorm/decoder.layer1': 1.4571360532484232, 'gradnorm/decoder.layer2': 0.8544310106448613, 'gradnorm/max_param': 437.84442138671875, 'epoch': 7.45}

 74%|███████▍  | 14220/19090 [6:56:32<2:36:25,  1.93s/it]
 74%|███████▍  | 14221/19090 [6:56:33<2:20:53,  1.74s/it]
 74%|███████▍  | 14222/19090 [6:56:34<2:10:02,  1.60s/it]
 75%|███████▍  | 14223/19090 [6:56:37<2:43:34,  2.02s/it]
 75%|███████▍  | 14224/19090 [6:56:39<2:25:43,  1.80s/it]
 75%|███████▍  | 14225/19090 [6:56:40<2:13:12,  1.64s/it]
 75%|███████▍  | 14226/19090 [6:56:43<2:39:26,  1.97s/it]
 75%|███████▍  | 14227/19090 [6:56:44<2:22:53,  1.76s/it]
 75%|███████▍  | 14228/19090 [6:56:45<2:11:14,  1.62s/it]
 75%|███████▍  | 14229/19090 [6:56:48<2:29:46,  1.85s/it][gradnorm] step 14229: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 417.8

 75%|███████▍  | 14230/19090 [6:56:49<2:16:02,  1.68s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 600.4986572265625, 'learning_rate': 9.242312494052533e-06, 'recon_loss': 0.0015935327799525111, 'recon_static': 0.0006247547367820516, 'recon_motion': 0.0009687780329841189, 'recon_norm': 1.3121211290359498, 'recon_norm_std': 0.03927171463146806, 'target_norm': 1.6178507596254348, 'recon_cos': 0.6259020075201989, 'gradnorm/compressor.query': 0.29013660986868434, 'gradnorm/compressor.layer0': 11.189649663082376, 'gradnorm/compressor.layer1': 592.1384181506888, 'gradnorm/compressor.layer2': 95.33105352205544, 'gradnorm/compressor.layer3': 25.9881682250849, 'gradnorm/compressor.layer4': 6.663267349799372, 'gradnorm/compressor.layer5': 4.014682050493256, 'gradnorm/compressor.layer6': 2.454012376202229, 'gradnorm/compressor.layer7': 2.586176558678365, 'gradnorm/decoder.layer0': 2.4194048682546754, 'gradnorm/decoder.layer1': 1.5080680828206945, 'gradnorm/decoder.layer2': 0.8575188097545116, 'gradnorm/max_param': 417.8299865722656, 'epoch': 7.45}

 75%|███████▍  | 14230/19090 [6:56:49<2:16:02,  1.68s/it]
 75%|███████▍  | 14231/19090 [6:56:50<2:06:39,  1.56s/it]
 75%|███████▍  | 14232/19090 [6:56:54<2:47:51,  2.07s/it]
 75%|███████▍  | 14233/19090 [6:56:55<2:29:00,  1.84s/it]
 75%|███████▍  | 14234/19090 [6:56:56<2:15:20,  1.67s/it]
 75%|███████▍  | 14235/19090 [6:56:59<2:39:07,  1.97s/it]
 75%|███████▍  | 14236/19090 [6:57:00<2:22:25,  1.76s/it]
 75%|███████▍  | 14237/19090 [6:57:01<2:10:53,  1.62s/it]
 75%|███████▍  | 14238/19090 [6:57:04<2:42:26,  2.01s/it]
 75%|███████▍  | 14239/19090 [6:57:06<2:24:55,  1.79s/it][gradnorm] step 14239: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 398.0

 75%|███████▍  | 14240/19090 [6:57:07<2:12:54,  1.64s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 575.5047607421875, 'learning_rate': 9.206849630914184e-06, 'recon_loss': 0.0015464457828784361, 'recon_static': 0.0005619176416075788, 'recon_motion': 0.0009845281761954538, 'recon_norm': 1.2859274744987488, 'recon_norm_std': 0.041110779717564584, 'target_norm': 1.6221600115299224, 'recon_cos': 0.637819553911686, 'gradnorm/compressor.query': 0.3035577835594691, 'gradnorm/compressor.layer0': 11.01842377732047, 'gradnorm/compressor.layer1': 564.0726903511638, 'gradnorm/compressor.layer2': 109.34559426138303, 'gradnorm/compressor.layer3': 28.985849019122675, 'gradnorm/compressor.layer4': 7.82620453262465, 'gradnorm/compressor.layer5': 4.429216201710978, 'gradnorm/compressor.layer6': 2.9341086432293944, 'gradnorm/compressor.layer7': 2.8578965998018004, 'gradnorm/decoder.layer0': 2.865573355938417, 'gradnorm/decoder.layer1': 1.689860423830743, 'gradnorm/decoder.layer2': 0.9731672545601838, 'gradnorm/max_param': 397.9844055175781, 'epoch': 7.46}

 75%|███████▍  | 14240/19090 [6:57:07<2:12:54,  1.64s/it]
 75%|███████▍  | 14241/19090 [6:57:09<2:37:01,  1.94s/it]
 75%|███████▍  | 14242/19090 [6:57:11<2:21:01,  1.75s/it]
 75%|███████▍  | 14243/19090 [6:57:12<2:09:57,  1.61s/it]
 75%|███████▍  | 14244/19090 [6:57:15<2:39:30,  1.97s/it]
 75%|███████▍  | 14245/19090 [6:57:16<2:22:42,  1.77s/it]
 75%|███████▍  | 14246/19090 [6:57:17<2:11:03,  1.62s/it]
 75%|███████▍  | 14247/19090 [6:57:20<2:30:37,  1.87s/it]
 75%|███████▍  | 14248/19090 [6:57:21<2:16:29,  1.69s/it]
 75%|███████▍  | 14249/19090 [6:57:22<2:06:44,  1.57s/it][gradnorm] step 14249: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 410.2

 75%|███████▍  | 14250/19090 [6:57:26<2:44:19,  2.04s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 591.0472412109375, 'learning_rate': 9.171439572346898e-06, 'recon_loss': 0.0015736855304567144, 'recon_static': 0.0005990091085550375, 'recon_motion': 0.000974676494661253, 'recon_norm': 1.3019761085510253, 'recon_norm_std': 0.040813132841140035, 'target_norm': 1.6206576377153397, 'recon_cos': 0.6315412744879723, 'gradnorm/compressor.query': 0.36817377618982716, 'gradnorm/compressor.layer0': 11.815083861955406, 'gradnorm/compressor.layer1': 581.2590321878878, 'gradnorm/compressor.layer2': 101.01960186487324, 'gradnorm/compressor.layer3': 31.813503406940857, 'gradnorm/compressor.layer4': 8.245162933016678, 'gradnorm/compressor.layer5': 4.59431360640935, 'gradnorm/compressor.layer6': 2.6112999126362584, 'gradnorm/compressor.layer7': 3.089360986701351, 'gradnorm/decoder.layer0': 2.878861197917503, 'gradnorm/decoder.layer1': 1.7651678523144712, 'gradnorm/decoder.layer2': 1.0577875988056449, 'gradnorm/max_param': 410.1551208496094, 'epoch': 7.46}

 75%|███████▍  | 14250/19090 [6:57:26<2:44:19,  2.04s/it]
 75%|███████▍  | 14251/19090 [6:57:27<2:26:15,  1.81s/it]
 75%|███████▍  | 14252/19090 [6:57:28<2:13:34,  1.66s/it]
 75%|███████▍  | 14253/19090 [6:57:31<2:37:22,  1.95s/it]
 75%|███████▍  | 14254/19090 [6:57:32<2:21:14,  1.75s/it]
 75%|███████▍  | 14255/19090 [6:57:33<2:09:57,  1.61s/it]
 75%|███████▍  | 14256/19090 [6:57:36<2:34:39,  1.92s/it]
 75%|███████▍  | 14257/19090 [6:57:37<2:19:16,  1.73s/it]
 75%|███████▍  | 14258/19090 [6:57:39<2:08:37,  1.60s/it]
 75%|███████▍  | 14259/19090 [6:57:42<2:43:53,  2.04s/it][gradnorm] step 14259: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 394.0

 75%|███████▍  | 14260/19090 [6:57:43<2:25:45,  1.81s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 567.5716552734375, 'learning_rate': 9.13608243674457e-06, 'recon_loss': 0.0016490669513586909, 'recon_static': 0.0006760091899195686, 'recon_motion': 0.0009730577585287392, 'recon_norm': 1.3332801789045334, 'recon_norm_std': 0.03956419071182608, 'target_norm': 1.6225574344396592, 'recon_cos': 0.6166216611862183, 'gradnorm/compressor.query': 0.31418565072159843, 'gradnorm/compressor.layer0': 14.163011565569143, 'gradnorm/compressor.layer1': 558.4981815100207, 'gradnorm/compressor.layer2': 95.84261900213026, 'gradnorm/compressor.layer3': 27.272794096893747, 'gradnorm/compressor.layer4': 6.940904486472353, 'gradnorm/compressor.layer5': 3.6575217358353336, 'gradnorm/compressor.layer6': 2.388113560153743, 'gradnorm/compressor.layer7': 2.7914616713532494, 'gradnorm/decoder.layer0': 2.966970452460133, 'gradnorm/decoder.layer1': 1.556138937790349, 'gradnorm/decoder.layer2': 0.8881917473920513, 'gradnorm/max_param': 394.0393371582031, 'epoch': 7.47}

 75%|███████▍  | 14260/19090 [6:57:43<2:25:45,  1.81s/it]
 75%|███████▍  | 14261/19090 [6:57:44<2:13:05,  1.65s/it]
 75%|███████▍  | 14262/19090 [6:57:47<2:35:17,  1.93s/it]
 75%|███████▍  | 14263/19090 [6:57:48<2:19:40,  1.74s/it]
 75%|███████▍  | 14264/19090 [6:57:49<2:08:45,  1.60s/it]
 75%|███████▍  | 14265/19090 [6:57:52<2:40:00,  1.99s/it]
 75%|███████▍  | 14266/19090 [6:57:54<2:23:14,  1.78s/it]
 75%|███████▍  | 14267/19090 [6:57:55<2:11:18,  1.63s/it]
 75%|███████▍  | 14268/19090 [6:57:58<2:43:27,  2.03s/it]
 75%|███████▍  | 14269/19090 [6:57:59<2:25:19,  1.81s/it][gradnorm] step 14269: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 355.3

 75%|███████▍  | 14270/19090 [6:58:00<2:12:45,  1.65s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 513.679931640625, 'learning_rate': 9.100778342324196e-06, 'recon_loss': 0.0016743728367146105, 'recon_static': 0.0006978716133744456, 'recon_motion': 0.000976501195691526, 'recon_norm': 1.3427651405334473, 'recon_norm_std': 0.03937633084133267, 'target_norm': 1.6191815793514253, 'recon_cos': 0.6094155728816986, 'gradnorm/compressor.query': 0.24754924869392525, 'gradnorm/compressor.layer0': 8.8524183514419, 'gradnorm/compressor.layer1': 503.44120498326316, 'gradnorm/compressor.layer2': 98.13268779597601, 'gradnorm/compressor.layer3': 24.50003114036605, 'gradnorm/compressor.layer4': 7.811514830462348, 'gradnorm/compressor.layer5': 4.128878301924687, 'gradnorm/compressor.layer6': 2.8439503693468535, 'gradnorm/compressor.layer7': 2.674516676974331, 'gradnorm/decoder.layer0': 2.90232617050741, 'gradnorm/decoder.layer1': 1.6150838096768456, 'gradnorm/decoder.layer2': 0.9787769083000035, 'gradnorm/max_param': 355.27978515625, 'epoch': 7.48}

 75%|███████▍  | 14270/19090 [6:58:00<2:12:45,  1.65s/it]
 75%|███████▍  | 14271/19090 [6:58:03<2:41:27,  2.01s/it]
 75%|███████▍  | 14272/19090 [6:58:05<2:23:57,  1.79s/it]
 75%|███████▍  | 14273/19090 [6:58:06<2:11:57,  1.64s/it]
 75%|███████▍  | 14274/19090 [6:58:09<2:42:45,  2.03s/it]
 75%|███████▍  | 14275/19090 [6:58:10<2:24:47,  1.80s/it]
 75%|███████▍  | 14276/19090 [6:58:11<2:12:17,  1.65s/it]
 75%|███████▍  | 14277/19090 [6:58:14<2:35:53,  1.94s/it]
 75%|███████▍  | 14278/19090 [6:58:15<2:19:58,  1.75s/it]
 75%|███████▍  | 14279/19090 [6:58:16<2:08:58,  1.61s/it][gradnorm] step 14279: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 468.5

 75%|███████▍  | 14280/19090 [6:58:19<2:37:29,  1.96s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 672.8889770507812, 'learning_rate': 9.065527407125399e-06, 'recon_loss': 0.001639336405787617, 'recon_static': 0.0006586091534700244, 'recon_motion': 0.0009807272857869976, 'recon_norm': 1.3290681779384612, 'recon_norm_std': 0.040325173269957305, 'target_norm': 1.6226672142744065, 'recon_cos': 0.6184854298830033, 'gradnorm/compressor.query': 0.33725652566227887, 'gradnorm/compressor.layer0': 13.425832393705322, 'gradnorm/compressor.layer1': 663.8319158491854, 'gradnorm/compressor.layer2': 104.90976870774475, 'gradnorm/compressor.layer3': 28.73222743809967, 'gradnorm/compressor.layer4': 7.4537893195157165, 'gradnorm/compressor.layer5': 3.973497809861068, 'gradnorm/compressor.layer6': 2.4406895432381743, 'gradnorm/compressor.layer7': 2.725355170665727, 'gradnorm/decoder.layer0': 2.6109896478290673, 'gradnorm/decoder.layer1': 1.548293725045378, 'gradnorm/decoder.layer2': 0.9039649766260429, 'gradnorm/max_param': 468.481689453125, 'epoch': 7.48}

 75%|███████▍  | 14280/19090 [6:58:19<2:37:29,  1.96s/it]
 75%|███████▍  | 14281/19090 [6:58:21<2:21:10,  1.76s/it]
 75%|███████▍  | 14282/19090 [6:58:22<2:09:49,  1.62s/it]
 75%|███████▍  | 14283/19090 [6:58:25<2:37:02,  1.96s/it]
 75%|███████▍  | 14284/19090 [6:58:26<2:20:52,  1.76s/it]
 75%|███████▍  | 14285/19090 [6:58:27<2:09:33,  1.62s/it]
 75%|███████▍  | 14286/19090 [6:58:30<2:44:59,  2.06s/it]
 75%|███████▍  | 14287/19090 [6:58:32<2:26:20,  1.83s/it]
 75%|███████▍  | 14288/19090 [6:58:33<2:13:20,  1.67s/it]
 75%|███████▍  | 14289/19090 [6:58:35<2:32:11,  1.90s/it][gradnorm] step 14289: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 403.5

 75%|███████▍  | 14290/19090 [6:58:37<2:17:19,  1.72s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 580.0953369140625, 'learning_rate': 9.03032974901004e-06, 'recon_loss': 0.0015854679775657132, 'recon_static': 0.0006225370336323977, 'recon_motion': 0.0009629309453885071, 'recon_norm': 1.3073483675718307, 'recon_norm_std': 0.04092808207497001, 'target_norm': 1.6172140449285508, 'recon_cos': 0.6265636250376702, 'gradnorm/compressor.query': 0.26105177998011747, 'gradnorm/compressor.layer0': 10.167826185131938, 'gradnorm/compressor.layer1': 571.852253645992, 'gradnorm/compressor.layer2': 92.85025536785022, 'gradnorm/compressor.layer3': 26.179527799050348, 'gradnorm/compressor.layer4': 6.770396464021092, 'gradnorm/compressor.layer5': 3.4527484118835186, 'gradnorm/compressor.layer6': 2.3802254061987416, 'gradnorm/compressor.layer7': 2.983490498272735, 'gradnorm/decoder.layer0': 3.112382364156578, 'gradnorm/decoder.layer1': 1.617943775789193, 'gradnorm/decoder.layer2': 0.9908198745579186, 'gradnorm/max_param': 403.5355529785156, 'epoch': 7.49}

 75%|███████▍  | 14290/19090 [6:58:37<2:17:19,  1.72s/it]
 75%|███████▍  | 14291/19090 [6:58:38<2:06:54,  1.59s/it]
 75%|███████▍  | 14292/19090 [6:58:41<2:35:35,  1.95s/it]
 75%|███████▍  | 14293/19090 [6:58:42<2:19:41,  1.75s/it]
 75%|███████▍  | 14294/19090 [6:58:43<2:08:40,  1.61s/it]
 75%|███████▍  | 14295/19090 [6:58:46<2:41:10,  2.02s/it]
 75%|███████▍  | 14296/19090 [6:58:47<2:23:37,  1.80s/it]
 75%|███████▍  | 14297/19090 [6:58:49<2:11:25,  1.65s/it]
 75%|███████▍  | 14298/19090 [6:58:51<2:34:04,  1.93s/it]
 75%|███████▍  | 14299/19090 [6:58:53<2:19:04,  1.74s/it][gradnorm] step 14299: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 429.7

 75%|███████▍  | 14300/19090 [6:58:54<2:08:21,  1.61s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 617.5331420898438, 'learning_rate': 8.995185485661906e-06, 'recon_loss': 0.0016341364796971902, 'recon_static': 0.0006562580034369603, 'recon_motion': 0.000977878534467891, 'recon_norm': 1.3304318219423295, 'recon_norm_std': 0.04060034109279513, 'target_norm': 1.62166728079319, 'recon_cos': 0.6200598269701004, 'gradnorm/compressor.query': 0.3041415213398497, 'gradnorm/compressor.layer0': 12.472332050280771, 'gradnorm/compressor.layer1': 609.037149934222, 'gradnorm/compressor.layer2': 97.90029007554064, 'gradnorm/compressor.layer3': 24.670009627075544, 'gradnorm/compressor.layer4': 6.032166406455958, 'gradnorm/compressor.layer5': 3.270506646106839, 'gradnorm/compressor.layer6': 2.515537994965246, 'gradnorm/compressor.layer7': 2.7260822211982894, 'gradnorm/decoder.layer0': 2.731211898153856, 'gradnorm/decoder.layer1': 1.6910647428538375, 'gradnorm/decoder.layer2': 1.0053259759295605, 'gradnorm/max_param': 429.7494812011719, 'epoch': 7.49}

 75%|███████▍  | 14300/19090 [6:58:54<2:08:21,  1.61s/it]
 75%|███████▍  | 14301/19090 [6:58:56<2:29:46,  1.88s/it]
 75%|███████▍  | 14302/19090 [6:58:58<2:15:33,  1.70s/it]
 75%|███████▍  | 14303/19090 [6:58:59<2:05:33,  1.57s/it]
 75%|███████▍  | 14304/19090 [6:59:02<2:33:48,  1.93s/it]
 75%|███████▍  | 14305/19090 [6:59:03<2:18:23,  1.74s/it]
 75%|███████▍  | 14306/19090 [6:59:04<2:07:36,  1.60s/it]
 75%|███████▍  | 14307/19090 [6:59:07<2:37:22,  1.97s/it]
 75%|███████▍  | 14308/19090 [6:59:09<2:20:53,  1.77s/it]
 75%|███████▍  | 14309/19090 [6:59:10<2:09:12,  1.62s/it][gradnorm] step 14309: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 538.5

 75%|███████▍  | 14310/19090 [6:59:12<2:33:46,  1.93s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 773.3753662109375, 'learning_rate': 8.960094734586211e-06, 'recon_loss': 0.001718268496915698, 'recon_static': 0.0007339194737141952, 'recon_motion': 0.0009843489213380962, 'recon_norm': 1.3579667270183564, 'recon_norm_std': 0.03944708360359073, 'target_norm': 1.6238462299108505, 'recon_cos': 0.6034331321716309, 'gradnorm/compressor.query': 0.4004297935121757, 'gradnorm/compressor.layer0': 15.215966943458245, 'gradnorm/compressor.layer1': 763.234031932539, 'gradnorm/compressor.layer2': 118.27432691099958, 'gradnorm/compressor.layer3': 35.477310021578404, 'gradnorm/compressor.layer4': 7.8121479412862485, 'gradnorm/compressor.layer5': 3.4951710767447266, 'gradnorm/compressor.layer6': 2.9689215359571093, 'gradnorm/compressor.layer7': 2.9708665229758924, 'gradnorm/decoder.layer0': 3.0614975826260737, 'gradnorm/decoder.layer1': 1.690516777593922, 'gradnorm/decoder.layer2': 1.0196260722412966, 'gradnorm/max_param': 538.4761352539062, 'epoch': 7.5}

 75%|███████▍  | 14310/19090 [6:59:12<2:33:46,  1.93s/it]
 75%|███████▍  | 14311/19090 [6:59:14<2:18:28,  1.74s/it]
 75%|███████▍  | 14312/19090 [6:59:15<2:07:52,  1.61s/it]
 75%|███████▍  | 14313/19090 [6:59:18<2:31:42,  1.91s/it]
 75%|███████▍  | 14314/19090 [6:59:19<2:17:14,  1.72s/it]
 75%|███████▍  | 14315/19090 [6:59:20<2:06:44,  1.59s/it]
 75%|███████▍  | 14316/19090 [6:59:23<2:42:34,  2.04s/it]
 75%|███████▍  | 14317/19090 [6:59:25<2:24:43,  1.82s/it]
 75%|███████▌  | 14318/19090 [6:59:26<2:11:57,  1.66s/it]
 75%|███████▌  | 14319/19090 [6:59:29<2:38:37,  1.99s/it][gradnorm] step 14319: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 486.7

 75%|███████▌  | 14320/19090 [6:59:30<2:22:14,  1.79s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 698.001953125, 'learning_rate': 8.925057613109267e-06, 'recon_loss': 0.0016378291184082627, 'recon_static': 0.0006542802075273357, 'recon_motion': 0.0009835488934186287, 'recon_norm': 1.3301525920629502, 'recon_norm_std': 0.04030491877347231, 'target_norm': 1.6247334420680999, 'recon_cos': 0.6202352181077003, 'gradnorm/compressor.query': 0.3890962823502452, 'gradnorm/compressor.layer0': 15.94132632530197, 'gradnorm/compressor.layer1': 689.930725689181, 'gradnorm/compressor.layer2': 100.71196746477054, 'gradnorm/compressor.layer3': 26.797650544535266, 'gradnorm/compressor.layer4': 6.6301678348891615, 'gradnorm/compressor.layer5': 3.986442956761048, 'gradnorm/compressor.layer6': 2.6810067268751046, 'gradnorm/compressor.layer7': 2.709735174999197, 'gradnorm/decoder.layer0': 2.9897082543053957, 'gradnorm/decoder.layer1': 1.6521222482389797, 'gradnorm/decoder.layer2': 0.9704406813329037, 'gradnorm/max_param': 486.7071228027344, 'epoch': 7.5}

 75%|███████▌  | 14320/19090 [6:59:30<2:22:14,  1.79s/it]
 75%|███████▌  | 14321/19090 [6:59:31<2:10:15,  1.64s/it]
 75%|███████▌  | 14322/19090 [6:59:34<2:34:43,  1.95s/it]
 75%|███████▌  | 14323/19090 [6:59:35<2:19:05,  1.75s/it]
 75%|███████▌  | 14324/19090 [6:59:37<2:08:02,  1.61s/it]
 75%|███████▌  | 14325/19090 [6:59:39<2:39:36,  2.01s/it]
 75%|███████▌  | 14326/19090 [6:59:41<2:22:25,  1.79s/it]
 75%|███████▌  | 14327/19090 [6:59:42<2:10:15,  1.64s/it]
 75%|███████▌  | 14328/19090 [6:59:45<2:36:21,  1.97s/it]
 75%|███████▌  | 14329/19090 [6:59:46<2:20:08,  1.77s/it][gradnorm] step 14329: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 331.8

 75%|███████▌  | 14330/19090 [6:59:47<2:08:45,  1.62s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 478.5931396484375, 'learning_rate': 8.890074238378074e-06, 'recon_loss': 0.0016537016141228377, 'recon_static': 0.0006700589088723064, 'recon_motion': 0.0009836427299887874, 'recon_norm': 1.325714758038521, 'recon_norm_std': 0.039476117584854364, 'target_norm': 1.6228225171566009, 'recon_cos': 0.6141112238168717, 'gradnorm/compressor.query': 0.3941231874024402, 'gradnorm/compressor.layer0': 14.679411003915485, 'gradnorm/compressor.layer1': 470.2755674867662, 'gradnorm/compressor.layer2': 82.84948171892054, 'gradnorm/compressor.layer3': 27.23962204725579, 'gradnorm/compressor.layer4': 5.832150620952661, 'gradnorm/compressor.layer5': 3.6167724898600135, 'gradnorm/compressor.layer6': 2.537309998412818, 'gradnorm/compressor.layer7': 2.4593174036981416, 'gradnorm/decoder.layer0': 2.7600836115072647, 'gradnorm/decoder.layer1': 1.5592773773887154, 'gradnorm/decoder.layer2': 0.9986220342993405, 'gradnorm/max_param': 331.8178405761719, 'epoch': 7.51}

 75%|███████▌  | 14330/19090 [6:59:47<2:08:45,  1.62s/it]
 75%|███████▌  | 14331/19090 [6:59:50<2:34:56,  1.95s/it]
 75%|███████▌  | 14332/19090 [6:59:51<2:19:13,  1.76s/it]
 75%|███████▌  | 14333/19090 [6:59:53<2:08:01,  1.61s/it]
 75%|███████▌  | 14334/19090 [6:59:55<2:32:24,  1.92s/it]
 75%|███████▌  | 14335/19090 [6:59:57<2:17:12,  1.73s/it]
 75%|███████▌  | 14336/19090 [6:59:58<2:06:42,  1.60s/it]
 75%|███████▌  | 14337/19090 [7:00:01<2:40:15,  2.02s/it]
 75%|███████▌  | 14338/19090 [7:00:02<2:22:48,  1.80s/it]
 75%|███████▌  | 14339/19090 [7:00:03<2:10:43,  1.65s/it][gradnorm] step 14339: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 399.4

 75%|███████▌  | 14340/19090 [7:00:06<2:36:31,  1.98s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 576.3817749023438, 'learning_rate': 8.855144727359929e-06, 'recon_loss': 0.0017710774060105904, 'recon_static': 0.0008096002551610581, 'recon_motion': 0.0009614771202905104, 'recon_norm': 1.3909984320402144, 'recon_norm_std': 0.037784531433135274, 'target_norm': 1.618338018655777, 'recon_cos': 0.5937779366970062, 'gradnorm/compressor.query': 0.346999791611793, 'gradnorm/compressor.layer0': 14.630401396391393, 'gradnorm/compressor.layer1': 566.0872006590504, 'gradnorm/compressor.layer2': 103.35669048856973, 'gradnorm/compressor.layer3': 27.968837290366576, 'gradnorm/compressor.layer4': 6.795038365360328, 'gradnorm/compressor.layer5': 3.729246952037994, 'gradnorm/compressor.layer6': 2.6722958774245495, 'gradnorm/compressor.layer7': 2.2795471733886856, 'gradnorm/decoder.layer0': 2.5884953781891316, 'gradnorm/decoder.layer1': 1.4883219395479415, 'gradnorm/decoder.layer2': 0.917283407002717, 'gradnorm/max_param': 399.4230651855469, 'epoch': 7.51}

 75%|███████▌  | 14340/19090 [7:00:06<2:36:31,  1.98s/it]
 75%|███████▌  | 14341/19090 [7:00:08<2:20:17,  1.77s/it]
 75%|███████▌  | 14342/19090 [7:00:09<2:08:43,  1.63s/it]
 75%|███████▌  | 14343/19090 [7:00:11<2:32:21,  1.93s/it]
 75%|███████▌  | 14344/19090 [7:00:13<2:17:12,  1.73s/it]
 75%|███████▌  | 14345/19090 [7:00:14<2:06:29,  1.60s/it]
 75%|███████▌  | 14346/19090 [7:00:17<2:35:58,  1.97s/it]
 75%|███████▌  | 14347/19090 [7:00:18<2:19:37,  1.77s/it]
 75%|███████▌  | 14348/19090 [7:00:19<2:08:11,  1.62s/it]
 75%|███████▌  | 14349/19090 [7:00:22<2:37:32,  1.99s/it][gradnorm] step 14349: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 420.9

 75%|███████▌  | 14350/19090 [7:00:24<2:20:57,  1.78s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 602.78076171875, 'learning_rate': 8.820269196842032e-06, 'recon_loss': 0.0019137331808451563, 'recon_static': 0.0009302912672865205, 'recon_motion': 0.0009834419572143816, 'recon_norm': 1.435595878958702, 'recon_norm_std': 0.03655350804328918, 'target_norm': 1.6227063477039336, 'recon_cos': 0.5691045477986336, 'gradnorm/compressor.query': 0.30741569338331043, 'gradnorm/compressor.layer0': 11.254859383449041, 'gradnorm/compressor.layer1': 596.5812235144515, 'gradnorm/compressor.layer2': 80.9871318626839, 'gradnorm/compressor.layer3': 25.863272695305184, 'gradnorm/compressor.layer4': 6.76052008394472, 'gradnorm/compressor.layer5': 3.7306880866474907, 'gradnorm/compressor.layer6': 2.6988404115106506, 'gradnorm/compressor.layer7': 2.238431542694971, 'gradnorm/decoder.layer0': 2.330615663517414, 'gradnorm/decoder.layer1': 1.6334279921732804, 'gradnorm/decoder.layer2': 0.9792821626657433, 'gradnorm/max_param': 420.93218994140625, 'epoch': 7.52}

 75%|███████▌  | 14350/19090 [7:00:24<2:20:57,  1.78s/it]
 75%|███████▌  | 14351/19090 [7:00:25<2:09:18,  1.64s/it]
 75%|███████▌  | 14352/19090 [7:00:27<2:27:36,  1.87s/it]
 75%|███████▌  | 14353/19090 [7:00:29<2:13:41,  1.69s/it]
 75%|███████▌  | 14354/19090 [7:00:30<2:04:07,  1.57s/it]
 75%|███████▌  | 14355/19090 [7:00:32<2:28:44,  1.88s/it]
 75%|███████▌  | 14356/19090 [7:00:34<2:14:28,  1.70s/it]
 75%|███████▌  | 14357/19090 [7:00:35<2:04:33,  1.58s/it]
 75%|███████▌  | 14358/19090 [7:00:38<2:33:11,  1.94s/it]
 75%|███████▌  | 14359/19090 [7:00:39<2:17:40,  1.75s/it][gradnorm] step 14359: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 456.6

 75%|███████▌  | 14360/19090 [7:00:40<2:06:51,  1.61s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 654.1937255859375, 'learning_rate': 8.785447763431101e-06, 'recon_loss': 0.0017584497516509145, 'recon_static': 0.000777398850186728, 'recon_motion': 0.000981050921836868, 'recon_norm': 1.3799589455127717, 'recon_norm_std': 0.037610862869769335, 'target_norm': 1.6212876915931702, 'recon_cos': 0.5955919429659844, 'gradnorm/compressor.query': 0.36981722470286293, 'gradnorm/compressor.layer0': 14.201541892141776, 'gradnorm/compressor.layer1': 646.9669620622061, 'gradnorm/compressor.layer2': 91.84016133683019, 'gradnorm/compressor.layer3': 26.00177700132363, 'gradnorm/compressor.layer4': 7.2993050231375305, 'gradnorm/compressor.layer5': 3.6169387369322683, 'gradnorm/compressor.layer6': 2.6377809998577306, 'gradnorm/compressor.layer7': 2.645264652042533, 'gradnorm/decoder.layer0': 2.5229859317140866, 'gradnorm/decoder.layer1': 1.7204773978910775, 'gradnorm/decoder.layer2': 1.0591352140757342, 'gradnorm/max_param': 456.5712890625, 'epoch': 7.52}

 75%|███████▌  | 14360/19090 [7:00:40<2:06:51,  1.61s/it]
 75%|███████▌  | 14361/19090 [7:00:43<2:33:14,  1.94s/it]
 75%|███████▌  | 14362/19090 [7:00:44<2:17:34,  1.75s/it]
 75%|███████▌  | 14363/19090 [7:00:46<2:06:50,  1.61s/it]
 75%|███████▌  | 14364/19090 [7:00:49<2:43:34,  2.08s/it]
 75%|███████▌  | 14365/19090 [7:00:50<2:24:45,  1.84s/it]
 75%|███████▌  | 14366/19090 [7:00:51<2:11:48,  1.67s/it]
 75%|███████▌  | 14367/19090 [7:00:54<2:29:15,  1.90s/it]
 75%|███████▌  | 14368/19090 [7:00:55<2:14:46,  1.71s/it]
 75%|███████▌  | 14369/19090 [7:00:56<2:04:44,  1.59s/it][gradnorm] step 14369: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 370.8

 75%|███████▌  | 14370/19090 [7:00:59<2:34:32,  1.96s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 532.9591674804688, 'learning_rate': 8.750680543552983e-06, 'recon_loss': 0.0017185575765324757, 'recon_static': 0.0007425846051773988, 'recon_motion': 0.0009759729859069921, 'recon_norm': 1.3541080087423325, 'recon_norm_std': 0.038947055023163556, 'target_norm': 1.6212441474199295, 'recon_cos': 0.6007129535079002, 'gradnorm/compressor.query': 0.3305410490917837, 'gradnorm/compressor.layer0': 12.61325282108131, 'gradnorm/compressor.layer1': 525.4620585732142, 'gradnorm/compressor.layer2': 83.79097644958554, 'gradnorm/compressor.layer3': 25.573936832587734, 'gradnorm/compressor.layer4': 7.624686813176226, 'gradnorm/compressor.layer5': 4.1400853219201075, 'gradnorm/compressor.layer6': 2.8113732518121175, 'gradnorm/compressor.layer7': 2.6848374418069008, 'gradnorm/decoder.layer0': 2.6226640025624546, 'gradnorm/decoder.layer1': 1.57168769402999, 'gradnorm/decoder.layer2': 1.1235369599819833, 'gradnorm/max_param': 370.8115234375, 'epoch': 7.53}

 75%|███████▌  | 14370/19090 [7:00:59<2:34:32,  1.96s/it]
 75%|███████▌  | 14371/19090 [7:01:01<2:18:30,  1.76s/it]
 75%|███████▌  | 14372/19090 [7:01:02<2:07:22,  1.62s/it]
 75%|███████▌  | 14373/19090 [7:01:05<2:41:57,  2.06s/it]
 75%|███████▌  | 14374/19090 [7:01:06<2:23:51,  1.83s/it]
 75%|███████▌  | 14375/19090 [7:01:08<2:11:03,  1.67s/it]
 75%|███████▌  | 14376/19090 [7:01:10<2:31:29,  1.93s/it]
 75%|███████▌  | 14377/19090 [7:01:11<2:16:13,  1.73s/it]
 75%|███████▌  | 14378/19090 [7:01:13<2:05:47,  1.60s/it]
 75%|███████▌  | 14379/19090 [7:01:15<2:27:39,  1.88s/it][gradnorm] step 14379: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 424.8

 75%|███████▌  | 14380/19090 [7:01:16<2:13:49,  1.70s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 612.8749389648438, 'learning_rate': 8.715967653452253e-06, 'recon_loss': 0.0017132659471826629, 'recon_static': 0.0007467009621905163, 'recon_motion': 0.0009665650082752109, 'recon_norm': 1.36839220225811, 'recon_norm_std': 0.03800082420930266, 'target_norm': 1.6173265546560287, 'recon_cos': 0.6033789768815041, 'gradnorm/compressor.query': 0.3368219111372355, 'gradnorm/compressor.layer0': 13.465145186283559, 'gradnorm/compressor.layer1': 601.9700210558663, 'gradnorm/compressor.layer2': 110.8177214294379, 'gradnorm/compressor.layer3': 26.456381546079175, 'gradnorm/compressor.layer4': 6.874372418976525, 'gradnorm/compressor.layer5': 4.040685685388739, 'gradnorm/compressor.layer6': 2.221859968646962, 'gradnorm/compressor.layer7': 2.586002314466959, 'gradnorm/decoder.layer0': 2.675524622730267, 'gradnorm/decoder.layer1': 1.5408403226791205, 'gradnorm/decoder.layer2': 1.070459954344384, 'gradnorm/max_param': 424.7760925292969, 'epoch': 7.53}

 75%|███████▌  | 14380/19090 [7:01:16<2:13:49,  1.70s/it]
 75%|███████▌  | 14381/19090 [7:01:18<2:04:06,  1.58s/it]
 75%|███████▌  | 14382/19090 [7:01:20<2:31:52,  1.94s/it]
 75%|███████▌  | 14383/19090 [7:01:22<2:16:39,  1.74s/it]
 75%|███████▌  | 14384/19090 [7:01:23<2:05:54,  1.61s/it]
 75%|███████▌  | 14385/19090 [7:01:26<2:28:00,  1.89s/it]
 75%|███████▌  | 14386/19090 [7:01:27<2:13:58,  1.71s/it]
 75%|███████▌  | 14387/19090 [7:01:28<2:04:04,  1.58s/it]
 75%|███████▌  | 14388/19090 [7:01:31<2:29:42,  1.91s/it]
 75%|███████▌  | 14389/19090 [7:01:32<2:15:01,  1.72s/it][gradnorm] step 14389: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 555.8

 75%|███████▌  | 14390/19090 [7:01:33<2:04:57,  1.60s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 798.492431640625, 'learning_rate': 8.681309209191843e-06, 'recon_loss': 0.0017696655209874733, 'recon_static': 0.0007971255443408154, 'recon_motion': 0.0009725399169838056, 'recon_norm': 1.3853169143199922, 'recon_norm_std': 0.03765651760622859, 'target_norm': 1.6194676637649537, 'recon_cos': 0.5925828143954277, 'gradnorm/compressor.query': 0.30026168039965245, 'gradnorm/compressor.layer0': 11.694782782759905, 'gradnorm/compressor.layer1': 787.7125903525981, 'gradnorm/compressor.layer2': 126.42748971742657, 'gradnorm/compressor.layer3': 29.534650721850745, 'gradnorm/compressor.layer4': 7.853694372033243, 'gradnorm/compressor.layer5': 4.432203113404953, 'gradnorm/compressor.layer6': 2.5259706991917086, 'gradnorm/compressor.layer7': 2.7615927919749037, 'gradnorm/decoder.layer0': 2.6302440577935635, 'gradnorm/decoder.layer1': 1.6494128338463083, 'gradnorm/decoder.layer2': 0.9942897242647655, 'gradnorm/max_param': 555.8378295898438, 'epoch': 7.54}

 75%|███████▌  | 14390/19090 [7:01:33<2:04:57,  1.60s/it]
 75%|███████▌  | 14391/19090 [7:01:36<2:31:19,  1.93s/it]
 75%|███████▌  | 14392/19090 [7:01:37<2:16:10,  1.74s/it]
 75%|███████▌  | 14393/19090 [7:01:39<2:05:39,  1.61s/it]
 75%|███████▌  | 14394/19090 [7:01:42<2:38:01,  2.02s/it]
 75%|███████▌  | 14395/19090 [7:01:43<2:20:43,  1.80s/it]
 75%|███████▌  | 14396/19090 [7:01:44<2:08:37,  1.64s/it]
 75%|███████▌  | 14397/19090 [7:01:47<2:39:25,  2.04s/it]
 75%|███████▌  | 14398/19090 [7:01:49<2:21:38,  1.81s/it]
 75%|███████▌  | 14399/19090 [7:01:50<2:09:16,  1.65s/it][gradnorm] step 14399: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 404.1

 75%|███████▌  | 14400/19090 [7:01:52<2:26:26,  1.87s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 581.5797119140625, 'learning_rate': 8.646705326652637e-06, 'recon_loss': 0.0016195818199776113, 'recon_static': 0.0006518350797705353, 'recon_motion': 0.0009677467489382253, 'recon_norm': 1.3257561892271041, 'recon_norm_std': 0.04027102729305625, 'target_norm': 1.6203162491321563, 'recon_cos': 0.6219434142112732, 'gradnorm/compressor.query': 0.3505615608432627, 'gradnorm/compressor.layer0': 10.611954882385302, 'gradnorm/compressor.layer1': 572.6337180082919, 'gradnorm/compressor.layer2': 97.1824657056251, 'gradnorm/compressor.layer3': 25.670709993346502, 'gradnorm/compressor.layer4': 8.09880043848967, 'gradnorm/compressor.layer5': 4.1192600889504325, 'gradnorm/compressor.layer6': 2.5518034074332583, 'gradnorm/compressor.layer7': 2.903692453235376, 'gradnorm/decoder.layer0': 2.8726463635160138, 'gradnorm/decoder.layer1': 1.6469606409841888, 'gradnorm/decoder.layer2': 1.002417622175836, 'gradnorm/max_param': 404.0916442871094, 'epoch': 7.54}

 75%|███████▌  | 14400/19090 [7:01:52<2:26:26,  1.87s/it]
 75%|███████▌  | 14401/19090 [7:01:54<2:12:36,  1.70s/it]
 75%|███████▌  | 14402/19090 [7:01:55<2:02:53,  1.57s/it]
 75%|███████▌  | 14403/19090 [7:01:58<2:34:05,  1.97s/it]
 75%|███████▌  | 14404/19090 [7:01:59<2:17:55,  1.77s/it]
 75%|███████▌  | 14405/19090 [7:02:00<2:06:36,  1.62s/it]
 75%|███████▌  | 14406/19090 [7:02:03<2:39:38,  2.04s/it]
 75%|███████▌  | 14407/19090 [7:02:05<2:21:46,  1.82s/it]
 75%|███████▌  | 14408/19090 [7:02:06<2:09:18,  1.66s/it]
 75%|███████▌  | 14409/19090 [7:02:09<2:40:09,  2.05s/it][gradnorm] step 14409: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 427.9

 75%|███████▌  | 14410/19090 [7:02:10<2:22:10,  1.82s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 615.1043090820312, 'learning_rate': 8.612156121533099e-06, 'recon_loss': 0.001589050656184554, 'recon_static': 0.0006097591307479888, 'recon_motion': 0.0009792915516300127, 'recon_norm': 1.3007392287254333, 'recon_norm_std': 0.04075511302798986, 'target_norm': 1.6208706349134445, 'recon_cos': 0.6269531428813935, 'gradnorm/compressor.query': 0.26136532448894406, 'gradnorm/compressor.layer0': 10.577228818089115, 'gradnorm/compressor.layer1': 606.407221881468, 'gradnorm/compressor.layer2': 99.3913950705115, 'gradnorm/compressor.layer3': 23.2637951030707, 'gradnorm/compressor.layer4': 6.967880400041831, 'gradnorm/compressor.layer5': 3.915083396091589, 'gradnorm/compressor.layer6': 2.589271815972071, 'gradnorm/compressor.layer7': 3.0019388291583806, 'gradnorm/decoder.layer0': 2.9041459865919257, 'gradnorm/decoder.layer1': 1.6235454358227202, 'gradnorm/decoder.layer2': 1.0569824681158897, 'gradnorm/max_param': 427.8886413574219, 'epoch': 7.55}

 75%|███████▌  | 14410/19090 [7:02:10<2:22:10,  1.82s/it]
 75%|███████▌  | 14411/19090 [7:02:11<2:09:40,  1.66s/it]
 75%|███████▌  | 14412/19090 [7:02:14<2:29:02,  1.91s/it]
 76%|███████▌  | 14413/19090 [7:02:15<2:14:19,  1.72s/it]
 76%|███████▌  | 14414/19090 [7:02:16<2:03:59,  1.59s/it]
 76%|███████▌  | 14415/19090 [7:02:19<2:28:07,  1.90s/it]
 76%|███████▌  | 14416/19090 [7:02:20<2:13:35,  1.71s/it]
 76%|███████▌  | 14417/19090 [7:02:22<2:03:32,  1.59s/it]
 76%|███████▌  | 14418/19090 [7:02:24<2:27:48,  1.90s/it]
 76%|███████▌  | 14419/19090 [7:02:26<2:13:30,  1.72s/it][gradnorm] step 14419: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 425.7

 76%|███████▌  | 14420/19090 [7:02:27<2:03:31,  1.59s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 613.0768432617188, 'learning_rate': 8.577661709348873e-06, 'recon_loss': 0.001657302217790857, 'recon_static': 0.0006905041242134758, 'recon_motion': 0.0009667980775702745, 'recon_norm': 1.3402214765548706, 'recon_norm_std': 0.03950673136860132, 'target_norm': 1.619042119383812, 'recon_cos': 0.6132732421159744, 'gradnorm/compressor.query': 0.2657926675406201, 'gradnorm/compressor.layer0': 12.070259405223029, 'gradnorm/compressor.layer1': 603.3778097096379, 'gradnorm/compressor.layer2': 103.46154365572264, 'gradnorm/compressor.layer3': 29.45336745589861, 'gradnorm/compressor.layer4': 6.324460322459247, 'gradnorm/compressor.layer5': 4.029352258920141, 'gradnorm/compressor.layer6': 2.4410908976019767, 'gradnorm/compressor.layer7': 2.697707110383161, 'gradnorm/decoder.layer0': 2.878810016571297, 'gradnorm/decoder.layer1': 1.4971549550792773, 'gradnorm/decoder.layer2': 0.9589106576093174, 'gradnorm/max_param': 425.7304992675781, 'epoch': 7.55}

 76%|███████▌  | 14420/19090 [7:02:27<2:03:31,  1.59s/it]
 76%|███████▌  | 14421/19090 [7:02:30<2:33:28,  1.97s/it]
 76%|███████▌  | 14422/19090 [7:02:31<2:17:25,  1.77s/it]
 76%|███████▌  | 14423/19090 [7:02:32<2:06:14,  1.62s/it]
 76%|███████▌  | 14424/19090 [7:02:35<2:41:02,  2.07s/it]
 76%|███████▌  | 14425/19090 [7:02:37<2:22:41,  1.84s/it]
 76%|███████▌  | 14426/19090 [7:02:38<2:09:46,  1.67s/it]
 76%|███████▌  | 14427/19090 [7:02:41<2:34:33,  1.99s/it]
 76%|███████▌  | 14428/19090 [7:02:42<2:18:16,  1.78s/it]
 76%|███████▌  | 14429/19090 [7:02:43<2:06:51,  1.63s/it][gradnorm] step 14429: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 469.1

 76%|███████▌  | 14430/19090 [7:02:46<2:36:29,  2.01s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 676.4176025390625, 'learning_rate': 8.543222205432402e-06, 'recon_loss': 0.0016751006536651404, 'recon_static': 0.0006901926681166515, 'recon_motion': 0.0009849079986452125, 'recon_norm': 1.3383480995893478, 'recon_norm_std': 0.03930507870391011, 'target_norm': 1.6224935531616211, 'recon_cos': 0.610081335902214, 'gradnorm/compressor.query': 0.28629756519809246, 'gradnorm/compressor.layer0': 10.558036628614804, 'gradnorm/compressor.layer1': 664.6843705850168, 'gradnorm/compressor.layer2': 121.89017934317965, 'gradnorm/compressor.layer3': 26.001163163209522, 'gradnorm/compressor.layer4': 7.128060737530909, 'gradnorm/compressor.layer5': 3.7582681423265245, 'gradnorm/compressor.layer6': 2.5187956926541317, 'gradnorm/compressor.layer7': 2.9263480316127852, 'gradnorm/decoder.layer0': 2.74708541394045, 'gradnorm/decoder.layer1': 1.5045811315474786, 'gradnorm/decoder.layer2': 0.9747927384283331, 'gradnorm/max_param': 469.0664367675781, 'epoch': 7.56}

 76%|███████▌  | 14430/19090 [7:02:46<2:36:29,  2.01s/it]
 76%|███████▌  | 14431/19090 [7:02:48<2:19:22,  1.79s/it]
 76%|███████▌  | 14432/19090 [7:02:49<2:07:28,  1.64s/it]
 76%|███████▌  | 14433/19090 [7:02:52<2:33:14,  1.97s/it]
 76%|███████▌  | 14434/19090 [7:02:53<2:17:21,  1.77s/it]
 76%|███████▌  | 14435/19090 [7:02:54<2:06:11,  1.63s/it]
 76%|███████▌  | 14436/19090 [7:02:57<2:32:24,  1.96s/it]
 76%|███████▌  | 14437/19090 [7:02:58<2:16:42,  1.76s/it]
 76%|███████▌  | 14438/19090 [7:02:59<2:05:36,  1.62s/it]
 76%|███████▌  | 14439/19090 [7:03:02<2:29:11,  1.92s/it][gradnorm] step 14439: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 531.7

 76%|███████▌  | 14440/19090 [7:03:03<2:14:28,  1.74s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 763.8569946289062, 'learning_rate': 8.508837724932542e-06, 'recon_loss': 0.0016053639323217795, 'recon_static': 0.0006290922814514488, 'recon_motion': 0.0009762716508703306, 'recon_norm': 1.3105804562568664, 'recon_norm_std': 0.04032314596697688, 'target_norm': 1.6209917455911635, 'recon_cos': 0.6241805732250214, 'gradnorm/compressor.query': 0.3290158168576403, 'gradnorm/compressor.layer0': 12.104403591408829, 'gradnorm/compressor.layer1': 753.7703977339519, 'gradnorm/compressor.layer2': 118.61115237931887, 'gradnorm/compressor.layer3': 31.328122077618318, 'gradnorm/compressor.layer4': 7.985389719644832, 'gradnorm/compressor.layer5': 4.113838817341229, 'gradnorm/compressor.layer6': 2.8945075277367995, 'gradnorm/compressor.layer7': 2.939888328196019, 'gradnorm/decoder.layer0': 3.102645262238501, 'gradnorm/decoder.layer1': 1.6503538590348217, 'gradnorm/decoder.layer2': 0.9616562551092374, 'gradnorm/max_param': 531.7218627929688, 'epoch': 7.56}

 76%|███████▌  | 14440/19090 [7:03:03<2:14:28,  1.74s/it]
 76%|███████▌  | 14441/19090 [7:03:05<2:04:06,  1.60s/it]
 76%|███████▌  | 14442/19090 [7:03:08<2:36:37,  2.02s/it]
 76%|███████▌  | 14443/19090 [7:03:09<2:19:23,  1.80s/it]
 76%|███████▌  | 14444/19090 [7:03:10<2:07:29,  1.65s/it]
 76%|███████▌  | 14445/19090 [7:03:13<2:27:29,  1.91s/it]
 76%|███████▌  | 14446/19090 [7:03:14<2:13:03,  1.72s/it]
 76%|███████▌  | 14447/19090 [7:03:15<2:02:56,  1.59s/it]
 76%|███████▌  | 14448/19090 [7:03:18<2:17:03,  1.77s/it]
 76%|███████▌  | 14449/19090 [7:03:19<2:05:45,  1.63s/it][gradnorm] step 14449: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 499.0

 76%|███████▌  | 14450/19090 [7:03:20<1:57:48,  1.52s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 717.506591796875, 'learning_rate': 8.474508382814186e-06, 'recon_loss': 0.0016696496430085972, 'recon_static': 0.0007064439880196006, 'recon_motion': 0.0009632056331611239, 'recon_norm': 1.3508109837770461, 'recon_norm_std': 0.0399119172245264, 'target_norm': 1.6205657601356507, 'recon_cos': 0.6132856398820877, 'gradnorm/compressor.query': 0.31006821745107266, 'gradnorm/compressor.layer0': 11.62367552728791, 'gradnorm/compressor.layer1': 707.110007353877, 'gradnorm/compressor.layer2': 117.23036299445847, 'gradnorm/compressor.layer3': 28.90302980374095, 'gradnorm/compressor.layer4': 7.472114858442934, 'gradnorm/compressor.layer5': 3.8053476751743824, 'gradnorm/compressor.layer6': 2.4619083487698967, 'gradnorm/compressor.layer7': 2.9638655786651316, 'gradnorm/decoder.layer0': 2.8703476124434415, 'gradnorm/decoder.layer1': 1.776107592957897, 'gradnorm/decoder.layer2': 0.9920683661598483, 'gradnorm/max_param': 498.981689453125, 'epoch': 7.57}

 76%|███████▌  | 14450/19090 [7:03:20<1:57:48,  1.52s/it]
 76%|███████▌  | 14451/19090 [7:03:22<2:04:18,  1.61s/it]
 76%|███████▌  | 14452/19090 [7:03:23<1:56:48,  1.51s/it]
 76%|███████▌  | 14453/19090 [7:03:24<1:51:35,  1.44s/it]
 76%|███████▌  | 14454/19090 [7:03:27<2:09:11,  1.67s/it]
 76%|███████▌  | 14455/19090 [7:03:28<2:00:20,  1.56s/it]
 76%|███████▌  | 14456/19090 [7:03:29<1:54:05,  1.48s/it]
 76%|███████▌  | 14457/19090 [7:03:31<2:08:29,  1.66s/it]
 76%|███████▌  | 14458/19090 [7:03:33<1:59:53,  1.55s/it]
 76%|███████▌  | 14459/19090 [7:03:34<1:54:16,  1.48s/it][gradnorm] step 14459: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 510.4

 76%|███████▌  | 14460/19090 [7:03:36<2:16:57,  1.77s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 736.89599609375, 'learning_rate': 8.44023429385786e-06, 'recon_loss': 0.0017118236923124642, 'recon_static': 0.0007371462153969333, 'recon_motion': 0.0009746775001985952, 'recon_norm': 1.3613410711288452, 'recon_norm_std': 0.03830684889107942, 'target_norm': 1.618684071302414, 'recon_cos': 0.6029017582535744, 'gradnorm/compressor.query': 0.35715416226205315, 'gradnorm/compressor.layer0': 14.918089445374543, 'gradnorm/compressor.layer1': 723.2212446755142, 'gradnorm/compressor.layer2': 136.04772829782937, 'gradnorm/compressor.layer3': 33.549469362886356, 'gradnorm/compressor.layer4': 8.294194997363931, 'gradnorm/compressor.layer5': 3.832575114908196, 'gradnorm/compressor.layer6': 2.5400138129211665, 'gradnorm/compressor.layer7': 2.7824116541846005, 'gradnorm/decoder.layer0': 2.8146737916887252, 'gradnorm/decoder.layer1': 1.7379110220640102, 'gradnorm/decoder.layer2': 0.9539523989099975, 'gradnorm/max_param': 510.3926086425781, 'epoch': 7.57}

 76%|███████▌  | 14460/19090 [7:03:36<2:16:57,  1.77s/it]
 76%|███████▌  | 14461/19090 [7:03:38<2:05:41,  1.63s/it]
 76%|███████▌  | 14462/19090 [7:03:39<1:57:49,  1.53s/it]
 76%|███████▌  | 14463/19090 [7:03:41<2:17:06,  1.78s/it]
 76%|███████▌  | 14464/19090 [7:03:43<2:05:48,  1.63s/it]
 76%|███████▌  | 14465/19090 [7:03:44<1:57:55,  1.53s/it]
 76%|███████▌  | 14466/19090 [7:03:46<2:21:16,  1.83s/it]
 76%|███████▌  | 14467/19090 [7:03:48<2:08:40,  1.67s/it]
 76%|███████▌  | 14468/19090 [7:03:49<1:59:46,  1.55s/it]
 76%|███████▌  | 14469/19090 [7:03:52<2:25:11,  1.89s/it][gradnorm] step 14469: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 474.0

 76%|███████▌  | 14470/19090 [7:03:53<2:11:19,  1.71s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 681.7734985351562, 'learning_rate': 8.406015572659357e-06, 'recon_loss': 0.0017110579530708492, 'recon_static': 0.0007390907645458356, 'recon_motion': 0.0009719671739730984, 'recon_norm': 1.3613463014364242, 'recon_norm_std': 0.038409820012748244, 'target_norm': 1.620400607585907, 'recon_cos': 0.6037049725651741, 'gradnorm/compressor.query': 0.3058254681107696, 'gradnorm/compressor.layer0': 12.310869602752543, 'gradnorm/compressor.layer1': 671.7190400383184, 'gradnorm/compressor.layer2': 112.45992341718893, 'gradnorm/compressor.layer3': 26.839953756960558, 'gradnorm/compressor.layer4': 6.499381696197057, 'gradnorm/compressor.layer5': 4.636360855237701, 'gradnorm/compressor.layer6': 2.472049102246154, 'gradnorm/compressor.layer7': 2.742196194452919, 'gradnorm/decoder.layer0': 2.902797001273604, 'gradnorm/decoder.layer1': 1.6306217957096338, 'gradnorm/decoder.layer2': 0.921462952937679, 'gradnorm/max_param': 474.0184020996094, 'epoch': 7.58}

 76%|███████▌  | 14470/19090 [7:03:53<2:11:19,  1.71s/it]
 76%|███████▌  | 14471/19090 [7:03:54<2:01:43,  1.58s/it]
 76%|███████▌  | 14472/19090 [7:03:57<2:20:09,  1.82s/it]
 76%|███████▌  | 14473/19090 [7:03:58<2:07:47,  1.66s/it]
 76%|███████▌  | 14474/19090 [7:03:59<1:59:16,  1.55s/it]
 76%|███████▌  | 14475/19090 [7:04:02<2:22:36,  1.85s/it]
 76%|███████▌  | 14476/19090 [7:04:03<2:09:31,  1.68s/it]
 76%|███████▌  | 14477/19090 [7:04:04<2:00:24,  1.57s/it]
 76%|███████▌  | 14478/19090 [7:04:07<2:18:35,  1.80s/it]
 76%|███████▌  | 14479/19090 [7:04:08<2:06:40,  1.65s/it][gradnorm] step 14479: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 497.6

 76%|███████▌  | 14480/19090 [7:04:09<1:58:30,  1.54s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 714.4571533203125, 'learning_rate': 8.371852333629346e-06, 'recon_loss': 0.0017598685255507008, 'recon_static': 0.0007829739741282537, 'recon_motion': 0.000976894584891852, 'recon_norm': 1.3792167544364928, 'recon_norm_std': 0.037601604126393795, 'target_norm': 1.618503412604332, 'recon_cos': 0.5937386348843574, 'gradnorm/compressor.query': 0.30786711213186485, 'gradnorm/compressor.layer0': 11.917276077871609, 'gradnorm/compressor.layer1': 705.2114709432909, 'gradnorm/compressor.layer2': 110.20048314038827, 'gradnorm/compressor.layer3': 27.111186223716423, 'gradnorm/compressor.layer4': 7.782116209787459, 'gradnorm/compressor.layer5': 4.318232389090069, 'gradnorm/compressor.layer6': 2.4601852962830093, 'gradnorm/compressor.layer7': 2.6852241460913833, 'gradnorm/decoder.layer0': 2.8554495459874922, 'gradnorm/decoder.layer1': 1.7187678943049316, 'gradnorm/decoder.layer2': 0.9473069958648676, 'gradnorm/max_param': 497.6227722167969, 'epoch': 7.59}

 76%|███████▌  | 14480/19090 [7:04:09<1:58:30,  1.54s/it]
 76%|███████▌  | 14481/19090 [7:04:12<2:18:17,  1.80s/it]
 76%|███████▌  | 14482/19090 [7:04:13<2:06:28,  1.65s/it]
 76%|███████▌  | 14483/19090 [7:04:14<1:58:09,  1.54s/it]
 76%|███████▌  | 14484/19090 [7:04:17<2:19:32,  1.82s/it]
 76%|███████▌  | 14485/19090 [7:04:18<2:07:19,  1.66s/it]
 76%|███████▌  | 14486/19090 [7:04:19<1:59:02,  1.55s/it]
 76%|███████▌  | 14487/19090 [7:04:22<2:21:39,  1.85s/it]
 76%|███████▌  | 14488/19090 [7:04:23<2:08:48,  1.68s/it]
 76%|███████▌  | 14489/19090 [7:04:24<1:59:44,  1.56s/it][gradnorm] step 14489: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 521.2

 76%|███████▌  | 14490/19090 [7:04:27<2:14:47,  1.76s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 751.1480102539062, 'learning_rate': 8.33774469099299e-06, 'recon_loss': 0.0017347693821648136, 'recon_static': 0.0007662821590201929, 'recon_motion': 0.0009684872056823224, 'recon_norm': 1.3716053664684296, 'recon_norm_std': 0.03898097751662135, 'target_norm': 1.6203664243221283, 'recon_cos': 0.5995295271277428, 'gradnorm/compressor.query': 0.4326435709673254, 'gradnorm/compressor.layer0': 19.28946435385373, 'gradnorm/compressor.layer1': 738.7486040595949, 'gradnorm/compressor.layer2': 129.88636108874172, 'gradnorm/compressor.layer3': 33.31130355203677, 'gradnorm/compressor.layer4': 8.04676122492382, 'gradnorm/compressor.layer5': 5.167636206113994, 'gradnorm/compressor.layer6': 2.7333496787195024, 'gradnorm/compressor.layer7': 2.7935554795103053, 'gradnorm/decoder.layer0': 3.1430058351244696, 'gradnorm/decoder.layer1': 1.905815950599909, 'gradnorm/decoder.layer2': 1.0829856877178183, 'gradnorm/max_param': 521.2067260742188, 'epoch': 7.59}

 76%|███████▌  | 14490/19090 [7:04:27<2:14:47,  1.76s/it]
 76%|███████▌  | 14491/19090 [7:04:28<2:03:56,  1.62s/it]
 76%|███████▌  | 14492/19090 [7:04:29<1:56:14,  1.52s/it]
 76%|███████▌  | 14493/19090 [7:04:32<2:16:13,  1.78s/it]
 76%|███████▌  | 14494/19090 [7:04:33<2:04:50,  1.63s/it]
 76%|███████▌  | 14495/19090 [7:04:34<1:57:00,  1.53s/it]
 76%|███████▌  | 14496/19090 [7:04:37<2:22:51,  1.87s/it]
 76%|███████▌  | 14497/19090 [7:04:38<2:09:31,  1.69s/it]
 76%|███████▌  | 14498/19090 [7:04:39<2:00:12,  1.57s/it]
 76%|███████▌  | 14499/19090 [7:04:42<2:27:58,  1.93s/it][gradnorm] step 14499: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 347.8

 76%|███████▌  | 14500/19090 [7:04:44<2:13:02,  1.74s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 501.04559326171875, 'learning_rate': 8.303692758789566e-06, 'recon_loss': 0.0018693121906835585, 'recon_static': 0.0008856818560161628, 'recon_motion': 0.0009836302793701178, 'recon_norm': 1.4246632277965545, 'recon_norm_std': 0.0366537538357079, 'target_norm': 1.6236267149448396, 'recon_cos': 0.577171765267849, 'gradnorm/compressor.query': 0.32928356492191, 'gradnorm/compressor.layer0': 11.060047977060625, 'gradnorm/compressor.layer1': 493.0929330511643, 'gradnorm/compressor.layer2': 83.38863562129585, 'gradnorm/compressor.layer3': 27.137170548725692, 'gradnorm/compressor.layer4': 7.287762394102227, 'gradnorm/compressor.layer5': 4.161482192230898, 'gradnorm/compressor.layer6': 2.6149719881559594, 'gradnorm/compressor.layer7': 2.3230500689007174, 'gradnorm/decoder.layer0': 2.6422335448671714, 'gradnorm/decoder.layer1': 1.7486439628164614, 'gradnorm/decoder.layer2': 0.9015156963503138, 'gradnorm/max_param': 347.8344421386719, 'epoch': 7.6}

 76%|███████▌  | 14500/19090 [7:04:44<2:13:02,  1.74s/it]
 76%|███████▌  | 14501/19090 [7:04:47<2:52:47,  2.26s/it]
 76%|███████▌  | 14502/19090 [7:04:48<2:30:18,  1.97s/it]
 76%|███████▌  | 14503/19090 [7:04:50<2:14:40,  1.76s/it]
 76%|███████▌  | 14504/19090 [7:04:51<2:03:34,  1.62s/it]
 76%|███████▌  | 14505/19090 [7:04:52<1:56:24,  1.52s/it]
 76%|███████▌  | 14506/19090 [7:04:53<1:50:58,  1.45s/it]
 76%|███████▌  | 14507/19090 [7:04:55<1:47:06,  1.40s/it]
 76%|███████▌  | 14508/19090 [7:04:58<2:27:11,  1.93s/it]
 76%|███████▌  | 14509/19090 [7:04:59<2:12:40,  1.74s/it][gradnorm] step 14509: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 468.3

 76%|███████▌  | 14510/19090 [7:05:00<2:02:45,  1.61s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 671.5151977539062, 'learning_rate': 8.269696650872082e-06, 'recon_loss': 0.0017223234928678722, 'recon_static': 0.0007520724146161228, 'recon_motion': 0.0009702510869828984, 'recon_norm': 1.3634005278348922, 'recon_norm_std': 0.03881688332185149, 'target_norm': 1.6204368323087692, 'recon_cos': 0.6011253729462623, 'gradnorm/compressor.query': 0.42456016079850545, 'gradnorm/compressor.layer0': 16.669597724037914, 'gradnorm/compressor.layer1': 663.7930306202378, 'gradnorm/compressor.layer2': 95.67029666548808, 'gradnorm/compressor.layer3': 28.139325778586763, 'gradnorm/compressor.layer4': 6.740599125797871, 'gradnorm/compressor.layer5': 4.330756864231684, 'gradnorm/compressor.layer6': 2.43258013978463, 'gradnorm/compressor.layer7': 2.6188088021243057, 'gradnorm/decoder.layer0': 2.7749348641389084, 'gradnorm/decoder.layer1': 1.7790240383228846, 'gradnorm/decoder.layer2': 0.995541647698501, 'gradnorm/max_param': 468.2793273925781, 'epoch': 7.6}

 76%|███████▌  | 14510/19090 [7:05:00<2:02:45,  1.61s/it]
 76%|███████▌  | 14511/19090 [7:05:03<2:26:53,  1.92s/it]
 76%|███████▌  | 14512/19090 [7:05:04<2:12:26,  1.74s/it]
 76%|███████▌  | 14513/19090 [7:05:06<2:02:29,  1.61s/it]
 76%|███████▌  | 14514/19090 [7:05:09<2:29:41,  1.96s/it]
 76%|███████▌  | 14515/19090 [7:05:10<2:14:12,  1.76s/it]
 76%|███████▌  | 14516/19090 [7:05:11<2:03:22,  1.62s/it]
 76%|███████▌  | 14517/19090 [7:05:14<2:35:20,  2.04s/it]
 76%|███████▌  | 14518/19090 [7:05:15<2:18:05,  1.81s/it]
 76%|███████▌  | 14519/19090 [7:05:17<2:06:02,  1.65s/it][gradnorm] step 14519: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 453.4

 76%|███████▌  | 14520/19090 [7:05:19<2:27:55,  1.94s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 652.90234375, 'learning_rate': 8.235756480906898e-06, 'recon_loss': 0.0016716786194592714, 'recon_static': 0.0007065760903060437, 'recon_motion': 0.0009651024985942059, 'recon_norm': 1.3505286544561386, 'recon_norm_std': 0.03901052437722683, 'target_norm': 1.616540190577507, 'recon_cos': 0.6103105291724205, 'gradnorm/compressor.query': 0.3223910001146993, 'gradnorm/compressor.layer0': 10.814815912677277, 'gradnorm/compressor.layer1': 642.4826845915771, 'gradnorm/compressor.layer2': 112.64222346798957, 'gradnorm/compressor.layer3': 24.139413905704867, 'gradnorm/compressor.layer4': 7.930444660587982, 'gradnorm/compressor.layer5': 4.628205503063673, 'gradnorm/compressor.layer6': 2.54414717739157, 'gradnorm/compressor.layer7': 2.7908501545978246, 'gradnorm/decoder.layer0': 2.4908722184039767, 'gradnorm/decoder.layer1': 1.9599632487452117, 'gradnorm/decoder.layer2': 0.9119206627439214, 'gradnorm/max_param': 453.40130615234375, 'epoch': 7.61}

 76%|███████▌  | 14520/19090 [7:05:19<2:27:55,  1.94s/it]
 76%|███████▌  | 14521/19090 [7:05:21<2:12:56,  1.75s/it]
 76%|███████▌  | 14522/19090 [7:05:22<2:02:30,  1.61s/it]
 76%|███████▌  | 14523/19090 [7:05:24<2:22:00,  1.87s/it]
 76%|███████▌  | 14524/19090 [7:05:26<2:08:45,  1.69s/it]
 76%|███████▌  | 14525/19090 [7:05:27<1:59:23,  1.57s/it]
 76%|███████▌  | 14526/19090 [7:05:29<2:20:04,  1.84s/it]
 76%|███████▌  | 14527/19090 [7:05:31<2:07:23,  1.68s/it]
 76%|███████▌  | 14528/19090 [7:05:32<1:58:26,  1.56s/it]
 76%|███████▌  | 14529/19090 [7:05:35<2:25:31,  1.91s/it][gradnorm] step 14529: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 434.8

 76%|███████▌  | 14530/19090 [7:05:36<2:11:18,  1.73s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 622.9151611328125, 'learning_rate': 8.20187236237334e-06, 'recon_loss': 0.0015840794891119002, 'recon_static': 0.0006144324390334078, 'recon_motion': 0.0009696470195194706, 'recon_norm': 1.3041667252779008, 'recon_norm_std': 0.04047061894088984, 'target_norm': 1.6185525089502335, 'recon_cos': 0.6274040699005127, 'gradnorm/compressor.query': 0.3993659328571652, 'gradnorm/compressor.layer0': 16.737554225021853, 'gradnorm/compressor.layer1': 616.2683060161378, 'gradnorm/compressor.layer2': 85.6177831128338, 'gradnorm/compressor.layer3': 22.97676704261217, 'gradnorm/compressor.layer4': 7.51962205708157, 'gradnorm/compressor.layer5': 3.980084431581764, 'gradnorm/compressor.layer6': 2.3387772767817427, 'gradnorm/compressor.layer7': 2.9098395053611545, 'gradnorm/decoder.layer0': 2.779424604106636, 'gradnorm/decoder.layer1': 1.6937192822588214, 'gradnorm/decoder.layer2': 1.0370835697965146, 'gradnorm/max_param': 434.80572509765625, 'epoch': 7.61}

 76%|███████▌  | 14530/19090 [7:05:36<2:11:18,  1.73s/it]
 76%|███████▌  | 14531/19090 [7:05:37<2:01:14,  1.60s/it]
 76%|███████▌  | 14532/19090 [7:05:40<2:20:28,  1.85s/it]
 76%|███████▌  | 14533/19090 [7:05:41<2:07:36,  1.68s/it]
 76%|███████▌  | 14534/19090 [7:05:42<1:58:31,  1.56s/it]
 76%|███████▌  | 14535/19090 [7:05:45<2:23:28,  1.89s/it]
 76%|███████▌  | 14536/19090 [7:05:46<2:09:42,  1.71s/it]
 76%|███████▌  | 14537/19090 [7:05:48<2:00:13,  1.58s/it]
 76%|███████▌  | 14538/19090 [7:05:50<2:24:38,  1.91s/it]
 76%|███████▌  | 14539/19090 [7:05:51<2:10:26,  1.72s/it][gradnorm] step 14539: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 382.5

 76%|███████▌  | 14540/19090 [7:05:53<2:00:47,  1.59s/it]
                                                         
{'loss': 0.002, 'grad_norm': 549.1376342773438, 'learning_rate': 8.16804440856333e-06, 'recon_loss': 0.001975592173403129, 'recon_static': 0.0010116445540916176, 'recon_motion': 0.0009639476629672572, 'recon_norm': 1.457780858874321, 'recon_norm_std': 0.03758090874180198, 'target_norm': 1.621765348315239, 'recon_cos': 0.5652682162821293, 'gradnorm/compressor.query': 0.6295171341552612, 'gradnorm/compressor.layer0': 20.915064939087568, 'gradnorm/compressor.layer1': 542.1378399447875, 'gradnorm/compressor.layer2': 81.55706998208372, 'gradnorm/compressor.layer3': 21.411918301761943, 'gradnorm/compressor.layer4': 7.483612212809332, 'gradnorm/compressor.layer5': 3.5881609370989005, 'gradnorm/compressor.layer6': 2.4874950466394226, 'gradnorm/compressor.layer7': 2.4288393361717153, 'gradnorm/decoder.layer0': 2.5359367967605793, 'gradnorm/decoder.layer1': 1.6330833392955537, 'gradnorm/decoder.layer2': 0.8919089248182466, 'gradnorm/max_param': 382.4670715332031, 'epoch': 7.62}

 76%|███████▌  | 14540/19090 [7:05:53<2:00:47,  1.59s/it]
 76%|███████▌  | 14541/19090 [7:05:56<2:30:50,  1.99s/it]
 76%|███████▌  | 14542/19090 [7:05:57<2:14:47,  1.78s/it]
 76%|███████▌  | 14543/19090 [7:05:58<2:03:45,  1.63s/it]
 76%|███████▌  | 14544/19090 [7:06:01<2:25:54,  1.93s/it]
 76%|███████▌  | 14545/19090 [7:06:02<2:11:16,  1.73s/it]
 76%|███████▌  | 14546/19090 [7:06:03<2:01:08,  1.60s/it]
 76%|███████▌  | 14547/19090 [7:06:06<2:26:00,  1.93s/it]
 76%|███████▌  | 14548/19090 [7:06:07<2:11:20,  1.73s/it]
 76%|███████▌  | 14549/19090 [7:06:09<2:01:05,  1.60s/it][gradnorm] step 14549: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 358.7

 76%|███████▌  | 14550/19090 [7:06:11<2:20:47,  1.86s/it]
                                                         
{'loss': 0.0028, 'grad_norm': 514.7290649414062, 'learning_rate': 8.134272732581003e-06, 'recon_loss': 0.002776316093513742, 'recon_static': 0.0018061903479974716, 'recon_motion': 0.0009701258197310381, 'recon_norm': 1.757901945710182, 'recon_norm_std': 0.03248031847178936, 'target_norm': 1.620115864276886, 'recon_cos': 0.4706847868859768, 'gradnorm/compressor.query': 0.4137856079464401, 'gradnorm/compressor.layer0': 16.441189544039897, 'gradnorm/compressor.layer1': 508.31802914120607, 'gradnorm/compressor.layer2': 75.36045542715026, 'gradnorm/compressor.layer3': 23.26047993395262, 'gradnorm/compressor.layer4': 6.0323896369055365, 'gradnorm/compressor.layer5': 3.512927431863705, 'gradnorm/compressor.layer6': 2.041515876754997, 'gradnorm/compressor.layer7': 2.4383794714038505, 'gradnorm/decoder.layer0': 2.540296706494174, 'gradnorm/decoder.layer1': 1.4583247684045595, 'gradnorm/decoder.layer2': 0.791436956775837, 'gradnorm/max_param': 358.6976623535156, 'epoch': 7.62}

 76%|███████▌  | 14550/19090 [7:06:11<2:20:47,  1.86s/it]
 76%|███████▌  | 14551/19090 [7:06:12<2:07:43,  1.69s/it]
 76%|███████▌  | 14552/19090 [7:06:14<1:58:29,  1.57s/it]
 76%|███████▌  | 14553/19090 [7:06:16<2:21:06,  1.87s/it]
 76%|███████▌  | 14554/19090 [7:06:18<2:07:53,  1.69s/it]
 76%|███████▌  | 14555/19090 [7:06:19<2:10:50,  1.73s/it]
 76%|███████▌  | 14556/19090 [7:06:22<2:28:38,  1.97s/it]
 76%|███████▋  | 14557/19090 [7:06:23<2:13:10,  1.76s/it]
 76%|███████▋  | 14558/19090 [7:06:25<2:14:58,  1.79s/it]
 76%|███████▋  | 14559/19090 [7:06:27<2:26:19,  1.94s/it][gradnorm] step 14559: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 366.0

 76%|███████▋  | 14560/19090 [7:06:29<2:11:34,  1.74s/it]
                                                         
{'loss': 0.0023, 'grad_norm': 526.7212524414062, 'learning_rate': 8.100557447342327e-06, 'recon_loss': 0.0023143533093389125, 'recon_static': 0.0013415735244052484, 'recon_motion': 0.0009727797660161741, 'recon_norm': 1.5920497208833695, 'recon_norm_std': 0.0338941786903888, 'target_norm': 1.6207090973854066, 'recon_cos': 0.5155063301324845, 'gradnorm/compressor.query': 0.31378719352546425, 'gradnorm/compressor.layer0': 12.718938952642814, 'gradnorm/compressor.layer1': 518.7406926875893, 'gradnorm/compressor.layer2': 86.0790623814453, 'gradnorm/compressor.layer3': 26.62923831069549, 'gradnorm/compressor.layer4': 5.542937780701012, 'gradnorm/compressor.layer5': 3.2716120544114506, 'gradnorm/compressor.layer6': 2.372542264057433, 'gradnorm/compressor.layer7': 2.63398757203457, 'gradnorm/decoder.layer0': 2.3837489554888514, 'gradnorm/decoder.layer1': 1.5062141445846682, 'gradnorm/decoder.layer2': 0.919473656314841, 'gradnorm/max_param': 366.0379943847656, 'epoch': 7.63}

 76%|███████▋  | 14560/19090 [7:06:29<2:11:34,  1.74s/it]
 76%|███████▋  | 14561/19090 [7:06:31<2:21:12,  1.87s/it]
 76%|███████▋  | 14562/19090 [7:06:33<2:19:59,  1.85s/it]
 76%|███████▋  | 14563/19090 [7:06:34<2:07:08,  1.69s/it]
 76%|███████▋  | 14564/19090 [7:06:36<2:20:23,  1.86s/it]
 76%|███████▋  | 14565/19090 [7:06:38<2:24:36,  1.92s/it]
 76%|███████▋  | 14566/19090 [7:06:40<2:10:16,  1.73s/it]
 76%|███████▋  | 14567/19090 [7:06:42<2:18:36,  1.84s/it]
 76%|███████▋  | 14568/19090 [7:06:43<2:18:54,  1.84s/it]
 76%|███████▋  | 14569/19090 [7:06:45<2:06:14,  1.68s/it][gradnorm] step 14569: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 499.2

 76%|███████▋  | 14570/19090 [7:06:47<2:24:35,  1.92s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 715.1260375976562, 'learning_rate': 8.066898665574726e-06, 'recon_loss': 0.0018476018391083925, 'recon_static': 0.0008799788149190136, 'recon_motion': 0.0009676230896729976, 'recon_norm': 1.4169063568115234, 'recon_norm_std': 0.036795821972191334, 'target_norm': 1.6197682976722718, 'recon_cos': 0.5793737739324569, 'gradnorm/compressor.query': 0.4229217541518616, 'gradnorm/compressor.layer0': 16.936649540456635, 'gradnorm/compressor.layer1': 707.404233801297, 'gradnorm/compressor.layer2': 99.2842156072656, 'gradnorm/compressor.layer3': 27.60094222519858, 'gradnorm/compressor.layer4': 6.79853691742049, 'gradnorm/compressor.layer5': 3.167635334498863, 'gradnorm/compressor.layer6': 2.3145166831001407, 'gradnorm/compressor.layer7': 2.5270823324337264, 'gradnorm/decoder.layer0': 2.6281891478195227, 'gradnorm/decoder.layer1': 1.601991698278093, 'gradnorm/decoder.layer2': 0.918624156974876, 'gradnorm/max_param': 499.1597900390625, 'epoch': 7.63}

 76%|███████▋  | 14570/19090 [7:06:47<2:24:35,  1.92s/it]
 76%|███████▋  | 14571/19090 [7:06:49<2:20:12,  1.86s/it]
 76%|███████▋  | 14572/19090 [7:06:50<2:07:08,  1.69s/it]
 76%|███████▋  | 14573/19090 [7:06:53<2:27:53,  1.96s/it]
 76%|███████▋  | 14574/19090 [7:06:54<2:12:26,  1.76s/it]
 76%|███████▋  | 14575/19090 [7:06:55<2:01:41,  1.62s/it]
 76%|███████▋  | 14576/19090 [7:06:58<2:33:43,  2.04s/it]
 76%|███████▋  | 14577/19090 [7:07:00<2:16:30,  1.81s/it]
 76%|███████▋  | 14578/19090 [7:07:01<2:04:30,  1.66s/it]
 76%|███████▋  | 14579/19090 [7:07:04<2:27:47,  1.97s/it][gradnorm] step 14579: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 467.9

 76%|███████▋  | 14580/19090 [7:07:05<2:12:30,  1.76s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 671.8671875, 'learning_rate': 8.033296499816709e-06, 'recon_loss': 0.0016738179489038884, 'recon_static': 0.0006987660526647232, 'recon_motion': 0.0009750518613145687, 'recon_norm': 1.3442955255508422, 'recon_norm_std': 0.03935097837820649, 'target_norm': 1.622206348180771, 'recon_cos': 0.6114871487021446, 'gradnorm/compressor.query': 0.3711632713648763, 'gradnorm/compressor.layer0': 12.943567299946418, 'gradnorm/compressor.layer1': 663.0413825697458, 'gradnorm/compressor.layer2': 103.63625636239713, 'gradnorm/compressor.layer3': 28.36674534046755, 'gradnorm/compressor.layer4': 6.152360491048651, 'gradnorm/compressor.layer5': 3.103489812667988, 'gradnorm/compressor.layer6': 1.9490172257149045, 'gradnorm/compressor.layer7': 2.616035686232014, 'gradnorm/decoder.layer0': 2.6723424491439762, 'gradnorm/decoder.layer1': 1.6625084582809169, 'gradnorm/decoder.layer2': 0.8916952495509589, 'gradnorm/max_param': 467.8976135253906, 'epoch': 7.64}

 76%|███████▋  | 14580/19090 [7:07:05<2:12:30,  1.76s/it]
 76%|███████▋  | 14581/19090 [7:07:06<2:01:41,  1.62s/it]
 76%|███████▋  | 14582/19090 [7:07:09<2:29:20,  1.99s/it]
 76%|███████▋  | 14583/19090 [7:07:10<2:13:27,  1.78s/it]
 76%|███████▋  | 14584/19090 [7:07:12<2:02:25,  1.63s/it]
 76%|███████▋  | 14585/19090 [7:07:15<2:34:57,  2.06s/it]
 76%|███████▋  | 14586/19090 [7:07:16<2:17:25,  1.83s/it]
 76%|███████▋  | 14587/19090 [7:07:17<2:05:21,  1.67s/it]
 76%|███████▋  | 14588/19090 [7:07:20<2:25:31,  1.94s/it]
 76%|███████▋  | 14589/19090 [7:07:21<2:10:49,  1.74s/it][gradnorm] step 14589: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 488.5

 76%|███████▋  | 14590/19090 [7:07:23<2:00:37,  1.61s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 700.6596069335938, 'learning_rate': 7.999751062417479e-06, 'recon_loss': 0.0016953109909081832, 'recon_static': 0.0007248608322697692, 'recon_motion': 0.0009704501295345836, 'recon_norm': 1.3550082325935364, 'recon_norm_std': 0.03897049007937312, 'target_norm': 1.619488748908043, 'recon_cos': 0.6059515193104744, 'gradnorm/compressor.query': 0.3059654821549225, 'gradnorm/compressor.layer0': 10.971324651089738, 'gradnorm/compressor.layer1': 692.4212274692335, 'gradnorm/compressor.layer2': 102.51665261068564, 'gradnorm/compressor.layer3': 27.547512001639436, 'gradnorm/compressor.layer4': 7.25398697240694, 'gradnorm/compressor.layer5': 3.476494331977401, 'gradnorm/compressor.layer6': 2.4703517498104897, 'gradnorm/compressor.layer7': 2.63278678179668, 'gradnorm/decoder.layer0': 2.5584286236076093, 'gradnorm/decoder.layer1': 1.6018242598827594, 'gradnorm/decoder.layer2': 0.9400725354960746, 'gradnorm/max_param': 488.4593811035156, 'epoch': 7.64}

 76%|███████▋  | 14590/19090 [7:07:23<2:00:37,  1.61s/it]
 76%|███████▋  | 14591/19090 [7:07:25<2:19:56,  1.87s/it]
 76%|███████▋  | 14592/19090 [7:07:26<2:06:50,  1.69s/it]
 76%|███████▋  | 14593/19090 [7:07:28<1:57:40,  1.57s/it]
 76%|███████▋  | 14594/19090 [7:07:30<2:26:48,  1.96s/it]
 76%|███████▋  | 14595/19090 [7:07:32<2:11:42,  1.76s/it]
 76%|███████▋  | 14596/19090 [7:07:33<2:01:03,  1.62s/it]
 76%|███████▋  | 14597/19090 [7:07:36<2:30:51,  2.01s/it]
 76%|███████▋  | 14598/19090 [7:07:37<2:14:25,  1.80s/it]
 76%|███████▋  | 14599/19090 [7:07:39<2:03:08,  1.65s/it][gradnorm] step 14599: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 467.7

 76%|███████▋  | 14600/19090 [7:07:42<2:33:05,  2.05s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 670.6172485351562, 'learning_rate': 7.966262465536575e-06, 'recon_loss': 0.0016623276751488447, 'recon_static': 0.0006830246740719304, 'recon_motion': 0.000979303011263255, 'recon_norm': 1.3360977649688721, 'recon_norm_std': 0.03799550514668226, 'target_norm': 1.620740568637848, 'recon_cos': 0.61259925365448, 'gradnorm/compressor.query': 0.2918923503938478, 'gradnorm/compressor.layer0': 12.865865188962104, 'gradnorm/compressor.layer1': 662.7956651563135, 'gradnorm/compressor.layer2': 97.15003799696889, 'gradnorm/compressor.layer3': 27.340374791688575, 'gradnorm/compressor.layer4': 6.306323665192425, 'gradnorm/compressor.layer5': 3.8708151092483547, 'gradnorm/compressor.layer6': 2.4866906656158867, 'gradnorm/compressor.layer7': 2.7499679216768085, 'gradnorm/decoder.layer0': 2.6122238232227164, 'gradnorm/decoder.layer1': 1.4539342288236408, 'gradnorm/decoder.layer2': 0.8365898082837363, 'gradnorm/max_param': 467.73028564453125, 'epoch': 7.65}

 76%|███████▋  | 14600/19090 [7:07:42<2:33:05,  2.05s/it]
 76%|███████▋  | 14601/19090 [7:07:43<2:16:12,  1.82s/it]
 76%|███████▋  | 14602/19090 [7:07:44<2:04:18,  1.66s/it]
 76%|███████▋  | 14603/19090 [7:07:47<2:22:14,  1.90s/it]
 77%|███████▋  | 14604/19090 [7:07:48<2:08:29,  1.72s/it]
 77%|███████▋  | 14605/19090 [7:07:49<1:58:44,  1.59s/it]
 77%|███████▋  | 14606/19090 [7:07:52<2:25:56,  1.95s/it]
 77%|███████▋  | 14607/19090 [7:07:53<2:11:00,  1.75s/it]
 77%|███████▋  | 14608/19090 [7:07:55<2:00:29,  1.61s/it]
 77%|███████▋  | 14609/19090 [7:07:57<2:27:11,  1.97s/it][gradnorm] step 14609: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 562.3

 77%|███████▋  | 14610/19090 [7:07:59<2:11:50,  1.77s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 805.975341796875, 'learning_rate': 7.932830821143489e-06, 'recon_loss': 0.0015897226840024814, 'recon_static': 0.0006248943609534763, 'recon_motion': 0.000964828344876878, 'recon_norm': 1.3094309121370316, 'recon_norm_std': 0.03959100367501378, 'target_norm': 1.6186628550291062, 'recon_cos': 0.6270710989832878, 'gradnorm/compressor.query': 0.3413122938225708, 'gradnorm/compressor.layer0': 13.079627093211066, 'gradnorm/compressor.layer1': 796.9348702999512, 'gradnorm/compressor.layer2': 115.18962372567461, 'gradnorm/compressor.layer3': 30.643488320764718, 'gradnorm/compressor.layer4': 7.468331425002618, 'gradnorm/compressor.layer5': 4.868278564820668, 'gradnorm/compressor.layer6': 3.1480197676700827, 'gradnorm/compressor.layer7': 2.823917694718706, 'gradnorm/decoder.layer0': 3.100309258078545, 'gradnorm/decoder.layer1': 1.8175322949296342, 'gradnorm/decoder.layer2': 1.3576404423012374, 'gradnorm/max_param': 562.3400268554688, 'epoch': 7.65}

 77%|███████▋  | 14610/19090 [7:07:59<2:11:50,  1.77s/it]
 77%|███████▋  | 14611/19090 [7:08:00<2:01:05,  1.62s/it]
 77%|███████▋  | 14612/19090 [7:08:03<2:27:23,  1.97s/it]
 77%|███████▋  | 14613/19090 [7:08:04<2:12:01,  1.77s/it]
 77%|███████▋  | 14614/19090 [7:08:05<2:01:06,  1.62s/it]
 77%|███████▋  | 14615/19090 [7:08:08<2:34:26,  2.07s/it]
 77%|███████▋  | 14616/19090 [7:08:10<2:16:51,  1.84s/it]
 77%|███████▋  | 14617/19090 [7:08:11<2:04:30,  1.67s/it]
 77%|███████▋  | 14618/19090 [7:08:14<2:31:42,  2.04s/it]
 77%|███████▋  | 14619/19090 [7:08:15<2:14:59,  1.81s/it][gradnorm] step 14619: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 495.2

 77%|███████▋  | 14620/19090 [7:08:16<2:03:21,  1.66s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 710.3572998046875, 'learning_rate': 7.899456241017288e-06, 'recon_loss': 0.001580210227984935, 'recon_static': 0.0006191763561218977, 'recon_motion': 0.0009610338107449934, 'recon_norm': 1.3068526148796082, 'recon_norm_std': 0.040579554811120036, 'target_norm': 1.6178478598594666, 'recon_cos': 0.6287049800157547, 'gradnorm/compressor.query': 0.4814801464671934, 'gradnorm/compressor.layer0': 21.250781235639316, 'gradnorm/compressor.layer1': 701.891462407116, 'gradnorm/compressor.layer2': 103.4848550966517, 'gradnorm/compressor.layer3': 26.313660947146946, 'gradnorm/compressor.layer4': 7.669618480598592, 'gradnorm/compressor.layer5': 4.066302823117333, 'gradnorm/compressor.layer6': 2.692045315858464, 'gradnorm/compressor.layer7': 2.7413324397780796, 'gradnorm/decoder.layer0': 2.920843095169044, 'gradnorm/decoder.layer1': 1.665978416751852, 'gradnorm/decoder.layer2': 1.0458131215750182, 'gradnorm/max_param': 495.1711730957031, 'epoch': 7.66}

 77%|███████▋  | 14620/19090 [7:08:16<2:03:21,  1.66s/it]
 77%|███████▋  | 14621/19090 [7:08:19<2:25:49,  1.96s/it]
 77%|███████▋  | 14622/19090 [7:08:20<2:10:51,  1.76s/it]
 77%|███████▋  | 14623/19090 [7:08:22<2:00:24,  1.62s/it]
 77%|███████▋  | 14624/19090 [7:08:24<2:20:29,  1.89s/it]
 77%|███████▋  | 14625/19090 [7:08:25<2:07:05,  1.71s/it]
 77%|███████▋  | 14626/19090 [7:08:27<1:58:04,  1.59s/it]
 77%|███████▋  | 14627/19090 [7:08:29<2:21:11,  1.90s/it]
 77%|███████▋  | 14628/19090 [7:08:31<2:14:46,  1.81s/it]
 77%|███████▋  | 14629/19090 [7:08:32<2:02:55,  1.65s/it][gradnorm] step 14629: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 500.4

 77%|███████▋  | 14630/19090 [7:08:35<2:17:44,  1.85s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 719.2680053710938, 'learning_rate': 7.86613883674625e-06, 'recon_loss': 0.0016929791774600745, 'recon_static': 0.0007181527849752456, 'recon_motion': 0.0009748263721121476, 'recon_norm': 1.3498946249485015, 'recon_norm_std': 0.038868540246039626, 'target_norm': 1.6204717725515365, 'recon_cos': 0.6065436959266662, 'gradnorm/compressor.query': 0.33428683722987745, 'gradnorm/compressor.layer0': 11.35601631054273, 'gradnorm/compressor.layer1': 709.1320178570419, 'gradnorm/compressor.layer2': 115.46770128075859, 'gradnorm/compressor.layer3': 29.874501499502045, 'gradnorm/compressor.layer4': 8.15688230906308, 'gradnorm/compressor.layer5': 5.113933804271206, 'gradnorm/compressor.layer6': 2.8687051607504745, 'gradnorm/compressor.layer7': 2.9425641499983426, 'gradnorm/decoder.layer0': 3.113328423615314, 'gradnorm/decoder.layer1': 1.830160328454953, 'gradnorm/decoder.layer2': 1.123356731625779, 'gradnorm/max_param': 500.43682861328125, 'epoch': 7.66}

 77%|███████▋  | 14630/19090 [7:08:35<2:17:44,  1.85s/it]
 77%|███████▋  | 14631/19090 [7:08:36<2:17:22,  1.85s/it]
 77%|███████▋  | 14632/19090 [7:08:38<2:04:52,  1.68s/it]
 77%|███████▋  | 14633/19090 [7:08:40<2:13:15,  1.79s/it]
 77%|███████▋  | 14634/19090 [7:08:42<2:15:33,  1.83s/it]
 77%|███████▋  | 14635/19090 [7:08:43<2:03:32,  1.66s/it]
 77%|███████▋  | 14636/19090 [7:08:45<2:12:07,  1.78s/it]
 77%|███████▋  | 14637/19090 [7:08:47<2:17:44,  1.86s/it]
 77%|███████▋  | 14638/19090 [7:08:48<2:05:13,  1.69s/it]
 77%|███████▋  | 14639/19090 [7:08:51<2:19:00,  1.87s/it][gradnorm] step 14639: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 369.6

 77%|███████▋  | 14640/19090 [7:08:53<2:24:28,  1.95s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 531.393310546875, 'learning_rate': 7.832878719727474e-06, 'recon_loss': 0.0016995289857732131, 'recon_static': 0.0007290182053111494, 'recon_motion': 0.0009705108139314688, 'recon_norm': 1.3550148665905, 'recon_norm_std': 0.039826774969697, 'target_norm': 1.6199987709522248, 'recon_cos': 0.6051328614354133, 'gradnorm/compressor.query': 0.2670633138978547, 'gradnorm/compressor.layer0': 10.47551779548372, 'gradnorm/compressor.layer1': 523.7294745381436, 'gradnorm/compressor.layer2': 85.49059364376001, 'gradnorm/compressor.layer3': 24.10007592660174, 'gradnorm/compressor.layer4': 7.173951224525784, 'gradnorm/compressor.layer5': 3.6218438714099244, 'gradnorm/compressor.layer6': 2.3798334228499933, 'gradnorm/compressor.layer7': 2.3123656568940087, 'gradnorm/decoder.layer0': 2.8101941298900432, 'gradnorm/decoder.layer1': 1.603470201116101, 'gradnorm/decoder.layer2': 0.9748369459601712, 'gradnorm/max_param': 369.55804443359375, 'epoch': 7.67}

 77%|███████▋  | 14640/19090 [7:08:53<2:24:28,  1.95s/it]
 77%|███████▋  | 14641/19090 [7:08:54<2:09:56,  1.75s/it]
 77%|███████▋  | 14642/19090 [7:08:56<2:11:51,  1.78s/it]
 77%|███████▋  | 14643/19090 [7:08:58<2:25:27,  1.96s/it]
 77%|███████▋  | 14644/19090 [7:09:00<2:10:28,  1.76s/it]
 77%|███████▋  | 14645/19090 [7:09:01<2:12:21,  1.79s/it]
 77%|███████▋  | 14646/19090 [7:09:04<2:17:57,  1.86s/it]
 77%|███████▋  | 14647/19090 [7:09:05<2:05:12,  1.69s/it]
 77%|███████▋  | 14648/19090 [7:09:07<2:13:07,  1.80s/it]
 77%|███████▋  | 14649/19090 [7:09:09<2:21:20,  1.91s/it][gradnorm] step 14649: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 458.2

 77%|███████▋  | 14650/19090 [7:09:10<2:07:41,  1.73s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 656.0147705078125, 'learning_rate': 7.79967600116653e-06, 'recon_loss': 0.0017252402903977782, 'recon_static': 0.0007636102978722193, 'recon_motion': 0.0009616299881599844, 'recon_norm': 1.3737048298120498, 'recon_norm_std': 0.0393640891648829, 'target_norm': 1.6180611193180083, 'recon_cos': 0.6009560987353325, 'gradnorm/compressor.query': 0.27765852567119054, 'gradnorm/compressor.layer0': 10.644133091750701, 'gradnorm/compressor.layer1': 649.2389775729735, 'gradnorm/compressor.layer2': 89.13665772284712, 'gradnorm/compressor.layer3': 25.88279127677971, 'gradnorm/compressor.layer4': 8.447070310242, 'gradnorm/compressor.layer5': 4.226476079653105, 'gradnorm/compressor.layer6': 2.7478464537455176, 'gradnorm/compressor.layer7': 2.4819249001575314, 'gradnorm/decoder.layer0': 2.889454702655629, 'gradnorm/decoder.layer1': 1.8290170302768771, 'gradnorm/decoder.layer2': 0.9374037693226248, 'gradnorm/max_param': 458.1926574707031, 'epoch': 7.67}

 77%|███████▋  | 14650/19090 [7:09:10<2:07:41,  1.73s/it]
 77%|███████▋  | 14651/19090 [7:09:12<2:15:40,  1.83s/it]
 77%|███████▋  | 14652/19090 [7:09:14<2:13:50,  1.81s/it]
 77%|███████▋  | 14653/19090 [7:09:15<2:02:17,  1.65s/it]
 77%|███████▋  | 14654/19090 [7:09:18<2:15:25,  1.83s/it]
 77%|███████▋  | 14655/19090 [7:09:20<2:15:18,  1.83s/it]
 77%|███████▋  | 14656/19090 [7:09:21<2:03:13,  1.67s/it]
 77%|███████▋  | 14657/19090 [7:09:23<2:14:43,  1.82s/it]
 77%|███████▋  | 14658/19090 [7:09:25<2:22:39,  1.93s/it]
 77%|███████▋  | 14659/19090 [7:09:26<2:08:29,  1.74s/it][gradnorm] step 14659: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 575.1

 77%|███████▋  | 14660/19090 [7:09:29<2:18:15,  1.87s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 822.0281372070312, 'learning_rate': 7.766530792077068e-06, 'recon_loss': 0.0017551999888382852, 'recon_static': 0.0007621232332894578, 'recon_motion': 0.0009930766900652088, 'recon_norm': 1.3658049911260606, 'recon_norm_std': 0.03863055408000946, 'target_norm': 1.6265027523040771, 'recon_cos': 0.5958430141210556, 'gradnorm/compressor.query': 0.39438694496020216, 'gradnorm/compressor.layer0': 17.367683022620454, 'gradnorm/compressor.layer1': 815.0239643716005, 'gradnorm/compressor.layer2': 101.50850006742046, 'gradnorm/compressor.layer3': 27.50399473826484, 'gradnorm/compressor.layer4': 7.795937766580772, 'gradnorm/compressor.layer5': 4.101373228519826, 'gradnorm/compressor.layer6': 2.634712054681495, 'gradnorm/compressor.layer7': 2.5467180074972364, 'gradnorm/decoder.layer0': 2.9108728537801207, 'gradnorm/decoder.layer1': 1.8031065650671907, 'gradnorm/decoder.layer2': 1.1160446403558273, 'gradnorm/max_param': 575.1087646484375, 'epoch': 7.68}

 77%|███████▋  | 14660/19090 [7:09:29<2:18:15,  1.87s/it]
 77%|███████▋  | 14661/19090 [7:09:30<2:12:26,  1.79s/it]
 77%|███████▋  | 14662/19090 [7:09:32<2:01:07,  1.64s/it]
 77%|███████▋  | 14663/19090 [7:09:34<2:21:59,  1.92s/it]
 77%|███████▋  | 14664/19090 [7:09:36<2:16:13,  1.85s/it]
 77%|███████▋  | 14665/19090 [7:09:37<2:04:05,  1.68s/it]
 77%|███████▋  | 14666/19090 [7:09:39<2:14:50,  1.83s/it]
 77%|███████▋  | 14667/19090 [7:09:41<2:17:02,  1.86s/it]
 77%|███████▋  | 14668/19090 [7:09:43<2:04:51,  1.69s/it]
 77%|███████▋  | 14669/19090 [7:09:45<2:18:25,  1.88s/it][gradnorm] step 14669: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 549.7

 77%|███████▋  | 14670/19090 [7:09:47<2:22:12,  1.93s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 788.8002319335938, 'learning_rate': 7.73344320328046e-06, 'recon_loss': 0.0017565781221492217, 'recon_static': 0.000777298814500682, 'recon_motion': 0.0009792792829102837, 'recon_norm': 1.3774522125720978, 'recon_norm_std': 0.038645392097532746, 'target_norm': 1.6212893724441528, 'recon_cos': 0.5954195246100425, 'gradnorm/compressor.query': 0.34202409003977174, 'gradnorm/compressor.layer0': 12.683250279265343, 'gradnorm/compressor.layer1': 779.0318510818412, 'gradnorm/compressor.layer2': 119.12779131698237, 'gradnorm/compressor.layer3': 29.241985038038095, 'gradnorm/compressor.layer4': 7.651903312124095, 'gradnorm/compressor.layer5': 4.569494242656073, 'gradnorm/compressor.layer6': 2.688009258191877, 'gradnorm/compressor.layer7': 2.5666950591692648, 'gradnorm/decoder.layer0': 3.0000707300113443, 'gradnorm/decoder.layer1': 1.9726199911581048, 'gradnorm/decoder.layer2': 1.2461010684572755, 'gradnorm/max_param': 549.7227783203125, 'epoch': 7.68}

 77%|███████▋  | 14670/19090 [7:09:47<2:22:12,  1.93s/it]
 77%|███████▋  | 14671/19090 [7:09:48<2:08:05,  1.74s/it]
 77%|███████▋  | 14672/19090 [7:09:50<2:12:17,  1.80s/it]
 77%|███████▋  | 14673/19090 [7:09:52<2:20:13,  1.90s/it]
 77%|███████▋  | 14674/19090 [7:09:54<2:06:34,  1.72s/it]
 77%|███████▋  | 14675/19090 [7:09:56<2:15:15,  1.84s/it]
 77%|███████▋  | 14676/19090 [7:09:58<2:21:38,  1.93s/it]
 77%|███████▋  | 14677/19090 [7:09:59<2:07:43,  1.74s/it]
 77%|███████▋  | 14678/19090 [7:10:01<2:16:45,  1.86s/it]
 77%|███████▋  | 14679/19090 [7:10:03<2:16:48,  1.86s/it][gradnorm] step 14679: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 484.2

 77%|███████▋  | 14680/19090 [7:10:04<2:04:25,  1.69s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 694.0089111328125, 'learning_rate': 7.700413345405418e-06, 'recon_loss': 0.0016785618499852717, 'recon_static': 0.0007058193776174448, 'recon_motion': 0.0009727424680022523, 'recon_norm': 1.3445968270301818, 'recon_norm_std': 0.040099300909787416, 'target_norm': 1.6220465689897536, 'recon_cos': 0.6107877880334854, 'gradnorm/compressor.query': 0.4584334455595561, 'gradnorm/compressor.layer0': 17.151012406342083, 'gradnorm/compressor.layer1': 686.363742850684, 'gradnorm/compressor.layer2': 98.37729805924485, 'gradnorm/compressor.layer3': 22.114941740672965, 'gradnorm/compressor.layer4': 7.270544736831186, 'gradnorm/compressor.layer5': 3.723876368903305, 'gradnorm/compressor.layer6': 2.6186476548618987, 'gradnorm/compressor.layer7': 2.6246782741482133, 'gradnorm/decoder.layer0': 2.6591424788814804, 'gradnorm/decoder.layer1': 1.7577036506012018, 'gradnorm/decoder.layer2': 0.9936179115271957, 'gradnorm/max_param': 484.192138671875, 'epoch': 7.69}

 77%|███████▋  | 14680/19090 [7:10:04<2:04:25,  1.69s/it]
 77%|███████▋  | 14681/19090 [7:10:06<2:09:52,  1.77s/it]
 77%|███████▋  | 14682/19090 [7:10:09<2:18:59,  1.89s/it]
 77%|███████▋  | 14683/19090 [7:10:10<2:05:39,  1.71s/it]
 77%|███████▋  | 14684/19090 [7:10:12<2:14:36,  1.83s/it]
 77%|███████▋  | 14685/19090 [7:10:14<2:23:08,  1.95s/it]
 77%|███████▋  | 14686/19090 [7:10:15<2:08:32,  1.75s/it]
 77%|███████▋  | 14687/19090 [7:10:17<2:06:48,  1.73s/it]
 77%|███████▋  | 14688/19090 [7:10:19<2:17:29,  1.87s/it]
 77%|███████▋  | 14689/19090 [7:10:21<2:04:34,  1.70s/it][gradnorm] step 14689: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 630.8

 77%|███████▋  | 14690/19090 [7:10:23<2:11:07,  1.79s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 903.9933471679688, 'learning_rate': 7.667441328887634e-06, 'recon_loss': 0.0018092008511302992, 'recon_static': 0.0008295878054923378, 'recon_motion': 0.0009796131154871546, 'recon_norm': 1.3988973647356033, 'recon_norm_std': 0.03806900465860963, 'target_norm': 1.6233899533748626, 'recon_cos': 0.586927492916584, 'gradnorm/compressor.query': 0.4263770283617518, 'gradnorm/compressor.layer0': 15.755950288142632, 'gradnorm/compressor.layer1': 893.8027257174818, 'gradnorm/compressor.layer2': 130.75537004985685, 'gradnorm/compressor.layer3': 29.118393832335602, 'gradnorm/compressor.layer4': 9.188651291731293, 'gradnorm/compressor.layer5': 3.7749479126020913, 'gradnorm/compressor.layer6': 2.6351057530124526, 'gradnorm/compressor.layer7': 2.959522241234606, 'gradnorm/decoder.layer0': 2.940157560680791, 'gradnorm/decoder.layer1': 1.7862077616989271, 'gradnorm/decoder.layer2': 1.0815465509428543, 'gradnorm/max_param': 630.7504272460938, 'epoch': 7.7}

 77%|███████▋  | 14690/19090 [7:10:23<2:11:07,  1.79s/it]
 77%|███████▋  | 14691/19090 [7:10:25<2:20:43,  1.92s/it]
 77%|███████▋  | 14692/19090 [7:10:26<2:06:49,  1.73s/it]
 77%|███████▋  | 14693/19090 [7:10:28<2:15:13,  1.85s/it]
 77%|███████▋  | 14694/19090 [7:10:30<2:20:07,  1.91s/it]
 77%|███████▋  | 14695/19090 [7:10:32<2:06:18,  1.72s/it]
 77%|███████▋  | 14696/19090 [7:10:33<2:09:40,  1.77s/it]
 77%|███████▋  | 14697/19090 [7:10:35<2:15:26,  1.85s/it]
 77%|███████▋  | 14698/19090 [7:10:37<2:03:03,  1.68s/it]
 77%|███████▋  | 14699/19090 [7:10:39<2:15:38,  1.85s/it][gradnorm] step 14699: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 477.8

 77%|███████▋  | 14700/19090 [7:10:41<2:18:44,  1.90s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 683.3638305664062, 'learning_rate': 7.63452726396941e-06, 'recon_loss': 0.001774502289481461, 'recon_static': 0.0007879849945311435, 'recon_motion': 0.0009865173051366583, 'recon_norm': 1.388017588853836, 'recon_norm_std': 0.03806714164093137, 'target_norm': 1.6212345659732819, 'recon_cos': 0.5926488831639289, 'gradnorm/compressor.query': 0.3124197737710648, 'gradnorm/compressor.layer0': 11.337100431389247, 'gradnorm/compressor.layer1': 677.0790897302915, 'gradnorm/compressor.layer2': 87.15759455757427, 'gradnorm/compressor.layer3': 27.121644827776624, 'gradnorm/compressor.layer4': 6.9347946803276574, 'gradnorm/compressor.layer5': 3.7299075678195486, 'gradnorm/compressor.layer6': 2.5933725473075437, 'gradnorm/compressor.layer7': 3.0204336143085193, 'gradnorm/decoder.layer0': 2.7144737375717143, 'gradnorm/decoder.layer1': 1.7689028670500428, 'gradnorm/decoder.layer2': 1.016493616593522, 'gradnorm/max_param': 477.79669189453125, 'epoch': 7.7}

 77%|███████▋  | 14700/19090 [7:10:41<2:18:44,  1.90s/it]
 77%|███████▋  | 14701/19090 [7:10:42<2:05:24,  1.71s/it]
 77%|███████▋  | 14702/19090 [7:10:45<2:16:38,  1.87s/it]
 77%|███████▋  | 14703/19090 [7:10:46<2:11:45,  1.80s/it]
 77%|███████▋  | 14704/19090 [7:10:47<2:00:19,  1.65s/it]
 77%|███████▋  | 14705/19090 [7:10:50<2:18:47,  1.90s/it]
 77%|███████▋  | 14706/19090 [7:10:52<2:12:50,  1.82s/it]
 77%|███████▋  | 14707/19090 [7:10:53<2:01:05,  1.66s/it]
 77%|███████▋  | 14708/19090 [7:10:55<2:18:35,  1.90s/it]
 77%|███████▋  | 14709/19090 [7:10:57<2:16:25,  1.87s/it][gradnorm] step 14709: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 521.8

 77%|███████▋  | 14710/19090 [7:10:58<2:03:48,  1.70s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 749.4031982421875, 'learning_rate': 7.601671260699275e-06, 'recon_loss': 0.0016884146112715825, 'recon_static': 0.0007063066470436752, 'recon_motion': 0.000982107968593482, 'recon_norm': 1.3386250585317612, 'recon_norm_std': 0.03900488717481494, 'target_norm': 1.623155128955841, 'recon_cos': 0.6069630950689315, 'gradnorm/compressor.query': 0.32364287421194915, 'gradnorm/compressor.layer0': 11.222653870627141, 'gradnorm/compressor.layer1': 739.5426796338396, 'gradnorm/compressor.layer2': 116.03227679319448, 'gradnorm/compressor.layer3': 31.681031298564523, 'gradnorm/compressor.layer4': 6.443987790002035, 'gradnorm/compressor.layer5': 4.152885279686911, 'gradnorm/compressor.layer6': 2.854524761670537, 'gradnorm/compressor.layer7': 2.917714303194526, 'gradnorm/decoder.layer0': 2.8334554945097823, 'gradnorm/decoder.layer1': 1.9122364909826606, 'gradnorm/decoder.layer2': 1.1956239182446895, 'gradnorm/max_param': 521.8456420898438, 'epoch': 7.71}

 77%|███████▋  | 14710/19090 [7:10:58<2:03:48,  1.70s/it]
 77%|███████▋  | 14711/19090 [7:11:01<2:13:14,  1.83s/it]
 77%|███████▋  | 14712/19090 [7:11:03<2:18:19,  1.90s/it]
 77%|███████▋  | 14713/19090 [7:11:04<2:05:13,  1.72s/it]
 77%|███████▋  | 14714/19090 [7:11:06<2:05:38,  1.72s/it]
 77%|███████▋  | 14715/19090 [7:11:08<2:13:52,  1.84s/it]
 77%|███████▋  | 14716/19090 [7:11:09<2:02:04,  1.67s/it]
 77%|███████▋  | 14717/19090 [7:11:11<2:09:43,  1.78s/it]
 77%|███████▋  | 14718/19090 [7:11:13<2:15:51,  1.86s/it]
 77%|███████▋  | 14719/19090 [7:11:14<2:03:15,  1.69s/it][gradnorm] step 14719: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 459.4

 77%|███████▋  | 14720/19090 [7:11:16<2:10:05,  1.79s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 659.05712890625, 'learning_rate': 7.568873428931642e-06, 'recon_loss': 0.0016102886118460446, 'recon_static': 0.0006355146295391023, 'recon_motion': 0.0009747739692102187, 'recon_norm': 1.3159562796354294, 'recon_norm_std': 0.04031324554234743, 'target_norm': 1.6201298594474793, 'recon_cos': 0.6231836661696434, 'gradnorm/compressor.query': 0.3880129457051825, 'gradnorm/compressor.layer0': 13.572557789568327, 'gradnorm/compressor.layer1': 650.942465775586, 'gradnorm/compressor.layer2': 99.22374495301011, 'gradnorm/compressor.layer3': 22.930519309869815, 'gradnorm/compressor.layer4': 6.668551973157379, 'gradnorm/compressor.layer5': 3.140939269782203, 'gradnorm/compressor.layer6': 2.2590471211821668, 'gradnorm/compressor.layer7': 2.639180708248493, 'gradnorm/decoder.layer0': 2.3224905561207914, 'gradnorm/decoder.layer1': 1.4579406027815447, 'gradnorm/decoder.layer2': 0.8200597101080797, 'gradnorm/max_param': 459.421142578125, 'epoch': 7.71}

 77%|███████▋  | 14720/19090 [7:11:16<2:10:05,  1.79s/it]
 77%|███████▋  | 14721/19090 [7:11:18<2:10:13,  1.79s/it]
 77%|███████▋  | 14722/19090 [7:11:20<1:59:17,  1.64s/it]
 77%|███████▋  | 14723/19090 [7:11:22<2:07:13,  1.75s/it]
 77%|███████▋  | 14724/19090 [7:11:24<2:14:29,  1.85s/it]
 77%|███████▋  | 14725/19090 [7:11:25<2:02:29,  1.68s/it]
 77%|███████▋  | 14726/19090 [7:11:27<2:02:47,  1.69s/it]
 77%|███████▋  | 14727/19090 [7:11:29<2:14:00,  1.84s/it]
 77%|███████▋  | 14728/19090 [7:11:30<2:01:57,  1.68s/it]
 77%|███████▋  | 14729/19090 [7:11:32<2:00:38,  1.66s/it][gradnorm] step 14729: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 571.5

 77%|███████▋  | 14730/19090 [7:11:34<2:18:48,  1.91s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 819.0454711914062, 'learning_rate': 7.536133878326418e-06, 'recon_loss': 0.0016549213381949813, 'recon_static': 0.0006826982673374005, 'recon_motion': 0.0009722230446641333, 'recon_norm': 1.3371916174888612, 'recon_norm_std': 0.03939460422843695, 'target_norm': 1.6188188076019288, 'recon_cos': 0.613978810608387, 'gradnorm/compressor.query': 0.4817718883880885, 'gradnorm/compressor.layer0': 18.487537207059837, 'gradnorm/compressor.layer1': 809.84589583451, 'gradnorm/compressor.layer2': 116.66749627681224, 'gradnorm/compressor.layer3': 30.463128487823194, 'gradnorm/compressor.layer4': 7.658575958072519, 'gradnorm/compressor.layer5': 4.530457315497943, 'gradnorm/compressor.layer6': 2.3817730747137, 'gradnorm/compressor.layer7': 2.8318382881460518, 'gradnorm/decoder.layer0': 2.6946828797840605, 'gradnorm/decoder.layer1': 1.6794578217788851, 'gradnorm/decoder.layer2': 0.937467510932, 'gradnorm/max_param': 571.519287109375, 'epoch': 7.72}

 77%|███████▋  | 14730/19090 [7:11:34<2:18:48,  1.91s/it]
 77%|███████▋  | 14731/19090 [7:11:36<2:05:27,  1.73s/it]
 77%|███████▋  | 14732/19090 [7:11:37<1:56:37,  1.61s/it]
 77%|███████▋  | 14733/19090 [7:11:40<2:21:46,  1.95s/it]
 77%|███████▋  | 14734/19090 [7:11:41<2:07:45,  1.76s/it]
 77%|███████▋  | 14735/19090 [7:11:42<2:01:43,  1.68s/it]
 77%|███████▋  | 14736/19090 [7:11:45<2:24:27,  1.99s/it]
 77%|███████▋  | 14737/19090 [7:11:46<2:09:18,  1.78s/it]
 77%|███████▋  | 14738/19090 [7:11:48<2:04:57,  1.72s/it]
 77%|███████▋  | 14739/19090 [7:11:50<2:20:02,  1.93s/it][gradnorm] step 14739: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 570.8

 77%|███████▋  | 14740/19090 [7:11:52<2:06:13,  1.74s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 818.3707885742188, 'learning_rate': 7.503452718348647e-06, 'recon_loss': 0.0016133866825839504, 'recon_static': 0.000642845885886345, 'recon_motion': 0.0009705408112495207, 'recon_norm': 1.3198911279439927, 'recon_norm_std': 0.03942675702273846, 'target_norm': 1.6176254242658614, 'recon_cos': 0.621488943696022, 'gradnorm/compressor.query': 0.38703341157361176, 'gradnorm/compressor.layer0': 13.780680326767422, 'gradnorm/compressor.layer1': 808.8220215226586, 'gradnorm/compressor.layer2': 118.09812855735692, 'gradnorm/compressor.layer3': 35.715583507391244, 'gradnorm/compressor.layer4': 8.68460276203105, 'gradnorm/compressor.layer5': 3.949005151362282, 'gradnorm/compressor.layer6': 3.0552398884640812, 'gradnorm/compressor.layer7': 3.3074206870236735, 'gradnorm/decoder.layer0': 2.812500678168321, 'gradnorm/decoder.layer1': 2.0652101224164685, 'gradnorm/decoder.layer2': 1.2001976466330793, 'gradnorm/max_param': 570.7560424804688, 'epoch': 7.72}

 77%|███████▋  | 14740/19090 [7:11:52<2:06:13,  1.74s/it]
 77%|███████▋  | 14741/19090 [7:11:54<2:09:07,  1.78s/it]
 77%|███████▋  | 14742/19090 [7:11:56<2:18:17,  1.91s/it]
 77%|███████▋  | 14743/19090 [7:11:57<2:04:55,  1.72s/it]
 77%|███████▋  | 14744/19090 [7:11:59<2:04:53,  1.72s/it]
 77%|███████▋  | 14745/19090 [7:12:01<2:15:53,  1.88s/it]
 77%|███████▋  | 14746/19090 [7:12:02<2:03:19,  1.70s/it]
 77%|███████▋  | 14747/19090 [7:12:04<2:02:28,  1.69s/it]
 77%|███████▋  | 14748/19090 [7:12:07<2:25:18,  2.01s/it]
 77%|███████▋  | 14749/19090 [7:12:08<2:10:12,  1.80s/it][gradnorm] step 14749: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 619.7

 77%|███████▋  | 14750/19090 [7:12:09<1:59:59,  1.66s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 885.8479614257812, 'learning_rate': 7.470830058268147e-06, 'recon_loss': 0.0015645812323782592, 'recon_static': 0.0005935299894190393, 'recon_motion': 0.0009710513011668808, 'recon_norm': 1.2975374430418014, 'recon_norm_std': 0.04077411061152816, 'target_norm': 1.620324859023094, 'recon_cos': 0.6329346016049385, 'gradnorm/compressor.query': 0.38043633465864957, 'gradnorm/compressor.layer0': 14.832114065972377, 'gradnorm/compressor.layer1': 878.1859498420594, 'gradnorm/compressor.layer2': 109.12427065191547, 'gradnorm/compressor.layer3': 35.38090201703544, 'gradnorm/compressor.layer4': 9.015700315164183, 'gradnorm/compressor.layer5': 4.352380973507467, 'gradnorm/compressor.layer6': 2.953381239658222, 'gradnorm/compressor.layer7': 3.2882126822542657, 'gradnorm/decoder.layer0': 3.3091805578353486, 'gradnorm/decoder.layer1': 1.9906922357483772, 'gradnorm/decoder.layer2': 1.1400112434301228, 'gradnorm/max_param': 619.6630249023438, 'epoch': 7.73}

 77%|███████▋  | 14750/19090 [7:12:09<1:59:59,  1.66s/it]
 77%|███████▋  | 14751/19090 [7:12:12<2:25:03,  2.01s/it]
 77%|███████▋  | 14752/19090 [7:12:14<2:09:46,  1.79s/it]
 77%|███████▋  | 14753/19090 [7:12:15<2:01:42,  1.68s/it]
 77%|███████▋  | 14754/19090 [7:12:17<2:20:21,  1.94s/it]
 77%|███████▋  | 14755/19090 [7:12:19<2:06:24,  1.75s/it]
 77%|███████▋  | 14756/19090 [7:12:20<2:04:28,  1.72s/it]
 77%|███████▋  | 14757/19090 [7:12:23<2:12:50,  1.84s/it]
 77%|███████▋  | 14758/19090 [7:12:24<2:01:12,  1.68s/it]
 77%|███████▋  | 14759/19090 [7:12:26<2:13:23,  1.85s/it][gradnorm] step 14759: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 623.3

 77%|███████▋  | 14760/19090 [7:12:28<2:05:32,  1.74s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 893.7464599609375, 'learning_rate': 7.438266007159137e-06, 'recon_loss': 0.0016263056866591796, 'recon_static': 0.0006530027894768864, 'recon_motion': 0.0009733029117342084, 'recon_norm': 1.3288673251867293, 'recon_norm_std': 0.03923390116542578, 'target_norm': 1.6174940854310988, 'recon_cos': 0.6193737894296646, 'gradnorm/compressor.query': 0.3054195315620008, 'gradnorm/compressor.layer0': 12.337012970654037, 'gradnorm/compressor.layer1': 883.3581167906932, 'gradnorm/compressor.layer2': 130.91741343543646, 'gradnorm/compressor.layer3': 32.27927034406621, 'gradnorm/compressor.layer4': 8.543804167971345, 'gradnorm/compressor.layer5': 4.580394820140947, 'gradnorm/compressor.layer6': 2.9726694034079886, 'gradnorm/compressor.layer7': 3.309900091225051, 'gradnorm/decoder.layer0': 3.0423378051555536, 'gradnorm/decoder.layer1': 1.82018197700892, 'gradnorm/decoder.layer2': 1.0859565458754434, 'gradnorm/max_param': 623.3392944335938, 'epoch': 7.73}

 77%|███████▋  | 14760/19090 [7:12:28<2:05:32,  1.74s/it]
 77%|███████▋  | 14761/19090 [7:12:29<1:56:04,  1.61s/it]
 77%|███████▋  | 14762/19090 [7:12:31<2:14:34,  1.87s/it]
 77%|███████▋  | 14763/19090 [7:12:33<2:12:49,  1.84s/it]
 77%|███████▋  | 14764/19090 [7:12:34<2:01:07,  1.68s/it]
 77%|███████▋  | 14765/19090 [7:12:37<2:19:36,  1.94s/it]
 77%|███████▋  | 14766/19090 [7:12:39<2:11:11,  1.82s/it]
 77%|███████▋  | 14767/19090 [7:12:40<1:59:53,  1.66s/it]
 77%|███████▋  | 14768/19090 [7:12:42<2:16:24,  1.89s/it]
 77%|███████▋  | 14769/19090 [7:12:44<2:04:30,  1.73s/it][gradnorm] step 14769: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 564.3

 77%|███████▋  | 14770/19090 [7:12:45<1:55:04,  1.60s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 810.4948120117188, 'learning_rate': 7.4057606738998805e-06, 'recon_loss': 0.0016134581790538504, 'recon_static': 0.0006253298779483885, 'recon_motion': 0.0009881283272989093, 'recon_norm': 1.3134815782308578, 'recon_norm_std': 0.03990709688514471, 'target_norm': 1.6224753141403199, 'recon_cos': 0.623192435503006, 'gradnorm/compressor.query': 0.4097121604754392, 'gradnorm/compressor.layer0': 19.304064639891095, 'gradnorm/compressor.layer1': 799.6208867332068, 'gradnorm/compressor.layer2': 126.76525425959592, 'gradnorm/compressor.layer3': 30.943716823288113, 'gradnorm/compressor.layer4': 7.991478911876988, 'gradnorm/compressor.layer5': 4.100992333881958, 'gradnorm/compressor.layer6': 2.522543545092859, 'gradnorm/compressor.layer7': 3.0395345836039303, 'gradnorm/decoder.layer0': 2.903040107988746, 'gradnorm/decoder.layer1': 1.7295991594176672, 'gradnorm/decoder.layer2': 0.9224839543030018, 'gradnorm/max_param': 564.2539672851562, 'epoch': 7.74}

 77%|███████▋  | 14770/19090 [7:12:45<1:55:04,  1.60s/it]
 77%|███████▋  | 14771/19090 [7:12:48<2:19:58,  1.94s/it]
 77%|███████▋  | 14772/19090 [7:12:49<2:08:22,  1.78s/it]
 77%|███████▋  | 14773/19090 [7:12:50<1:57:39,  1.64s/it]
 77%|███████▋  | 14774/19090 [7:12:53<2:25:02,  2.02s/it]
 77%|███████▋  | 14775/19090 [7:12:55<2:09:24,  1.80s/it]
 77%|███████▋  | 14776/19090 [7:12:56<1:58:29,  1.65s/it]
 77%|███████▋  | 14777/19090 [7:12:59<2:24:26,  2.01s/it]
 77%|███████▋  | 14778/19090 [7:13:00<2:09:02,  1.80s/it]
 77%|███████▋  | 14779/19090 [7:13:01<1:57:59,  1.64s/it][gradnorm] step 14779: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 444.7

 77%|███████▋  | 14780/19090 [7:13:04<2:27:48,  2.06s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 641.5874633789062, 'learning_rate': 7.373314167172312e-06, 'recon_loss': 0.0016270583582809195, 'recon_static': 0.0006492239466751925, 'recon_motion': 0.0009778344567166642, 'recon_norm': 1.3204534024000167, 'recon_norm_std': 0.04084356324747205, 'target_norm': 1.6204860508441925, 'recon_cos': 0.6190510347485543, 'gradnorm/compressor.query': 0.4008797797729045, 'gradnorm/compressor.layer0': 15.275116296003945, 'gradnorm/compressor.layer1': 630.1843976170784, 'gradnorm/compressor.layer2': 114.86801614380089, 'gradnorm/compressor.layer3': 30.926561552695382, 'gradnorm/compressor.layer4': 8.233344170978945, 'gradnorm/compressor.layer5': 4.651157743650847, 'gradnorm/compressor.layer6': 2.6799291384993613, 'gradnorm/compressor.layer7': 2.853770448976522, 'gradnorm/decoder.layer0': 2.9641651286615502, 'gradnorm/decoder.layer1': 1.7188280261275382, 'gradnorm/decoder.layer2': 0.9950211677983339, 'gradnorm/max_param': 444.6678466796875, 'epoch': 7.74}

 77%|███████▋  | 14780/19090 [7:13:04<2:27:48,  2.06s/it]
 77%|███████▋  | 14781/19090 [7:13:06<2:11:18,  1.83s/it]
 77%|███████▋  | 14782/19090 [7:13:07<1:59:48,  1.67s/it]
 77%|███████▋  | 14783/19090 [7:13:09<2:19:18,  1.94s/it]
 77%|███████▋  | 14784/19090 [7:13:11<2:05:14,  1.75s/it]
 77%|███████▋  | 14785/19090 [7:13:12<1:55:35,  1.61s/it]
 77%|███████▋  | 14786/19090 [7:13:15<2:16:04,  1.90s/it]
 77%|███████▋  | 14787/19090 [7:13:16<2:06:56,  1.77s/it]
 77%|███████▋  | 14788/19090 [7:13:17<1:56:34,  1.63s/it]
 77%|███████▋  | 14789/19090 [7:13:20<2:13:05,  1.86s/it][gradnorm] step 14789: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 507.9

 77%|███████▋  | 14790/19090 [7:13:21<2:07:52,  1.78s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 727.068115234375, 'learning_rate': 7.340926595461687e-06, 'recon_loss': 0.0015919747529551387, 'recon_static': 0.0006054764322470874, 'recon_motion': 0.0009864982697763481, 'recon_norm': 1.3051881045103073, 'recon_norm_std': 0.04058844745159149, 'target_norm': 1.6222416162490845, 'recon_cos': 0.6275699302554131, 'gradnorm/compressor.query': 0.3066343683467057, 'gradnorm/compressor.layer0': 10.567214545678276, 'gradnorm/compressor.layer1': 719.7331102568508, 'gradnorm/compressor.layer2': 97.46710683129206, 'gradnorm/compressor.layer3': 29.742875985267464, 'gradnorm/compressor.layer4': 8.290836874177344, 'gradnorm/compressor.layer5': 4.514718251778768, 'gradnorm/compressor.layer6': 2.55460079874259, 'gradnorm/compressor.layer7': 2.687250879850961, 'gradnorm/decoder.layer0': 3.0510452293061756, 'gradnorm/decoder.layer1': 1.7482683606192122, 'gradnorm/decoder.layer2': 0.9004629110252571, 'gradnorm/max_param': 507.87957763671875, 'epoch': 7.75}

 77%|███████▋  | 14790/19090 [7:13:21<2:07:52,  1.78s/it]
 77%|███████▋  | 14791/19090 [7:13:23<1:57:18,  1.64s/it]
 77%|███████▋  | 14792/19090 [7:13:25<2:18:29,  1.93s/it]
 77%|███████▋  | 14793/19090 [7:13:27<2:04:41,  1.74s/it]
 77%|███████▋  | 14794/19090 [7:13:29<2:16:37,  1.91s/it]
 78%|███████▊  | 14795/19090 [7:13:31<2:15:36,  1.89s/it]
 78%|███████▊  | 14796/19090 [7:13:32<2:02:34,  1.71s/it]
 78%|███████▊  | 14797/19090 [7:13:33<1:53:38,  1.59s/it]
 78%|███████▊  | 14798/19090 [7:13:36<2:14:54,  1.89s/it]
 78%|███████▊  | 14799/19090 [7:13:37<2:02:03,  1.71s/it][gradnorm] step 14799: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 648.2

 78%|███████▊  | 14800/19090 [7:13:39<1:53:08,  1.58s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 930.30322265625, 'learning_rate': 7.3085980670561994e-06, 'recon_loss': 0.0016035862819990144, 'recon_static': 0.0006240791350137443, 'recon_motion': 0.0009795071178814395, 'recon_norm': 1.3120883494615554, 'recon_norm_std': 0.039994518272578716, 'target_norm': 1.6207703918218612, 'recon_cos': 0.624375693500042, 'gradnorm/compressor.query': 0.3880564931403681, 'gradnorm/compressor.layer0': 15.81645882848139, 'gradnorm/compressor.layer1': 918.5184061846556, 'gradnorm/compressor.layer2': 140.9845126689453, 'gradnorm/compressor.layer3': 39.338503050645244, 'gradnorm/compressor.layer4': 7.795662520015638, 'gradnorm/compressor.layer5': 4.758129992485559, 'gradnorm/compressor.layer6': 2.9514250529947095, 'gradnorm/compressor.layer7': 3.0864745626410204, 'gradnorm/decoder.layer0': 2.816607548772329, 'gradnorm/decoder.layer1': 1.6909649914276423, 'gradnorm/decoder.layer2': 1.0731366620238256, 'gradnorm/max_param': 648.1911010742188, 'epoch': 7.75}

 78%|███████▊  | 14800/19090 [7:13:39<1:53:08,  1.58s/it]
 78%|███████▊  | 14801/19090 [7:13:41<2:20:16,  1.96s/it]
 78%|███████▊  | 14802/19090 [7:13:43<2:05:53,  1.76s/it]
 78%|███████▊  | 14803/19090 [7:13:44<1:55:53,  1.62s/it]
 78%|███████▊  | 14804/19090 [7:13:47<2:18:26,  1.94s/it]
 78%|███████▊  | 14805/19090 [7:13:48<2:04:34,  1.74s/it]
 78%|███████▊  | 14806/19090 [7:13:49<1:54:49,  1.61s/it]
 78%|███████▊  | 14807/19090 [7:13:52<2:23:16,  2.01s/it]
 78%|███████▊  | 14808/19090 [7:13:53<2:07:54,  1.79s/it]
 78%|███████▊  | 14809/19090 [7:13:55<1:57:03,  1.64s/it][gradnorm] step 14809: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 602.9

 78%|███████▊  | 14810/19090 [7:13:58<2:23:43,  2.01s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 864.3182373046875, 'learning_rate': 7.276328690046641e-06, 'recon_loss': 0.0016195724834688007, 'recon_static': 0.0006544132178532891, 'recon_motion': 0.000965159249608405, 'recon_norm': 1.3266197979450225, 'recon_norm_std': 0.04019304057583213, 'target_norm': 1.6170298010110855, 'recon_cos': 0.6205150797963143, 'gradnorm/compressor.query': 0.37971261337805506, 'gradnorm/compressor.layer0': 14.312364602593735, 'gradnorm/compressor.layer1': 854.462916105784, 'gradnorm/compressor.layer2': 124.24218725437025, 'gradnorm/compressor.layer3': 34.41539884607434, 'gradnorm/compressor.layer4': 8.259175602226772, 'gradnorm/compressor.layer5': 4.298079997941078, 'gradnorm/compressor.layer6': 2.6347219182172203, 'gradnorm/compressor.layer7': 2.740380107242722, 'gradnorm/decoder.layer0': 2.902106500018766, 'gradnorm/decoder.layer1': 1.669267635570524, 'gradnorm/decoder.layer2': 1.1131386113392165, 'gradnorm/max_param': 602.9010009765625, 'epoch': 7.76}

 78%|███████▊  | 14810/19090 [7:13:58<2:23:43,  2.01s/it]
 78%|███████▊  | 14811/19090 [7:13:59<2:08:04,  1.80s/it]
 78%|███████▊  | 14812/19090 [7:14:00<1:57:15,  1.64s/it]
 78%|███████▊  | 14813/19090 [7:14:03<2:15:14,  1.90s/it]
 78%|███████▊  | 14814/19090 [7:14:04<2:02:12,  1.71s/it]
 78%|███████▊  | 14815/19090 [7:14:05<1:53:05,  1.59s/it]
 78%|███████▊  | 14816/19090 [7:14:08<2:12:36,  1.86s/it]
 78%|███████▊  | 14817/19090 [7:14:09<2:00:16,  1.69s/it]
 78%|███████▊  | 14818/19090 [7:14:10<1:51:44,  1.57s/it]
 78%|███████▊  | 14819/19090 [7:14:13<2:17:30,  1.93s/it][gradnorm] step 14819: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 710.2

 78%|███████▊  | 14820/19090 [7:14:14<2:03:46,  1.74s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1016.3700561523438, 'learning_rate': 7.244118572326028e-06, 'recon_loss': 0.0016294732369715348, 'recon_static': 0.0006570769735844805, 'recon_motion': 0.000972396245924756, 'recon_norm': 1.3260041326284409, 'recon_norm_std': 0.039315538760274646, 'target_norm': 1.6188476890325547, 'recon_cos': 0.618511950969696, 'gradnorm/compressor.query': 0.4810461745939615, 'gradnorm/compressor.layer0': 20.22683785178548, 'gradnorm/compressor.layer1': 1006.3750543410741, 'gradnorm/compressor.layer2': 135.15775649449424, 'gradnorm/compressor.layer3': 37.81709521406539, 'gradnorm/compressor.layer4': 7.699138731905976, 'gradnorm/compressor.layer5': 4.802685923911768, 'gradnorm/compressor.layer6': 2.8352934471187305, 'gradnorm/compressor.layer7': 2.6604592575053556, 'gradnorm/decoder.layer0': 2.959447641901071, 'gradnorm/decoder.layer1': 1.6913209677817955, 'gradnorm/decoder.layer2': 1.0588759718246423, 'gradnorm/max_param': 710.203125, 'epoch': 7.76}

 78%|███████▊  | 14820/19090 [7:14:14<2:03:46,  1.74s/it]
 78%|███████▊  | 14821/19090 [7:14:16<1:54:16,  1.61s/it]
 78%|███████▊  | 14822/19090 [7:14:19<2:20:27,  1.97s/it]
 78%|███████▊  | 14823/19090 [7:14:20<2:05:54,  1.77s/it]
 78%|███████▊  | 14824/19090 [7:14:21<1:55:36,  1.63s/it]
 78%|███████▊  | 14825/19090 [7:14:24<2:15:09,  1.90s/it]
 78%|███████▊  | 14826/19090 [7:14:25<2:01:59,  1.72s/it]
 78%|███████▊  | 14827/19090 [7:14:26<1:52:46,  1.59s/it]
 78%|███████▊  | 14828/19090 [7:14:29<2:22:51,  2.01s/it]
 78%|███████▊  | 14829/19090 [7:14:31<2:07:25,  1.79s/it][gradnorm] step 14829: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 615.5

 78%|███████▊  | 14830/19090 [7:14:32<1:56:31,  1.64s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 883.9085693359375, 'learning_rate': 7.211967821589241e-06, 'recon_loss': 0.0015959229320287704, 'recon_static': 0.0006190517393406481, 'recon_motion': 0.0009768712799996138, 'recon_norm': 1.3102180272340775, 'recon_norm_std': 0.040258510317653415, 'target_norm': 1.619403812289238, 'recon_cos': 0.6260672226548195, 'gradnorm/compressor.query': 0.39067334829576794, 'gradnorm/compressor.layer0': 15.210210201859987, 'gradnorm/compressor.layer1': 872.0361231050007, 'gradnorm/compressor.layer2': 139.61058424909982, 'gradnorm/compressor.layer3': 31.647248535121502, 'gradnorm/compressor.layer4': 8.535079376084015, 'gradnorm/compressor.layer5': 4.542747065743405, 'gradnorm/compressor.layer6': 2.7702194241262545, 'gradnorm/compressor.layer7': 2.7331907487136275, 'gradnorm/decoder.layer0': 3.2685026056968756, 'gradnorm/decoder.layer1': 1.6720020566772515, 'gradnorm/decoder.layer2': 1.0351924098194984, 'gradnorm/max_param': 615.477783203125, 'epoch': 7.77}

 78%|███████▊  | 14830/19090 [7:14:32<1:56:31,  1.64s/it]
 78%|███████▊  | 14831/19090 [7:14:35<2:21:28,  1.99s/it]
 78%|███████▊  | 14832/19090 [7:14:36<2:06:24,  1.78s/it]
 78%|███████▊  | 14833/19090 [7:14:37<1:55:54,  1.63s/it]
 78%|███████▊  | 14834/19090 [7:14:40<2:21:16,  1.99s/it]
 78%|███████▊  | 14835/19090 [7:14:41<2:06:22,  1.78s/it]
 78%|███████▊  | 14836/19090 [7:14:43<1:55:55,  1.64s/it]
 78%|███████▊  | 14837/19090 [7:14:46<2:25:57,  2.06s/it]
 78%|███████▊  | 14838/19090 [7:14:47<2:09:36,  1.83s/it]
 78%|███████▊  | 14839/19090 [7:14:48<1:58:11,  1.67s/it][gradnorm] step 14839: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 446.4

 78%|███████▊  | 14840/19090 [7:14:51<2:23:20,  2.02s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 641.5732421875, 'learning_rate': 7.179876545332664e-06, 'recon_loss': 0.0016014235152397305, 'recon_static': 0.0006249036945519037, 'recon_motion': 0.0009765198497916572, 'recon_norm': 1.315196967124939, 'recon_norm_std': 0.04059810321778059, 'target_norm': 1.622458964586258, 'recon_cos': 0.6263816833496094, 'gradnorm/compressor.query': 0.30379021398015343, 'gradnorm/compressor.layer0': 12.16511051903885, 'gradnorm/compressor.layer1': 632.5195649147938, 'gradnorm/compressor.layer2': 102.50126714155782, 'gradnorm/compressor.layer3': 28.370286726526267, 'gradnorm/compressor.layer4': 6.406219035748339, 'gradnorm/compressor.layer5': 3.688529048805808, 'gradnorm/compressor.layer6': 2.207762362483888, 'gradnorm/compressor.layer7': 2.349368773561484, 'gradnorm/decoder.layer0': 2.6341427122541647, 'gradnorm/decoder.layer1': 1.6712135361179412, 'gradnorm/decoder.layer2': 0.9984190185460166, 'gradnorm/max_param': 446.39874267578125, 'epoch': 7.77}

 78%|███████▊  | 14840/19090 [7:14:51<2:23:20,  2.02s/it]
 78%|███████▊  | 14841/19090 [7:14:52<2:07:42,  1.80s/it]
 78%|███████▊  | 14842/19090 [7:14:54<1:56:47,  1.65s/it]
 78%|███████▊  | 14843/19090 [7:14:57<2:23:44,  2.03s/it]
 78%|███████▊  | 14844/19090 [7:14:58<2:07:54,  1.81s/it]
 78%|███████▊  | 14845/19090 [7:14:59<1:56:51,  1.65s/it]
 78%|███████▊  | 14846/19090 [7:15:02<2:20:48,  1.99s/it]
 78%|███████▊  | 14847/19090 [7:15:03<2:05:48,  1.78s/it]
 78%|███████▊  | 14848/19090 [7:15:05<1:55:21,  1.63s/it]
 78%|███████▊  | 14849/19090 [7:15:07<2:10:18,  1.84s/it][gradnorm] step 14849: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 677.1

 78%|███████▊  | 14850/19090 [7:15:08<1:58:32,  1.68s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 969.5281982421875, 'learning_rate': 7.147844850853832e-06, 'recon_loss': 0.0016599521768512204, 'recon_static': 0.0006835152045823634, 'recon_motion': 0.0009764369824551977, 'recon_norm': 1.3362843722105027, 'recon_norm_std': 0.0384769550524652, 'target_norm': 1.6204972952604293, 'recon_cos': 0.612619186937809, 'gradnorm/compressor.query': 0.45599654029173703, 'gradnorm/compressor.layer0': 18.059624994214044, 'gradnorm/compressor.layer1': 959.6587284029672, 'gradnorm/compressor.layer2': 132.12556173012095, 'gradnorm/compressor.layer3': 33.601194669284084, 'gradnorm/compressor.layer4': 9.01402842435382, 'gradnorm/compressor.layer5': 4.381698929319785, 'gradnorm/compressor.layer6': 2.519909639555001, 'gradnorm/compressor.layer7': 2.6536250888860926, 'gradnorm/decoder.layer0': 3.0774098451425784, 'gradnorm/decoder.layer1': 1.7240245687278561, 'gradnorm/decoder.layer2': 1.0372982104671513, 'gradnorm/max_param': 677.1053466796875, 'epoch': 7.78}

 78%|███████▊  | 14850/19090 [7:15:08<1:58:32,  1.68s/it]
 78%|███████▊  | 14851/19090 [7:15:09<1:50:11,  1.56s/it]
 78%|███████▊  | 14852/19090 [7:15:12<2:10:31,  1.85s/it]
 78%|███████▊  | 14853/19090 [7:15:13<1:58:36,  1.68s/it]
 78%|███████▊  | 14854/19090 [7:15:15<1:50:19,  1.56s/it]
 78%|███████▊  | 14855/19090 [7:15:17<2:12:38,  1.88s/it]
 78%|███████▊  | 14856/19090 [7:15:18<2:00:04,  1.70s/it]
 78%|███████▊  | 14857/19090 [7:15:20<1:51:32,  1.58s/it]
 78%|███████▊  | 14858/19090 [7:15:23<2:18:00,  1.96s/it]
 78%|███████▊  | 14859/19090 [7:15:24<2:03:56,  1.76s/it][gradnorm] step 14859: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 551.4

 78%|███████▊  | 14860/19090 [7:15:25<1:54:05,  1.62s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 793.270263671875, 'learning_rate': 7.115872845251068e-06, 'recon_loss': 0.0016420148604083806, 'recon_static': 0.0006631072756135836, 'recon_motion': 0.00097890758770518, 'recon_norm': 1.326625829935074, 'recon_norm_std': 0.038703143037855624, 'target_norm': 1.6208520203828811, 'recon_cos': 0.6166252315044403, 'gradnorm/compressor.query': 0.503016521772164, 'gradnorm/compressor.layer0': 14.247072153302454, 'gradnorm/compressor.layer1': 781.4624111241692, 'gradnorm/compressor.layer2': 132.7646237237164, 'gradnorm/compressor.layer3': 25.954882870554794, 'gradnorm/compressor.layer4': 7.016055087561615, 'gradnorm/compressor.layer5': 3.949960761539309, 'gradnorm/compressor.layer6': 2.319509245169705, 'gradnorm/compressor.layer7': 2.63665482055022, 'gradnorm/decoder.layer0': 3.1719210015329544, 'gradnorm/decoder.layer1': 1.6383203213479949, 'gradnorm/decoder.layer2': 0.9357402498133214, 'gradnorm/max_param': 551.4349365234375, 'epoch': 7.78}

 78%|███████▊  | 14860/19090 [7:15:25<1:54:05,  1.62s/it]
 78%|███████▊  | 14861/19090 [7:15:28<2:20:04,  1.99s/it]
 78%|███████▊  | 14862/19090 [7:15:29<2:05:19,  1.78s/it]
 78%|███████▊  | 14863/19090 [7:15:31<1:54:54,  1.63s/it]
 78%|███████▊  | 14864/19090 [7:15:33<2:17:13,  1.95s/it]
 78%|███████▊  | 14865/19090 [7:15:35<2:03:12,  1.75s/it]
 78%|███████▊  | 14866/19090 [7:15:36<1:53:27,  1.61s/it]
 78%|███████▊  | 14867/19090 [7:15:38<2:11:09,  1.86s/it]
 78%|███████▊  | 14868/19090 [7:15:40<1:59:11,  1.69s/it]
 78%|███████▊  | 14869/19090 [7:15:41<1:50:40,  1.57s/it][gradnorm] step 14869: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 675.2

 78%|███████▊  | 14870/19090 [7:15:44<2:17:41,  1.96s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 967.4334106445312, 'learning_rate': 7.083960635423117e-06, 'recon_loss': 0.0017315452656475826, 'recon_static': 0.0007709239798714407, 'recon_motion': 0.0009606213352526538, 'recon_norm': 1.3738635778427124, 'recon_norm_std': 0.03851979095488787, 'target_norm': 1.6178737610578537, 'recon_cos': 0.5992743462324143, 'gradnorm/compressor.query': 0.4539545785253499, 'gradnorm/compressor.layer0': 16.179011295321498, 'gradnorm/compressor.layer1': 956.7686501971101, 'gradnorm/compressor.layer2': 138.29366116619374, 'gradnorm/compressor.layer3': 32.31203910454662, 'gradnorm/compressor.layer4': 7.08842449852842, 'gradnorm/compressor.layer5': 3.6028156872312618, 'gradnorm/compressor.layer6': 2.4862927406592004, 'gradnorm/compressor.layer7': 2.883728874139082, 'gradnorm/decoder.layer0': 2.9131299692343573, 'gradnorm/decoder.layer1': 1.6134975925613189, 'gradnorm/decoder.layer2': 1.0405393731340666, 'gradnorm/max_param': 675.2348022460938, 'epoch': 7.79}

 78%|███████▊  | 14870/19090 [7:15:44<2:17:41,  1.96s/it]
 78%|███████▊  | 14871/19090 [7:15:45<2:03:36,  1.76s/it]
 78%|███████▊  | 14872/19090 [7:15:46<1:53:46,  1.62s/it]
 78%|███████▊  | 14873/19090 [7:15:49<2:16:27,  1.94s/it]
 78%|███████▊  | 14874/19090 [7:15:50<2:06:27,  1.80s/it]
 78%|███████▊  | 14875/19090 [7:15:52<1:55:44,  1.65s/it]
 78%|███████▊  | 14876/19090 [7:15:54<2:16:11,  1.94s/it]
 78%|███████▊  | 14877/19090 [7:15:56<2:02:32,  1.75s/it]
 78%|███████▊  | 14878/19090 [7:15:57<1:52:59,  1.61s/it]
 78%|███████▊  | 14879/19090 [7:15:59<2:10:18,  1.86s/it][gradnorm] step 14879: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 597.7

 78%|███████▊  | 14880/19090 [7:16:01<2:00:20,  1.71s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 856.7918090820312, 'learning_rate': 7.052108328068807e-06, 'recon_loss': 0.0016322626557666807, 'recon_static': 0.0006634951787418686, 'recon_motion': 0.0009687674537417479, 'recon_norm': 1.3312182933092118, 'recon_norm_std': 0.03929083170369267, 'target_norm': 1.621531143784523, 'recon_cos': 0.6203742235898971, 'gradnorm/compressor.query': 0.48228472717708515, 'gradnorm/compressor.layer0': 22.12719026058494, 'gradnorm/compressor.layer1': 846.9388260081125, 'gradnorm/compressor.layer2': 122.1687645482674, 'gradnorm/compressor.layer3': 35.64706519246623, 'gradnorm/compressor.layer4': 7.468411872903122, 'gradnorm/compressor.layer5': 3.9025689442890306, 'gradnorm/compressor.layer6': 2.7271385130947223, 'gradnorm/compressor.layer7': 3.125808001010249, 'gradnorm/decoder.layer0': 3.0136036794070393, 'gradnorm/decoder.layer1': 1.7110938431885476, 'gradnorm/decoder.layer2': 1.011291472594835, 'gradnorm/max_param': 597.6843872070312, 'epoch': 7.79}

 78%|███████▊  | 14880/19090 [7:16:01<2:00:20,  1.71s/it]
 78%|███████▊  | 14881/19090 [7:16:02<1:51:22,  1.59s/it]
 78%|███████▊  | 14882/19090 [7:16:05<2:19:19,  1.99s/it]
 78%|███████▊  | 14883/19090 [7:16:06<2:04:51,  1.78s/it]
 78%|███████▊  | 14884/19090 [7:16:08<1:54:24,  1.63s/it]
 78%|███████▊  | 14885/19090 [7:16:10<2:18:49,  1.98s/it]
 78%|███████▊  | 14886/19090 [7:16:12<2:04:16,  1.77s/it]
 78%|███████▊  | 14887/19090 [7:16:13<1:54:12,  1.63s/it]
 78%|███████▊  | 14888/19090 [7:16:16<2:18:03,  1.97s/it]
 78%|███████▊  | 14889/19090 [7:16:17<2:03:39,  1.77s/it][gradnorm] step 14889: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 435.3

 78%|███████▊  | 14890/19090 [7:16:18<1:53:38,  1.62s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 627.3024291992188, 'learning_rate': 7.020316029686677e-06, 'recon_loss': 0.0017479919421020894, 'recon_static': 0.0007863213133532554, 'recon_motion': 0.0009616706403903664, 'recon_norm': 1.3772126525640487, 'recon_norm_std': 0.03954124534502625, 'target_norm': 1.6206510663032532, 'recon_cos': 0.5975718930363655, 'gradnorm/compressor.query': 0.46755951069147916, 'gradnorm/compressor.layer0': 16.942612668164184, 'gradnorm/compressor.layer1': 616.9632029140798, 'gradnorm/compressor.layer2': 108.19902180848263, 'gradnorm/compressor.layer3': 28.066348349366162, 'gradnorm/compressor.layer4': 6.955556105890276, 'gradnorm/compressor.layer5': 3.461868195438395, 'gradnorm/compressor.layer6': 2.449352692709086, 'gradnorm/compressor.layer7': 2.589262515936599, 'gradnorm/decoder.layer0': 2.505591338786262, 'gradnorm/decoder.layer1': 1.5276823717248242, 'gradnorm/decoder.layer2': 1.0011793573154244, 'gradnorm/max_param': 435.2980041503906, 'epoch': 7.8}

 78%|███████▊  | 14890/19090 [7:16:18<1:53:38,  1.62s/it]
 78%|███████▊  | 14891/19090 [7:16:21<2:11:46,  1.88s/it]
 78%|███████▊  | 14892/19090 [7:16:22<1:59:20,  1.71s/it]
 78%|███████▊  | 14893/19090 [7:16:23<1:50:29,  1.58s/it]
 78%|███████▊  | 14894/19090 [7:16:26<2:13:19,  1.91s/it]
 78%|███████▊  | 14895/19090 [7:16:27<2:00:18,  1.72s/it]
 78%|███████▊  | 14896/19090 [7:16:29<1:51:16,  1.59s/it]
 78%|███████▊  | 14897/19090 [7:16:32<2:18:53,  1.99s/it]
 78%|███████▊  | 14898/19090 [7:16:33<2:04:06,  1.78s/it]
 78%|███████▊  | 14899/19090 [7:16:34<1:53:56,  1.63s/it][gradnorm] step 14899: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 514.1

 78%|███████▊  | 14900/19090 [7:16:37<2:16:27,  1.95s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 735.7387084960938, 'learning_rate': 6.98858384657462e-06, 'recon_loss': 0.0018093465332640335, 'recon_static': 0.0008252730724052526, 'recon_motion': 0.0009840734608587808, 'recon_norm': 1.3963196873664856, 'recon_norm_std': 0.038744577392935756, 'target_norm': 1.6256284713745117, 'recon_cos': 0.5882979646325112, 'gradnorm/compressor.query': 0.35261339370885825, 'gradnorm/compressor.layer0': 14.1106407974911, 'gradnorm/compressor.layer1': 728.5329007670141, 'gradnorm/compressor.layer2': 96.86865402601866, 'gradnorm/compressor.layer3': 29.43565255946541, 'gradnorm/compressor.layer4': 7.699806597373916, 'gradnorm/compressor.layer5': 3.612870938485328, 'gradnorm/compressor.layer6': 2.828369888746799, 'gradnorm/compressor.layer7': 2.894039632255648, 'gradnorm/decoder.layer0': 3.0267164798732193, 'gradnorm/decoder.layer1': 1.7819428016824832, 'gradnorm/decoder.layer2': 0.9820018196399868, 'gradnorm/max_param': 514.1046142578125, 'epoch': 7.81}

 78%|███████▊  | 14900/19090 [7:16:37<2:16:27,  1.95s/it]
 78%|███████▊  | 14901/19090 [7:16:38<2:02:33,  1.76s/it]
 78%|███████▊  | 14902/19090 [7:16:39<1:52:58,  1.62s/it]
 78%|███████▊  | 14903/19090 [7:16:42<2:17:13,  1.97s/it]
 78%|███████▊  | 14904/19090 [7:16:43<2:02:58,  1.76s/it]
 78%|███████▊  | 14905/19090 [7:16:45<1:53:12,  1.62s/it]
 78%|███████▊  | 14906/19090 [7:16:48<2:20:37,  2.02s/it]
 78%|███████▊  | 14907/19090 [7:16:49<2:05:21,  1.80s/it]
 78%|███████▊  | 14908/19090 [7:16:50<1:54:49,  1.65s/it]
 78%|███████▊  | 14909/19090 [7:16:53<2:19:58,  2.01s/it][gradnorm] step 14909: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 468.8

 78%|███████▊  | 14910/19090 [7:16:54<2:04:54,  1.79s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 672.9557495117188, 'learning_rate': 6.95691188482954e-06, 'recon_loss': 0.0016227204032475129, 'recon_static': 0.0006473166024079547, 'recon_motion': 0.0009754038357641548, 'recon_norm': 1.3150742799043655, 'recon_norm_std': 0.03938859757035971, 'target_norm': 1.619510719180107, 'recon_cos': 0.6189839884638786, 'gradnorm/compressor.query': 0.34861635655479634, 'gradnorm/compressor.layer0': 15.321198402324748, 'gradnorm/compressor.layer1': 664.2596442657043, 'gradnorm/compressor.layer2': 103.55174959289678, 'gradnorm/compressor.layer3': 24.22345216449598, 'gradnorm/compressor.layer4': 7.005341535433955, 'gradnorm/compressor.layer5': 3.353354661796597, 'gradnorm/compressor.layer6': 2.609526943876819, 'gradnorm/compressor.layer7': 2.667570874775072, 'gradnorm/decoder.layer0': 2.4963793285260336, 'gradnorm/decoder.layer1': 1.6762118553242138, 'gradnorm/decoder.layer2': 0.8588545523774295, 'gradnorm/max_param': 468.7705078125, 'epoch': 7.81}

 78%|███████▊  | 14910/19090 [7:16:54<2:04:54,  1.79s/it]
 78%|███████▊  | 14911/19090 [7:16:56<1:54:32,  1.64s/it]
 78%|███████▊  | 14912/19090 [7:16:58<2:16:03,  1.95s/it]
 78%|███████▊  | 14913/19090 [7:17:00<2:02:10,  1.75s/it]
 78%|███████▊  | 14914/19090 [7:17:01<1:52:35,  1.62s/it]
 78%|███████▊  | 14915/19090 [7:17:04<2:12:23,  1.90s/it]
 78%|███████▊  | 14916/19090 [7:17:05<1:59:37,  1.72s/it]
 78%|███████▊  | 14917/19090 [7:17:06<1:50:42,  1.59s/it]
 78%|███████▊  | 14918/19090 [7:17:09<2:20:54,  2.03s/it]
 78%|███████▊  | 14919/19090 [7:17:10<2:05:35,  1.81s/it][gradnorm] step 14919: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 492.8

 78%|███████▊  | 14920/19090 [7:17:12<1:54:57,  1.65s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 706.9249267578125, 'learning_rate': 6.925300250346986e-06, 'recon_loss': 0.0016255495342193172, 'recon_static': 0.0006615140664507635, 'recon_motion': 0.0009640354794100859, 'recon_norm': 1.3341488540172577, 'recon_norm_std': 0.03964727846905589, 'target_norm': 1.6178437799215317, 'recon_cos': 0.6206309959292412, 'gradnorm/compressor.query': 0.3224669665599576, 'gradnorm/compressor.layer0': 12.602384650412521, 'gradnorm/compressor.layer1': 698.354807744602, 'gradnorm/compressor.layer2': 105.04539569276704, 'gradnorm/compressor.layer3': 27.330893182769728, 'gradnorm/compressor.layer4': 7.686381436978428, 'gradnorm/compressor.layer5': 4.10088257013096, 'gradnorm/compressor.layer6': 2.7860219072958, 'gradnorm/compressor.layer7': 2.9463510478336272, 'gradnorm/decoder.layer0': 2.6943312475890226, 'gradnorm/decoder.layer1': 1.592661167774525, 'gradnorm/decoder.layer2': 0.945361585367295, 'gradnorm/max_param': 492.8035583496094, 'epoch': 7.82}

 78%|███████▊  | 14920/19090 [7:17:12<1:54:57,  1.65s/it]
 78%|███████▊  | 14921/19090 [7:17:15<2:21:08,  2.03s/it]
 78%|███████▊  | 14922/19090 [7:17:16<2:05:47,  1.81s/it]
 78%|███████▊  | 14923/19090 [7:17:17<1:55:00,  1.66s/it]
 78%|███████▊  | 14924/19090 [7:17:20<2:12:04,  1.90s/it]
 78%|███████▊  | 14925/19090 [7:17:21<1:59:04,  1.72s/it]
 78%|███████▊  | 14926/19090 [7:17:22<1:50:03,  1.59s/it]
 78%|███████▊  | 14927/19090 [7:17:25<2:15:56,  1.96s/it]
 78%|███████▊  | 14928/19090 [7:17:26<2:01:56,  1.76s/it]
 78%|███████▊  | 14929/19090 [7:17:28<1:52:08,  1.62s/it][gradnorm] step 14929: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 699.1

 78%|███████▊  | 14930/19090 [7:17:30<2:15:29,  1.95s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 1000.9105834960938, 'learning_rate': 6.893749048820805e-06, 'recon_loss': 0.0016523189464351163, 'recon_static': 0.0006824267606134526, 'recon_motion': 0.0009698920912342146, 'recon_norm': 1.3350291818380355, 'recon_norm_std': 0.03866030285134912, 'target_norm': 1.6191420257091522, 'recon_cos': 0.6142198458313942, 'gradnorm/compressor.query': 0.535182061234227, 'gradnorm/compressor.layer0': 18.06901543658802, 'gradnorm/compressor.layer1': 990.7659600026637, 'gradnorm/compressor.layer2': 137.11936662311783, 'gradnorm/compressor.layer3': 30.91329446859949, 'gradnorm/compressor.layer4': 8.635163441907572, 'gradnorm/compressor.layer5': 4.210275912995466, 'gradnorm/compressor.layer6': 2.7334055026536506, 'gradnorm/compressor.layer7': 3.1602701701789595, 'gradnorm/decoder.layer0': 2.6617468113339564, 'gradnorm/decoder.layer1': 1.7186187694041806, 'gradnorm/decoder.layer2': 0.9026493034075688, 'gradnorm/max_param': 699.116455078125, 'epoch': 7.82}

 78%|███████▊  | 14930/19090 [7:17:30<2:15:29,  1.95s/it]
 78%|███████▊  | 14931/19090 [7:17:32<2:01:30,  1.75s/it]
 78%|███████▊  | 14932/19090 [7:17:33<1:51:45,  1.61s/it]
 78%|███████▊  | 14933/19090 [7:17:36<2:20:33,  2.03s/it]
 78%|███████▊  | 14934/19090 [7:17:37<2:05:02,  1.81s/it]
 78%|███████▊  | 14935/19090 [7:17:39<1:54:15,  1.65s/it]
 78%|███████▊  | 14936/19090 [7:17:42<2:33:39,  2.22s/it]
 78%|███████▊  | 14937/19090 [7:17:43<2:14:18,  1.94s/it]
 78%|███████▊  | 14938/19090 [7:17:45<2:00:46,  1.75s/it]
 78%|███████▊  | 14939/19090 [7:17:46<2:00:22,  1.74s/it][gradnorm] step 14939: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 525.2

 78%|███████▊  | 14940/19090 [7:17:48<1:50:58,  1.60s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 752.4239501953125, 'learning_rate': 6.862258385742784e-06, 'recon_loss': 0.0016075644554803148, 'recon_static': 0.0006184183948789723, 'recon_motion': 0.000989146064966917, 'recon_norm': 1.3055971622467042, 'recon_norm_std': 0.0393105786293745, 'target_norm': 1.6234479755163194, 'recon_cos': 0.6240265145897865, 'gradnorm/compressor.query': 0.4212107904450398, 'gradnorm/compressor.layer0': 15.148603197424105, 'gradnorm/compressor.layer1': 744.2169542546044, 'gradnorm/compressor.layer2': 105.34320768920036, 'gradnorm/compressor.layer3': 29.073253848777725, 'gradnorm/compressor.layer4': 8.11713627495266, 'gradnorm/compressor.layer5': 4.315847231193204, 'gradnorm/compressor.layer6': 2.8150796715690825, 'gradnorm/compressor.layer7': 2.7856163297624827, 'gradnorm/decoder.layer0': 2.7286111300225717, 'gradnorm/decoder.layer1': 1.5955219516134662, 'gradnorm/decoder.layer2': 0.8741524201593803, 'gradnorm/max_param': 525.2058715820312, 'epoch': 7.83}

 78%|███████▊  | 14940/19090 [7:17:48<1:50:58,  1.60s/it]
 78%|███████▊  | 14941/19090 [7:17:49<1:44:26,  1.51s/it]
 78%|███████▊  | 14942/19090 [7:17:52<2:15:52,  1.97s/it]
 78%|███████▊  | 14943/19090 [7:17:54<2:21:59,  2.05s/it]
 78%|███████▊  | 14944/19090 [7:17:56<2:06:25,  1.83s/it]
 78%|███████▊  | 14945/19090 [7:17:57<2:00:59,  1.75s/it]
 78%|███████▊  | 14946/19090 [7:17:59<1:51:24,  1.61s/it]
 78%|███████▊  | 14947/19090 [7:18:00<1:44:35,  1.51s/it]
 78%|███████▊  | 14948/19090 [7:18:03<2:15:43,  1.97s/it]
 78%|███████▊  | 14949/19090 [7:18:04<2:01:37,  1.76s/it][gradnorm] step 14949: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 461.5

 78%|███████▊  | 14950/19090 [7:18:05<1:51:48,  1.62s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 661.7230224609375, 'learning_rate': 6.8308283664023005e-06, 'recon_loss': 0.0015729457954876125, 'recon_static': 0.0005896113521885127, 'recon_motion': 0.0009833344141952694, 'recon_norm': 1.3051577657461166, 'recon_norm_std': 0.038729694206267595, 'target_norm': 1.620028156042099, 'recon_cos': 0.6318436816334725, 'gradnorm/compressor.query': 0.29169538333850176, 'gradnorm/compressor.layer0': 10.381260832724067, 'gradnorm/compressor.layer1': 654.0129968127545, 'gradnorm/compressor.layer2': 94.35759876826296, 'gradnorm/compressor.layer3': 32.32393716833548, 'gradnorm/compressor.layer4': 6.85235973839487, 'gradnorm/compressor.layer5': 3.9372149167475197, 'gradnorm/compressor.layer6': 2.620669289343638, 'gradnorm/compressor.layer7': 2.8988296778142315, 'gradnorm/decoder.layer0': 2.6458219593033903, 'gradnorm/decoder.layer1': 1.6544053223428392, 'gradnorm/decoder.layer2': 0.9643830346139297, 'gradnorm/max_param': 461.52557373046875, 'epoch': 7.83}

 78%|███████▊  | 14950/19090 [7:18:05<1:51:48,  1.62s/it]
 78%|███████▊  | 14951/19090 [7:18:08<2:21:39,  2.05s/it]
 78%|███████▊  | 14952/19090 [7:18:10<2:05:44,  1.82s/it]
 78%|███████▊  | 14953/19090 [7:18:11<1:54:45,  1.66s/it]
 78%|███████▊  | 14954/19090 [7:18:14<2:17:25,  1.99s/it]
 78%|███████▊  | 14955/19090 [7:18:15<2:02:46,  1.78s/it]
 78%|███████▊  | 14956/19090 [7:18:16<1:52:34,  1.63s/it]
 78%|███████▊  | 14957/19090 [7:18:19<2:16:08,  1.98s/it]
 78%|███████▊  | 14958/19090 [7:18:20<2:01:56,  1.77s/it]
 78%|███████▊  | 14959/19090 [7:18:22<1:52:01,  1.63s/it][gradnorm] step 14959: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 741.3

 78%|███████▊  | 14960/19090 [7:18:25<2:15:59,  1.98s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1060.03076171875, 'learning_rate': 6.799459095885968e-06, 'recon_loss': 0.0015873826865572481, 'recon_static': 0.0006182222496136092, 'recon_motion': 0.0009691604194813408, 'recon_norm': 1.3073540896177291, 'recon_norm_std': 0.03968976717442274, 'target_norm': 1.6200096428394317, 'recon_cos': 0.6279443070292473, 'gradnorm/compressor.query': 0.3965071497798086, 'gradnorm/compressor.layer0': 13.956007366061488, 'gradnorm/compressor.layer1': 1050.3680307397021, 'gradnorm/compressor.layer2': 137.96427300210732, 'gradnorm/compressor.layer3': 32.57573807202993, 'gradnorm/compressor.layer4': 7.8654686301097305, 'gradnorm/compressor.layer5': 3.889606388448022, 'gradnorm/compressor.layer6': 2.7643779809648077, 'gradnorm/compressor.layer7': 2.7469273654429394, 'gradnorm/decoder.layer0': 2.464004204350142, 'gradnorm/decoder.layer1': 1.670387890134046, 'gradnorm/decoder.layer2': 0.9181458322694779, 'gradnorm/max_param': 741.284423828125, 'epoch': 7.84}

 78%|███████▊  | 14960/19090 [7:18:25<2:15:59,  1.98s/it]
 78%|███████▊  | 14961/19090 [7:18:26<2:01:54,  1.77s/it]
 78%|███████▊  | 14962/19090 [7:18:27<1:51:53,  1.63s/it]
 78%|███████▊  | 14963/19090 [7:18:30<2:14:34,  1.96s/it]
 78%|███████▊  | 14964/19090 [7:18:31<2:00:51,  1.76s/it]
 78%|███████▊  | 14965/19090 [7:18:32<1:51:09,  1.62s/it]
 78%|███████▊  | 14966/19090 [7:18:35<2:12:56,  1.93s/it]
 78%|███████▊  | 14967/19090 [7:18:36<1:59:32,  1.74s/it]
 78%|███████▊  | 14968/19090 [7:18:38<1:50:16,  1.61s/it]
 78%|███████▊  | 14969/19090 [7:18:40<2:08:37,  1.87s/it][gradnorm] step 14969: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 657.4

 78%|███████▊  | 14970/19090 [7:18:41<1:56:33,  1.70s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 943.4664306640625, 'learning_rate': 6.768150679077284e-06, 'recon_loss': 0.0015755902393721044, 'recon_static': 0.0006003027185215615, 'recon_motion': 0.0009752875092090107, 'recon_norm': 1.3000544488430024, 'recon_norm_std': 0.03964487267658114, 'target_norm': 1.6208815693855285, 'recon_cos': 0.630723275244236, 'gradnorm/compressor.query': 0.3998556517379633, 'gradnorm/compressor.layer0': 16.477185747162178, 'gradnorm/compressor.layer1': 931.6615533550797, 'gradnorm/compressor.layer2': 144.12848272027983, 'gradnorm/compressor.layer3': 31.628741016688938, 'gradnorm/compressor.layer4': 6.445571135909246, 'gradnorm/compressor.layer5': 4.334639621349468, 'gradnorm/compressor.layer6': 3.041459498905729, 'gradnorm/compressor.layer7': 2.956142760704814, 'gradnorm/decoder.layer0': 2.973531947437532, 'gradnorm/decoder.layer1': 1.6369157549666482, 'gradnorm/decoder.layer2': 0.9115973920555522, 'gradnorm/max_param': 657.4310302734375, 'epoch': 7.84}

 78%|███████▊  | 14970/19090 [7:18:41<1:56:33,  1.70s/it]
 78%|███████▊  | 14971/19090 [7:18:43<1:48:13,  1.58s/it]
 78%|███████▊  | 14972/19090 [7:18:46<2:20:59,  2.05s/it]
 78%|███████▊  | 14973/19090 [7:18:47<2:05:15,  1.83s/it]
 78%|███████▊  | 14974/19090 [7:18:48<1:54:15,  1.67s/it]
 78%|███████▊  | 14975/19090 [7:18:51<2:10:49,  1.91s/it]
 78%|███████▊  | 14976/19090 [7:18:52<1:58:10,  1.72s/it]
 78%|███████▊  | 14977/19090 [7:18:54<1:49:11,  1.59s/it]
 78%|███████▊  | 14978/19090 [7:18:56<2:07:08,  1.86s/it]
 78%|███████▊  | 14979/19090 [7:18:57<1:55:38,  1.69s/it][gradnorm] step 14979: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 698.9

 78%|███████▊  | 14980/19090 [7:18:59<1:47:41,  1.57s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 997.8331909179688, 'learning_rate': 6.736903220656285e-06, 'recon_loss': 0.0015674715366913005, 'recon_static': 0.0005932094762101769, 'recon_motion': 0.0009742620779434219, 'recon_norm': 1.293869322538376, 'recon_norm_std': 0.04014457361772657, 'target_norm': 1.6184372752904892, 'recon_cos': 0.6304293587803841, 'gradnorm/compressor.query': 0.5930186586146368, 'gradnorm/compressor.layer0': 18.094096600469154, 'gradnorm/compressor.layer1': 990.367135207949, 'gradnorm/compressor.layer2': 115.27022853251181, 'gradnorm/compressor.layer3': 33.57824860377436, 'gradnorm/compressor.layer4': 6.9876370930060965, 'gradnorm/compressor.layer5': 4.597752079900832, 'gradnorm/compressor.layer6': 3.127211589245853, 'gradnorm/compressor.layer7': 2.9979201100541504, 'gradnorm/decoder.layer0': 2.952659933618118, 'gradnorm/decoder.layer1': 1.7941531588903308, 'gradnorm/decoder.layer2': 0.9880269141739872, 'gradnorm/max_param': 698.8623657226562, 'epoch': 7.85}

 78%|███████▊  | 14980/19090 [7:18:59<1:47:41,  1.57s/it]
 78%|███████▊  | 14981/19090 [7:19:01<2:12:29,  1.93s/it]
 78%|███████▊  | 14982/19090 [7:19:03<1:59:30,  1.75s/it]
 78%|███████▊  | 14983/19090 [7:19:04<1:50:02,  1.61s/it]
 78%|███████▊  | 14984/19090 [7:19:07<2:14:00,  1.96s/it]
 78%|███████▊  | 14985/19090 [7:19:08<2:00:19,  1.76s/it]
 79%|███████▊  | 14986/19090 [7:19:09<1:50:39,  1.62s/it]
 79%|███████▊  | 14987/19090 [7:19:12<2:10:46,  1.91s/it]
 79%|███████▊  | 14988/19090 [7:19:13<1:57:52,  1.72s/it]
 79%|███████▊  | 14989/19090 [7:19:15<1:48:52,  1.59s/it][gradnorm] step 14989: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 702.2

 79%|███████▊  | 14990/19090 [7:19:18<2:17:51,  2.02s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1003.8023681640625, 'learning_rate': 6.705716825099187e-06, 'recon_loss': 0.0015958691423293204, 'recon_static': 0.0006191303822561167, 'recon_motion': 0.0009767387658939698, 'recon_norm': 1.3103515356779099, 'recon_norm_std': 0.03847946198657155, 'target_norm': 1.620176723599434, 'recon_cos': 0.6261748299002647, 'gradnorm/compressor.query': 0.4690574750180005, 'gradnorm/compressor.layer0': 15.308725319988323, 'gradnorm/compressor.layer1': 994.9883165143196, 'gradnorm/compressor.layer2': 127.43409755144029, 'gradnorm/compressor.layer3': 32.42159826689128, 'gradnorm/compressor.layer4': 6.008746765275516, 'gradnorm/compressor.layer5': 5.0690619244536075, 'gradnorm/compressor.layer6': 2.858632475952157, 'gradnorm/compressor.layer7': 3.015567126731752, 'gradnorm/decoder.layer0': 3.0350619857466206, 'gradnorm/decoder.layer1': 1.7085484935207178, 'gradnorm/decoder.layer2': 1.009686760248346, 'gradnorm/max_param': 702.1796264648438, 'epoch': 7.85}

 79%|███████▊  | 14990/19090 [7:19:18<2:17:51,  2.02s/it]
 79%|███████▊  | 14991/19090 [7:19:19<2:02:54,  1.80s/it]
 79%|███████▊  | 14992/19090 [7:19:20<1:52:21,  1.65s/it]
 79%|███████▊  | 14993/19090 [7:19:23<2:11:31,  1.93s/it]
 79%|███████▊  | 14994/19090 [7:19:24<1:58:30,  1.74s/it]
 79%|███████▊  | 14995/19090 [7:19:25<1:49:22,  1.60s/it]
 79%|███████▊  | 14996/19090 [7:19:28<2:05:08,  1.83s/it]
 79%|███████▊  | 14997/19090 [7:19:29<1:54:05,  1.67s/it]
 79%|███████▊  | 14998/19090 [7:19:30<1:46:13,  1.56s/it]
 79%|███████▊  | 14999/19090 [7:19:33<2:06:17,  1.85s/it][gradnorm] step 14999: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 514.7

 79%|███████▊  | 15000/19090 [7:19:34<1:54:50,  1.68s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 737.8106689453125, 'learning_rate': 6.674591596678048e-06, 'recon_loss': 0.0015972024732036517, 'recon_static': 0.000609406948206015, 'recon_motion': 0.000987795536639169, 'recon_norm': 1.3080953240394593, 'recon_norm_std': 0.03935676496475935, 'target_norm': 1.6225958913564682, 'recon_cos': 0.6269050806760788, 'gradnorm/compressor.query': 0.3483690925982864, 'gradnorm/compressor.layer0': 10.670064345275723, 'gradnorm/compressor.layer1': 729.5227035534946, 'gradnorm/compressor.layer2': 106.49616245286728, 'gradnorm/compressor.layer3': 24.78019829499075, 'gradnorm/compressor.layer4': 6.817554243856071, 'gradnorm/compressor.layer5': 4.403350092249322, 'gradnorm/compressor.layer6': 2.6412941350106394, 'gradnorm/compressor.layer7': 2.6291972618519397, 'gradnorm/decoder.layer0': 2.8331596283865537, 'gradnorm/decoder.layer1': 1.6434107064041252, 'gradnorm/decoder.layer2': 1.0617810228960938, 'gradnorm/max_param': 514.7286987304688, 'epoch': 7.86}

 79%|███████▊  | 15000/19090 [7:19:34<1:54:50,  1.68s/it]
 79%|███████▊  | 15001/19090 [7:19:38<2:40:31,  2.36s/it]
 79%|███████▊  | 15002/19090 [7:19:39<2:18:41,  2.04s/it]
 79%|███████▊  | 15003/19090 [7:19:41<2:03:24,  1.81s/it]
 79%|███████▊  | 15004/19090 [7:19:42<1:52:42,  1.66s/it]
 79%|███████▊  | 15005/19090 [7:19:43<1:45:15,  1.55s/it]
 79%|███████▊  | 15006/19090 [7:19:44<1:39:57,  1.47s/it]
 79%|███████▊  | 15007/19090 [7:19:46<1:36:12,  1.41s/it]
 79%|███████▊  | 15008/19090 [7:19:47<1:33:42,  1.38s/it]
 79%|███████▊  | 15009/19090 [7:19:48<1:31:50,  1.35s/it][gradnorm] step 15009: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 603.1

 79%|███████▊  | 15010/19090 [7:19:50<1:30:34,  1.33s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 865.6954956054688, 'learning_rate': 6.643527639460409e-06, 'recon_loss': 0.0015297824080334975, 'recon_static': 0.0005580711949733086, 'recon_motion': 0.0009717111825011671, 'recon_norm': 1.2804517656564713, 'recon_norm_std': 0.04029282713308931, 'target_norm': 1.6189832597970963, 'recon_cos': 0.639705953001976, 'gradnorm/compressor.query': 0.3194504978578479, 'gradnorm/compressor.layer0': 12.13007681046309, 'gradnorm/compressor.layer1': 854.6721959324523, 'gradnorm/compressor.layer2': 132.41937405455442, 'gradnorm/compressor.layer3': 34.29841788955523, 'gradnorm/compressor.layer4': 7.500619989836413, 'gradnorm/compressor.layer5': 4.365535416772799, 'gradnorm/compressor.layer6': 3.021788312424042, 'gradnorm/compressor.layer7': 2.7927152932144477, 'gradnorm/decoder.layer0': 3.0611837244238926, 'gradnorm/decoder.layer1': 1.7101514569629683, 'gradnorm/decoder.layer2': 0.9704002352991786, 'gradnorm/max_param': 603.1439819335938, 'epoch': 7.86}

 79%|███████▊  | 15010/19090 [7:19:50<1:30:34,  1.33s/it]
 79%|███████▊  | 15011/19090 [7:19:52<1:56:42,  1.72s/it]
 79%|███████▊  | 15012/19090 [7:19:53<1:47:52,  1.59s/it]
 79%|███████▊  | 15013/19090 [7:19:55<1:41:43,  1.50s/it]
 79%|███████▊  | 15014/19090 [7:19:58<2:07:35,  1.88s/it]
 79%|███████▊  | 15015/19090 [7:19:59<1:55:30,  1.70s/it]
 79%|███████▊  | 15016/19090 [7:20:00<1:47:00,  1.58s/it]
 79%|███████▊  | 15017/19090 [7:20:03<2:13:59,  1.97s/it]
 79%|███████▊  | 15018/19090 [7:20:04<1:59:56,  1.77s/it]
 79%|███████▊  | 15019/19090 [7:20:06<1:50:11,  1.62s/it][gradnorm] step 15019: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 705.0

 79%|███████▊  | 15020/19090 [7:20:08<2:10:07,  1.92s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1009.0343017578125, 'learning_rate': 6.612525057308949e-06, 'recon_loss': 0.0016038231755374, 'recon_static': 0.0006202059666975401, 'recon_motion': 0.0009836171651841142, 'recon_norm': 1.3092450380325318, 'recon_norm_std': 0.03937775939702988, 'target_norm': 1.6222120076417923, 'recon_cos': 0.6249359548091888, 'gradnorm/compressor.query': 0.6379597277800979, 'gradnorm/compressor.layer0': 20.331193321470337, 'gradnorm/compressor.layer1': 999.2364272283112, 'gradnorm/compressor.layer2': 136.65594628939496, 'gradnorm/compressor.layer3': 21.671832050921754, 'gradnorm/compressor.layer4': 7.862772334344819, 'gradnorm/compressor.layer5': 4.882883983851742, 'gradnorm/compressor.layer6': 3.1333412237947282, 'gradnorm/compressor.layer7': 2.8049226200085875, 'gradnorm/decoder.layer0': 3.213732169186043, 'gradnorm/decoder.layer1': 1.9003660351376739, 'gradnorm/decoder.layer2': 1.0827156408896565, 'gradnorm/max_param': 705.0137939453125, 'epoch': 7.87}

 79%|███████▊  | 15020/19090 [7:20:08<2:10:07,  1.92s/it]
 79%|███████▊  | 15021/19090 [7:20:09<1:57:17,  1.73s/it]
 79%|███████▊  | 15022/19090 [7:20:11<1:48:16,  1.60s/it]
 79%|███████▊  | 15023/19090 [7:20:13<2:09:53,  1.92s/it]
 79%|███████▊  | 15024/19090 [7:20:15<1:56:59,  1.73s/it]
 79%|███████▊  | 15025/19090 [7:20:16<1:48:07,  1.60s/it]
 79%|███████▊  | 15026/19090 [7:20:19<2:20:19,  2.07s/it]
 79%|███████▊  | 15027/19090 [7:20:20<2:04:24,  1.84s/it]
 79%|███████▊  | 15028/19090 [7:20:22<1:53:13,  1.67s/it]
 79%|███████▊  | 15029/19090 [7:20:24<2:12:32,  1.96s/it][gradnorm] step 15029: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 524.3

 79%|███████▊  | 15030/19090 [7:20:26<1:58:57,  1.76s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 747.8812866210938, 'learning_rate': 6.58158395388114e-06, 'recon_loss': 0.0015898196608759463, 'recon_static': 0.0006185255377204157, 'recon_motion': 0.0009712940649478696, 'recon_norm': 1.3117518931627274, 'recon_norm_std': 0.040105495415627954, 'target_norm': 1.6202625513076783, 'recon_cos': 0.6281747609376908, 'gradnorm/compressor.query': 0.4713683117452972, 'gradnorm/compressor.layer0': 15.128330336959849, 'gradnorm/compressor.layer1': 743.2017727373906, 'gradnorm/compressor.layer2': 75.70462621060551, 'gradnorm/compressor.layer3': 30.42078155135585, 'gradnorm/compressor.layer4': 6.633037945226342, 'gradnorm/compressor.layer5': 4.1665711709841196, 'gradnorm/compressor.layer6': 3.0771509168706292, 'gradnorm/compressor.layer7': 2.7454185469700016, 'gradnorm/decoder.layer0': 2.977388362505241, 'gradnorm/decoder.layer1': 1.8480961540068623, 'gradnorm/decoder.layer2': 1.0492878906010783, 'gradnorm/max_param': 524.3499755859375, 'epoch': 7.87}

 79%|███████▊  | 15030/19090 [7:20:26<1:58:57,  1.76s/it]
 79%|███████▊  | 15031/19090 [7:20:27<1:49:22,  1.62s/it]
 79%|███████▊  | 15032/19090 [7:20:30<2:08:37,  1.90s/it]
 79%|███████▊  | 15033/19090 [7:20:31<1:56:08,  1.72s/it]
 79%|███████▉  | 15034/19090 [7:20:32<1:47:29,  1.59s/it]
 79%|███████▉  | 15035/19090 [7:20:35<2:08:02,  1.89s/it]
 79%|███████▉  | 15036/19090 [7:20:36<1:55:50,  1.71s/it]
 79%|███████▉  | 15037/19090 [7:20:37<1:47:15,  1.59s/it]
 79%|███████▉  | 15038/19090 [7:20:40<2:09:03,  1.91s/it]
 79%|███████▉  | 15039/19090 [7:20:41<1:56:29,  1.73s/it][gradnorm] step 15039: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 543.0

 79%|███████▉  | 15040/19090 [7:20:43<1:47:42,  1.60s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 777.5325317382812, 'learning_rate': 6.550704432628904e-06, 'recon_loss': 0.0017559951142175124, 'recon_static': 0.0007785495123243891, 'recon_motion': 0.0009774456673767417, 'recon_norm': 1.3703453332185744, 'recon_norm_std': 0.03871732894331217, 'target_norm': 1.62269247174263, 'recon_cos': 0.5948240414261818, 'gradnorm/compressor.query': 0.3935690751770647, 'gradnorm/compressor.layer0': 12.435905622721112, 'gradnorm/compressor.layer1': 769.4688508965129, 'gradnorm/compressor.layer2': 106.99906906721198, 'gradnorm/compressor.layer3': 28.077375042154717, 'gradnorm/compressor.layer4': 6.746499178067687, 'gradnorm/compressor.layer5': 3.7651827857239692, 'gradnorm/compressor.layer6': 2.428962231294906, 'gradnorm/compressor.layer7': 2.601576063332699, 'gradnorm/decoder.layer0': 2.5773573223773054, 'gradnorm/decoder.layer1': 1.6765802074965543, 'gradnorm/decoder.layer2': 0.8823297581712526, 'gradnorm/max_param': 542.9837036132812, 'epoch': 7.88}

 79%|███████▉  | 15040/19090 [7:20:43<1:47:42,  1.60s/it]
 79%|███████▉  | 15041/19090 [7:20:46<2:15:53,  2.01s/it]
 79%|███████▉  | 15042/19090 [7:20:47<2:01:16,  1.80s/it]
 79%|███████▉  | 15043/19090 [7:20:48<1:51:11,  1.65s/it]
 79%|███████▉  | 15044/19090 [7:20:51<2:16:29,  2.02s/it]
 79%|███████▉  | 15045/19090 [7:20:52<2:01:49,  1.81s/it]
 79%|███████▉  | 15046/19090 [7:20:54<1:51:24,  1.65s/it]
 79%|███████▉  | 15047/19090 [7:20:56<2:15:02,  2.00s/it]
 79%|███████▉  | 15048/19090 [7:20:58<2:00:49,  1.79s/it]
 79%|███████▉  | 15049/19090 [7:20:59<1:50:50,  1.65s/it][gradnorm] step 15049: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 522.6

 79%|███████▉  | 15050/19090 [7:21:02<2:11:45,  1.96s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 750.4360961914062, 'learning_rate': 6.519886596798258e-06, 'recon_loss': 0.0017063019418856128, 'recon_static': 0.0007387335776002146, 'recon_motion': 0.0009675683817476965, 'recon_norm': 1.3600072979927063, 'recon_norm_std': 0.03896446609869599, 'target_norm': 1.6183848679065704, 'recon_cos': 0.6036271005868912, 'gradnorm/compressor.query': 0.38217440772226224, 'gradnorm/compressor.layer0': 14.523751635588706, 'gradnorm/compressor.layer1': 740.4712604146092, 'gradnorm/compressor.layer2': 118.65835690781539, 'gradnorm/compressor.layer3': 21.821358617064412, 'gradnorm/compressor.layer4': 6.930776000949807, 'gradnorm/compressor.layer5': 3.947967783950106, 'gradnorm/compressor.layer6': 2.4087067979977768, 'gradnorm/compressor.layer7': 2.830254527531542, 'gradnorm/decoder.layer0': 2.8028143362940314, 'gradnorm/decoder.layer1': 1.825854595181824, 'gradnorm/decoder.layer2': 0.9852853531610921, 'gradnorm/max_param': 522.5795288085938, 'epoch': 7.88}

 79%|███████▉  | 15050/19090 [7:21:02<2:11:45,  1.96s/it]
 79%|███████▉  | 15051/19090 [7:21:03<1:58:19,  1.76s/it]
 79%|███████▉  | 15052/19090 [7:21:04<1:48:49,  1.62s/it]
 79%|███████▉  | 15053/19090 [7:21:07<2:14:09,  1.99s/it]
 79%|███████▉  | 15054/19090 [7:21:08<1:59:58,  1.78s/it]
 79%|███████▉  | 15055/19090 [7:21:10<1:50:08,  1.64s/it]
 79%|███████▉  | 15056/19090 [7:21:12<2:08:31,  1.91s/it]
 79%|███████▉  | 15057/19090 [7:21:14<1:56:13,  1.73s/it]
 79%|███████▉  | 15058/19090 [7:21:15<1:47:39,  1.60s/it]
 79%|███████▉  | 15059/19090 [7:21:18<2:12:16,  1.97s/it][gradnorm] step 15059: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 567.6

 79%|███████▉  | 15060/19090 [7:21:19<1:58:37,  1.77s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 812.9230346679688, 'learning_rate': 6.489130549428968e-06, 'recon_loss': 0.0017257459869142622, 'recon_static': 0.000754207062709611, 'recon_motion': 0.00097153897077078, 'recon_norm': 1.3655899614095688, 'recon_norm_std': 0.03838264876976609, 'target_norm': 1.6210042417049408, 'recon_cos': 0.6010214135050773, 'gradnorm/compressor.query': 0.61088514274387, 'gradnorm/compressor.layer0': 23.840018804145924, 'gradnorm/compressor.layer1': 804.4394787676696, 'gradnorm/compressor.layer2': 109.11936644249957, 'gradnorm/compressor.layer3': 33.36883648026816, 'gradnorm/compressor.layer4': 8.936126816904121, 'gradnorm/compressor.layer5': 4.520841865121318, 'gradnorm/compressor.layer6': 2.8748202060374166, 'gradnorm/compressor.layer7': 3.0497487136634223, 'gradnorm/decoder.layer0': 3.015597486617527, 'gradnorm/decoder.layer1': 1.911709705700855, 'gradnorm/decoder.layer2': 1.0653971274599106, 'gradnorm/max_param': 567.5635986328125, 'epoch': 7.89}

 79%|███████▉  | 15060/19090 [7:21:19<1:58:37,  1.77s/it]
 79%|███████▉  | 15061/19090 [7:21:20<1:49:03,  1.62s/it]
 79%|███████▉  | 15062/19090 [7:21:23<2:12:06,  1.97s/it]
 79%|███████▉  | 15063/19090 [7:21:24<1:58:35,  1.77s/it]
 79%|███████▉  | 15064/19090 [7:21:26<1:48:58,  1.62s/it]
 79%|███████▉  | 15065/19090 [7:21:29<2:15:46,  2.02s/it]
 79%|███████▉  | 15066/19090 [7:21:30<2:00:59,  1.80s/it]
 79%|███████▉  | 15067/19090 [7:21:31<1:50:38,  1.65s/it]
 79%|███████▉  | 15068/19090 [7:21:34<2:10:30,  1.95s/it]
 79%|███████▉  | 15069/19090 [7:21:35<1:57:17,  1.75s/it][gradnorm] step 15069: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 629.5

 79%|███████▉  | 15070/19090 [7:21:36<1:48:00,  1.61s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 901.0980224609375, 'learning_rate': 6.45843639335422e-06, 'recon_loss': 0.0017287070775637403, 'recon_static': 0.0007672656225622632, 'recon_motion': 0.0009614414782845415, 'recon_norm': 1.3765056014060975, 'recon_norm_std': 0.03876425316557288, 'target_norm': 1.6168406933546067, 'recon_cos': 0.6001378849148751, 'gradnorm/compressor.query': 0.495023338557593, 'gradnorm/compressor.layer0': 15.287618296953891, 'gradnorm/compressor.layer1': 891.9490246084695, 'gradnorm/compressor.layer2': 122.97487675334341, 'gradnorm/compressor.layer3': 30.615427194867276, 'gradnorm/compressor.layer4': 7.900321818689868, 'gradnorm/compressor.layer5': 3.9673610870137566, 'gradnorm/compressor.layer6': 2.965122456072328, 'gradnorm/compressor.layer7': 3.0687408695745506, 'gradnorm/decoder.layer0': 3.015165778263304, 'gradnorm/decoder.layer1': 1.9853623704667693, 'gradnorm/decoder.layer2': 1.080327534635809, 'gradnorm/max_param': 629.5066528320312, 'epoch': 7.89}

 79%|███████▉  | 15070/19090 [7:21:36<1:48:00,  1.61s/it]
 79%|███████▉  | 15071/19090 [7:21:39<2:07:11,  1.90s/it]
 79%|███████▉  | 15072/19090 [7:21:40<1:55:05,  1.72s/it]
 79%|███████▉  | 15073/19090 [7:21:42<1:46:28,  1.59s/it]
 79%|███████▉  | 15074/19090 [7:21:44<2:03:42,  1.85s/it]
 79%|███████▉  | 15075/19090 [7:21:45<1:52:20,  1.68s/it]
 79%|███████▉  | 15076/19090 [7:21:47<1:44:23,  1.56s/it]
 79%|███████▉  | 15077/19090 [7:21:49<2:05:40,  1.88s/it]
 79%|███████▉  | 15078/19090 [7:21:51<1:53:52,  1.70s/it]
 79%|███████▉  | 15079/19090 [7:21:52<1:45:35,  1.58s/it][gradnorm] step 15079: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 640.9

 79%|███████▉  | 15080/19090 [7:21:55<2:07:34,  1.91s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 918.9139404296875, 'learning_rate': 6.4278042312002575e-06, 'recon_loss': 0.0017555955389980226, 'recon_static': 0.0007731051402515732, 'recon_motion': 0.0009824904322158546, 'recon_norm': 1.3671948730945587, 'recon_norm_std': 0.03768983688205481, 'target_norm': 1.6218319058418273, 'recon_cos': 0.5937609225511551, 'gradnorm/compressor.query': 0.5027907392535902, 'gradnorm/compressor.layer0': 21.468097526930077, 'gradnorm/compressor.layer1': 908.3108292319321, 'gradnorm/compressor.layer2': 132.4905362186296, 'gradnorm/compressor.layer3': 35.021057769643214, 'gradnorm/compressor.layer4': 8.699188492734763, 'gradnorm/compressor.layer5': 4.70932703025399, 'gradnorm/compressor.layer6': 2.9256088144905505, 'gradnorm/compressor.layer7': 3.354065571716659, 'gradnorm/decoder.layer0': 3.1808292672785323, 'gradnorm/decoder.layer1': 1.9771717926698964, 'gradnorm/decoder.layer2': 1.1579533994452735, 'gradnorm/max_param': 640.9063110351562, 'epoch': 7.9}

 79%|███████▉  | 15080/19090 [7:21:55<2:07:34,  1.91s/it]
 79%|███████▉  | 15081/19090 [7:21:56<1:55:12,  1.72s/it]
 79%|███████▉  | 15082/19090 [7:21:57<1:46:34,  1.60s/it]
 79%|███████▉  | 15083/19090 [7:21:59<2:01:06,  1.81s/it]
 79%|███████▉  | 15084/19090 [7:22:01<1:50:43,  1.66s/it]
 79%|███████▉  | 15085/19090 [7:22:02<1:43:20,  1.55s/it]
 79%|███████▉  | 15086/19090 [7:22:04<1:59:43,  1.79s/it]
 79%|███████▉  | 15087/19090 [7:22:06<1:49:43,  1.64s/it]
 79%|███████▉  | 15088/19090 [7:22:07<1:42:46,  1.54s/it]
 79%|███████▉  | 15089/19090 [7:22:10<2:05:32,  1.88s/it][gradnorm] step 15089: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 582.5

 79%|███████▉  | 15090/19090 [7:22:11<1:53:53,  1.71s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 838.1820678710938, 'learning_rate': 6.397234165386051e-06, 'recon_loss': 0.0017733785847667606, 'recon_static': 0.000795816199388355, 'recon_motion': 0.0009775623882887885, 'recon_norm': 1.3861635476350784, 'recon_norm_std': 0.038503385800868276, 'target_norm': 1.6217792630195618, 'recon_cos': 0.592819957435131, 'gradnorm/compressor.query': 0.5291023319164158, 'gradnorm/compressor.layer0': 20.873508788429056, 'gradnorm/compressor.layer1': 825.665943345128, 'gradnorm/compressor.layer2': 138.9281373363582, 'gradnorm/compressor.layer3': 31.0206062986641, 'gradnorm/compressor.layer4': 8.660058020558154, 'gradnorm/compressor.layer5': 4.7112962070627225, 'gradnorm/compressor.layer6': 2.646740214801679, 'gradnorm/compressor.layer7': 3.0871649123994165, 'gradnorm/decoder.layer0': 2.7498733317940056, 'gradnorm/decoder.layer1': 1.747387639200372, 'gradnorm/decoder.layer2': 1.1221702803392997, 'gradnorm/max_param': 582.5298461914062, 'epoch': 7.9}

 79%|███████▉  | 15090/19090 [7:22:11<1:53:53,  1.71s/it]
 79%|███████▉  | 15091/19090 [7:22:12<1:45:39,  1.59s/it]
 79%|███████▉  | 15092/19090 [7:22:14<1:56:06,  1.74s/it]
 79%|███████▉  | 15093/19090 [7:22:16<1:47:24,  1.61s/it]
 79%|███████▉  | 15094/19090 [7:22:17<1:41:06,  1.52s/it]
 79%|███████▉  | 15095/19090 [7:22:19<1:53:55,  1.71s/it]
 79%|███████▉  | 15096/19090 [7:22:20<1:45:43,  1.59s/it]
 79%|███████▉  | 15097/19090 [7:22:22<1:39:50,  1.50s/it]
 79%|███████▉  | 15098/19090 [7:22:24<1:49:42,  1.65s/it]
 79%|███████▉  | 15099/19090 [7:22:25<1:42:39,  1.54s/it][gradnorm] step 15099: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 456.5

 79%|███████▉  | 15100/19090 [7:22:26<1:37:55,  1.47s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 657.0209350585938, 'learning_rate': 6.366726298122952e-06, 'recon_loss': 0.0017996129783568904, 'recon_static': 0.0008182947363820859, 'recon_motion': 0.0009813182536163368, 'recon_norm': 1.3971885651350022, 'recon_norm_std': 0.038054489996284246, 'target_norm': 1.6218973249197006, 'recon_cos': 0.5894205823540688, 'gradnorm/compressor.query': 0.38333979122623885, 'gradnorm/compressor.layer0': 12.398197510673553, 'gradnorm/compressor.layer1': 646.9590645860061, 'gradnorm/compressor.layer2': 108.4214055059009, 'gradnorm/compressor.layer3': 32.95844218304033, 'gradnorm/compressor.layer4': 8.858698097487117, 'gradnorm/compressor.layer5': 4.300951546355941, 'gradnorm/compressor.layer6': 2.807173793113357, 'gradnorm/compressor.layer7': 3.0641269643986813, 'gradnorm/decoder.layer0': 2.6566036213374624, 'gradnorm/decoder.layer1': 1.6692659216306158, 'gradnorm/decoder.layer2': 1.0686805144805132, 'gradnorm/max_param': 456.5397644042969, 'epoch': 7.91}

 79%|███████▉  | 15100/19090 [7:22:26<1:37:55,  1.47s/it]
 79%|███████▉  | 15101/19090 [7:22:29<1:55:09,  1.73s/it]
 79%|███████▉  | 15102/19090 [7:22:30<1:46:23,  1.60s/it]
 79%|███████▉  | 15103/19090 [7:22:31<1:40:22,  1.51s/it]
 79%|███████▉  | 15104/19090 [7:22:33<1:51:53,  1.68s/it]
 79%|███████▉  | 15105/19090 [7:22:35<1:44:04,  1.57s/it]
 79%|███████▉  | 15106/19090 [7:22:36<1:38:40,  1.49s/it]
 79%|███████▉  | 15107/19090 [7:22:38<1:54:36,  1.73s/it]
 79%|███████▉  | 15108/19090 [7:22:40<1:45:53,  1.60s/it]
 79%|███████▉  | 15109/19090 [7:22:41<1:39:49,  1.50s/it][gradnorm] step 15109: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 649.4

 79%|███████▉  | 15110/19090 [7:22:43<1:57:54,  1.78s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 929.9127807617188, 'learning_rate': 6.3362807314143426e-06, 'recon_loss': 0.0017035183089319616, 'recon_static': 0.0007347888997173869, 'recon_motion': 0.0009687294601462782, 'recon_norm': 1.3549618035554887, 'recon_norm_std': 0.038001203816384076, 'target_norm': 1.6183808535337447, 'recon_cos': 0.6031812101602554, 'gradnorm/compressor.query': 0.4182907361624631, 'gradnorm/compressor.layer0': 17.799963893210833, 'gradnorm/compressor.layer1': 920.3177779984477, 'gradnorm/compressor.layer2': 127.07581859612394, 'gradnorm/compressor.layer3': 33.99273065384256, 'gradnorm/compressor.layer4': 9.107226895497565, 'gradnorm/compressor.layer5': 4.6637486235596874, 'gradnorm/compressor.layer6': 2.6284876769359133, 'gradnorm/compressor.layer7': 2.885808612814176, 'gradnorm/decoder.layer0': 2.8028733700834834, 'gradnorm/decoder.layer1': 1.7991619649416029, 'gradnorm/decoder.layer2': 1.1214340703584704, 'gradnorm/max_param': 649.4257202148438, 'epoch': 7.92}

 79%|███████▉  | 15110/19090 [7:22:43<1:57:54,  1.78s/it]
 79%|███████▉  | 15111/19090 [7:22:45<1:48:10,  1.63s/it]
 79%|███████▉  | 15112/19090 [7:22:46<1:41:28,  1.53s/it]
 79%|███████▉  | 15113/19090 [7:22:48<2:02:10,  1.84s/it]
 79%|███████▉  | 15114/19090 [7:22:50<1:51:15,  1.68s/it]
 79%|███████▉  | 15115/19090 [7:22:51<1:43:38,  1.56s/it]
 79%|███████▉  | 15116/19090 [7:22:54<2:06:32,  1.91s/it]
 79%|███████▉  | 15117/19090 [7:22:55<1:54:21,  1.73s/it]
 79%|███████▉  | 15118/19090 [7:22:56<1:45:44,  1.60s/it]
 79%|███████▉  | 15119/19090 [7:22:59<2:03:15,  1.86s/it][gradnorm] step 15119: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 686.0

 79%|███████▉  | 15120/19090 [7:23:00<1:52:01,  1.69s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 981.2960205078125, 'learning_rate': 6.305897567055316e-06, 'recon_loss': 0.0017184128111694009, 'recon_static': 0.0007463669957360253, 'recon_motion': 0.0009720458299852907, 'recon_norm': 1.3639940947294236, 'recon_norm_std': 0.03885070644319057, 'target_norm': 1.6180917471647263, 'recon_cos': 0.6011432722210884, 'gradnorm/compressor.query': 0.41464651056130325, 'gradnorm/compressor.layer0': 15.974521855465415, 'gradnorm/compressor.layer1': 972.0429517258998, 'gradnorm/compressor.layer2': 126.99519783612882, 'gradnorm/compressor.layer3': 39.157702736418926, 'gradnorm/compressor.layer4': 10.132952545774694, 'gradnorm/compressor.layer5': 4.441988112583597, 'gradnorm/compressor.layer6': 2.9895143046811246, 'gradnorm/compressor.layer7': 3.210715775828165, 'gradnorm/decoder.layer0': 3.317821618525987, 'gradnorm/decoder.layer1': 1.9003586330249742, 'gradnorm/decoder.layer2': 1.3024598708899346, 'gradnorm/max_param': 686.035400390625, 'epoch': 7.92}

 79%|███████▉  | 15120/19090 [7:23:00<1:52:01,  1.69s/it]
 79%|███████▉  | 15121/19090 [7:23:01<1:44:00,  1.57s/it]
 79%|███████▉  | 15122/19090 [7:23:04<2:02:21,  1.85s/it]
 79%|███████▉  | 15123/19090 [7:23:05<1:51:16,  1.68s/it]
 79%|███████▉  | 15124/19090 [7:23:06<1:43:32,  1.57s/it]
 79%|███████▉  | 15125/19090 [7:23:09<2:07:20,  1.93s/it]
 79%|███████▉  | 15126/19090 [7:23:11<1:54:46,  1.74s/it]
 79%|███████▉  | 15127/19090 [7:23:12<1:45:56,  1.60s/it]
 79%|███████▉  | 15128/19090 [7:23:14<2:00:17,  1.82s/it]
 79%|███████▉  | 15129/19090 [7:23:15<1:49:55,  1.67s/it][gradnorm] step 15129: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 542.3

 79%|███████▉  | 15130/19090 [7:23:17<1:42:31,  1.55s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 777.84619140625, 'learning_rate': 6.275576906632308e-06, 'recon_loss': 0.0016328593803336844, 'recon_static': 0.0006486698126536794, 'recon_motion': 0.0009841895443969406, 'recon_norm': 1.3252494394779206, 'recon_norm_std': 0.03834963794797659, 'target_norm': 1.619181004166603, 'recon_cos': 0.6180491209030151, 'gradnorm/compressor.query': 0.28308880570407396, 'gradnorm/compressor.layer0': 11.4797729405933, 'gradnorm/compressor.layer1': 768.393982602675, 'gradnorm/compressor.layer2': 113.38810477597507, 'gradnorm/compressor.layer3': 38.85888651934635, 'gradnorm/compressor.layer4': 8.01611231462141, 'gradnorm/compressor.layer5': 4.3664244392904825, 'gradnorm/compressor.layer6': 3.072386063278209, 'gradnorm/compressor.layer7': 3.1617009898809316, 'gradnorm/decoder.layer0': 3.106439479521402, 'gradnorm/decoder.layer1': 1.7008155773730622, 'gradnorm/decoder.layer2': 1.1509008652312178, 'gradnorm/max_param': 542.2782592773438, 'epoch': 7.93}

 79%|███████▉  | 15130/19090 [7:23:17<1:42:31,  1.55s/it]
 79%|███████▉  | 15131/19090 [7:23:20<2:11:00,  1.99s/it]
 79%|███████▉  | 15132/19090 [7:23:21<1:57:16,  1.78s/it]
 79%|███████▉  | 15133/19090 [7:23:22<1:47:36,  1.63s/it]
 79%|███████▉  | 15134/19090 [7:23:25<2:03:59,  1.88s/it]
 79%|███████▉  | 15135/19090 [7:23:26<1:52:21,  1.70s/it]
 79%|███████▉  | 15136/19090 [7:23:27<1:44:15,  1.58s/it]
 79%|███████▉  | 15137/19090 [7:23:30<2:00:52,  1.83s/it]
 79%|███████▉  | 15138/19090 [7:23:31<1:50:19,  1.68s/it]
 79%|███████▉  | 15139/19090 [7:23:32<1:42:58,  1.56s/it][gradnorm] step 15139: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 491.4

 79%|███████▉  | 15140/19090 [7:23:35<2:05:14,  1.90s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 706.8333740234375, 'learning_rate': 6.245318851522783e-06, 'recon_loss': 0.0016500110243214295, 'recon_static': 0.0006709532302920706, 'recon_motion': 0.0009790577852982097, 'recon_norm': 1.3275926887989045, 'recon_norm_std': 0.03941962504759431, 'target_norm': 1.6213735491037369, 'recon_cos': 0.6146954074501991, 'gradnorm/compressor.query': 0.4742980061245662, 'gradnorm/compressor.layer0': 15.26122933611868, 'gradnorm/compressor.layer1': 696.3709948727044, 'gradnorm/compressor.layer2': 114.57796199531566, 'gradnorm/compressor.layer3': 34.821810287227855, 'gradnorm/compressor.layer4': 7.5851819087364705, 'gradnorm/compressor.layer5': 4.134848426706935, 'gradnorm/compressor.layer6': 3.025415527556485, 'gradnorm/compressor.layer7': 3.157476054846065, 'gradnorm/decoder.layer0': 3.0419299564762206, 'gradnorm/decoder.layer1': 1.7421853788217538, 'gradnorm/decoder.layer2': 1.0233350513228452, 'gradnorm/max_param': 491.3779602050781, 'epoch': 7.93}

 79%|███████▉  | 15140/19090 [7:23:35<2:05:14,  1.90s/it]
 79%|███████▉  | 15141/19090 [7:23:36<1:53:11,  1.72s/it]
 79%|███████▉  | 15142/19090 [7:23:38<1:44:46,  1.59s/it]
 79%|███████▉  | 15143/19090 [7:23:40<2:06:39,  1.93s/it]
 79%|███████▉  | 15144/19090 [7:23:42<1:54:24,  1.74s/it]
 79%|███████▉  | 15145/19090 [7:23:43<1:45:48,  1.61s/it]
 79%|███████▉  | 15146/19090 [7:23:46<2:14:20,  2.04s/it]
 79%|███████▉  | 15147/19090 [7:23:47<1:59:40,  1.82s/it]
 79%|███████▉  | 15148/19090 [7:23:49<1:49:24,  1.67s/it]
 79%|███████▉  | 15149/19090 [7:23:52<2:12:42,  2.02s/it][gradnorm] step 15149: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 644.0

 79%|███████▉  | 15150/19090 [7:23:53<1:58:21,  1.80s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 938.515380859375, 'learning_rate': 6.215123502894879e-06, 'recon_loss': 0.001632033905480057, 'recon_static': 0.0006621499924222007, 'recon_motion': 0.0009698838548501953, 'recon_norm': 1.3296673566102981, 'recon_norm_std': 0.03963761497288942, 'target_norm': 1.6190096229314803, 'recon_cos': 0.618675771355629, 'gradnorm/compressor.query': 0.5727000231490905, 'gradnorm/compressor.layer0': 19.343656918697697, 'gradnorm/compressor.layer1': 912.6519394051601, 'gradnorm/compressor.layer2': 214.59910844118156, 'gradnorm/compressor.layer3': 36.60149162880012, 'gradnorm/compressor.layer4': 7.417824501213988, 'gradnorm/compressor.layer5': 4.892672661574186, 'gradnorm/compressor.layer6': 3.1465269591411498, 'gradnorm/compressor.layer7': 3.153862078198852, 'gradnorm/decoder.layer0': 2.69616561877869, 'gradnorm/decoder.layer1': 1.7710422018699734, 'gradnorm/decoder.layer2': 1.0881555532584293, 'gradnorm/max_param': 644.0421752929688, 'epoch': 7.94}

 79%|███████▉  | 15150/19090 [7:23:53<1:58:21,  1.80s/it]
 79%|███████▉  | 15151/19090 [7:23:54<1:48:16,  1.65s/it]
 79%|███████▉  | 15152/19090 [7:23:57<2:15:32,  2.07s/it]
 79%|███████▉  | 15153/19090 [7:23:58<2:00:17,  1.83s/it]
 79%|███████▉  | 15154/19090 [7:24:00<1:49:32,  1.67s/it]
 79%|███████▉  | 15155/19090 [7:24:03<2:13:30,  2.04s/it]
 79%|███████▉  | 15156/19090 [7:24:04<1:58:53,  1.81s/it]
 79%|███████▉  | 15157/19090 [7:24:05<1:48:36,  1.66s/it]
 79%|███████▉  | 15158/19090 [7:24:08<2:15:42,  2.07s/it]
 79%|███████▉  | 15159/19090 [7:24:10<2:00:19,  1.84s/it][gradnorm] step 15159: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 548.0

 79%|███████▉  | 15160/19090 [7:24:11<1:49:54,  1.68s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 790.8003540039062, 'learning_rate': 6.1849909617070716e-06, 'recon_loss': 0.0016510059620486572, 'recon_static': 0.0006663299165666104, 'recon_motion': 0.0009846760163782164, 'recon_norm': 1.3276702076196671, 'recon_norm_std': 0.03952139616012573, 'target_norm': 1.6217907875776292, 'recon_cos': 0.614810261130333, 'gradnorm/compressor.query': 0.44072502029092586, 'gradnorm/compressor.layer0': 18.05374514119015, 'gradnorm/compressor.layer1': 776.5326779988077, 'gradnorm/compressor.layer2': 143.24493473810864, 'gradnorm/compressor.layer3': 37.199518803591396, 'gradnorm/compressor.layer4': 8.90377269891397, 'gradnorm/compressor.layer5': 5.004540098784482, 'gradnorm/compressor.layer6': 2.804450659992701, 'gradnorm/compressor.layer7': 2.8720462386890584, 'gradnorm/decoder.layer0': 2.898195817027457, 'gradnorm/decoder.layer1': 1.7635373627155178, 'gradnorm/decoder.layer2': 0.9973660888751882, 'gradnorm/max_param': 548.019775390625, 'epoch': 7.94}

 79%|███████▉  | 15160/19090 [7:24:11<1:49:54,  1.68s/it]
 79%|███████▉  | 15161/19090 [7:24:13<2:02:10,  1.87s/it]
 79%|███████▉  | 15162/19090 [7:24:14<1:50:55,  1.69s/it]
 79%|███████▉  | 15163/19090 [7:24:16<1:43:01,  1.57s/it]
 79%|███████▉  | 15164/19090 [7:24:19<2:07:22,  1.95s/it]
 79%|███████▉  | 15165/19090 [7:24:20<1:54:28,  1.75s/it]
 79%|███████▉  | 15166/19090 [7:24:21<1:45:29,  1.61s/it]
 79%|███████▉  | 15167/19090 [7:24:24<2:15:02,  2.07s/it]
 79%|███████▉  | 15168/19090 [7:24:26<1:59:58,  1.84s/it]
 79%|███████▉  | 15169/19090 [7:24:27<1:49:16,  1.67s/it][gradnorm] step 15169: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 513.9

 79%|███████▉  | 15170/19090 [7:24:29<2:08:17,  1.96s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 737.0545043945312, 'learning_rate': 6.154921328707841e-06, 'recon_loss': 0.00172970722487662, 'recon_static': 0.0007570279674837365, 'recon_motion': 0.0009726792923174798, 'recon_norm': 1.3696797788143158, 'recon_norm_std': 0.03892973875626922, 'target_norm': 1.6190871924161911, 'recon_cos': 0.599899785220623, 'gradnorm/compressor.query': 0.4077176396622417, 'gradnorm/compressor.layer0': 15.300044759049113, 'gradnorm/compressor.layer1': 728.221463567231, 'gradnorm/compressor.layer2': 108.9360567173468, 'gradnorm/compressor.layer3': 26.549848779509908, 'gradnorm/compressor.layer4': 8.874757145527411, 'gradnorm/compressor.layer5': 4.923994110711277, 'gradnorm/compressor.layer6': 3.0548771557872314, 'gradnorm/compressor.layer7': 3.1764949733238614, 'gradnorm/decoder.layer0': 3.157919715899673, 'gradnorm/decoder.layer1': 1.7477154124464407, 'gradnorm/decoder.layer2': 1.0980875925760856, 'gradnorm/max_param': 513.9159545898438, 'epoch': 7.95}

 79%|███████▉  | 15170/19090 [7:24:29<2:08:17,  1.96s/it]
 79%|███████▉  | 15171/19090 [7:24:31<1:55:27,  1.77s/it]
 79%|███████▉  | 15172/19090 [7:24:32<1:45:57,  1.62s/it]
 79%|███████▉  | 15173/19090 [7:24:35<2:04:17,  1.90s/it]
 79%|███████▉  | 15174/19090 [7:24:36<1:52:07,  1.72s/it]
 79%|███████▉  | 15175/19090 [7:24:37<1:43:44,  1.59s/it]
 79%|███████▉  | 15176/19090 [7:24:40<2:12:36,  2.03s/it]
 80%|███████▉  | 15177/19090 [7:24:42<1:58:03,  1.81s/it]
 80%|███████▉  | 15178/19090 [7:24:43<1:47:41,  1.65s/it]
 80%|███████▉  | 15179/19090 [7:24:46<2:11:31,  2.02s/it][gradnorm] step 15179: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 480.7

 80%|███████▉  | 15180/19090 [7:24:47<1:57:11,  1.80s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 688.9183349609375, 'learning_rate': 6.1249147044353415e-06, 'recon_loss': 0.0016563441255129873, 'recon_static': 0.0006770874984795228, 'recon_motion': 0.0009792565935640596, 'recon_norm': 1.331492617726326, 'recon_norm_std': 0.03952688435092568, 'target_norm': 1.6230542540550232, 'recon_cos': 0.6145654484629631, 'gradnorm/compressor.query': 0.46632725543482545, 'gradnorm/compressor.layer0': 16.151113257997302, 'gradnorm/compressor.layer1': 681.4549462363598, 'gradnorm/compressor.layer2': 96.1334064513346, 'gradnorm/compressor.layer3': 24.955810555234425, 'gradnorm/compressor.layer4': 7.17118483071108, 'gradnorm/compressor.layer5': 4.651473699868291, 'gradnorm/compressor.layer6': 2.6906513102061393, 'gradnorm/compressor.layer7': 2.8990538730885675, 'gradnorm/decoder.layer0': 3.0335970571825586, 'gradnorm/decoder.layer1': 1.72044808860805, 'gradnorm/decoder.layer2': 1.135466157493713, 'gradnorm/max_param': 480.654052734375, 'epoch': 7.95}

 80%|███████▉  | 15180/19090 [7:24:47<1:57:11,  1.80s/it]
 80%|███████▉  | 15181/19090 [7:24:48<1:47:22,  1.65s/it]
 80%|███████▉  | 15182/19090 [7:24:51<2:08:32,  1.97s/it]
 80%|███████▉  | 15183/19090 [7:24:52<1:55:09,  1.77s/it]
 80%|███████▉  | 15184/19090 [7:24:54<1:45:51,  1.63s/it]
 80%|███████▉  | 15185/19090 [7:24:57<2:12:01,  2.03s/it]
 80%|███████▉  | 15186/19090 [7:24:58<1:57:29,  1.81s/it]
 80%|███████▉  | 15187/19090 [7:24:59<1:47:22,  1.65s/it]
 80%|███████▉  | 15188/19090 [7:25:02<2:08:23,  1.97s/it]
 80%|███████▉  | 15189/19090 [7:25:03<1:55:01,  1.77s/it][gradnorm] step 15189: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 367.8

 80%|███████▉  | 15190/19090 [7:25:04<1:45:36,  1.62s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 530.1837768554688, 'learning_rate': 6.094971189217042e-06, 'recon_loss': 0.0017060022364603356, 'recon_static': 0.0007292259717360139, 'recon_motion': 0.0009767762137926183, 'recon_norm': 1.3646425902843475, 'recon_norm_std': 0.038552239444106816, 'target_norm': 1.6184818923473359, 'recon_cos': 0.6048386126756669, 'gradnorm/compressor.query': 0.556012331090006, 'gradnorm/compressor.layer0': 16.27177027285292, 'gradnorm/compressor.layer1': 521.575198317558, 'gradnorm/compressor.layer2': 89.37580856277609, 'gradnorm/compressor.layer3': 26.384327762231194, 'gradnorm/compressor.layer4': 7.476326264064342, 'gradnorm/compressor.layer5': 4.368010032320903, 'gradnorm/compressor.layer6': 2.906505778552535, 'gradnorm/compressor.layer7': 2.9448745361441753, 'gradnorm/decoder.layer0': 2.8980859095948293, 'gradnorm/decoder.layer1': 1.7755349817119794, 'gradnorm/decoder.layer2': 1.1339301022410921, 'gradnorm/max_param': 367.84979248046875, 'epoch': 7.96}

 80%|███████▉  | 15190/19090 [7:25:04<1:45:36,  1.62s/it]
 80%|███████▉  | 15191/19090 [7:25:07<2:07:44,  1.97s/it]
 80%|███████▉  | 15192/19090 [7:25:09<1:54:31,  1.76s/it]
 80%|███████▉  | 15193/19090 [7:25:10<1:45:20,  1.62s/it]
 80%|███████▉  | 15194/19090 [7:25:13<2:11:11,  2.02s/it]
 80%|███████▉  | 15195/19090 [7:25:14<1:56:56,  1.80s/it]
 80%|███████▉  | 15196/19090 [7:25:15<1:46:55,  1.65s/it]
 80%|███████▉  | 15197/19090 [7:25:18<2:08:58,  1.99s/it]
 80%|███████▉  | 15198/19090 [7:25:19<1:55:18,  1.78s/it]
 80%|███████▉  | 15199/19090 [7:25:21<1:45:48,  1.63s/it][gradnorm] step 15199: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 647.3

 80%|███████▉  | 15200/19090 [7:25:24<2:14:10,  2.07s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 925.2404174804688, 'learning_rate': 6.0650908831694166e-06, 'recon_loss': 0.001730479917023331, 'recon_static': 0.0007563427672721446, 'recon_motion': 0.0009741371584823355, 'recon_norm': 1.3635943979024887, 'recon_norm_std': 0.03861935157328844, 'target_norm': 1.6213307976722717, 'recon_cos': 0.5994739949703216, 'gradnorm/compressor.query': 0.5505207067782408, 'gradnorm/compressor.layer0': 23.240125958992643, 'gradnorm/compressor.layer1': 917.1792695542132, 'gradnorm/compressor.layer2': 115.3699171076997, 'gradnorm/compressor.layer3': 29.831472441816796, 'gradnorm/compressor.layer4': 7.920622416227731, 'gradnorm/compressor.layer5': 4.229526573363742, 'gradnorm/compressor.layer6': 3.036504850909138, 'gradnorm/compressor.layer7': 2.8638292649487425, 'gradnorm/decoder.layer0': 2.9473004095756288, 'gradnorm/decoder.layer1': 1.769435113649038, 'gradnorm/decoder.layer2': 1.2407649790013215, 'gradnorm/max_param': 647.2628173828125, 'epoch': 7.96}

 80%|███████▉  | 15200/19090 [7:25:24<2:14:10,  2.07s/it]
 80%|███████▉  | 15201/19090 [7:25:25<1:58:54,  1.83s/it]
 80%|███████▉  | 15202/19090 [7:25:26<1:48:17,  1.67s/it]
 80%|███████▉  | 15203/19090 [7:25:29<2:09:09,  1.99s/it]
 80%|███████▉  | 15204/19090 [7:25:30<1:55:21,  1.78s/it]
 80%|███████▉  | 15205/19090 [7:25:32<1:45:41,  1.63s/it]
 80%|███████▉  | 15206/19090 [7:25:34<2:06:39,  1.96s/it]
 80%|███████▉  | 15207/19090 [7:25:36<1:53:32,  1.75s/it]
 80%|███████▉  | 15208/19090 [7:25:37<1:44:24,  1.61s/it]
 80%|███████▉  | 15209/19090 [7:25:40<2:08:43,  1.99s/it][gradnorm] step 15209: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 470.4

 80%|███████▉  | 15210/19090 [7:25:41<1:55:02,  1.78s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 675.1663208007812, 'learning_rate': 6.0352738861975924e-06, 'recon_loss': 0.0017182157578645274, 'recon_static': 0.0007389512160443701, 'recon_motion': 0.0009792645636480302, 'recon_norm': 1.3625579714775085, 'recon_norm_std': 0.03920853165909648, 'target_norm': 1.6208212971687317, 'recon_cos': 0.6022039234638215, 'gradnorm/compressor.query': 0.32387822434319513, 'gradnorm/compressor.layer0': 12.205042337971618, 'gradnorm/compressor.layer1': 666.5834062591117, 'gradnorm/compressor.layer2': 102.63542882455356, 'gradnorm/compressor.layer3': 26.947638707938655, 'gradnorm/compressor.layer4': 7.595584349563115, 'gradnorm/compressor.layer5': 4.494464012158936, 'gradnorm/compressor.layer6': 2.5231459127319678, 'gradnorm/compressor.layer7': 2.8975295626668456, 'gradnorm/decoder.layer0': 3.1033200297768766, 'gradnorm/decoder.layer1': 1.743092596902512, 'gradnorm/decoder.layer2': 1.2200060998263846, 'gradnorm/max_param': 470.4330749511719, 'epoch': 7.97}

 80%|███████▉  | 15210/19090 [7:25:41<1:55:02,  1.78s/it]
 80%|███████▉  | 15211/19090 [7:25:42<1:45:30,  1.63s/it]
 80%|███████▉  | 15212/19090 [7:25:45<2:11:50,  2.04s/it]
 80%|███████▉  | 15213/19090 [7:25:47<1:57:06,  1.81s/it]
 80%|███████▉  | 15214/19090 [7:25:48<1:46:51,  1.65s/it]
 80%|███████▉  | 15215/19090 [7:25:51<2:14:24,  2.08s/it]
 80%|███████▉  | 15216/19090 [7:25:52<1:58:52,  1.84s/it]
 80%|███████▉  | 15217/19090 [7:25:54<1:48:07,  1.68s/it]
 80%|███████▉  | 15218/19090 [7:25:56<2:11:04,  2.03s/it]
 80%|███████▉  | 15219/19090 [7:25:58<1:56:32,  1.81s/it][gradnorm] step 15219: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 431.4

 80%|███████▉  | 15220/19090 [7:25:59<1:46:34,  1.65s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 623.8560180664062, 'learning_rate': 6.005520297995032e-06, 'recon_loss': 0.0018203941464889795, 'recon_static': 0.0008489767933497205, 'recon_motion': 0.0009714173749671318, 'recon_norm': 1.4088427662849425, 'recon_norm_std': 0.037501603923738004, 'target_norm': 1.618884837627411, 'recon_cos': 0.5839168354868889, 'gradnorm/compressor.query': 0.5054570958770647, 'gradnorm/compressor.layer0': 17.600262102429575, 'gradnorm/compressor.layer1': 611.4788426429814, 'gradnorm/compressor.layer2': 118.18678113213423, 'gradnorm/compressor.layer3': 29.991832384261304, 'gradnorm/compressor.layer4': 8.21247699798196, 'gradnorm/compressor.layer5': 4.022467694481761, 'gradnorm/compressor.layer6': 2.6615325459524373, 'gradnorm/compressor.layer7': 2.741497855124921, 'gradnorm/decoder.layer0': 3.06484973059933, 'gradnorm/decoder.layer1': 1.8656204223576651, 'gradnorm/decoder.layer2': 1.0978810361887132, 'gradnorm/max_param': 431.3771667480469, 'epoch': 7.97}

 80%|███████▉  | 15220/19090 [7:25:59<1:46:34,  1.65s/it]
 80%|███████▉  | 15221/19090 [7:26:02<2:10:07,  2.02s/it]
 80%|███████▉  | 15222/19090 [7:26:03<1:55:55,  1.80s/it]
 80%|███████▉  | 15223/19090 [7:26:04<1:46:07,  1.65s/it]
 80%|███████▉  | 15224/19090 [7:26:07<2:07:02,  1.97s/it]
 80%|███████▉  | 15225/19090 [7:26:09<1:53:45,  1.77s/it]
 80%|███████▉  | 15226/19090 [7:26:10<1:44:32,  1.62s/it]
 80%|███████▉  | 15227/19090 [7:26:13<2:09:52,  2.02s/it]
 80%|███████▉  | 15228/19090 [7:26:14<1:55:41,  1.80s/it]
 80%|███████▉  | 15229/19090 [7:26:15<1:45:51,  1.65s/it][gradnorm] step 15229: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 389.0

 80%|███████▉  | 15230/19090 [7:26:18<2:12:18,  2.06s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 559.5281372070312, 'learning_rate': 5.975830218043174e-06, 'recon_loss': 0.0019005470705451444, 'recon_static': 0.0009252896663383581, 'recon_motion': 0.0009752574638696388, 'recon_norm': 1.4320978611707686, 'recon_norm_std': 0.037528601195663217, 'target_norm': 1.6192166954278946, 'recon_cos': 0.5693612962961196, 'gradnorm/compressor.query': 0.24062241639570903, 'gradnorm/compressor.layer0': 9.717324302582604, 'gradnorm/compressor.layer1': 551.2891142585712, 'gradnorm/compressor.layer2': 91.41523994252819, 'gradnorm/compressor.layer3': 25.090548908528287, 'gradnorm/compressor.layer4': 6.052527021294978, 'gradnorm/compressor.layer5': 3.5712727730692215, 'gradnorm/compressor.layer6': 2.2846087460099094, 'gradnorm/compressor.layer7': 2.569861375357805, 'gradnorm/decoder.layer0': 2.612015809816005, 'gradnorm/decoder.layer1': 1.5373251931906047, 'gradnorm/decoder.layer2': 0.94229875100713, 'gradnorm/max_param': 388.9907531738281, 'epoch': 7.98}

 80%|███████▉  | 15230/19090 [7:26:18<2:12:18,  2.06s/it]
 80%|███████▉  | 15231/19090 [7:26:20<1:57:24,  1.83s/it]
 80%|███████▉  | 15232/19090 [7:26:21<1:47:05,  1.67s/it]
 80%|███████▉  | 15233/19090 [7:26:24<2:10:10,  2.02s/it]
 80%|███████▉  | 15234/19090 [7:26:25<1:56:04,  1.81s/it]
 80%|███████▉  | 15235/19090 [7:26:26<1:46:11,  1.65s/it]
 80%|███████▉  | 15236/19090 [7:26:29<2:09:25,  2.01s/it]
 80%|███████▉  | 15237/19090 [7:26:31<1:55:27,  1.80s/it]
 80%|███████▉  | 15238/19090 [7:26:32<1:45:38,  1.65s/it]
 80%|███████▉  | 15239/19090 [7:26:35<2:14:40,  2.10s/it][gradnorm] step 15239: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 347.5

 80%|███████▉  | 15240/19090 [7:26:36<1:59:08,  1.86s/it]
                                                         
{'loss': 0.0021, 'grad_norm': 500.8725891113281, 'learning_rate': 5.9462037456111345e-06, 'recon_loss': 0.0021149539097677916, 'recon_static': 0.0011415266111725942, 'recon_motion': 0.0009734273524372838, 'recon_norm': 1.5318660646677018, 'recon_norm_std': 0.03561762375757098, 'target_norm': 1.6211455464363098, 'recon_cos': 0.5434721261262894, 'gradnorm/compressor.query': 0.3927375887118379, 'gradnorm/compressor.layer0': 12.320264198145646, 'gradnorm/compressor.layer1': 492.50341052118614, 'gradnorm/compressor.layer2': 87.11318797708674, 'gradnorm/compressor.layer3': 22.139426490364777, 'gradnorm/compressor.layer4': 6.923078238658291, 'gradnorm/compressor.layer5': 3.7042544932014096, 'gradnorm/compressor.layer6': 2.2888568876965545, 'gradnorm/compressor.layer7': 2.4262619156621024, 'gradnorm/decoder.layer0': 2.5359374548723426, 'gradnorm/decoder.layer1': 1.6915044954236245, 'gradnorm/decoder.layer2': 0.9320404346457979, 'gradnorm/max_param': 347.4797668457031, 'epoch': 7.98}

 80%|███████▉  | 15240/19090 [7:26:36<1:59:08,  1.86s/it]
 80%|███████▉  | 15241/19090 [7:26:38<1:48:15,  1.69s/it]
 80%|███████▉  | 15242/19090 [7:26:41<2:12:50,  2.07s/it]
 80%|███████▉  | 15243/19090 [7:26:42<1:57:36,  1.83s/it]
 80%|███████▉  | 15244/19090 [7:26:43<1:46:58,  1.67s/it]
 80%|███████▉  | 15245/19090 [7:26:46<2:18:39,  2.16s/it]
 80%|███████▉  | 15246/19090 [7:26:48<2:01:41,  1.90s/it]
 80%|███████▉  | 15247/19090 [7:26:49<1:49:50,  1.72s/it]
 80%|███████▉  | 15248/19090 [7:26:52<2:14:06,  2.09s/it]
 80%|███████▉  | 15249/19090 [7:26:53<1:58:41,  1.85s/it][gradnorm] step 15249: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 757.7

 80%|███████▉  | 15250/19090 [7:26:55<1:47:55,  1.69s/it]
                                                         
{'loss': 0.002, 'grad_norm': 1086.6597900390625, 'learning_rate': 5.9166409797553415e-06, 'recon_loss': 0.001970695855561644, 'recon_static': 0.0009958540525985883, 'recon_motion': 0.0009748418116942049, 'recon_norm': 1.4558631271123885, 'recon_norm_std': 0.03646215293556452, 'target_norm': 1.619415709376335, 'recon_cos': 0.5575123578310013, 'gradnorm/compressor.query': 0.6040676271444996, 'gradnorm/compressor.layer0': 22.3192480659299, 'gradnorm/compressor.layer1': 1073.624294155083, 'gradnorm/compressor.layer2': 161.59988687341337, 'gradnorm/compressor.layer3': 37.595418967525156, 'gradnorm/compressor.layer4': 9.383321908658143, 'gradnorm/compressor.layer5': 4.174623648304176, 'gradnorm/compressor.layer6': 2.5848500711137423, 'gradnorm/compressor.layer7': 2.906630132282247, 'gradnorm/decoder.layer0': 2.911823465906964, 'gradnorm/decoder.layer1': 1.890767241872847, 'gradnorm/decoder.layer2': 1.1892487549245017, 'gradnorm/max_param': 757.6527709960938, 'epoch': 7.99}

 80%|███████▉  | 15250/19090 [7:26:55<1:47:55,  1.69s/it]
 80%|███████▉  | 15251/19090 [7:26:58<2:14:24,  2.10s/it]
 80%|███████▉  | 15252/19090 [7:26:59<1:58:54,  1.86s/it]
 80%|███████▉  | 15253/19090 [7:27:00<1:47:54,  1.69s/it]
 80%|███████▉  | 15254/19090 [7:27:04<2:22:12,  2.22s/it]
 80%|███████▉  | 15255/19090 [7:27:05<2:04:22,  1.95s/it]
 80%|███████▉  | 15256/19090 [7:27:06<1:51:48,  1.75s/it]
 80%|███████▉  | 15257/19090 [7:27:09<2:04:37,  1.95s/it]
 80%|███████▉  | 15258/19090 [7:27:10<1:51:58,  1.75s/it]
 80%|███████▉  | 15259/19090 [7:27:11<1:43:01,  1.61s/it][gradnorm] step 15259: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 651.5

 80%|███████▉  | 15260/19090 [7:27:14<2:02:10,  1.91s/it]
                                                         
{'loss': 0.002, 'grad_norm': 933.0736083984375, 'learning_rate': 5.887142019319228e-06, 'recon_loss': 0.0019932069088099526, 'recon_static': 0.0010195726572419517, 'recon_motion': 0.0009736342442920431, 'recon_norm': 1.4746820092201234, 'recon_norm_std': 0.03620046367868781, 'target_norm': 1.619725489616394, 'recon_cos': 0.5571002572774887, 'gradnorm/compressor.query': 0.46205880476078875, 'gradnorm/compressor.layer0': 16.286437513311498, 'gradnorm/compressor.layer1': 923.343584479797, 'gradnorm/compressor.layer2': 128.26626358867713, 'gradnorm/compressor.layer3': 35.03230756376445, 'gradnorm/compressor.layer4': 8.26756958901343, 'gradnorm/compressor.layer5': 3.9426051151096737, 'gradnorm/compressor.layer6': 2.620379241111587, 'gradnorm/compressor.layer7': 3.435347837025028, 'gradnorm/decoder.layer0': 3.2151648545306477, 'gradnorm/decoder.layer1': 1.9428288794813864, 'gradnorm/decoder.layer2': 1.1423799261355707, 'gradnorm/max_param': 651.5457763671875, 'epoch': 7.99}

 80%|███████▉  | 15260/19090 [7:27:14<2:02:10,  1.91s/it]
 80%|███████▉  | 15261/19090 [7:27:15<1:50:15,  1.73s/it]
 80%|███████▉  | 15262/19090 [7:27:16<1:41:53,  1.60s/it]
 80%|███████▉  | 15263/19090 [7:27:19<2:05:11,  1.96s/it]
 80%|███████▉  | 15264/19090 [7:27:21<1:52:27,  1.76s/it]
 80%|███████▉  | 15265/19090 [7:27:22<1:43:15,  1.62s/it]
 80%|███████▉  | 15266/19090 [7:27:25<2:11:50,  2.07s/it]
 80%|███████▉  | 15267/19090 [7:27:26<1:56:55,  1.84s/it]
 80%|███████▉  | 15268/19090 [7:27:28<1:46:27,  1.67s/it]
 80%|███████▉  | 15269/19090 [7:27:29<1:38:45,  1.55s/it][gradnorm] step 15269: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 457.6

 80%|███████▉  | 15270/19090 [7:27:30<1:33:23,  1.47s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 660.2677001953125, 'learning_rate': 5.857706962932894e-06, 'recon_loss': 0.0017805339535698294, 'recon_static': 0.0008075881632976234, 'recon_motion': 0.0009729457378853112, 'recon_norm': 1.3878807872533798, 'recon_norm_std': 0.038037578016519545, 'target_norm': 1.6189752846956253, 'recon_cos': 0.58981152176857, 'gradnorm/compressor.query': 0.4199389254634808, 'gradnorm/compressor.layer0': 14.78423048620847, 'gradnorm/compressor.layer1': 648.3906037258714, 'gradnorm/compressor.layer2': 119.67120483067345, 'gradnorm/compressor.layer3': 29.602996509538524, 'gradnorm/compressor.layer4': 8.545539486608643, 'gradnorm/compressor.layer5': 4.601024447267746, 'gradnorm/compressor.layer6': 2.730149148073698, 'gradnorm/compressor.layer7': 3.248779214314481, 'gradnorm/decoder.layer0': 3.237093460303509, 'gradnorm/decoder.layer1': 1.7886360693049268, 'gradnorm/decoder.layer2': 0.980783936194679, 'gradnorm/max_param': 457.5779113769531, 'epoch': 8.0}

 80%|███████▉  | 15270/19090 [7:27:30<1:33:23,  1.47s/it]
 80%|███████▉  | 15271/19090 [7:27:31<1:29:36,  1.41s/it]
 80%|████████  | 15272/19090 [7:27:33<1:35:05,  1.49s/it]
 80%|████████  | 15273/19090 [7:27:41<3:37:59,  3.43s/it]
 80%|████████  | 15274/19090 [7:27:42<2:57:17,  2.79s/it]
 80%|████████  | 15275/19090 [7:27:46<3:17:18,  3.10s/it]
 80%|████████  | 15276/19090 [7:27:47<2:42:37,  2.56s/it]
 80%|████████  | 15277/19090 [7:27:49<2:18:23,  2.18s/it]
 80%|████████  | 15278/19090 [7:27:51<2:20:42,  2.21s/it]
 80%|████████  | 15279/19090 [7:27:52<2:03:01,  1.94s/it][gradnorm] step 15279: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 664.3

 80%|████████  | 15280/19090 [7:27:54<1:50:53,  1.75s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 951.5503540039062, 'learning_rate': 5.828335909012769e-06, 'recon_loss': 0.0017366324260365217, 'recon_static': 0.0007624447753187269, 'recon_motion': 0.0009741876318003051, 'recon_norm': 1.3714295476675034, 'recon_norm_std': 0.03827938102185726, 'target_norm': 1.620059186220169, 'recon_cos': 0.5985773131251335, 'gradnorm/compressor.query': 0.41701493212382085, 'gradnorm/compressor.layer0': 16.20139402820113, 'gradnorm/compressor.layer1': 941.3571984108902, 'gradnorm/compressor.layer2': 133.58150163149463, 'gradnorm/compressor.layer3': 33.03910037092887, 'gradnorm/compressor.layer4': 7.3482610604681815, 'gradnorm/compressor.layer5': 4.127243125681045, 'gradnorm/compressor.layer6': 2.5313290242233886, 'gradnorm/compressor.layer7': 2.719936068288402, 'gradnorm/decoder.layer0': 2.7988081916051333, 'gradnorm/decoder.layer1': 1.7048383721179485, 'gradnorm/decoder.layer2': 0.9199154604548453, 'gradnorm/max_param': 664.2788696289062, 'epoch': 8.0}

 80%|████████  | 15280/19090 [7:27:54<1:50:53,  1.75s/it]
 80%|████████  | 15281/19090 [7:27:56<2:05:35,  1.98s/it]
 80%|████████  | 15282/19090 [7:27:57<1:52:16,  1.77s/it]
 80%|████████  | 15283/19090 [7:27:59<1:43:00,  1.62s/it]
 80%|████████  | 15284/19090 [7:28:00<1:46:11,  1.67s/it]
 80%|████████  | 15285/19090 [7:28:02<1:38:49,  1.56s/it]
 80%|████████  | 15286/19090 [7:28:03<1:33:44,  1.48s/it]
 80%|████████  | 15287/19090 [7:28:05<1:48:23,  1.71s/it]
 80%|████████  | 15288/19090 [7:28:07<1:40:13,  1.58s/it]
 80%|████████  | 15289/19090 [7:28:08<1:34:43,  1.50s/it][gradnorm] step 15289: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 616.0

 80%|████████  | 15290/19090 [7:28:10<1:52:00,  1.77s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 882.5211791992188, 'learning_rate': 5.799028955761293e-06, 'recon_loss': 0.0016835619229823352, 'recon_static': 0.0007152512596803717, 'recon_motion': 0.0009683106356533245, 'recon_norm': 1.3532233446836472, 'recon_norm_std': 0.03854089817032218, 'target_norm': 1.6170815140008927, 'recon_cos': 0.6081379145383835, 'gradnorm/compressor.query': 0.4068183224839083, 'gradnorm/compressor.layer0': 14.54361190828699, 'gradnorm/compressor.layer1': 872.8768527117671, 'gradnorm/compressor.layer2': 124.14921009489952, 'gradnorm/compressor.layer3': 34.3935283872108, 'gradnorm/compressor.layer4': 8.48161301872551, 'gradnorm/compressor.layer5': 4.564904820263661, 'gradnorm/compressor.layer6': 2.7324725699204064, 'gradnorm/compressor.layer7': 2.956058236287926, 'gradnorm/decoder.layer0': 2.9113409920636766, 'gradnorm/decoder.layer1': 1.8285355718217795, 'gradnorm/decoder.layer2': 1.078662848725967, 'gradnorm/max_param': 616.0269775390625, 'epoch': 8.01}

 80%|████████  | 15290/19090 [7:28:10<1:52:00,  1.77s/it]
 80%|████████  | 15291/19090 [7:28:12<1:42:45,  1.62s/it]
 80%|████████  | 15292/19090 [7:28:13<1:36:30,  1.52s/it]
 80%|████████  | 15293/19090 [7:28:15<1:48:08,  1.71s/it]
 80%|████████  | 15294/19090 [7:28:16<1:40:00,  1.58s/it]
 80%|████████  | 15295/19090 [7:28:18<1:34:20,  1.49s/it]
 80%|████████  | 15296/19090 [7:28:19<1:39:35,  1.58s/it]
 80%|████████  | 15297/19090 [7:28:21<1:34:03,  1.49s/it]
 80%|████████  | 15298/19090 [7:28:22<1:30:12,  1.43s/it]
 80%|████████  | 15299/19090 [7:28:24<1:45:05,  1.66s/it][gradnorm] step 15299: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 526.5

 80%|████████  | 15300/19090 [7:28:25<1:37:54,  1.55s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 756.7734985351562, 'learning_rate': 5.769786201166588e-06, 'recon_loss': 0.001671550763421692, 'recon_static': 0.0006980100937653333, 'recon_motion': 0.0009735406652907841, 'recon_norm': 1.3423657715320587, 'recon_norm_std': 0.03902143621817231, 'target_norm': 1.6219589203596114, 'recon_cos': 0.6113962560892106, 'gradnorm/compressor.query': 0.32344415353759176, 'gradnorm/compressor.layer0': 10.943365475245935, 'gradnorm/compressor.layer1': 745.9732736499344, 'gradnorm/compressor.layer2': 124.135109419636, 'gradnorm/compressor.layer3': 24.463603485746177, 'gradnorm/compressor.layer4': 7.4208261491618295, 'gradnorm/compressor.layer5': 4.642600721048292, 'gradnorm/compressor.layer6': 2.525026464795706, 'gradnorm/compressor.layer7': 3.0092658953036526, 'gradnorm/decoder.layer0': 2.495317556319969, 'gradnorm/decoder.layer1': 1.6836159122716563, 'gradnorm/decoder.layer2': 0.9884349417012648, 'gradnorm/max_param': 526.4658203125, 'epoch': 8.01}

 80%|████████  | 15300/19090 [7:28:25<1:37:54,  1.55s/it]
 80%|████████  | 15301/19090 [7:28:27<1:32:51,  1.47s/it]
 80%|████████  | 15302/19090 [7:28:29<1:47:00,  1.70s/it]
 80%|████████  | 15303/19090 [7:28:30<1:41:38,  1.61s/it]
 80%|████████  | 15304/19090 [7:28:32<1:35:47,  1.52s/it]
 80%|████████  | 15305/19090 [7:28:34<1:54:30,  1.82s/it]
 80%|████████  | 15306/19090 [7:28:35<1:45:22,  1.67s/it]
 80%|████████  | 15307/19090 [7:28:37<1:37:58,  1.55s/it]
 80%|████████  | 15308/19090 [7:28:39<1:55:04,  1.83s/it]
 80%|████████  | 15309/19090 [7:28:40<1:44:49,  1.66s/it][gradnorm] step 15309: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 553.6

 80%|████████  | 15310/19090 [7:28:42<1:37:40,  1.55s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 795.5791625976562, 'learning_rate': 5.740607743002119e-06, 'recon_loss': 0.0015976812283042818, 'recon_static': 0.000622207454580348, 'recon_motion': 0.0009754738290212117, 'recon_norm': 1.3099331319332124, 'recon_norm_std': 0.039929352980107066, 'target_norm': 1.620256531238556, 'recon_cos': 0.6257929086685181, 'gradnorm/compressor.query': 0.2704181729192962, 'gradnorm/compressor.layer0': 10.937583007497517, 'gradnorm/compressor.layer1': 784.5257723618772, 'gradnorm/compressor.layer2': 128.69914044002005, 'gradnorm/compressor.layer3': 26.164093992590782, 'gradnorm/compressor.layer4': 7.713172534573287, 'gradnorm/compressor.layer5': 4.182271169546532, 'gradnorm/compressor.layer6': 2.37237664950194, 'gradnorm/compressor.layer7': 2.49866287713118, 'gradnorm/decoder.layer0': 2.3430915416201965, 'gradnorm/decoder.layer1': 1.5959567331888467, 'gradnorm/decoder.layer2': 0.9437117777277921, 'gradnorm/max_param': 553.599365234375, 'epoch': 8.02}

 80%|████████  | 15310/19090 [7:28:42<1:37:40,  1.55s/it]
 80%|████████  | 15311/19090 [7:28:44<1:48:09,  1.72s/it]
 80%|████████  | 15312/19090 [7:28:45<1:39:58,  1.59s/it]
 80%|████████  | 15313/19090 [7:28:46<1:34:26,  1.50s/it]
 80%|████████  | 15314/19090 [7:28:48<1:44:01,  1.65s/it]
 80%|████████  | 15315/19090 [7:28:50<1:37:05,  1.54s/it]
 80%|████████  | 15316/19090 [7:28:51<1:32:15,  1.47s/it]
 80%|████████  | 15317/19090 [7:28:53<1:40:46,  1.60s/it]
 80%|████████  | 15318/19090 [7:28:54<1:34:44,  1.51s/it]
 80%|████████  | 15319/19090 [7:28:56<1:30:33,  1.44s/it][gradnorm] step 15319: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 829.6

 80%|████████  | 15320/19090 [7:28:57<1:38:57,  1.57s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1190.921630859375, 'learning_rate': 5.711493678826385e-06, 'recon_loss': 0.0016015491943107917, 'recon_static': 0.0006368461879901588, 'recon_motion': 0.0009647029859479517, 'recon_norm': 1.321460375189781, 'recon_norm_std': 0.03959324155002832, 'target_norm': 1.6198669075965881, 'recon_cos': 0.6260058626532554, 'gradnorm/compressor.query': 0.5969001125625093, 'gradnorm/compressor.layer0': 21.402158738819796, 'gradnorm/compressor.layer1': 1175.6208466167993, 'gradnorm/compressor.layer2': 186.02242808328248, 'gradnorm/compressor.layer3': 31.890840311926265, 'gradnorm/compressor.layer4': 8.447286623890085, 'gradnorm/compressor.layer5': 5.198297500674387, 'gradnorm/compressor.layer6': 2.9770106981909814, 'gradnorm/compressor.layer7': 3.2860947945423398, 'gradnorm/decoder.layer0': 2.8535002512847325, 'gradnorm/decoder.layer1': 1.7618979893619273, 'gradnorm/decoder.layer2': 0.9405015579283738, 'gradnorm/max_param': 829.627685546875, 'epoch': 8.03}

 80%|████████  | 15320/19090 [7:28:57<1:38:57,  1.57s/it]
 80%|████████  | 15321/19090 [7:28:59<1:33:43,  1.49s/it]
 80%|████████  | 15322/19090 [7:29:00<1:30:04,  1.43s/it]
 80%|████████  | 15323/19090 [7:29:02<1:39:58,  1.59s/it]
 80%|████████  | 15324/19090 [7:29:03<1:34:16,  1.50s/it]
 80%|████████  | 15325/19090 [7:29:05<1:30:20,  1.44s/it]
 80%|████████  | 15326/19090 [7:29:06<1:39:13,  1.58s/it]
 80%|████████  | 15327/19090 [7:29:08<1:33:48,  1.50s/it]
 80%|████████  | 15328/19090 [7:29:09<1:30:22,  1.44s/it]
 80%|████████  | 15329/19090 [7:29:11<1:38:52,  1.58s/it][gradnorm] step 15329: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 524.0

 80%|████████  | 15330/19090 [7:29:12<1:35:39,  1.53s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 750.7282104492188, 'learning_rate': 5.682444105982576e-06, 'recon_loss': 0.0015758755151182413, 'recon_static': 0.0005987438737065532, 'recon_motion': 0.000977131682157051, 'recon_norm': 1.2990545213222504, 'recon_norm_std': 0.039588537719100717, 'target_norm': 1.6207640051841736, 'recon_cos': 0.6304905757308006, 'gradnorm/compressor.query': 0.3732372616972332, 'gradnorm/compressor.layer0': 13.11341739205468, 'gradnorm/compressor.layer1': 742.4305523077562, 'gradnorm/compressor.layer2': 105.04561416207486, 'gradnorm/compressor.layer3': 32.610965911120296, 'gradnorm/compressor.layer4': 8.188476038405618, 'gradnorm/compressor.layer5': 4.984899415557704, 'gradnorm/compressor.layer6': 2.5587534614394216, 'gradnorm/compressor.layer7': 3.0682316297604237, 'gradnorm/decoder.layer0': 2.7543291348905425, 'gradnorm/decoder.layer1': 1.7936418055425576, 'gradnorm/decoder.layer2': 0.9469730200564714, 'gradnorm/max_param': 523.98681640625, 'epoch': 8.03}

 80%|████████  | 15330/19090 [7:29:12<1:35:39,  1.53s/it]
 80%|████████  | 15331/19090 [7:29:14<1:31:15,  1.46s/it]
 80%|████████  | 15332/19090 [7:29:16<1:39:32,  1.59s/it]
 80%|████████  | 15333/19090 [7:29:17<1:33:53,  1.50s/it]
 80%|████████  | 15334/19090 [7:29:18<1:34:02,  1.50s/it]
 80%|████████  | 15335/19090 [7:29:21<1:53:58,  1.82s/it]
 80%|████████  | 15336/19090 [7:29:22<1:44:04,  1.66s/it]
 80%|████████  | 15337/19090 [7:29:24<1:37:03,  1.55s/it]
 80%|████████  | 15338/19090 [7:29:26<1:51:16,  1.78s/it]
 80%|████████  | 15339/19090 [7:29:27<1:42:05,  1.63s/it][gradnorm] step 15339: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 589.7

 80%|████████  | 15340/19090 [7:29:28<1:35:44,  1.53s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 842.7640380859375, 'learning_rate': 5.653459121598259e-06, 'recon_loss': 0.0016133334604091943, 'recon_static': 0.0006461941782617941, 'recon_motion': 0.0009671392777818255, 'recon_norm': 1.3171724170446395, 'recon_norm_std': 0.03957528825849295, 'target_norm': 1.6196091532707215, 'recon_cos': 0.6220195770263672, 'gradnorm/compressor.query': 0.34528770918772644, 'gradnorm/compressor.layer0': 13.138597982966363, 'gradnorm/compressor.layer1': 835.5610016629546, 'gradnorm/compressor.layer2': 104.46606712080005, 'gradnorm/compressor.layer3': 29.954431130234095, 'gradnorm/compressor.layer4': 7.854130050884131, 'gradnorm/compressor.layer5': 4.31792319038917, 'gradnorm/compressor.layer6': 2.373353487779286, 'gradnorm/compressor.layer7': 2.8316882537444057, 'gradnorm/decoder.layer0': 2.8074527805124574, 'gradnorm/decoder.layer1': 1.7405005528669129, 'gradnorm/decoder.layer2': 0.9922290703182503, 'gradnorm/max_param': 589.6600341796875, 'epoch': 8.04}

 80%|████████  | 15340/19090 [7:29:28<1:35:44,  1.53s/it]
 80%|████████  | 15341/19090 [7:29:31<1:52:01,  1.79s/it]
 80%|████████  | 15342/19090 [7:29:32<1:42:46,  1.65s/it]
 80%|████████  | 15343/19090 [7:29:33<1:36:08,  1.54s/it]
 80%|████████  | 15344/19090 [7:29:36<1:50:34,  1.77s/it]
 80%|████████  | 15345/19090 [7:29:37<1:41:31,  1.63s/it]
 80%|████████  | 15346/19090 [7:29:39<1:40:22,  1.61s/it]
 80%|████████  | 15347/19090 [7:29:41<1:50:11,  1.77s/it]
 80%|████████  | 15348/19090 [7:29:42<1:41:14,  1.62s/it]
 80%|████████  | 15349/19090 [7:29:43<1:37:02,  1.56s/it][gradnorm] step 15349: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 668.9

 80%|████████  | 15350/19090 [7:29:46<1:52:20,  1.80s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 955.706787109375, 'learning_rate': 5.624538822585043e-06, 'recon_loss': 0.0016543525940505787, 'recon_static': 0.0006707785127218812, 'recon_motion': 0.0009835740303969942, 'recon_norm': 1.3260134547948836, 'recon_norm_std': 0.040200062282383445, 'target_norm': 1.6246761202812194, 'recon_cos': 0.6147895634174347, 'gradnorm/compressor.query': 0.4371177842982109, 'gradnorm/compressor.layer0': 14.770466716110183, 'gradnorm/compressor.layer1': 947.7758503464836, 'gradnorm/compressor.layer2': 116.20457361776901, 'gradnorm/compressor.layer3': 35.484192673463774, 'gradnorm/compressor.layer4': 8.043742753553943, 'gradnorm/compressor.layer5': 4.694113173574762, 'gradnorm/compressor.layer6': 2.582425316822884, 'gradnorm/compressor.layer7': 2.932838637368131, 'gradnorm/decoder.layer0': 3.0320655276280832, 'gradnorm/decoder.layer1': 1.7406961532651077, 'gradnorm/decoder.layer2': 1.0296618051752742, 'gradnorm/max_param': 668.9116821289062, 'epoch': 8.04}

 80%|████████  | 15350/19090 [7:29:46<1:52:20,  1.80s/it]
 80%|████████  | 15351/19090 [7:29:47<1:44:27,  1.68s/it]
 80%|████████  | 15352/19090 [7:29:48<1:37:09,  1.56s/it]
 80%|████████  | 15353/19090 [7:29:51<1:55:20,  1.85s/it]
 80%|████████  | 15354/19090 [7:29:52<1:44:45,  1.68s/it]
 80%|████████  | 15355/19090 [7:29:54<1:37:17,  1.56s/it]
 80%|████████  | 15356/19090 [7:29:56<1:46:48,  1.72s/it]
 80%|████████  | 15357/19090 [7:29:57<1:45:07,  1.69s/it]
 80%|████████  | 15358/19090 [7:29:59<1:40:57,  1.62s/it]
 80%|████████  | 15359/19090 [7:30:01<1:49:38,  1.76s/it][gradnorm] step 15359: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 558.7

 80%|████████  | 15360/19090 [7:30:02<1:40:44,  1.62s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 798.86328125, 'learning_rate': 5.595683305638263e-06, 'recon_loss': 0.001739344169618562, 'recon_static': 0.0007814645214239135, 'recon_motion': 0.0009578796350979246, 'recon_norm': 1.3740883231163026, 'recon_norm_std': 0.03883608290925622, 'target_norm': 1.6183853626251221, 'recon_cos': 0.5982682764530182, 'gradnorm/compressor.query': 0.36104968915956853, 'gradnorm/compressor.layer0': 13.661703247403567, 'gradnorm/compressor.layer1': 791.7388300443524, 'gradnorm/compressor.layer2': 100.44595386792591, 'gradnorm/compressor.layer3': 30.86747322539233, 'gradnorm/compressor.layer4': 7.403284179212871, 'gradnorm/compressor.layer5': 4.4888880546073855, 'gradnorm/compressor.layer6': 2.7949382533587195, 'gradnorm/compressor.layer7': 2.947884405563911, 'gradnorm/decoder.layer0': 2.8406190509125344, 'gradnorm/decoder.layer1': 1.6698812557879696, 'gradnorm/decoder.layer2': 1.010159797396704, 'gradnorm/max_param': 558.7081298828125, 'epoch': 8.05}

 80%|████████  | 15360/19090 [7:30:02<1:40:44,  1.62s/it]
 80%|████████  | 15361/19090 [7:30:04<1:45:23,  1.70s/it]
 80%|████████  | 15362/19090 [7:30:06<1:42:19,  1.65s/it]
 80%|████████  | 15363/19090 [7:30:07<1:36:47,  1.56s/it]
 80%|████████  | 15364/19090 [7:30:10<1:58:27,  1.91s/it]
 80%|████████  | 15365/19090 [7:30:11<1:46:51,  1.72s/it]
 80%|████████  | 15366/19090 [7:30:12<1:41:08,  1.63s/it]
 80%|████████  | 15367/19090 [7:30:15<1:54:02,  1.84s/it]
 81%|████████  | 15368/19090 [7:30:16<1:43:52,  1.67s/it]
 81%|████████  | 15369/19090 [7:30:17<1:40:31,  1.62s/it][gradnorm] step 15369: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 451.8

 81%|████████  | 15370/19090 [7:30:20<1:53:03,  1.82s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 649.403076171875, 'learning_rate': 5.566892667236662e-06, 'recon_loss': 0.0018233287322800607, 'recon_static': 0.0008495718589983881, 'recon_motion': 0.0009737569082062692, 'recon_norm': 1.4056357324123383, 'recon_norm_std': 0.038115890324115755, 'target_norm': 1.6200874358415605, 'recon_cos': 0.5844131410121918, 'gradnorm/compressor.query': 0.4214762109738281, 'gradnorm/compressor.layer0': 11.518935826210656, 'gradnorm/compressor.layer1': 640.124328744034, 'gradnorm/compressor.layer2': 105.42798749324109, 'gradnorm/compressor.layer3': 25.131378426990928, 'gradnorm/compressor.layer4': 6.542642249316331, 'gradnorm/compressor.layer5': 4.183476807057788, 'gradnorm/compressor.layer6': 2.5253974224096454, 'gradnorm/compressor.layer7': 2.839131219984302, 'gradnorm/decoder.layer0': 2.670975740140473, 'gradnorm/decoder.layer1': 1.6898065980027313, 'gradnorm/decoder.layer2': 0.9637275138537396, 'gradnorm/max_param': 451.7605285644531, 'epoch': 8.05}

 81%|████████  | 15370/19090 [7:30:20<1:53:03,  1.82s/it]
 81%|████████  | 15371/19090 [7:30:21<1:43:04,  1.66s/it]
 81%|████████  | 15372/19090 [7:30:23<1:45:27,  1.70s/it]
 81%|████████  | 15373/19090 [7:30:24<1:45:26,  1.70s/it]
 81%|████████  | 15374/19090 [7:30:26<1:37:49,  1.58s/it]
 81%|████████  | 15375/19090 [7:30:28<1:44:21,  1.69s/it]
 81%|████████  | 15376/19090 [7:30:30<1:54:54,  1.86s/it]
 81%|████████  | 15377/19090 [7:30:31<1:44:25,  1.69s/it]
 81%|████████  | 15378/19090 [7:30:33<1:37:02,  1.57s/it]
 81%|████████  | 15379/19090 [7:30:35<1:54:04,  1.84s/it][gradnorm] step 15379: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 441.3

 81%|████████  | 15380/19090 [7:30:36<1:43:46,  1.68s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 634.9232177734375, 'learning_rate': 5.538167003642047e-06, 'recon_loss': 0.0018221093487227336, 'recon_static': 0.0008480371296172961, 'recon_motion': 0.0009740722351125441, 'recon_norm': 1.4095336675643921, 'recon_norm_std': 0.037824783381074666, 'target_norm': 1.6218097120523454, 'recon_cos': 0.5848848193883895, 'gradnorm/compressor.query': 0.33113897145303545, 'gradnorm/compressor.layer0': 11.163073592704333, 'gradnorm/compressor.layer1': 625.3571729419916, 'gradnorm/compressor.layer2': 104.04468222084923, 'gradnorm/compressor.layer3': 31.69537412182819, 'gradnorm/compressor.layer4': 7.253351686445092, 'gradnorm/compressor.layer5': 4.305707364157685, 'gradnorm/compressor.layer6': 2.860652617308089, 'gradnorm/compressor.layer7': 2.8460198031800084, 'gradnorm/decoder.layer0': 3.092757711319348, 'gradnorm/decoder.layer1': 1.7341544896067485, 'gradnorm/decoder.layer2': 1.0822545451216186, 'gradnorm/max_param': 441.3167724609375, 'epoch': 8.06}

 81%|████████  | 15380/19090 [7:30:36<1:43:46,  1.68s/it]
 81%|████████  | 15381/19090 [7:30:38<1:36:36,  1.56s/it]
 81%|████████  | 15382/19090 [7:30:40<1:55:06,  1.86s/it]
 81%|████████  | 15383/19090 [7:30:41<1:44:27,  1.69s/it]
 81%|████████  | 15384/19090 [7:30:43<1:37:08,  1.57s/it]
 81%|████████  | 15385/19090 [7:30:45<1:55:56,  1.88s/it]
 81%|████████  | 15386/19090 [7:30:47<1:45:05,  1.70s/it]
 81%|████████  | 15387/19090 [7:30:48<1:37:30,  1.58s/it]
 81%|████████  | 15388/19090 [7:30:50<1:52:55,  1.83s/it]
 81%|████████  | 15389/19090 [7:30:52<1:42:49,  1.67s/it][gradnorm] step 15389: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 588.2

 81%|████████  | 15390/19090 [7:30:53<1:35:52,  1.55s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 842.8472900390625, 'learning_rate': 5.5095064108989956e-06, 'recon_loss': 0.001754387075197883, 'recon_static': 0.0007870144094340503, 'recon_motion': 0.0009673726861365139, 'recon_norm': 1.3693301200866699, 'recon_norm_std': 0.03846509158611298, 'target_norm': 1.618600457906723, 'recon_cos': 0.5930240690708161, 'gradnorm/compressor.query': 0.31543872217992563, 'gradnorm/compressor.layer0': 11.117570432178656, 'gradnorm/compressor.layer1': 833.4330357023292, 'gradnorm/compressor.layer2': 120.13689294665066, 'gradnorm/compressor.layer3': 33.38272016866473, 'gradnorm/compressor.layer4': 7.8814060858230395, 'gradnorm/compressor.layer5': 4.563215225805319, 'gradnorm/compressor.layer6': 2.8151061804685114, 'gradnorm/compressor.layer7': 2.648022284353479, 'gradnorm/decoder.layer0': 2.8966675981433045, 'gradnorm/decoder.layer1': 1.6637086447284175, 'gradnorm/decoder.layer2': 0.9334821434826313, 'gradnorm/max_param': 588.1558837890625, 'epoch': 8.06}

 81%|████████  | 15390/19090 [7:30:53<1:35:52,  1.55s/it]
 81%|████████  | 15391/19090 [7:30:55<1:52:57,  1.83s/it]
 81%|████████  | 15392/19090 [7:30:57<1:42:50,  1.67s/it]
 81%|████████  | 15393/19090 [7:30:58<1:35:49,  1.56s/it]
 81%|████████  | 15394/19090 [7:31:00<1:45:40,  1.72s/it]
 81%|████████  | 15395/19090 [7:31:02<1:41:27,  1.65s/it]
 81%|████████  | 15396/19090 [7:31:03<1:34:53,  1.54s/it]
 81%|████████  | 15397/19090 [7:31:05<1:52:39,  1.83s/it]
 81%|████████  | 15398/19090 [7:31:07<1:48:55,  1.77s/it]
 81%|████████  | 15399/19090 [7:31:08<1:40:34,  1.63s/it][gradnorm] step 15399: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 656.8

 81%|████████  | 15400/19090 [7:31:10<1:46:43,  1.74s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 940.5457763671875, 'learning_rate': 5.480910984834506e-06, 'recon_loss': 0.0016017791524063796, 'recon_static': 0.0006166763312648981, 'recon_motion': 0.0009851028327830136, 'recon_norm': 1.315729370713234, 'recon_norm_std': 0.03893789751455188, 'target_norm': 1.6213672190904618, 'recon_cos': 0.6260450527071952, 'gradnorm/compressor.query': 0.4550572992233803, 'gradnorm/compressor.layer0': 20.033004993322894, 'gradnorm/compressor.layer1': 930.7730322156955, 'gradnorm/compressor.layer2': 130.43207194024404, 'gradnorm/compressor.layer3': 27.718309607550026, 'gradnorm/compressor.layer4': 8.163940078328544, 'gradnorm/compressor.layer5': 3.8942654288434992, 'gradnorm/compressor.layer6': 2.5627794811148386, 'gradnorm/compressor.layer7': 2.564136912714923, 'gradnorm/decoder.layer0': 2.549798983710348, 'gradnorm/decoder.layer1': 1.7649631441282918, 'gradnorm/decoder.layer2': 0.9674903925337349, 'gradnorm/max_param': 656.8497924804688, 'epoch': 8.07}

 81%|████████  | 15400/19090 [7:31:10<1:46:43,  1.74s/it]
 81%|████████  | 15401/19090 [7:31:12<1:47:29,  1.75s/it]
 81%|████████  | 15402/19090 [7:31:13<1:39:08,  1.61s/it]
 81%|████████  | 15403/19090 [7:31:16<1:50:24,  1.80s/it]
 81%|████████  | 15404/19090 [7:31:17<1:41:07,  1.65s/it]
 81%|████████  | 15405/19090 [7:31:18<1:34:24,  1.54s/it]
 81%|████████  | 15406/19090 [7:31:21<1:53:45,  1.85s/it]
 81%|████████  | 15407/19090 [7:31:22<1:43:20,  1.68s/it]
 81%|████████  | 15408/19090 [7:31:23<1:36:06,  1.57s/it]
 81%|████████  | 15409/19090 [7:31:26<1:52:28,  1.83s/it][gradnorm] step 15409: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 456.7

 81%|████████  | 15410/19090 [7:31:27<1:47:02,  1.75s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 654.3837280273438, 'learning_rate': 5.4523808210577036e-06, 'recon_loss': 0.001603056612657383, 'recon_static': 0.0006300624474533834, 'recon_motion': 0.0009729942001285962, 'recon_norm': 1.3158772051334382, 'recon_norm_std': 0.03876652242615819, 'target_norm': 1.619106525182724, 'recon_cos': 0.6244682937860488, 'gradnorm/compressor.query': 0.4129786430960721, 'gradnorm/compressor.layer0': 13.212187163692372, 'gradnorm/compressor.layer1': 647.1969850825944, 'gradnorm/compressor.layer2': 91.45822240407092, 'gradnorm/compressor.layer3': 26.62036092397427, 'gradnorm/compressor.layer4': 8.262505041631641, 'gradnorm/compressor.layer5': 3.8197323099182743, 'gradnorm/compressor.layer6': 2.5546400899891717, 'gradnorm/compressor.layer7': 2.548870498480392, 'gradnorm/decoder.layer0': 2.533457135220518, 'gradnorm/decoder.layer1': 1.669326051303613, 'gradnorm/decoder.layer2': 1.032520725051477, 'gradnorm/max_param': 456.73040771484375, 'epoch': 8.07}

 81%|████████  | 15410/19090 [7:31:27<1:47:02,  1.75s/it]
 81%|████████  | 15411/19090 [7:31:29<1:38:40,  1.61s/it]
 81%|████████  | 15412/19090 [7:31:31<1:47:03,  1.75s/it]
 81%|████████  | 15413/19090 [7:31:33<1:48:31,  1.77s/it]
 81%|████████  | 15414/19090 [7:31:34<1:39:44,  1.63s/it]
 81%|████████  | 15415/19090 [7:31:35<1:40:32,  1.64s/it]
 81%|████████  | 15416/19090 [7:31:37<1:42:57,  1.68s/it]
 81%|████████  | 15417/19090 [7:31:39<1:35:43,  1.56s/it]
 81%|████████  | 15418/19090 [7:31:41<1:49:01,  1.78s/it]
 81%|████████  | 15419/19090 [7:31:43<1:47:24,  1.76s/it][gradnorm] step 15419: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 599.3

 81%|████████  | 15420/19090 [7:31:44<1:39:00,  1.62s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 859.531494140625, 'learning_rate': 5.423916014959502e-06, 'recon_loss': 0.001653771082055755, 'recon_static': 0.0006727952772052958, 'recon_motion': 0.0009809757902985439, 'recon_norm': 1.33359836935997, 'recon_norm_std': 0.038708473555743694, 'target_norm': 1.621928483247757, 'recon_cos': 0.6146587044000625, 'gradnorm/compressor.query': 0.4893176561864256, 'gradnorm/compressor.layer0': 18.289260230936677, 'gradnorm/compressor.layer1': 849.1415150020637, 'gradnorm/compressor.layer2': 127.80872549438868, 'gradnorm/compressor.layer3': 31.029836755016433, 'gradnorm/compressor.layer4': 8.67646143143571, 'gradnorm/compressor.layer5': 4.11170643857707, 'gradnorm/compressor.layer6': 2.8311677023945494, 'gradnorm/compressor.layer7': 2.7659480019437797, 'gradnorm/decoder.layer0': 2.983045831249572, 'gradnorm/decoder.layer1': 1.668054011218291, 'gradnorm/decoder.layer2': 0.9864432692594414, 'gradnorm/max_param': 599.2579345703125, 'epoch': 8.08}

 81%|████████  | 15420/19090 [7:31:44<1:39:00,  1.62s/it]
 81%|████████  | 15421/19090 [7:31:46<1:53:28,  1.86s/it]
 81%|████████  | 15422/19090 [7:31:48<1:47:37,  1.76s/it]
 81%|████████  | 15423/19090 [7:31:49<1:38:59,  1.62s/it]
 81%|████████  | 15424/19090 [7:31:51<1:51:01,  1.82s/it]
 81%|████████  | 15425/19090 [7:31:53<1:42:16,  1.67s/it]
 81%|████████  | 15426/19090 [7:31:54<1:35:15,  1.56s/it]
 81%|████████  | 15427/19090 [7:31:56<1:42:57,  1.69s/it]
 81%|████████  | 15428/19090 [7:31:58<1:44:54,  1.72s/it]
 81%|████████  | 15429/19090 [7:31:59<1:37:05,  1.59s/it][gradnorm] step 15429: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 853.4

 81%|████████  | 15430/19090 [7:32:01<1:40:04,  1.64s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1219.7799072265625, 'learning_rate': 5.395516661712291e-06, 'recon_loss': 0.001573270553490147, 'recon_static': 0.0005966290991636924, 'recon_motion': 0.0009766414339537732, 'recon_norm': 1.293725025653839, 'recon_norm_std': 0.038988866843283176, 'target_norm': 1.6215147972106934, 'recon_cos': 0.6304624125361442, 'gradnorm/compressor.query': 0.9447100903090591, 'gradnorm/compressor.layer0': 36.52688068663012, 'gradnorm/compressor.layer1': 1209.8120205221967, 'gradnorm/compressor.layer2': 147.8807519485717, 'gradnorm/compressor.layer3': 30.093408251888714, 'gradnorm/compressor.layer4': 8.081658835517088, 'gradnorm/compressor.layer5': 3.8632792751511635, 'gradnorm/compressor.layer6': 2.7470466186933664, 'gradnorm/compressor.layer7': 2.7603483563494082, 'gradnorm/decoder.layer0': 3.002344328110714, 'gradnorm/decoder.layer1': 1.7560639903980728, 'gradnorm/decoder.layer2': 1.0379135548532077, 'gradnorm/max_param': 853.3727416992188, 'epoch': 8.08}

 81%|████████  | 15430/19090 [7:32:01<1:40:04,  1.64s/it]
 81%|████████  | 15431/19090 [7:32:03<1:48:22,  1.78s/it]
 81%|████████  | 15432/19090 [7:32:04<1:39:32,  1.63s/it]
 81%|████████  | 15433/19090 [7:32:06<1:39:27,  1.63s/it]
 81%|████████  | 15434/19090 [7:32:08<1:46:36,  1.75s/it]
 81%|████████  | 15435/19090 [7:32:09<1:38:19,  1.61s/it]
 81%|████████  | 15436/19090 [7:32:11<1:50:08,  1.81s/it]
 81%|████████  | 15437/19090 [7:32:13<1:47:45,  1.77s/it]
 81%|████████  | 15438/19090 [7:32:14<1:39:00,  1.63s/it]
 81%|████████  | 15439/19090 [7:32:17<1:54:20,  1.88s/it][gradnorm] step 15439: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 527.2

 81%|████████  | 15440/19090 [7:32:18<1:46:39,  1.75s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 754.8953857421875, 'learning_rate': 5.3671828562696204e-06, 'recon_loss': 0.0015798820939380675, 'recon_static': 0.0006090851165936328, 'recon_motion': 0.0009707970151794143, 'recon_norm': 1.3088891685009003, 'recon_norm_std': 0.039363282080739734, 'target_norm': 1.6187284469604493, 'recon_cos': 0.6295585572719574, 'gradnorm/compressor.query': 0.37627418925067657, 'gradnorm/compressor.layer0': 12.272241447928328, 'gradnorm/compressor.layer1': 746.9963604329006, 'gradnorm/compressor.layer2': 100.30534923889901, 'gradnorm/compressor.layer3': 39.23720619016312, 'gradnorm/compressor.layer4': 7.981717438179546, 'gradnorm/compressor.layer5': 4.667300635372364, 'gradnorm/compressor.layer6': 2.5738914177436607, 'gradnorm/compressor.layer7': 2.833780103876097, 'gradnorm/decoder.layer0': 2.8056701808800324, 'gradnorm/decoder.layer1': 1.7527476266657995, 'gradnorm/decoder.layer2': 0.9962801590570084, 'gradnorm/max_param': 527.1976318359375, 'epoch': 8.09}

 81%|████████  | 15440/19090 [7:32:18<1:46:39,  1.75s/it]
 81%|████████  | 15441/19090 [7:32:20<1:38:14,  1.62s/it]
 81%|████████  | 15442/19090 [7:32:23<2:03:07,  2.02s/it]
 81%|████████  | 15443/19090 [7:32:24<1:49:42,  1.80s/it]
 81%|████████  | 15444/19090 [7:32:25<1:40:28,  1.65s/it]
 81%|████████  | 15445/19090 [7:32:27<1:51:53,  1.84s/it]
 81%|████████  | 15446/19090 [7:32:29<1:41:54,  1.68s/it]
 81%|████████  | 15447/19090 [7:32:30<1:34:53,  1.56s/it]
 81%|████████  | 15448/19090 [7:32:33<2:01:38,  2.00s/it]
 81%|████████  | 15449/19090 [7:32:34<1:48:39,  1.79s/it][gradnorm] step 15449: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 750.7

 81%|████████  | 15450/19090 [7:32:36<1:39:32,  1.64s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1074.525634765625, 'learning_rate': 5.338914693365868e-06, 'recon_loss': 0.0016179873957298696, 'recon_static': 0.0006328357441816479, 'recon_motion': 0.0009851516180788167, 'recon_norm': 1.316160473227501, 'recon_norm_std': 0.0386945073492825, 'target_norm': 1.6221391081809997, 'recon_cos': 0.6220066919922829, 'gradnorm/compressor.query': 0.3965210357403877, 'gradnorm/compressor.layer0': 15.545249110177068, 'gradnorm/compressor.layer1': 1063.7487367795084, 'gradnorm/compressor.layer2': 147.31830705567452, 'gradnorm/compressor.layer3': 31.0289722540181, 'gradnorm/compressor.layer4': 9.133859174678115, 'gradnorm/compressor.layer5': 4.685510034960946, 'gradnorm/compressor.layer6': 2.6513728526355225, 'gradnorm/compressor.layer7': 3.2301628955120774, 'gradnorm/decoder.layer0': 3.0023954682334, 'gradnorm/decoder.layer1': 1.8578813467912538, 'gradnorm/decoder.layer2': 1.1373178273299636, 'gradnorm/max_param': 750.73828125, 'epoch': 8.09}

 81%|████████  | 15450/19090 [7:32:36<1:39:32,  1.64s/it]
 81%|████████  | 15451/19090 [7:32:38<1:53:55,  1.88s/it]
 81%|████████  | 15452/19090 [7:32:39<1:43:17,  1.70s/it]
 81%|████████  | 15453/19090 [7:32:41<1:35:55,  1.58s/it]
 81%|████████  | 15454/19090 [7:32:43<1:57:51,  1.94s/it]
 81%|████████  | 15455/19090 [7:32:45<1:45:59,  1.75s/it]
 81%|████████  | 15456/19090 [7:32:46<1:37:44,  1.61s/it]
 81%|████████  | 15457/19090 [7:32:49<1:55:53,  1.91s/it]
 81%|████████  | 15458/19090 [7:32:50<1:44:41,  1.73s/it]
 81%|████████  | 15459/19090 [7:32:51<1:36:46,  1.60s/it][gradnorm] step 15459: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 806.8

 81%|████████  | 15460/19090 [7:32:54<1:51:19,  1.84s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1153.9193115234375, 'learning_rate': 5.310712267515971e-06, 'recon_loss': 0.0015902138780802487, 'recon_static': 0.0006063114313292317, 'recon_motion': 0.0009839024263783359, 'recon_norm': 1.29642294049263, 'recon_norm_std': 0.03945710780099034, 'target_norm': 1.6211865305900575, 'recon_cos': 0.6260320484638214, 'gradnorm/compressor.query': 0.4205956841635565, 'gradnorm/compressor.layer0': 14.478811106901558, 'gradnorm/compressor.layer1': 1143.2852334391448, 'gradnorm/compressor.layer2': 150.97004695675562, 'gradnorm/compressor.layer3': 35.919514861027324, 'gradnorm/compressor.layer4': 9.130107965405498, 'gradnorm/compressor.layer5': 4.683682323951352, 'gradnorm/compressor.layer6': 2.767732065424727, 'gradnorm/compressor.layer7': 3.1508778090735925, 'gradnorm/decoder.layer0': 3.116465243462294, 'gradnorm/decoder.layer1': 1.7532639038761768, 'gradnorm/decoder.layer2': 1.100068777275045, 'gradnorm/max_param': 806.843017578125, 'epoch': 8.1}

 81%|████████  | 15460/19090 [7:32:54<1:51:19,  1.84s/it]
 81%|████████  | 15461/19090 [7:32:55<1:41:25,  1.68s/it]
 81%|████████  | 15462/19090 [7:32:56<1:34:23,  1.56s/it]
 81%|████████  | 15463/19090 [7:32:59<1:55:51,  1.92s/it]
 81%|████████  | 15464/19090 [7:33:00<1:44:26,  1.73s/it]
 81%|████████  | 15465/19090 [7:33:02<1:36:34,  1.60s/it]
 81%|████████  | 15466/19090 [7:33:05<2:00:38,  2.00s/it]
 81%|████████  | 15467/19090 [7:33:06<1:47:43,  1.78s/it]
 81%|████████  | 15468/19090 [7:33:07<1:38:39,  1.63s/it]
 81%|████████  | 15469/19090 [7:33:10<1:59:27,  1.98s/it][gradnorm] step 15469: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 616.2

 81%|████████  | 15470/19090 [7:33:11<1:46:52,  1.77s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 885.4615478515625, 'learning_rate': 5.282575673015025e-06, 'recon_loss': 0.0015591605595545843, 'recon_static': 0.0005819221987621859, 'recon_motion': 0.0009772383666131646, 'recon_norm': 1.2951608061790467, 'recon_norm_std': 0.04033373966813088, 'target_norm': 1.6205187350511552, 'recon_cos': 0.6343429267406464, 'gradnorm/compressor.query': 0.3373259638717486, 'gradnorm/compressor.layer0': 13.068137964469095, 'gradnorm/compressor.layer1': 873.0165734394737, 'gradnorm/compressor.layer2': 142.27326396981618, 'gradnorm/compressor.layer3': 36.73683231625738, 'gradnorm/compressor.layer4': 8.750442057750837, 'gradnorm/compressor.layer5': 4.218162891366404, 'gradnorm/compressor.layer6': 2.440064670456154, 'gradnorm/compressor.layer7': 3.0994726101581547, 'gradnorm/decoder.layer0': 2.8758641478454425, 'gradnorm/decoder.layer1': 1.721388992980677, 'gradnorm/decoder.layer2': 0.9292593539128698, 'gradnorm/max_param': 616.1585693359375, 'epoch': 8.1}

 81%|████████  | 15470/19090 [7:33:11<1:46:52,  1.77s/it]
 81%|████████  | 15471/19090 [7:33:12<1:38:01,  1.63s/it]
 81%|████████  | 15472/19090 [7:33:16<2:03:56,  2.06s/it]
 81%|████████  | 15473/19090 [7:33:17<1:50:04,  1.83s/it]
 81%|████████  | 15474/19090 [7:33:18<1:40:33,  1.67s/it]
 81%|████████  | 15475/19090 [7:33:20<1:53:19,  1.88s/it]
 81%|████████  | 15476/19090 [7:33:22<1:42:32,  1.70s/it]
 81%|████████  | 15477/19090 [7:33:23<1:35:05,  1.58s/it]
 81%|████████  | 15478/19090 [7:33:26<2:00:52,  2.01s/it]
 81%|████████  | 15479/19090 [7:33:27<1:47:59,  1.79s/it][gradnorm] step 15479: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 744.0

 81%|████████  | 15480/19090 [7:33:29<1:38:52,  1.64s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1062.9764404296875, 'learning_rate': 5.254505003938043e-06, 'recon_loss': 0.0015401055512484162, 'recon_static': 0.0005666701807058416, 'recon_motion': 0.000973435424384661, 'recon_norm': 1.2813008904457093, 'recon_norm_std': 0.04055239222943783, 'target_norm': 1.6204266905784608, 'recon_cos': 0.6376737937331199, 'gradnorm/compressor.query': 0.5394987469606458, 'gradnorm/compressor.layer0': 13.972880663734617, 'gradnorm/compressor.layer1': 1054.3423305549295, 'gradnorm/compressor.layer2': 130.15140672261288, 'gradnorm/compressor.layer3': 31.88192326773904, 'gradnorm/compressor.layer4': 8.63440976103096, 'gradnorm/compressor.layer5': 4.834924545567192, 'gradnorm/compressor.layer6': 2.801837716570383, 'gradnorm/compressor.layer7': 3.194471626350376, 'gradnorm/decoder.layer0': 3.157612723590817, 'gradnorm/decoder.layer1': 1.7628209626701865, 'gradnorm/decoder.layer2': 0.9221211444246166, 'gradnorm/max_param': 744.0094604492188, 'epoch': 8.11}

 81%|████████  | 15480/19090 [7:33:29<1:38:52,  1.64s/it]
 81%|████████  | 15481/19090 [7:33:31<1:55:04,  1.91s/it]
 81%|████████  | 15482/19090 [7:33:33<1:43:55,  1.73s/it]
 81%|████████  | 15483/19090 [7:33:34<1:36:01,  1.60s/it]
 81%|████████  | 15484/19090 [7:33:36<1:55:22,  1.92s/it]
 81%|████████  | 15485/19090 [7:33:38<1:44:16,  1.74s/it]
 81%|████████  | 15486/19090 [7:33:39<1:36:11,  1.60s/it]
 81%|████████  | 15487/19090 [7:33:42<1:58:00,  1.97s/it]
 81%|████████  | 15488/19090 [7:33:43<1:46:10,  1.77s/it]
 81%|████████  | 15489/19090 [7:33:44<1:37:32,  1.63s/it][gradnorm] step 15489: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 745.0

 81%|████████  | 15490/19090 [7:33:47<1:58:43,  1.98s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1067.9088134765625, 'learning_rate': 5.226500354139635e-06, 'recon_loss': 0.0015207940741674975, 'recon_static': 0.0005471634038258344, 'recon_motion': 0.0009736306630657055, 'recon_norm': 1.2822131484746933, 'recon_norm_std': 0.04034536294639111, 'target_norm': 1.618553677201271, 'recon_cos': 0.642233869433403, 'gradnorm/compressor.query': 0.39694407079799343, 'gradnorm/compressor.layer0': 14.329751624372662, 'gradnorm/compressor.layer1': 1055.7265862902193, 'gradnorm/compressor.layer2': 155.45379010939232, 'gradnorm/compressor.layer3': 36.958360490642036, 'gradnorm/compressor.layer4': 8.56684486807438, 'gradnorm/compressor.layer5': 5.390333438673171, 'gradnorm/compressor.layer6': 2.774856848933648, 'gradnorm/compressor.layer7': 3.320886898845228, 'gradnorm/decoder.layer0': 2.830817358379715, 'gradnorm/decoder.layer1': 1.7841337184485853, 'gradnorm/decoder.layer2': 1.0679243412007484, 'gradnorm/max_param': 745.0335693359375, 'epoch': 8.11}

 81%|████████  | 15490/19090 [7:33:47<1:58:43,  1.98s/it]
 81%|████████  | 15491/19090 [7:33:49<1:46:15,  1.77s/it]
 81%|████████  | 15492/19090 [7:33:50<1:37:36,  1.63s/it]
 81%|████████  | 15493/19090 [7:33:53<1:58:49,  1.98s/it]
 81%|████████  | 15494/19090 [7:33:54<1:46:27,  1.78s/it]
 81%|████████  | 15495/19090 [7:33:55<1:37:41,  1.63s/it]
 81%|████████  | 15496/19090 [7:33:58<1:56:58,  1.95s/it]
 81%|████████  | 15497/19090 [7:33:59<1:45:04,  1.75s/it]
 81%|████████  | 15498/19090 [7:34:01<1:36:50,  1.62s/it]
 81%|████████  | 15499/19090 [7:34:03<1:58:43,  1.98s/it][gradnorm] step 15499: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 474.1

 81%|████████  | 15500/19090 [7:34:05<1:46:20,  1.78s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 682.462158203125, 'learning_rate': 5.198561817253633e-06, 'recon_loss': 0.001554944831877947, 'recon_static': 0.0005785232380731032, 'recon_motion': 0.0009764215938048437, 'recon_norm': 1.2889501243829726, 'recon_norm_std': 0.04051136868074536, 'target_norm': 1.6209725528955459, 'recon_cos': 0.6347013249993324, 'gradnorm/compressor.query': 0.3899127761491551, 'gradnorm/compressor.layer0': 13.26772691589062, 'gradnorm/compressor.layer1': 671.8483366802362, 'gradnorm/compressor.layer2': 113.60962320320846, 'gradnorm/compressor.layer3': 33.91583605849744, 'gradnorm/compressor.layer4': 8.951062562405301, 'gradnorm/compressor.layer5': 5.621749277707629, 'gradnorm/compressor.layer6': 2.6788211233831913, 'gradnorm/compressor.layer7': 3.016508615333196, 'gradnorm/decoder.layer0': 2.79912148726415, 'gradnorm/decoder.layer1': 1.8918979237342473, 'gradnorm/decoder.layer2': 1.1896102626088498, 'gradnorm/max_param': 474.0628967285156, 'epoch': 8.12}

 81%|████████  | 15500/19090 [7:34:05<1:46:20,  1.78s/it]
 81%|████████  | 15501/19090 [7:34:09<2:32:49,  2.55s/it]
 81%|████████  | 15502/19090 [7:34:10<2:10:10,  2.18s/it]
 81%|████████  | 15503/19090 [7:34:12<1:54:17,  1.91s/it]
 81%|████████  | 15504/19090 [7:34:13<1:43:04,  1.72s/it]
 81%|████████  | 15505/19090 [7:34:14<1:35:08,  1.59s/it]
 81%|████████  | 15506/19090 [7:34:15<1:29:34,  1.50s/it]
 81%|████████  | 15507/19090 [7:34:17<1:25:47,  1.44s/it]
 81%|████████  | 15508/19090 [7:34:18<1:23:01,  1.39s/it]
 81%|████████  | 15509/19090 [7:34:19<1:21:01,  1.36s/it][gradnorm] step 15509: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 648.4

 81%|████████  | 15510/19090 [7:34:21<1:19:44,  1.34s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 933.89501953125, 'learning_rate': 5.1706894866928396e-06, 'recon_loss': 0.0015681716002291068, 'recon_static': 0.0006004075519740581, 'recon_motion': 0.0009677640438894742, 'recon_norm': 1.2993214458227158, 'recon_norm_std': 0.03965166490525007, 'target_norm': 1.620902308821678, 'recon_cos': 0.6323261767625808, 'gradnorm/compressor.query': 0.5075629868436352, 'gradnorm/compressor.layer0': 16.26278793701667, 'gradnorm/compressor.layer1': 918.7841150128794, 'gradnorm/compressor.layer2': 162.18444120429677, 'gradnorm/compressor.layer3': 36.047408359188076, 'gradnorm/compressor.layer4': 8.261730755014918, 'gradnorm/compressor.layer5': 5.1205413776517945, 'gradnorm/compressor.layer6': 3.0835653338054443, 'gradnorm/compressor.layer7': 3.0766218358910384, 'gradnorm/decoder.layer0': 2.955656228306871, 'gradnorm/decoder.layer1': 2.0526500506734666, 'gradnorm/decoder.layer2': 1.1134145556496664, 'gradnorm/max_param': 648.4356079101562, 'epoch': 8.12}

 81%|████████  | 15510/19090 [7:34:21<1:19:44,  1.34s/it]
 81%|████████▏ | 15511/19090 [7:34:22<1:18:50,  1.32s/it]
 81%|████████▏ | 15512/19090 [7:34:23<1:18:14,  1.31s/it]
 81%|████████▏ | 15513/19090 [7:34:24<1:17:40,  1.30s/it]
 81%|████████▏ | 15514/19090 [7:34:26<1:24:34,  1.42s/it]
 81%|████████▏ | 15515/19090 [7:34:27<1:22:06,  1.38s/it]
 81%|████████▏ | 15516/19090 [7:34:30<1:38:39,  1.66s/it]
 81%|████████▏ | 15517/19090 [7:34:31<1:37:23,  1.64s/it]
 81%|████████▏ | 15518/19090 [7:34:33<1:38:16,  1.65s/it]
 81%|████████▏ | 15519/19090 [7:34:35<1:47:22,  1.80s/it][gradnorm] step 15519: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 673.2

 81%|████████▏ | 15520/19090 [7:34:37<1:43:28,  1.74s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 960.6339721679688, 'learning_rate': 5.14288345564872e-06, 'recon_loss': 0.001580521569121629, 'recon_static': 0.0006104556247009896, 'recon_motion': 0.00097006595460698, 'recon_norm': 1.3068033427000045, 'recon_norm_std': 0.03988624783232808, 'target_norm': 1.619011190533638, 'recon_cos': 0.6290901318192482, 'gradnorm/compressor.query': 0.48999234794461166, 'gradnorm/compressor.layer0': 15.328141923092085, 'gradnorm/compressor.layer1': 954.0491667099762, 'gradnorm/compressor.layer2': 104.68217337195001, 'gradnorm/compressor.layer3': 35.028148558070406, 'gradnorm/compressor.layer4': 10.583859127730515, 'gradnorm/compressor.layer5': 6.000731741471657, 'gradnorm/compressor.layer6': 3.2586880977766666, 'gradnorm/compressor.layer7': 3.7459589483452684, 'gradnorm/decoder.layer0': 3.1513655230763344, 'gradnorm/decoder.layer1': 1.846161187062585, 'gradnorm/decoder.layer2': 1.1255535247494033, 'gradnorm/max_param': 673.2189331054688, 'epoch': 8.13}

 81%|████████▏ | 15520/19090 [7:34:37<1:43:28,  1.74s/it]
 81%|████████▏ | 15521/19090 [7:34:39<1:43:54,  1.75s/it]
 81%|████████▏ | 15522/19090 [7:34:41<1:47:30,  1.81s/it]
 81%|████████▏ | 15523/19090 [7:34:42<1:49:04,  1.83s/it]
 81%|████████▏ | 15524/19090 [7:34:44<1:52:46,  1.90s/it]
 81%|████████▏ | 15525/19090 [7:34:46<1:47:28,  1.81s/it]
 81%|████████▏ | 15526/19090 [7:34:48<1:41:18,  1.71s/it]
 81%|████████▏ | 15527/19090 [7:34:50<1:48:14,  1.82s/it]
 81%|████████▏ | 15528/19090 [7:34:52<1:51:02,  1.87s/it]
 81%|████████▏ | 15529/19090 [7:34:53<1:40:52,  1.70s/it][gradnorm] step 15529: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 659.3

 81%|████████▏ | 15530/19090 [7:34:55<1:45:31,  1.78s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 941.1295776367188, 'learning_rate': 5.11514381709102e-06, 'recon_loss': 0.0015820077300304548, 'recon_static': 0.0006099444944993593, 'recon_motion': 0.0009720632471726276, 'recon_norm': 1.3048835694789886, 'recon_norm_std': 0.04036896117031574, 'target_norm': 1.621241146326065, 'recon_cos': 0.6296283349394798, 'gradnorm/compressor.query': 0.41712749349679334, 'gradnorm/compressor.layer0': 13.164090898522781, 'gradnorm/compressor.layer1': 934.460205145195, 'gradnorm/compressor.layer2': 106.88521972336727, 'gradnorm/compressor.layer3': 28.504426311293084, 'gradnorm/compressor.layer4': 7.370560683891143, 'gradnorm/compressor.layer5': 4.483489583824925, 'gradnorm/compressor.layer6': 2.414669713561679, 'gradnorm/compressor.layer7': 2.635977272045693, 'gradnorm/decoder.layer0': 2.652393099026597, 'gradnorm/decoder.layer1': 1.8139227183239528, 'gradnorm/decoder.layer2': 0.9973371038584484, 'gradnorm/max_param': 659.3347778320312, 'epoch': 8.14}

 81%|████████▏ | 15530/19090 [7:34:55<1:45:31,  1.78s/it]
 81%|████████▏ | 15531/19090 [7:34:57<1:54:44,  1.93s/it]
 81%|████████▏ | 15532/19090 [7:34:58<1:43:07,  1.74s/it]
 81%|████████▏ | 15533/19090 [7:35:00<1:41:18,  1.71s/it]
 81%|████████▏ | 15534/19090 [7:35:03<1:56:24,  1.96s/it]
 81%|████████▏ | 15535/19090 [7:35:04<1:44:13,  1.76s/it]
 81%|████████▏ | 15536/19090 [7:35:06<1:42:36,  1.73s/it]
 81%|████████▏ | 15537/19090 [7:35:08<1:55:44,  1.95s/it]
 81%|████████▏ | 15538/19090 [7:35:09<1:44:51,  1.77s/it]
 81%|████████▏ | 15539/19090 [7:35:11<1:42:48,  1.74s/it][gradnorm] step 15539: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 741.0

 81%|████████▏ | 15540/19090 [7:35:13<1:55:09,  1.95s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1058.326171875, 'learning_rate': 5.087470663767532e-06, 'recon_loss': 0.0015948942920658738, 'recon_static': 0.0006072828487958759, 'recon_motion': 0.0009876114461803808, 'recon_norm': 1.3054384410381317, 'recon_norm_std': 0.03979519745334983, 'target_norm': 1.6222217291593553, 'recon_cos': 0.6265856146812439, 'gradnorm/compressor.query': 0.5451768121534342, 'gradnorm/compressor.layer0': 20.55154785861069, 'gradnorm/compressor.layer1': 1050.2078960853419, 'gradnorm/compressor.layer2': 123.2433282930561, 'gradnorm/compressor.layer3': 36.94777974025469, 'gradnorm/compressor.layer4': 8.895687438578472, 'gradnorm/compressor.layer5': 5.176755325498129, 'gradnorm/compressor.layer6': 3.0609743638907667, 'gradnorm/compressor.layer7': 3.3184889868558836, 'gradnorm/decoder.layer0': 2.9754332154871532, 'gradnorm/decoder.layer1': 1.9937084301289056, 'gradnorm/decoder.layer2': 1.2985200215361568, 'gradnorm/max_param': 740.9786376953125, 'epoch': 8.14}

 81%|████████▏ | 15540/19090 [7:35:14<1:55:09,  1.95s/it]
 81%|████████▏ | 15541/19090 [7:35:15<1:43:19,  1.75s/it]
 81%|████████▏ | 15542/19090 [7:35:17<1:43:34,  1.75s/it]
 81%|████████▏ | 15543/19090 [7:35:19<1:54:12,  1.93s/it]
 81%|████████▏ | 15544/19090 [7:35:20<1:42:36,  1.74s/it]
 81%|████████▏ | 15545/19090 [7:35:22<1:35:22,  1.61s/it]
 81%|████████▏ | 15546/19090 [7:35:24<1:58:17,  2.00s/it]
 81%|████████▏ | 15547/19090 [7:35:26<1:45:29,  1.79s/it]
 81%|████████▏ | 15548/19090 [7:35:27<1:36:31,  1.64s/it]
 81%|████████▏ | 15549/19090 [7:35:30<1:57:32,  1.99s/it][gradnorm] step 15549: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 657.6

 81%|████████▏ | 15550/19090 [7:35:31<1:44:57,  1.78s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 941.6822509765625, 'learning_rate': 5.05986408820375e-06, 'recon_loss': 0.0016530115710338578, 'recon_static': 0.0006711211302899755, 'recon_motion': 0.0009818903912673705, 'recon_norm': 1.3305675685405731, 'recon_norm_std': 0.0386549879796803, 'target_norm': 1.6202945917844773, 'recon_cos': 0.613759596645832, 'gradnorm/compressor.query': 0.4893962184646886, 'gradnorm/compressor.layer0': 20.812134152424534, 'gradnorm/compressor.layer1': 932.0107295519725, 'gradnorm/compressor.layer2': 128.72553620737028, 'gradnorm/compressor.layer3': 31.44216113867354, 'gradnorm/compressor.layer4': 8.391413305322423, 'gradnorm/compressor.layer5': 4.896338492230002, 'gradnorm/compressor.layer6': 3.1472744367158203, 'gradnorm/compressor.layer7': 3.3470147852552556, 'gradnorm/decoder.layer0': 2.8913975688450466, 'gradnorm/decoder.layer1': 1.9404879265134158, 'gradnorm/decoder.layer2': 1.1989898463177997, 'gradnorm/max_param': 657.5838012695312, 'epoch': 8.15}

 81%|████████▏ | 15550/19090 [7:35:31<1:44:57,  1.78s/it]
 81%|████████▏ | 15551/19090 [7:35:32<1:36:11,  1.63s/it]
 81%|████████▏ | 15552/19090 [7:35:35<1:57:58,  2.00s/it]
 81%|████████▏ | 15553/19090 [7:35:37<1:45:16,  1.79s/it]
 81%|████████▏ | 15554/19090 [7:35:38<1:36:24,  1.64s/it]
 81%|████████▏ | 15555/19090 [7:35:41<1:57:32,  2.00s/it]
 81%|████████▏ | 15556/19090 [7:35:42<1:44:54,  1.78s/it]
 81%|████████▏ | 15557/19090 [7:35:43<1:36:02,  1.63s/it]
 81%|████████▏ | 15558/19090 [7:35:46<2:02:01,  2.07s/it]
 82%|████████▏ | 15559/19090 [7:35:48<1:47:57,  1.83s/it][gradnorm] step 15559: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 576.6

 82%|████████▏ | 15560/19090 [7:35:49<1:38:10,  1.67s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 822.5715942382812, 'learning_rate': 5.032324182702558e-06, 'recon_loss': 0.0016312010178808124, 'recon_static': 0.000641519452619832, 'recon_motion': 0.000989681636565365, 'recon_norm': 1.3201592773199082, 'recon_norm_std': 0.039373948611319064, 'target_norm': 1.6243708342313767, 'recon_cos': 0.620283916592598, 'gradnorm/compressor.query': 0.3505440689630935, 'gradnorm/compressor.layer0': 12.174998596316653, 'gradnorm/compressor.layer1': 817.1994478094072, 'gradnorm/compressor.layer2': 87.59617761084812, 'gradnorm/compressor.layer3': 29.742199812842212, 'gradnorm/compressor.layer4': 7.933606491942624, 'gradnorm/compressor.layer5': 4.048609299365787, 'gradnorm/compressor.layer6': 2.280951179895067, 'gradnorm/compressor.layer7': 2.594257442978711, 'gradnorm/decoder.layer0': 2.758250431378505, 'gradnorm/decoder.layer1': 1.8108663102278053, 'gradnorm/decoder.layer2': 0.9943343837458397, 'gradnorm/max_param': 576.6455688476562, 'epoch': 8.15}

 82%|████████▏ | 15560/19090 [7:35:49<1:38:10,  1.67s/it]
 82%|████████▏ | 15561/19090 [7:35:52<1:57:31,  2.00s/it]
 82%|████████▏ | 15562/19090 [7:35:53<1:44:49,  1.78s/it]
 82%|████████▏ | 15563/19090 [7:35:54<1:36:07,  1.64s/it]
 82%|████████▏ | 15564/19090 [7:35:57<1:58:05,  2.01s/it]
 82%|████████▏ | 15565/19090 [7:35:58<1:45:21,  1.79s/it]
 82%|████████▏ | 15566/19090 [7:36:00<1:36:31,  1.64s/it]
 82%|████████▏ | 15567/19090 [7:36:03<1:58:41,  2.02s/it]
 82%|████████▏ | 15568/19090 [7:36:04<1:45:44,  1.80s/it]
 82%|████████▏ | 15569/19090 [7:36:05<1:36:48,  1.65s/it][gradnorm] step 15569: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 599.0

 82%|████████▏ | 15570/19090 [7:36:08<1:58:15,  2.02s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 862.0529174804688, 'learning_rate': 5.004851039343933e-06, 'recon_loss': 0.001658333433442749, 'recon_static': 0.0006787217527744361, 'recon_motion': 0.000979611683578696, 'recon_norm': 1.3366955131292344, 'recon_norm_std': 0.0387686600908637, 'target_norm': 1.6203496247529983, 'recon_cos': 0.6134528964757919, 'gradnorm/compressor.query': 0.5014243699244383, 'gradnorm/compressor.layer0': 17.104138218263238, 'gradnorm/compressor.layer1': 848.8575778656865, 'gradnorm/compressor.layer2': 144.21514418569222, 'gradnorm/compressor.layer3': 36.981201345138686, 'gradnorm/compressor.layer4': 8.324244871058266, 'gradnorm/compressor.layer5': 4.151975340841638, 'gradnorm/compressor.layer6': 2.4480561778036503, 'gradnorm/compressor.layer7': 2.976135543679593, 'gradnorm/decoder.layer0': 3.3542869183010633, 'gradnorm/decoder.layer1': 1.8775694725298517, 'gradnorm/decoder.layer2': 1.221490078367771, 'gradnorm/max_param': 599.03369140625, 'epoch': 8.16}

 82%|████████▏ | 15570/19090 [7:36:08<1:58:15,  2.02s/it]
 82%|████████▏ | 15571/19090 [7:36:09<1:45:26,  1.80s/it]
 82%|████████▏ | 15572/19090 [7:36:11<1:36:18,  1.64s/it]
 82%|████████▏ | 15573/19090 [7:36:14<1:59:23,  2.04s/it]
 82%|████████▏ | 15574/19090 [7:36:15<1:46:17,  1.81s/it]
 82%|████████▏ | 15575/19090 [7:36:16<1:36:54,  1.65s/it]
 82%|████████▏ | 15576/19090 [7:36:19<1:56:07,  1.98s/it]
 82%|████████▏ | 15577/19090 [7:36:20<1:43:59,  1.78s/it]
 82%|████████▏ | 15578/19090 [7:36:21<1:35:18,  1.63s/it]
 82%|████████▏ | 15579/19090 [7:36:24<1:49:16,  1.87s/it][gradnorm] step 15579: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 599.9

 82%|████████▏ | 15580/19090 [7:36:25<1:39:09,  1.70s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 858.7024536132812, 'learning_rate': 4.977444749984628e-06, 'recon_loss': 0.0015923855011351407, 'recon_static': 0.0006211230953340419, 'recon_motion': 0.0009712624116218649, 'recon_norm': 1.3051064491271973, 'recon_norm_std': 0.040141843073070047, 'target_norm': 1.619650763273239, 'recon_cos': 0.6258304581046105, 'gradnorm/compressor.query': 0.4442825956080727, 'gradnorm/compressor.layer0': 11.928440351067879, 'gradnorm/compressor.layer1': 850.3440847680425, 'gradnorm/compressor.layer2': 111.91460978269325, 'gradnorm/compressor.layer3': 38.58154064212979, 'gradnorm/compressor.layer4': 8.945074966142114, 'gradnorm/compressor.layer5': 4.577205630743324, 'gradnorm/compressor.layer6': 2.531264976174591, 'gradnorm/compressor.layer7': 2.9775584399480164, 'gradnorm/decoder.layer0': 2.901495048301793, 'gradnorm/decoder.layer1': 1.7733128037495294, 'gradnorm/decoder.layer2': 1.0294394349137979, 'gradnorm/max_param': 599.8908081054688, 'epoch': 8.16}

 82%|████████▏ | 15580/19090 [7:36:25<1:39:09,  1.70s/it]
 82%|████████▏ | 15581/19090 [7:36:26<1:32:03,  1.57s/it]
 82%|████████▏ | 15582/19090 [7:36:29<1:55:55,  1.98s/it]
 82%|████████▏ | 15583/19090 [7:36:31<1:43:45,  1.78s/it]
 82%|████████▏ | 15584/19090 [7:36:32<1:35:11,  1.63s/it]
 82%|████████▏ | 15585/19090 [7:36:35<1:57:41,  2.01s/it]
 82%|████████▏ | 15586/19090 [7:36:36<1:45:01,  1.80s/it]
 82%|████████▏ | 15587/19090 [7:36:37<1:35:58,  1.64s/it]
 82%|████████▏ | 15588/19090 [7:36:40<1:56:49,  2.00s/it]
 82%|████████▏ | 15589/19090 [7:36:42<1:44:17,  1.79s/it][gradnorm] step 15589: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 752.9

 82%|████████▏ | 15590/19090 [7:36:43<1:35:30,  1.64s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 1074.7691650390625, 'learning_rate': 4.950105406257874e-06, 'recon_loss': 0.0017695837013889104, 'recon_static': 0.0007913672627182678, 'recon_motion': 0.0009782164634088986, 'recon_norm': 1.385163250565529, 'recon_norm_std': 0.03808615757152438, 'target_norm': 1.6196512311697007, 'recon_cos': 0.5951627045869827, 'gradnorm/compressor.query': 0.5693447053473, 'gradnorm/compressor.layer0': 18.671798182373365, 'gradnorm/compressor.layer1': 1066.9284184049088, 'gradnorm/compressor.layer2': 120.73724537968803, 'gradnorm/compressor.layer3': 41.76673124470226, 'gradnorm/compressor.layer4': 8.656240745567029, 'gradnorm/compressor.layer5': 4.2992307551537134, 'gradnorm/compressor.layer6': 2.555415711746108, 'gradnorm/compressor.layer7': 2.9183179449797314, 'gradnorm/decoder.layer0': 2.890058802234151, 'gradnorm/decoder.layer1': 1.82526199602721, 'gradnorm/decoder.layer2': 1.1792377347164769, 'gradnorm/max_param': 752.9368896484375, 'epoch': 8.17}

 82%|████████▏ | 15590/19090 [7:36:43<1:35:30,  1.64s/it]
 82%|████████▏ | 15591/19090 [7:36:45<1:50:51,  1.90s/it]
 82%|████████▏ | 15592/19090 [7:36:47<1:40:02,  1.72s/it]
 82%|████████▏ | 15593/19090 [7:36:48<1:32:34,  1.59s/it]
 82%|████████▏ | 15594/19090 [7:36:51<1:54:22,  1.96s/it]
 82%|████████▏ | 15595/19090 [7:36:52<1:42:29,  1.76s/it]
 82%|████████▏ | 15596/19090 [7:36:53<1:34:12,  1.62s/it]
 82%|████████▏ | 15597/19090 [7:36:56<1:48:21,  1.86s/it]
 82%|████████▏ | 15598/19090 [7:36:57<1:38:28,  1.69s/it]
 82%|████████▏ | 15599/19090 [7:36:58<1:31:42,  1.58s/it][gradnorm] step 15599: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 634.3

 82%|████████▏ | 15600/19090 [7:37:01<1:51:06,  1.91s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 909.4523315429688, 'learning_rate': 4.922833099573068e-06, 'recon_loss': 0.0018309625826077535, 'recon_static': 0.000864129449473694, 'recon_motion': 0.0009668332015280612, 'recon_norm': 1.4118097394704818, 'recon_norm_std': 0.03754087463021279, 'target_norm': 1.6199854552745818, 'recon_cos': 0.5826728984713554, 'gradnorm/compressor.query': 0.4834916612533281, 'gradnorm/compressor.layer0': 17.91446283078633, 'gradnorm/compressor.layer1': 898.7957220636956, 'gradnorm/compressor.layer2': 133.97428624553297, 'gradnorm/compressor.layer3': 29.313213152638713, 'gradnorm/compressor.layer4': 9.568764622911957, 'gradnorm/compressor.layer5': 4.156291531233407, 'gradnorm/compressor.layer6': 2.766712901483607, 'gradnorm/compressor.layer7': 2.9472292220975023, 'gradnorm/decoder.layer0': 3.129535435578499, 'gradnorm/decoder.layer1': 1.9297050444388424, 'gradnorm/decoder.layer2': 1.1922229906551527, 'gradnorm/max_param': 634.291015625, 'epoch': 8.17}

 82%|████████▏ | 15600/19090 [7:37:01<1:51:06,  1.91s/it]
 82%|████████▏ | 15601/19090 [7:37:02<1:40:26,  1.73s/it]
 82%|████████▏ | 15602/19090 [7:37:04<1:33:00,  1.60s/it]
 82%|████████▏ | 15603/19090 [7:37:06<1:48:51,  1.87s/it]
 82%|████████▏ | 15604/19090 [7:37:07<1:38:32,  1.70s/it]
 82%|████████▏ | 15605/19090 [7:37:09<1:31:24,  1.57s/it]
 82%|████████▏ | 15606/19090 [7:37:11<1:49:43,  1.89s/it]
 82%|████████▏ | 15607/19090 [7:37:13<1:39:29,  1.71s/it]
 82%|████████▏ | 15608/19090 [7:37:14<1:32:09,  1.59s/it]
 82%|████████▏ | 15609/19090 [7:37:17<1:52:30,  1.94s/it][gradnorm] step 15609: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 685.7

 82%|████████▏ | 15610/19090 [7:37:18<1:41:16,  1.75s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 982.3510131835938, 'learning_rate': 4.895627921115458e-06, 'recon_loss': 0.0018328010570257901, 'recon_static': 0.0008578483902965672, 'recon_motion': 0.0009749527176609263, 'recon_norm': 1.4155696153640747, 'recon_norm_std': 0.036389713548123835, 'target_norm': 1.6188262015581132, 'recon_cos': 0.5822558164596557, 'gradnorm/compressor.query': 0.47871100660344607, 'gradnorm/compressor.layer0': 15.8553826178678, 'gradnorm/compressor.layer1': 971.7739577700156, 'gradnorm/compressor.layer2': 136.83328416818037, 'gradnorm/compressor.layer3': 39.490580920585515, 'gradnorm/compressor.layer4': 8.772886590144385, 'gradnorm/compressor.layer5': 4.487747253576933, 'gradnorm/compressor.layer6': 3.1086559902037547, 'gradnorm/compressor.layer7': 3.1931804828523767, 'gradnorm/decoder.layer0': 3.3698558457329457, 'gradnorm/decoder.layer1': 2.0883618385985687, 'gradnorm/decoder.layer2': 1.33649887871829, 'gradnorm/max_param': 685.6996459960938, 'epoch': 8.18}

 82%|████████▏ | 15610/19090 [7:37:18<1:41:16,  1.75s/it]
 82%|████████▏ | 15611/19090 [7:37:19<1:33:28,  1.61s/it]
 82%|████████▏ | 15612/19090 [7:37:22<1:56:14,  2.01s/it]
 82%|████████▏ | 15613/19090 [7:37:24<1:43:52,  1.79s/it]
 82%|████████▏ | 15614/19090 [7:37:25<1:35:15,  1.64s/it]
 82%|████████▏ | 15615/19090 [7:37:28<1:54:35,  1.98s/it]
 82%|████████▏ | 15616/19090 [7:37:29<1:42:47,  1.78s/it]
 82%|████████▏ | 15617/19090 [7:37:30<1:34:22,  1.63s/it]
 82%|████████▏ | 15618/19090 [7:37:33<1:48:27,  1.87s/it]
 82%|████████▏ | 15619/19090 [7:37:34<1:38:30,  1.70s/it][gradnorm] step 15619: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 817.2

 82%|████████▏ | 15620/19090 [7:37:35<1:31:38,  1.58s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 1169.841552734375, 'learning_rate': 4.868489961845865e-06, 'recon_loss': 0.0017274205893045291, 'recon_static': 0.0007417113476549276, 'recon_motion': 0.000985709225642495, 'recon_norm': 1.3566136837005616, 'recon_norm_std': 0.03794486382976174, 'target_norm': 1.6229463934898376, 'recon_cos': 0.5997405111789703, 'gradnorm/compressor.query': 0.502038584755614, 'gradnorm/compressor.layer0': 20.783455592542534, 'gradnorm/compressor.layer1': 1158.1610423425577, 'gradnorm/compressor.layer2': 157.43259346264674, 'gradnorm/compressor.layer3': 42.89070895938959, 'gradnorm/compressor.layer4': 9.032416984373, 'gradnorm/compressor.layer5': 4.466689695232487, 'gradnorm/compressor.layer6': 2.9948963622437845, 'gradnorm/compressor.layer7': 2.9918738616628815, 'gradnorm/decoder.layer0': 3.120199712817077, 'gradnorm/decoder.layer1': 2.021828619352772, 'gradnorm/decoder.layer2': 1.4447318070416926, 'gradnorm/max_param': 817.2021484375, 'epoch': 8.18}

 82%|████████▏ | 15620/19090 [7:37:35<1:31:38,  1.58s/it]
 82%|████████▏ | 15621/19090 [7:37:38<1:50:35,  1.91s/it]
 82%|████████▏ | 15622/19090 [7:37:39<1:39:49,  1.73s/it]
 82%|████████▏ | 15623/19090 [7:37:41<1:32:15,  1.60s/it]
 82%|████████▏ | 15624/19090 [7:37:43<1:49:45,  1.90s/it]
 82%|████████▏ | 15625/19090 [7:37:44<1:39:18,  1.72s/it]
 82%|████████▏ | 15626/19090 [7:37:46<1:32:01,  1.59s/it]
 82%|████████▏ | 15627/19090 [7:37:49<1:53:45,  1.97s/it]
 82%|████████▏ | 15628/19090 [7:37:50<1:41:57,  1.77s/it]
 82%|████████▏ | 15629/19090 [7:37:51<1:33:42,  1.62s/it][gradnorm] step 15629: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 688.4

 82%|████████▏ | 15630/19090 [7:37:54<1:54:28,  1.99s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 986.8530883789062, 'learning_rate': 4.841419312500348e-06, 'recon_loss': 0.0016056895081419497, 'recon_static': 0.0006272412880207412, 'recon_motion': 0.0009784482113900594, 'recon_norm': 1.3140134125947953, 'recon_norm_std': 0.039807432889938356, 'target_norm': 1.6223040044307708, 'recon_cos': 0.6251617178320885, 'gradnorm/compressor.query': 0.3265503611886663, 'gradnorm/compressor.layer0': 10.451952402350294, 'gradnorm/compressor.layer1': 975.4233375822008, 'gradnorm/compressor.layer2': 144.47802736272737, 'gradnorm/compressor.layer3': 36.54097757302324, 'gradnorm/compressor.layer4': 8.088607273625858, 'gradnorm/compressor.layer5': 4.203630268117104, 'gradnorm/compressor.layer6': 2.8055352335110975, 'gradnorm/compressor.layer7': 2.7597181096073253, 'gradnorm/decoder.layer0': 2.79712646035922, 'gradnorm/decoder.layer1': 1.6298207156338556, 'gradnorm/decoder.layer2': 0.9061586399725017, 'gradnorm/max_param': 688.4096069335938, 'epoch': 8.19}

 82%|████████▏ | 15630/19090 [7:37:54<1:54:28,  1.99s/it]
 82%|████████▏ | 15631/19090 [7:37:55<1:43:28,  1.79s/it]
 82%|████████▏ | 15632/19090 [7:37:57<1:34:35,  1.64s/it]
 82%|████████▏ | 15633/19090 [7:37:59<1:46:33,  1.85s/it]
 82%|████████▏ | 15634/19090 [7:38:00<1:36:42,  1.68s/it]
 82%|████████▏ | 15635/19090 [7:38:02<1:29:45,  1.56s/it]
 82%|████████▏ | 15636/19090 [7:38:04<1:50:11,  1.91s/it]
 82%|████████▏ | 15637/19090 [7:38:06<1:39:15,  1.72s/it]
 82%|████████▏ | 15638/19090 [7:38:07<1:31:42,  1.59s/it]
 82%|████████▏ | 15639/19090 [7:38:09<1:48:18,  1.88s/it][gradnorm] step 15639: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 730.1

 82%|████████▏ | 15640/19090 [7:38:11<1:38:04,  1.71s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1043.8958740234375, 'learning_rate': 4.814416063589919e-06, 'recon_loss': 0.0015712742897449061, 'recon_static': 0.0005909603205509484, 'recon_motion': 0.0009803139415453187, 'recon_norm': 1.2873859643936156, 'recon_norm_std': 0.0398006746545434, 'target_norm': 1.6242428183555604, 'recon_cos': 0.6317394971847534, 'gradnorm/compressor.query': 0.46338826402193345, 'gradnorm/compressor.layer0': 13.793904625216543, 'gradnorm/compressor.layer1': 1034.6438396859085, 'gradnorm/compressor.layer2': 133.49336506218202, 'gradnorm/compressor.layer3': 32.95162655139823, 'gradnorm/compressor.layer4': 9.209735148014953, 'gradnorm/compressor.layer5': 4.126880101696595, 'gradnorm/compressor.layer6': 2.8514708621833855, 'gradnorm/compressor.layer7': 3.153336946379169, 'gradnorm/decoder.layer0': 3.0491293367983614, 'gradnorm/decoder.layer1': 1.8665827326410096, 'gradnorm/decoder.layer2': 1.1354045283984053, 'gradnorm/max_param': 730.0989379882812, 'epoch': 8.19}

 82%|████████▏ | 15640/19090 [7:38:11<1:38:04,  1.71s/it]
 82%|████████▏ | 15641/19090 [7:38:12<1:30:52,  1.58s/it]
 82%|████████▏ | 15642/19090 [7:38:15<1:53:12,  1.97s/it]
 82%|████████▏ | 15643/19090 [7:38:16<1:41:21,  1.76s/it]
 82%|████████▏ | 15644/19090 [7:38:17<1:33:04,  1.62s/it]
 82%|████████▏ | 15645/19090 [7:38:20<1:53:22,  1.97s/it]
 82%|████████▏ | 15646/19090 [7:38:22<1:41:32,  1.77s/it]
 82%|████████▏ | 15647/19090 [7:38:23<1:33:13,  1.62s/it]
 82%|████████▏ | 15648/19090 [7:38:26<1:52:55,  1.97s/it]
 82%|████████▏ | 15649/19090 [7:38:27<1:41:06,  1.76s/it][gradnorm] step 15649: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 766.8

 82%|████████▏ | 15650/19090 [7:38:28<1:32:59,  1.62s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1096.8642578125, 'learning_rate': 4.787480305400235e-06, 'recon_loss': 0.0015516313345870004, 'recon_static': 0.0005786457331851125, 'recon_motion': 0.0009729856072226539, 'recon_norm': 1.3003343373537064, 'recon_norm_std': 0.0413237021304667, 'target_norm': 1.6217997074127197, 'recon_cos': 0.6379358932375908, 'gradnorm/compressor.query': 0.6673572038577371, 'gradnorm/compressor.layer0': 20.643898923026644, 'gradnorm/compressor.layer1': 1086.8381894283987, 'gradnorm/compressor.layer2': 142.0542232520737, 'gradnorm/compressor.layer3': 34.329108215892205, 'gradnorm/compressor.layer4': 7.964155480775891, 'gradnorm/compressor.layer5': 4.234900860667533, 'gradnorm/compressor.layer6': 2.5595513102594905, 'gradnorm/compressor.layer7': 2.9276868168669017, 'gradnorm/decoder.layer0': 2.918663830679812, 'gradnorm/decoder.layer1': 1.8517607892570473, 'gradnorm/decoder.layer2': 1.0104819732220793, 'gradnorm/max_param': 766.7550659179688, 'epoch': 8.2}

 82%|████████▏ | 15650/19090 [7:38:28<1:32:59,  1.62s/it]
 82%|████████▏ | 15651/19090 [7:38:31<1:49:53,  1.92s/it]
 82%|████████▏ | 15652/19090 [7:38:32<1:39:03,  1.73s/it]
 82%|████████▏ | 15653/19090 [7:38:33<1:31:41,  1.60s/it]
 82%|████████▏ | 15654/19090 [7:38:36<1:45:59,  1.85s/it]
 82%|████████▏ | 15655/19090 [7:38:37<1:36:28,  1.69s/it]
 82%|████████▏ | 15656/19090 [7:38:38<1:29:51,  1.57s/it]
 82%|████████▏ | 15657/19090 [7:38:41<1:47:37,  1.88s/it]
 82%|████████▏ | 15658/19090 [7:38:42<1:37:37,  1.71s/it]
 82%|████████▏ | 15659/19090 [7:38:44<1:36:48,  1.69s/it][gradnorm] step 15659: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 503.4

 82%|████████▏ | 15660/19090 [7:38:46<1:45:27,  1.84s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 724.4222412109375, 'learning_rate': 4.7606121279913016e-06, 'recon_loss': 0.0015858799946727232, 'recon_static': 0.0006183927282108926, 'recon_motion': 0.0009674872475443408, 'recon_norm': 1.3030620723962785, 'recon_norm_std': 0.03961210809648037, 'target_norm': 1.6177476644515991, 'recon_cos': 0.6263797834515572, 'gradnorm/compressor.query': 0.5083680268731676, 'gradnorm/compressor.layer0': 14.396118573437382, 'gradnorm/compressor.layer1': 713.5415413625194, 'gradnorm/compressor.layer2': 119.15650816258422, 'gradnorm/compressor.layer3': 33.062619993284144, 'gradnorm/compressor.layer4': 9.908442787055666, 'gradnorm/compressor.layer5': 4.284679737563965, 'gradnorm/compressor.layer6': 3.0406422566534816, 'gradnorm/compressor.layer7': 2.7236911990369213, 'gradnorm/decoder.layer0': 3.0613790523668425, 'gradnorm/decoder.layer1': 1.8773812749432568, 'gradnorm/decoder.layer2': 1.007665519937899, 'gradnorm/max_param': 503.37005615234375, 'epoch': 8.2}

 82%|████████▏ | 15660/19090 [7:38:46<1:45:27,  1.84s/it]
 82%|████████▏ | 15661/19090 [7:38:47<1:35:54,  1.68s/it]
 82%|████████▏ | 15662/19090 [7:38:50<1:42:05,  1.79s/it]
 82%|████████▏ | 15663/19090 [7:38:51<1:43:37,  1.81s/it]
 82%|████████▏ | 15664/19090 [7:38:53<1:34:33,  1.66s/it]
 82%|████████▏ | 15665/19090 [7:38:55<1:39:16,  1.74s/it]
 82%|████████▏ | 15666/19090 [7:38:56<1:40:56,  1.77s/it]
 82%|████████▏ | 15667/19090 [7:38:58<1:32:52,  1.63s/it]
 82%|████████▏ | 15668/19090 [7:39:00<1:37:38,  1.71s/it]
 82%|████████▏ | 15669/19090 [7:39:02<1:43:06,  1.81s/it][gradnorm] step 15669: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 523.3

 82%|████████▏ | 15670/19090 [7:39:03<1:34:22,  1.66s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 749.9777221679688, 'learning_rate': 4.733811621197162e-06, 'recon_loss': 0.0016944897099165246, 'recon_static': 0.0007146430405555293, 'recon_motion': 0.0009798466795473359, 'recon_norm': 1.348995825648308, 'recon_norm_std': 0.038614144548773766, 'target_norm': 1.620895516872406, 'recon_cos': 0.606943553686142, 'gradnorm/compressor.query': 0.6202215632508775, 'gradnorm/compressor.layer0': 27.711507958605278, 'gradnorm/compressor.layer1': 741.6565074210567, 'gradnorm/compressor.layer2': 101.88664619014114, 'gradnorm/compressor.layer3': 34.096333182960265, 'gradnorm/compressor.layer4': 7.2555802015606785, 'gradnorm/compressor.layer5': 4.2738304724624525, 'gradnorm/compressor.layer6': 2.957159767734361, 'gradnorm/compressor.layer7': 2.8884103900413507, 'gradnorm/decoder.layer0': 2.919968743091666, 'gradnorm/decoder.layer1': 1.714954763655557, 'gradnorm/decoder.layer2': 0.9998800086511053, 'gradnorm/max_param': 523.2821655273438, 'epoch': 8.21}

 82%|████████▏ | 15670/19090 [7:39:03<1:34:22,  1.66s/it]
 82%|████████▏ | 15671/19090 [7:39:05<1:37:05,  1.70s/it]
 82%|████████▏ | 15672/19090 [7:39:07<1:42:54,  1.81s/it]
 82%|████████▏ | 15673/19090 [7:39:08<1:34:14,  1.65s/it]
 82%|████████▏ | 15674/19090 [7:39:10<1:38:05,  1.72s/it]
 82%|████████▏ | 15675/19090 [7:39:12<1:42:40,  1.80s/it]
 82%|████████▏ | 15676/19090 [7:39:13<1:33:58,  1.65s/it]
 82%|████████▏ | 15677/19090 [7:39:15<1:39:53,  1.76s/it]
 82%|████████▏ | 15678/19090 [7:39:17<1:46:25,  1.87s/it]
 82%|████████▏ | 15679/19090 [7:39:19<1:36:37,  1.70s/it][gradnorm] step 15679: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 636.1

 82%|████████▏ | 15680/19090 [7:39:20<1:30:45,  1.60s/it]
                                                         
{'loss': 0.0021, 'grad_norm': 912.4923706054688, 'learning_rate': 4.707078874625606e-06, 'recon_loss': 0.002125776041066274, 'recon_static': 0.001148968338384293, 'recon_motion': 0.0009768077288754284, 'recon_norm': 1.527231791615486, 'recon_norm_std': 0.0363101189956069, 'target_norm': 1.6237233310937882, 'recon_cos': 0.5398567393422127, 'gradnorm/compressor.query': 0.5857921165665188, 'gradnorm/compressor.layer0': 18.827423969055005, 'gradnorm/compressor.layer1': 901.4311537771479, 'gradnorm/compressor.layer2': 136.00189566693547, 'gradnorm/compressor.layer3': 33.32908460878239, 'gradnorm/compressor.layer4': 7.492162296702265, 'gradnorm/compressor.layer5': 4.159433164125392, 'gradnorm/compressor.layer6': 2.8505374267883234, 'gradnorm/compressor.layer7': 2.8095677667154786, 'gradnorm/decoder.layer0': 2.856736096310333, 'gradnorm/decoder.layer1': 1.8173742852684822, 'gradnorm/decoder.layer2': 0.9858126964865161, 'gradnorm/max_param': 636.0653686523438, 'epoch': 8.21}

 82%|████████▏ | 15680/19090 [7:39:20<1:30:45,  1.60s/it]
 82%|████████▏ | 15681/19090 [7:39:23<1:50:14,  1.94s/it]
 82%|████████▏ | 15682/19090 [7:39:24<1:39:26,  1.75s/it]
 82%|████████▏ | 15683/19090 [7:39:26<1:35:28,  1.68s/it]
 82%|████████▏ | 15684/19090 [7:39:28<1:45:53,  1.87s/it]
 82%|████████▏ | 15685/19090 [7:39:29<1:36:15,  1.70s/it]
 82%|████████▏ | 15686/19090 [7:39:31<1:38:31,  1.74s/it]
 82%|████████▏ | 15687/19090 [7:39:33<1:48:33,  1.91s/it]
 82%|████████▏ | 15688/19090 [7:39:35<1:38:05,  1.73s/it]
 82%|████████▏ | 15689/19090 [7:39:37<1:39:07,  1.75s/it][gradnorm] step 15689: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 558.1

 82%|████████▏ | 15690/19090 [7:39:38<1:42:15,  1.80s/it]
                                                         
{'loss': 0.0019, 'grad_norm': 799.87744140625, 'learning_rate': 4.680413977657863e-06, 'recon_loss': 0.0019078415469266474, 'recon_static': 0.0009502949498710223, 'recon_motion': 0.0009575466436217539, 'recon_norm': 1.4435169249773026, 'recon_norm_std': 0.03766376776620746, 'target_norm': 1.617247039079666, 'recon_cos': 0.5690941825509072, 'gradnorm/compressor.query': 0.4471109280983364, 'gradnorm/compressor.layer0': 16.949137275818238, 'gradnorm/compressor.layer1': 790.9368653691646, 'gradnorm/compressor.layer2': 114.76077359555399, 'gradnorm/compressor.layer3': 25.846171808760552, 'gradnorm/compressor.layer4': 7.081134619450424, 'gradnorm/compressor.layer5': 4.405786368144569, 'gradnorm/compressor.layer6': 2.511689418781645, 'gradnorm/compressor.layer7': 2.9673641785002864, 'gradnorm/decoder.layer0': 2.889550911616751, 'gradnorm/decoder.layer1': 1.7846982272357934, 'gradnorm/decoder.layer2': 1.069583059523136, 'gradnorm/max_param': 558.060302734375, 'epoch': 8.22}

 82%|████████▏ | 15690/19090 [7:39:38<1:42:15,  1.80s/it]
 82%|████████▏ | 15691/19090 [7:39:40<1:33:41,  1.65s/it]
 82%|████████▏ | 15692/19090 [7:39:42<1:41:00,  1.78s/it]
 82%|████████▏ | 15693/19090 [7:39:44<1:44:43,  1.85s/it]
 82%|████████▏ | 15694/19090 [7:39:45<1:35:24,  1.69s/it]
 82%|████████▏ | 15695/19090 [7:39:47<1:38:59,  1.75s/it]
 82%|████████▏ | 15696/19090 [7:39:49<1:42:00,  1.80s/it]
 82%|████████▏ | 15697/19090 [7:39:50<1:33:31,  1.65s/it]
 82%|████████▏ | 15698/19090 [7:39:52<1:37:29,  1.72s/it]
 82%|████████▏ | 15699/19090 [7:39:54<1:42:42,  1.82s/it][gradnorm] step 15699: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 672.4

 82%|████████▏ | 15700/19090 [7:39:55<1:33:39,  1.66s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 961.826904296875, 'learning_rate': 4.653817019448306e-06, 'recon_loss': 0.001671765820356086, 'recon_static': 0.0006926144735189155, 'recon_motion': 0.0009791513613890857, 'recon_norm': 1.334505444765091, 'recon_norm_std': 0.03892292371019721, 'target_norm': 1.6238985270261765, 'recon_cos': 0.6111129835247994, 'gradnorm/compressor.query': 0.44162964020879153, 'gradnorm/compressor.layer0': 13.642723385277131, 'gradnorm/compressor.layer1': 952.8215008594212, 'gradnorm/compressor.layer2': 125.39541754859904, 'gradnorm/compressor.layer3': 34.65878773677482, 'gradnorm/compressor.layer4': 8.356938258311034, 'gradnorm/compressor.layer5': 4.978892979887776, 'gradnorm/compressor.layer6': 3.079436202032394, 'gradnorm/compressor.layer7': 3.127560443980182, 'gradnorm/decoder.layer0': 3.391693649996958, 'gradnorm/decoder.layer1': 1.9105596135973812, 'gradnorm/decoder.layer2': 1.244159979796959, 'gradnorm/max_param': 672.4425048828125, 'epoch': 8.22}

 82%|████████▏ | 15700/19090 [7:39:56<1:33:39,  1.66s/it]
 82%|████████▏ | 15701/19090 [7:39:57<1:38:43,  1.75s/it]
 82%|████████▏ | 15702/19090 [7:40:00<1:46:36,  1.89s/it]
 82%|████████▏ | 15703/19090 [7:40:01<1:36:20,  1.71s/it]
 82%|████████▏ | 15704/19090 [7:40:03<1:43:28,  1.83s/it]
 82%|████████▏ | 15705/19090 [7:40:05<1:42:24,  1.82s/it]
 82%|████████▏ | 15706/19090 [7:40:06<1:33:23,  1.66s/it]
 82%|████████▏ | 15707/19090 [7:40:09<1:45:46,  1.88s/it]
 82%|████████▏ | 15708/19090 [7:40:10<1:42:00,  1.81s/it]
 82%|████████▏ | 15709/19090 [7:40:11<1:33:05,  1.65s/it][gradnorm] step 15709: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 667.3

 82%|████████▏ | 15710/19090 [7:40:14<1:50:33,  1.96s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 954.6349487304688, 'learning_rate': 4.627288088924156e-06, 'recon_loss': 0.0016561535594519228, 'recon_static': 0.0006849831872386858, 'recon_motion': 0.000971170405682642, 'recon_norm': 1.3428590416908264, 'recon_norm_std': 0.03986494950950146, 'target_norm': 1.6205922454595565, 'recon_cos': 0.614772766828537, 'gradnorm/compressor.query': 0.4113746348692097, 'gradnorm/compressor.layer0': 17.703987767204545, 'gradnorm/compressor.layer1': 945.5233669772524, 'gradnorm/compressor.layer2': 124.1453910743367, 'gradnorm/compressor.layer3': 38.11329386113607, 'gradnorm/compressor.layer4': 8.742060083410152, 'gradnorm/compressor.layer5': 5.077077340847899, 'gradnorm/compressor.layer6': 2.7914834508034647, 'gradnorm/compressor.layer7': 3.0262696557271305, 'gradnorm/decoder.layer0': 3.3190086766813716, 'gradnorm/decoder.layer1': 1.8896532319868118, 'gradnorm/decoder.layer2': 1.1800809292600536, 'gradnorm/max_param': 667.2553100585938, 'epoch': 8.23}

 82%|████████▏ | 15710/19090 [7:40:14<1:50:33,  1.96s/it]
 82%|████████▏ | 15711/19090 [7:40:16<1:42:38,  1.82s/it]
 82%|████████▏ | 15712/19090 [7:40:17<1:33:30,  1.66s/it]
 82%|████████▏ | 15713/19090 [7:40:20<1:52:12,  1.99s/it]
 82%|████████▏ | 15714/19090 [7:40:21<1:41:50,  1.81s/it]
 82%|████████▏ | 15715/19090 [7:40:22<1:32:56,  1.65s/it]
 82%|████████▏ | 15716/19090 [7:40:25<1:48:29,  1.93s/it]
 82%|████████▏ | 15717/19090 [7:40:26<1:41:13,  1.80s/it]
 82%|████████▏ | 15718/19090 [7:40:28<1:32:26,  1.64s/it]
 82%|████████▏ | 15719/19090 [7:40:30<1:48:04,  1.92s/it][gradnorm] step 15719: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 522.1

 82%|████████▏ | 15720/19090 [7:40:32<1:44:51,  1.87s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 749.4883422851562, 'learning_rate': 4.600827274785186e-06, 'recon_loss': 0.0016287354665109888, 'recon_static': 0.0006526655328343622, 'recon_motion': 0.000976069932221435, 'recon_norm': 1.3247547447681427, 'recon_norm_std': 0.040083049517124894, 'target_norm': 1.6225268423557282, 'recon_cos': 0.6203181818127632, 'gradnorm/compressor.query': 0.28344023414058556, 'gradnorm/compressor.layer0': 10.290611758717377, 'gradnorm/compressor.layer1': 739.8983460584299, 'gradnorm/compressor.layer2': 114.55264977429593, 'gradnorm/compressor.layer3': 30.338125444212107, 'gradnorm/compressor.layer4': 9.01229654153905, 'gradnorm/compressor.layer5': 4.778658494923905, 'gradnorm/compressor.layer6': 2.835798780343065, 'gradnorm/compressor.layer7': 2.961215285503918, 'gradnorm/decoder.layer0': 2.906555815983203, 'gradnorm/decoder.layer1': 1.9871310821361414, 'gradnorm/decoder.layer2': 1.1525485616453146, 'gradnorm/max_param': 522.1215209960938, 'epoch': 8.23}

 82%|████████▏ | 15720/19090 [7:40:32<1:44:51,  1.87s/it]
 82%|████████▏ | 15721/19090 [7:40:33<1:35:03,  1.69s/it]
 82%|████████▏ | 15722/19090 [7:40:36<1:48:05,  1.93s/it]
 82%|████████▏ | 15723/19090 [7:40:38<1:47:50,  1.92s/it]
 82%|████████▏ | 15724/19090 [7:40:39<1:37:02,  1.73s/it]
 82%|████████▏ | 15725/19090 [7:40:41<1:49:20,  1.95s/it]
 82%|████████▏ | 15726/19090 [7:40:43<1:44:26,  1.86s/it]
 82%|████████▏ | 15727/19090 [7:40:44<1:34:37,  1.69s/it]
 82%|████████▏ | 15728/19090 [7:40:46<1:40:29,  1.79s/it]
 82%|████████▏ | 15729/19090 [7:40:48<1:40:44,  1.80s/it][gradnorm] step 15729: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 520.1

 82%|████████▏ | 15730/19090 [7:40:50<1:32:07,  1.65s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 744.9940185546875, 'learning_rate': 4.574434665503407e-06, 'recon_loss': 0.0015892862400505691, 'recon_static': 0.0006205057929037139, 'recon_motion': 0.0009687804369605147, 'recon_norm': 1.3095591872930528, 'recon_norm_std': 0.03988398909568787, 'target_norm': 1.6181194305419921, 'recon_cos': 0.6269346192479134, 'gradnorm/compressor.query': 0.3515549552955626, 'gradnorm/compressor.layer0': 13.134512224553797, 'gradnorm/compressor.layer1': 737.1066323809601, 'gradnorm/compressor.layer2': 101.46225822319352, 'gradnorm/compressor.layer3': 33.24639276426748, 'gradnorm/compressor.layer4': 7.922128767354326, 'gradnorm/compressor.layer5': 4.698662855796986, 'gradnorm/compressor.layer6': 2.7692356140143244, 'gradnorm/compressor.layer7': 3.129073273085794, 'gradnorm/decoder.layer0': 3.1644657584459583, 'gradnorm/decoder.layer1': 1.9380266642788195, 'gradnorm/decoder.layer2': 1.1361995179698898, 'gradnorm/max_param': 520.1083374023438, 'epoch': 8.24}

 82%|████████▏ | 15730/19090 [7:40:50<1:32:07,  1.65s/it]
 82%|████████▏ | 15731/19090 [7:40:52<1:47:43,  1.92s/it]
 82%|████████▏ | 15732/19090 [7:40:53<1:38:14,  1.76s/it]
 82%|████████▏ | 15733/19090 [7:40:55<1:30:14,  1.61s/it]
 82%|████████▏ | 15734/19090 [7:40:57<1:48:40,  1.94s/it]
 82%|████████▏ | 15735/19090 [7:40:59<1:49:06,  1.95s/it]
 82%|████████▏ | 15736/19090 [7:41:01<1:38:08,  1.76s/it]
 82%|████████▏ | 15737/19090 [7:41:02<1:36:18,  1.72s/it]
 82%|████████▏ | 15738/19090 [7:41:05<1:49:24,  1.96s/it]
 82%|████████▏ | 15739/19090 [7:41:06<1:38:18,  1.76s/it][gradnorm] step 15739: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 503.5

 82%|████████▏ | 15740/19090 [7:41:07<1:30:42,  1.62s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 725.1236572265625, 'learning_rate': 4.54811034932282e-06, 'recon_loss': 0.0016845686332089826, 'recon_static': 0.0007136790940421633, 'recon_motion': 0.0009708895406220109, 'recon_norm': 1.345527672767639, 'recon_norm_std': 0.03894396908581257, 'target_norm': 1.6216615170240403, 'recon_cos': 0.608442808687687, 'gradnorm/compressor.query': 0.6953759110928385, 'gradnorm/compressor.layer0': 20.19363357382872, 'gradnorm/compressor.layer1': 713.9863225230579, 'gradnorm/compressor.layer2': 122.38759990037798, 'gradnorm/compressor.layer3': 22.889158442781376, 'gradnorm/compressor.layer4': 8.122345945729558, 'gradnorm/compressor.layer5': 4.54831591449858, 'gradnorm/compressor.layer6': 2.9467533543581315, 'gradnorm/compressor.layer7': 2.7301921131322198, 'gradnorm/decoder.layer0': 3.07234260671195, 'gradnorm/decoder.layer1': 1.8338847920414623, 'gradnorm/decoder.layer2': 1.1461423399440733, 'gradnorm/max_param': 503.48553466796875, 'epoch': 8.25}

 82%|████████▏ | 15740/19090 [7:41:08<1:30:42,  1.62s/it]
 82%|████████▏ | 15741/19090 [7:41:10<1:44:21,  1.87s/it]
 82%|████████▏ | 15742/19090 [7:41:11<1:34:47,  1.70s/it]
 82%|████████▏ | 15743/19090 [7:41:13<1:33:16,  1.67s/it]
 82%|████████▏ | 15744/19090 [7:41:15<1:42:43,  1.84s/it]
 82%|████████▏ | 15745/19090 [7:41:16<1:33:37,  1.68s/it]
 82%|████████▏ | 15746/19090 [7:41:18<1:35:38,  1.72s/it]
 82%|████████▏ | 15747/19090 [7:41:20<1:45:02,  1.89s/it]
 82%|████████▏ | 15748/19090 [7:41:22<1:35:12,  1.71s/it]
 82%|████████▏ | 15749/19090 [7:41:24<1:37:48,  1.76s/it][gradnorm] step 15749: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 537.8

 83%|████████▎ | 15750/19090 [7:41:26<1:41:36,  1.83s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 770.07666015625, 'learning_rate': 4.521854414259041e-06, 'recon_loss': 0.0016367608855944126, 'recon_static': 0.000672598859819118, 'recon_motion': 0.000964162046147976, 'recon_norm': 1.3280284881591797, 'recon_norm_std': 0.040657124295830724, 'target_norm': 1.6199406772851943, 'recon_cos': 0.617383961379528, 'gradnorm/compressor.query': 0.3604241274831387, 'gradnorm/compressor.layer0': 12.020720077501206, 'gradnorm/compressor.layer1': 762.2230234649173, 'gradnorm/compressor.layer2': 103.72109913730186, 'gradnorm/compressor.layer3': 31.940250882935214, 'gradnorm/compressor.layer4': 7.979890822043763, 'gradnorm/compressor.layer5': 4.1203716784477935, 'gradnorm/compressor.layer6': 2.9915621031726336, 'gradnorm/compressor.layer7': 2.7700499572105004, 'gradnorm/decoder.layer0': 3.1341566593571972, 'gradnorm/decoder.layer1': 1.7911945948314338, 'gradnorm/decoder.layer2': 0.949517693560135, 'gradnorm/max_param': 537.7783203125, 'epoch': 8.25}

 83%|████████▎ | 15750/19090 [7:41:26<1:41:36,  1.83s/it]
 83%|████████▎ | 15751/19090 [7:41:27<1:32:54,  1.67s/it]
 83%|████████▎ | 15752/19090 [7:41:29<1:35:52,  1.72s/it]
 83%|████████▎ | 15753/19090 [7:41:31<1:39:04,  1.78s/it]
 83%|████████▎ | 15754/19090 [7:41:32<1:30:51,  1.63s/it]
 83%|████████▎ | 15755/19090 [7:41:34<1:39:30,  1.79s/it]
 83%|████████▎ | 15756/19090 [7:41:36<1:35:45,  1.72s/it]
 83%|████████▎ | 15757/19090 [7:41:37<1:28:39,  1.60s/it]
 83%|████████▎ | 15758/19090 [7:41:39<1:40:54,  1.82s/it]
 83%|████████▎ | 15759/19090 [7:41:41<1:40:40,  1.81s/it][gradnorm] step 15759: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 613.4

 83%|████████▎ | 15760/19090 [7:41:42<1:32:06,  1.66s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 877.2894287109375, 'learning_rate': 4.495666948099078e-06, 'recon_loss': 0.0016504552826518194, 'recon_static': 0.0006798115457058885, 'recon_motion': 0.00097064374567708, 'recon_norm': 1.337278100848198, 'recon_norm_std': 0.03887948039919138, 'target_norm': 1.619160234928131, 'recon_cos': 0.6150323778390885, 'gradnorm/compressor.query': 0.33573579277264365, 'gradnorm/compressor.layer0': 11.953213081783309, 'gradnorm/compressor.layer1': 869.157170769476, 'gradnorm/compressor.layer2': 113.95456794129623, 'gradnorm/compressor.layer3': 31.368370541202108, 'gradnorm/compressor.layer4': 6.839469232768455, 'gradnorm/compressor.layer5': 3.959305466520824, 'gradnorm/compressor.layer6': 2.811650041788454, 'gradnorm/compressor.layer7': 2.766357324916847, 'gradnorm/decoder.layer0': 2.877790009494285, 'gradnorm/decoder.layer1': 1.6537144524061722, 'gradnorm/decoder.layer2': 0.8874424741127247, 'gradnorm/max_param': 613.398193359375, 'epoch': 8.26}

 83%|████████▎ | 15760/19090 [7:41:42<1:32:06,  1.66s/it]
 83%|████████▎ | 15761/19090 [7:41:45<1:41:47,  1.83s/it]
 83%|████████▎ | 15762/19090 [7:41:47<1:42:04,  1.84s/it]
 83%|████████▎ | 15763/19090 [7:41:48<1:32:51,  1.67s/it]
 83%|████████▎ | 15764/19090 [7:41:50<1:40:52,  1.82s/it]
 83%|████████▎ | 15765/19090 [7:41:52<1:46:21,  1.92s/it]
 83%|████████▎ | 15766/19090 [7:41:53<1:35:56,  1.73s/it]
 83%|████████▎ | 15767/19090 [7:41:55<1:37:21,  1.76s/it]
 83%|████████▎ | 15768/19090 [7:41:57<1:40:30,  1.82s/it]
 83%|████████▎ | 15769/19090 [7:41:58<1:31:45,  1.66s/it][gradnorm] step 15769: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 733.4

 83%|████████▎ | 15770/19090 [7:42:00<1:36:47,  1.75s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1047.8619384765625, 'learning_rate': 4.46954803840103e-06, 'recon_loss': 0.0016117734456202015, 'recon_static': 0.0006294074933975935, 'recon_motion': 0.0009823659653193316, 'recon_norm': 1.3173485189676284, 'recon_norm_std': 0.03869941057637334, 'target_norm': 1.6232814908027648, 'recon_cos': 0.624286837875843, 'gradnorm/compressor.query': 0.4702790910634641, 'gradnorm/compressor.layer0': 18.34929493893047, 'gradnorm/compressor.layer1': 1039.3270659421894, 'gradnorm/compressor.layer2': 128.68052585521633, 'gradnorm/compressor.layer3': 28.145983925469945, 'gradnorm/compressor.layer4': 8.581444936873384, 'gradnorm/compressor.layer5': 4.6499831250612615, 'gradnorm/compressor.layer6': 2.842073889795294, 'gradnorm/compressor.layer7': 2.950513381252771, 'gradnorm/decoder.layer0': 3.16243628637493, 'gradnorm/decoder.layer1': 1.72550635162396, 'gradnorm/decoder.layer2': 1.0662301725129417, 'gradnorm/max_param': 733.3983154296875, 'epoch': 8.26}

 83%|████████▎ | 15770/19090 [7:42:00<1:36:47,  1.75s/it]
 83%|████████▎ | 15771/19090 [7:42:02<1:41:48,  1.84s/it]
 83%|████████▎ | 15772/19090 [7:42:04<1:32:40,  1.68s/it]
 83%|████████▎ | 15773/19090 [7:42:06<1:33:14,  1.69s/it]
 83%|████████▎ | 15774/19090 [7:42:08<1:42:50,  1.86s/it]
 83%|████████▎ | 15775/19090 [7:42:09<1:33:31,  1.69s/it]
 83%|████████▎ | 15776/19090 [7:42:11<1:36:41,  1.75s/it]
 83%|████████▎ | 15777/19090 [7:42:13<1:42:41,  1.86s/it]
 83%|████████▎ | 15778/19090 [7:42:14<1:33:18,  1.69s/it]
 83%|████████▎ | 15779/19090 [7:42:16<1:33:51,  1.70s/it][gradnorm] step 15779: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 745.1

 83%|████████▎ | 15780/19090 [7:42:18<1:37:17,  1.76s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1067.18994140625, 'learning_rate': 4.4434977724937404e-06, 'recon_loss': 0.0015976040594978258, 'recon_static': 0.0006279586072196252, 'recon_motion': 0.000969645420263987, 'recon_norm': 1.310177180171013, 'recon_norm_std': 0.03913014214485884, 'target_norm': 1.6201236367225647, 'recon_cos': 0.6255399525165558, 'gradnorm/compressor.query': 0.44278192095544494, 'gradnorm/compressor.layer0': 16.261994146268815, 'gradnorm/compressor.layer1': 1055.7043857065291, 'gradnorm/compressor.layer2': 150.37036957750354, 'gradnorm/compressor.layer3': 36.974737667574296, 'gradnorm/compressor.layer4': 9.359600796785903, 'gradnorm/compressor.layer5': 4.477328942213997, 'gradnorm/compressor.layer6': 3.1260735765756618, 'gradnorm/compressor.layer7': 2.7768364508728705, 'gradnorm/decoder.layer0': 3.062567418193272, 'gradnorm/decoder.layer1': 1.888116417580892, 'gradnorm/decoder.layer2': 1.1491440137345985, 'gradnorm/max_param': 745.0606689453125, 'epoch': 8.27}

 83%|████████▎ | 15780/19090 [7:42:18<1:37:17,  1.76s/it]
 83%|████████▎ | 15781/19090 [7:42:19<1:29:40,  1.63s/it]
 83%|████████▎ | 15782/19090 [7:42:21<1:36:55,  1.76s/it]
 83%|████████▎ | 15783/19090 [7:42:23<1:40:07,  1.82s/it]
 83%|████████▎ | 15784/19090 [7:42:25<1:31:25,  1.66s/it]
 83%|████████▎ | 15785/19090 [7:42:27<1:45:22,  1.91s/it]
 83%|████████▎ | 15786/19090 [7:42:29<1:38:25,  1.79s/it]
 83%|████████▎ | 15787/19090 [7:42:30<1:30:21,  1.64s/it]
 83%|████████▎ | 15788/19090 [7:42:32<1:43:05,  1.87s/it]
 83%|████████▎ | 15789/19090 [7:42:34<1:33:28,  1.70s/it][gradnorm] step 15789: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 566.9

 83%|████████▎ | 15790/19090 [7:42:35<1:26:48,  1.58s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 812.1805419921875, 'learning_rate': 4.417516237476554e-06, 'recon_loss': 0.0015864947868976741, 'recon_static': 0.000616915785940364, 'recon_motion': 0.0009695789762190543, 'recon_norm': 1.3072310864925385, 'recon_norm_std': 0.039880294166505334, 'target_norm': 1.619280594587326, 'recon_cos': 0.627784126996994, 'gradnorm/compressor.query': 0.36096602157923563, 'gradnorm/compressor.layer0': 13.20730506592818, 'gradnorm/compressor.layer1': 803.4050192773257, 'gradnorm/compressor.layer2': 114.798396656922, 'gradnorm/compressor.layer3': 26.527877251758667, 'gradnorm/compressor.layer4': 8.557942089237013, 'gradnorm/compressor.layer5': 4.337261375877154, 'gradnorm/compressor.layer6': 2.8773735863088734, 'gradnorm/compressor.layer7': 2.810013583276031, 'gradnorm/decoder.layer0': 2.86312803350295, 'gradnorm/decoder.layer1': 1.818264771324778, 'gradnorm/decoder.layer2': 0.982210595408286, 'gradnorm/max_param': 566.9406127929688, 'epoch': 8.27}

 83%|████████▎ | 15790/19090 [7:42:35<1:26:48,  1.58s/it]
 83%|████████▎ | 15791/19090 [7:42:38<1:52:03,  2.04s/it]
 83%|████████▎ | 15792/19090 [7:42:39<1:39:49,  1.82s/it]
 83%|████████▎ | 15793/19090 [7:42:41<1:31:16,  1.66s/it]
 83%|████████▎ | 15794/19090 [7:42:44<1:56:09,  2.11s/it]
 83%|████████▎ | 15795/19090 [7:42:45<1:42:33,  1.87s/it]
 83%|████████▎ | 15796/19090 [7:42:46<1:33:01,  1.69s/it]
 83%|████████▎ | 15797/19090 [7:42:49<1:50:34,  2.01s/it]
 83%|████████▎ | 15798/19090 [7:42:50<1:38:33,  1.80s/it]
 83%|████████▎ | 15799/19090 [7:42:52<1:30:16,  1.65s/it][gradnorm] step 15799: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 629.6

 83%|████████▎ | 15800/19090 [7:42:55<1:50:23,  2.01s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 903.1939086914062, 'learning_rate': 4.39160352021904e-06, 'recon_loss': 0.0015850133931962773, 'recon_static': 0.0006106668006395921, 'recon_motion': 0.0009743466100189835, 'recon_norm': 1.3010795176029206, 'recon_norm_std': 0.039445372950285676, 'target_norm': 1.6197621077299118, 'recon_cos': 0.6274270281195641, 'gradnorm/compressor.query': 0.4748400217997588, 'gradnorm/compressor.layer0': 17.98090791134881, 'gradnorm/compressor.layer1': 892.2434505223337, 'gradnorm/compressor.layer2': 133.7749027634668, 'gradnorm/compressor.layer3': 36.25008250260008, 'gradnorm/compressor.layer4': 8.914240521885825, 'gradnorm/compressor.layer5': 4.010281937883355, 'gradnorm/compressor.layer6': 3.1868201634083855, 'gradnorm/compressor.layer7': 2.898522224113384, 'gradnorm/decoder.layer0': 3.0901462103069197, 'gradnorm/decoder.layer1': 1.7225862173087336, 'gradnorm/decoder.layer2': 0.925515052091839, 'gradnorm/max_param': 629.609375, 'epoch': 8.28}

 83%|████████▎ | 15800/19090 [7:42:55<1:50:23,  2.01s/it]
 83%|████████▎ | 15801/19090 [7:42:56<1:38:28,  1.80s/it]
 83%|████████▎ | 15802/19090 [7:42:57<1:30:04,  1.64s/it]
 83%|████████▎ | 15803/19090 [7:43:00<1:43:14,  1.88s/it]
 83%|████████▎ | 15804/19090 [7:43:01<1:33:27,  1.71s/it]
 83%|████████▎ | 15805/19090 [7:43:02<1:26:30,  1.58s/it]
 83%|████████▎ | 15806/19090 [7:43:05<1:46:00,  1.94s/it]
 83%|████████▎ | 15807/19090 [7:43:06<1:35:35,  1.75s/it]
 83%|████████▎ | 15808/19090 [7:43:08<1:28:16,  1.61s/it]
 83%|████████▎ | 15809/19090 [7:43:10<1:42:41,  1.88s/it][gradnorm] step 15809: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 911.5

 83%|████████▎ | 15810/19090 [7:43:12<1:38:08,  1.80s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 1304.644775390625, 'learning_rate': 4.365759707360628e-06, 'recon_loss': 0.0017443488119170069, 'recon_static': 0.0007723453876678832, 'recon_motion': 0.0009720034766360186, 'recon_norm': 1.370054453611374, 'recon_norm_std': 0.03980076853185892, 'target_norm': 1.6224038928747178, 'recon_cos': 0.5995360493659974, 'gradnorm/compressor.query': 1.289614333898169, 'gradnorm/compressor.layer0': 44.30431071850896, 'gradnorm/compressor.layer1': 1291.667526881434, 'gradnorm/compressor.layer2': 172.1570242994023, 'gradnorm/compressor.layer3': 44.53003699005804, 'gradnorm/compressor.layer4': 7.8952857425673395, 'gradnorm/compressor.layer5': 4.273284853592305, 'gradnorm/compressor.layer6': 2.5091476926547562, 'gradnorm/compressor.layer7': 2.851831878061885, 'gradnorm/decoder.layer0': 2.7403770621706594, 'gradnorm/decoder.layer1': 1.931230553587748, 'gradnorm/decoder.layer2': 1.0312132973352968, 'gradnorm/max_param': 911.4779052734375, 'epoch': 8.28}

 83%|████████▎ | 15810/19090 [7:43:12<1:38:08,  1.80s/it]
 83%|████████▎ | 15811/19090 [7:43:13<1:30:03,  1.65s/it]
 83%|████████▎ | 15812/19090 [7:43:15<1:37:28,  1.78s/it]
 83%|████████▎ | 15813/19090 [7:43:17<1:38:44,  1.81s/it]
 83%|████████▎ | 15814/19090 [7:43:18<1:30:31,  1.66s/it]
 83%|████████▎ | 15815/19090 [7:43:20<1:36:34,  1.77s/it]
 83%|████████▎ | 15816/19090 [7:43:22<1:36:44,  1.77s/it]
 83%|████████▎ | 15817/19090 [7:43:23<1:28:59,  1.63s/it]
 83%|████████▎ | 15818/19090 [7:43:25<1:35:58,  1.76s/it]
 83%|████████▎ | 15819/19090 [7:43:28<1:42:55,  1.89s/it][gradnorm] step 15819: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 615.8

 83%|████████▎ | 15820/19090 [7:43:29<1:33:25,  1.71s/it]
                                                         
{'loss': 0.0022, 'grad_norm': 884.8449096679688, 'learning_rate': 4.339984885310397e-06, 'recon_loss': 0.002174659859156236, 'recon_static': 0.0012014098581857979, 'recon_motion': 0.0009732499864185229, 'recon_norm': 1.5451491206884385, 'recon_norm_std': 0.03478375058621168, 'target_norm': 1.6215429455041885, 'recon_cos': 0.533801032602787, 'gradnorm/compressor.query': 0.5530438315109356, 'gradnorm/compressor.layer0': 19.838518164886096, 'gradnorm/compressor.layer1': 872.7057780260195, 'gradnorm/compressor.layer2': 139.6530088849682, 'gradnorm/compressor.layer3': 36.23495444771699, 'gradnorm/compressor.layer4': 8.989972780587443, 'gradnorm/compressor.layer5': 4.381381584945039, 'gradnorm/compressor.layer6': 2.532250571587636, 'gradnorm/compressor.layer7': 2.72127275603719, 'gradnorm/decoder.layer0': 2.79008982141218, 'gradnorm/decoder.layer1': 1.724177996728962, 'gradnorm/decoder.layer2': 0.8506394729672323, 'gradnorm/max_param': 615.7951049804688, 'epoch': 8.29}

 83%|████████▎ | 15820/19090 [7:43:29<1:33:25,  1.71s/it]
 83%|████████▎ | 15821/19090 [7:43:30<1:30:09,  1.65s/it]
 83%|████████▎ | 15822/19090 [7:43:33<1:44:19,  1.92s/it]
 83%|████████▎ | 15823/19090 [7:43:34<1:34:10,  1.73s/it]
 83%|████████▎ | 15824/19090 [7:43:36<1:33:46,  1.72s/it]
 83%|████████▎ | 15825/19090 [7:43:38<1:46:59,  1.97s/it]
 83%|████████▎ | 15826/19090 [7:43:40<1:36:06,  1.77s/it]
 83%|████████▎ | 15827/19090 [7:43:41<1:30:16,  1.66s/it]
 83%|████████▎ | 15828/19090 [7:43:44<1:43:55,  1.91s/it]
 83%|████████▎ | 15829/19090 [7:43:45<1:33:52,  1.73s/it][gradnorm] step 15829: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 669.5

 83%|████████▎ | 15830/19090 [7:43:46<1:29:08,  1.64s/it]
                                                         
{'loss': 0.002, 'grad_norm': 959.8352661132812, 'learning_rate': 4.314279140246757e-06, 'recon_loss': 0.0020228202745784076, 'recon_static': 0.0010439560457598418, 'recon_motion': 0.0009788642782950774, 'recon_norm': 1.4808071672916412, 'recon_norm_std': 0.036221450800076126, 'target_norm': 1.623880273103714, 'recon_cos': 0.5539055868983269, 'gradnorm/compressor.query': 0.363736718253206, 'gradnorm/compressor.layer0': 15.678079276486795, 'gradnorm/compressor.layer1': 948.7961121863854, 'gradnorm/compressor.layer2': 139.28961790546344, 'gradnorm/compressor.layer3': 36.41776407248339, 'gradnorm/compressor.layer4': 7.5836204502401605, 'gradnorm/compressor.layer5': 3.9353249006879483, 'gradnorm/compressor.layer6': 2.3702192877872053, 'gradnorm/compressor.layer7': 2.4613402022014146, 'gradnorm/decoder.layer0': 2.696020149546185, 'gradnorm/decoder.layer1': 1.7480526034298596, 'gradnorm/decoder.layer2': 0.9391388555830098, 'gradnorm/max_param': 669.5457763671875, 'epoch': 8.29}

 83%|████████▎ | 15830/19090 [7:43:46<1:29:08,  1.64s/it]
 83%|████████▎ | 15831/19090 [7:43:49<1:49:58,  2.02s/it]
 83%|████████▎ | 15832/19090 [7:43:51<1:38:17,  1.81s/it]
 83%|████████▎ | 15833/19090 [7:43:52<1:29:53,  1.66s/it]
 83%|████████▎ | 15834/19090 [7:43:54<1:32:37,  1.71s/it]
 83%|████████▎ | 15835/19090 [7:43:55<1:26:05,  1.59s/it]
 83%|████████▎ | 15836/19090 [7:43:56<1:21:30,  1.50s/it]
 83%|████████▎ | 15837/19090 [7:43:58<1:28:31,  1.63s/it]
 83%|████████▎ | 15838/19090 [7:44:00<1:23:09,  1.53s/it]
 83%|████████▎ | 15839/19090 [7:44:01<1:23:39,  1.54s/it][gradnorm] step 15839: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 565.4

 83%|████████▎ | 15840/19090 [7:44:03<1:32:35,  1.71s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 811.4021606445312, 'learning_rate': 4.288642558117134e-06, 'recon_loss': 0.0016970487748039886, 'recon_static': 0.0007242007879540325, 'recon_motion': 0.0009728480406920425, 'recon_norm': 1.3488540351390839, 'recon_norm_std': 0.03913815086707473, 'target_norm': 1.619269785284996, 'recon_cos': 0.6047646045684815, 'gradnorm/compressor.query': 0.4485360638696286, 'gradnorm/compressor.layer0': 15.484369087410576, 'gradnorm/compressor.layer1': 801.2592043153077, 'gradnorm/compressor.layer2': 122.28812991762528, 'gradnorm/compressor.layer3': 32.543561176113435, 'gradnorm/compressor.layer4': 7.886425388844959, 'gradnorm/compressor.layer5': 4.017175753504681, 'gradnorm/compressor.layer6': 2.415722322662512, 'gradnorm/compressor.layer7': 2.8068566395157593, 'gradnorm/decoder.layer0': 2.7319572004537487, 'gradnorm/decoder.layer1': 1.7824548360873103, 'gradnorm/decoder.layer2': 0.9957423469178486, 'gradnorm/max_param': 565.3754272460938, 'epoch': 8.3}

 83%|████████▎ | 15840/19090 [7:44:03<1:32:35,  1.71s/it]
 83%|████████▎ | 15841/19090 [7:44:05<1:25:56,  1.59s/it]
 83%|████████▎ | 15842/19090 [7:44:06<1:25:50,  1.59s/it]
 83%|████████▎ | 15843/19090 [7:44:08<1:35:22,  1.76s/it]
 83%|████████▎ | 15844/19090 [7:44:10<1:27:48,  1.62s/it]
 83%|████████▎ | 15845/19090 [7:44:11<1:28:46,  1.64s/it]
 83%|████████▎ | 15846/19090 [7:44:14<1:39:44,  1.84s/it]
 83%|████████▎ | 15847/19090 [7:44:15<1:30:53,  1.68s/it]
 83%|████████▎ | 15848/19090 [7:44:16<1:24:49,  1.57s/it]
 83%|████████▎ | 15849/19090 [7:44:18<1:32:19,  1.71s/it][gradnorm] step 15849: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 705.6

 83%|████████▎ | 15850/19090 [7:44:20<1:25:38,  1.59s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1010.4547729492188, 'learning_rate': 4.2630752246377224e-06, 'recon_loss': 0.0016002334916265682, 'recon_static': 0.0006295538070844487, 'recon_motion': 0.0009706796161481179, 'recon_norm': 1.3160547882318496, 'recon_norm_std': 0.04024017294868827, 'target_norm': 1.6206696659326554, 'recon_cos': 0.6259837508201599, 'gradnorm/compressor.query': 0.5160869494228895, 'gradnorm/compressor.layer0': 14.774223480215783, 'gradnorm/compressor.layer1': 1000.0005312498589, 'gradnorm/compressor.layer2': 140.11033598560815, 'gradnorm/compressor.layer3': 32.48195863427497, 'gradnorm/compressor.layer4': 7.796402850026565, 'gradnorm/compressor.layer5': 4.756980785716788, 'gradnorm/compressor.layer6': 2.9400211634490523, 'gradnorm/compressor.layer7': 2.7676157709871236, 'gradnorm/decoder.layer0': 2.971237104529938, 'gradnorm/decoder.layer1': 1.8426348012942098, 'gradnorm/decoder.layer2': 1.108260494840786, 'gradnorm/max_param': 705.6386108398438, 'epoch': 8.3}

 83%|████████▎ | 15850/19090 [7:44:20<1:25:38,  1.59s/it]
 83%|████████▎ | 15851/19090 [7:44:21<1:23:37,  1.55s/it]
 83%|████████▎ | 15852/19090 [7:44:24<1:42:44,  1.90s/it]
 83%|████████▎ | 15853/19090 [7:44:25<1:32:55,  1.72s/it]
 83%|████████▎ | 15854/19090 [7:44:27<1:34:33,  1.75s/it]
 83%|████████▎ | 15855/19090 [7:44:29<1:34:41,  1.76s/it]
 83%|████████▎ | 15856/19090 [7:44:30<1:27:03,  1.62s/it]
 83%|████████▎ | 15857/19090 [7:44:32<1:36:19,  1.79s/it]
 83%|████████▎ | 15858/19090 [7:44:34<1:34:37,  1.76s/it]
 83%|████████▎ | 15859/19090 [7:44:35<1:26:59,  1.62s/it][gradnorm] step 15859: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 690.5

 83%|████████▎ | 15860/19090 [7:44:38<1:41:02,  1.88s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 984.9418334960938, 'learning_rate': 4.237577225293196e-06, 'recon_loss': 0.0015690296742832289, 'recon_static': 0.000594302496756427, 'recon_motion': 0.0009747272211825475, 'recon_norm': 1.3014258027076722, 'recon_norm_std': 0.040463111363351345, 'target_norm': 1.6196794718503953, 'recon_cos': 0.6321636497974396, 'gradnorm/compressor.query': 0.35705343131646994, 'gradnorm/compressor.layer0': 9.930935977030705, 'gradnorm/compressor.layer1': 978.3931661147271, 'gradnorm/compressor.layer2': 106.68393923466175, 'gradnorm/compressor.layer3': 35.6736253266375, 'gradnorm/compressor.layer4': 7.502415840162468, 'gradnorm/compressor.layer5': 4.193752801612794, 'gradnorm/compressor.layer6': 2.700079199724322, 'gradnorm/compressor.layer7': 3.1182259954603433, 'gradnorm/decoder.layer0': 3.030017562916029, 'gradnorm/decoder.layer1': 1.8268116202699634, 'gradnorm/decoder.layer2': 1.0597306793784245, 'gradnorm/max_param': 690.5198974609375, 'epoch': 8.31}

 83%|████████▎ | 15860/19090 [7:44:38<1:41:02,  1.88s/it]
 83%|████████▎ | 15861/19090 [7:44:39<1:35:43,  1.78s/it]
 83%|████████▎ | 15862/19090 [7:44:40<1:27:53,  1.63s/it]
 83%|████████▎ | 15863/19090 [7:44:43<1:39:28,  1.85s/it]
 83%|████████▎ | 15864/19090 [7:44:45<1:37:52,  1.82s/it]
 83%|████████▎ | 15865/19090 [7:44:46<1:29:13,  1.66s/it]
 83%|████████▎ | 15866/19090 [7:44:48<1:40:44,  1.87s/it]
 83%|████████▎ | 15867/19090 [7:44:50<1:36:55,  1.80s/it]
 83%|████████▎ | 15868/19090 [7:44:51<1:28:38,  1.65s/it]
 83%|████████▎ | 15869/19090 [7:44:54<1:46:25,  1.98s/it][gradnorm] step 15869: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 677.7

 83%|████████▎ | 15870/19090 [7:44:55<1:35:15,  1.77s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 968.8732299804688, 'learning_rate': 4.2121486453363755e-06, 'recon_loss': 0.0015881060564424843, 'recon_static': 0.0006082629843149334, 'recon_motion': 0.0009798430764931254, 'recon_norm': 1.306788730621338, 'recon_norm_std': 0.04034981345757842, 'target_norm': 1.6220663249492646, 'recon_cos': 0.6286567911505699, 'gradnorm/compressor.query': 0.41325283060990997, 'gradnorm/compressor.layer0': 17.58016915197659, 'gradnorm/compressor.layer1': 960.4353505051758, 'gradnorm/compressor.layer2': 120.66329136407849, 'gradnorm/compressor.layer3': 35.837404762037735, 'gradnorm/compressor.layer4': 9.076163679184486, 'gradnorm/compressor.layer5': 4.139848053212403, 'gradnorm/compressor.layer6': 2.5418922032415914, 'gradnorm/compressor.layer7': 2.9030819925627305, 'gradnorm/decoder.layer0': 2.7085074931170112, 'gradnorm/decoder.layer1': 1.8445256428227392, 'gradnorm/decoder.layer2': 0.9894723461888035, 'gradnorm/max_param': 677.665283203125, 'epoch': 8.31}

 83%|████████▎ | 15870/19090 [7:44:55<1:35:15,  1.77s/it]
 83%|████████▎ | 15871/19090 [7:44:57<1:27:21,  1.63s/it]
 83%|████████▎ | 15872/19090 [7:45:00<1:51:02,  2.07s/it]
 83%|████████▎ | 15873/19090 [7:45:01<1:38:23,  1.84s/it]
 83%|████████▎ | 15874/19090 [7:45:02<1:29:39,  1.67s/it]
 83%|████████▎ | 15875/19090 [7:45:05<1:46:48,  1.99s/it]
 83%|████████▎ | 15876/19090 [7:45:06<1:35:17,  1.78s/it]
 83%|████████▎ | 15877/19090 [7:45:07<1:27:22,  1.63s/it]
 83%|████████▎ | 15878/19090 [7:45:10<1:44:38,  1.95s/it]
 83%|████████▎ | 15879/19090 [7:45:11<1:33:58,  1.76s/it][gradnorm] step 15879: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 773.6

 83%|████████▎ | 15880/19090 [7:45:13<1:26:31,  1.62s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 1105.6842041015625, 'learning_rate': 4.186789569787999e-06, 'recon_loss': 0.0017152509739389644, 'recon_static': 0.0007317681884160265, 'recon_motion': 0.0009834827476879581, 'recon_norm': 1.3515818059444427, 'recon_norm_std': 0.038085075374692676, 'target_norm': 1.622839370369911, 'recon_cos': 0.6029422879219055, 'gradnorm/compressor.query': 0.7298209842241979, 'gradnorm/compressor.layer0': 21.63096509920551, 'gradnorm/compressor.layer1': 1096.4206537638736, 'gradnorm/compressor.layer2': 134.88873453007483, 'gradnorm/compressor.layer3': 40.4103342870051, 'gradnorm/compressor.layer4': 7.528533572285073, 'gradnorm/compressor.layer5': 4.217556933079226, 'gradnorm/compressor.layer6': 2.592837253510787, 'gradnorm/compressor.layer7': 2.8443788943916806, 'gradnorm/decoder.layer0': 2.9279053824515686, 'gradnorm/decoder.layer1': 1.9452992021343618, 'gradnorm/decoder.layer2': 1.1358007545491524, 'gradnorm/max_param': 773.6246948242188, 'epoch': 8.32}

 83%|████████▎ | 15880/19090 [7:45:13<1:26:31,  1.62s/it]
 83%|████████▎ | 15881/19090 [7:45:16<1:44:50,  1.96s/it]
 83%|████████▎ | 15882/19090 [7:45:17<1:34:17,  1.76s/it]
 83%|████████▎ | 15883/19090 [7:45:18<1:26:54,  1.63s/it]
 83%|████████▎ | 15884/19090 [7:45:21<1:40:14,  1.88s/it]
 83%|████████▎ | 15885/19090 [7:45:22<1:30:55,  1.70s/it]
 83%|████████▎ | 15886/19090 [7:45:23<1:24:17,  1.58s/it]
 83%|████████▎ | 15887/19090 [7:45:26<1:44:16,  1.95s/it]
 83%|████████▎ | 15888/19090 [7:45:27<1:33:37,  1.75s/it]
 83%|████████▎ | 15889/19090 [7:45:29<1:26:05,  1.61s/it][gradnorm] step 15889: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 620.8

 83%|████████▎ | 15890/19090 [7:45:31<1:41:23,  1.90s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 891.795166015625, 'learning_rate': 4.161500083436434e-06, 'recon_loss': 0.0017975043883780017, 'recon_static': 0.0008266489778179676, 'recon_motion': 0.0009708553960081189, 'recon_norm': 1.4001775085926056, 'recon_norm_std': 0.03852072665467858, 'target_norm': 1.6206244260072709, 'recon_cos': 0.5889539688825607, 'gradnorm/compressor.query': 0.43863117915834665, 'gradnorm/compressor.layer0': 17.53078173115702, 'gradnorm/compressor.layer1': 879.7801216781384, 'gradnorm/compressor.layer2': 140.18040134724612, 'gradnorm/compressor.layer3': 34.9918778940869, 'gradnorm/compressor.layer4': 7.429694174714404, 'gradnorm/compressor.layer5': 4.477373459144475, 'gradnorm/compressor.layer6': 2.692983577588957, 'gradnorm/compressor.layer7': 2.7107238685356676, 'gradnorm/decoder.layer0': 2.9508212352706855, 'gradnorm/decoder.layer1': 1.8782102441722217, 'gradnorm/decoder.layer2': 0.9433025398271966, 'gradnorm/max_param': 620.846923828125, 'epoch': 8.32}

 83%|████████▎ | 15890/19090 [7:45:31<1:41:23,  1.90s/it]
 83%|████████▎ | 15891/19090 [7:45:32<1:31:28,  1.72s/it]
 83%|████████▎ | 15892/19090 [7:45:34<1:24:31,  1.59s/it]
 83%|████████▎ | 15893/19090 [7:45:37<1:46:32,  2.00s/it]
 83%|████████▎ | 15894/19090 [7:45:38<1:35:02,  1.78s/it]
 83%|████████▎ | 15895/19090 [7:45:39<1:27:05,  1.64s/it]
 83%|████████▎ | 15896/19090 [7:45:42<1:41:28,  1.91s/it]
 83%|████████▎ | 15897/19090 [7:45:43<1:31:32,  1.72s/it]
 83%|████████▎ | 15898/19090 [7:45:44<1:24:38,  1.59s/it]
 83%|████████▎ | 15899/19090 [7:45:47<1:42:50,  1.93s/it][gradnorm] step 15899: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 628.3

 83%|████████▎ | 15900/19090 [7:45:48<1:32:24,  1.74s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 904.0659790039062, 'learning_rate': 4.136280270837331e-06, 'recon_loss': 0.001805501026683487, 'recon_static': 0.0008185085403965786, 'recon_motion': 0.0009869924309896306, 'recon_norm': 1.3907731592655181, 'recon_norm_std': 0.037579051405191424, 'target_norm': 1.6235451251268387, 'recon_cos': 0.5866400554776192, 'gradnorm/compressor.query': 0.35633396865679845, 'gradnorm/compressor.layer0': 14.014233029923265, 'gradnorm/compressor.layer1': 890.297387112868, 'gradnorm/compressor.layer2': 151.28133069095142, 'gradnorm/compressor.layer3': 38.53940062313664, 'gradnorm/compressor.layer4': 8.76389626168332, 'gradnorm/compressor.layer5': 5.266672273624832, 'gradnorm/compressor.layer6': 2.8181105706781833, 'gradnorm/compressor.layer7': 3.270098674580279, 'gradnorm/decoder.layer0': 3.1004319905187456, 'gradnorm/decoder.layer1': 1.9985646819564953, 'gradnorm/decoder.layer2': 1.0820997195838626, 'gradnorm/max_param': 628.3063354492188, 'epoch': 8.33}

 83%|████████▎ | 15900/19090 [7:45:48<1:32:24,  1.74s/it]
 83%|████████▎ | 15901/19090 [7:45:50<1:25:08,  1.60s/it]
 83%|████████▎ | 15902/19090 [7:45:52<1:37:23,  1.83s/it]
 83%|████████▎ | 15903/19090 [7:45:53<1:28:35,  1.67s/it]
 83%|████████▎ | 15904/19090 [7:45:55<1:22:24,  1.55s/it]
 83%|████████▎ | 15905/19090 [7:45:58<1:44:37,  1.97s/it]
 83%|████████▎ | 15906/19090 [7:45:59<1:33:53,  1.77s/it]
 83%|████████▎ | 15907/19090 [7:46:00<1:26:21,  1.63s/it]
 83%|████████▎ | 15908/19090 [7:46:03<1:45:24,  1.99s/it]
 83%|████████▎ | 15909/19090 [7:46:04<1:34:20,  1.78s/it][gradnorm] step 15909: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 597.1

 83%|████████▎ | 15910/19090 [7:46:06<1:26:45,  1.64s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 855.6466674804688, 'learning_rate': 4.1111302163134105e-06, 'recon_loss': 0.0016559366020374, 'recon_static': 0.0006870121345855295, 'recon_motion': 0.0009689244165201671, 'recon_norm': 1.3376334339380265, 'recon_norm_std': 0.03900890229269862, 'target_norm': 1.6188665091991425, 'recon_cos': 0.6136320799589157, 'gradnorm/compressor.query': 0.4156653312962722, 'gradnorm/compressor.layer0': 15.047105195214773, 'gradnorm/compressor.layer1': 846.1515452328856, 'gradnorm/compressor.layer2': 118.60028802794916, 'gradnorm/compressor.layer3': 41.70966807747111, 'gradnorm/compressor.layer4': 8.401710390616483, 'gradnorm/compressor.layer5': 4.779724276582934, 'gradnorm/compressor.layer6': 2.749356888219063, 'gradnorm/compressor.layer7': 3.1369070944290955, 'gradnorm/decoder.layer0': 3.266515874950512, 'gradnorm/decoder.layer1': 1.9198779142824451, 'gradnorm/decoder.layer2': 1.106810493506624, 'gradnorm/max_param': 597.057373046875, 'epoch': 8.33}

 83%|████████▎ | 15910/19090 [7:46:06<1:26:45,  1.64s/it]
 83%|████████▎ | 15911/19090 [7:46:08<1:38:37,  1.86s/it]
 83%|████████▎ | 15912/19090 [7:46:09<1:29:35,  1.69s/it]
 83%|████████▎ | 15913/19090 [7:46:11<1:23:20,  1.57s/it]
 83%|████████▎ | 15914/19090 [7:46:13<1:42:07,  1.93s/it]
 83%|████████▎ | 15915/19090 [7:46:15<1:31:49,  1.74s/it]
 83%|████████▎ | 15916/19090 [7:46:16<1:24:37,  1.60s/it]
 83%|████████▎ | 15917/19090 [7:46:19<1:44:28,  1.98s/it]
 83%|████████▎ | 15918/19090 [7:46:20<1:33:28,  1.77s/it]
 83%|████████▎ | 15919/19090 [7:46:21<1:25:46,  1.62s/it][gradnorm] step 15919: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 720.6

 83%|████████▎ | 15920/19090 [7:46:24<1:39:15,  1.88s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1030.3685302734375, 'learning_rate': 4.0860500039541756e-06, 'recon_loss': 0.0015641137433703989, 'recon_static': 0.0005860229372046888, 'recon_motion': 0.0009780907508684321, 'recon_norm': 1.2980003207921982, 'recon_norm_std': 0.03996947817504406, 'target_norm': 1.6214881896972657, 'recon_cos': 0.6339879527688026, 'gradnorm/compressor.query': 0.3992078663934376, 'gradnorm/compressor.layer0': 10.5604444681899, 'gradnorm/compressor.layer1': 1021.2447062286296, 'gradnorm/compressor.layer2': 130.87531339504025, 'gradnorm/compressor.layer3': 36.743718239695404, 'gradnorm/compressor.layer4': 8.293330295639873, 'gradnorm/compressor.layer5': 5.1355220430510355, 'gradnorm/compressor.layer6': 2.8644902069675484, 'gradnorm/compressor.layer7': 3.1775046314172735, 'gradnorm/decoder.layer0': 3.0644469008859674, 'gradnorm/decoder.layer1': 2.0538855318379032, 'gradnorm/decoder.layer2': 1.1719241831313132, 'gradnorm/max_param': 720.6256713867188, 'epoch': 8.34}

 83%|████████▎ | 15920/19090 [7:46:24<1:39:15,  1.88s/it]
 83%|████████▎ | 15921/19090 [7:46:25<1:29:51,  1.70s/it]
 83%|████████▎ | 15922/19090 [7:46:26<1:23:12,  1.58s/it]
 83%|████████▎ | 15923/19090 [7:46:29<1:45:43,  2.00s/it]
 83%|████████▎ | 15924/19090 [7:46:31<1:34:18,  1.79s/it]
 83%|████████▎ | 15925/19090 [7:46:32<1:26:21,  1.64s/it]
 83%|████████▎ | 15926/19090 [7:46:35<1:49:59,  2.09s/it]
 83%|████████▎ | 15927/19090 [7:46:36<1:37:15,  1.84s/it]
 83%|████████▎ | 15928/19090 [7:46:38<1:28:20,  1.68s/it]
 83%|████████▎ | 15929/19090 [7:46:40<1:38:46,  1.87s/it][gradnorm] step 15929: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 689.3

 83%|████████▎ | 15930/19090 [7:46:41<1:29:28,  1.70s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 986.2630615234375, 'learning_rate': 4.0610397176155724e-06, 'recon_loss': 0.0015392473695101216, 'recon_static': 0.0005650658160448075, 'recon_motion': 0.0009741815869347192, 'recon_norm': 1.2809574782848359, 'recon_norm_std': 0.041481124609708785, 'target_norm': 1.620512354373932, 'recon_cos': 0.6378122955560684, 'gradnorm/compressor.query': 0.3485304417696139, 'gradnorm/compressor.layer0': 11.10905325164792, 'gradnorm/compressor.layer1': 976.7649090236606, 'gradnorm/compressor.layer2': 130.26734829284734, 'gradnorm/compressor.layer3': 37.765169241147326, 'gradnorm/compressor.layer4': 8.26265785885983, 'gradnorm/compressor.layer5': 5.051531931356628, 'gradnorm/compressor.layer6': 2.8960103774438046, 'gradnorm/compressor.layer7': 3.151581890782253, 'gradnorm/decoder.layer0': 3.097293250424631, 'gradnorm/decoder.layer1': 1.7626746855398039, 'gradnorm/decoder.layer2': 1.0352488350399918, 'gradnorm/max_param': 689.3339233398438, 'epoch': 8.34}

 83%|████████▎ | 15930/19090 [7:46:41<1:29:28,  1.70s/it]
 83%|████████▎ | 15931/19090 [7:46:43<1:22:53,  1.57s/it]
 83%|████████▎ | 15932/19090 [7:46:45<1:37:14,  1.85s/it]
 83%|████████▎ | 15933/19090 [7:46:46<1:28:16,  1.68s/it]
 83%|████████▎ | 15934/19090 [7:46:48<1:21:59,  1.56s/it]
 83%|████████▎ | 15935/19090 [7:46:51<1:45:12,  2.00s/it]
 83%|████████▎ | 15936/19090 [7:46:52<1:33:50,  1.79s/it]
 83%|████████▎ | 15937/19090 [7:46:53<1:25:53,  1.63s/it]
 83%|████████▎ | 15938/19090 [7:46:56<1:39:51,  1.90s/it]
 83%|████████▎ | 15939/19090 [7:46:57<1:30:01,  1.71s/it][gradnorm] step 15939: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 654.1

 83%|████████▎ | 15940/19090 [7:46:58<1:23:13,  1.59s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 938.073486328125, 'learning_rate': 4.036099440919763e-06, 'recon_loss': 0.0015564757050015032, 'recon_static': 0.0005720601897337474, 'recon_motion': 0.0009844155021710322, 'recon_norm': 1.2866668343544005, 'recon_norm_std': 0.04020013585686684, 'target_norm': 1.621576452255249, 'recon_cos': 0.6345912098884583, 'gradnorm/compressor.query': 0.36286401887231695, 'gradnorm/compressor.layer0': 12.561980734953845, 'gradnorm/compressor.layer1': 926.7619705188598, 'gradnorm/compressor.layer2': 138.88980850389996, 'gradnorm/compressor.layer3': 39.05597445494095, 'gradnorm/compressor.layer4': 8.062165482552837, 'gradnorm/compressor.layer5': 4.985610282678247, 'gradnorm/compressor.layer6': 2.9657912116307728, 'gradnorm/compressor.layer7': 2.909868182575599, 'gradnorm/decoder.layer0': 2.993794540042092, 'gradnorm/decoder.layer1': 1.8311900991596295, 'gradnorm/decoder.layer2': 1.018795113061987, 'gradnorm/max_param': 654.0653686523438, 'epoch': 8.35}

 83%|████████▎ | 15940/19090 [7:46:58<1:23:13,  1.59s/it]
 84%|████████▎ | 15941/19090 [7:47:01<1:45:33,  2.01s/it]
 84%|████████▎ | 15942/19090 [7:47:03<1:34:01,  1.79s/it]
 84%|████████▎ | 15943/19090 [7:47:04<1:25:59,  1.64s/it]
 84%|████████▎ | 15944/19090 [7:47:07<1:49:19,  2.09s/it]
 84%|████████▎ | 15945/19090 [7:47:08<1:36:39,  1.84s/it]
 84%|████████▎ | 15946/19090 [7:47:10<1:27:48,  1.68s/it]
 84%|████████▎ | 15947/19090 [7:47:12<1:43:30,  1.98s/it]
 84%|████████▎ | 15948/19090 [7:47:14<1:32:46,  1.77s/it]
 84%|████████▎ | 15949/19090 [7:47:15<1:25:05,  1.63s/it][gradnorm] step 15949: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 714.4

 84%|████████▎ | 15950/19090 [7:47:17<1:39:02,  1.89s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1022.279296875, 'learning_rate': 4.011229257254859e-06, 'recon_loss': 0.0016010761668439955, 'recon_static': 0.0006291336874710395, 'recon_motion': 0.0009719425113871693, 'recon_norm': 1.3178123891353608, 'recon_norm_std': 0.03929794682189822, 'target_norm': 1.6193138152360915, 'recon_cos': 0.6253690719604492, 'gradnorm/compressor.query': 0.381666960730016, 'gradnorm/compressor.layer0': 11.827259994584214, 'gradnorm/compressor.layer1': 1012.635669922801, 'gradnorm/compressor.layer2': 134.71212796877273, 'gradnorm/compressor.layer3': 34.986543520889036, 'gradnorm/compressor.layer4': 7.546007568549701, 'gradnorm/compressor.layer5': 4.76524416543127, 'gradnorm/compressor.layer6': 2.984661787543599, 'gradnorm/compressor.layer7': 3.1701423911476545, 'gradnorm/decoder.layer0': 3.1217758330975363, 'gradnorm/decoder.layer1': 1.7968793786037005, 'gradnorm/decoder.layer2': 1.0115205430694494, 'gradnorm/max_param': 714.406494140625, 'epoch': 8.36}

 84%|████████▎ | 15950/19090 [7:47:17<1:39:02,  1.89s/it]
 84%|████████▎ | 15951/19090 [7:47:19<1:29:28,  1.71s/it]
 84%|████████▎ | 15952/19090 [7:47:20<1:22:47,  1.58s/it]
 84%|████████▎ | 15953/19090 [7:47:23<1:40:48,  1.93s/it]
 84%|████████▎ | 15954/19090 [7:47:24<1:30:40,  1.73s/it]
 84%|████████▎ | 15955/19090 [7:47:25<1:23:34,  1.60s/it]
 84%|████████▎ | 15956/19090 [7:47:28<1:40:25,  1.92s/it]
 84%|████████▎ | 15957/19090 [7:47:29<1:30:22,  1.73s/it]
 84%|████████▎ | 15958/19090 [7:47:30<1:23:25,  1.60s/it]
 84%|████████▎ | 15959/19090 [7:47:34<1:46:32,  2.04s/it][gradnorm] step 15959: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1044.8

 84%|████████▎ | 15960/19090 [7:47:35<1:34:54,  1.82s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1490.39697265625, 'learning_rate': 3.986429249774568e-06, 'recon_loss': 0.0015851849952014164, 'recon_static': 0.0006117933327914216, 'recon_motion': 0.0009733916478580796, 'recon_norm': 1.3015020072460175, 'recon_norm_std': 0.04006344759836793, 'target_norm': 1.6205268204212189, 'recon_cos': 0.6276398718357086, 'gradnorm/compressor.query': 0.6857823241984398, 'gradnorm/compressor.layer0': 21.21308690766961, 'gradnorm/compressor.layer1': 1480.8766829145497, 'gradnorm/compressor.layer2': 160.28656588474593, 'gradnorm/compressor.layer3': 44.84186761075315, 'gradnorm/compressor.layer4': 8.362869359456708, 'gradnorm/compressor.layer5': 5.506056831773242, 'gradnorm/compressor.layer6': 2.972739179660006, 'gradnorm/compressor.layer7': 3.318690435589826, 'gradnorm/decoder.layer0': 3.1071017596152997, 'gradnorm/decoder.layer1': 1.880574776259967, 'gradnorm/decoder.layer2': 1.076777293089413, 'gradnorm/max_param': 1044.770263671875, 'epoch': 8.36}

 84%|████████▎ | 15960/19090 [7:47:35<1:34:54,  1.82s/it]
 84%|████████▎ | 15961/19090 [7:47:36<1:26:32,  1.66s/it]
 84%|████████▎ | 15962/19090 [7:47:39<1:41:38,  1.95s/it]
 84%|████████▎ | 15963/19090 [7:47:40<1:31:11,  1.75s/it]
 84%|████████▎ | 15964/19090 [7:47:41<1:24:01,  1.61s/it]
 84%|████████▎ | 15965/19090 [7:47:44<1:46:29,  2.04s/it]
 84%|████████▎ | 15966/19090 [7:47:46<1:34:37,  1.82s/it]
 84%|████████▎ | 15967/19090 [7:47:47<1:26:31,  1.66s/it]
 84%|████████▎ | 15968/19090 [7:47:49<1:40:23,  1.93s/it]
 84%|████████▎ | 15969/19090 [7:47:51<1:30:19,  1.74s/it][gradnorm] step 15969: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 747.7

 84%|████████▎ | 15970/19090 [7:47:52<1:23:25,  1.60s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1068.083251953125, 'learning_rate': 3.961699501397989e-06, 'recon_loss': 0.0016223134065512569, 'recon_static': 0.0006434112743590959, 'recon_motion': 0.0009789021758479066, 'recon_norm': 1.3156360059976577, 'recon_norm_std': 0.039397585578262805, 'target_norm': 1.620689034461975, 'recon_cos': 0.6201847240328788, 'gradnorm/compressor.query': 0.3602522839917878, 'gradnorm/compressor.layer0': 12.559486748966334, 'gradnorm/compressor.layer1': 1059.4322772126588, 'gradnorm/compressor.layer2': 129.3947297757911, 'gradnorm/compressor.layer3': 37.26049165377787, 'gradnorm/compressor.layer4': 7.7870052762198, 'gradnorm/compressor.layer5': 4.899180378776542, 'gradnorm/compressor.layer6': 2.9767749139882587, 'gradnorm/compressor.layer7': 3.1452596448691765, 'gradnorm/decoder.layer0': 2.9063771537939145, 'gradnorm/decoder.layer1': 1.7492866424565205, 'gradnorm/decoder.layer2': 1.0187531640875422, 'gradnorm/max_param': 747.6946411132812, 'epoch': 8.37}

 84%|████████▎ | 15970/19090 [7:47:52<1:23:25,  1.60s/it]
 84%|████████▎ | 15971/19090 [7:47:55<1:39:01,  1.90s/it]
 84%|████████▎ | 15972/19090 [7:47:56<1:29:17,  1.72s/it]
 84%|████████▎ | 15973/19090 [7:47:57<1:22:38,  1.59s/it]
 84%|████████▎ | 15974/19090 [7:48:00<1:39:21,  1.91s/it]
 84%|████████▎ | 15975/19090 [7:48:01<1:29:37,  1.73s/it]
 84%|████████▎ | 15976/19090 [7:48:03<1:22:51,  1.60s/it]
 84%|████████▎ | 15977/19090 [7:48:06<1:45:30,  2.03s/it]
 84%|████████▎ | 15978/19090 [7:48:07<1:33:57,  1.81s/it]
 84%|████████▎ | 15979/19090 [7:48:08<1:25:49,  1.66s/it][gradnorm] step 15979: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 690.2

 84%|████████▎ | 15980/19090 [7:48:11<1:38:10,  1.89s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 989.80078125, 'learning_rate': 3.937040094809327e-06, 'recon_loss': 0.0016562034463277087, 'recon_static': 0.0006788402184611187, 'recon_motion': 0.0009773632438736968, 'recon_norm': 1.3436647355556488, 'recon_norm_std': 0.038229058310389516, 'target_norm': 1.6192319095134735, 'recon_cos': 0.6145351991057396, 'gradnorm/compressor.query': 0.47862323407385116, 'gradnorm/compressor.layer0': 16.189945934598867, 'gradnorm/compressor.layer1': 978.3352569543837, 'gradnorm/compressor.layer2': 143.07283437990597, 'gradnorm/compressor.layer3': 41.3376672585065, 'gradnorm/compressor.layer4': 8.064792041119468, 'gradnorm/compressor.layer5': 5.252263444079204, 'gradnorm/compressor.layer6': 3.0031267402318385, 'gradnorm/compressor.layer7': 3.020272897933351, 'gradnorm/decoder.layer0': 3.0733992030710864, 'gradnorm/decoder.layer1': 1.8417144789322175, 'gradnorm/decoder.layer2': 1.0758154637167314, 'gradnorm/max_param': 690.1727294921875, 'epoch': 8.37}

 84%|████████▎ | 15980/19090 [7:48:11<1:38:10,  1.89s/it]
 84%|████████▎ | 15981/19090 [7:48:12<1:28:52,  1.72s/it]
 84%|████████▎ | 15982/19090 [7:48:13<1:23:23,  1.61s/it]
 84%|████████▎ | 15983/19090 [7:48:16<1:36:05,  1.86s/it]
 84%|████████▎ | 15984/19090 [7:48:17<1:28:43,  1.71s/it]
 84%|████████▎ | 15985/19090 [7:48:19<1:28:05,  1.70s/it]
 84%|████████▎ | 15986/19090 [7:48:21<1:37:52,  1.89s/it]
 84%|████████▎ | 15987/19090 [7:48:22<1:28:31,  1.71s/it]
 84%|████████▍ | 15988/19090 [7:48:24<1:27:49,  1.70s/it]
 84%|████████▍ | 15989/19090 [7:48:26<1:33:34,  1.81s/it][gradnorm] step 15989: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 674.8

 84%|████████▍ | 15990/19090 [7:48:27<1:25:34,  1.66s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 968.3140258789062, 'learning_rate': 3.912451112457558e-06, 'recon_loss': 0.0016415485326433555, 'recon_static': 0.0006497655936982483, 'recon_motion': 0.0009917829491314478, 'recon_norm': 1.3229096710681916, 'recon_norm_std': 0.038654162921011447, 'target_norm': 1.622202953696251, 'recon_cos': 0.6167772531509399, 'gradnorm/compressor.query': 0.5335131851075701, 'gradnorm/compressor.layer0': 16.60468261121364, 'gradnorm/compressor.layer1': 956.3808733972047, 'gradnorm/compressor.layer2': 145.3667506257879, 'gradnorm/compressor.layer3': 37.58161041528489, 'gradnorm/compressor.layer4': 9.342705607653977, 'gradnorm/compressor.layer5': 4.753315621691248, 'gradnorm/compressor.layer6': 3.358362515482864, 'gradnorm/compressor.layer7': 3.370823430509289, 'gradnorm/decoder.layer0': 3.2011729176992656, 'gradnorm/decoder.layer1': 1.9163042914556059, 'gradnorm/decoder.layer2': 1.1302532357057977, 'gradnorm/max_param': 674.8468017578125, 'epoch': 8.38}

 84%|████████▍ | 15990/19090 [7:48:27<1:25:34,  1.66s/it]
 84%|████████▍ | 15991/19090 [7:48:29<1:21:07,  1.57s/it]
 84%|████████▍ | 15992/19090 [7:48:31<1:28:23,  1.71s/it]
 84%|████████▍ | 15993/19090 [7:48:32<1:21:53,  1.59s/it]
 84%|████████▍ | 15994/19090 [7:48:34<1:29:30,  1.73s/it]
 84%|████████▍ | 15995/19090 [7:48:36<1:30:18,  1.75s/it]
 84%|████████▍ | 15996/19090 [7:48:37<1:24:33,  1.64s/it]
 84%|████████▍ | 15997/19090 [7:48:40<1:32:22,  1.79s/it]
 84%|████████▍ | 15998/19090 [7:48:42<1:36:46,  1.88s/it]
 84%|████████▍ | 15999/19090 [7:48:43<1:32:29,  1.80s/it][gradnorm] step 15999: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 698.6

 84%|████████▍ | 16000/19090 [7:48:45<1:27:13,  1.69s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1002.3013916015625, 'learning_rate': 3.887932636556216e-06, 'recon_loss': 0.0016085610463051126, 'recon_static': 0.0006244768912438304, 'recon_motion': 0.0009840841972618363, 'recon_norm': 1.3107569128274918, 'recon_norm_std': 0.039479071833193304, 'target_norm': 1.6223484575748444, 'recon_cos': 0.62406025826931, 'gradnorm/compressor.query': 0.6711393587292623, 'gradnorm/compressor.layer0': 21.94327862482361, 'gradnorm/compressor.layer1': 990.1254339728881, 'gradnorm/compressor.layer2': 147.4255505967334, 'gradnorm/compressor.layer3': 43.651752128261556, 'gradnorm/compressor.layer4': 8.751166129743572, 'gradnorm/compressor.layer5': 5.104769529694161, 'gradnorm/compressor.layer6': 3.0533513027256536, 'gradnorm/compressor.layer7': 3.1805819066569585, 'gradnorm/decoder.layer0': 3.2121129912363355, 'gradnorm/decoder.layer1': 1.9579843115406983, 'gradnorm/decoder.layer2': 1.3427643820717385, 'gradnorm/max_param': 698.625244140625, 'epoch': 8.38}

 84%|████████▍ | 16000/19090 [7:48:45<1:27:13,  1.69s/it]
 84%|████████▍ | 16001/19090 [7:48:49<2:02:45,  2.38s/it]
 84%|████████▍ | 16002/19090 [7:48:50<1:45:55,  2.06s/it]
 84%|████████▍ | 16003/19090 [7:48:51<1:34:03,  1.83s/it]
 84%|████████▍ | 16004/19090 [7:48:53<1:25:45,  1.67s/it]
 84%|████████▍ | 16005/19090 [7:48:54<1:19:51,  1.55s/it]
 84%|████████▍ | 16006/19090 [7:48:55<1:15:49,  1.48s/it]
 84%|████████▍ | 16007/19090 [7:48:56<1:12:53,  1.42s/it]
 84%|████████▍ | 16008/19090 [7:48:58<1:14:21,  1.45s/it]
 84%|████████▍ | 16009/19090 [7:49:00<1:17:28,  1.51s/it][gradnorm] step 16009: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 666.6

 84%|████████▍ | 16010/19090 [7:49:02<1:26:55,  1.69s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 954.6251220703125, 'learning_rate': 3.8634847490830965e-06, 'recon_loss': 0.0016540903190616519, 'recon_static': 0.0006726630075718276, 'recon_motion': 0.0009814273071242496, 'recon_norm': 1.3253660887479781, 'recon_norm_std': 0.039695034921169284, 'target_norm': 1.621180784702301, 'recon_cos': 0.6135401010513306, 'gradnorm/compressor.query': 0.5623302468482929, 'gradnorm/compressor.layer0': 15.864933216280214, 'gradnorm/compressor.layer1': 944.6228414028533, 'gradnorm/compressor.layer2': 132.53652208919624, 'gradnorm/compressor.layer3': 32.52153733489547, 'gradnorm/compressor.layer4': 8.474074375613982, 'gradnorm/compressor.layer5': 4.167751603337707, 'gradnorm/compressor.layer6': 2.9994274228779574, 'gradnorm/compressor.layer7': 2.868172707575497, 'gradnorm/decoder.layer0': 3.1768772913091317, 'gradnorm/decoder.layer1': 1.8880409678120003, 'gradnorm/decoder.layer2': 1.065424260895753, 'gradnorm/max_param': 666.5531005859375, 'epoch': 8.39}

 84%|████████▍ | 16010/19090 [7:49:02<1:26:55,  1.69s/it]
 84%|████████▍ | 16011/19090 [7:49:03<1:25:03,  1.66s/it]
 84%|████████▍ | 16012/19090 [7:49:05<1:22:36,  1.61s/it]
 84%|████████▍ | 16013/19090 [7:49:07<1:34:09,  1.84s/it]
 84%|████████▍ | 16014/19090 [7:49:08<1:26:39,  1.69s/it]
 84%|████████▍ | 16015/19090 [7:49:10<1:27:58,  1.72s/it]
 84%|████████▍ | 16016/19090 [7:49:13<1:37:28,  1.90s/it]
 84%|████████▍ | 16017/19090 [7:49:14<1:28:02,  1.72s/it]
 84%|████████▍ | 16018/19090 [7:49:15<1:23:19,  1.63s/it]
 84%|████████▍ | 16019/19090 [7:49:18<1:39:05,  1.94s/it][gradnorm] step 16019: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 568.1

 84%|████████▍ | 16020/19090 [7:49:19<1:29:17,  1.75s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 812.6390991210938, 'learning_rate': 3.839107531779978e-06, 'recon_loss': 0.0017592598625924439, 'recon_static': 0.0007856012452975847, 'recon_motion': 0.0009736586245708168, 'recon_norm': 1.3802930563688278, 'recon_norm_std': 0.037800658214837315, 'target_norm': 1.6193505972623825, 'recon_cos': 0.5943256765604019, 'gradnorm/compressor.query': 0.36792591598882984, 'gradnorm/compressor.layer0': 12.885179957952785, 'gradnorm/compressor.layer1': 804.9283353193624, 'gradnorm/compressor.layer2': 106.05354540130895, 'gradnorm/compressor.layer3': 31.050425218505463, 'gradnorm/compressor.layer4': 7.29277233641844, 'gradnorm/compressor.layer5': 3.8752090643813752, 'gradnorm/compressor.layer6': 2.5917061715830707, 'gradnorm/compressor.layer7': 2.8108624247374294, 'gradnorm/decoder.layer0': 2.887910465279135, 'gradnorm/decoder.layer1': 1.7326656544193033, 'gradnorm/decoder.layer2': 0.974253159368853, 'gradnorm/max_param': 568.0546875, 'epoch': 8.39}

 84%|████████▍ | 16020/19090 [7:49:19<1:29:17,  1.75s/it]
 84%|████████▍ | 16021/19090 [7:49:21<1:22:16,  1.61s/it]
 84%|████████▍ | 16022/19090 [7:49:23<1:36:39,  1.89s/it]
 84%|████████▍ | 16023/19090 [7:49:24<1:27:24,  1.71s/it]
 84%|████████▍ | 16024/19090 [7:49:26<1:21:52,  1.60s/it]
 84%|████████▍ | 16025/19090 [7:49:29<1:45:13,  2.06s/it]
 84%|████████▍ | 16026/19090 [7:49:30<1:33:21,  1.83s/it]
 84%|████████▍ | 16027/19090 [7:49:31<1:25:04,  1.67s/it]
 84%|████████▍ | 16028/19090 [7:49:34<1:43:16,  2.02s/it]
 84%|████████▍ | 16029/19090 [7:49:36<1:31:57,  1.80s/it][gradnorm] step 16029: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 753.8

 84%|████████▍ | 16030/19090 [7:49:37<1:24:11,  1.65s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1078.5782470703125, 'learning_rate': 3.814801066152354e-06, 'recon_loss': 0.0016391134558944032, 'recon_static': 0.000661061450955458, 'recon_motion': 0.0009780519787454978, 'recon_norm': 1.3287306129932404, 'recon_norm_std': 0.03939701179042458, 'target_norm': 1.621971347928047, 'recon_cos': 0.6182594195008277, 'gradnorm/compressor.query': 0.517709564217755, 'gradnorm/compressor.layer0': 17.73085386299649, 'gradnorm/compressor.layer1': 1068.241662733672, 'gradnorm/compressor.layer2': 144.1312674808107, 'gradnorm/compressor.layer3': 31.23922568168754, 'gradnorm/compressor.layer4': 9.016369661621024, 'gradnorm/compressor.layer5': 3.990886916823569, 'gradnorm/compressor.layer6': 2.6561309787787617, 'gradnorm/compressor.layer7': 2.852309037819886, 'gradnorm/decoder.layer0': 3.1015309829455817, 'gradnorm/decoder.layer1': 1.8203099459519703, 'gradnorm/decoder.layer2': 1.0434112753004605, 'gradnorm/max_param': 753.7860107421875, 'epoch': 8.4}

 84%|████████▍ | 16030/19090 [7:49:37<1:24:11,  1.65s/it]
 84%|████████▍ | 16031/19090 [7:49:40<1:40:30,  1.97s/it]
 84%|████████▍ | 16032/19090 [7:49:41<1:29:55,  1.76s/it]
 84%|████████▍ | 16033/19090 [7:49:42<1:22:40,  1.62s/it]
 84%|████████▍ | 16034/19090 [7:49:45<1:34:10,  1.85s/it]
 84%|████████▍ | 16035/19090 [7:49:46<1:25:27,  1.68s/it]
 84%|████████▍ | 16036/19090 [7:49:47<1:19:22,  1.56s/it]
 84%|████████▍ | 16037/19090 [7:49:50<1:37:54,  1.92s/it]
 84%|████████▍ | 16038/19090 [7:49:51<1:29:05,  1.75s/it]
 84%|████████▍ | 16039/19090 [7:49:53<1:22:00,  1.61s/it][gradnorm] step 16039: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 450.8

 84%|████████▍ | 16040/19090 [7:49:55<1:38:15,  1.93s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 648.9190673828125, 'learning_rate': 3.790565433469154e-06, 'recon_loss': 0.0016206007072469219, 'recon_static': 0.0006447738720453345, 'recon_motion': 0.0009758268613950349, 'recon_norm': 1.3201242923736571, 'recon_norm_std': 0.03897027801722288, 'target_norm': 1.6206792086362838, 'recon_cos': 0.6214868783950805, 'gradnorm/compressor.query': 0.5916618218245964, 'gradnorm/compressor.layer0': 21.684662998880107, 'gradnorm/compressor.layer1': 639.5297833173995, 'gradnorm/compressor.layer2': 101.53270235957231, 'gradnorm/compressor.layer3': 34.34398991897545, 'gradnorm/compressor.layer4': 9.321918000733666, 'gradnorm/compressor.layer5': 4.574102828644293, 'gradnorm/compressor.layer6': 2.675102292538215, 'gradnorm/compressor.layer7': 3.1205288754218103, 'gradnorm/decoder.layer0': 3.1822072385752507, 'gradnorm/decoder.layer1': 1.7010489593492808, 'gradnorm/decoder.layer2': 0.9229284183853355, 'gradnorm/max_param': 450.81707763671875, 'epoch': 8.4}

 84%|████████▍ | 16040/19090 [7:49:55<1:38:15,  1.93s/it]
 84%|████████▍ | 16041/19090 [7:49:57<1:31:06,  1.79s/it]
 84%|████████▍ | 16042/19090 [7:49:58<1:23:25,  1.64s/it]
 84%|████████▍ | 16043/19090 [7:50:00<1:35:58,  1.89s/it]
 84%|████████▍ | 16044/19090 [7:50:02<1:31:44,  1.81s/it]
 84%|████████▍ | 16045/19090 [7:50:03<1:23:41,  1.65s/it]
 84%|████████▍ | 16046/19090 [7:50:06<1:33:37,  1.85s/it]
 84%|████████▍ | 16047/19090 [7:50:07<1:32:10,  1.82s/it]
 84%|████████▍ | 16048/19090 [7:50:09<1:24:06,  1.66s/it]
 84%|████████▍ | 16049/19090 [7:50:11<1:34:42,  1.87s/it][gradnorm] step 16049: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 625.3

 84%|████████▍ | 16050/19090 [7:50:13<1:30:23,  1.78s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 895.36767578125, 'learning_rate': 3.7664007147624943e-06, 'recon_loss': 0.0017566681635798886, 'recon_static': 0.0007882582503953017, 'recon_motion': 0.0009684099015430548, 'recon_norm': 1.3753373712301253, 'recon_norm_std': 0.038157024700194594, 'target_norm': 1.618985664844513, 'recon_cos': 0.5936844393610954, 'gradnorm/compressor.query': 0.38166596514971923, 'gradnorm/compressor.layer0': 12.315670321733723, 'gradnorm/compressor.layer1': 886.0520230212219, 'gradnorm/compressor.layer2': 123.28343629240102, 'gradnorm/compressor.layer3': 33.740354162685335, 'gradnorm/compressor.layer4': 7.766984847377055, 'gradnorm/compressor.layer5': 4.25396229660622, 'gradnorm/compressor.layer6': 2.654563278516072, 'gradnorm/compressor.layer7': 2.8561997423821377, 'gradnorm/decoder.layer0': 2.955777546266826, 'gradnorm/decoder.layer1': 1.715830108666721, 'gradnorm/decoder.layer2': 0.9074192725286716, 'gradnorm/max_param': 625.2548828125, 'epoch': 8.41}

 84%|████████▍ | 16050/19090 [7:50:13<1:30:23,  1.78s/it]
 84%|████████▍ | 16051/19090 [7:50:14<1:22:43,  1.63s/it]
 84%|████████▍ | 16052/19090 [7:50:16<1:33:14,  1.84s/it]
 84%|████████▍ | 16053/19090 [7:50:18<1:34:54,  1.88s/it]
 84%|████████▍ | 16054/19090 [7:50:19<1:25:51,  1.70s/it]
 84%|████████▍ | 16055/19090 [7:50:22<1:31:52,  1.82s/it]
 84%|████████▍ | 16056/19090 [7:50:24<1:34:56,  1.88s/it]
 84%|████████▍ | 16057/19090 [7:50:25<1:25:54,  1.70s/it]
 84%|████████▍ | 16058/19090 [7:50:27<1:30:33,  1.79s/it]
 84%|████████▍ | 16059/19090 [7:50:29<1:35:54,  1.90s/it][gradnorm] step 16059: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 640.9

 84%|████████▍ | 16060/19090 [7:50:30<1:26:43,  1.72s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 919.808349609375, 'learning_rate': 3.74230699082736e-06, 'recon_loss': 0.0017197471694089471, 'recon_static': 0.0007641497635631822, 'recon_motion': 0.0009555974538670853, 'recon_norm': 1.3723444879055022, 'recon_norm_std': 0.038412649743258954, 'target_norm': 1.61864815056324, 'recon_cos': 0.6026203915476799, 'gradnorm/compressor.query': 0.35179713154952225, 'gradnorm/compressor.layer0': 10.786516136310496, 'gradnorm/compressor.layer1': 908.0988382329316, 'gradnorm/compressor.layer2': 140.73062005787867, 'gradnorm/compressor.layer3': 36.746164965182935, 'gradnorm/compressor.layer4': 8.965220325015679, 'gradnorm/compressor.layer5': 4.366387090881544, 'gradnorm/compressor.layer6': 3.139072926695289, 'gradnorm/compressor.layer7': 2.931944280524294, 'gradnorm/decoder.layer0': 3.145645607574405, 'gradnorm/decoder.layer1': 1.7526303006434765, 'gradnorm/decoder.layer2': 1.0766118806495617, 'gradnorm/max_param': 640.917236328125, 'epoch': 8.41}

 84%|████████▍ | 16060/19090 [7:50:30<1:26:43,  1.72s/it]
 84%|████████▍ | 16061/19090 [7:50:32<1:30:26,  1.79s/it]
 84%|████████▍ | 16062/19090 [7:50:35<1:40:03,  1.98s/it]
 84%|████████▍ | 16063/19090 [7:50:36<1:29:26,  1.77s/it]
 84%|████████▍ | 16064/19090 [7:50:38<1:30:44,  1.80s/it]
 84%|████████▍ | 16065/19090 [7:50:40<1:34:48,  1.88s/it]
 84%|████████▍ | 16066/19090 [7:50:41<1:25:52,  1.70s/it]
 84%|████████▍ | 16067/19090 [7:50:43<1:31:45,  1.82s/it]
 84%|████████▍ | 16068/19090 [7:50:45<1:34:39,  1.88s/it]
 84%|████████▍ | 16069/19090 [7:50:47<1:25:44,  1.70s/it][gradnorm] step 16069: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 599.7

 84%|████████▍ | 16070/19090 [7:50:49<1:30:49,  1.80s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 859.9435424804688, 'learning_rate': 3.718284342221384e-06, 'recon_loss': 0.0015781295689521358, 'recon_static': 0.0005927349498961121, 'recon_motion': 0.0009853946263319812, 'recon_norm': 1.300644251704216, 'recon_norm_std': 0.03891995288431645, 'target_norm': 1.6221973210573197, 'recon_cos': 0.6312249600887299, 'gradnorm/compressor.query': 0.5893463412180908, 'gradnorm/compressor.layer0': 11.071699379979774, 'gradnorm/compressor.layer1': 849.7202113048742, 'gradnorm/compressor.layer2': 126.06147743759789, 'gradnorm/compressor.layer3': 36.373410953800615, 'gradnorm/compressor.layer4': 9.334528119951644, 'gradnorm/compressor.layer5': 4.533380678341039, 'gradnorm/compressor.layer6': 3.1010841750278098, 'gradnorm/compressor.layer7': 2.932608569802939, 'gradnorm/decoder.layer0': 3.2137986404494883, 'gradnorm/decoder.layer1': 1.9028211304056504, 'gradnorm/decoder.layer2': 1.1088519945507118, 'gradnorm/max_param': 599.7063598632812, 'epoch': 8.42}

 84%|████████▍ | 16070/19090 [7:50:49<1:30:49,  1.80s/it]
 84%|████████▍ | 16071/19090 [7:50:50<1:31:16,  1.81s/it]
 84%|████████▍ | 16072/19090 [7:50:52<1:25:35,  1.70s/it]
 84%|████████▍ | 16073/19090 [7:50:54<1:28:36,  1.76s/it]
 84%|████████▍ | 16074/19090 [7:50:56<1:31:32,  1.82s/it]
 84%|████████▍ | 16075/19090 [7:50:57<1:23:20,  1.66s/it]
 84%|████████▍ | 16076/19090 [7:50:59<1:30:45,  1.81s/it]
 84%|████████▍ | 16077/19090 [7:51:01<1:32:04,  1.83s/it]
 84%|████████▍ | 16078/19090 [7:51:03<1:26:22,  1.72s/it]
 84%|████████▍ | 16079/19090 [7:51:04<1:27:47,  1.75s/it][gradnorm] step 16079: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 795.2

 84%|████████▍ | 16080/19090 [7:51:07<1:34:08,  1.88s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1137.3470458984375, 'learning_rate': 3.6943328492645703e-06, 'recon_loss': 0.001559746521525085, 'recon_static': 0.0005845963460160419, 'recon_motion': 0.0009751501391292549, 'recon_norm': 1.2900217682123185, 'recon_norm_std': 0.03986751437187195, 'target_norm': 1.6208215326070785, 'recon_cos': 0.6333746418356896, 'gradnorm/compressor.query': 0.46261929958060755, 'gradnorm/compressor.layer0': 13.868891290578452, 'gradnorm/compressor.layer1': 1126.8422693527252, 'gradnorm/compressor.layer2': 148.84634484931095, 'gradnorm/compressor.layer3': 36.147717365044294, 'gradnorm/compressor.layer4': 9.12959362049007, 'gradnorm/compressor.layer5': 4.2778435959369565, 'gradnorm/compressor.layer6': 2.6562745934638134, 'gradnorm/compressor.layer7': 2.878239837906869, 'gradnorm/decoder.layer0': 2.991991479892125, 'gradnorm/decoder.layer1': 1.8601755935066933, 'gradnorm/decoder.layer2': 1.0428037477194017, 'gradnorm/max_param': 795.1531982421875, 'epoch': 8.42}

 84%|████████▍ | 16080/19090 [7:51:07<1:34:08,  1.88s/it]
 84%|████████▍ | 16081/19090 [7:51:08<1:25:17,  1.70s/it]
 84%|████████▍ | 16082/19090 [7:51:10<1:26:47,  1.73s/it]
 84%|████████▍ | 16083/19090 [7:51:12<1:36:20,  1.92s/it]
 84%|████████▍ | 16084/19090 [7:51:13<1:26:41,  1.73s/it]
 84%|████████▍ | 16085/19090 [7:51:15<1:25:53,  1.72s/it]
 84%|████████▍ | 16086/19090 [7:51:17<1:35:44,  1.91s/it]
 84%|████████▍ | 16087/19090 [7:51:19<1:26:23,  1.73s/it]
 84%|████████▍ | 16088/19090 [7:51:20<1:26:49,  1.74s/it]
 84%|████████▍ | 16089/19090 [7:51:23<1:35:42,  1.91s/it][gradnorm] step 16089: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 597.7

 84%|████████▍ | 16090/19090 [7:51:24<1:26:28,  1.73s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 853.4097290039062, 'learning_rate': 3.670452592038978e-06, 'recon_loss': 0.0016625748976366594, 'recon_static': 0.0006994387149461545, 'recon_motion': 0.0009631361914216541, 'recon_norm': 1.3385315984487534, 'recon_norm_std': 0.03940419815480709, 'target_norm': 1.6186901926994324, 'recon_cos': 0.611690990626812, 'gradnorm/compressor.query': 0.3391891730692355, 'gradnorm/compressor.layer0': 13.12647231790382, 'gradnorm/compressor.layer1': 846.8873154086085, 'gradnorm/compressor.layer2': 97.34334369899156, 'gradnorm/compressor.layer3': 36.41420748603404, 'gradnorm/compressor.layer4': 8.326440121837127, 'gradnorm/compressor.layer5': 4.617051175811093, 'gradnorm/compressor.layer6': 2.5284783530173334, 'gradnorm/compressor.layer7': 2.5397609278288624, 'gradnorm/decoder.layer0': 2.8490559469978005, 'gradnorm/decoder.layer1': 1.7770323092733271, 'gradnorm/decoder.layer2': 1.0210767118496182, 'gradnorm/max_param': 597.7266235351562, 'epoch': 8.43}

 84%|████████▍ | 16090/19090 [7:51:24<1:26:28,  1.73s/it]
 84%|████████▍ | 16091/19090 [7:51:25<1:20:33,  1.61s/it]
 84%|████████▍ | 16092/19090 [7:51:28<1:38:29,  1.97s/it]
 84%|████████▍ | 16093/19090 [7:51:29<1:28:15,  1.77s/it]
 84%|████████▍ | 16094/19090 [7:51:31<1:23:59,  1.68s/it]
 84%|████████▍ | 16095/19090 [7:51:34<1:38:41,  1.98s/it]
 84%|████████▍ | 16096/19090 [7:51:35<1:28:18,  1.77s/it]
 84%|████████▍ | 16097/19090 [7:51:36<1:21:53,  1.64s/it]
 84%|████████▍ | 16098/19090 [7:51:39<1:37:56,  1.96s/it]
 84%|████████▍ | 16099/19090 [7:51:40<1:27:45,  1.76s/it][gradnorm] step 16099: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 581.5

 84%|████████▍ | 16100/19090 [7:51:42<1:20:39,  1.62s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 834.0895385742188, 'learning_rate': 3.646643650388512e-06, 'recon_loss': 0.0016661577799823135, 'recon_static': 0.0006889238808071241, 'recon_motion': 0.000977233875892125, 'recon_norm': 1.3365708202123643, 'recon_norm_std': 0.0389978451654315, 'target_norm': 1.6209228396415711, 'recon_cos': 0.6117170557379723, 'gradnorm/compressor.query': 0.3580992278633752, 'gradnorm/compressor.layer0': 13.508810488545233, 'gradnorm/compressor.layer1': 824.0624976298824, 'gradnorm/compressor.layer2': 122.29493358782284, 'gradnorm/compressor.layer3': 36.956964982830875, 'gradnorm/compressor.layer4': 8.714117535435628, 'gradnorm/compressor.layer5': 4.299921904453546, 'gradnorm/compressor.layer6': 2.6741376966518193, 'gradnorm/compressor.layer7': 2.692578596067754, 'gradnorm/decoder.layer0': 2.965887034409589, 'gradnorm/decoder.layer1': 1.7702656078946195, 'gradnorm/decoder.layer2': 0.9664018228440272, 'gradnorm/max_param': 581.5401000976562, 'epoch': 8.43}

 84%|████████▍ | 16100/19090 [7:51:42<1:20:39,  1.62s/it]
 84%|████████▍ | 16101/19090 [7:51:44<1:37:55,  1.97s/it]
 84%|████████▍ | 16102/19090 [7:51:46<1:27:42,  1.76s/it]
 84%|████████▍ | 16103/19090 [7:51:47<1:21:29,  1.64s/it]
 84%|████████▍ | 16104/19090 [7:51:50<1:41:20,  2.04s/it]
 84%|████████▍ | 16105/19090 [7:51:51<1:30:05,  1.81s/it]
 84%|████████▍ | 16106/19090 [7:51:52<1:22:10,  1.65s/it]
 84%|████████▍ | 16107/19090 [7:51:56<1:49:02,  2.19s/it]
 84%|████████▍ | 16108/19090 [7:51:57<1:35:32,  1.92s/it]
 84%|████████▍ | 16109/19090 [7:51:58<1:26:04,  1.73s/it][gradnorm] step 16109: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 602.9

 84%|████████▍ | 16110/19090 [7:52:01<1:33:58,  1.89s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 862.7301025390625, 'learning_rate': 3.622906103918647e-06, 'recon_loss': 0.0016157428501173854, 'recon_static': 0.0006456154340412467, 'recon_motion': 0.000970127405889798, 'recon_norm': 1.3225466758012772, 'recon_norm_std': 0.03938578460365534, 'target_norm': 1.6194828003644943, 'recon_cos': 0.622190760076046, 'gradnorm/compressor.query': 0.42440143606799546, 'gradnorm/compressor.layer0': 14.820297570992745, 'gradnorm/compressor.layer1': 854.4070312210686, 'gradnorm/compressor.layer2': 113.93651004095439, 'gradnorm/compressor.layer3': 31.56175741700706, 'gradnorm/compressor.layer4': 7.303317572911696, 'gradnorm/compressor.layer5': 4.053833388259871, 'gradnorm/compressor.layer6': 2.4680932052546596, 'gradnorm/compressor.layer7': 2.733547412428336, 'gradnorm/decoder.layer0': 2.6829989688969413, 'gradnorm/decoder.layer1': 1.716058461613927, 'gradnorm/decoder.layer2': 0.9160228151474372, 'gradnorm/max_param': 602.8721923828125, 'epoch': 8.44}

 84%|████████▍ | 16110/19090 [7:52:01<1:33:58,  1.89s/it]
 84%|████████▍ | 16111/19090 [7:52:02<1:24:56,  1.71s/it]
 84%|████████▍ | 16112/19090 [7:52:03<1:18:35,  1.58s/it]
 84%|████████▍ | 16113/19090 [7:52:06<1:36:42,  1.95s/it]
 84%|████████▍ | 16114/19090 [7:52:07<1:26:47,  1.75s/it]
 84%|████████▍ | 16115/19090 [7:52:09<1:19:48,  1.61s/it]
 84%|████████▍ | 16116/19090 [7:52:11<1:30:23,  1.82s/it]
 84%|████████▍ | 16117/19090 [7:52:12<1:22:22,  1.66s/it]
 84%|████████▍ | 16118/19090 [7:52:14<1:16:58,  1.55s/it]
 84%|████████▍ | 16119/19090 [7:52:16<1:34:20,  1.91s/it][gradnorm] step 16119: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 533.9

 84%|████████▍ | 16120/19090 [7:52:18<1:25:20,  1.72s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 767.4661865234375, 'learning_rate': 3.5992400319961068e-06, 'recon_loss': 0.0015589497430482878, 'recon_static': 0.0005795138466055505, 'recon_motion': 0.0009794359415536747, 'recon_norm': 1.290027904510498, 'recon_norm_std': 0.03988775666803122, 'target_norm': 1.6205858319997788, 'recon_cos': 0.6336942046880722, 'gradnorm/compressor.query': 0.36640162271859716, 'gradnorm/compressor.layer0': 13.20834944249724, 'gradnorm/compressor.layer1': 756.6410972184897, 'gradnorm/compressor.layer2': 122.47793168567144, 'gradnorm/compressor.layer3': 34.80314162532152, 'gradnorm/compressor.layer4': 7.804545999796494, 'gradnorm/compressor.layer5': 4.7527307390751075, 'gradnorm/compressor.layer6': 2.76590602330606, 'gradnorm/compressor.layer7': 2.834356534618048, 'gradnorm/decoder.layer0': 2.934572099885063, 'gradnorm/decoder.layer1': 1.672001129811649, 'gradnorm/decoder.layer2': 1.0455005790165193, 'gradnorm/max_param': 533.9386596679688, 'epoch': 8.44}

 84%|████████▍ | 16120/19090 [7:52:18<1:25:20,  1.72s/it]
 84%|████████▍ | 16121/19090 [7:52:19<1:18:47,  1.59s/it]
 84%|████████▍ | 16122/19090 [7:52:22<1:38:47,  2.00s/it]
 84%|████████▍ | 16123/19090 [7:52:23<1:28:16,  1.78s/it]
 84%|████████▍ | 16124/19090 [7:52:24<1:20:48,  1.63s/it]
 84%|████████▍ | 16125/19090 [7:52:27<1:39:50,  2.02s/it]
 84%|████████▍ | 16126/19090 [7:52:29<1:28:52,  1.80s/it]
 84%|████████▍ | 16127/19090 [7:52:30<1:21:15,  1.65s/it]
 84%|████████▍ | 16128/19090 [7:52:33<1:36:02,  1.95s/it]
 84%|████████▍ | 16129/19090 [7:52:34<1:26:13,  1.75s/it][gradnorm] step 16129: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 639.2

 84%|████████▍ | 16130/19090 [7:52:35<1:19:28,  1.61s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 911.3243408203125, 'learning_rate': 3.5756455137486657e-06, 'recon_loss': 0.0015377114061266184, 'recon_static': 0.0005625421428703703, 'recon_motion': 0.0009751692850841209, 'recon_norm': 1.284998020529747, 'recon_norm_std': 0.04041391871869564, 'target_norm': 1.620753464102745, 'recon_cos': 0.63941111266613, 'gradnorm/compressor.query': 0.4862649102854634, 'gradnorm/compressor.layer0': 14.830446340952621, 'gradnorm/compressor.layer1': 905.907280023734, 'gradnorm/compressor.layer2': 92.63852064779802, 'gradnorm/compressor.layer3': 30.63319903798461, 'gradnorm/compressor.layer4': 7.718586908385801, 'gradnorm/compressor.layer5': 4.288557512749612, 'gradnorm/compressor.layer6': 2.532044933438661, 'gradnorm/compressor.layer7': 2.618066076588648, 'gradnorm/decoder.layer0': 2.828698874570789, 'gradnorm/decoder.layer1': 1.7973977696789598, 'gradnorm/decoder.layer2': 1.0485393332587996, 'gradnorm/max_param': 639.1815795898438, 'epoch': 8.45}

 84%|████████▍ | 16130/19090 [7:52:35<1:19:28,  1.61s/it]
 84%|████████▍ | 16131/19090 [7:52:38<1:39:24,  2.02s/it]
 85%|████████▍ | 16132/19090 [7:52:39<1:28:32,  1.80s/it]
 85%|████████▍ | 16133/19090 [7:52:41<1:21:01,  1.64s/it]
 85%|████████▍ | 16134/19090 [7:52:43<1:35:59,  1.95s/it]
 85%|████████▍ | 16135/19090 [7:52:45<1:26:06,  1.75s/it]
 85%|████████▍ | 16136/19090 [7:52:46<1:19:16,  1.61s/it]
 85%|████████▍ | 16137/19090 [7:52:49<1:36:50,  1.97s/it]
 85%|████████▍ | 16138/19090 [7:52:50<1:26:44,  1.76s/it]
 85%|████████▍ | 16139/19090 [7:52:51<1:19:44,  1.62s/it][gradnorm] step 16139: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 658.0

 85%|████████▍ | 16140/19090 [7:52:54<1:39:55,  2.03s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 942.4548950195312, 'learning_rate': 3.552122628064869e-06, 'recon_loss': 0.0015746574412332849, 'recon_static': 0.0006072836593375541, 'recon_motion': 0.0009673738386482, 'recon_norm': 1.3027488470077515, 'recon_norm_std': 0.0406941139139235, 'target_norm': 1.6197374254465102, 'recon_cos': 0.6307182297110557, 'gradnorm/compressor.query': 0.3928095007624874, 'gradnorm/compressor.layer0': 12.43529088323172, 'gradnorm/compressor.layer1': 932.5001675603066, 'gradnorm/compressor.layer2': 130.5689173865664, 'gradnorm/compressor.layer3': 36.37904529428222, 'gradnorm/compressor.layer4': 8.925149987934596, 'gradnorm/compressor.layer5': 5.024746879920957, 'gradnorm/compressor.layer6': 2.856873632416933, 'gradnorm/compressor.layer7': 2.9996642878568514, 'gradnorm/decoder.layer0': 3.297750203182147, 'gradnorm/decoder.layer1': 1.9290298391420213, 'gradnorm/decoder.layer2': 1.2704870294136277, 'gradnorm/max_param': 657.9960327148438, 'epoch': 8.45}

 85%|████████▍ | 16140/19090 [7:52:54<1:39:55,  2.03s/it]
 85%|████████▍ | 16141/19090 [7:52:56<1:28:56,  1.81s/it]
 85%|████████▍ | 16142/19090 [7:52:57<1:21:16,  1.65s/it]
 85%|████████▍ | 16143/19090 [7:53:00<1:38:04,  2.00s/it]
 85%|████████▍ | 16144/19090 [7:53:01<1:27:35,  1.78s/it]
 85%|████████▍ | 16145/19090 [7:53:02<1:20:21,  1.64s/it]
 85%|████████▍ | 16146/19090 [7:53:05<1:41:02,  2.06s/it]
 85%|████████▍ | 16147/19090 [7:53:07<1:29:42,  1.83s/it]
 85%|████████▍ | 16148/19090 [7:53:08<1:21:44,  1.67s/it]
 85%|████████▍ | 16149/19090 [7:53:11<1:37:26,  1.99s/it][gradnorm] step 16149: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 745.1

 85%|████████▍ | 16150/19090 [7:53:12<1:27:10,  1.78s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1068.5035400390625, 'learning_rate': 3.5286714535937293e-06, 'recon_loss': 0.0015807345975190402, 'recon_static': 0.0005985807161778212, 'recon_motion': 0.0009821539075346663, 'recon_norm': 1.295096442103386, 'recon_norm_std': 0.03996014352887869, 'target_norm': 1.622558480501175, 'recon_cos': 0.6289432495832443, 'gradnorm/compressor.query': 0.3129480725912411, 'gradnorm/compressor.layer0': 10.684668093262589, 'gradnorm/compressor.layer1': 1055.755949545159, 'gradnorm/compressor.layer2': 158.3890520107845, 'gradnorm/compressor.layer3': 41.72288427548634, 'gradnorm/compressor.layer4': 8.66985521593827, 'gradnorm/compressor.layer5': 5.306434489750962, 'gradnorm/compressor.layer6': 3.070142435503569, 'gradnorm/compressor.layer7': 3.014406580687384, 'gradnorm/decoder.layer0': 3.2029402516747547, 'gradnorm/decoder.layer1': 1.950629509282301, 'gradnorm/decoder.layer2': 1.111373873608698, 'gradnorm/max_param': 745.0988159179688, 'epoch': 8.46}

 85%|████████▍ | 16150/19090 [7:53:12<1:27:10,  1.78s/it]
 85%|████████▍ | 16151/19090 [7:53:13<1:19:57,  1.63s/it]
 85%|████████▍ | 16152/19090 [7:53:16<1:43:13,  2.11s/it]
 85%|████████▍ | 16153/19090 [7:53:18<1:31:05,  1.86s/it]
 85%|████████▍ | 16154/19090 [7:53:19<1:22:37,  1.69s/it]
 85%|████████▍ | 16155/19090 [7:53:21<1:33:33,  1.91s/it]
 85%|████████▍ | 16156/19090 [7:53:23<1:24:19,  1.72s/it]
 85%|████████▍ | 16157/19090 [7:53:24<1:17:50,  1.59s/it]
 85%|████████▍ | 16158/19090 [7:53:27<1:34:03,  1.92s/it]
 85%|████████▍ | 16159/19090 [7:53:28<1:24:41,  1.73s/it][gradnorm] step 16159: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 760.4

 85%|████████▍ | 16160/19090 [7:53:29<1:18:05,  1.60s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1088.8001708984375, 'learning_rate': 3.505292068744498e-06, 'recon_loss': 0.0015624723513610662, 'recon_static': 0.0005780336359748617, 'recon_motion': 0.0009844386513577774, 'recon_norm': 1.294187492132187, 'recon_norm_std': 0.04040138050913811, 'target_norm': 1.622054323554039, 'recon_cos': 0.6342324465513229, 'gradnorm/compressor.query': 0.40345645481793635, 'gradnorm/compressor.layer0': 15.488449715351145, 'gradnorm/compressor.layer1': 1077.400517449291, 'gradnorm/compressor.layer2': 151.66564789035124, 'gradnorm/compressor.layer3': 36.33353118375986, 'gradnorm/compressor.layer4': 8.30078125, 'gradnorm/compressor.layer5': 5.35596911696518, 'gradnorm/compressor.layer6': 3.123347035978122, 'gradnorm/compressor.layer7': 3.1449729467056238, 'gradnorm/decoder.layer0': 3.148632261248866, 'gradnorm/decoder.layer1': 1.6984348637508648, 'gradnorm/decoder.layer2': 1.09484018762117, 'gradnorm/max_param': 760.3834838867188, 'epoch': 8.47}

 85%|████████▍ | 16160/19090 [7:53:29<1:18:05,  1.60s/it]
 85%|████████▍ | 16161/19090 [7:53:32<1:35:38,  1.96s/it]
 85%|████████▍ | 16162/19090 [7:53:33<1:25:48,  1.76s/it]
 85%|████████▍ | 16163/19090 [7:53:35<1:18:50,  1.62s/it]
 85%|████████▍ | 16164/19090 [7:53:37<1:35:25,  1.96s/it]
 85%|████████▍ | 16165/19090 [7:53:39<1:25:44,  1.76s/it]
 85%|████████▍ | 16166/19090 [7:53:40<1:18:47,  1.62s/it]
 85%|████████▍ | 16167/19090 [7:53:43<1:34:17,  1.94s/it]
 85%|████████▍ | 16168/19090 [7:53:44<1:24:51,  1.74s/it]
 85%|████████▍ | 16169/19090 [7:53:45<1:18:07,  1.60s/it][gradnorm] step 16169: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 753.3

 85%|████████▍ | 16170/19090 [7:53:48<1:35:39,  1.97s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1074.8321533203125, 'learning_rate': 3.481984551686429e-06, 'recon_loss': 0.0016385401715524495, 'recon_static': 0.0006618260609684512, 'recon_motion': 0.0009767141047632322, 'recon_norm': 1.3247329771518708, 'recon_norm_std': 0.039783574733883144, 'target_norm': 1.6209550082683564, 'recon_cos': 0.6170688822865487, 'gradnorm/compressor.query': 0.4294465516510117, 'gradnorm/compressor.layer0': 13.69239197147053, 'gradnorm/compressor.layer1': 1067.6604680327919, 'gradnorm/compressor.layer2': 118.24929802910037, 'gradnorm/compressor.layer3': 33.217930522582606, 'gradnorm/compressor.layer4': 6.762590886490686, 'gradnorm/compressor.layer5': 4.425179153522655, 'gradnorm/compressor.layer6': 2.4054094307880565, 'gradnorm/compressor.layer7': 2.941491843874146, 'gradnorm/decoder.layer0': 2.801666758225451, 'gradnorm/decoder.layer1': 1.6362750560453627, 'gradnorm/decoder.layer2': 0.9763670153464135, 'gradnorm/max_param': 753.342041015625, 'epoch': 8.47}

 85%|████████▍ | 16170/19090 [7:53:48<1:35:39,  1.97s/it]
 85%|████████▍ | 16171/19090 [7:53:49<1:26:04,  1.77s/it]
 85%|████████▍ | 16172/19090 [7:53:51<1:18:56,  1.62s/it]
 85%|████████▍ | 16173/19090 [7:53:53<1:34:23,  1.94s/it]
 85%|████████▍ | 16174/19090 [7:53:55<1:24:54,  1.75s/it]
 85%|████████▍ | 16175/19090 [7:53:56<1:18:08,  1.61s/it]
 85%|████████▍ | 16176/19090 [7:53:59<1:33:12,  1.92s/it]
 85%|████████▍ | 16177/19090 [7:54:00<1:24:04,  1.73s/it]
 85%|████████▍ | 16178/19090 [7:54:01<1:17:32,  1.60s/it]
 85%|████████▍ | 16179/19090 [7:54:04<1:38:27,  2.03s/it][gradnorm] step 16179: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 860.0

 85%|████████▍ | 16180/19090 [7:54:05<1:27:42,  1.81s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1232.1968994140625, 'learning_rate': 3.458748980348436e-06, 'recon_loss': 0.0016070468118414283, 'recon_static': 0.0006336821854347364, 'recon_motion': 0.0009733646031236276, 'recon_norm': 1.3203739553689957, 'recon_norm_std': 0.039886020869016645, 'target_norm': 1.6197541326284408, 'recon_cos': 0.624585309624672, 'gradnorm/compressor.query': 0.5978012657007274, 'gradnorm/compressor.layer0': 17.6413818413265, 'gradnorm/compressor.layer1': 1218.529082541734, 'gradnorm/compressor.layer2': 176.84641443036102, 'gradnorm/compressor.layer3': 42.34538309812018, 'gradnorm/compressor.layer4': 7.209973304900168, 'gradnorm/compressor.layer5': 5.447017539561806, 'gradnorm/compressor.layer6': 3.0010353527026945, 'gradnorm/compressor.layer7': 3.3930069252166097, 'gradnorm/decoder.layer0': 3.143250691556934, 'gradnorm/decoder.layer1': 1.7962565809114726, 'gradnorm/decoder.layer2': 1.1488799715454394, 'gradnorm/max_param': 859.9799194335938, 'epoch': 8.48}

 85%|████████▍ | 16180/19090 [7:54:05<1:27:42,  1.81s/it]
 85%|████████▍ | 16181/19090 [7:54:07<1:20:04,  1.65s/it]
 85%|████████▍ | 16182/19090 [7:54:09<1:32:36,  1.91s/it]
 85%|████████▍ | 16183/19090 [7:54:11<1:23:33,  1.72s/it]
 85%|████████▍ | 16184/19090 [7:54:12<1:17:12,  1.59s/it]
 85%|████████▍ | 16185/19090 [7:54:15<1:37:07,  2.01s/it]
 85%|████████▍ | 16186/19090 [7:54:16<1:26:38,  1.79s/it]
 85%|████████▍ | 16187/19090 [7:54:17<1:19:19,  1.64s/it]
 85%|████████▍ | 16188/19090 [7:54:20<1:31:16,  1.89s/it]
 85%|████████▍ | 16189/19090 [7:54:21<1:22:31,  1.71s/it][gradnorm] step 16189: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 547.7

 85%|████████▍ | 16190/19090 [7:54:22<1:16:34,  1.58s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 785.7150268554688, 'learning_rate': 3.435585432418914e-06, 'recon_loss': 0.001674319471931085, 'recon_static': 0.0006930557632585987, 'recon_motion': 0.0009812636984861457, 'recon_norm': 1.3372999399900436, 'recon_norm_std': 0.03982360092923045, 'target_norm': 1.6214339226484298, 'recon_cos': 0.6110824748873711, 'gradnorm/compressor.query': 0.46980277095564565, 'gradnorm/compressor.layer0': 22.715880578765816, 'gradnorm/compressor.layer1': 776.337877473462, 'gradnorm/compressor.layer2': 114.80157388811139, 'gradnorm/compressor.layer3': 29.062597656085927, 'gradnorm/compressor.layer4': 7.471837893800324, 'gradnorm/compressor.layer5': 4.626020782819837, 'gradnorm/compressor.layer6': 2.788521081000348, 'gradnorm/compressor.layer7': 3.114676704835828, 'gradnorm/decoder.layer0': 2.949345668057568, 'gradnorm/decoder.layer1': 1.698570039941765, 'gradnorm/decoder.layer2': 0.9893789113703012, 'gradnorm/max_param': 547.7474975585938, 'epoch': 8.48}

 85%|████████▍ | 16190/19090 [7:54:22<1:16:34,  1.58s/it]
 85%|████████▍ | 16191/19090 [7:54:25<1:28:51,  1.84s/it]
 85%|████████▍ | 16192/19090 [7:54:26<1:20:45,  1.67s/it]
 85%|████████▍ | 16193/19090 [7:54:27<1:15:09,  1.56s/it]
 85%|████████▍ | 16194/19090 [7:54:30<1:32:13,  1.91s/it]
 85%|████████▍ | 16195/19090 [7:54:31<1:23:20,  1.73s/it]
 85%|████████▍ | 16196/19090 [7:54:33<1:16:58,  1.60s/it]
 85%|████████▍ | 16197/19090 [7:54:36<1:34:25,  1.96s/it]
 85%|████████▍ | 16198/19090 [7:54:37<1:24:37,  1.76s/it]
 85%|████████▍ | 16199/19090 [7:54:38<1:17:57,  1.62s/it][gradnorm] step 16199: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 452.2

 85%|████████▍ | 16200/19090 [7:54:41<1:36:35,  2.01s/it]
                                                         
{'loss': 0.0022, 'grad_norm': 648.8322143554688, 'learning_rate': 3.4124939853454497e-06, 'recon_loss': 0.002212526780203916, 'recon_static': 0.0012381556050968356, 'recon_motion': 0.0009743711838382296, 'recon_norm': 1.5617076575756073, 'recon_norm_std': 0.03574381829239428, 'target_norm': 1.6207332462072372, 'recon_cos': 0.5300394862890243, 'gradnorm/compressor.query': 0.5163220693331555, 'gradnorm/compressor.layer0': 14.112757415721422, 'gradnorm/compressor.layer1': 640.8080592111806, 'gradnorm/compressor.layer2': 95.6982970721005, 'gradnorm/compressor.layer3': 29.625063867942757, 'gradnorm/compressor.layer4': 7.765728648547482, 'gradnorm/compressor.layer5': 4.413329152991706, 'gradnorm/compressor.layer6': 2.824167591438881, 'gradnorm/compressor.layer7': 3.479211924976184, 'gradnorm/decoder.layer0': 3.0131745339076446, 'gradnorm/decoder.layer1': 1.7441644100946785, 'gradnorm/decoder.layer2': 1.0252898248445703, 'gradnorm/max_param': 452.1734924316406, 'epoch': 8.49}

 85%|████████▍ | 16200/19090 [7:54:41<1:36:35,  2.01s/it]
 85%|████████▍ | 16201/19090 [7:54:42<1:26:15,  1.79s/it]
 85%|████████▍ | 16202/19090 [7:54:44<1:19:02,  1.64s/it]
 85%|████████▍ | 16203/19090 [7:54:45<1:19:25,  1.65s/it]
 85%|████████▍ | 16204/19090 [7:54:47<1:14:14,  1.54s/it]
 85%|████████▍ | 16205/19090 [7:54:48<1:10:32,  1.47s/it]
 85%|████████▍ | 16206/19090 [7:54:50<1:15:21,  1.57s/it]
 85%|████████▍ | 16207/19090 [7:54:51<1:11:20,  1.48s/it]
 85%|████████▍ | 16208/19090 [7:54:52<1:08:31,  1.43s/it]
 85%|████████▍ | 16209/19090 [7:54:55<1:25:01,  1.77s/it][gradnorm] step 16209: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 679.3

 85%|████████▍ | 16210/19090 [7:54:56<1:18:02,  1.63s/it]
                                                         
{'loss': 0.0022, 'grad_norm': 974.8427734375, 'learning_rate': 3.389474716334534e-06, 'recon_loss': 0.0022326167498249562, 'recon_static': 0.001246706716483459, 'recon_motion': 0.0009859100260655395, 'recon_norm': 1.5530796080827713, 'recon_norm_std': 0.03483850220218301, 'target_norm': 1.624839836359024, 'recon_cos': 0.523085369169712, 'gradnorm/compressor.query': 0.6486271040980274, 'gradnorm/compressor.layer0': 25.21871611676782, 'gradnorm/compressor.layer1': 962.6301860008339, 'gradnorm/compressor.layer2': 147.69387417476054, 'gradnorm/compressor.layer3': 32.98876576945836, 'gradnorm/compressor.layer4': 8.512669713685412, 'gradnorm/compressor.layer5': 4.39717780550187, 'gradnorm/compressor.layer6': 2.8292389251365484, 'gradnorm/compressor.layer7': 3.0533083560982557, 'gradnorm/decoder.layer0': 3.0925241218680797, 'gradnorm/decoder.layer1': 1.7868774263539609, 'gradnorm/decoder.layer2': 1.2024192226188077, 'gradnorm/max_param': 679.3021240234375, 'epoch': 8.49}

 85%|████████▍ | 16210/19090 [7:54:56<1:18:02,  1.63s/it]
 85%|████████▍ | 16211/19090 [7:54:57<1:13:05,  1.52s/it]
 85%|████████▍ | 16212/19090 [7:55:00<1:26:05,  1.79s/it]
 85%|████████▍ | 16213/19090 [7:55:01<1:18:49,  1.64s/it]
 85%|████████▍ | 16214/19090 [7:55:02<1:13:46,  1.54s/it]
 85%|████████▍ | 16215/19090 [7:55:04<1:19:17,  1.65s/it]
 85%|████████▍ | 16216/19090 [7:55:06<1:14:02,  1.55s/it]
 85%|████████▍ | 16217/19090 [7:55:07<1:10:26,  1.47s/it]
 85%|████████▍ | 16218/19090 [7:55:09<1:20:34,  1.68s/it]
 85%|████████▍ | 16219/19090 [7:55:10<1:14:56,  1.57s/it][gradnorm] step 16219: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 615.0

 85%|████████▍ | 16220/19090 [7:55:12<1:10:53,  1.48s/it]
                                                         
{'loss': 0.0027, 'grad_norm': 880.9231567382812, 'learning_rate': 3.3665277023513435e-06, 'recon_loss': 0.0026759963831864296, 'recon_static': 0.001700543920742348, 'recon_motion': 0.0009754523911396973, 'recon_norm': 1.7178122222423553, 'recon_norm_std': 0.03304150202311575, 'target_norm': 1.623346322774887, 'recon_cos': 0.4828583724796772, 'gradnorm/compressor.query': 0.5586098355198038, 'gradnorm/compressor.layer0': 22.16579181216309, 'gradnorm/compressor.layer1': 871.4648587292547, 'gradnorm/compressor.layer2': 122.7117828108817, 'gradnorm/compressor.layer3': 30.708134133371324, 'gradnorm/compressor.layer4': 5.789930270867163, 'gradnorm/compressor.layer5': 4.502431848191311, 'gradnorm/compressor.layer6': 2.3999113344662466, 'gradnorm/compressor.layer7': 2.9577261789726705, 'gradnorm/decoder.layer0': 3.144814652650284, 'gradnorm/decoder.layer1': 1.679137738769845, 'gradnorm/decoder.layer2': 0.8770101142156537, 'gradnorm/max_param': 614.9989013671875, 'epoch': 8.5}

 85%|████████▍ | 16220/19090 [7:55:12<1:10:53,  1.48s/it]
 85%|████████▍ | 16221/19090 [7:55:14<1:16:45,  1.61s/it]
 85%|████████▍ | 16222/19090 [7:55:15<1:12:09,  1.51s/it]
 85%|████████▍ | 16223/19090 [7:55:16<1:08:53,  1.44s/it]
 85%|████████▍ | 16224/19090 [7:55:18<1:17:06,  1.61s/it]
 85%|████████▍ | 16225/19090 [7:55:19<1:12:21,  1.52s/it]
 85%|████████▍ | 16226/19090 [7:55:21<1:09:01,  1.45s/it]
 85%|████████▌ | 16227/19090 [7:55:23<1:23:26,  1.75s/it]
 85%|████████▌ | 16228/19090 [7:55:24<1:16:49,  1.61s/it]
 85%|████████▌ | 16229/19090 [7:55:26<1:12:05,  1.51s/it][gradnorm] step 16229: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 673.0

 85%|████████▌ | 16230/19090 [7:55:28<1:26:34,  1.82s/it]
                                                         
{'loss': 0.0025, 'grad_norm': 965.4536743164062, 'learning_rate': 3.3436530201194915e-06, 'recon_loss': 0.002504271786892787, 'recon_static': 0.0015240479493513702, 'recon_motion': 0.0009802237880649045, 'recon_norm': 1.6547871947288513, 'recon_norm_std': 0.03337272326461971, 'target_norm': 1.6237054109573363, 'recon_cos': 0.49797305911779405, 'gradnorm/compressor.query': 0.5832858804293638, 'gradnorm/compressor.layer0': 20.642275701641477, 'gradnorm/compressor.layer1': 953.6245658014478, 'gradnorm/compressor.layer2': 143.6337165544706, 'gradnorm/compressor.layer3': 38.85683357692698, 'gradnorm/compressor.layer4': 8.999418875688455, 'gradnorm/compressor.layer5': 4.861013056659251, 'gradnorm/compressor.layer6': 2.5863637889759015, 'gradnorm/compressor.layer7': 3.0757719234047065, 'gradnorm/decoder.layer0': 2.997295591178968, 'gradnorm/decoder.layer1': 1.7477787089171897, 'gradnorm/decoder.layer2': 1.208200672145935, 'gradnorm/max_param': 673.035888671875, 'epoch': 8.5}

 85%|████████▌ | 16230/19090 [7:55:28<1:26:34,  1.82s/it]
 85%|████████▌ | 16231/19090 [7:55:30<1:18:57,  1.66s/it]
 85%|████████▌ | 16232/19090 [7:55:31<1:13:36,  1.55s/it]
 85%|████████▌ | 16233/19090 [7:55:34<1:31:07,  1.91s/it]
 85%|████████▌ | 16234/19090 [7:55:35<1:22:06,  1.72s/it]
 85%|████████▌ | 16235/19090 [7:55:36<1:15:43,  1.59s/it]
 85%|████████▌ | 16236/19090 [7:55:38<1:21:56,  1.72s/it]
 85%|████████▌ | 16237/19090 [7:55:40<1:15:39,  1.59s/it]
 85%|████████▌ | 16238/19090 [7:55:41<1:11:12,  1.50s/it]
 85%|████████▌ | 16239/19090 [7:55:43<1:23:54,  1.77s/it][gradnorm] step 16239: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 636.2

 85%|████████▌ | 16240/19090 [7:55:44<1:17:01,  1.62s/it]
                                                         
{'loss': 0.0018, 'grad_norm': 915.9090576171875, 'learning_rate': 3.3208507461207074e-06, 'recon_loss': 0.0018368306889897212, 'recon_static': 0.0008588473530835472, 'recon_motion': 0.0009779833286302163, 'recon_norm': 1.4027554750442506, 'recon_norm_std': 0.03779567312449217, 'target_norm': 1.623006609082222, 'recon_cos': 0.5813763454556465, 'gradnorm/compressor.query': 0.4705556108099885, 'gradnorm/compressor.layer0': 15.807318641434287, 'gradnorm/compressor.layer1': 901.5325839923923, 'gradnorm/compressor.layer2': 153.42140317806053, 'gradnorm/compressor.layer3': 46.808241524430024, 'gradnorm/compressor.layer4': 9.361102166125976, 'gradnorm/compressor.layer5': 5.195780568083098, 'gradnorm/compressor.layer6': 2.8498434659567784, 'gradnorm/compressor.layer7': 3.335209429797894, 'gradnorm/decoder.layer0': 3.1245394558099417, 'gradnorm/decoder.layer1': 2.0679668453886557, 'gradnorm/decoder.layer2': 1.1448787122729354, 'gradnorm/max_param': 636.16259765625, 'epoch': 8.51}

 85%|████████▌ | 16240/19090 [7:55:44<1:17:01,  1.62s/it]
 85%|████████▌ | 16241/19090 [7:55:46<1:12:09,  1.52s/it]
 85%|████████▌ | 16242/19090 [7:55:48<1:20:55,  1.70s/it]
 85%|████████▌ | 16243/19090 [7:55:49<1:14:55,  1.58s/it]
 85%|████████▌ | 16244/19090 [7:55:50<1:10:42,  1.49s/it]
 85%|████████▌ | 16245/19090 [7:55:53<1:29:38,  1.89s/it]
 85%|████████▌ | 16246/19090 [7:55:55<1:21:06,  1.71s/it]
 85%|████████▌ | 16247/19090 [7:55:56<1:15:01,  1.58s/it]
 85%|████████▌ | 16248/19090 [7:55:58<1:25:34,  1.81s/it]
 85%|████████▌ | 16249/19090 [7:55:59<1:18:12,  1.65s/it][gradnorm] step 16249: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 640.3

 85%|████████▌ | 16250/19090 [7:56:01<1:13:04,  1.54s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 919.0889282226562, 'learning_rate': 3.2981209565946443e-06, 'recon_loss': 0.001628815138246864, 'recon_static': 0.0006507997721200808, 'recon_motion': 0.0009780153806786985, 'recon_norm': 1.3216258734464645, 'recon_norm_std': 0.03862406350672245, 'target_norm': 1.6197308033704758, 'recon_cos': 0.6188012421131134, 'gradnorm/compressor.query': 0.4418318561766452, 'gradnorm/compressor.layer0': 16.949023841498587, 'gradnorm/compressor.layer1': 907.2253027776507, 'gradnorm/compressor.layer2': 139.93015138141243, 'gradnorm/compressor.layer3': 40.53250437775604, 'gradnorm/compressor.layer4': 9.399822687952225, 'gradnorm/compressor.layer5': 5.625322290300439, 'gradnorm/compressor.layer6': 3.005113376503397, 'gradnorm/compressor.layer7': 3.283711554596643, 'gradnorm/decoder.layer0': 3.2120317882328258, 'gradnorm/decoder.layer1': 2.08498186373427, 'gradnorm/decoder.layer2': 1.2331871884765975, 'gradnorm/max_param': 640.2566528320312, 'epoch': 8.51}

 85%|████████▌ | 16250/19090 [7:56:01<1:13:04,  1.54s/it]
 85%|████████▌ | 16251/19090 [7:56:03<1:24:49,  1.79s/it]
 85%|████████▌ | 16252/19090 [7:56:04<1:17:46,  1.64s/it]
 85%|████████▌ | 16253/19090 [7:56:06<1:12:40,  1.54s/it]
 85%|████████▌ | 16254/19090 [7:56:08<1:25:29,  1.81s/it]
 85%|████████▌ | 16255/19090 [7:56:09<1:18:13,  1.66s/it]
 85%|████████▌ | 16256/19090 [7:56:11<1:13:07,  1.55s/it]
 85%|████████▌ | 16257/19090 [7:56:13<1:24:27,  1.79s/it]
 85%|████████▌ | 16258/19090 [7:56:14<1:17:17,  1.64s/it]
 85%|████████▌ | 16259/19090 [7:56:16<1:12:23,  1.53s/it][gradnorm] step 16259: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 760.1

 85%|████████▌ | 16260/19090 [7:56:18<1:22:22,  1.75s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1087.71630859375, 'learning_rate': 3.275463727538619e-06, 'recon_loss': 0.0015889533213339746, 'recon_static': 0.0006123220751760528, 'recon_motion': 0.0009766312839929014, 'recon_norm': 1.3015841454267503, 'recon_norm_std': 0.04050563843920827, 'target_norm': 1.6220970362424851, 'recon_cos': 0.6274927794933319, 'gradnorm/compressor.query': 0.38934863419151766, 'gradnorm/compressor.layer0': 11.981044420877467, 'gradnorm/compressor.layer1': 1076.946319460724, 'gradnorm/compressor.layer2': 146.47453383609042, 'gradnorm/compressor.layer3': 39.68267225804057, 'gradnorm/compressor.layer4': 9.250150524667363, 'gradnorm/compressor.layer5': 4.8136799344273395, 'gradnorm/compressor.layer6': 2.722093127124202, 'gradnorm/compressor.layer7': 3.2879304727317558, 'gradnorm/decoder.layer0': 2.8918408285529043, 'gradnorm/decoder.layer1': 1.9397808309023046, 'gradnorm/decoder.layer2': 1.0319345109808606, 'gradnorm/max_param': 760.0804443359375, 'epoch': 8.52}

 85%|████████▌ | 16260/19090 [7:56:18<1:22:22,  1.75s/it]
 85%|████████▌ | 16261/19090 [7:56:19<1:16:23,  1.62s/it]
 85%|████████▌ | 16262/19090 [7:56:21<1:11:49,  1.52s/it]
 85%|████████▌ | 16263/19090 [7:56:23<1:30:12,  1.91s/it]
 85%|████████▌ | 16264/19090 [7:56:25<1:21:34,  1.73s/it]
 85%|████████▌ | 16265/19090 [7:56:26<1:15:25,  1.60s/it]
 85%|████████▌ | 16266/19090 [7:56:28<1:27:51,  1.87s/it]
 85%|████████▌ | 16267/19090 [7:56:30<1:20:06,  1.70s/it]
 85%|████████▌ | 16268/19090 [7:56:31<1:14:39,  1.59s/it]
 85%|████████▌ | 16269/19090 [7:56:34<1:29:35,  1.91s/it][gradnorm] step 16269: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 828.3

 85%|████████▌ | 16270/19090 [7:56:35<1:21:07,  1.73s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1189.5020751953125, 'learning_rate': 3.2528791347073095e-06, 'recon_loss': 0.0015741948736831546, 'recon_static': 0.0006048465918865986, 'recon_motion': 0.0009693483181763441, 'recon_norm': 1.3087096422910691, 'recon_norm_std': 0.04031119840219617, 'target_norm': 1.6189828246831894, 'recon_cos': 0.6313580319285392, 'gradnorm/compressor.query': 0.399749167593542, 'gradnorm/compressor.layer0': 13.179503416393745, 'gradnorm/compressor.layer1': 1173.6353777898823, 'gradnorm/compressor.layer2': 188.26011551507665, 'gradnorm/compressor.layer3': 41.28066598044492, 'gradnorm/compressor.layer4': 10.397948106674777, 'gradnorm/compressor.layer5': 5.256928549568517, 'gradnorm/compressor.layer6': 3.085575439846336, 'gradnorm/compressor.layer7': 3.6520160385230898, 'gradnorm/decoder.layer0': 3.2042618385602806, 'gradnorm/decoder.layer1': 2.2063365281727347, 'gradnorm/decoder.layer2': 1.3298030350334793, 'gradnorm/max_param': 828.2938842773438, 'epoch': 8.52}

 85%|████████▌ | 16270/19090 [7:56:35<1:21:07,  1.73s/it]
 85%|████████▌ | 16271/19090 [7:56:36<1:15:04,  1.60s/it]
 85%|████████▌ | 16272/19090 [7:56:39<1:27:58,  1.87s/it]
 85%|████████▌ | 16273/19090 [7:56:40<1:19:42,  1.70s/it]
 85%|████████▌ | 16274/19090 [7:56:41<1:13:55,  1.58s/it]
 85%|████████▌ | 16275/19090 [7:56:44<1:28:06,  1.88s/it]
 85%|████████▌ | 16276/19090 [7:56:45<1:19:46,  1.70s/it]
 85%|████████▌ | 16277/19090 [7:56:47<1:13:55,  1.58s/it]
 85%|████████▌ | 16278/19090 [7:56:49<1:30:42,  1.94s/it]
 85%|████████▌ | 16279/19090 [7:56:51<1:21:38,  1.74s/it][gradnorm] step 16279: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 771.3

 85%|████████▌ | 16280/19090 [7:56:52<1:15:15,  1.61s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1102.80029296875, 'learning_rate': 3.230367253612551e-06, 'recon_loss': 0.0015517239953624085, 'recon_static': 0.0005636837086058222, 'recon_motion': 0.0009880403042188846, 'recon_norm': 1.2853636026382447, 'recon_norm_std': 0.039737116917967795, 'target_norm': 1.6235263437032699, 'recon_cos': 0.6367820397019386, 'gradnorm/compressor.query': 0.46064412754050044, 'gradnorm/compressor.layer0': 14.67611863715857, 'gradnorm/compressor.layer1': 1093.106753707066, 'gradnorm/compressor.layer2': 138.77710883823744, 'gradnorm/compressor.layer3': 40.6471138730776, 'gradnorm/compressor.layer4': 9.967365419863262, 'gradnorm/compressor.layer5': 4.893914721994279, 'gradnorm/compressor.layer6': 3.0957396303545544, 'gradnorm/compressor.layer7': 3.5621646924078907, 'gradnorm/decoder.layer0': 2.953666677700201, 'gradnorm/decoder.layer1': 1.909581010009542, 'gradnorm/decoder.layer2': 1.162075523075754, 'gradnorm/max_param': 771.291748046875, 'epoch': 8.53}

 85%|████████▌ | 16280/19090 [7:56:52<1:15:15,  1.61s/it]
 85%|████████▌ | 16281/19090 [7:56:55<1:31:06,  1.95s/it]
 85%|████████▌ | 16282/19090 [7:56:56<1:21:46,  1.75s/it]
 85%|████████▌ | 16283/19090 [7:56:57<1:15:17,  1.61s/it]
 85%|████████▌ | 16284/19090 [7:57:00<1:25:07,  1.82s/it]
 85%|████████▌ | 16285/19090 [7:57:01<1:17:34,  1.66s/it]
 85%|████████▌ | 16286/19090 [7:57:02<1:12:16,  1.55s/it]
 85%|████████▌ | 16287/19090 [7:57:05<1:26:54,  1.86s/it]
 85%|████████▌ | 16288/19090 [7:57:06<1:18:45,  1.69s/it]
 85%|████████▌ | 16289/19090 [7:57:07<1:13:06,  1.57s/it][gradnorm] step 16289: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 763.3

 85%|████████▌ | 16290/19090 [7:57:10<1:34:19,  2.02s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1092.292724609375, 'learning_rate': 3.207928159523074e-06, 'recon_loss': 0.0015186984761385246, 'recon_static': 0.0005402218477684074, 'recon_motion': 0.0009784766109078192, 'recon_norm': 1.2745141863822937, 'recon_norm_std': 0.041319516208022836, 'target_norm': 1.6225830793380738, 'recon_cos': 0.643799164891243, 'gradnorm/compressor.query': 0.48420998623534683, 'gradnorm/compressor.layer0': 14.976886489062071, 'gradnorm/compressor.layer1': 1081.731944614746, 'gradnorm/compressor.layer2': 144.05013504419216, 'gradnorm/compressor.layer3': 42.51404868908041, 'gradnorm/compressor.layer4': 10.197819824590189, 'gradnorm/compressor.layer5': 5.154068178897579, 'gradnorm/compressor.layer6': 3.2936618568743814, 'gradnorm/compressor.layer7': 3.964041973657809, 'gradnorm/decoder.layer0': 3.60189665528883, 'gradnorm/decoder.layer1': 2.189857520428631, 'gradnorm/decoder.layer2': 1.442023416527442, 'gradnorm/max_param': 763.3419799804688, 'epoch': 8.53}

 85%|████████▌ | 16290/19090 [7:57:10<1:34:19,  2.02s/it]
 85%|████████▌ | 16291/19090 [7:57:12<1:24:00,  1.80s/it]
 85%|████████▌ | 16292/19090 [7:57:13<1:16:43,  1.65s/it]
 85%|████████▌ | 16293/19090 [7:57:16<1:31:31,  1.96s/it]
 85%|████████▌ | 16294/19090 [7:57:17<1:21:57,  1.76s/it]
 85%|████████▌ | 16295/19090 [7:57:18<1:15:17,  1.62s/it]
 85%|████████▌ | 16296/19090 [7:57:21<1:31:01,  1.95s/it]
 85%|████████▌ | 16297/19090 [7:57:22<1:21:37,  1.75s/it]
 85%|████████▌ | 16298/19090 [7:57:24<1:15:05,  1.61s/it]
 85%|████████▌ | 16299/19090 [7:57:26<1:32:38,  1.99s/it][gradnorm] step 16299: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 825.3

 85%|████████▌ | 16300/19090 [7:57:28<1:22:50,  1.78s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1179.8917236328125, 'learning_rate': 3.185561927464226e-06, 'recon_loss': 0.0014849208906525745, 'recon_static': 0.0005122044211020693, 'recon_motion': 0.000972716452088207, 'recon_norm': 1.2575950294733047, 'recon_norm_std': 0.040434387139976026, 'target_norm': 1.6173677921295166, 'recon_cos': 0.6488282337784768, 'gradnorm/compressor.query': 0.46996409225768415, 'gradnorm/compressor.layer0': 15.722067147348728, 'gradnorm/compressor.layer1': 1169.6362682475267, 'gradnorm/compressor.layer2': 148.76048413683318, 'gradnorm/compressor.layer3': 39.543511870529244, 'gradnorm/compressor.layer4': 9.52096292642644, 'gradnorm/compressor.layer5': 5.062625271224488, 'gradnorm/compressor.layer6': 3.062873506151545, 'gradnorm/compressor.layer7': 3.546207314719449, 'gradnorm/decoder.layer0': 3.3143270959760587, 'gradnorm/decoder.layer1': 2.069772441168699, 'gradnorm/decoder.layer2': 1.1311083372998487, 'gradnorm/max_param': 825.3148803710938, 'epoch': 8.54}

 85%|████████▌ | 16300/19090 [7:57:28<1:22:50,  1.78s/it]
 85%|████████▌ | 16301/19090 [7:57:29<1:16:01,  1.64s/it]
 85%|████████▌ | 16302/19090 [7:57:32<1:29:38,  1.93s/it]
 85%|████████▌ | 16303/19090 [7:57:33<1:20:43,  1.74s/it]
 85%|████████▌ | 16304/19090 [7:57:34<1:14:26,  1.60s/it]
 85%|████████▌ | 16305/19090 [7:57:37<1:28:53,  1.92s/it]
 85%|████████▌ | 16306/19090 [7:57:38<1:20:12,  1.73s/it]
 85%|████████▌ | 16307/19090 [7:57:39<1:13:57,  1.59s/it]
 85%|████████▌ | 16308/19090 [7:57:42<1:26:44,  1.87s/it]
 85%|████████▌ | 16309/19090 [7:57:43<1:18:35,  1.70s/it][gradnorm] step 16309: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 544.6

 85%|████████▌ | 16310/19090 [7:57:45<1:12:52,  1.57s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 780.3580322265625, 'learning_rate': 3.1632686322177446e-06, 'recon_loss': 0.0015257003426086158, 'recon_static': 0.0005470259071444161, 'recon_motion': 0.0009786744398297743, 'recon_norm': 1.2827457517385483, 'recon_norm_std': 0.039612894877791405, 'target_norm': 1.6189910292625427, 'recon_cos': 0.6414001807570457, 'gradnorm/compressor.query': 0.33059406023978244, 'gradnorm/compressor.layer0': 12.704460646265153, 'gradnorm/compressor.layer1': 771.6570643751018, 'gradnorm/compressor.layer2': 108.86820981983216, 'gradnorm/compressor.layer3': 36.76677876164585, 'gradnorm/compressor.layer4': 9.445531396088427, 'gradnorm/compressor.layer5': 4.535545680685185, 'gradnorm/compressor.layer6': 2.6176452236459054, 'gradnorm/compressor.layer7': 2.9500978485948868, 'gradnorm/decoder.layer0': 2.8019089390445497, 'gradnorm/decoder.layer1': 1.9868549859741973, 'gradnorm/decoder.layer2': 1.11456711778105, 'gradnorm/max_param': 544.5833740234375, 'epoch': 8.54}

 85%|████████▌ | 16310/19090 [7:57:45<1:12:52,  1.57s/it]
 85%|████████▌ | 16311/19090 [7:57:47<1:27:21,  1.89s/it]
 85%|████████▌ | 16312/19090 [7:57:48<1:18:55,  1.70s/it]
 85%|████████▌ | 16313/19090 [7:57:50<1:13:07,  1.58s/it]
 85%|████████▌ | 16314/19090 [7:57:53<1:32:41,  2.00s/it]
 85%|████████▌ | 16315/19090 [7:57:54<1:22:42,  1.79s/it]
 85%|████████▌ | 16316/19090 [7:57:55<1:15:39,  1.64s/it]
 85%|████████▌ | 16317/19090 [7:57:58<1:34:45,  2.05s/it]
 85%|████████▌ | 16318/19090 [7:58:00<1:24:21,  1.83s/it]
 85%|████████▌ | 16319/19090 [7:58:01<1:17:06,  1.67s/it][gradnorm] step 16319: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 685.2

 85%|████████▌ | 16320/19090 [7:58:04<1:30:38,  1.96s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 981.2817993164062, 'learning_rate': 3.1410483483215207e-06, 'recon_loss': 0.0015401780110551044, 'recon_static': 0.0005613301764242351, 'recon_motion': 0.0009788478244445286, 'recon_norm': 1.2823859006166458, 'recon_norm_std': 0.039326677192002536, 'target_norm': 1.6193304300308227, 'recon_cos': 0.6371196925640106, 'gradnorm/compressor.query': 0.7282520052446494, 'gradnorm/compressor.layer0': 19.458280096314798, 'gradnorm/compressor.layer1': 971.1061347762148, 'gradnorm/compressor.layer2': 134.28962404072774, 'gradnorm/compressor.layer3': 36.10446100542407, 'gradnorm/compressor.layer4': 9.645735166858962, 'gradnorm/compressor.layer5': 4.931974191998777, 'gradnorm/compressor.layer6': 2.962241014193794, 'gradnorm/compressor.layer7': 2.9788306698138833, 'gradnorm/decoder.layer0': 3.2014154105330253, 'gradnorm/decoder.layer1': 2.004881147633868, 'gradnorm/decoder.layer2': 1.096153615457302, 'gradnorm/max_param': 685.2350463867188, 'epoch': 8.55}

 85%|████████▌ | 16320/19090 [7:58:04<1:30:38,  1.96s/it]
 85%|████████▌ | 16321/19090 [7:58:05<1:21:19,  1.76s/it]
 86%|████████▌ | 16322/19090 [7:58:06<1:14:44,  1.62s/it]
 86%|████████▌ | 16323/19090 [7:58:09<1:27:58,  1.91s/it]
 86%|████████▌ | 16324/19090 [7:58:10<1:19:20,  1.72s/it]
 86%|████████▌ | 16325/19090 [7:58:11<1:13:15,  1.59s/it]
 86%|████████▌ | 16326/19090 [7:58:14<1:26:49,  1.88s/it]
 86%|████████▌ | 16327/19090 [7:58:15<1:18:28,  1.70s/it]
 86%|████████▌ | 16328/19090 [7:58:16<1:12:39,  1.58s/it]
 86%|████████▌ | 16329/19090 [7:58:19<1:25:09,  1.85s/it][gradnorm] step 16329: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 751.7

 86%|████████▌ | 16330/19090 [7:58:20<1:17:21,  1.68s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1074.4146728515625, 'learning_rate': 3.118901150069306e-06, 'recon_loss': 0.0015525470429565758, 'recon_static': 0.0005816614066134207, 'recon_motion': 0.0009708856217912398, 'recon_norm': 1.2977712839841842, 'recon_norm_std': 0.040562549233436586, 'target_norm': 1.6206222623586655, 'recon_cos': 0.6363110303878784, 'gradnorm/compressor.query': 0.5857891403566325, 'gradnorm/compressor.layer0': 16.243301654307, 'gradnorm/compressor.layer1': 1065.5421742005335, 'gradnorm/compressor.layer2': 132.26974391844115, 'gradnorm/compressor.layer3': 32.74927502477631, 'gradnorm/compressor.layer4': 9.94952011101305, 'gradnorm/compressor.layer5': 4.920719265349214, 'gradnorm/compressor.layer6': 3.138419976182189, 'gradnorm/compressor.layer7': 2.9592594434338206, 'gradnorm/decoder.layer0': 2.941237162356523, 'gradnorm/decoder.layer1': 1.8299911624924354, 'gradnorm/decoder.layer2': 1.0396412549800034, 'gradnorm/max_param': 751.6776123046875, 'epoch': 8.55}

 86%|████████▌ | 16330/19090 [7:58:20<1:17:21,  1.68s/it]
 86%|████████▌ | 16331/19090 [7:58:21<1:11:54,  1.56s/it]
 86%|████████▌ | 16332/19090 [7:58:24<1:23:34,  1.82s/it]
 86%|████████▌ | 16333/19090 [7:58:25<1:16:25,  1.66s/it]
 86%|████████▌ | 16334/19090 [7:58:27<1:11:51,  1.56s/it]
 86%|████████▌ | 16335/19090 [7:58:29<1:21:14,  1.77s/it]
 86%|████████▌ | 16336/19090 [7:58:30<1:14:43,  1.63s/it]
 86%|████████▌ | 16337/19090 [7:58:31<1:09:57,  1.52s/it]
 86%|████████▌ | 16338/19090 [7:58:34<1:30:39,  1.98s/it]
 86%|████████▌ | 16339/19090 [7:58:36<1:21:25,  1.78s/it][gradnorm] step 16339: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 476.6

 86%|████████▌ | 16340/19090 [7:58:37<1:14:55,  1.63s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 685.3350830078125, 'learning_rate': 3.0968271115104974e-06, 'recon_loss': 0.0015666623803554103, 'recon_static': 0.0005934004715527408, 'recon_motion': 0.0009732619160786271, 'recon_norm': 1.288610076904297, 'recon_norm_std': 0.040663902461528775, 'target_norm': 1.622071349620819, 'recon_cos': 0.6319848865270614, 'gradnorm/compressor.query': 0.3654233882000237, 'gradnorm/compressor.layer0': 14.013967358600835, 'gradnorm/compressor.layer1': 675.6058854835413, 'gradnorm/compressor.layer2': 108.41690636583162, 'gradnorm/compressor.layer3': 34.27668646983781, 'gradnorm/compressor.layer4': 8.490561237453807, 'gradnorm/compressor.layer5': 4.136115155297403, 'gradnorm/compressor.layer6': 2.6507391816389494, 'gradnorm/compressor.layer7': 2.576381856994995, 'gradnorm/decoder.layer0': 2.8647277425721454, 'gradnorm/decoder.layer1': 1.791687152065297, 'gradnorm/decoder.layer2': 0.952131034496406, 'gradnorm/max_param': 476.6492004394531, 'epoch': 8.56}

 86%|████████▌ | 16340/19090 [7:58:37<1:14:55,  1.63s/it]
 86%|████████▌ | 16341/19090 [7:58:40<1:28:27,  1.93s/it]
 86%|████████▌ | 16342/19090 [7:58:41<1:19:45,  1.74s/it]
 86%|████████▌ | 16343/19090 [7:58:42<1:13:42,  1.61s/it]
 86%|████████▌ | 16344/19090 [7:58:45<1:26:18,  1.89s/it]
 86%|████████▌ | 16345/19090 [7:58:46<1:18:15,  1.71s/it]
 86%|████████▌ | 16346/19090 [7:58:47<1:12:35,  1.59s/it]
 86%|████████▌ | 16347/19090 [7:58:50<1:28:45,  1.94s/it]
 86%|████████▌ | 16348/19090 [7:58:51<1:19:46,  1.75s/it]
 86%|████████▌ | 16349/19090 [7:58:53<1:13:31,  1.61s/it][gradnorm] step 16349: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 783.2

 86%|████████▌ | 16350/19090 [7:58:55<1:24:30,  1.85s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 1121.68505859375, 'learning_rate': 3.074826306449893e-06, 'recon_loss': 0.0017263333254959434, 'recon_static': 0.0007567936612758785, 'recon_motion': 0.0009695396511233412, 'recon_norm': 1.3726285994052887, 'recon_norm_std': 0.038903312385082246, 'target_norm': 1.6203088968992234, 'recon_cos': 0.6020844712853431, 'gradnorm/compressor.query': 0.49193661971480546, 'gradnorm/compressor.layer0': 16.451215352437206, 'gradnorm/compressor.layer1': 1109.7069432962921, 'gradnorm/compressor.layer2': 159.12581009220347, 'gradnorm/compressor.layer3': 32.08173993269965, 'gradnorm/compressor.layer4': 8.21510146929902, 'gradnorm/compressor.layer5': 3.742096519558834, 'gradnorm/compressor.layer6': 2.3144801142281857, 'gradnorm/compressor.layer7': 2.739664770349475, 'gradnorm/decoder.layer0': 2.8558432859171847, 'gradnorm/decoder.layer1': 1.8547526373026282, 'gradnorm/decoder.layer2': 0.9974727285572238, 'gradnorm/max_param': 783.162841796875, 'epoch': 8.56}

 86%|████████▌ | 16350/19090 [7:58:55<1:24:30,  1.85s/it]
 86%|████████▌ | 16351/19090 [7:58:56<1:16:46,  1.68s/it]
 86%|████████▌ | 16352/19090 [7:58:58<1:11:21,  1.56s/it]
 86%|████████▌ | 16353/19090 [7:59:01<1:29:05,  1.95s/it]
 86%|████████▌ | 16354/19090 [7:59:02<1:19:57,  1.75s/it]
 86%|████████▌ | 16355/19090 [7:59:03<1:13:30,  1.61s/it]
 86%|████████▌ | 16356/19090 [7:59:06<1:30:47,  1.99s/it]
 86%|████████▌ | 16357/19090 [7:59:07<1:21:07,  1.78s/it]
 86%|████████▌ | 16358/19090 [7:59:09<1:14:21,  1.63s/it]
 86%|████████▌ | 16359/19090 [7:59:11<1:28:43,  1.95s/it][gradnorm] step 16359: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 631.1

 86%|████████▌ | 16360/19090 [7:59:13<1:19:42,  1.75s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 903.1712646484375, 'learning_rate': 3.052898808447427e-06, 'recon_loss': 0.0017227244999958203, 'recon_static': 0.0007429558478179388, 'recon_motion': 0.0009797686376259661, 'recon_norm': 1.3597546607255935, 'recon_norm_std': 0.03842898169532418, 'target_norm': 1.6220240414142608, 'recon_cos': 0.6011245250701904, 'gradnorm/compressor.query': 0.5242928306208735, 'gradnorm/compressor.layer0': 12.666536497819202, 'gradnorm/compressor.layer1': 894.3749213836444, 'gradnorm/compressor.layer2': 118.9948052306276, 'gradnorm/compressor.layer3': 37.08071755316788, 'gradnorm/compressor.layer4': 8.331385219475388, 'gradnorm/compressor.layer5': 4.359769413651365, 'gradnorm/compressor.layer6': 2.633617334704402, 'gradnorm/compressor.layer7': 2.722821574069581, 'gradnorm/decoder.layer0': 2.917188833815643, 'gradnorm/decoder.layer1': 1.9406999803883636, 'gradnorm/decoder.layer2': 1.1311891170578463, 'gradnorm/max_param': 631.1408081054688, 'epoch': 8.57}

 86%|████████▌ | 16360/19090 [7:59:13<1:19:42,  1.75s/it]
 86%|████████▌ | 16361/19090 [7:59:14<1:13:21,  1.61s/it]
 86%|████████▌ | 16362/19090 [7:59:17<1:35:48,  2.11s/it]
 86%|████████▌ | 16363/19090 [7:59:18<1:24:40,  1.86s/it]
 86%|████████▌ | 16364/19090 [7:59:20<1:16:50,  1.69s/it]
 86%|████████▌ | 16365/19090 [7:59:23<1:40:03,  2.20s/it]
 86%|████████▌ | 16366/19090 [7:59:24<1:27:31,  1.93s/it]
 86%|████████▌ | 16367/19090 [7:59:26<1:18:46,  1.74s/it]
 86%|████████▌ | 16368/19090 [7:59:28<1:33:38,  2.06s/it]
 86%|████████▌ | 16369/19090 [7:59:30<1:23:04,  1.83s/it][gradnorm] step 16369: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 564.3

 86%|████████▌ | 16370/19090 [7:59:31<1:15:40,  1.67s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 810.1553344726562, 'learning_rate': 3.031044690817916e-06, 'recon_loss': 0.001598188851494342, 'recon_static': 0.0006150982255348936, 'recon_motion': 0.000983090652152896, 'recon_norm': 1.3065205246210099, 'recon_norm_std': 0.03949285447597504, 'target_norm': 1.6222505182027818, 'recon_cos': 0.62609793394804, 'gradnorm/compressor.query': 0.310227127528498, 'gradnorm/compressor.layer0': 11.616535345472728, 'gradnorm/compressor.layer1': 799.5307608216209, 'gradnorm/compressor.layer2': 125.33665602588096, 'gradnorm/compressor.layer3': 34.1270585888923, 'gradnorm/compressor.layer4': 7.267822458352489, 'gradnorm/compressor.layer5': 3.577467012499892, 'gradnorm/compressor.layer6': 2.7192657683030084, 'gradnorm/compressor.layer7': 2.7993886723049286, 'gradnorm/decoder.layer0': 2.782289546469211, 'gradnorm/decoder.layer1': 1.8781486776322442, 'gradnorm/decoder.layer2': 1.061868648319209, 'gradnorm/max_param': 564.2807006835938, 'epoch': 8.58}

 86%|████████▌ | 16370/19090 [7:59:31<1:15:40,  1.67s/it]
 86%|████████▌ | 16371/19090 [7:59:34<1:33:20,  2.06s/it]
 86%|████████▌ | 16372/19090 [7:59:35<1:22:47,  1.83s/it]
 86%|████████▌ | 16373/19090 [7:59:37<1:15:28,  1.67s/it]
 86%|████████▌ | 16374/19090 [7:59:39<1:24:34,  1.87s/it]
 86%|████████▌ | 16375/19090 [7:59:40<1:16:34,  1.69s/it]
 86%|████████▌ | 16376/19090 [7:59:42<1:11:05,  1.57s/it]
 86%|████████▌ | 16377/19090 [7:59:45<1:32:38,  2.05s/it]
 86%|████████▌ | 16378/19090 [7:59:46<1:22:13,  1.82s/it]
 86%|████████▌ | 16379/19090 [7:59:47<1:14:57,  1.66s/it][gradnorm] step 16379: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 666.4

 86%|████████▌ | 16380/19090 [7:59:50<1:32:09,  2.04s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 954.7467651367188, 'learning_rate': 3.009264026630851e-06, 'recon_loss': 0.0016413688252214342, 'recon_static': 0.0006687555796816014, 'recon_motion': 0.0009726132760988548, 'recon_norm': 1.3236523151397706, 'recon_norm_std': 0.039528759010136125, 'target_norm': 1.6208125561475755, 'recon_cos': 0.6163228675723076, 'gradnorm/compressor.query': 0.4827045276734945, 'gradnorm/compressor.layer0': 13.260553152016733, 'gradnorm/compressor.layer1': 944.6069949984491, 'gradnorm/compressor.layer2': 133.37089965580947, 'gradnorm/compressor.layer3': 34.43936445724209, 'gradnorm/compressor.layer4': 7.756041356307478, 'gradnorm/compressor.layer5': 4.148943451741307, 'gradnorm/compressor.layer6': 2.7621744473258234, 'gradnorm/compressor.layer7': 3.0485533175618054, 'gradnorm/decoder.layer0': 3.1352266442052747, 'gradnorm/decoder.layer1': 1.9012431695478087, 'gradnorm/decoder.layer2': 1.1686039348861827, 'gradnorm/max_param': 666.4142456054688, 'epoch': 8.58}

 86%|████████▌ | 16380/19090 [7:59:50<1:32:09,  2.04s/it]
 86%|████████▌ | 16381/19090 [7:59:51<1:21:55,  1.81s/it]
 86%|████████▌ | 16382/19090 [7:59:53<1:14:54,  1.66s/it]
 86%|████████▌ | 16383/19090 [7:59:56<1:30:34,  2.01s/it]
 86%|████████▌ | 16384/19090 [7:59:57<1:20:59,  1.80s/it]
 86%|████████▌ | 16385/19090 [7:59:58<1:14:13,  1.65s/it]
 86%|████████▌ | 16386/19090 [8:00:01<1:29:26,  1.98s/it]
 86%|████████▌ | 16387/19090 [8:00:02<1:20:02,  1.78s/it]
 86%|████████▌ | 16388/19090 [8:00:04<1:13:29,  1.63s/it]
 86%|████████▌ | 16389/19090 [8:00:06<1:30:36,  2.01s/it][gradnorm] step 16389: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 678.6

 86%|████████▌ | 16390/19090 [8:00:08<1:20:51,  1.80s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 972.8004760742188, 'learning_rate': 2.9875568887101273e-06, 'recon_loss': 0.0016705468675354495, 'recon_static': 0.0007019729615421966, 'recon_motion': 0.000968573967111297, 'recon_norm': 1.3492897480726243, 'recon_norm_std': 0.03867811923846602, 'target_norm': 1.6198806643486023, 'recon_cos': 0.6119219705462455, 'gradnorm/compressor.query': 0.4408073413243603, 'gradnorm/compressor.layer0': 17.89986626772233, 'gradnorm/compressor.layer1': 961.6593406711131, 'gradnorm/compressor.layer2': 139.98258122664762, 'gradnorm/compressor.layer3': 39.4168954940724, 'gradnorm/compressor.layer4': 6.6913423997245545, 'gradnorm/compressor.layer5': 3.604019882358998, 'gradnorm/compressor.layer6': 2.5105520715773975, 'gradnorm/compressor.layer7': 2.608031789474754, 'gradnorm/decoder.layer0': 2.7341856972697633, 'gradnorm/decoder.layer1': 1.845791672293142, 'gradnorm/decoder.layer2': 0.9730241378522787, 'gradnorm/max_param': 678.6034545898438, 'epoch': 8.59}

 86%|████████▌ | 16390/19090 [8:00:08<1:20:51,  1.80s/it]
 86%|████████▌ | 16391/19090 [8:00:09<1:13:59,  1.64s/it]
 86%|████████▌ | 16392/19090 [8:00:12<1:32:05,  2.05s/it]
 86%|████████▌ | 16393/19090 [8:00:13<1:21:53,  1.82s/it]
 86%|████████▌ | 16394/19090 [8:00:15<1:14:38,  1.66s/it]
 86%|████████▌ | 16395/19090 [8:00:18<1:32:13,  2.05s/it]
 86%|████████▌ | 16396/19090 [8:00:19<1:22:02,  1.83s/it]
 86%|████████▌ | 16397/19090 [8:00:20<1:14:53,  1.67s/it]
 86%|████████▌ | 16398/19090 [8:00:23<1:27:15,  1.95s/it]
 86%|████████▌ | 16399/19090 [8:00:24<1:18:30,  1.75s/it][gradnorm] step 16399: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 577.4

 86%|████████▌ | 16400/19090 [8:00:25<1:12:14,  1.61s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 827.0722045898438, 'learning_rate': 2.9659233496337786e-06, 'recon_loss': 0.0016513978480361402, 'recon_static': 0.0006712045898893848, 'recon_motion': 0.0009801933076232672, 'recon_norm': 1.3348161548376083, 'recon_norm_std': 0.038620331324636936, 'target_norm': 1.620389896631241, 'recon_cos': 0.6150673866271973, 'gradnorm/compressor.query': 0.38841394271678403, 'gradnorm/compressor.layer0': 14.580297768146318, 'gradnorm/compressor.layer1': 818.194429827043, 'gradnorm/compressor.layer2': 113.80419811154815, 'gradnorm/compressor.layer3': 36.30036573830042, 'gradnorm/compressor.layer4': 8.389233692260538, 'gradnorm/compressor.layer5': 4.703041253342315, 'gradnorm/compressor.layer6': 2.8632514400413607, 'gradnorm/compressor.layer7': 2.946931995799396, 'gradnorm/decoder.layer0': 3.1590630110312796, 'gradnorm/decoder.layer1': 1.8923677971242523, 'gradnorm/decoder.layer2': 1.2088444494766046, 'gradnorm/max_param': 577.4072265625, 'epoch': 8.59}

 86%|████████▌ | 16400/19090 [8:00:25<1:12:14,  1.61s/it]
 86%|████████▌ | 16401/19090 [8:00:28<1:25:26,  1.91s/it]
 86%|████████▌ | 16402/19090 [8:00:29<1:17:08,  1.72s/it]
 86%|████████▌ | 16403/19090 [8:00:31<1:11:18,  1.59s/it]
 86%|████████▌ | 16404/19090 [8:00:33<1:27:10,  1.95s/it]
 86%|████████▌ | 16405/19090 [8:00:35<1:18:21,  1.75s/it]
 86%|████████▌ | 16406/19090 [8:00:36<1:12:09,  1.61s/it]
 86%|████████▌ | 16407/19090 [8:00:39<1:33:15,  2.09s/it]
 86%|████████▌ | 16408/19090 [8:00:40<1:22:32,  1.85s/it]
 86%|████████▌ | 16409/19090 [8:00:42<1:15:05,  1.68s/it][gradnorm] step 16409: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 757.8

 86%|████████▌ | 16410/19090 [8:00:44<1:25:08,  1.91s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1084.1522216796875, 'learning_rate': 2.9443634817337945e-06, 'recon_loss': 0.001584435542463325, 'recon_static': 0.0006124095714767463, 'recon_motion': 0.0009720259346067906, 'recon_norm': 1.2994395732879638, 'recon_norm_std': 0.03911012746393681, 'target_norm': 1.6208717852830887, 'recon_cos': 0.6281330466270447, 'gradnorm/compressor.query': 0.2932942807123124, 'gradnorm/compressor.layer0': 9.226625343774126, 'gradnorm/compressor.layer1': 1073.7053832406727, 'gradnorm/compressor.layer2': 143.01803453804348, 'gradnorm/compressor.layer3': 43.45913636052982, 'gradnorm/compressor.layer4': 8.236333451281773, 'gradnorm/compressor.layer5': 4.238843523690743, 'gradnorm/compressor.layer6': 2.922919591538336, 'gradnorm/compressor.layer7': 2.8954231648672533, 'gradnorm/decoder.layer0': 2.787612751959805, 'gradnorm/decoder.layer1': 1.7274890329589738, 'gradnorm/decoder.layer2': 1.1371980685597058, 'gradnorm/max_param': 757.808349609375, 'epoch': 8.6}

 86%|████████▌ | 16410/19090 [8:00:44<1:25:08,  1.91s/it]
 86%|████████▌ | 16411/19090 [8:00:45<1:16:49,  1.72s/it]
 86%|████████▌ | 16412/19090 [8:00:47<1:11:01,  1.59s/it]
 86%|████████▌ | 16413/19090 [8:00:49<1:25:57,  1.93s/it]
 86%|████████▌ | 16414/19090 [8:00:51<1:17:22,  1.74s/it]
 86%|████████▌ | 16415/19090 [8:00:52<1:11:19,  1.60s/it]
 86%|████████▌ | 16416/19090 [8:00:54<1:20:07,  1.80s/it]
 86%|████████▌ | 16417/19090 [8:00:55<1:13:14,  1.64s/it]
 86%|████████▌ | 16418/19090 [8:00:57<1:08:24,  1.54s/it]
 86%|████████▌ | 16419/19090 [8:01:00<1:28:43,  1.99s/it][gradnorm] step 16419: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 632.6

 86%|████████▌ | 16420/19090 [8:01:01<1:19:14,  1.78s/it]
                                                         
{'loss': 0.0021, 'grad_norm': 904.9856567382812, 'learning_rate': 2.9228773570958324e-06, 'recon_loss': 0.0021423193713417275, 'recon_static': 0.0011724909869371913, 'recon_motion': 0.0009698283465695567, 'recon_norm': 1.4971796482801438, 'recon_norm_std': 0.03701949473470449, 'target_norm': 1.620338362455368, 'recon_cos': 0.5661212280392647, 'gradnorm/compressor.query': 1.0222002077726788, 'gradnorm/compressor.layer0': 33.74737583837538, 'gradnorm/compressor.layer1': 896.7690965348884, 'gradnorm/compressor.layer2': 112.82361873877517, 'gradnorm/compressor.layer3': 28.854698541653274, 'gradnorm/compressor.layer4': 7.606144087970125, 'gradnorm/compressor.layer5': 4.126227687439081, 'gradnorm/compressor.layer6': 2.7254494742575868, 'gradnorm/compressor.layer7': 2.724686378840161, 'gradnorm/decoder.layer0': 2.8952209227603607, 'gradnorm/decoder.layer1': 1.811789801003422, 'gradnorm/decoder.layer2': 1.0686406910639306, 'gradnorm/max_param': 632.6458129882812, 'epoch': 8.6}

 86%|████████▌ | 16420/19090 [8:01:01<1:19:14,  1.78s/it]
 86%|████████▌ | 16421/19090 [8:01:02<1:12:38,  1.63s/it]
 86%|████████▌ | 16422/19090 [8:01:06<1:32:58,  2.09s/it]
 86%|████████▌ | 16423/19090 [8:01:07<1:22:14,  1.85s/it]
 86%|████████▌ | 16424/19090 [8:01:08<1:14:38,  1.68s/it]
 86%|████████▌ | 16425/19090 [8:01:11<1:26:19,  1.94s/it]
 86%|████████▌ | 16426/19090 [8:01:12<1:17:47,  1.75s/it]
 86%|████████▌ | 16427/19090 [8:01:13<1:11:29,  1.61s/it]
 86%|████████▌ | 16428/19090 [8:01:16<1:28:20,  1.99s/it]
 86%|████████▌ | 16429/19090 [8:01:17<1:19:07,  1.78s/it][gradnorm] step 16429: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 261.9

 86%|████████▌ | 16430/19090 [8:01:19<1:12:28,  1.63s/it]
                                                         
{'loss': 0.0072, 'grad_norm': 380.9798278808594, 'learning_rate': 2.9014650475589655e-06, 'recon_loss': 0.007194284343859181, 'recon_static': 0.006223854492418468, 'recon_motion': 0.0009704300158773549, 'recon_norm': 2.8110715627670286, 'recon_norm_std': 0.024115724861621855, 'target_norm': 1.6192118316888808, 'recon_cos': 0.3007855784147978, 'gradnorm/compressor.query': 0.8219561141566795, 'gradnorm/compressor.layer0': 23.087710115956444, 'gradnorm/compressor.layer1': 372.7171766232407, 'gradnorm/compressor.layer2': 66.65711438726927, 'gradnorm/compressor.layer3': 33.006228697209565, 'gradnorm/compressor.layer4': 8.559387532364639, 'gradnorm/compressor.layer5': 6.754371357882652, 'gradnorm/compressor.layer6': 3.732195229987468, 'gradnorm/compressor.layer7': 3.251696950542363, 'gradnorm/decoder.layer0': 3.1976186176455714, 'gradnorm/decoder.layer1': 2.299666181471479, 'gradnorm/decoder.layer2': 1.5356127809962965, 'gradnorm/max_param': 261.8504943847656, 'epoch': 8.61}

 86%|████████▌ | 16430/19090 [8:01:19<1:12:28,  1.63s/it]
 86%|████████▌ | 16431/19090 [8:01:22<1:29:25,  2.02s/it]
 86%|████████▌ | 16432/19090 [8:01:23<1:19:40,  1.80s/it]
 86%|████████▌ | 16433/19090 [8:01:24<1:12:51,  1.65s/it]
 86%|████████▌ | 16434/19090 [8:01:27<1:25:40,  1.94s/it]
 86%|████████▌ | 16435/19090 [8:01:28<1:17:02,  1.74s/it]
 86%|████████▌ | 16436/19090 [8:01:29<1:10:56,  1.60s/it]
 86%|████████▌ | 16437/19090 [8:01:32<1:26:33,  1.96s/it]
 86%|████████▌ | 16438/19090 [8:01:33<1:17:33,  1.75s/it]
 86%|████████▌ | 16439/19090 [8:01:35<1:11:20,  1.61s/it][gradnorm] step 16439: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 587.8

 86%|████████▌ | 16440/19090 [8:01:38<1:26:22,  1.96s/it]
                                                         
{'loss': 0.0064, 'grad_norm': 841.0340576171875, 'learning_rate': 2.880126624715504e-06, 'recon_loss': 0.006434977159369737, 'recon_static': 0.005463405878981575, 'recon_motion': 0.0009715712716570124, 'recon_norm': 2.710092270374298, 'recon_norm_std': 0.023674609279260038, 'target_norm': 1.6206036269664765, 'recon_cos': 0.31830798760056495, 'gradnorm/compressor.query': 0.9420964097357388, 'gradnorm/compressor.layer0': 33.576736243966224, 'gradnorm/compressor.layer1': 833.2117077910032, 'gradnorm/compressor.layer2': 102.6930014122798, 'gradnorm/compressor.layer3': 36.046908864310055, 'gradnorm/compressor.layer4': 8.528635148291652, 'gradnorm/compressor.layer5': 4.607710198230139, 'gradnorm/compressor.layer6': 2.476202135311876, 'gradnorm/compressor.layer7': 3.2354430863054415, 'gradnorm/decoder.layer0': 2.8602081919301052, 'gradnorm/decoder.layer1': 1.6755548070756392, 'gradnorm/decoder.layer2': 1.1669800825568581, 'gradnorm/max_param': 587.76220703125, 'epoch': 8.61}

 86%|████████▌ | 16440/19090 [8:01:38<1:26:22,  1.96s/it]
 86%|████████▌ | 16441/19090 [8:01:39<1:17:23,  1.75s/it]
 86%|████████▌ | 16442/19090 [8:01:40<1:11:12,  1.61s/it]
 86%|████████▌ | 16443/19090 [8:01:43<1:25:25,  1.94s/it]
 86%|████████▌ | 16444/19090 [8:01:44<1:16:43,  1.74s/it]
 86%|████████▌ | 16445/19090 [8:01:45<1:10:40,  1.60s/it]
 86%|████████▌ | 16446/19090 [8:01:48<1:28:05,  2.00s/it]
 86%|████████▌ | 16447/19090 [8:01:50<1:18:34,  1.78s/it]
 86%|████████▌ | 16448/19090 [8:01:51<1:11:59,  1.63s/it]
 86%|████████▌ | 16449/19090 [8:01:54<1:28:30,  2.01s/it][gradnorm] step 16449: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 450.4

 86%|████████▌ | 16450/19090 [8:01:55<1:19:01,  1.80s/it]
                                                         
{'loss': 0.005, 'grad_norm': 643.2345581054688, 'learning_rate': 2.8588621599106997e-06, 'recon_loss': 0.005003603192744776, 'recon_static': 0.004033890756545588, 'recon_motion': 0.0009697124827653169, 'recon_norm': 2.3396601021289825, 'recon_norm_std': 0.026749441400170326, 'target_norm': 1.6230939388275147, 'recon_cos': 0.3459602534770966, 'gradnorm/compressor.query': 0.6237816857219425, 'gradnorm/compressor.layer0': 23.55411742676652, 'gradnorm/compressor.layer1': 638.4766538409999, 'gradnorm/compressor.layer2': 68.03737196846303, 'gradnorm/compressor.layer3': 28.857313940682243, 'gradnorm/compressor.layer4': 6.574809419023823, 'gradnorm/compressor.layer5': 3.7904096902329716, 'gradnorm/compressor.layer6': 2.4462043671403766, 'gradnorm/compressor.layer7': 2.5556998857109092, 'gradnorm/decoder.layer0': 2.564315710867607, 'gradnorm/decoder.layer1': 1.801354174384445, 'gradnorm/decoder.layer2': 1.152758559228926, 'gradnorm/max_param': 450.4052429199219, 'epoch': 8.62}

 86%|████████▌ | 16450/19090 [8:01:55<1:19:01,  1.80s/it]
 86%|████████▌ | 16451/19090 [8:01:56<1:12:18,  1.64s/it]
 86%|████████▌ | 16452/19090 [8:01:59<1:28:01,  2.00s/it]
 86%|████████▌ | 16453/19090 [8:02:00<1:18:32,  1.79s/it]
 86%|████████▌ | 16454/19090 [8:02:02<1:11:57,  1.64s/it]
 86%|████████▌ | 16455/19090 [8:02:05<1:29:53,  2.05s/it]
 86%|████████▌ | 16456/19090 [8:02:06<1:19:51,  1.82s/it]
 86%|████████▌ | 16457/19090 [8:02:07<1:12:47,  1.66s/it]
 86%|████████▌ | 16458/19090 [8:02:10<1:26:04,  1.96s/it]
 86%|████████▌ | 16459/19090 [8:02:11<1:17:05,  1.76s/it][gradnorm] step 16459: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 542.9

 86%|████████▌ | 16460/19090 [8:02:13<1:10:55,  1.62s/it]
                                                         
{'loss': 0.0049, 'grad_norm': 780.3582763671875, 'learning_rate': 2.8376717242425075e-06, 'recon_loss': 0.004866699472768232, 'recon_static': 0.003902257661684416, 'recon_motion': 0.0009644419144024142, 'recon_norm': 2.3133031487464906, 'recon_norm_std': 0.027201097318902613, 'target_norm': 1.6195759057998658, 'recon_cos': 0.37710094451904297, 'gradnorm/compressor.query': 0.7600919125843741, 'gradnorm/compressor.layer0': 27.61459941904316, 'gradnorm/compressor.layer1': 769.4177181479512, 'gradnorm/compressor.layer2': 122.0855193028129, 'gradnorm/compressor.layer3': 34.416040841263246, 'gradnorm/compressor.layer4': 7.549464569305768, 'gradnorm/compressor.layer5': 4.179848355914648, 'gradnorm/compressor.layer6': 2.3453586398471375, 'gradnorm/compressor.layer7': 2.7885887106878076, 'gradnorm/decoder.layer0': 3.109941133880378, 'gradnorm/decoder.layer1': 2.0450880097941884, 'gradnorm/decoder.layer2': 1.147906485556831, 'gradnorm/max_param': 542.9146728515625, 'epoch': 8.62}

 86%|████████▌ | 16460/19090 [8:02:13<1:10:55,  1.62s/it]
 86%|████████▌ | 16461/19090 [8:02:15<1:22:42,  1.89s/it]
 86%|████████▌ | 16462/19090 [8:02:16<1:14:42,  1.71s/it]
 86%|████████▌ | 16463/19090 [8:02:18<1:09:10,  1.58s/it]
 86%|████████▌ | 16464/19090 [8:02:20<1:25:54,  1.96s/it]
 86%|████████▌ | 16465/19090 [8:02:22<1:17:11,  1.76s/it]
 86%|████████▋ | 16466/19090 [8:02:23<1:10:48,  1.62s/it]
 86%|████████▋ | 16467/19090 [8:02:26<1:25:35,  1.96s/it]
 86%|████████▋ | 16468/19090 [8:02:27<1:16:44,  1.76s/it]
 86%|████████▋ | 16469/19090 [8:02:28<1:10:29,  1.61s/it][gradnorm] step 16469: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 440.9

 86%|████████▋ | 16470/19090 [8:02:31<1:26:00,  1.97s/it]
                                                         
{'loss': 0.0048, 'grad_norm': 630.9623413085938, 'learning_rate': 2.8165553885613844e-06, 'recon_loss': 0.004812086891615763, 'recon_static': 0.003832658310420811, 'recon_motion': 0.0009794286845135503, 'recon_norm': 2.290616562962532, 'recon_norm_std': 0.02623789724893868, 'target_norm': 1.6232017785310746, 'recon_cos': 0.35910557955503464, 'gradnorm/compressor.query': 0.5865806356074379, 'gradnorm/compressor.layer0': 22.008223643878768, 'gradnorm/compressor.layer1': 624.8652104654251, 'gradnorm/compressor.layer2': 76.58320510847989, 'gradnorm/compressor.layer3': 34.57638806075795, 'gradnorm/compressor.layer4': 7.549616155909021, 'gradnorm/compressor.layer5': 5.003489611728772, 'gradnorm/compressor.layer6': 2.9376532332041316, 'gradnorm/compressor.layer7': 2.859615961980246, 'gradnorm/decoder.layer0': 2.932583204358963, 'gradnorm/decoder.layer1': 1.7718666278298196, 'gradnorm/decoder.layer2': 1.076689330936212, 'gradnorm/max_param': 440.8970642089844, 'epoch': 8.63}

 86%|████████▋ | 16470/19090 [8:02:31<1:26:00,  1.97s/it]
 86%|████████▋ | 16471/19090 [8:02:32<1:17:05,  1.77s/it]
 86%|████████▋ | 16472/19090 [8:02:34<1:10:56,  1.63s/it]
 86%|████████▋ | 16473/19090 [8:02:37<1:28:31,  2.03s/it]
 86%|████████▋ | 16474/19090 [8:02:38<1:18:54,  1.81s/it]
 86%|████████▋ | 16475/19090 [8:02:39<1:12:16,  1.66s/it]
 86%|████████▋ | 16476/19090 [8:02:42<1:25:38,  1.97s/it]
 86%|████████▋ | 16477/19090 [8:02:43<1:16:51,  1.76s/it]
 86%|████████▋ | 16478/19090 [8:02:45<1:10:47,  1.63s/it]
 86%|████████▋ | 16479/19090 [8:02:48<1:29:16,  2.05s/it][gradnorm] step 16479: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 392.0

 86%|████████▋ | 16480/19090 [8:02:49<1:19:33,  1.83s/it]
                                                         
{'loss': 0.0034, 'grad_norm': 560.9237060546875, 'learning_rate': 2.795513223470045e-06, 'recon_loss': 0.003367617988260463, 'recon_static': 0.00239669420698192, 'recon_motion': 0.0009709237958304584, 'recon_norm': 1.917766308784485, 'recon_norm_std': 0.029839681135490535, 'target_norm': 1.6214557826519012, 'recon_cos': 0.4207158885896206, 'gradnorm/compressor.query': 0.360851757384571, 'gradnorm/compressor.layer0': 13.69217354750801, 'gradnorm/compressor.layer1': 555.507397565505, 'gradnorm/compressor.layer2': 71.2174918987165, 'gradnorm/compressor.layer3': 26.37011094904312, 'gradnorm/compressor.layer4': 7.144914069329698, 'gradnorm/compressor.layer5': 3.6965337755436987, 'gradnorm/compressor.layer6': 2.793335705105936, 'gradnorm/compressor.layer7': 2.9344042434291464, 'gradnorm/decoder.layer0': 2.588005783109969, 'gradnorm/decoder.layer1': 1.7662031189351992, 'gradnorm/decoder.layer2': 1.0194748903094355, 'gradnorm/max_param': 391.97967529296875, 'epoch': 8.63}

 86%|████████▋ | 16480/19090 [8:02:49<1:19:33,  1.83s/it]
 86%|████████▋ | 16481/19090 [8:02:50<1:12:34,  1.67s/it]
 86%|████████▋ | 16482/19090 [8:02:53<1:29:47,  2.07s/it]
 86%|████████▋ | 16483/19090 [8:02:55<1:19:39,  1.83s/it]
 86%|████████▋ | 16484/19090 [8:02:56<1:12:28,  1.67s/it]
 86%|████████▋ | 16485/19090 [8:02:59<1:28:19,  2.03s/it]
 86%|████████▋ | 16486/19090 [8:03:00<1:18:37,  1.81s/it]
 86%|████████▋ | 16487/19090 [8:03:01<1:11:47,  1.65s/it]
 86%|████████▋ | 16488/19090 [8:03:04<1:21:10,  1.87s/it]
 86%|████████▋ | 16489/19090 [8:03:05<1:13:34,  1.70s/it][gradnorm] step 16489: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 589.0

 86%|████████▋ | 16490/19090 [8:03:06<1:08:11,  1.57s/it]
                                                         
{'loss': 0.0031, 'grad_norm': 847.7701416015625, 'learning_rate': 2.7745452993231807e-06, 'recon_loss': 0.0030986671277787536, 'recon_static': 0.00212258082756307, 'recon_motion': 0.0009760862158145756, 'recon_norm': 1.8779381811618805, 'recon_norm_std': 0.031124343816190957, 'target_norm': 1.622628030180931, 'recon_cos': 0.45298941880464555, 'gradnorm/compressor.query': 0.7679965599201564, 'gradnorm/compressor.layer0': 24.309841295020057, 'gradnorm/compressor.layer1': 835.4359266275302, 'gradnorm/compressor.layer2': 134.38541384065832, 'gradnorm/compressor.layer3': 44.607655122274835, 'gradnorm/compressor.layer4': 8.601517707507123, 'gradnorm/compressor.layer5': 3.910631453899114, 'gradnorm/compressor.layer6': 2.745752609139156, 'gradnorm/compressor.layer7': 3.1658973930634207, 'gradnorm/decoder.layer0': 3.0508623686298564, 'gradnorm/decoder.layer1': 1.7760880614305634, 'gradnorm/decoder.layer2': 1.018772295842433, 'gradnorm/max_param': 589.0428466796875, 'epoch': 8.64}

 86%|████████▋ | 16490/19090 [8:03:06<1:08:11,  1.57s/it]
 86%|████████▋ | 16491/19090 [8:03:09<1:23:10,  1.92s/it]
 86%|████████▋ | 16492/19090 [8:03:10<1:14:55,  1.73s/it]
 86%|████████▋ | 16493/19090 [8:03:12<1:09:09,  1.60s/it]
 86%|████████▋ | 16494/19090 [8:03:14<1:23:23,  1.93s/it]
 86%|████████▋ | 16495/19090 [8:03:16<1:15:05,  1.74s/it]
 86%|████████▋ | 16496/19090 [8:03:17<1:09:14,  1.60s/it]
 86%|████████▋ | 16497/19090 [8:03:19<1:22:30,  1.91s/it]
 86%|████████▋ | 16498/19090 [8:03:21<1:14:32,  1.73s/it]
 86%|████████▋ | 16499/19090 [8:03:22<1:08:47,  1.59s/it][gradnorm] step 16499: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 611.2

 86%|████████▋ | 16500/19090 [8:03:25<1:22:39,  1.92s/it]
                                                         
{'loss': 0.0023, 'grad_norm': 877.4600830078125, 'learning_rate': 2.753651686227279e-06, 'recon_loss': 0.0022692511294735594, 'recon_static': 0.0012956655118614435, 'recon_motion': 0.0009735855928738602, 'recon_norm': 1.5588153451681137, 'recon_norm_std': 0.03468662304803729, 'target_norm': 1.622154739499092, 'recon_cos': 0.519148787856102, 'gradnorm/compressor.query': 0.5068730100035262, 'gradnorm/compressor.layer0': 15.221836286294693, 'gradnorm/compressor.layer1': 866.0643379680288, 'gradnorm/compressor.layer2': 135.44088885931012, 'gradnorm/compressor.layer3': 33.96575280015519, 'gradnorm/compressor.layer4': 9.936456589532511, 'gradnorm/compressor.layer5': 3.5443681773155546, 'gradnorm/compressor.layer6': 2.2292940216286827, 'gradnorm/compressor.layer7': 3.116563624618298, 'gradnorm/decoder.layer0': 2.9716567415623314, 'gradnorm/decoder.layer1': 1.7542088848600106, 'gradnorm/decoder.layer2': 0.9015651497892846, 'gradnorm/max_param': 611.2406005859375, 'epoch': 8.64}

 86%|████████▋ | 16500/19090 [8:03:25<1:22:39,  1.92s/it]
 86%|████████▋ | 16501/19090 [8:03:28<1:45:24,  2.44s/it]
 86%|████████▋ | 16502/19090 [8:03:30<1:30:25,  2.10s/it]
 86%|████████▋ | 16503/19090 [8:03:31<1:19:55,  1.85s/it]
 86%|████████▋ | 16504/19090 [8:03:32<1:12:33,  1.68s/it]
 86%|████████▋ | 16505/19090 [8:03:34<1:07:24,  1.56s/it]
 86%|████████▋ | 16506/19090 [8:03:35<1:03:48,  1.48s/it]
 86%|████████▋ | 16507/19090 [8:03:36<1:01:19,  1.42s/it]
 86%|████████▋ | 16508/19090 [8:03:37<59:32,  1.38s/it]  
 86%|████████▋ | 16509/19090 [8:03:39<58:20,  1.36s/it][gradnorm] step 16509: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 544.0

 86%|████████▋ | 16510/19090 [8:03:40<57:30,  1.34s/it]
                                                       
{'loss': 0.0019, 'grad_norm': 779.0339965820312, 'learning_rate': 2.7328324540403794e-06, 'recon_loss': 0.0018609492894029246, 'recon_static': 0.0008829557380522601, 'recon_motion': 0.0009779935164260678, 'recon_norm': 1.4202609717845918, 'recon_norm_std': 0.03690948151051998, 'target_norm': 1.621775558590889, 'recon_cos': 0.5776953712105751, 'gradnorm/compressor.query': 0.5738978884301006, 'gradnorm/compressor.layer0': 23.017576716297157, 'gradnorm/compressor.layer1': 771.2031590313929, 'gradnorm/compressor.layer2': 102.97900176583332, 'gradnorm/compressor.layer3': 30.32292628356369, 'gradnorm/compressor.layer4': 7.12773936297573, 'gradnorm/compressor.layer5': 3.537831342835829, 'gradnorm/compressor.layer6': 2.077428485911057, 'gradnorm/compressor.layer7': 2.987858681315386, 'gradnorm/decoder.layer0': 2.2909034382831583, 'gradnorm/decoder.layer1': 1.6852407520030823, 'gradnorm/decoder.layer2': 0.8457357563304437, 'gradnorm/max_param': 544.0363159179688, 'epoch': 8.65}

 86%|████████▋ | 16510/19090 [8:03:40<57:30,  1.34s/it]
 86%|████████▋ | 16511/19090 [8:03:41<56:54,  1.32s/it]
 86%|████████▋ | 16512/19090 [8:03:44<1:11:39,  1.67s/it]
 87%|████████▋ | 16513/19090 [8:03:45<1:06:39,  1.55s/it]
 87%|████████▋ | 16514/19090 [8:03:46<1:03:09,  1.47s/it]
 87%|████████▋ | 16515/19090 [8:03:49<1:20:18,  1.87s/it]
 87%|████████▋ | 16516/19090 [8:03:50<1:12:48,  1.70s/it]
 87%|████████▋ | 16517/19090 [8:03:52<1:07:31,  1.57s/it]
 87%|████████▋ | 16518/19090 [8:03:54<1:18:21,  1.83s/it]
 87%|████████▋ | 16519/19090 [8:03:55<1:11:21,  1.67s/it][gradnorm] step 16519: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 635.2

 87%|████████▋ | 16520/19090 [8:03:57<1:06:28,  1.55s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 909.259033203125, 'learning_rate': 2.7120876723717907e-06, 'recon_loss': 0.0017238816013559698, 'recon_static': 0.0007520016471971758, 'recon_motion': 0.0009718799687107093, 'recon_norm': 1.3591742545366288, 'recon_norm_std': 0.03865930987522006, 'target_norm': 1.61994209587574, 'recon_cos': 0.5998409777879715, 'gradnorm/compressor.query': 0.5838909265775217, 'gradnorm/compressor.layer0': 20.041731212403644, 'gradnorm/compressor.layer1': 900.1195406722376, 'gradnorm/compressor.layer2': 121.52434732261474, 'gradnorm/compressor.layer3': 35.14879955775962, 'gradnorm/compressor.layer4': 8.69305812953961, 'gradnorm/compressor.layer5': 4.757066790391132, 'gradnorm/compressor.layer6': 3.2985768919837386, 'gradnorm/compressor.layer7': 3.083666465564914, 'gradnorm/decoder.layer0': 2.7425973265355634, 'gradnorm/decoder.layer1': 1.9389550682634151, 'gradnorm/decoder.layer2': 1.1238812075725535, 'gradnorm/max_param': 635.1689453125, 'epoch': 8.65}

 87%|████████▋ | 16520/19090 [8:03:57<1:06:28,  1.55s/it]
 87%|████████▋ | 16521/19090 [8:03:59<1:16:42,  1.79s/it]
 87%|████████▋ | 16522/19090 [8:04:00<1:10:15,  1.64s/it]
 87%|████████▋ | 16523/19090 [8:04:02<1:05:39,  1.53s/it]
 87%|████████▋ | 16524/19090 [8:04:05<1:24:52,  1.98s/it]
 87%|████████▋ | 16525/19090 [8:04:06<1:15:55,  1.78s/it]
 87%|████████▋ | 16526/19090 [8:04:07<1:09:41,  1.63s/it]
 87%|████████▋ | 16527/19090 [8:04:10<1:18:40,  1.84s/it]
 87%|████████▋ | 16528/19090 [8:04:11<1:11:34,  1.68s/it]
 87%|████████▋ | 16529/19090 [8:04:12<1:06:38,  1.56s/it][gradnorm] step 16529: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 651.9

 87%|████████▋ | 16530/19090 [8:04:15<1:19:40,  1.87s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 933.9052734375, 'learning_rate': 2.6914174105819256e-06, 'recon_loss': 0.0016365701943868772, 'recon_static': 0.0006625755515415222, 'recon_motion': 0.0009739946195622906, 'recon_norm': 1.325897428393364, 'recon_norm_std': 0.038840333092957736, 'target_norm': 1.6192783772945405, 'recon_cos': 0.6171368151903153, 'gradnorm/compressor.query': 0.36169046328451165, 'gradnorm/compressor.layer0': 11.716573365564868, 'gradnorm/compressor.layer1': 923.7666777926123, 'gradnorm/compressor.layer2': 130.9307874174367, 'gradnorm/compressor.layer3': 37.65812041516577, 'gradnorm/compressor.layer4': 9.216098830911744, 'gradnorm/compressor.layer5': 3.991811357608709, 'gradnorm/compressor.layer6': 3.107538188721154, 'gradnorm/compressor.layer7': 3.444150006278337, 'gradnorm/decoder.layer0': 2.857817679140332, 'gradnorm/decoder.layer1': 1.9721326076343666, 'gradnorm/decoder.layer2': 1.1264524091241794, 'gradnorm/max_param': 651.881591796875, 'epoch': 8.66}

 87%|████████▋ | 16530/19090 [8:04:15<1:19:40,  1.87s/it]
 87%|████████▋ | 16531/19090 [8:04:17<1:18:36,  1.84s/it]
 87%|████████▋ | 16532/19090 [8:04:18<1:11:31,  1.68s/it]
 87%|████████▋ | 16533/19090 [8:04:20<1:20:09,  1.88s/it]
 87%|████████▋ | 16534/19090 [8:04:22<1:21:30,  1.91s/it]
 87%|████████▋ | 16535/19090 [8:04:23<1:13:30,  1.73s/it]
 87%|████████▋ | 16536/19090 [8:04:26<1:17:57,  1.83s/it]
 87%|████████▋ | 16537/19090 [8:04:27<1:18:45,  1.85s/it]
 87%|████████▋ | 16538/19090 [8:04:29<1:11:37,  1.68s/it]
 87%|████████▋ | 16539/19090 [8:04:31<1:16:16,  1.79s/it][gradnorm] step 16539: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 642.5

 87%|████████▋ | 16540/19090 [8:04:33<1:18:57,  1.86s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 920.0433959960938, 'learning_rate': 2.670821737782042e-06, 'recon_loss': 0.0015747275901958346, 'recon_static': 0.0006056376820197329, 'recon_motion': 0.0009690899052657187, 'recon_norm': 1.3020530611276626, 'recon_norm_std': 0.040946285054087636, 'target_norm': 1.6183343172073363, 'recon_cos': 0.629556542634964, 'gradnorm/compressor.query': 0.23893302359542512, 'gradnorm/compressor.layer0': 8.093642068386293, 'gradnorm/compressor.layer1': 910.4219145539062, 'gradnorm/compressor.layer2': 125.60713103636871, 'gradnorm/compressor.layer3': 40.65929441906549, 'gradnorm/compressor.layer4': 8.173828358348265, 'gradnorm/compressor.layer5': 4.228647334842515, 'gradnorm/compressor.layer6': 2.891339682817982, 'gradnorm/compressor.layer7': 3.2211095068909565, 'gradnorm/decoder.layer0': 2.9061459758558925, 'gradnorm/decoder.layer1': 1.7438959204213442, 'gradnorm/decoder.layer2': 0.9875780702465239, 'gradnorm/max_param': 642.5382080078125, 'epoch': 8.66}

 87%|████████▋ | 16540/19090 [8:04:33<1:18:57,  1.86s/it]
 87%|████████▋ | 16541/19090 [8:04:35<1:23:09,  1.96s/it]
 87%|████████▋ | 16542/19090 [8:04:36<1:14:40,  1.76s/it]
 87%|████████▋ | 16543/19090 [8:04:38<1:17:24,  1.82s/it]
 87%|████████▋ | 16544/19090 [8:04:40<1:10:32,  1.66s/it]
 87%|████████▋ | 16545/19090 [8:04:42<1:18:00,  1.84s/it]
 87%|████████▋ | 16546/19090 [8:04:44<1:25:04,  2.01s/it]
 87%|████████▋ | 16547/19090 [8:04:45<1:15:51,  1.79s/it]
 87%|████████▋ | 16548/19090 [8:04:47<1:17:01,  1.82s/it]
 87%|████████▋ | 16549/19090 [8:04:50<1:25:08,  2.01s/it][gradnorm] step 16549: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 937.3

 87%|████████▋ | 16550/19090 [8:04:51<1:15:57,  1.79s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1339.1392822265625, 'learning_rate': 2.6503007228339816e-06, 'recon_loss': 0.0015329982328694315, 'recon_static': 0.0005463335182867013, 'recon_motion': 0.0009866647436865606, 'recon_norm': 1.2760330259799957, 'recon_norm_std': 0.040008056163787845, 'target_norm': 1.6226763725280762, 'recon_cos': 0.6405629619956017, 'gradnorm/compressor.query': 0.5725694182008895, 'gradnorm/compressor.layer0': 15.76653669006668, 'gradnorm/compressor.layer1': 1328.5382474735156, 'gradnorm/compressor.layer2': 162.65869054365032, 'gradnorm/compressor.layer3': 37.833147449008756, 'gradnorm/compressor.layer4': 9.554275865647513, 'gradnorm/compressor.layer5': 4.565356679730422, 'gradnorm/compressor.layer6': 2.678929346785458, 'gradnorm/compressor.layer7': 3.079021342714929, 'gradnorm/decoder.layer0': 2.963065396851799, 'gradnorm/decoder.layer1': 1.7861188634595762, 'gradnorm/decoder.layer2': 0.9926771022264329, 'gradnorm/max_param': 937.3346557617188, 'epoch': 8.67}

 87%|████████▋ | 16550/19090 [8:04:51<1:15:57,  1.79s/it]
 87%|████████▋ | 16551/19090 [8:04:53<1:16:24,  1.81s/it]
 87%|████████▋ | 16552/19090 [8:04:55<1:20:44,  1.91s/it]
 87%|████████▋ | 16553/19090 [8:04:56<1:12:48,  1.72s/it]
 87%|████████▋ | 16554/19090 [8:04:58<1:14:47,  1.77s/it]
 87%|████████▋ | 16555/19090 [8:05:00<1:19:40,  1.89s/it]
 87%|████████▋ | 16556/19090 [8:05:02<1:12:03,  1.71s/it]
 87%|████████▋ | 16557/19090 [8:05:04<1:17:58,  1.85s/it]
 87%|████████▋ | 16558/19090 [8:05:06<1:17:44,  1.84s/it]
 87%|████████▋ | 16559/19090 [8:05:07<1:10:46,  1.68s/it][gradnorm] step 16559: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 705.8

 87%|████████▋ | 16560/19090 [8:05:09<1:19:06,  1.88s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1010.5098266601562, 'learning_rate': 2.629854434349982e-06, 'recon_loss': 0.0014877581998007372, 'recon_static': 0.0005046772363130003, 'recon_motion': 0.0009830809329287149, 'recon_norm': 1.2594807267189025, 'recon_norm_std': 0.04065016070380807, 'target_norm': 1.6213653713464737, 'recon_cos': 0.650369007885456, 'gradnorm/compressor.query': 0.38901922146101947, 'gradnorm/compressor.layer0': 10.322975340726815, 'gradnorm/compressor.layer1': 1000.1208676955001, 'gradnorm/compressor.layer2': 139.41643862678103, 'gradnorm/compressor.layer3': 34.655839647593126, 'gradnorm/compressor.layer4': 9.323447532386298, 'gradnorm/compressor.layer5': 4.93188853027901, 'gradnorm/compressor.layer6': 2.657565610116976, 'gradnorm/compressor.layer7': 3.11698067727095, 'gradnorm/decoder.layer0': 3.27881655603324, 'gradnorm/decoder.layer1': 1.9999448649436142, 'gradnorm/decoder.layer2': 0.9930497275287424, 'gradnorm/max_param': 705.8445434570312, 'epoch': 8.67}

 87%|████████▋ | 16560/19090 [8:05:09<1:19:06,  1.88s/it]
 87%|████████▋ | 16561/19090 [8:05:11<1:11:48,  1.70s/it]
 87%|████████▋ | 16562/19090 [8:05:12<1:06:36,  1.58s/it]
 87%|████████▋ | 16563/19090 [8:05:14<1:18:26,  1.86s/it]
 87%|████████▋ | 16564/19090 [8:05:16<1:14:56,  1.78s/it]
 87%|████████▋ | 16565/19090 [8:05:17<1:08:48,  1.64s/it]
 87%|████████▋ | 16566/19090 [8:05:19<1:15:09,  1.79s/it]
 87%|████████▋ | 16567/19090 [8:05:22<1:19:14,  1.88s/it]
 87%|████████▋ | 16568/19090 [8:05:23<1:11:41,  1.71s/it]
 87%|████████▋ | 16569/19090 [8:05:25<1:14:40,  1.78s/it][gradnorm] step 16569: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 902.6

 87%|████████▋ | 16570/19090 [8:05:27<1:21:33,  1.94s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1290.418212890625, 'learning_rate': 2.6094829406924483e-06, 'recon_loss': 0.0014705425535794347, 'recon_static': 0.0004837282685912214, 'recon_motion': 0.0009868142966297454, 'recon_norm': 1.2512785136699676, 'recon_norm_std': 0.042111022397875786, 'target_norm': 1.6217438012361527, 'recon_cos': 0.654824098944664, 'gradnorm/compressor.query': 0.5909702079351125, 'gradnorm/compressor.layer0': 18.350931402781967, 'gradnorm/compressor.layer1': 1279.308553477229, 'gradnorm/compressor.layer2': 162.39285530196517, 'gradnorm/compressor.layer3': 41.50149129003529, 'gradnorm/compressor.layer4': 8.000702350303037, 'gradnorm/compressor.layer5': 4.79667678075541, 'gradnorm/compressor.layer6': 2.5709585751816952, 'gradnorm/compressor.layer7': 2.996318147255928, 'gradnorm/decoder.layer0': 3.1941135087646084, 'gradnorm/decoder.layer1': 2.0099943307065957, 'gradnorm/decoder.layer2': 1.0493394113382395, 'gradnorm/max_param': 902.6253051757812, 'epoch': 8.68}

 87%|████████▋ | 16570/19090 [8:05:27<1:21:33,  1.94s/it]
 87%|████████▋ | 16571/19090 [8:05:28<1:13:22,  1.75s/it]
 87%|████████▋ | 16572/19090 [8:05:30<1:13:39,  1.76s/it]
 87%|████████▋ | 16573/19090 [8:05:33<1:20:52,  1.93s/it]
 87%|████████▋ | 16574/19090 [8:05:34<1:12:47,  1.74s/it]
 87%|████████▋ | 16575/19090 [8:05:36<1:12:54,  1.74s/it]
 87%|████████▋ | 16576/19090 [8:05:38<1:20:16,  1.92s/it]
 87%|████████▋ | 16577/19090 [8:05:39<1:12:23,  1.73s/it]
 87%|████████▋ | 16578/19090 [8:05:41<1:15:01,  1.79s/it]
 87%|████████▋ | 16579/19090 [8:05:43<1:14:21,  1.78s/it][gradnorm] step 16579: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 803.7

 87%|████████▋ | 16580/19090 [8:05:44<1:08:11,  1.63s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1148.7064208984375, 'learning_rate': 2.589186309973679e-06, 'recon_loss': 0.0014677002443931996, 'recon_static': 0.0004837141023017466, 'recon_motion': 0.00098398615809856, 'recon_norm': 1.2462544649839402, 'recon_norm_std': 0.041276916209608314, 'target_norm': 1.6209102421998978, 'recon_cos': 0.654652689397335, 'gradnorm/compressor.query': 0.6393181566940425, 'gradnorm/compressor.layer0': 25.182806974002, 'gradnorm/compressor.layer1': 1139.3197531860842, 'gradnorm/compressor.layer2': 138.37534581031767, 'gradnorm/compressor.layer3': 39.61226875524173, 'gradnorm/compressor.layer4': 8.370841744777719, 'gradnorm/compressor.layer5': 4.723138650468336, 'gradnorm/compressor.layer6': 2.7729204582419626, 'gradnorm/compressor.layer7': 3.164546675194547, 'gradnorm/decoder.layer0': 3.2134089924253493, 'gradnorm/decoder.layer1': 1.963711603907489, 'gradnorm/decoder.layer2': 1.1358539660775184, 'gradnorm/max_param': 803.74365234375, 'epoch': 8.69}

 87%|████████▋ | 16580/19090 [8:05:44<1:08:11,  1.63s/it]
 87%|████████▋ | 16581/19090 [8:05:46<1:13:14,  1.75s/it]
 87%|████████▋ | 16582/19090 [8:05:48<1:18:37,  1.88s/it]
 87%|████████▋ | 16583/19090 [8:05:50<1:11:12,  1.70s/it]
 87%|████████▋ | 16584/19090 [8:05:52<1:15:43,  1.81s/it]
 87%|████████▋ | 16585/19090 [8:05:54<1:17:19,  1.85s/it]
 87%|████████▋ | 16586/19090 [8:05:55<1:10:24,  1.69s/it]
 87%|████████▋ | 16587/19090 [8:05:57<1:13:59,  1.77s/it]
 87%|████████▋ | 16588/19090 [8:05:59<1:17:06,  1.85s/it]
 87%|████████▋ | 16589/19090 [8:06:00<1:10:06,  1.68s/it][gradnorm] step 16589: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 867.2

 87%|████████▋ | 16590/19090 [8:06:02<1:15:27,  1.81s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1240.8533935546875, 'learning_rate': 2.5689646100556798e-06, 'recon_loss': 0.00147334466164466, 'recon_static': 0.0004894022953521926, 'recon_motion': 0.000983942361199297, 'recon_norm': 1.2537572771310805, 'recon_norm_std': 0.041961163748055696, 'target_norm': 1.6210620015859605, 'recon_cos': 0.6539573088288307, 'gradnorm/compressor.query': 0.49545249645240563, 'gradnorm/compressor.layer0': 18.230567933493855, 'gradnorm/compressor.layer1': 1229.045462950822, 'gradnorm/compressor.layer2': 164.45097459995486, 'gradnorm/compressor.layer3': 40.56090346193533, 'gradnorm/compressor.layer4': 9.220650376821693, 'gradnorm/compressor.layer5': 4.792777731523658, 'gradnorm/compressor.layer6': 3.1256170045185554, 'gradnorm/compressor.layer7': 3.029567448173859, 'gradnorm/decoder.layer0': 3.158049721846427, 'gradnorm/decoder.layer1': 2.030003091593676, 'gradnorm/decoder.layer2': 1.1902763637226201, 'gradnorm/max_param': 867.1807250976562, 'epoch': 8.69}

 87%|████████▋ | 16590/19090 [8:06:02<1:15:27,  1.81s/it]
 87%|████████▋ | 16591/19090 [8:06:05<1:19:34,  1.91s/it]
 87%|████████▋ | 16592/19090 [8:06:06<1:11:49,  1.73s/it]
 87%|████████▋ | 16593/19090 [8:06:08<1:12:59,  1.75s/it]
 87%|████████▋ | 16594/19090 [8:06:10<1:18:50,  1.90s/it]
 87%|████████▋ | 16595/19090 [8:06:11<1:11:17,  1.71s/it]
 87%|████████▋ | 16596/19090 [8:06:13<1:16:36,  1.84s/it]
 87%|████████▋ | 16597/19090 [8:06:15<1:17:47,  1.87s/it]
 87%|████████▋ | 16598/19090 [8:06:17<1:10:30,  1.70s/it]
 87%|████████▋ | 16599/19090 [8:06:18<1:14:02,  1.78s/it][gradnorm] step 16599: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 647.6

 87%|████████▋ | 16600/19090 [8:06:21<1:19:22,  1.91s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 928.4976806640625, 'learning_rate': 2.5488179085499414e-06, 'recon_loss': 0.0014548477542120964, 'recon_static': 0.00047774441190995276, 'recon_motion': 0.0009771033190190792, 'recon_norm': 1.240738210082054, 'recon_norm_std': 0.04195213401690125, 'target_norm': 1.6201210230588914, 'recon_cos': 0.6573568820953369, 'gradnorm/compressor.query': 0.2605902554201888, 'gradnorm/compressor.layer0': 7.503571486464978, 'gradnorm/compressor.layer1': 917.693474151364, 'gradnorm/compressor.layer2': 134.91878344026082, 'gradnorm/compressor.layer3': 39.46469044591871, 'gradnorm/compressor.layer4': 8.632311389584936, 'gradnorm/compressor.layer5': 4.963235157819094, 'gradnorm/compressor.layer6': 2.7149885619258027, 'gradnorm/compressor.layer7': 2.8715961042765383, 'gradnorm/decoder.layer0': 3.056529082679267, 'gradnorm/decoder.layer1': 2.0580606946679225, 'gradnorm/decoder.layer2': 1.1478042413636176, 'gradnorm/max_param': 647.60302734375, 'epoch': 8.7}

 87%|████████▋ | 16600/19090 [8:06:21<1:19:22,  1.91s/it]
 87%|████████▋ | 16601/19090 [8:06:22<1:11:54,  1.73s/it]
 87%|████████▋ | 16602/19090 [8:06:24<1:14:44,  1.80s/it]
 87%|████████▋ | 16603/19090 [8:06:26<1:16:55,  1.86s/it]
 87%|████████▋ | 16604/19090 [8:06:27<1:09:52,  1.69s/it]
 87%|████████▋ | 16605/19090 [8:06:29<1:11:38,  1.73s/it]
 87%|████████▋ | 16606/19090 [8:06:31<1:16:09,  1.84s/it]
 87%|████████▋ | 16607/19090 [8:06:32<1:09:35,  1.68s/it]
 87%|████████▋ | 16608/19090 [8:06:34<1:05:35,  1.59s/it]
 87%|████████▋ | 16609/19090 [8:06:36<1:16:50,  1.86s/it][gradnorm] step 16609: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 723.2

 87%|████████▋ | 16610/19090 [8:06:38<1:09:44,  1.69s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1036.153076171875, 'learning_rate': 2.5287462728171617e-06, 'recon_loss': 0.0014654145430540665, 'recon_static': 0.0004881738270341884, 'recon_motion': 0.0009772407516720704, 'recon_norm': 1.2535544484853745, 'recon_norm_std': 0.04202403016388416, 'target_norm': 1.6215654820203782, 'recon_cos': 0.6566023990511894, 'gradnorm/compressor.query': 0.36993190183816854, 'gradnorm/compressor.layer0': 8.730352959942444, 'gradnorm/compressor.layer1': 1024.7825379074334, 'gradnorm/compressor.layer2': 147.24175618901384, 'gradnorm/compressor.layer3': 39.28399857378717, 'gradnorm/compressor.layer4': 8.615594573134974, 'gradnorm/compressor.layer5': 4.823072833285413, 'gradnorm/compressor.layer6': 2.999237281481724, 'gradnorm/compressor.layer7': 3.2713155856643885, 'gradnorm/decoder.layer0': 3.3652492529506435, 'gradnorm/decoder.layer1': 2.080215320629546, 'gradnorm/decoder.layer2': 1.1309772225809087, 'gradnorm/max_param': 723.2431640625, 'epoch': 8.7}

 87%|████████▋ | 16610/19090 [8:06:38<1:09:44,  1.69s/it]
 87%|████████▋ | 16611/19090 [8:06:39<1:05:18,  1.58s/it]
 87%|████████▋ | 16612/19090 [8:06:41<1:16:03,  1.84s/it]
 87%|████████▋ | 16613/19090 [8:06:43<1:09:10,  1.68s/it]
 87%|████████▋ | 16614/19090 [8:06:44<1:09:05,  1.67s/it]
 87%|████████▋ | 16615/19090 [8:06:47<1:17:41,  1.88s/it]
 87%|████████▋ | 16616/19090 [8:06:48<1:10:18,  1.70s/it]
 87%|████████▋ | 16617/19090 [8:06:50<1:09:49,  1.69s/it]
 87%|████████▋ | 16618/19090 [8:06:52<1:21:25,  1.98s/it]
 87%|████████▋ | 16619/19090 [8:06:54<1:12:55,  1.77s/it][gradnorm] step 16619: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 559.4

 87%|████████▋ | 16620/19090 [8:06:55<1:10:11,  1.70s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 804.5491333007812, 'learning_rate': 2.508749769967081e-06, 'recon_loss': 0.0014705825131386518, 'recon_static': 0.0004900245199678466, 'recon_motion': 0.000980557948059868, 'recon_norm': 1.2532145410776139, 'recon_norm_std': 0.04170141899958253, 'target_norm': 1.6203948110342026, 'recon_cos': 0.654333807528019, 'gradnorm/compressor.query': 0.44172004078803634, 'gradnorm/compressor.layer0': 14.716703430567637, 'gradnorm/compressor.layer1': 792.7812592386376, 'gradnorm/compressor.layer2': 130.89376507066714, 'gradnorm/compressor.layer3': 36.43414301645848, 'gradnorm/compressor.layer4': 7.962228297893338, 'gradnorm/compressor.layer5': 4.857169776329008, 'gradnorm/compressor.layer6': 2.904534766915218, 'gradnorm/compressor.layer7': 3.0556690561041013, 'gradnorm/decoder.layer0': 3.247106878318549, 'gradnorm/decoder.layer1': 1.8672042909769015, 'gradnorm/decoder.layer2': 1.007657120429663, 'gradnorm/max_param': 559.4197387695312, 'epoch': 8.71}

 87%|████████▋ | 16620/19090 [8:06:55<1:10:11,  1.70s/it]
 87%|████████▋ | 16621/19090 [8:06:58<1:18:26,  1.91s/it]
 87%|████████▋ | 16622/19090 [8:06:59<1:10:50,  1.72s/it]
 87%|████████▋ | 16623/19090 [8:07:01<1:15:42,  1.84s/it]
 87%|████████▋ | 16624/19090 [8:07:03<1:13:39,  1.79s/it]
 87%|████████▋ | 16625/19090 [8:07:04<1:07:28,  1.64s/it]
 87%|████████▋ | 16626/19090 [8:07:06<1:15:57,  1.85s/it]
 87%|████████▋ | 16627/19090 [8:07:08<1:15:12,  1.83s/it]
 87%|████████▋ | 16628/19090 [8:07:09<1:08:33,  1.67s/it]
 87%|████████▋ | 16629/19090 [8:07:12<1:15:57,  1.85s/it][gradnorm] step 16629: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 712.3

 87%|████████▋ | 16630/19090 [8:07:14<1:16:22,  1.86s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1020.1026000976562, 'learning_rate': 2.4888284668582285e-06, 'recon_loss': 0.0015300682396627962, 'recon_static': 0.0005431773097370751, 'recon_motion': 0.0009868909110082313, 'recon_norm': 1.272557109594345, 'recon_norm_std': 0.04078750954940915, 'target_norm': 1.6236323326826096, 'recon_cos': 0.6410851493477822, 'gradnorm/compressor.query': 0.3246679641052624, 'gradnorm/compressor.layer0': 10.656988481903223, 'gradnorm/compressor.layer1': 1009.3793575757333, 'gradnorm/compressor.layer2': 142.1940245755777, 'gradnorm/compressor.layer3': 36.255246118234396, 'gradnorm/compressor.layer4': 7.739897265085129, 'gradnorm/compressor.layer5': 4.960917122866615, 'gradnorm/compressor.layer6': 2.58842362583103, 'gradnorm/compressor.layer7': 2.974883960416254, 'gradnorm/decoder.layer0': 3.174234048560998, 'gradnorm/decoder.layer1': 1.9641279429698515, 'gradnorm/decoder.layer2': 1.1155961981295872, 'gradnorm/max_param': 712.2921752929688, 'epoch': 8.71}

 87%|████████▋ | 16630/19090 [8:07:14<1:16:22,  1.86s/it]
 87%|████████▋ | 16631/19090 [8:07:15<1:09:22,  1.69s/it]
 87%|████████▋ | 16632/19090 [8:07:17<1:11:18,  1.74s/it]
 87%|████████▋ | 16633/19090 [8:07:19<1:18:36,  1.92s/it]
 87%|████████▋ | 16634/19090 [8:07:20<1:10:56,  1.73s/it]
 87%|████████▋ | 16635/19090 [8:07:22<1:10:27,  1.72s/it]
 87%|████████▋ | 16636/19090 [8:07:25<1:20:44,  1.97s/it]
 87%|████████▋ | 16637/19090 [8:07:26<1:12:28,  1.77s/it]
 87%|████████▋ | 16638/19090 [8:07:27<1:09:05,  1.69s/it]
 87%|████████▋ | 16639/19090 [8:07:30<1:19:03,  1.94s/it][gradnorm] step 16639: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 843.7

 87%|████████▋ | 16640/19090 [8:07:31<1:11:11,  1.74s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1209.4261474609375, 'learning_rate': 2.468982430097705e-06, 'recon_loss': 0.0015651681052986533, 'recon_static': 0.0005962966766674072, 'recon_motion': 0.0009688714315416291, 'recon_norm': 1.29786759018898, 'recon_norm_std': 0.040447833202779296, 'target_norm': 1.6186325997114182, 'recon_cos': 0.631817689538002, 'gradnorm/compressor.query': 0.5511375789664685, 'gradnorm/compressor.layer0': 16.529233986240108, 'gradnorm/compressor.layer1': 1195.7624659605267, 'gradnorm/compressor.layer2': 173.5662022471541, 'gradnorm/compressor.layer3': 48.41883965155635, 'gradnorm/compressor.layer4': 8.380184561579346, 'gradnorm/compressor.layer5': 4.636607887851686, 'gradnorm/compressor.layer6': 2.6068156591341065, 'gradnorm/compressor.layer7': 2.997888934917549, 'gradnorm/decoder.layer0': 2.999518196835447, 'gradnorm/decoder.layer1': 1.8020730505265885, 'gradnorm/decoder.layer2': 0.9747560195834819, 'gradnorm/max_param': 843.7371215820312, 'epoch': 8.72}

 87%|████████▋ | 16640/19090 [8:07:31<1:11:11,  1.74s/it]
 87%|████████▋ | 16641/19090 [8:07:33<1:09:21,  1.70s/it]
 87%|████████▋ | 16642/19090 [8:07:35<1:20:36,  1.98s/it]
 87%|████████▋ | 16643/19090 [8:07:37<1:12:15,  1.77s/it]
 87%|████████▋ | 16644/19090 [8:07:38<1:12:25,  1.78s/it]
 87%|████████▋ | 16645/19090 [8:07:41<1:16:27,  1.88s/it]
 87%|████████▋ | 16646/19090 [8:07:42<1:09:20,  1.70s/it]
 87%|████████▋ | 16647/19090 [8:07:44<1:14:47,  1.84s/it]
 87%|████████▋ | 16648/19090 [8:07:46<1:14:17,  1.83s/it]
 87%|████████▋ | 16649/19090 [8:07:47<1:07:53,  1.67s/it][gradnorm] step 16649: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1007.2

 87%|████████▋ | 16650/19090 [8:07:49<1:14:46,  1.84s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1439.2296142578125, 'learning_rate': 2.4492117260409343e-06, 'recon_loss': 0.0015654805640224368, 'recon_static': 0.000593879418738652, 'recon_motion': 0.000971601162746083, 'recon_norm': 1.2949339985847472, 'recon_norm_std': 0.03984585991129279, 'target_norm': 1.620004940032959, 'recon_cos': 0.6323188498616219, 'gradnorm/compressor.query': 0.47793478454440996, 'gradnorm/compressor.layer0': 15.045004406538851, 'gradnorm/compressor.layer1': 1427.1218851240421, 'gradnorm/compressor.layer2': 179.56927088696162, 'gradnorm/compressor.layer3': 45.49862903583607, 'gradnorm/compressor.layer4': 9.81504021098547, 'gradnorm/compressor.layer5': 5.441525602435222, 'gradnorm/compressor.layer6': 2.7956005581308174, 'gradnorm/compressor.layer7': 3.3821786899023665, 'gradnorm/decoder.layer0': 3.5021106622837244, 'gradnorm/decoder.layer1': 2.0402883301593246, 'gradnorm/decoder.layer2': 1.1638359002883472, 'gradnorm/max_param': 1007.1748046875, 'epoch': 8.72}

 87%|████████▋ | 16650/19090 [8:07:49<1:14:46,  1.84s/it]
 87%|████████▋ | 16651/19090 [8:07:51<1:16:15,  1.88s/it]
 87%|████████▋ | 16652/19090 [8:07:53<1:10:05,  1.72s/it]
 87%|████████▋ | 16653/19090 [8:07:54<1:07:33,  1.66s/it]
 87%|████████▋ | 16654/19090 [8:07:57<1:15:26,  1.86s/it]
 87%|████████▋ | 16655/19090 [8:07:58<1:13:12,  1.80s/it]
 87%|████████▋ | 16656/19090 [8:08:00<1:08:17,  1.68s/it]
 87%|████████▋ | 16657/19090 [8:08:02<1:15:49,  1.87s/it]
 87%|████████▋ | 16658/19090 [8:08:04<1:18:08,  1.93s/it]
 87%|████████▋ | 16659/19090 [8:08:05<1:10:28,  1.74s/it][gradnorm] step 16659: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 804.6

 87%|████████▋ | 16660/19090 [8:08:07<1:11:59,  1.78s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1153.212890625, 'learning_rate': 2.4295164207914943e-06, 'recon_loss': 0.0015288341004634277, 'recon_static': 0.0005546779066207819, 'recon_motion': 0.000974156214215327, 'recon_norm': 1.290176346898079, 'recon_norm_std': 0.04005058705806732, 'target_norm': 1.6185630142688752, 'recon_cos': 0.6409870117902756, 'gradnorm/compressor.query': 0.4333283583037573, 'gradnorm/compressor.layer0': 14.638734286266223, 'gradnorm/compressor.layer1': 1140.0132126427307, 'gradnorm/compressor.layer2': 167.6301092155732, 'gradnorm/compressor.layer3': 42.51595518523017, 'gradnorm/compressor.layer4': 9.405725204681273, 'gradnorm/compressor.layer5': 4.941024874116768, 'gradnorm/compressor.layer6': 2.884376329150984, 'gradnorm/compressor.layer7': 3.1250109863088134, 'gradnorm/decoder.layer0': 3.415188182600016, 'gradnorm/decoder.layer1': 1.9847504868781274, 'gradnorm/decoder.layer2': 1.051805755457606, 'gradnorm/max_param': 804.5780639648438, 'epoch': 8.73}

 87%|████████▋ | 16660/19090 [8:08:07<1:11:59,  1.78s/it]
 87%|████████▋ | 16661/19090 [8:08:09<1:17:42,  1.92s/it]
 87%|████████▋ | 16662/19090 [8:08:11<1:10:09,  1.73s/it]
 87%|████████▋ | 16663/19090 [8:08:12<1:09:47,  1.73s/it]
 87%|████████▋ | 16664/19090 [8:08:15<1:19:09,  1.96s/it]
 87%|████████▋ | 16665/19090 [8:08:16<1:11:08,  1.76s/it]
 87%|████████▋ | 16666/19090 [8:08:18<1:05:33,  1.62s/it]
 87%|████████▋ | 16667/19090 [8:08:20<1:20:18,  1.99s/it]
 87%|████████▋ | 16668/19090 [8:08:22<1:11:52,  1.78s/it]
 87%|████████▋ | 16669/19090 [8:08:23<1:06:01,  1.64s/it][gradnorm] step 16669: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 852.4

 87%|████████▋ | 16670/19090 [8:08:25<1:16:07,  1.89s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1219.7835693359375, 'learning_rate': 2.4098965802008572e-06, 'recon_loss': 0.0015054314833832904, 'recon_static': 0.0005244750966085121, 'recon_motion': 0.0009809563736780547, 'recon_norm': 1.264194393157959, 'recon_norm_std': 0.04108753837645054, 'target_norm': 1.6222089499235153, 'recon_cos': 0.6462800592184067, 'gradnorm/compressor.query': 0.550934844548863, 'gradnorm/compressor.layer0': 15.919962490171784, 'gradnorm/compressor.layer1': 1207.854709805778, 'gradnorm/compressor.layer2': 160.66805113727494, 'gradnorm/compressor.layer3': 52.25001168136831, 'gradnorm/compressor.layer4': 9.50568821919317, 'gradnorm/compressor.layer5': 5.11161097325915, 'gradnorm/compressor.layer6': 3.4051125403360043, 'gradnorm/compressor.layer7': 3.6262762519651486, 'gradnorm/decoder.layer0': 3.677330088058194, 'gradnorm/decoder.layer1': 2.114519077571267, 'gradnorm/decoder.layer2': 1.248776981951972, 'gradnorm/max_param': 852.3851928710938, 'epoch': 8.73}

 87%|████████▋ | 16670/19090 [8:08:25<1:16:07,  1.89s/it]
 87%|████████▋ | 16671/19090 [8:08:27<1:09:41,  1.73s/it]
 87%|████████▋ | 16672/19090 [8:08:28<1:04:55,  1.61s/it]
 87%|████████▋ | 16673/19090 [8:08:31<1:22:38,  2.05s/it]
 87%|████████▋ | 16674/19090 [8:08:32<1:13:26,  1.82s/it]
 87%|████████▋ | 16675/19090 [8:08:34<1:06:56,  1.66s/it]
 87%|████████▋ | 16676/19090 [8:08:37<1:25:05,  2.11s/it]
 87%|████████▋ | 16677/19090 [8:08:38<1:15:03,  1.87s/it]
 87%|████████▋ | 16678/19090 [8:08:40<1:07:59,  1.69s/it]
 87%|████████▋ | 16679/19090 [8:08:42<1:21:24,  2.03s/it][gradnorm] step 16679: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 754.9

 87%|████████▋ | 16680/19090 [8:08:44<1:12:44,  1.81s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1082.1231689453125, 'learning_rate': 2.3903522698681567e-06, 'recon_loss': 0.001521895866608247, 'recon_static': 0.000544806435937062, 'recon_motion': 0.0009770894481334834, 'recon_norm': 1.2684432744979859, 'recon_norm_std': 0.041127741895616056, 'target_norm': 1.6213643312454225, 'recon_cos': 0.6418314024806022, 'gradnorm/compressor.query': 0.5989087761542798, 'gradnorm/compressor.layer0': 16.463213979985234, 'gradnorm/compressor.layer1': 1069.7804097103292, 'gradnorm/compressor.layer2': 156.0393204652276, 'gradnorm/compressor.layer3': 42.37562078960835, 'gradnorm/compressor.layer4': 9.384124791140787, 'gradnorm/compressor.layer5': 5.172346295321928, 'gradnorm/compressor.layer6': 2.848603686215512, 'gradnorm/compressor.layer7': 2.9945640588043387, 'gradnorm/decoder.layer0': 2.9552381917157526, 'gradnorm/decoder.layer1': 1.8508848728914407, 'gradnorm/decoder.layer2': 1.023596074289901, 'gradnorm/max_param': 754.9118041992188, 'epoch': 8.74}

 87%|████████▋ | 16680/19090 [8:08:44<1:12:44,  1.81s/it]
 87%|████████▋ | 16681/19090 [8:08:45<1:06:36,  1.66s/it]
 87%|████████▋ | 16682/19090 [8:08:47<1:16:19,  1.90s/it]
 87%|████████▋ | 16683/19090 [8:08:49<1:09:07,  1.72s/it]
 87%|████████▋ | 16684/19090 [8:08:50<1:04:02,  1.60s/it]
 87%|████████▋ | 16685/19090 [8:08:52<1:12:09,  1.80s/it]
 87%|████████▋ | 16686/19090 [8:08:54<1:07:29,  1.68s/it]
 87%|████████▋ | 16687/19090 [8:08:55<1:05:48,  1.64s/it]
 87%|████████▋ | 16688/19090 [8:08:57<1:11:56,  1.80s/it]
 87%|████████▋ | 16689/19090 [8:08:59<1:11:30,  1.79s/it][gradnorm] step 16689: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 851.7

 87%|████████▋ | 16690/19090 [8:09:01<1:09:53,  1.75s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1219.5972900390625, 'learning_rate': 2.370883555140027e-06, 'recon_loss': 0.001529116736492142, 'recon_static': 0.0005590915170614608, 'recon_motion': 0.000970025199058, 'recon_norm': 1.2880147874355317, 'recon_norm_std': 0.04028448006138206, 'target_norm': 1.620642325282097, 'recon_cos': 0.6419867232441903, 'gradnorm/compressor.query': 0.432021443426436, 'gradnorm/compressor.layer0': 12.875047405859327, 'gradnorm/compressor.layer1': 1206.6069679062855, 'gradnorm/compressor.layer2': 170.32181167205215, 'gradnorm/compressor.layer3': 47.055283518763865, 'gradnorm/compressor.layer4': 8.355416703028641, 'gradnorm/compressor.layer5': 5.0707307945714915, 'gradnorm/compressor.layer6': 2.714962831808308, 'gradnorm/compressor.layer7': 3.0477005842632177, 'gradnorm/decoder.layer0': 3.2500152587532427, 'gradnorm/decoder.layer1': 1.897828432189307, 'gradnorm/decoder.layer2': 1.0187330958362268, 'gradnorm/max_param': 851.6555786132812, 'epoch': 8.74}

 87%|████████▋ | 16690/19090 [8:09:01<1:09:53,  1.75s/it]
 87%|████████▋ | 16691/19090 [8:09:03<1:09:16,  1.73s/it]
 87%|████████▋ | 16692/19090 [8:09:05<1:12:27,  1.81s/it]
 87%|████████▋ | 16693/19090 [8:09:06<1:12:35,  1.82s/it]
 87%|████████▋ | 16694/19090 [8:09:08<1:08:04,  1.70s/it]
 87%|████████▋ | 16695/19090 [8:09:10<1:11:55,  1.80s/it]
 87%|████████▋ | 16696/19090 [8:09:12<1:12:24,  1.81s/it]
 87%|████████▋ | 16697/19090 [8:09:13<1:12:12,  1.81s/it]
 87%|████████▋ | 16698/19090 [8:09:15<1:08:50,  1.73s/it]
 87%|████████▋ | 16699/19090 [8:09:17<1:12:59,  1.83s/it][gradnorm] step 16699: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 655.3

 87%|████████▋ | 16700/19090 [8:09:19<1:12:29,  1.82s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 941.5047607421875, 'learning_rate': 2.3514905011103235e-06, 'recon_loss': 0.0015202477399725467, 'recon_static': 0.0005458241255837492, 'recon_motion': 0.0009744236900587566, 'recon_norm': 1.2737833917140962, 'recon_norm_std': 0.04105539731681347, 'target_norm': 1.6199223011732102, 'recon_cos': 0.6419081747531891, 'gradnorm/compressor.query': 0.32791325003691196, 'gradnorm/compressor.layer0': 10.260921707245727, 'gradnorm/compressor.layer1': 928.6067857279528, 'gradnorm/compressor.layer2': 147.35230290658848, 'gradnorm/compressor.layer3': 46.41994747344616, 'gradnorm/compressor.layer4': 9.265615119092539, 'gradnorm/compressor.layer5': 5.227421483271607, 'gradnorm/compressor.layer6': 2.6995633867243924, 'gradnorm/compressor.layer7': 3.39251037775981, 'gradnorm/decoder.layer0': 3.2695379587558113, 'gradnorm/decoder.layer1': 2.007494830817269, 'gradnorm/decoder.layer2': 1.0957155277957793, 'gradnorm/max_param': 655.3015747070312, 'epoch': 8.75}

 87%|████████▋ | 16700/19090 [8:09:19<1:12:29,  1.82s/it]
 87%|████████▋ | 16701/19090 [8:09:20<1:07:46,  1.70s/it]
 87%|████████▋ | 16702/19090 [8:09:23<1:16:22,  1.92s/it]
 87%|████████▋ | 16703/19090 [8:09:24<1:10:43,  1.78s/it]
 88%|████████▊ | 16704/19090 [8:09:26<1:09:34,  1.75s/it]
 88%|████████▊ | 16705/19090 [8:09:28<1:18:20,  1.97s/it]
 88%|████████▊ | 16706/19090 [8:09:30<1:10:10,  1.77s/it]
 88%|████████▊ | 16707/19090 [8:09:31<1:06:48,  1.68s/it]
 88%|████████▊ | 16708/19090 [8:09:34<1:19:03,  1.99s/it]
 88%|████████▊ | 16709/19090 [8:09:35<1:10:37,  1.78s/it][gradnorm] step 16709: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 740.8

 88%|████████▊ | 16710/19090 [8:09:36<1:04:50,  1.63s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1060.0037841796875, 'learning_rate': 2.332173172619925e-06, 'recon_loss': 0.0015008467016741633, 'recon_static': 0.0005230111288256012, 'recon_motion': 0.0009778355612070299, 'recon_norm': 1.271369817852974, 'recon_norm_std': 0.04021305013448, 'target_norm': 1.6203357100486755, 'recon_cos': 0.6477180480957031, 'gradnorm/compressor.query': 0.4688046900316464, 'gradnorm/compressor.layer0': 12.708938000578138, 'gradnorm/compressor.layer1': 1049.7076974091406, 'gradnorm/compressor.layer2': 138.16620567499493, 'gradnorm/compressor.layer3': 48.07975247082965, 'gradnorm/compressor.layer4': 9.323723296037599, 'gradnorm/compressor.layer5': 5.788391650991961, 'gradnorm/compressor.layer6': 2.838788409693432, 'gradnorm/compressor.layer7': 3.525508838478558, 'gradnorm/decoder.layer0': 3.508192283954272, 'gradnorm/decoder.layer1': 1.9522000983894072, 'gradnorm/decoder.layer2': 1.125662767368695, 'gradnorm/max_param': 740.7627563476562, 'epoch': 8.75}

 88%|████████▊ | 16710/19090 [8:09:36<1:04:50,  1.63s/it]
 88%|████████▊ | 16711/19090 [8:09:39<1:19:52,  2.01s/it]
 88%|████████▊ | 16712/19090 [8:09:41<1:11:14,  1.80s/it]
 88%|████████▊ | 16713/19090 [8:09:42<1:06:11,  1.67s/it]
 88%|████████▊ | 16714/19090 [8:09:45<1:20:27,  2.03s/it]
 88%|████████▊ | 16715/19090 [8:09:46<1:11:34,  1.81s/it]
 88%|████████▊ | 16716/19090 [8:09:47<1:05:20,  1.65s/it]
 88%|████████▊ | 16717/19090 [8:09:50<1:17:00,  1.95s/it]
 88%|████████▊ | 16718/19090 [8:09:51<1:09:05,  1.75s/it]
 88%|████████▊ | 16719/19090 [8:09:53<1:06:54,  1.69s/it][gradnorm] step 16719: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 566.7

 88%|████████▊ | 16720/19090 [8:09:55<1:15:39,  1.92s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 810.9878540039062, 'learning_rate': 2.312931634256543e-06, 'recon_loss': 0.0014725017215823754, 'recon_static': 0.0004973447474185377, 'recon_motion': 0.0009751569625223056, 'recon_norm': 1.2541823983192444, 'recon_norm_std': 0.041046032682061195, 'target_norm': 1.6187917679548263, 'recon_cos': 0.6529423326253891, 'gradnorm/compressor.query': 0.31566327778827613, 'gradnorm/compressor.layer0': 9.659801237537472, 'gradnorm/compressor.layer1': 803.063392578693, 'gradnorm/compressor.layer2': 105.81832345168299, 'gradnorm/compressor.layer3': 37.11354252914292, 'gradnorm/compressor.layer4': 8.165873378709064, 'gradnorm/compressor.layer5': 4.658459235294371, 'gradnorm/compressor.layer6': 2.7346880924266497, 'gradnorm/compressor.layer7': 3.2512109407852843, 'gradnorm/decoder.layer0': 3.3627852762022696, 'gradnorm/decoder.layer1': 1.8305229045494624, 'gradnorm/decoder.layer2': 1.0660880595724442, 'gradnorm/max_param': 566.7357788085938, 'epoch': 8.76}

 88%|████████▊ | 16720/19090 [8:09:55<1:15:39,  1.92s/it]
 88%|████████▊ | 16721/19090 [8:09:57<1:08:11,  1.73s/it]
 88%|████████▊ | 16722/19090 [8:09:58<1:03:46,  1.62s/it]
 88%|████████▊ | 16723/19090 [8:10:01<1:16:23,  1.94s/it]
 88%|████████▊ | 16724/19090 [8:10:02<1:08:39,  1.74s/it]
 88%|████████▊ | 16725/19090 [8:10:03<1:05:26,  1.66s/it]
 88%|████████▊ | 16726/19090 [8:10:06<1:13:34,  1.87s/it]
 88%|████████▊ | 16727/19090 [8:10:07<1:06:39,  1.69s/it]
 88%|████████▊ | 16728/19090 [8:10:09<1:06:16,  1.68s/it]
 88%|████████▊ | 16729/19090 [8:10:11<1:14:31,  1.89s/it][gradnorm] step 16729: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 658.5

 88%|████████▊ | 16730/19090 [8:10:12<1:07:21,  1.71s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 940.943359375, 'learning_rate': 2.2937659503544705e-06, 'recon_loss': 0.0014866636774968356, 'recon_static': 0.00052796148811467, 'recon_motion': 0.0009587022010236979, 'recon_norm': 1.2583679109811783, 'recon_norm_std': 0.04089445946738124, 'target_norm': 1.6163074910640716, 'recon_cos': 0.6475314036011696, 'gradnorm/compressor.query': 0.3577582230692509, 'gradnorm/compressor.layer0': 15.059957075931733, 'gradnorm/compressor.layer1': 933.1447837822382, 'gradnorm/compressor.layer2': 113.70296503527294, 'gradnorm/compressor.layer3': 36.60438139491894, 'gradnorm/compressor.layer4': 8.036053953712148, 'gradnorm/compressor.layer5': 5.05641505628598, 'gradnorm/compressor.layer6': 2.570169369500995, 'gradnorm/compressor.layer7': 3.0733794989880123, 'gradnorm/decoder.layer0': 3.0760833630619224, 'gradnorm/decoder.layer1': 1.8476805735755257, 'gradnorm/decoder.layer2': 0.9493357896272979, 'gradnorm/max_param': 658.5259399414062, 'epoch': 8.76}

 88%|████████▊ | 16730/19090 [8:10:12<1:07:21,  1.71s/it]
 88%|████████▊ | 16731/19090 [8:10:14<1:07:58,  1.73s/it]
 88%|████████▊ | 16732/19090 [8:10:16<1:14:39,  1.90s/it]
 88%|████████▊ | 16733/19090 [8:10:18<1:07:23,  1.72s/it]
 88%|████████▊ | 16734/19090 [8:10:19<1:02:20,  1.59s/it]
 88%|████████▊ | 16735/19090 [8:10:21<1:12:08,  1.84s/it]
 88%|████████▊ | 16736/19090 [8:10:23<1:07:01,  1.71s/it]
 88%|████████▊ | 16737/19090 [8:10:24<1:05:35,  1.67s/it]
 88%|████████▊ | 16738/19090 [8:10:27<1:17:49,  1.99s/it]
 88%|████████▊ | 16739/19090 [8:10:28<1:09:44,  1.78s/it][gradnorm] step 16739: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 596.1

 88%|████████▊ | 16740/19090 [8:10:30<1:05:41,  1.68s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 853.2833862304688, 'learning_rate': 2.2746761849943676e-06, 'recon_loss': 0.001580608417862095, 'recon_static': 0.0006008482101606205, 'recon_motion': 0.0009797602411708794, 'recon_norm': 1.309136089682579, 'recon_norm_std': 0.03988908110186458, 'target_norm': 1.621642169356346, 'recon_cos': 0.6315896183252334, 'gradnorm/compressor.query': 0.2953054573591798, 'gradnorm/compressor.layer0': 11.256211664027584, 'gradnorm/compressor.layer1': 844.8094829605075, 'gradnorm/compressor.layer2': 114.43782427589228, 'gradnorm/compressor.layer3': 32.453634670408064, 'gradnorm/compressor.layer4': 7.685248316548511, 'gradnorm/compressor.layer5': 4.731804581141978, 'gradnorm/compressor.layer6': 2.5488147486777835, 'gradnorm/compressor.layer7': 3.3066122163114406, 'gradnorm/decoder.layer0': 3.1806923967785385, 'gradnorm/decoder.layer1': 1.839131276052751, 'gradnorm/decoder.layer2': 0.9931192603644611, 'gradnorm/max_param': 596.146484375, 'epoch': 8.77}

 88%|████████▊ | 16740/19090 [8:10:30<1:05:41,  1.68s/it]
 88%|████████▊ | 16741/19090 [8:10:33<1:18:43,  2.01s/it]
 88%|████████▊ | 16742/19090 [8:10:34<1:10:10,  1.79s/it]
 88%|████████▊ | 16743/19090 [8:10:35<1:04:15,  1.64s/it]
 88%|████████▊ | 16744/19090 [8:10:38<1:14:15,  1.90s/it]
 88%|████████▊ | 16745/19090 [8:10:40<1:13:41,  1.89s/it]
 88%|████████▊ | 16746/19090 [8:10:41<1:06:33,  1.70s/it]
 88%|████████▊ | 16747/19090 [8:10:43<1:15:05,  1.92s/it]
 88%|████████▊ | 16748/19090 [8:10:45<1:07:42,  1.73s/it]
 88%|████████▊ | 16749/19090 [8:10:46<1:02:21,  1.60s/it][gradnorm] step 16749: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 749.8

 88%|████████▊ | 16750/19090 [8:10:49<1:15:39,  1.94s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1072.69482421875, 'learning_rate': 2.255662402003092e-06, 'recon_loss': 0.0016104921262012795, 'recon_static': 0.0006260170106543228, 'recon_motion': 0.0009844751490163616, 'recon_norm': 1.3076580256223678, 'recon_norm_std': 0.03934649312868714, 'target_norm': 1.6213324069976807, 'recon_cos': 0.6225007846951485, 'gradnorm/compressor.query': 0.3114151003929794, 'gradnorm/compressor.layer0': 8.907946829861121, 'gradnorm/compressor.layer1': 1062.5537045250937, 'gradnorm/compressor.layer2': 141.04396616170789, 'gradnorm/compressor.layer3': 39.213683998003965, 'gradnorm/compressor.layer4': 9.582498221015317, 'gradnorm/compressor.layer5': 4.441336464312673, 'gradnorm/compressor.layer6': 2.66287992748626, 'gradnorm/compressor.layer7': 3.3573951742321055, 'gradnorm/decoder.layer0': 2.961071323400229, 'gradnorm/decoder.layer1': 1.8905957432524534, 'gradnorm/decoder.layer2': 1.0732036441341581, 'gradnorm/max_param': 749.8228149414062, 'epoch': 8.77}

 88%|████████▊ | 16750/19090 [8:10:49<1:15:39,  1.94s/it]
 88%|████████▊ | 16751/19090 [8:10:50<1:07:58,  1.74s/it]
 88%|████████▊ | 16752/19090 [8:10:51<1:02:33,  1.61s/it]
 88%|████████▊ | 16753/19090 [8:10:53<1:10:09,  1.80s/it]
 88%|████████▊ | 16754/19090 [8:10:55<1:08:28,  1.76s/it]
 88%|████████▊ | 16755/19090 [8:10:56<1:02:59,  1.62s/it]
 88%|████████▊ | 16756/19090 [8:10:59<1:10:48,  1.82s/it]
 88%|████████▊ | 16757/19090 [8:11:01<1:11:57,  1.85s/it]
 88%|████████▊ | 16758/19090 [8:11:02<1:05:18,  1.68s/it]
 88%|████████▊ | 16759/19090 [8:11:04<1:07:37,  1.74s/it][gradnorm] step 16759: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 672.3

 88%|████████▊ | 16760/19090 [8:11:06<1:15:05,  1.93s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 965.513427734375, 'learning_rate': 2.236724664953427e-06, 'recon_loss': 0.0015014101431006567, 'recon_static': 0.0005237066121480894, 'recon_motion': 0.0009777035666047595, 'recon_norm': 1.2716129451990128, 'recon_norm_std': 0.041096166241914034, 'target_norm': 1.6210178285837173, 'recon_cos': 0.6482572913169861, 'gradnorm/compressor.query': 0.3638586767146299, 'gradnorm/compressor.layer0': 12.653269324936343, 'gradnorm/compressor.layer1': 952.6129526203179, 'gradnorm/compressor.layer2': 152.9122310040142, 'gradnorm/compressor.layer3': 32.707736401466434, 'gradnorm/compressor.layer4': 8.803897636071525, 'gradnorm/compressor.layer5': 4.566349861312447, 'gradnorm/compressor.layer6': 2.723261542881756, 'gradnorm/compressor.layer7': 3.3487118422159385, 'gradnorm/decoder.layer0': 3.261147674089856, 'gradnorm/decoder.layer1': 1.9207712120663365, 'gradnorm/decoder.layer2': 1.0684547172558287, 'gradnorm/max_param': 672.2681274414062, 'epoch': 8.78}

 88%|████████▊ | 16760/19090 [8:11:06<1:15:05,  1.93s/it]
 88%|████████▊ | 16761/19090 [8:11:07<1:07:32,  1.74s/it]
 88%|████████▊ | 16762/19090 [8:11:09<1:05:46,  1.70s/it]
 88%|████████▊ | 16763/19090 [8:11:12<1:16:04,  1.96s/it]
 88%|████████▊ | 16764/19090 [8:11:13<1:08:14,  1.76s/it]
 88%|████████▊ | 16765/19090 [8:11:14<1:05:34,  1.69s/it]
 88%|████████▊ | 16766/19090 [8:11:17<1:14:59,  1.94s/it]
 88%|████████▊ | 16767/19090 [8:11:18<1:07:29,  1.74s/it]
 88%|████████▊ | 16768/19090 [8:11:20<1:06:50,  1.73s/it]
 88%|████████▊ | 16769/19090 [8:11:22<1:16:15,  1.97s/it][gradnorm] step 16769: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 536.9

 88%|████████▊ | 16770/19090 [8:11:24<1:08:30,  1.77s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 772.654541015625, 'learning_rate': 2.2178630371638975e-06, 'recon_loss': 0.0014688694878714158, 'recon_static': 0.00048479559627594425, 'recon_motion': 0.000984073906147387, 'recon_norm': 1.249372461438179, 'recon_norm_std': 0.040636746771633626, 'target_norm': 1.621422040462494, 'recon_cos': 0.6546722084283829, 'gradnorm/compressor.query': 0.46917713096302616, 'gradnorm/compressor.layer0': 12.384869579726544, 'gradnorm/compressor.layer1': 760.9912121700223, 'gradnorm/compressor.layer2': 125.9390237190701, 'gradnorm/compressor.layer3': 40.88929652503819, 'gradnorm/compressor.layer4': 11.057570571484469, 'gradnorm/compressor.layer5': 5.472402344830472, 'gradnorm/compressor.layer6': 3.608402401892274, 'gradnorm/compressor.layer7': 3.838139230378442, 'gradnorm/decoder.layer0': 3.7854422037078184, 'gradnorm/decoder.layer1': 2.30517800494028, 'gradnorm/decoder.layer2': 1.2998668675843705, 'gradnorm/max_param': 536.936279296875, 'epoch': 8.78}

 88%|████████▊ | 16770/19090 [8:11:24<1:08:30,  1.77s/it]
 88%|████████▊ | 16771/19090 [8:11:25<1:03:23,  1.64s/it]
 88%|████████▊ | 16772/19090 [8:11:27<1:10:59,  1.84s/it]
 88%|████████▊ | 16773/19090 [8:11:29<1:04:35,  1.67s/it]
 88%|████████▊ | 16774/19090 [8:11:30<1:01:59,  1.61s/it]
 88%|████████▊ | 16775/19090 [8:11:33<1:15:58,  1.97s/it]
 88%|████████▊ | 16776/19090 [8:11:34<1:08:03,  1.76s/it]
 88%|████████▊ | 16777/19090 [8:11:36<1:02:40,  1.63s/it]
 88%|████████▊ | 16778/19090 [8:11:38<1:17:01,  2.00s/it]
 88%|████████▊ | 16779/19090 [8:11:40<1:08:53,  1.79s/it][gradnorm] step 16779: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 779.9

 88%|████████▊ | 16780/19090 [8:11:41<1:03:09,  1.64s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1115.149169921875, 'learning_rate': 2.199077581698572e-06, 'recon_loss': 0.0014858119568089024, 'recon_static': 0.0005129077391757165, 'recon_motion': 0.0009729042285471223, 'recon_norm': 1.266213595867157, 'recon_norm_std': 0.04010192025452852, 'target_norm': 1.6182194948196411, 'recon_cos': 0.6502924352884293, 'gradnorm/compressor.query': 0.41049706967245336, 'gradnorm/compressor.layer0': 12.043077080480979, 'gradnorm/compressor.layer1': 1105.0920210552604, 'gradnorm/compressor.layer2': 141.95147305725644, 'gradnorm/compressor.layer3': 43.83682929386046, 'gradnorm/compressor.layer4': 7.414393072928665, 'gradnorm/compressor.layer5': 5.035227654470911, 'gradnorm/compressor.layer6': 2.8490494196881166, 'gradnorm/compressor.layer7': 3.1933554926390775, 'gradnorm/decoder.layer0': 3.0058159084535925, 'gradnorm/decoder.layer1': 1.9214810843845704, 'gradnorm/decoder.layer2': 0.9995522092548164, 'gradnorm/max_param': 779.85791015625, 'epoch': 8.79}

 88%|████████▊ | 16780/19090 [8:11:41<1:03:09,  1.64s/it]
 88%|████████▊ | 16781/19090 [8:11:44<1:15:43,  1.97s/it]
 88%|████████▊ | 16782/19090 [8:11:45<1:07:57,  1.77s/it]
 88%|████████▊ | 16783/19090 [8:11:46<1:02:24,  1.62s/it]
 88%|████████▊ | 16784/19090 [8:11:49<1:15:00,  1.95s/it]
 88%|████████▊ | 16785/19090 [8:11:50<1:07:22,  1.75s/it]
 88%|████████▊ | 16786/19090 [8:11:52<1:02:08,  1.62s/it]
 88%|████████▊ | 16787/19090 [8:11:54<1:15:06,  1.96s/it]
 88%|████████▊ | 16788/19090 [8:11:56<1:07:26,  1.76s/it]
 88%|████████▊ | 16789/19090 [8:11:57<1:02:02,  1.62s/it][gradnorm] step 16789: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 797.0

 88%|████████▊ | 16790/19090 [8:12:00<1:12:35,  1.89s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1139.0970458984375, 'learning_rate': 2.180368361366825e-06, 'recon_loss': 0.0014959482970880345, 'recon_static': 0.0005192636570427566, 'recon_motion': 0.0009766846807906405, 'recon_norm': 1.256775477528572, 'recon_norm_std': 0.040272290166467425, 'target_norm': 1.6208565473556518, 'recon_cos': 0.6474417388439179, 'gradnorm/compressor.query': 0.34527147144284825, 'gradnorm/compressor.layer0': 10.93927929165804, 'gradnorm/compressor.layer1': 1129.3665259781699, 'gradnorm/compressor.layer2': 143.26177962038585, 'gradnorm/compressor.layer3': 36.17550602725724, 'gradnorm/compressor.layer4': 8.37990824792565, 'gradnorm/compressor.layer5': 4.592287829984934, 'gradnorm/compressor.layer6': 2.739779031433514, 'gradnorm/compressor.layer7': 2.9242219525910027, 'gradnorm/decoder.layer0': 3.0606928573293235, 'gradnorm/decoder.layer1': 1.8162645510296076, 'gradnorm/decoder.layer2': 0.9316956125982131, 'gradnorm/max_param': 797.0223388671875, 'epoch': 8.8}

 88%|████████▊ | 16790/19090 [8:12:00<1:12:35,  1.89s/it]
 88%|████████▊ | 16791/19090 [8:12:01<1:05:47,  1.72s/it]
 88%|████████▊ | 16792/19090 [8:12:02<1:00:59,  1.59s/it]
 88%|████████▊ | 16793/19090 [8:12:05<1:13:12,  1.91s/it]
 88%|████████▊ | 16794/19090 [8:12:06<1:06:12,  1.73s/it]
 88%|████████▊ | 16795/19090 [8:12:07<1:01:13,  1.60s/it]
 88%|████████▊ | 16796/19090 [8:12:10<1:13:57,  1.93s/it]
 88%|████████▊ | 16797/19090 [8:12:11<1:06:37,  1.74s/it]
 88%|████████▊ | 16798/19090 [8:12:13<1:01:20,  1.61s/it]
 88%|████████▊ | 16799/19090 [8:12:15<1:12:18,  1.89s/it][gradnorm] step 16799: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 633.2

 88%|████████▊ | 16800/19090 [8:12:17<1:05:19,  1.71s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 907.43505859375, 'learning_rate': 2.161735438723131e-06, 'recon_loss': 0.0015079316537594423, 'recon_static': 0.0005298582742398139, 'recon_motion': 0.000978073371516075, 'recon_norm': 1.2681740880012513, 'recon_norm_std': 0.04026435315608978, 'target_norm': 1.6195038408041, 'recon_cos': 0.6446097046136856, 'gradnorm/compressor.query': 0.3535747162554588, 'gradnorm/compressor.layer0': 9.687045178965379, 'gradnorm/compressor.layer1': 897.2635482398691, 'gradnorm/compressor.layer2': 129.57057228832866, 'gradnorm/compressor.layer3': 36.61099048132183, 'gradnorm/compressor.layer4': 9.276761494886447, 'gradnorm/compressor.layer5': 4.24427095117435, 'gradnorm/compressor.layer6': 2.7375763381695695, 'gradnorm/compressor.layer7': 3.02132734446705, 'gradnorm/decoder.layer0': 3.0344193397370307, 'gradnorm/decoder.layer1': 1.697734947855753, 'gradnorm/decoder.layer2': 0.9264227338750465, 'gradnorm/max_param': 633.2200317382812, 'epoch': 8.8}

 88%|████████▊ | 16800/19090 [8:12:17<1:05:19,  1.71s/it]
 88%|████████▊ | 16801/19090 [8:12:18<1:00:25,  1.58s/it]
 88%|████████▊ | 16802/19090 [8:12:21<1:13:52,  1.94s/it]
 88%|████████▊ | 16803/19090 [8:12:22<1:08:35,  1.80s/it]
 88%|████████▊ | 16804/19090 [8:12:23<1:02:46,  1.65s/it]
 88%|████████▊ | 16805/19090 [8:12:26<1:12:34,  1.91s/it]
 88%|████████▊ | 16806/19090 [8:12:27<1:06:30,  1.75s/it]
 88%|████████▊ | 16807/19090 [8:12:29<1:03:01,  1.66s/it]
 88%|████████▊ | 16808/19090 [8:12:31<1:11:02,  1.87s/it]
 88%|████████▊ | 16809/19090 [8:12:32<1:05:40,  1.73s/it][gradnorm] step 16809: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 766.8

 88%|████████▊ | 16810/19090 [8:12:34<1:03:56,  1.68s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1097.6435546875, 'learning_rate': 2.1431788760668814e-06, 'recon_loss': 0.001497263173223473, 'recon_static': 0.0005187888244108762, 'recon_motion': 0.000978474372823257, 'recon_norm': 1.269240403175354, 'recon_norm_std': 0.040307024773210284, 'target_norm': 1.6210776269435883, 'recon_cos': 0.648710410296917, 'gradnorm/compressor.query': 0.38033199476618895, 'gradnorm/compressor.layer0': 13.44477682692358, 'gradnorm/compressor.layer1': 1086.5082259237618, 'gradnorm/compressor.layer2': 149.4038871126685, 'gradnorm/compressor.layer3': 40.965575117732236, 'gradnorm/compressor.layer4': 9.021612284964082, 'gradnorm/compressor.layer5': 5.03140069013223, 'gradnorm/compressor.layer6': 3.0121203994598695, 'gradnorm/compressor.layer7': 3.3969971860301817, 'gradnorm/decoder.layer0': 2.9823032398023748, 'gradnorm/decoder.layer1': 1.7952768931497427, 'gradnorm/decoder.layer2': 1.0604521145294978, 'gradnorm/max_param': 766.8468017578125, 'epoch': 8.81}

 88%|████████▊ | 16810/19090 [8:12:34<1:03:56,  1.68s/it]
 88%|████████▊ | 16811/19090 [8:12:37<1:14:18,  1.96s/it]
 88%|████████▊ | 16812/19090 [8:12:38<1:07:39,  1.78s/it]
 88%|████████▊ | 16813/19090 [8:12:40<1:04:45,  1.71s/it]
 88%|████████▊ | 16814/19090 [8:12:42<1:14:34,  1.97s/it]
 88%|████████▊ | 16815/19090 [8:12:43<1:07:10,  1.77s/it]
 88%|████████▊ | 16816/19090 [8:12:45<1:02:33,  1.65s/it]
 88%|████████▊ | 16817/19090 [8:12:48<1:15:03,  1.98s/it]
 88%|████████▊ | 16818/19090 [8:12:49<1:07:14,  1.78s/it]
 88%|████████▊ | 16819/19090 [8:12:50<1:03:37,  1.68s/it][gradnorm] step 16819: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1018.9

 88%|████████▊ | 16820/19090 [8:12:53<1:16:13,  2.01s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1454.8458251953125, 'learning_rate': 2.124698735442143e-06, 'recon_loss': 0.001489402714651078, 'recon_static': 0.0005175605503609404, 'recon_motion': 0.0009718421788420528, 'recon_norm': 1.260433778166771, 'recon_norm_std': 0.04191509634256363, 'target_norm': 1.619882845878601, 'recon_cos': 0.6492148712277412, 'gradnorm/compressor.query': 0.4922789080071374, 'gradnorm/compressor.layer0': 15.872498322662683, 'gradnorm/compressor.layer1': 1443.9030438363927, 'gradnorm/compressor.layer2': 171.42489358498958, 'gradnorm/compressor.layer3': 43.49173912446262, 'gradnorm/compressor.layer4': 11.28465561962851, 'gradnorm/compressor.layer5': 5.167276695088383, 'gradnorm/compressor.layer6': 3.2274424389848457, 'gradnorm/compressor.layer7': 3.322604046090164, 'gradnorm/decoder.layer0': 3.171338266144977, 'gradnorm/decoder.layer1': 1.8921221015395384, 'gradnorm/decoder.layer2': 1.1668079438273573, 'gradnorm/max_param': 1018.932861328125, 'epoch': 8.81}

 88%|████████▊ | 16820/19090 [8:12:53<1:16:13,  2.01s/it]
 88%|████████▊ | 16821/19090 [8:12:54<1:08:03,  1.80s/it]
 88%|████████▊ | 16822/19090 [8:12:56<1:02:24,  1.65s/it]
 88%|████████▊ | 16823/19090 [8:12:58<1:11:08,  1.88s/it]
 88%|████████▊ | 16824/19090 [8:12:59<1:04:27,  1.71s/it]
 88%|████████▊ | 16825/19090 [8:13:01<1:01:39,  1.63s/it]
 88%|████████▊ | 16826/19090 [8:13:04<1:12:42,  1.93s/it]
 88%|████████▊ | 16827/19090 [8:13:05<1:05:27,  1.74s/it]
 88%|████████▊ | 16828/19090 [8:13:06<1:00:19,  1.60s/it]
 88%|████████▊ | 16829/19090 [8:13:09<1:14:09,  1.97s/it][gradnorm] step 16829: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 700.4

 88%|████████▊ | 16830/19090 [8:13:10<1:06:28,  1.76s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1001.7625122070312, 'learning_rate': 2.1062950786374603e-06, 'recon_loss': 0.0014825833000941201, 'recon_static': 0.0005053965331171639, 'recon_motion': 0.0009771867655217648, 'recon_norm': 1.2659497827291488, 'recon_norm_std': 0.04039153028279543, 'target_norm': 1.619620668888092, 'recon_cos': 0.6518110454082489, 'gradnorm/compressor.query': 0.40331641465490314, 'gradnorm/compressor.layer0': 13.608344020399361, 'gradnorm/compressor.layer1': 992.3274724101918, 'gradnorm/compressor.layer2': 128.40998792928843, 'gradnorm/compressor.layer3': 44.60300140545477, 'gradnorm/compressor.layer4': 9.739262855411434, 'gradnorm/compressor.layer5': 4.738031265947874, 'gradnorm/compressor.layer6': 2.9396907672314447, 'gradnorm/compressor.layer7': 3.240005531070839, 'gradnorm/decoder.layer0': 3.056892555262842, 'gradnorm/decoder.layer1': 1.8707244444879074, 'gradnorm/decoder.layer2': 1.0805199048999277, 'gradnorm/max_param': 700.400146484375, 'epoch': 8.82}

 88%|████████▊ | 16830/19090 [8:13:10<1:06:28,  1.76s/it]
 88%|████████▊ | 16831/19090 [8:13:12<1:03:32,  1.69s/it]
 88%|████████▊ | 16832/19090 [8:13:14<1:15:54,  2.02s/it]
 88%|████████▊ | 16833/19090 [8:13:16<1:07:41,  1.80s/it]
 88%|████████▊ | 16834/19090 [8:13:17<1:01:53,  1.65s/it]
 88%|████████▊ | 16835/19090 [8:13:20<1:11:18,  1.90s/it]
 88%|████████▊ | 16836/19090 [8:13:21<1:04:24,  1.71s/it]
 88%|████████▊ | 16837/19090 [8:13:22<59:34,  1.59s/it]  
 88%|████████▊ | 16838/19090 [8:13:25<1:11:14,  1.90s/it]
 88%|████████▊ | 16839/19090 [8:13:26<1:04:23,  1.72s/it][gradnorm] step 16839: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 772.7

 88%|████████▊ | 16840/19090 [8:13:27<59:36,  1.59s/it]  
                                                       
{'loss': 0.0015, 'grad_norm': 1103.875, 'learning_rate': 2.087967967185675e-06, 'recon_loss': 0.0014922591217327863, 'recon_static': 0.0005218011618126183, 'recon_motion': 0.0009704579380922951, 'recon_norm': 1.258057999610901, 'recon_norm_std': 0.04286355022341013, 'target_norm': 1.621339562535286, 'recon_cos': 0.6487712830305099, 'gradnorm/compressor.query': 0.5627530906114482, 'gradnorm/compressor.layer0': 18.476765656916932, 'gradnorm/compressor.layer1': 1095.3707135029674, 'gradnorm/compressor.layer2': 127.63795003470754, 'gradnorm/compressor.layer3': 43.81280508765032, 'gradnorm/compressor.layer4': 9.639075749833859, 'gradnorm/compressor.layer5': 5.054967669073263, 'gradnorm/compressor.layer6': 2.992856102886988, 'gradnorm/compressor.layer7': 3.257362389515987, 'gradnorm/decoder.layer0': 2.8655914936997995, 'gradnorm/decoder.layer1': 1.8164745030859442, 'gradnorm/decoder.layer2': 1.0348563281606207, 'gradnorm/max_param': 772.6675415039062, 'epoch': 8.82}

 88%|████████▊ | 16840/19090 [8:13:27<59:36,  1.59s/it]
 88%|████████▊ | 16841/19090 [8:13:30<1:09:25,  1.85s/it]
 88%|████████▊ | 16842/19090 [8:13:31<1:03:13,  1.69s/it]
 88%|████████▊ | 16843/19090 [8:13:32<58:45,  1.57s/it]  
 88%|████████▊ | 16844/19090 [8:13:35<1:10:54,  1.89s/it]
 88%|████████▊ | 16845/19090 [8:13:37<1:06:33,  1.78s/it]
 88%|████████▊ | 16846/19090 [8:13:38<1:01:03,  1.63s/it]
 88%|████████▊ | 16847/19090 [8:13:41<1:13:44,  1.97s/it]
 88%|████████▊ | 16848/19090 [8:13:42<1:06:06,  1.77s/it]
 88%|████████▊ | 16849/19090 [8:13:43<1:00:46,  1.63s/it][gradnorm] step 16849: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 876.4

 88%|████████▊ | 16850/19090 [8:13:46<1:11:22,  1.91s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1252.8394775390625, 'learning_rate': 2.0697174623636794e-06, 'recon_loss': 0.0015084013633895666, 'recon_static': 0.000534891257120762, 'recon_motion': 0.0009735101019032299, 'recon_norm': 1.2753714561462401, 'recon_norm_std': 0.04076064759865403, 'target_norm': 1.6209876447916032, 'recon_cos': 0.6461371541023254, 'gradnorm/compressor.query': 0.4058615220924544, 'gradnorm/compressor.layer0': 13.589106044938568, 'gradnorm/compressor.layer1': 1241.9375487519492, 'gradnorm/compressor.layer2': 155.79387799870378, 'gradnorm/compressor.layer3': 50.73919652943989, 'gradnorm/compressor.layer4': 10.391649972942519, 'gradnorm/compressor.layer5': 4.822183551916499, 'gradnorm/compressor.layer6': 3.126372836878759, 'gradnorm/compressor.layer7': 3.401834688524249, 'gradnorm/decoder.layer0': 3.042644831173128, 'gradnorm/decoder.layer1': 2.0568569319565224, 'gradnorm/decoder.layer2': 1.0742538168992273, 'gradnorm/max_param': 876.4295654296875, 'epoch': 8.83}

 88%|████████▊ | 16850/19090 [8:13:46<1:11:22,  1.91s/it]
 88%|████████▊ | 16851/19090 [8:13:47<1:04:23,  1.73s/it]
 88%|████████▊ | 16852/19090 [8:13:48<59:35,  1.60s/it]  
 88%|████████▊ | 16853/19090 [8:13:51<1:11:42,  1.92s/it]
 88%|████████▊ | 16854/19090 [8:13:52<1:04:40,  1.74s/it]
 88%|████████▊ | 16855/19090 [8:13:54<1:01:35,  1.65s/it]
 88%|████████▊ | 16856/19090 [8:13:56<1:11:48,  1.93s/it]
 88%|████████▊ | 16857/19090 [8:13:58<1:04:41,  1.74s/it]
 88%|████████▊ | 16858/19090 [8:13:59<1:01:40,  1.66s/it]
 88%|████████▊ | 16859/19090 [8:14:02<1:17:54,  2.10s/it][gradnorm] step 16859: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 981.9

 88%|████████▊ | 16860/19090 [8:14:04<1:08:57,  1.86s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1403.36376953125, 'learning_rate': 2.051543625192226e-06, 'recon_loss': 0.0014820419193711132, 'recon_static': 0.0005103240866446868, 'recon_motion': 0.000971717840002384, 'recon_norm': 1.2522439450025558, 'recon_norm_std': 0.04063697848469019, 'target_norm': 1.6200997143983842, 'recon_cos': 0.6504829600453377, 'gradnorm/compressor.query': 0.4627240038019795, 'gradnorm/compressor.layer0': 17.004888392695992, 'gradnorm/compressor.layer1': 1391.4194640725707, 'gradnorm/compressor.layer2': 175.21632277166702, 'gradnorm/compressor.layer3': 46.93200700215153, 'gradnorm/compressor.layer4': 11.258427832767104, 'gradnorm/compressor.layer5': 5.261648425430193, 'gradnorm/compressor.layer6': 3.252717862408549, 'gradnorm/compressor.layer7': 3.2203046877274386, 'gradnorm/decoder.layer0': 2.9904124918885464, 'gradnorm/decoder.layer1': 1.905489910738642, 'gradnorm/decoder.layer2': 1.0865457601412472, 'gradnorm/max_param': 981.8698120117188, 'epoch': 8.83}

 88%|████████▊ | 16860/19090 [8:14:04<1:08:57,  1.86s/it]
 88%|████████▊ | 16861/19090 [8:14:05<1:02:34,  1.68s/it]
 88%|████████▊ | 16862/19090 [8:14:08<1:16:12,  2.05s/it]
 88%|████████▊ | 16863/19090 [8:14:09<1:07:41,  1.82s/it]
 88%|████████▊ | 16864/19090 [8:14:10<1:01:40,  1.66s/it]
 88%|████████▊ | 16865/19090 [8:14:13<1:15:47,  2.04s/it]
 88%|████████▊ | 16866/19090 [8:14:15<1:07:23,  1.82s/it]
 88%|████████▊ | 16867/19090 [8:14:16<1:01:26,  1.66s/it]
 88%|████████▊ | 16868/19090 [8:14:19<1:14:20,  2.01s/it]
 88%|████████▊ | 16869/19090 [8:14:20<1:06:18,  1.79s/it][gradnorm] step 16869: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 704.0

 88%|████████▊ | 16870/19090 [8:14:21<1:00:42,  1.64s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1009.1835327148438, 'learning_rate': 2.03344651643575e-06, 'recon_loss': 0.0014896485663484782, 'recon_static': 0.000518075143918395, 'recon_motion': 0.0009715733889606781, 'recon_norm': 1.264518576860428, 'recon_norm_std': 0.04079511435702443, 'target_norm': 1.620483735203743, 'recon_cos': 0.6499531850218773, 'gradnorm/compressor.query': 0.388255620068307, 'gradnorm/compressor.layer0': 15.612640131007682, 'gradnorm/compressor.layer1': 997.598898606048, 'gradnorm/compressor.layer2': 146.76689505598665, 'gradnorm/compressor.layer3': 36.17659087779088, 'gradnorm/compressor.layer4': 9.597996470872596, 'gradnorm/compressor.layer5': 5.275349588245147, 'gradnorm/compressor.layer6': 3.086785076937825, 'gradnorm/compressor.layer7': 3.386851940593932, 'gradnorm/decoder.layer0': 3.0782932293614618, 'gradnorm/decoder.layer1': 1.834195028441362, 'gradnorm/decoder.layer2': 0.9858149335943991, 'gradnorm/max_param': 703.974365234375, 'epoch': 8.84}

 88%|████████▊ | 16870/19090 [8:14:21<1:00:42,  1.64s/it]
 88%|████████▊ | 16871/19090 [8:14:24<1:13:18,  1.98s/it]
 88%|████████▊ | 16872/19090 [8:14:25<1:05:36,  1.77s/it]
 88%|████████▊ | 16873/19090 [8:14:27<1:00:11,  1.63s/it]
 88%|████████▊ | 16874/19090 [8:14:30<1:15:01,  2.03s/it]
 88%|████████▊ | 16875/19090 [8:14:31<1:06:44,  1.81s/it]
 88%|████████▊ | 16876/19090 [8:14:32<1:01:03,  1.65s/it]
 88%|████████▊ | 16877/19090 [8:14:35<1:12:14,  1.96s/it]
 88%|████████▊ | 16878/19090 [8:14:36<1:04:58,  1.76s/it]
 88%|████████▊ | 16879/19090 [8:14:37<1:00:02,  1.63s/it][gradnorm] step 16879: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 800.0

 88%|████████▊ | 16880/19090 [8:14:40<1:12:45,  1.98s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1146.206298828125, 'learning_rate': 2.015426196602133e-06, 'recon_loss': 0.0015135812223888933, 'recon_static': 0.0005265105282887816, 'recon_motion': 0.0009870706679066643, 'recon_norm': 1.2628899812698364, 'recon_norm_std': 0.04081329824402928, 'target_norm': 1.6234589636325836, 'recon_cos': 0.6447774142026901, 'gradnorm/compressor.query': 0.3308961089008038, 'gradnorm/compressor.layer0': 11.604160457223616, 'gradnorm/compressor.layer1': 1133.6364055551498, 'gradnorm/compressor.layer2': 164.21300834396465, 'gradnorm/compressor.layer3': 37.55561305937669, 'gradnorm/compressor.layer4': 9.34989681084958, 'gradnorm/compressor.layer5': 4.97938273239062, 'gradnorm/compressor.layer6': 3.0785667760786657, 'gradnorm/compressor.layer7': 3.284077615822008, 'gradnorm/decoder.layer0': 3.0658921120530356, 'gradnorm/decoder.layer1': 1.8248580694530934, 'gradnorm/decoder.layer2': 0.975505898125912, 'gradnorm/max_param': 799.9974975585938, 'epoch': 8.84}

 88%|████████▊ | 16880/19090 [8:14:40<1:12:45,  1.98s/it]
 88%|████████▊ | 16881/19090 [8:14:42<1:05:09,  1.77s/it]
 88%|████████▊ | 16882/19090 [8:14:43<59:48,  1.63s/it]  
 88%|████████▊ | 16883/19090 [8:14:45<1:10:18,  1.91s/it]
 88%|████████▊ | 16884/19090 [8:14:47<1:03:32,  1.73s/it]
 88%|████████▊ | 16885/19090 [8:14:48<58:48,  1.60s/it]  
 88%|████████▊ | 16886/19090 [8:14:51<1:12:51,  1.98s/it]
 88%|████████▊ | 16887/19090 [8:14:52<1:05:15,  1.78s/it]
 88%|████████▊ | 16888/19090 [8:14:53<59:53,  1.63s/it]  
 88%|████████▊ | 16889/19090 [8:14:56<1:11:03,  1.94s/it][gradnorm] step 16889: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 658.6

 88%|████████▊ | 16890/19090 [8:14:57<1:03:53,  1.74s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 942.69873046875, 'learning_rate': 1.997482725942501e-06, 'recon_loss': 0.0014824489073362202, 'recon_static': 0.0005086723118438386, 'recon_motion': 0.0009737766071339138, 'recon_norm': 1.2578482896089553, 'recon_norm_std': 0.04073642361909151, 'target_norm': 1.620615404844284, 'recon_cos': 0.6514790445566178, 'gradnorm/compressor.query': 0.28923487034598067, 'gradnorm/compressor.layer0': 10.313201695469054, 'gradnorm/compressor.layer1': 933.1789753310991, 'gradnorm/compressor.layer2': 126.73131102972343, 'gradnorm/compressor.layer3': 39.54903101837247, 'gradnorm/compressor.layer4': 8.522908474932153, 'gradnorm/compressor.layer5': 4.499032870316517, 'gradnorm/compressor.layer6': 2.8110891406223066, 'gradnorm/compressor.layer7': 3.464371400652672, 'gradnorm/decoder.layer0': 3.058003920468399, 'gradnorm/decoder.layer1': 1.8100103840609105, 'gradnorm/decoder.layer2': 1.0033226246330973, 'gradnorm/max_param': 658.5796508789062, 'epoch': 8.85}

 88%|████████▊ | 16890/19090 [8:14:57<1:03:53,  1.74s/it]
 88%|████████▊ | 16891/19090 [8:14:59<58:53,  1.61s/it]  
 88%|████████▊ | 16892/19090 [8:15:01<1:11:47,  1.96s/it]
 88%|████████▊ | 16893/19090 [8:15:03<1:04:25,  1.76s/it]
 88%|████████▊ | 16894/19090 [8:15:04<59:14,  1.62s/it]  
 89%|████████▊ | 16895/19090 [8:15:07<1:14:25,  2.03s/it]
 89%|████████▊ | 16896/19090 [8:15:08<1:06:09,  1.81s/it]
 89%|████████▊ | 16897/19090 [8:15:10<1:00:24,  1.65s/it]
 89%|████████▊ | 16898/19090 [8:15:12<1:08:42,  1.88s/it]
 89%|████████▊ | 16899/19090 [8:15:13<1:02:12,  1.70s/it][gradnorm] step 16899: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 796.1

 89%|████████▊ | 16900/19090 [8:15:15<57:39,  1.58s/it]  
                                                       
{'loss': 0.0015, 'grad_norm': 1140.1741943359375, 'learning_rate': 1.9796161644510563e-06, 'recon_loss': 0.0015071266505401581, 'recon_static': 0.0005198875820497051, 'recon_motion': 0.0009872390786767938, 'recon_norm': 1.2674431502819061, 'recon_norm_std': 0.04078189739957452, 'target_norm': 1.6228635132312774, 'recon_cos': 0.6472213178873062, 'gradnorm/compressor.query': 0.5054859565270767, 'gradnorm/compressor.layer0': 18.76705035308101, 'gradnorm/compressor.layer1': 1128.1735238871722, 'gradnorm/compressor.layer2': 157.4798474210272, 'gradnorm/compressor.layer3': 43.84831030274998, 'gradnorm/compressor.layer4': 9.479843842069316, 'gradnorm/compressor.layer5': 4.455613040209721, 'gradnorm/compressor.layer6': 3.202807303699957, 'gradnorm/compressor.layer7': 3.7253087011855026, 'gradnorm/decoder.layer0': 2.9763007261654932, 'gradnorm/decoder.layer1': 1.6880164592356293, 'gradnorm/decoder.layer2': 0.9994888489403153, 'gradnorm/max_param': 796.106201171875, 'epoch': 8.85}

 89%|████████▊ | 16900/19090 [8:15:15<57:39,  1.58s/it]
 89%|████████▊ | 16901/19090 [8:15:17<1:09:50,  1.91s/it]
 89%|████████▊ | 16902/19090 [8:15:19<1:02:56,  1.73s/it]
 89%|████████▊ | 16903/19090 [8:15:20<58:16,  1.60s/it]  
 89%|████████▊ | 16904/19090 [8:15:23<1:10:35,  1.94s/it]
 89%|████████▊ | 16905/19090 [8:15:24<1:03:30,  1.74s/it]
 89%|████████▊ | 16906/19090 [8:15:25<58:31,  1.61s/it]  
 89%|████████▊ | 16907/19090 [8:15:28<1:11:18,  1.96s/it]
 89%|████████▊ | 16908/19090 [8:15:29<1:03:55,  1.76s/it]
 89%|████████▊ | 16909/19090 [8:15:31<58:44,  1.62s/it]  [gradnorm] step 16909: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 803.0

 89%|████████▊ | 16910/19090 [8:15:33<1:07:15,  1.85s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1148.2779541015625, 'learning_rate': 1.9618265718648388e-06, 'recon_loss': 0.0014775780873605981, 'recon_static': 0.0005071633364423178, 'recon_motion': 0.0009704148076707497, 'recon_norm': 1.2590159595012664, 'recon_norm_std': 0.04086615527048707, 'target_norm': 1.618786373734474, 'recon_cos': 0.6517410337924957, 'gradnorm/compressor.query': 0.40458609882880187, 'gradnorm/compressor.layer0': 15.734368696467207, 'gradnorm/compressor.layer1': 1137.8209437341184, 'gradnorm/compressor.layer2': 148.96726517510484, 'gradnorm/compressor.layer3': 36.67802761819312, 'gradnorm/compressor.layer4': 8.425879041504828, 'gradnorm/compressor.layer5': 4.134292539435949, 'gradnorm/compressor.layer6': 3.088853293914375, 'gradnorm/compressor.layer7': 3.4356233502627447, 'gradnorm/decoder.layer0': 2.9032493605734953, 'gradnorm/decoder.layer1': 1.6805497905888036, 'gradnorm/decoder.layer2': 1.013069341261425, 'gradnorm/max_param': 802.9561157226562, 'epoch': 8.86}

 89%|████████▊ | 16910/19090 [8:15:33<1:07:15,  1.85s/it]
 89%|████████▊ | 16911/19090 [8:15:34<1:02:10,  1.71s/it]
 89%|████████▊ | 16912/19090 [8:15:36<1:06:54,  1.84s/it]
 89%|████████▊ | 16913/19090 [8:15:38<1:06:31,  1.83s/it]
 89%|████████▊ | 16914/19090 [8:15:40<1:02:32,  1.72s/it]
 89%|████████▊ | 16915/19090 [8:15:41<57:51,  1.60s/it]  
 89%|████████▊ | 16916/19090 [8:15:44<1:09:19,  1.91s/it]
 89%|████████▊ | 16917/19090 [8:15:45<1:07:28,  1.86s/it]
 89%|████████▊ | 16918/19090 [8:15:47<1:01:12,  1.69s/it]
 89%|████████▊ | 16919/19090 [8:15:49<1:08:31,  1.89s/it][gradnorm] step 16919: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 650.8

 89%|████████▊ | 16920/19090 [8:15:51<1:09:09,  1.91s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 931.3922119140625, 'learning_rate': 1.9441140076635388e-06, 'recon_loss': 0.0014765566913411022, 'recon_static': 0.0005003895770641975, 'recon_motion': 0.0009761671113665216, 'recon_norm': 1.2618622720241546, 'recon_norm_std': 0.040650746412575246, 'target_norm': 1.6222472935914993, 'recon_cos': 0.6542833596467972, 'gradnorm/compressor.query': 0.4948573135863492, 'gradnorm/compressor.layer0': 19.17825382879982, 'gradnorm/compressor.layer1': 922.3991814827244, 'gradnorm/compressor.layer2': 124.6397543297282, 'gradnorm/compressor.layer3': 25.145801644251296, 'gradnorm/compressor.layer4': 8.924395577616517, 'gradnorm/compressor.layer5': 4.920957255707471, 'gradnorm/compressor.layer6': 2.579966032262947, 'gradnorm/compressor.layer7': 3.5941577182525415, 'gradnorm/decoder.layer0': 2.8663649873422647, 'gradnorm/decoder.layer1': 1.7305890939169393, 'gradnorm/decoder.layer2': 1.01125498861207, 'gradnorm/max_param': 650.8212890625, 'epoch': 8.86}

 89%|████████▊ | 16920/19090 [8:15:51<1:09:09,  1.91s/it]
 89%|████████▊ | 16921/19090 [8:15:52<1:02:24,  1.73s/it]
 89%|████████▊ | 16922/19090 [8:15:54<1:04:57,  1.80s/it]
 89%|████████▊ | 16923/19090 [8:15:56<1:04:36,  1.79s/it]
 89%|████████▊ | 16924/19090 [8:15:57<59:10,  1.64s/it]  
 89%|████████▊ | 16925/19090 [8:16:00<1:09:02,  1.91s/it]
 89%|████████▊ | 16926/19090 [8:16:02<1:07:01,  1.86s/it]
 89%|████████▊ | 16927/19090 [8:16:03<1:00:53,  1.69s/it]
 89%|████████▊ | 16928/19090 [8:16:05<1:06:52,  1.86s/it]
 89%|████████▊ | 16929/19090 [8:16:07<1:07:42,  1.88s/it][gradnorm] step 16929: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 823.6

 89%|████████▊ | 16930/19090 [8:16:08<1:01:18,  1.70s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1181.5347900390625, 'learning_rate': 1.926478531069317e-06, 'recon_loss': 0.0015025192056782543, 'recon_static': 0.0005200753308599815, 'recon_motion': 0.0009824438791838474, 'recon_norm': 1.2724044889211654, 'recon_norm_std': 0.0401230157352984, 'target_norm': 1.6208351880311966, 'recon_cos': 0.648071177303791, 'gradnorm/compressor.query': 0.49893459417564245, 'gradnorm/compressor.layer0': 16.330864752201197, 'gradnorm/compressor.layer1': 1167.1227977380959, 'gradnorm/compressor.layer2': 177.40274073946827, 'gradnorm/compressor.layer3': 44.2890631890545, 'gradnorm/compressor.layer4': 9.532370979802561, 'gradnorm/compressor.layer5': 4.6834420504910295, 'gradnorm/compressor.layer6': 2.844390294019855, 'gradnorm/compressor.layer7': 3.6819642751888684, 'gradnorm/decoder.layer0': 3.103769281199156, 'gradnorm/decoder.layer1': 1.8569442503967157, 'gradnorm/decoder.layer2': 0.9820284349496776, 'gradnorm/max_param': 823.6319580078125, 'epoch': 8.87}

 89%|████████▊ | 16930/19090 [8:16:08<1:01:18,  1.70s/it]
 89%|████████▊ | 16931/19090 [8:16:11<1:05:24,  1.82s/it]
 89%|████████▊ | 16932/19090 [8:16:12<1:06:42,  1.85s/it]
 89%|████████▊ | 16933/19090 [8:16:14<1:00:32,  1.68s/it]
 89%|████████▊ | 16934/19090 [8:16:16<1:05:21,  1.82s/it]
 89%|████████▊ | 16935/19090 [8:16:18<1:03:12,  1.76s/it]
 89%|████████▊ | 16936/19090 [8:16:19<58:06,  1.62s/it]  
 89%|████████▊ | 16937/19090 [8:16:22<1:10:21,  1.96s/it]
 89%|████████▊ | 16938/19090 [8:16:23<1:03:02,  1.76s/it]
 89%|████████▊ | 16939/19090 [8:16:24<58:02,  1.62s/it]  [gradnorm] step 16939: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 882.7

 89%|████████▊ | 16940/19090 [8:16:27<1:10:56,  1.98s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1263.2032470703125, 'learning_rate': 1.9089202010465796e-06, 'recon_loss': 0.0015768059500260278, 'recon_static': 0.0005944184304098599, 'recon_motion': 0.0009823875094298272, 'recon_norm': 1.29004348218441, 'recon_norm_std': 0.03928243322297931, 'target_norm': 1.6205617249011994, 'recon_cos': 0.628625912964344, 'gradnorm/compressor.query': 0.432170456435072, 'gradnorm/compressor.layer0': 14.277157253398915, 'gradnorm/compressor.layer1': 1250.9106682733184, 'gradnorm/compressor.layer2': 169.8025335223977, 'gradnorm/compressor.layer3': 41.27199934201759, 'gradnorm/compressor.layer4': 9.939259883106276, 'gradnorm/compressor.layer5': 5.199587057296593, 'gradnorm/compressor.layer6': 3.2420043893604347, 'gradnorm/compressor.layer7': 3.65188089793635, 'gradnorm/decoder.layer0': 3.2137342464338365, 'gradnorm/decoder.layer1': 1.8737893328568778, 'gradnorm/decoder.layer2': 1.0549591844808073, 'gradnorm/max_param': 882.689697265625, 'epoch': 8.87}

 89%|████████▊ | 16940/19090 [8:16:27<1:10:56,  1.98s/it]
 89%|████████▊ | 16941/19090 [8:16:29<1:06:36,  1.86s/it]
 89%|████████▊ | 16942/19090 [8:16:30<1:00:27,  1.69s/it]
 89%|████████▉ | 16943/19090 [8:16:32<1:04:38,  1.81s/it]
 89%|████████▉ | 16944/19090 [8:16:34<1:04:13,  1.80s/it]
 89%|████████▉ | 16945/19090 [8:16:35<58:42,  1.64s/it]  
 89%|████████▉ | 16946/19090 [8:16:37<1:03:49,  1.79s/it]
 89%|████████▉ | 16947/19090 [8:16:39<1:01:52,  1.73s/it]
 89%|████████▉ | 16948/19090 [8:16:40<57:01,  1.60s/it]  
 89%|████████▉ | 16949/19090 [8:16:42<1:05:57,  1.85s/it][gradnorm] step 16949: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 793.1

 89%|████████▉ | 16950/19090 [8:16:44<1:03:29,  1.78s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 1134.2547607421875, 'learning_rate': 1.8914390763018013e-06, 'recon_loss': 0.001592249609529972, 'recon_static': 0.0006188430765178054, 'recon_motion': 0.0009734064995427616, 'recon_norm': 1.3124959349632264, 'recon_norm_std': 0.03882956560701132, 'target_norm': 1.6196911662817002, 'recon_cos': 0.627560418844223, 'gradnorm/compressor.query': 0.4120319167107521, 'gradnorm/compressor.layer0': 12.558565804790058, 'gradnorm/compressor.layer1': 1123.9955515926208, 'gradnorm/compressor.layer2': 146.2732019522886, 'gradnorm/compressor.layer3': 38.27907200436846, 'gradnorm/compressor.layer4': 9.653443555185088, 'gradnorm/compressor.layer5': 4.4289567401812, 'gradnorm/compressor.layer6': 3.0318319804155873, 'gradnorm/compressor.layer7': 3.4997322797563046, 'gradnorm/decoder.layer0': 3.1262230577798875, 'gradnorm/decoder.layer1': 1.925083862063303, 'gradnorm/decoder.layer2': 1.1142500025402786, 'gradnorm/max_param': 793.1387939453125, 'epoch': 8.88}

 89%|████████▉ | 16950/19090 [8:16:44<1:03:29,  1.78s/it]
 89%|████████▉ | 16951/19090 [8:16:45<58:10,  1.63s/it]  
 89%|████████▉ | 16952/19090 [8:16:48<1:04:53,  1.82s/it]
 89%|████████▉ | 16953/19090 [8:16:49<1:02:18,  1.75s/it]
 89%|████████▉ | 16954/19090 [8:16:50<57:22,  1.61s/it]  
 89%|████████▉ | 16955/19090 [8:16:53<1:06:33,  1.87s/it]
 89%|████████▉ | 16956/19090 [8:16:55<1:05:14,  1.83s/it]
 89%|████████▉ | 16957/19090 [8:16:56<59:26,  1.67s/it]  
 89%|████████▉ | 16958/19090 [8:16:58<1:06:37,  1.88s/it]
 89%|████████▉ | 16959/19090 [8:17:00<1:05:14,  1.84s/it][gradnorm] step 16959: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 743.6

 89%|████████▉ | 16960/19090 [8:17:01<59:23,  1.67s/it]  
                                                       
{'loss': 0.0015, 'grad_norm': 1065.563720703125, 'learning_rate': 1.8740352152832986e-06, 'recon_loss': 0.0015135118272155523, 'recon_static': 0.0005336606249329634, 'recon_motion': 0.0009798512212000788, 'recon_norm': 1.2729946792125701, 'recon_norm_std': 0.03941589556634426, 'target_norm': 1.620782932639122, 'recon_cos': 0.6444245159626008, 'gradnorm/compressor.query': 0.41265859661713383, 'gradnorm/compressor.layer0': 7.598637518975056, 'gradnorm/compressor.layer1': 1053.6465251686639, 'gradnorm/compressor.layer2': 153.8325680224282, 'gradnorm/compressor.layer3': 37.36751499307671, 'gradnorm/compressor.layer4': 9.046024895292406, 'gradnorm/compressor.layer5': 4.214121221883572, 'gradnorm/compressor.layer6': 3.4241453677719043, 'gradnorm/compressor.layer7': 3.3202289166869354, 'gradnorm/decoder.layer0': 3.0076363648205087, 'gradnorm/decoder.layer1': 1.9442924674753592, 'gradnorm/decoder.layer2': 1.1865526234997104, 'gradnorm/max_param': 743.562744140625, 'epoch': 8.88}

 89%|████████▉ | 16960/19090 [8:17:01<59:23,  1.67s/it]
 89%|████████▉ | 16961/19090 [8:17:04<1:06:02,  1.86s/it]
 89%|████████▉ | 16962/19090 [8:17:05<1:03:29,  1.79s/it]
 89%|████████▉ | 16963/19090 [8:17:07<58:14,  1.64s/it]  
 89%|████████▉ | 16964/19090 [8:17:09<1:06:31,  1.88s/it]
 89%|████████▉ | 16965/19090 [8:17:11<1:06:18,  1.87s/it]
 89%|████████▉ | 16966/19090 [8:17:12<1:00:02,  1.70s/it]
 89%|████████▉ | 16967/19090 [8:17:14<1:02:03,  1.75s/it]
 89%|████████▉ | 16968/19090 [8:17:16<1:07:46,  1.92s/it]
 89%|████████▉ | 16969/19090 [8:17:18<1:01:02,  1.73s/it][gradnorm] step 16969: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1023.3

 89%|████████▉ | 16970/19090 [8:17:20<1:04:11,  1.82s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1460.0537109375, 'learning_rate': 1.8567086761810847e-06, 'recon_loss': 0.0014527511986671016, 'recon_static': 0.00048483262216905134, 'recon_motion': 0.000967918592505157, 'recon_norm': 1.2520172148942947, 'recon_norm_std': 0.040271548088639975, 'target_norm': 1.6173083424568175, 'recon_cos': 0.6575213477015496, 'gradnorm/compressor.query': 0.6048417567360943, 'gradnorm/compressor.layer0': 22.855536513422873, 'gradnorm/compressor.layer1': 1449.9468093692265, 'gradnorm/compressor.layer2': 163.86026225682357, 'gradnorm/compressor.layer3': 43.33384524312016, 'gradnorm/compressor.layer4': 9.760346229540621, 'gradnorm/compressor.layer5': 5.014582350313064, 'gradnorm/compressor.layer6': 3.1611397513053285, 'gradnorm/compressor.layer7': 3.6512632352647216, 'gradnorm/decoder.layer0': 3.264638359289335, 'gradnorm/decoder.layer1': 2.1115343132582276, 'gradnorm/decoder.layer2': 1.3530277402251538, 'gradnorm/max_param': 1023.2553100585938, 'epoch': 8.89}

 89%|████████▉ | 16970/19090 [8:17:20<1:04:11,  1.82s/it]
 89%|████████▉ | 16971/19090 [8:17:22<1:12:54,  2.06s/it]
 89%|████████▉ | 16972/19090 [8:17:24<1:04:46,  1.83s/it]
 89%|████████▉ | 16973/19090 [8:17:25<59:29,  1.69s/it]  
 89%|████████▉ | 16974/19090 [8:17:27<1:07:36,  1.92s/it]
 89%|████████▉ | 16975/19090 [8:17:29<1:01:05,  1.73s/it]
 89%|████████▉ | 16976/19090 [8:17:30<59:17,  1.68s/it]  
 89%|████████▉ | 16977/19090 [8:17:33<1:06:33,  1.89s/it]
 89%|████████▉ | 16978/19090 [8:17:34<1:00:32,  1.72s/it]
 89%|████████▉ | 16979/19090 [8:17:36<59:53,  1.70s/it]  [gradnorm] step 16979: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 899.1

 89%|████████▉ | 16980/19090 [8:17:38<1:04:29,  1.83s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1283.5550537109375, 'learning_rate': 1.8394595169266337e-06, 'recon_loss': 0.001449740954558365, 'recon_static': 0.00047123460317379797, 'recon_motion': 0.0009785063855815678, 'recon_norm': 1.2429614216089249, 'recon_norm_std': 0.040197024773806336, 'target_norm': 1.6190682768821716, 'recon_cos': 0.6582816690206528, 'gradnorm/compressor.query': 0.5277338633615991, 'gradnorm/compressor.layer0': 17.303783660086545, 'gradnorm/compressor.layer1': 1274.022811805189, 'gradnorm/compressor.layer2': 149.45947532901687, 'gradnorm/compressor.layer3': 39.787158032912295, 'gradnorm/compressor.layer4': 9.490901404643093, 'gradnorm/compressor.layer5': 5.178934770437383, 'gradnorm/compressor.layer6': 3.162201058352752, 'gradnorm/compressor.layer7': 3.8394685796740413, 'gradnorm/decoder.layer0': 3.2976912080221816, 'gradnorm/decoder.layer1': 2.1657973403943926, 'gradnorm/decoder.layer2': 1.2740764770040556, 'gradnorm/max_param': 899.1381225585938, 'epoch': 8.89}

 89%|████████▉ | 16980/19090 [8:17:38<1:04:29,  1.83s/it]
 89%|████████▉ | 16981/19090 [8:17:39<58:52,  1.67s/it]  
 89%|████████▉ | 16982/19090 [8:17:41<1:00:29,  1.72s/it]
 89%|████████▉ | 16983/19090 [8:17:43<1:06:06,  1.88s/it]
 89%|████████▉ | 16984/19090 [8:17:44<59:57,  1.71s/it]  
 89%|████████▉ | 16985/19090 [8:17:46<1:02:03,  1.77s/it]
 89%|████████▉ | 16986/19090 [8:17:48<1:05:19,  1.86s/it]
 89%|████████▉ | 16987/19090 [8:17:50<59:27,  1.70s/it]  
 89%|████████▉ | 16988/19090 [8:17:52<1:01:37,  1.76s/it]
 89%|████████▉ | 16989/19090 [8:17:54<1:02:36,  1.79s/it][gradnorm] step 16989: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 430.6

 89%|████████▉ | 16990/19090 [8:17:55<57:29,  1.64s/it]  
                                                       
{'loss': 0.0015, 'grad_norm': 620.8805541992188, 'learning_rate': 1.8222877951926814e-06, 'recon_loss': 0.0014875038410536944, 'recon_static': 0.000508048283518292, 'recon_motion': 0.0009794555633561685, 'recon_norm': 1.2562348961830139, 'recon_norm_std': 0.040337287914007904, 'target_norm': 1.6209354758262635, 'recon_cos': 0.6501221045851707, 'gradnorm/compressor.query': 0.48642215049627774, 'gradnorm/compressor.layer0': 14.146454038565084, 'gradnorm/compressor.layer1': 610.3397619359237, 'gradnorm/compressor.layer2': 108.21162985528635, 'gradnorm/compressor.layer3': 31.20046171189068, 'gradnorm/compressor.layer4': 6.720270029273929, 'gradnorm/compressor.layer5': 4.166451919107946, 'gradnorm/compressor.layer6': 2.903942709943548, 'gradnorm/compressor.layer7': 3.2930304361993605, 'gradnorm/decoder.layer0': 2.7774606672054847, 'gradnorm/decoder.layer1': 1.9286246519464858, 'gradnorm/decoder.layer2': 1.0859413970218688, 'gradnorm/max_param': 430.5845642089844, 'epoch': 8.9}

 89%|████████▉ | 16990/19090 [8:17:55<57:29,  1.64s/it]
 89%|████████▉ | 16991/19090 [8:17:57<1:03:51,  1.83s/it]
 89%|████████▉ | 16992/19090 [8:17:59<1:02:31,  1.79s/it]
 89%|████████▉ | 16993/19090 [8:18:01<1:04:00,  1.83s/it]
 89%|████████▉ | 16994/19090 [8:18:03<1:04:23,  1.84s/it]
 89%|████████▉ | 16995/19090 [8:18:04<1:01:07,  1.75s/it]
 89%|████████▉ | 16996/19090 [8:18:06<1:02:30,  1.79s/it]
 89%|████████▉ | 16997/19090 [8:18:08<1:03:38,  1.82s/it]
 89%|████████▉ | 16998/19090 [8:18:09<1:00:39,  1.74s/it]
 89%|████████▉ | 16999/19090 [8:18:11<1:03:35,  1.82s/it][gradnorm] step 16999: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 815.5

 89%|████████▉ | 17000/19090 [8:18:13<1:04:03,  1.84s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1166.3936767578125, 'learning_rate': 1.8051935683930877e-06, 'recon_loss': 0.001501470641233027, 'recon_static': 0.000527204638638068, 'recon_motion': 0.000974266008415725, 'recon_norm': 1.2627517372369765, 'recon_norm_std': 0.041101954970508815, 'target_norm': 1.6205458343029022, 'recon_cos': 0.6460860788822174, 'gradnorm/compressor.query': 0.3811171612932799, 'gradnorm/compressor.layer0': 13.920294983467764, 'gradnorm/compressor.layer1': 1155.4909670784969, 'gradnorm/compressor.layer2': 153.52383009243223, 'gradnorm/compressor.layer3': 37.67665746291893, 'gradnorm/compressor.layer4': 8.681750693551415, 'gradnorm/compressor.layer5': 4.85229742162868, 'gradnorm/compressor.layer6': 3.080521788057071, 'gradnorm/compressor.layer7': 3.1990062004904534, 'gradnorm/decoder.layer0': 3.004964376736693, 'gradnorm/decoder.layer1': 1.7017848490451357, 'gradnorm/decoder.layer2': 1.1647668026352653, 'gradnorm/max_param': 815.514892578125, 'epoch': 8.91}

 89%|████████▉ | 17000/19090 [8:18:13<1:04:03,  1.84s/it]
 89%|████████▉ | 17001/19090 [8:18:17<1:27:17,  2.51s/it]
 89%|████████▉ | 17002/19090 [8:18:19<1:14:28,  2.14s/it]
 89%|████████▉ | 17003/19090 [8:18:20<1:05:32,  1.88s/it]
 89%|████████▉ | 17004/19090 [8:18:21<59:15,  1.70s/it]  
 89%|████████▉ | 17005/19090 [8:18:23<54:50,  1.58s/it]
 89%|████████▉ | 17006/19090 [8:18:24<51:45,  1.49s/it]
 89%|████████▉ | 17007/19090 [8:18:25<49:38,  1.43s/it]
 89%|████████▉ | 17008/19090 [8:18:26<48:09,  1.39s/it]
 89%|████████▉ | 17009/19090 [8:18:28<47:04,  1.36s/it][gradnorm] step 17009: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 783.7

 89%|████████▉ | 17010/19090 [8:18:29<46:18,  1.34s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 1121.5523681640625, 'learning_rate': 1.7881768936825805e-06, 'recon_loss': 0.001484432930010371, 'recon_static': 0.0005125390256580431, 'recon_motion': 0.0009718939269077964, 'recon_norm': 1.261800652742386, 'recon_norm_std': 0.04127413053065539, 'target_norm': 1.6215882867574691, 'recon_cos': 0.6515675023198128, 'gradnorm/compressor.query': 0.44353946943561573, 'gradnorm/compressor.layer0': 14.405473071681655, 'gradnorm/compressor.layer1': 1110.6096861634153, 'gradnorm/compressor.layer2': 148.94013583483803, 'gradnorm/compressor.layer3': 43.66841723789059, 'gradnorm/compressor.layer4': 8.532935566373556, 'gradnorm/compressor.layer5': 4.659213563103899, 'gradnorm/compressor.layer6': 2.8483458888753908, 'gradnorm/compressor.layer7': 3.357285954518577, 'gradnorm/decoder.layer0': 3.1176272586401215, 'gradnorm/decoder.layer1': 1.7702262810285898, 'gradnorm/decoder.layer2': 1.1548155056794105, 'gradnorm/max_param': 783.6983032226562, 'epoch': 8.91}

 89%|████████▉ | 17010/19090 [8:18:29<46:18,  1.34s/it]
 89%|████████▉ | 17011/19090 [8:18:30<45:52,  1.32s/it]
 89%|████████▉ | 17012/19090 [8:18:32<54:59,  1.59s/it]
 89%|████████▉ | 17013/19090 [8:18:34<54:07,  1.56s/it]
 89%|████████▉ | 17014/19090 [8:18:35<53:06,  1.54s/it]
 89%|████████▉ | 17015/19090 [8:18:38<1:00:49,  1.76s/it]
 89%|████████▉ | 17016/19090 [8:18:39<59:24,  1.72s/it]  
 89%|████████▉ | 17017/19090 [8:18:41<54:50,  1.59s/it]
 89%|████████▉ | 17018/19090 [8:18:43<1:04:00,  1.85s/it]
 89%|████████▉ | 17019/19090 [8:18:45<1:01:26,  1.78s/it][gradnorm] step 17019: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 306.7

 89%|████████▉ | 17020/19090 [8:18:46<56:17,  1.63s/it]  
                                                       
{'loss': 0.0015, 'grad_norm': 445.32720947265625, 'learning_rate': 1.7712378279565873e-06, 'recon_loss': 0.0014866391953546553, 'recon_static': 0.0005038255469116848, 'recon_motion': 0.0009828136331634595, 'recon_norm': 1.2566632121801375, 'recon_norm_std': 0.040209188032895325, 'target_norm': 1.6217648059129715, 'recon_cos': 0.6510330960154533, 'gradnorm/compressor.query': 0.7769545150905504, 'gradnorm/compressor.layer0': 22.75231116994972, 'gradnorm/compressor.layer1': 435.58461075318075, 'gradnorm/compressor.layer2': 82.60703284224655, 'gradnorm/compressor.layer3': 33.39858643972638, 'gradnorm/compressor.layer4': 8.67440049796085, 'gradnorm/compressor.layer5': 4.414752521426252, 'gradnorm/compressor.layer6': 2.744746044341973, 'gradnorm/compressor.layer7': 3.1689745039102326, 'gradnorm/decoder.layer0': 2.814119508495199, 'gradnorm/decoder.layer1': 1.8044328386043669, 'gradnorm/decoder.layer2': 1.0926011727840883, 'gradnorm/max_param': 306.6925354003906, 'epoch': 8.92}

 89%|████████▉ | 17020/19090 [8:18:46<56:17,  1.63s/it]
 89%|████████▉ | 17021/19090 [8:18:48<1:04:15,  1.86s/it]
 89%|████████▉ | 17022/19090 [8:18:50<1:01:23,  1.78s/it]
 89%|████████▉ | 17023/19090 [8:18:51<56:13,  1.63s/it]  
 89%|████████▉ | 17024/19090 [8:18:54<1:04:34,  1.88s/it]
 89%|████████▉ | 17025/19090 [8:18:55<1:00:34,  1.76s/it]
 89%|████████▉ | 17026/19090 [8:18:57<55:38,  1.62s/it]  
 89%|████████▉ | 17027/19090 [8:18:59<1:04:33,  1.88s/it]
 89%|████████▉ | 17028/19090 [8:19:01<1:03:52,  1.86s/it]
 89%|████████▉ | 17029/19090 [8:19:02<57:56,  1.69s/it]  [gradnorm] step 17029: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 475.0

 89%|████████▉ | 17030/19090 [8:19:04<1:02:54,  1.83s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 683.0035400390625, 'learning_rate': 1.7543764278510706e-06, 'recon_loss': 0.0015274156001396477, 'recon_static': 0.000548855262604775, 'recon_motion': 0.000978560333896894, 'recon_norm': 1.2766649186611176, 'recon_norm_std': 0.04039426259696484, 'target_norm': 1.6211520671844482, 'recon_cos': 0.6414575934410095, 'gradnorm/compressor.query': 0.46075317770741303, 'gradnorm/compressor.layer0': 17.54345887271806, 'gradnorm/compressor.layer1': 673.2089896532874, 'gradnorm/compressor.layer2': 108.0561467912631, 'gradnorm/compressor.layer3': 34.243867422571995, 'gradnorm/compressor.layer4': 8.875224795651086, 'gradnorm/compressor.layer5': 4.149524266195125, 'gradnorm/compressor.layer6': 2.6397533896888974, 'gradnorm/compressor.layer7': 3.322945159184067, 'gradnorm/decoder.layer0': 2.854972911111619, 'gradnorm/decoder.layer1': 1.8286218212626015, 'gradnorm/decoder.layer2': 1.1499552137527291, 'gradnorm/max_param': 474.9959716796875, 'epoch': 8.92}

 89%|████████▉ | 17030/19090 [8:19:04<1:02:54,  1.83s/it]
 89%|████████▉ | 17031/19090 [8:19:06<1:05:55,  1.92s/it]
 89%|████████▉ | 17032/19090 [8:19:08<59:20,  1.73s/it]  
 89%|████████▉ | 17033/19090 [8:19:10<1:00:52,  1.78s/it]
 89%|████████▉ | 17034/19090 [8:19:12<1:05:48,  1.92s/it]
 89%|████████▉ | 17035/19090 [8:19:13<59:14,  1.73s/it]  
 89%|████████▉ | 17036/19090 [8:19:15<58:43,  1.72s/it]
 89%|████████▉ | 17037/19090 [8:19:17<1:05:29,  1.91s/it]
 89%|████████▉ | 17038/19090 [8:19:18<59:09,  1.73s/it]  
 89%|████████▉ | 17039/19090 [8:19:20<59:13,  1.73s/it][gradnorm] step 17039: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 593.4

 89%|████████▉ | 17040/19090 [8:19:23<1:06:05,  1.93s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 847.1200561523438, 'learning_rate': 1.737592749742309e-06, 'recon_loss': 0.001662430164287798, 'recon_static': 0.0006965404289076104, 'recon_motion': 0.000965889731014613, 'recon_norm': 1.3385448783636094, 'recon_norm_std': 0.03899468509480357, 'target_norm': 1.6198562026023864, 'recon_cos': 0.6120001927018166, 'gradnorm/compressor.query': 0.5061926314732882, 'gradnorm/compressor.layer0': 16.120017058029806, 'gradnorm/compressor.layer1': 841.0561440831402, 'gradnorm/compressor.layer2': 91.63501288229025, 'gradnorm/compressor.layer3': 37.81034181599245, 'gradnorm/compressor.layer4': 9.631679842675258, 'gradnorm/compressor.layer5': 4.347650546801066, 'gradnorm/compressor.layer6': 2.964864979543559, 'gradnorm/compressor.layer7': 3.685107085122602, 'gradnorm/decoder.layer0': 3.2063652211208535, 'gradnorm/decoder.layer1': 2.0222982491301966, 'gradnorm/decoder.layer2': 1.3187406313595036, 'gradnorm/max_param': 593.4114379882812, 'epoch': 8.93}

 89%|████████▉ | 17040/19090 [8:19:23<1:06:05,  1.93s/it]
 89%|████████▉ | 17041/19090 [8:19:24<59:29,  1.74s/it]  
 89%|████████▉ | 17042/19090 [8:19:26<58:34,  1.72s/it]
 89%|████████▉ | 17043/19090 [8:19:28<1:07:51,  1.99s/it]
 89%|████████▉ | 17044/19090 [8:19:29<1:00:40,  1.78s/it]
 89%|████████▉ | 17045/19090 [8:19:31<55:39,  1.63s/it]  
 89%|████████▉ | 17046/19090 [8:19:33<1:04:16,  1.89s/it]
 89%|████████▉ | 17047/19090 [8:19:35<58:07,  1.71s/it]  
 89%|████████▉ | 17048/19090 [8:19:36<57:08,  1.68s/it]
 89%|████████▉ | 17049/19090 [8:19:39<1:07:20,  1.98s/it][gradnorm] step 17049: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 749.1

 89%|████████▉ | 17050/19090 [8:19:40<1:00:12,  1.77s/it]
                                                         
{'loss': 0.0017, 'grad_norm': 1072.3822021484375, 'learning_rate': 1.720886849746703e-06, 'recon_loss': 0.0017222157795913518, 'recon_static': 0.0007517687554354779, 'recon_motion': 0.0009704469775897451, 'recon_norm': 1.362240931391716, 'recon_norm_std': 0.03911685841158032, 'target_norm': 1.6198956489562988, 'recon_cos': 0.6010044366121292, 'gradnorm/compressor.query': 0.4852988753706051, 'gradnorm/compressor.layer0': 13.329560572599298, 'gradnorm/compressor.layer1': 1061.5107747922298, 'gradnorm/compressor.layer2': 146.39542049061848, 'gradnorm/compressor.layer3': 38.15077253886872, 'gradnorm/compressor.layer4': 8.358601471950587, 'gradnorm/compressor.layer5': 5.164887562919222, 'gradnorm/compressor.layer6': 3.2019010857214405, 'gradnorm/compressor.layer7': 3.5592940779304705, 'gradnorm/decoder.layer0': 3.076539536560441, 'gradnorm/decoder.layer1': 1.8711812231844807, 'gradnorm/decoder.layer2': 1.080025863072876, 'gradnorm/max_param': 749.1433715820312, 'epoch': 8.93}

 89%|████████▉ | 17050/19090 [8:19:40<1:00:12,  1.77s/it]
 89%|████████▉ | 17051/19090 [8:19:41<55:36,  1.64s/it]  
 89%|████████▉ | 17052/19090 [8:19:44<1:09:18,  2.04s/it]
 89%|████████▉ | 17053/19090 [8:19:46<1:01:36,  1.81s/it]
 89%|████████▉ | 17054/19090 [8:19:47<56:09,  1.65s/it]  
 89%|████████▉ | 17055/19090 [8:19:50<1:05:04,  1.92s/it]
 89%|████████▉ | 17056/19090 [8:19:51<58:33,  1.73s/it]  
 89%|████████▉ | 17057/19090 [8:19:52<54:36,  1.61s/it]
 89%|████████▉ | 17058/19090 [8:19:55<1:04:42,  1.91s/it]
 89%|████████▉ | 17059/19090 [8:19:56<58:18,  1.72s/it]  [gradnorm] step 17059: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 705.6

 89%|████████▉ | 17060/19090 [8:19:57<53:50,  1.59s/it]
                                                       
{'loss': 0.0016, 'grad_norm': 1011.1927490234375, 'learning_rate': 1.7042587837206192e-06, 'recon_loss': 0.0015855597303016111, 'recon_static': 0.0006256846754695289, 'recon_motion': 0.0009598750286386349, 'recon_norm': 1.3185924500226975, 'recon_norm_std': 0.03892828384414315, 'target_norm': 1.6165258288383484, 'recon_cos': 0.6282525897026062, 'gradnorm/compressor.query': 0.44001246142080674, 'gradnorm/compressor.layer0': 13.469503925933656, 'gradnorm/compressor.layer1': 999.665225212921, 'gradnorm/compressor.layer2': 147.33834494328013, 'gradnorm/compressor.layer3': 33.7870221566053, 'gradnorm/compressor.layer4': 9.316287096329228, 'gradnorm/compressor.layer5': 5.13888492526798, 'gradnorm/compressor.layer6': 2.9953492991654964, 'gradnorm/compressor.layer7': 3.1812344473887695, 'gradnorm/decoder.layer0': 3.414237639869115, 'gradnorm/decoder.layer1': 1.8108563698738291, 'gradnorm/decoder.layer2': 1.0996983764839035, 'gradnorm/max_param': 705.5531005859375, 'epoch': 8.94}

 89%|████████▉ | 17060/19090 [8:19:57<53:50,  1.59s/it]
 89%|████████▉ | 17061/19090 [8:20:00<1:07:29,  2.00s/it]
 89%|████████▉ | 17062/19090 [8:20:02<1:00:13,  1.78s/it]
 89%|████████▉ | 17063/19090 [8:20:03<55:06,  1.63s/it]  
 89%|████████▉ | 17064/19090 [8:20:05<1:05:21,  1.94s/it]
 89%|████████▉ | 17065/19090 [8:20:07<59:36,  1.77s/it]  
 89%|████████▉ | 17066/19090 [8:20:08<54:40,  1.62s/it]
 89%|████████▉ | 17067/19090 [8:20:11<1:04:53,  1.92s/it]
 89%|████████▉ | 17068/19090 [8:20:12<58:33,  1.74s/it]  
 89%|████████▉ | 17069/19090 [8:20:14<57:45,  1.71s/it][gradnorm] step 17069: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 593.6

 89%|████████▉ | 17070/19090 [8:20:16<1:02:15,  1.85s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 850.750244140625, 'learning_rate': 1.68770860726018e-06, 'recon_loss': 0.0015209062199573963, 'recon_static': 0.0005499263294041157, 'recon_motion': 0.0009709798832773231, 'recon_norm': 1.2780809700489044, 'recon_norm_std': 0.039797034300863746, 'target_norm': 1.619863787293434, 'recon_cos': 0.6422848418354988, 'gradnorm/compressor.query': 0.364427746771299, 'gradnorm/compressor.layer0': 10.079998463221841, 'gradnorm/compressor.layer1': 841.1309351105807, 'gradnorm/compressor.layer2': 120.82978487965002, 'gradnorm/compressor.layer3': 37.81510515302549, 'gradnorm/compressor.layer4': 8.980289917069989, 'gradnorm/compressor.layer5': 5.264360496757354, 'gradnorm/compressor.layer6': 2.9495351457456427, 'gradnorm/compressor.layer7': 3.2310641384718664, 'gradnorm/decoder.layer0': 3.348724515276141, 'gradnorm/decoder.layer1': 1.8521694041771717, 'gradnorm/decoder.layer2': 0.9965886999390274, 'gradnorm/max_param': 593.6044311523438, 'epoch': 8.94}

 89%|████████▉ | 17070/19090 [8:20:16<1:02:15,  1.85s/it]
 89%|████████▉ | 17071/19090 [8:20:17<59:43,  1.77s/it]  
 89%|████████▉ | 17072/19090 [8:20:19<59:31,  1.77s/it]
 89%|████████▉ | 17073/19090 [8:20:22<1:04:41,  1.92s/it]
 89%|████████▉ | 17074/19090 [8:20:23<1:01:35,  1.83s/it]
 89%|████████▉ | 17075/19090 [8:20:25<58:06,  1.73s/it]  
 89%|████████▉ | 17076/19090 [8:20:27<1:04:38,  1.93s/it]
 89%|████████▉ | 17077/19090 [8:20:29<1:01:39,  1.84s/it]
 89%|████████▉ | 17078/19090 [8:20:31<1:07:43,  2.02s/it]
 89%|████████▉ | 17079/19090 [8:20:32<1:00:15,  1.80s/it][gradnorm] step 17079: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 529.4

 89%|████████▉ | 17080/19090 [8:20:35<1:09:11,  2.07s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 763.431396484375, 'learning_rate': 1.6712363757010713e-06, 'recon_loss': 0.0015203869319520892, 'recon_static': 0.0005404473587987014, 'recon_motion': 0.000979939526587259, 'recon_norm': 1.2736756563186646, 'recon_norm_std': 0.04043737836182117, 'target_norm': 1.6226553708314895, 'recon_cos': 0.6436406642198562, 'gradnorm/compressor.query': 0.7562167089401552, 'gradnorm/compressor.layer0': 29.060189760327617, 'gradnorm/compressor.layer1': 750.6808159930557, 'gradnorm/compressor.layer2': 128.8391605719123, 'gradnorm/compressor.layer3': 41.347118745030166, 'gradnorm/compressor.layer4': 9.937176009360789, 'gradnorm/compressor.layer5': 4.8168951004577085, 'gradnorm/compressor.layer6': 2.5821418680800807, 'gradnorm/compressor.layer7': 2.863507562413195, 'gradnorm/decoder.layer0': 3.1160031317816337, 'gradnorm/decoder.layer1': 1.9359512907434488, 'gradnorm/decoder.layer2': 1.1315381478736737, 'gradnorm/max_param': 529.373779296875, 'epoch': 8.95}

 89%|████████▉ | 17080/19090 [8:20:35<1:09:11,  2.07s/it]
 89%|████████▉ | 17081/19090 [8:20:36<1:01:16,  1.83s/it]
 89%|████████▉ | 17082/19090 [8:20:38<55:47,  1.67s/it]  
 89%|████████▉ | 17083/19090 [8:20:39<53:38,  1.60s/it]
 89%|████████▉ | 17084/19090 [8:20:41<56:10,  1.68s/it]
 89%|████████▉ | 17085/19090 [8:20:43<1:00:13,  1.80s/it]
 90%|████████▉ | 17086/19090 [8:20:44<55:55,  1.67s/it]  
 90%|████████▉ | 17087/19090 [8:20:46<55:40,  1.67s/it]
 90%|████████▉ | 17088/19090 [8:20:48<59:41,  1.79s/it]
 90%|████████▉ | 17089/19090 [8:20:50<57:25,  1.72s/it][gradnorm] step 17089: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 741.5

 90%|████████▉ | 17090/19090 [8:20:52<58:57,  1.77s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 1065.6396484375, 'learning_rate': 1.6548421441183875e-06, 'recon_loss': 0.0015369521803222596, 'recon_static': 0.0005572863199631683, 'recon_motion': 0.0009796658909181134, 'recon_norm': 1.2849628239870072, 'recon_norm_std': 0.039715620782226325, 'target_norm': 1.6209297686815263, 'recon_cos': 0.6394927874207497, 'gradnorm/compressor.query': 0.5433291330155666, 'gradnorm/compressor.layer0': 16.342288825398185, 'gradnorm/compressor.layer1': 1050.7313524398137, 'gradnorm/compressor.layer2': 167.1854497537839, 'gradnorm/compressor.layer3': 56.49753909532808, 'gradnorm/compressor.layer4': 9.280633462792458, 'gradnorm/compressor.layer5': 4.748904051585177, 'gradnorm/compressor.layer6': 2.824652378671268, 'gradnorm/compressor.layer7': 3.0950146797090876, 'gradnorm/decoder.layer0': 3.104563608754967, 'gradnorm/decoder.layer1': 1.9214661326107576, 'gradnorm/decoder.layer2': 1.2184720333710917, 'gradnorm/max_param': 741.51220703125, 'epoch': 8.95}

 90%|████████▉ | 17090/19090 [8:20:52<58:57,  1.77s/it]
 90%|████████▉ | 17091/19090 [8:20:53<58:52,  1.77s/it]
 90%|████████▉ | 17092/19090 [8:20:55<56:23,  1.69s/it]
 90%|████████▉ | 17093/19090 [8:20:57<1:01:08,  1.84s/it]
 90%|████████▉ | 17094/19090 [8:20:59<1:00:44,  1.83s/it]
 90%|████████▉ | 17095/19090 [8:21:01<1:00:56,  1.83s/it]
 90%|████████▉ | 17096/19090 [8:21:02<1:00:23,  1.82s/it]
 90%|████████▉ | 17097/19090 [8:21:04<56:42,  1.71s/it]  
 90%|████████▉ | 17098/19090 [8:21:06<1:04:38,  1.95s/it]
 90%|████████▉ | 17099/19090 [8:21:08<1:01:08,  1.84s/it][gradnorm] step 17099: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 618.2

 90%|████████▉ | 17100/19090 [8:21:09<57:11,  1.72s/it]  
                                                       
{'loss': 0.0016, 'grad_norm': 886.6556396484375, 'learning_rate': 1.6385259673264263e-06, 'recon_loss': 0.0015960094198817388, 'recon_static': 0.0006178651907248423, 'recon_motion': 0.0009781442757230252, 'recon_norm': 1.3072620153427124, 'recon_norm_std': 0.038965577445924285, 'target_norm': 1.6195795327425002, 'recon_cos': 0.6254285171627998, 'gradnorm/compressor.query': 0.4418968917193516, 'gradnorm/compressor.layer0': 9.798669094097678, 'gradnorm/compressor.layer1': 875.9703904813222, 'gradnorm/compressor.layer2': 132.8159778956395, 'gradnorm/compressor.layer3': 31.078941000788117, 'gradnorm/compressor.layer4': 9.011743724905791, 'gradnorm/compressor.layer5': 4.353004093503879, 'gradnorm/compressor.layer6': 2.7135714616561417, 'gradnorm/compressor.layer7': 3.0291945577296784, 'gradnorm/decoder.layer0': 3.197874799986572, 'gradnorm/decoder.layer1': 1.9410328803444121, 'gradnorm/decoder.layer2': 1.1059604265779834, 'gradnorm/max_param': 618.1680297851562, 'epoch': 8.96}

 90%|████████▉ | 17100/19090 [8:21:09<57:11,  1.72s/it]
 90%|████████▉ | 17101/19090 [8:21:12<1:02:10,  1.88s/it]
 90%|████████▉ | 17102/19090 [8:21:13<59:15,  1.79s/it]  
 90%|████████▉ | 17103/19090 [8:21:15<56:39,  1.71s/it]
 90%|████████▉ | 17104/19090 [8:21:17<1:01:29,  1.86s/it]
 90%|████████▉ | 17105/19090 [8:21:19<58:17,  1.76s/it]  
 90%|████████▉ | 17106/19090 [8:21:20<54:25,  1.65s/it]
 90%|████████▉ | 17107/19090 [8:21:22<1:01:16,  1.85s/it]
 90%|████████▉ | 17108/19090 [8:21:24<1:00:56,  1.85s/it]
 90%|████████▉ | 17109/19090 [8:21:25<55:21,  1.68s/it]  [gradnorm] step 17109: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 648.1

 90%|████████▉ | 17110/19090 [8:21:28<1:01:57,  1.88s/it]
                                                         
{'loss': 0.0016, 'grad_norm': 927.0941162109375, 'learning_rate': 1.6222878998784913e-06, 'recon_loss': 0.0015533609956037253, 'recon_static': 0.000581721737398766, 'recon_motion': 0.0009716393105918542, 'recon_norm': 1.2841688632965087, 'recon_norm_std': 0.0393754524178803, 'target_norm': 1.6199918925762176, 'recon_cos': 0.6339426204562187, 'gradnorm/compressor.query': 0.3115325017584985, 'gradnorm/compressor.layer0': 9.325366661901624, 'gradnorm/compressor.layer1': 918.2588142784147, 'gradnorm/compressor.layer2': 122.2789659255119, 'gradnorm/compressor.layer3': 34.13035099987986, 'gradnorm/compressor.layer4': 7.326451482600857, 'gradnorm/compressor.layer5': 4.369117269508989, 'gradnorm/compressor.layer6': 2.5308959148140238, 'gradnorm/compressor.layer7': 2.706226399871611, 'gradnorm/decoder.layer0': 3.0565933565135093, 'gradnorm/decoder.layer1': 1.833371581054296, 'gradnorm/decoder.layer2': 0.9919469947382159, 'gradnorm/max_param': 648.0587768554688, 'epoch': 8.96}

 90%|████████▉ | 17110/19090 [8:21:28<1:01:57,  1.88s/it]
 90%|████████▉ | 17111/19090 [8:21:30<1:01:20,  1.86s/it]
 90%|████████▉ | 17112/19090 [8:21:31<55:36,  1.69s/it]  
 90%|████████▉ | 17113/19090 [8:21:33<1:03:05,  1.91s/it]
 90%|████████▉ | 17114/19090 [8:21:35<58:51,  1.79s/it]  
 90%|████████▉ | 17115/19090 [8:21:36<57:11,  1.74s/it]
 90%|████████▉ | 17116/19090 [8:21:38<1:00:48,  1.85s/it]
 90%|████████▉ | 17117/19090 [8:21:40<58:11,  1.77s/it]  
 90%|████████▉ | 17118/19090 [8:21:42<55:52,  1.70s/it]
 90%|████████▉ | 17119/19090 [8:21:44<1:01:59,  1.89s/it][gradnorm] step 17119: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 632.5

 90%|████████▉ | 17120/19090 [8:21:46<1:01:25,  1.87s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 907.60888671875, 'learning_rate': 1.6061279960667446e-06, 'recon_loss': 0.0015002989064669236, 'recon_static': 0.0005214075463300105, 'recon_motion': 0.0009788913477677852, 'recon_norm': 1.2729896903038025, 'recon_norm_std': 0.039043949265033004, 'target_norm': 1.61911284327507, 'recon_cos': 0.6472270205616951, 'gradnorm/compressor.query': 0.38181712696932946, 'gradnorm/compressor.layer0': 13.234664967860859, 'gradnorm/compressor.layer1': 896.3070762858006, 'gradnorm/compressor.layer2': 137.62959801583378, 'gradnorm/compressor.layer3': 33.837677274309364, 'gradnorm/compressor.layer4': 8.40847102324379, 'gradnorm/compressor.layer5': 5.031161479289695, 'gradnorm/compressor.layer6': 2.483381156546245, 'gradnorm/compressor.layer7': 2.878241660273098, 'gradnorm/decoder.layer0': 3.2722884098246525, 'gradnorm/decoder.layer1': 1.7321850121246425, 'gradnorm/decoder.layer2': 1.071189298479015, 'gradnorm/max_param': 632.488525390625, 'epoch': 8.97}

 90%|████████▉ | 17120/19090 [8:21:46<1:01:25,  1.87s/it]
 90%|████████▉ | 17121/19090 [8:21:47<56:11,  1.71s/it]  
 90%|████████▉ | 17122/19090 [8:21:49<1:02:34,  1.91s/it]
 90%|████████▉ | 17123/19090 [8:21:51<57:37,  1.76s/it]  
 90%|████████▉ | 17124/19090 [8:21:53<58:12,  1.78s/it]
 90%|████████▉ | 17125/19090 [8:21:55<1:02:00,  1.89s/it]
 90%|████████▉ | 17126/19090 [8:21:56<56:36,  1.73s/it]  
 90%|████████▉ | 17127/19090 [8:21:58<56:16,  1.72s/it]
 90%|████████▉ | 17128/19090 [8:22:00<1:01:24,  1.88s/it]
 90%|████████▉ | 17129/19090 [8:22:02<56:34,  1.73s/it]  [gradnorm] step 17129: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 604.8

 90%|████████▉ | 17130/19090 [8:22:03<57:41,  1.77s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 865.598876953125, 'learning_rate': 1.590046309922008e-06, 'recon_loss': 0.0014568394544767216, 'recon_static': 0.0004836378284380771, 'recon_motion': 0.0009732016464113258, 'recon_norm': 1.2477747261524201, 'recon_norm_std': 0.04025638662278652, 'target_norm': 1.6184512764215468, 'recon_cos': 0.6566416621208191, 'gradnorm/compressor.query': 0.39521002213930506, 'gradnorm/compressor.layer0': 13.153400030906338, 'gradnorm/compressor.layer1': 857.0696441946827, 'gradnorm/compressor.layer2': 114.57390491512454, 'gradnorm/compressor.layer3': 35.36578874663868, 'gradnorm/compressor.layer4': 9.315639709031075, 'gradnorm/compressor.layer5': 5.115398810716109, 'gradnorm/compressor.layer6': 2.525714520116572, 'gradnorm/compressor.layer7': 2.942427540213478, 'gradnorm/decoder.layer0': 3.0917842283972243, 'gradnorm/decoder.layer1': 1.769844684252017, 'gradnorm/decoder.layer2': 1.1158318998248289, 'gradnorm/max_param': 604.7781372070312, 'epoch': 8.97}

 90%|████████▉ | 17130/19090 [8:22:03<57:41,  1.77s/it]
 90%|████████▉ | 17131/19090 [8:22:05<1:00:50,  1.86s/it]
 90%|████████▉ | 17132/19090 [8:22:07<59:00,  1.81s/it]  
 90%|████████▉ | 17133/19090 [8:22:09<57:49,  1.77s/it]
 90%|████████▉ | 17134/19090 [8:22:11<59:59,  1.84s/it]
 90%|████████▉ | 17135/19090 [8:22:13<58:53,  1.81s/it]
 90%|████████▉ | 17136/19090 [8:22:14<56:43,  1.74s/it]
 90%|████████▉ | 17137/19090 [8:22:17<1:03:04,  1.94s/it]
 90%|████████▉ | 17138/19090 [8:22:18<1:00:05,  1.85s/it]
 90%|████████▉ | 17139/19090 [8:22:19<54:36,  1.68s/it]  [gradnorm] step 17139: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 538.9

 90%|████████▉ | 17140/19090 [8:22:22<1:01:47,  1.90s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 774.9508056640625, 'learning_rate': 1.5740428952135594e-06, 'recon_loss': 0.0014909549208823591, 'recon_static': 0.0005174169040401466, 'recon_motion': 0.000973538015387021, 'recon_norm': 1.2635935485363006, 'recon_norm_std': 0.04084769757464528, 'target_norm': 1.621084028482437, 'recon_cos': 0.6498454481363296, 'gradnorm/compressor.query': 0.32036513384579113, 'gradnorm/compressor.layer0': 10.097745028018826, 'gradnorm/compressor.layer1': 763.6647251903155, 'gradnorm/compressor.layer2': 125.7055556399557, 'gradnorm/compressor.layer3': 36.24394143875118, 'gradnorm/compressor.layer4': 9.424880729750353, 'gradnorm/compressor.layer5': 5.205224882136064, 'gradnorm/compressor.layer6': 2.5006761590195596, 'gradnorm/compressor.layer7': 3.0340765919281, 'gradnorm/decoder.layer0': 3.377207846270949, 'gradnorm/decoder.layer1': 1.8638409107941918, 'gradnorm/decoder.layer2': 1.0335404362844536, 'gradnorm/max_param': 538.9313354492188, 'epoch': 8.98}

 90%|████████▉ | 17140/19090 [8:22:22<1:01:47,  1.90s/it]
 90%|████████▉ | 17141/19090 [8:22:24<59:26,  1.83s/it]  
 90%|████████▉ | 17142/19090 [8:22:25<54:08,  1.67s/it]
 90%|████████▉ | 17143/19090 [8:22:27<1:01:11,  1.89s/it]
 90%|████████▉ | 17144/19090 [8:22:29<56:34,  1.74s/it]  
 90%|████████▉ | 17145/19090 [8:22:30<52:04,  1.61s/it]
 90%|████████▉ | 17146/19090 [8:22:33<1:02:26,  1.93s/it]
 90%|████████▉ | 17147/19090 [8:22:34<58:23,  1.80s/it]  
 90%|████████▉ | 17148/19090 [8:22:35<53:20,  1.65s/it]
 90%|████████▉ | 17149/19090 [8:22:38<1:04:50,  2.00s/it][gradnorm] step 17149: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 709.4

 90%|████████▉ | 17150/19090 [8:22:40<57:53,  1.79s/it]  
                                                       
{'loss': 0.0015, 'grad_norm': 1015.823486328125, 'learning_rate': 1.5581178054490025e-06, 'recon_loss': 0.0014850450650556013, 'recon_static': 0.000506471136759501, 'recon_motion': 0.000978573938482441, 'recon_norm': 1.2654904574155807, 'recon_norm_std': 0.04009106494486332, 'target_norm': 1.6209419578313828, 'recon_cos': 0.6517827659845352, 'gradnorm/compressor.query': 0.40364301947156395, 'gradnorm/compressor.layer0': 12.432546271263558, 'gradnorm/compressor.layer1': 1005.3292060315367, 'gradnorm/compressor.layer2': 138.83594607260397, 'gradnorm/compressor.layer3': 40.75053620717002, 'gradnorm/compressor.layer4': 7.956751268392792, 'gradnorm/compressor.layer5': 5.040360820206821, 'gradnorm/compressor.layer6': 2.763647115751505, 'gradnorm/compressor.layer7': 2.9749639428782633, 'gradnorm/decoder.layer0': 3.160988753427002, 'gradnorm/decoder.layer1': 1.8798706371358969, 'gradnorm/decoder.layer2': 0.9899449273406478, 'gradnorm/max_param': 709.400390625, 'epoch': 8.98}

 90%|████████▉ | 17150/19090 [8:22:40<57:53,  1.79s/it]
 90%|████████▉ | 17151/19090 [8:22:41<54:45,  1.69s/it]
 90%|████████▉ | 17152/19090 [8:22:44<1:08:33,  2.12s/it]
 90%|████████▉ | 17153/19090 [8:22:45<1:00:29,  1.87s/it]
 90%|████████▉ | 17154/19090 [8:22:47<54:49,  1.70s/it]  
 90%|████████▉ | 17155/19090 [8:22:49<1:05:05,  2.02s/it]
 90%|████████▉ | 17156/19090 [8:22:51<58:01,  1.80s/it]  
 90%|████████▉ | 17157/19090 [8:22:52<53:06,  1.65s/it]
 90%|████████▉ | 17158/19090 [8:22:54<1:00:26,  1.88s/it]
 90%|████████▉ | 17159/19090 [8:22:56<56:03,  1.74s/it]  [gradnorm] step 17159: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 510.4

 90%|████████▉ | 17160/19090 [8:22:57<54:01,  1.68s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 734.1904296875, 'learning_rate': 1.5422710938740536e-06, 'recon_loss': 0.0014606655749958009, 'recon_static': 0.00048461415935889817, 'recon_motion': 0.0009760514294612221, 'recon_norm': 1.2422062844038009, 'recon_norm_std': 0.04015912301838398, 'target_norm': 1.618778157234192, 'recon_cos': 0.6548681020736694, 'gradnorm/compressor.query': 0.384244536922538, 'gradnorm/compressor.layer0': 12.716494348507572, 'gradnorm/compressor.layer1': 723.3692867408735, 'gradnorm/compressor.layer2': 120.0704383112263, 'gradnorm/compressor.layer3': 32.36320557663151, 'gradnorm/compressor.layer4': 9.391283872216551, 'gradnorm/compressor.layer5': 4.912679359326113, 'gradnorm/compressor.layer6': 2.824664195534648, 'gradnorm/compressor.layer7': 3.1331035069798157, 'gradnorm/decoder.layer0': 3.318404926528511, 'gradnorm/decoder.layer1': 2.014927351519919, 'gradnorm/decoder.layer2': 1.083448837920732, 'gradnorm/max_param': 510.4058532714844, 'epoch': 8.99}

 90%|████████▉ | 17160/19090 [8:22:57<54:01,  1.68s/it]
 90%|████████▉ | 17161/19090 [8:23:00<1:03:03,  1.96s/it]
 90%|████████▉ | 17162/19090 [8:23:01<56:29,  1.76s/it]  
 90%|████████▉ | 17163/19090 [8:23:03<53:37,  1.67s/it]
 90%|████████▉ | 17164/19090 [8:23:06<1:05:31,  2.04s/it]
 90%|████████▉ | 17165/19090 [8:23:07<58:12,  1.81s/it]  
 90%|████████▉ | 17166/19090 [8:23:08<53:06,  1.66s/it]
 90%|████████▉ | 17167/19090 [8:23:11<1:05:44,  2.05s/it]
 90%|████████▉ | 17168/19090 [8:23:13<58:18,  1.82s/it]  
 90%|████████▉ | 17169/19090 [8:23:14<53:07,  1.66s/it][gradnorm] step 17169: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 714.1

 90%|████████▉ | 17170/19090 [8:23:16<1:02:31,  1.95s/it]
                                                         
{'loss': 0.0015, 'grad_norm': 1023.3076782226562, 'learning_rate': 1.5265028134723497e-06, 'recon_loss': 0.0014649807795649394, 'recon_static': 0.0004901502616121434, 'recon_motion': 0.0009748304859385825, 'recon_norm': 1.2523604393005372, 'recon_norm_std': 0.04047695966437459, 'target_norm': 1.6210041135549544, 'recon_cos': 0.6559204280376434, 'gradnorm/compressor.query': 0.3173210318663554, 'gradnorm/compressor.layer0': 9.945004107060646, 'gradnorm/compressor.layer1': 1011.7925306108956, 'gradnorm/compressor.layer2': 145.19403773485328, 'gradnorm/compressor.layer3': 45.823255580034086, 'gradnorm/compressor.layer4': 9.68181339838526, 'gradnorm/compressor.layer5': 5.3026235888519215, 'gradnorm/compressor.layer6': 2.598766760687199, 'gradnorm/compressor.layer7': 3.1401432104314995, 'gradnorm/decoder.layer0': 3.23462237457913, 'gradnorm/decoder.layer1': 2.0344818227822645, 'gradnorm/decoder.layer2': 1.1203800282151102, 'gradnorm/max_param': 714.069091796875, 'epoch': 8.99}

 90%|████████▉ | 17170/19090 [8:23:16<1:02:31,  1.95s/it]
 90%|████████▉ | 17171/19090 [8:23:18<56:05,  1.75s/it]  
 90%|████████▉ | 17172/19090 [8:23:19<55:37,  1.74s/it]
 90%|████████▉ | 17173/19090 [8:23:22<59:46,  1.87s/it]
 90%|████████▉ | 17174/19090 [8:23:23<58:23,  1.83s/it]
 90%|████████▉ | 17175/19090 [8:23:25<56:54,  1.78s/it]
 90%|████████▉ | 17176/19090 [8:23:27<59:54,  1.88s/it]
 90%|████████▉ | 17177/19090 [8:23:28<54:12,  1.70s/it]
 90%|████████▉ | 17178/19090 [8:23:30<50:09,  1.57s/it]
 90%|████████▉ | 17179/19090 [8:23:31<47:12,  1.48s/it][gradnorm] step 17179: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 611.7

 90%|████████▉ | 17180/19090 [8:23:32<45:11,  1.42s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 877.490234375, 'learning_rate': 1.5108130169653234e-06, 'recon_loss': 0.0014630110148573295, 'recon_static': 0.00047964478726498784, 'recon_motion': 0.000983366217406001, 'recon_norm': 1.243884590268135, 'recon_norm_std': 0.04080806486308575, 'target_norm': 1.6230308949947356, 'recon_cos': 0.6569745004177093, 'gradnorm/compressor.query': 0.3079110694530531, 'gradnorm/compressor.layer0': 8.831981579498763, 'gradnorm/compressor.layer1': 866.6288853944345, 'gradnorm/compressor.layer2': 130.36004827975478, 'gradnorm/compressor.layer3': 41.58860200110422, 'gradnorm/compressor.layer4': 9.039846163765773, 'gradnorm/compressor.layer5': 5.14147997764006, 'gradnorm/compressor.layer6': 2.76158562626684, 'gradnorm/compressor.layer7': 3.1656803468947747, 'gradnorm/decoder.layer0': 3.4311985866676245, 'gradnorm/decoder.layer1': 1.8544351690525265, 'gradnorm/decoder.layer2': 1.0766672423944814, 'gradnorm/max_param': 611.6571044921875, 'epoch': 9.0}

 90%|████████▉ | 17180/19090 [8:23:32<45:11,  1.42s/it]
 90%|█████████ | 17181/19090 [8:23:34<47:53,  1.51s/it]
 90%|█████████ | 17182/19090 [8:23:42<1:51:45,  3.51s/it]
 90%|█████████ | 17183/19090 [8:23:43<1:30:28,  2.85s/it]
 90%|█████████ | 17184/19090 [8:23:47<1:33:19,  2.94s/it]
 90%|█████████ | 17185/19090 [8:23:48<1:17:38,  2.45s/it]
 90%|█████████ | 17186/19090 [8:23:49<1:06:35,  2.10s/it]
 90%|█████████ | 17187/19090 [8:23:51<1:06:23,  2.09s/it]
 90%|█████████ | 17188/19090 [8:23:53<58:43,  1.85s/it]  
 90%|█████████ | 17189/19090 [8:23:54<53:20,  1.68s/it][gradnorm] step 17189: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 768.5

 90%|█████████ | 17190/19090 [8:23:56<59:25,  1.88s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1100.159912109375, 'learning_rate': 1.4952017568119786e-06, 'recon_loss': 0.0014400396292330697, 'recon_static': 0.0004610563722962979, 'recon_motion': 0.000978983256209176, 'recon_norm': 1.240879660844803, 'recon_norm_std': 0.041488026920706036, 'target_norm': 1.6199481695890428, 'recon_cos': 0.661481462419033, 'gradnorm/compressor.query': 0.43460240620606616, 'gradnorm/compressor.layer0': 9.595233814915122, 'gradnorm/compressor.layer1': 1089.1921891934408, 'gradnorm/compressor.layer2': 145.76698416265256, 'gradnorm/compressor.layer3': 50.266795525693446, 'gradnorm/compressor.layer4': 8.757653077552103, 'gradnorm/compressor.layer5': 5.4193617742687445, 'gradnorm/compressor.layer6': 2.996833242497387, 'gradnorm/compressor.layer7': 3.487025326382224, 'gradnorm/decoder.layer0': 3.5498834026692183, 'gradnorm/decoder.layer1': 2.003553095404141, 'gradnorm/decoder.layer2': 1.2278875495108559, 'gradnorm/max_param': 768.5181884765625, 'epoch': 9.0}

 90%|█████████ | 17190/19090 [8:23:56<59:25,  1.88s/it]
 90%|█████████ | 17191/19090 [8:23:57<53:50,  1.70s/it]
 90%|█████████ | 17192/19090 [8:23:59<49:56,  1.58s/it]
 90%|█████████ | 17193/19090 [8:24:01<57:58,  1.83s/it]
 90%|█████████ | 17194/19090 [8:24:03<54:06,  1.71s/it]
 90%|█████████ | 17195/19090 [8:24:04<50:03,  1.59s/it]
 90%|█████████ | 17196/19090 [8:24:07<1:01:12,  1.94s/it]
 90%|█████████ | 17197/19090 [8:24:08<55:05,  1.75s/it]  
 90%|█████████ | 17198/19090 [8:24:09<50:56,  1.62s/it]
 90%|█████████ | 17199/19090 [8:24:11<55:49,  1.77s/it][gradnorm] step 17199: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 865.5

 90%|█████████ | 17200/19090 [8:24:13<51:23,  1.63s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1236.713134765625, 'learning_rate': 1.4796690852087235e-06, 'recon_loss': 0.0014279251889092847, 'recon_static': 0.0004467733400815632, 'recon_motion': 0.0009811518379137852, 'recon_norm': 1.2312258064746857, 'recon_norm_std': 0.04021841669455171, 'target_norm': 1.6185022175312043, 'recon_cos': 0.6635932177305222, 'gradnorm/compressor.query': 0.3723955510100604, 'gradnorm/compressor.layer0': 8.300735293990433, 'gradnorm/compressor.layer1': 1226.4423141754364, 'gradnorm/compressor.layer2': 151.9237961734435, 'gradnorm/compressor.layer3': 44.80060669624422, 'gradnorm/compressor.layer4': 8.926066735040605, 'gradnorm/compressor.layer5': 5.208478493257093, 'gradnorm/compressor.layer6': 2.8866166650378124, 'gradnorm/compressor.layer7': 3.1492509950516783, 'gradnorm/decoder.layer0': 3.301480949322627, 'gradnorm/decoder.layer1': 2.025186143961609, 'gradnorm/decoder.layer2': 1.2214058036164104, 'gradnorm/max_param': 865.5498046875, 'epoch': 9.01}

 90%|█████████ | 17200/19090 [8:24:13<51:23,  1.63s/it]
 90%|█████████ | 17201/19090 [8:24:14<48:15,  1.53s/it]
 90%|█████████ | 17202/19090 [8:24:17<57:31,  1.83s/it]
 90%|█████████ | 17203/19090 [8:24:18<52:32,  1.67s/it]
 90%|█████████ | 17204/19090 [8:24:19<49:01,  1.56s/it]
 90%|█████████ | 17205/19090 [8:24:21<51:13,  1.63s/it]
 90%|█████████ | 17206/19090 [8:24:22<47:58,  1.53s/it]
 90%|█████████ | 17207/19090 [8:24:24<45:41,  1.46s/it]
 90%|█████████ | 17208/19090 [8:24:26<58:58,  1.88s/it]
 90%|█████████ | 17209/19090 [8:24:28<53:55,  1.72s/it][gradnorm] step 17209: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 847.7

 90%|█████████ | 17210/19090 [8:24:29<49:50,  1.59s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1214.8927001953125, 'learning_rate': 1.464215054089227e-06, 'recon_loss': 0.0014257984614232555, 'recon_static': 0.0004523093717580196, 'recon_motion': 0.0009734890321851708, 'recon_norm': 1.2315767377614975, 'recon_norm_std': 0.04050894258543849, 'target_norm': 1.6193088620901108, 'recon_cos': 0.6648834019899368, 'gradnorm/compressor.query': 0.5664749629993714, 'gradnorm/compressor.layer0': 23.796754452255406, 'gradnorm/compressor.layer1': 1201.6584373273463, 'gradnorm/compressor.layer2': 171.20145297280627, 'gradnorm/compressor.layer3': 44.02420210627048, 'gradnorm/compressor.layer4': 9.887634693101214, 'gradnorm/compressor.layer5': 5.732933923574413, 'gradnorm/compressor.layer6': 2.9943433520022675, 'gradnorm/compressor.layer7': 3.109510715915772, 'gradnorm/decoder.layer0': 3.437951491789828, 'gradnorm/decoder.layer1': 2.0491332198809875, 'gradnorm/decoder.layer2': 1.1440426561499253, 'gradnorm/max_param': 847.734619140625, 'epoch': 9.02}

 90%|█████████ | 17210/19090 [8:24:29<49:50,  1.59s/it]
 90%|█████████ | 17211/19090 [8:24:31<56:37,  1.81s/it]
 90%|█████████ | 17212/19090 [8:24:33<51:46,  1.65s/it]
 90%|█████████ | 17213/19090 [8:24:34<48:18,  1.54s/it]
 90%|█████████ | 17214/19090 [8:24:37<59:27,  1.90s/it]
 90%|█████████ | 17215/19090 [8:24:38<53:36,  1.72s/it]
 90%|█████████ | 17216/19090 [8:24:39<49:31,  1.59s/it]
 90%|█████████ | 17217/19090 [8:24:42<56:11,  1.80s/it]
 90%|█████████ | 17218/19090 [8:24:43<51:19,  1.65s/it]
 90%|█████████ | 17219/19090 [8:24:44<47:55,  1.54s/it][gradnorm] step 17219: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 670.0

 90%|█████████ | 17220/19090 [8:24:46<51:43,  1.66s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 960.4759521484375, 'learning_rate': 1.4488397151242116e-06, 'recon_loss': 0.0014375869533978403, 'recon_static': 0.0004552034806692973, 'recon_motion': 0.0009823835105635225, 'recon_norm': 1.2368591338396073, 'recon_norm_std': 0.0406027608551085, 'target_norm': 1.6199195027351379, 'recon_cos': 0.6621544435620308, 'gradnorm/compressor.query': 0.3342126880183385, 'gradnorm/compressor.layer0': 9.280141847776022, 'gradnorm/compressor.layer1': 949.5803678467663, 'gradnorm/compressor.layer2': 135.75617935981774, 'gradnorm/compressor.layer3': 46.09789176307307, 'gradnorm/compressor.layer4': 10.40502409258083, 'gradnorm/compressor.layer5': 4.972598330991249, 'gradnorm/compressor.layer6': 3.173133187139745, 'gradnorm/compressor.layer7': 3.039523288342028, 'gradnorm/decoder.layer0': 3.5094078870881544, 'gradnorm/decoder.layer1': 2.067211318797902, 'gradnorm/decoder.layer2': 1.1733624999438212, 'gradnorm/max_param': 669.982421875, 'epoch': 9.02}

 90%|█████████ | 17220/19090 [8:24:46<51:43,  1.66s/it]
 90%|█████████ | 17221/19090 [8:24:47<48:16,  1.55s/it]
 90%|█████████ | 17222/19090 [8:24:49<45:51,  1.47s/it]
 90%|█████████ | 17223/19090 [8:24:50<49:17,  1.58s/it]
 90%|█████████ | 17224/19090 [8:24:52<49:44,  1.60s/it]
 90%|█████████ | 17225/19090 [8:24:53<46:46,  1.50s/it]
 90%|█████████ | 17226/19090 [8:24:55<50:54,  1.64s/it]
 90%|█████████ | 17227/19090 [8:24:57<48:37,  1.57s/it]
 90%|█████████ | 17228/19090 [8:24:58<45:58,  1.48s/it]
 90%|█████████ | 17229/19090 [8:25:01<55:46,  1.80s/it][gradnorm] step 17229: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 936.6

 90%|█████████ | 17230/19090 [8:25:02<50:56,  1.64s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 1341.795166015625, 'learning_rate': 1.433543119721284e-06, 'recon_loss': 0.0014548824430676178, 'recon_static': 0.0004657848367060069, 'recon_motion': 0.0009890976420138032, 'recon_norm': 1.2350068032741546, 'recon_norm_std': 0.040578663162887094, 'target_norm': 1.6223551332950592, 'recon_cos': 0.6578436538577079, 'gradnorm/compressor.query': 0.5151322351052655, 'gradnorm/compressor.layer0': 17.703705066897808, 'gradnorm/compressor.layer1': 1327.7297823729043, 'gradnorm/compressor.layer2': 185.96444132548567, 'gradnorm/compressor.layer3': 49.93317555978565, 'gradnorm/compressor.layer4': 9.267757794293688, 'gradnorm/compressor.layer5': 5.518505176544181, 'gradnorm/compressor.layer6': 3.2800792240251373, 'gradnorm/compressor.layer7': 3.4852745366155955, 'gradnorm/decoder.layer0': 3.5757946438379125, 'gradnorm/decoder.layer1': 2.126791367277942, 'gradnorm/decoder.layer2': 1.2190653075200784, 'gradnorm/max_param': 936.56591796875, 'epoch': 9.03}

 90%|█████████ | 17230/19090 [8:25:02<50:56,  1.64s/it]
 90%|█████████ | 17231/19090 [8:25:03<47:35,  1.54s/it]
 90%|█████████ | 17232/19090 [8:25:05<54:12,  1.75s/it]
 90%|█████████ | 17233/19090 [8:25:07<49:49,  1.61s/it]
 90%|█████████ | 17234/19090 [8:25:08<46:48,  1.51s/it]
 90%|█████████ | 17235/19090 [8:25:11<57:21,  1.86s/it]
 90%|█████████ | 17236/19090 [8:25:12<52:06,  1.69s/it]
 90%|█████████ | 17237/19090 [8:25:13<48:21,  1.57s/it]
 90%|█████████ | 17238/19090 [8:25:16<57:13,  1.85s/it]
 90%|█████████ | 17239/19090 [8:25:17<51:57,  1.68s/it][gradnorm] step 17239: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1138.8

 90%|█████████ | 17240/19090 [8:25:18<48:16,  1.57s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 1625.768310546875, 'learning_rate': 1.4183253190247853e-06, 'recon_loss': 0.0014719686703756451, 'recon_static': 0.0004824665367777925, 'recon_motion': 0.0009895021576085128, 'recon_norm': 1.2521326839923859, 'recon_norm_std': 0.04037237428128719, 'target_norm': 1.6223902761936189, 'recon_cos': 0.6554758861660958, 'gradnorm/compressor.query': 0.683536701546723, 'gradnorm/compressor.layer0': 26.99946537194938, 'gradnorm/compressor.layer1': 1613.895829971687, 'gradnorm/compressor.layer2': 187.54304714184954, 'gradnorm/compressor.layer3': 49.08058733078054, 'gradnorm/compressor.layer4': 9.092767511627459, 'gradnorm/compressor.layer5': 5.474946963258026, 'gradnorm/compressor.layer6': 3.0954352518666153, 'gradnorm/compressor.layer7': 3.563679683851948, 'gradnorm/decoder.layer0': 3.414580351708169, 'gradnorm/decoder.layer1': 2.0946864980142923, 'gradnorm/decoder.layer2': 1.137894903529171, 'gradnorm/max_param': 1138.82177734375, 'epoch': 9.03}

 90%|█████████ | 17240/19090 [8:25:18<48:16,  1.57s/it]
 90%|█████████ | 17241/19090 [8:25:20<48:46,  1.58s/it]
 90%|█████████ | 17242/19090 [8:25:22<51:54,  1.69s/it]
 90%|█████████ | 17243/19090 [8:25:23<48:13,  1.57s/it]
 90%|█████████ | 17244/19090 [8:25:25<48:20,  1.57s/it]
 90%|█████████ | 17245/19090 [8:25:27<53:57,  1.75s/it]
 90%|█████████ | 17246/19090 [8:25:28<49:39,  1.62s/it]
 90%|█████████ | 17247/19090 [8:25:30<47:49,  1.56s/it]
 90%|█████████ | 17248/19090 [8:25:32<53:38,  1.75s/it]
 90%|█████████ | 17249/19090 [8:25:33<49:22,  1.61s/it][gradnorm] step 17249: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 829.9

 90%|█████████ | 17250/19090 [8:25:34<46:23,  1.51s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 1187.3245849609375, 'learning_rate': 1.4031863639155928e-06, 'recon_loss': 0.0014751960552530363, 'recon_static': 0.0004919542436255143, 'recon_motion': 0.0009832418552832678, 'recon_norm': 1.2524741858243942, 'recon_norm_std': 0.04060248900204897, 'target_norm': 1.6210843414068221, 'recon_cos': 0.6527776479721069, 'gradnorm/compressor.query': 0.4201999896609175, 'gradnorm/compressor.layer0': 13.39903025747835, 'gradnorm/compressor.layer1': 1176.116756533976, 'gradnorm/compressor.layer2': 155.46607095681682, 'gradnorm/compressor.layer3': 44.663050566855595, 'gradnorm/compressor.layer4': 9.136142146593052, 'gradnorm/compressor.layer5': 5.196089653457451, 'gradnorm/compressor.layer6': 2.896363042964046, 'gradnorm/compressor.layer7': 3.189530716820655, 'gradnorm/decoder.layer0': 3.234962741690314, 'gradnorm/decoder.layer1': 2.015300281006712, 'gradnorm/decoder.layer2': 1.1663023118596785, 'gradnorm/max_param': 829.9278564453125, 'epoch': 9.04}

 90%|█████████ | 17250/19090 [8:25:34<46:23,  1.51s/it]
 90%|█████████ | 17251/19090 [8:25:37<52:30,  1.71s/it]
 90%|█████████ | 17252/19090 [8:25:38<48:34,  1.59s/it]
 90%|█████████ | 17253/19090 [8:25:39<49:16,  1.61s/it]
 90%|█████████ | 17254/19090 [8:25:42<53:30,  1.75s/it]
 90%|█████████ | 17255/19090 [8:25:43<49:15,  1.61s/it]
 90%|█████████ | 17256/19090 [8:25:45<51:51,  1.70s/it]
 90%|█████████ | 17257/19090 [8:25:47<56:17,  1.84s/it]
 90%|█████████ | 17258/19090 [8:25:48<51:09,  1.68s/it]
 90%|█████████ | 17259/19090 [8:25:50<51:44,  1.70s/it][gradnorm] step 17259: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 812.6

 90%|█████████ | 17260/19090 [8:25:52<54:20,  1.78s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 1161.5528564453125, 'learning_rate': 1.3881263050109728e-06, 'recon_loss': 0.001460685025085695, 'recon_static': 0.000482410732365679, 'recon_motion': 0.0009782742738025263, 'recon_norm': 1.2404283106327056, 'recon_norm_std': 0.04166863104328513, 'target_norm': 1.6203124791383743, 'recon_cos': 0.6557059854269027, 'gradnorm/compressor.query': 0.35725655406361634, 'gradnorm/compressor.layer0': 11.855935302804957, 'gradnorm/compressor.layer1': 1151.4923469133437, 'gradnorm/compressor.layer2': 146.4578517654482, 'gradnorm/compressor.layer3': 39.54018704560668, 'gradnorm/compressor.layer4': 7.907327088579997, 'gradnorm/compressor.layer5': 4.643398907251587, 'gradnorm/compressor.layer6': 2.8136968821015786, 'gradnorm/compressor.layer7': 3.1223940855927568, 'gradnorm/decoder.layer0': 3.0954395651311617, 'gradnorm/decoder.layer1': 1.9850408082894269, 'gradnorm/decoder.layer2': 1.1155061674541722, 'gradnorm/max_param': 812.5689697265625, 'epoch': 9.04}

 90%|█████████ | 17260/19090 [8:25:52<54:20,  1.78s/it]
 90%|█████████ | 17261/19090 [8:25:53<49:57,  1.64s/it]
 90%|█████████ | 17262/19090 [8:25:55<52:57,  1.74s/it]
 90%|█████████ | 17263/19090 [8:25:57<50:38,  1.66s/it]
 90%|█████████ | 17264/19090 [8:25:58<47:10,  1.55s/it]
 90%|█████████ | 17265/19090 [8:26:00<47:58,  1.58s/it]
 90%|█████████ | 17266/19090 [8:26:01<47:46,  1.57s/it]
 90%|█████████ | 17267/19090 [8:26:02<45:07,  1.48s/it]
 90%|█████████ | 17268/19090 [8:26:05<51:53,  1.71s/it]
 90%|█████████ | 17269/19090 [8:26:06<52:02,  1.71s/it][gradnorm] step 17269: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 817.7

 90%|█████████ | 17270/19090 [8:26:08<48:07,  1.59s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 1169.7637939453125, 'learning_rate': 1.3731451926643962e-06, 'recon_loss': 0.0014493354712612927, 'recon_static': 0.00046668533905176447, 'recon_motion': 0.0009826501525822095, 'recon_norm': 1.243741649389267, 'recon_norm_std': 0.0403929453343153, 'target_norm': 1.6196009546518326, 'recon_cos': 0.6592878520488739, 'gradnorm/compressor.query': 0.6968751556670011, 'gradnorm/compressor.layer0': 23.000448471341, 'gradnorm/compressor.layer1': 1158.7803933446578, 'gradnorm/compressor.layer2': 152.74243664785828, 'gradnorm/compressor.layer3': 39.80714157670471, 'gradnorm/compressor.layer4': 9.173421232260742, 'gradnorm/compressor.layer5': 3.9136020899760045, 'gradnorm/compressor.layer6': 2.5791865157522693, 'gradnorm/compressor.layer7': 3.086974768809522, 'gradnorm/decoder.layer0': 3.300972945385865, 'gradnorm/decoder.layer1': 1.897433859435102, 'gradnorm/decoder.layer2': 0.9703322993667983, 'gradnorm/max_param': 817.6686401367188, 'epoch': 9.05}

 90%|█████████ | 17270/19090 [8:26:08<48:07,  1.59s/it]
 90%|█████████ | 17271/19090 [8:26:10<51:35,  1.70s/it]
 90%|█████████ | 17272/19090 [8:26:11<48:10,  1.59s/it]
 90%|█████████ | 17273/19090 [8:26:12<45:23,  1.50s/it]
 90%|█████████ | 17274/19090 [8:26:15<55:02,  1.82s/it]
 90%|█████████ | 17275/19090 [8:26:16<50:17,  1.66s/it]
 90%|█████████ | 17276/19090 [8:26:17<46:53,  1.55s/it]
 91%|█████████ | 17277/19090 [8:26:20<57:29,  1.90s/it]
 91%|█████████ | 17278/19090 [8:26:21<51:52,  1.72s/it]
 91%|█████████ | 17279/19090 [8:26:23<47:56,  1.59s/it][gradnorm] step 17279: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 860.4

 91%|█████████ | 17280/19090 [8:26:25<52:04,  1.73s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1231.5753173828125, 'learning_rate': 1.3582430769653736e-06, 'recon_loss': 0.0014355752646224573, 'recon_static': 0.0004600787709932774, 'recon_motion': 0.0009754964674357324, 'recon_norm': 1.2419607549905778, 'recon_norm_std': 0.0405276071280241, 'target_norm': 1.6206490069627761, 'recon_cos': 0.6633470550179481, 'gradnorm/compressor.query': 0.3982492450951212, 'gradnorm/compressor.layer0': 12.186009399029782, 'gradnorm/compressor.layer1': 1219.2040948914173, 'gradnorm/compressor.layer2': 168.17912697478246, 'gradnorm/compressor.layer3': 42.17489684430923, 'gradnorm/compressor.layer4': 7.875115287784582, 'gradnorm/compressor.layer5': 4.107660784555083, 'gradnorm/compressor.layer6': 2.432972150499759, 'gradnorm/compressor.layer7': 2.842918358230665, 'gradnorm/decoder.layer0': 3.1436388727951607, 'gradnorm/decoder.layer1': 1.9008304286980864, 'gradnorm/decoder.layer2': 0.9518092720247029, 'gradnorm/max_param': 860.352783203125, 'epoch': 9.05}

 91%|█████████ | 17280/19090 [8:26:25<52:04,  1.73s/it]
 91%|█████████ | 17281/19090 [8:26:27<54:57,  1.82s/it]
 91%|█████████ | 17282/19090 [8:26:28<50:09,  1.66s/it]
 91%|█████████ | 17283/19090 [8:26:30<50:04,  1.66s/it]
 91%|█████████ | 17284/19090 [8:26:31<50:10,  1.67s/it]
 91%|█████████ | 17285/19090 [8:26:33<46:44,  1.55s/it]
 91%|█████████ | 17286/19090 [8:26:35<54:43,  1.82s/it]
 91%|█████████ | 17287/19090 [8:26:36<49:53,  1.66s/it]
 91%|█████████ | 17288/19090 [8:26:38<46:30,  1.55s/it]
 91%|█████████ | 17289/19090 [8:26:40<52:38,  1.75s/it][gradnorm] step 17289: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 637.4

 91%|█████████ | 17290/19090 [8:26:41<48:25,  1.61s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 909.6758422851562, 'learning_rate': 1.343420007739296e-06, 'recon_loss': 0.001449767465237528, 'recon_static': 0.0004803849638847169, 'recon_motion': 0.0009693825311842374, 'recon_norm': 1.237096408009529, 'recon_norm_std': 0.040829696040600535, 'target_norm': 1.61797394156456, 'recon_cos': 0.6568120241165161, 'gradnorm/compressor.query': 0.39117267362476127, 'gradnorm/compressor.layer0': 12.719084114353825, 'gradnorm/compressor.layer1': 903.1432264596796, 'gradnorm/compressor.layer2': 100.36538228362157, 'gradnorm/compressor.layer3': 38.409919284138866, 'gradnorm/compressor.layer4': 8.7481648291752, 'gradnorm/compressor.layer5': 4.713711505450478, 'gradnorm/compressor.layer6': 2.5719870819847133, 'gradnorm/compressor.layer7': 3.239011380813075, 'gradnorm/decoder.layer0': 3.4413663448310667, 'gradnorm/decoder.layer1': 1.8587935363965808, 'gradnorm/decoder.layer2': 0.96879285286906, 'gradnorm/max_param': 637.3543701171875, 'epoch': 9.06}

 91%|█████████ | 17290/19090 [8:26:41<48:25,  1.61s/it]
 91%|█████████ | 17291/19090 [8:26:43<45:26,  1.52s/it]
 91%|█████████ | 17292/19090 [8:26:45<55:20,  1.85s/it]
 91%|█████████ | 17293/19090 [8:26:47<51:34,  1.72s/it]
 91%|█████████ | 17294/19090 [8:26:48<47:43,  1.59s/it]
 91%|█████████ | 17295/19090 [8:26:50<55:06,  1.84s/it]
 91%|█████████ | 17296/19090 [8:26:52<50:09,  1.68s/it]
 91%|█████████ | 17297/19090 [8:26:53<46:41,  1.56s/it]
 91%|█████████ | 17298/19090 [8:26:55<49:20,  1.65s/it]
 91%|█████████ | 17299/19090 [8:26:57<52:23,  1.76s/it][gradnorm] step 17299: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 743.4

 91%|█████████ | 17300/19090 [8:26:58<48:10,  1.61s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 1067.392333984375, 'learning_rate': 1.328676034547252e-06, 'recon_loss': 0.001468754987581633, 'recon_static': 0.0004868483381869737, 'recon_motion': 0.0009819066850468515, 'recon_norm': 1.254408949613571, 'recon_norm_std': 0.04042128492146731, 'target_norm': 1.621896782517433, 'recon_cos': 0.6559046924114227, 'gradnorm/compressor.query': 0.5321572354862184, 'gradnorm/compressor.layer0': 13.036421593515115, 'gradnorm/compressor.layer1': 1053.4011581539105, 'gradnorm/compressor.layer2': 166.61640192557573, 'gradnorm/compressor.layer3': 40.16182779841295, 'gradnorm/compressor.layer4': 8.907134430455567, 'gradnorm/compressor.layer5': 4.21951264446347, 'gradnorm/compressor.layer6': 2.4980200556552403, 'gradnorm/compressor.layer7': 3.225950760382223, 'gradnorm/decoder.layer0': 3.166419839441601, 'gradnorm/decoder.layer1': 1.8567738651145662, 'gradnorm/decoder.layer2': 0.9585518726009229, 'gradnorm/max_param': 743.3549194335938, 'epoch': 9.06}

 91%|█████████ | 17300/19090 [8:26:58<48:10,  1.61s/it]
 91%|█████████ | 17301/19090 [8:27:00<53:55,  1.81s/it]
 91%|█████████ | 17302/19090 [8:27:02<52:20,  1.76s/it]
 91%|█████████ | 17303/19090 [8:27:03<48:06,  1.62s/it]
 91%|█████████ | 17304/19090 [8:27:06<53:36,  1.80s/it]
 91%|█████████ | 17305/19090 [8:27:07<53:02,  1.78s/it]
 91%|█████████ | 17306/19090 [8:27:09<48:37,  1.64s/it]
 91%|█████████ | 17307/19090 [8:27:11<51:54,  1.75s/it]
 91%|█████████ | 17308/19090 [8:27:12<52:16,  1.76s/it]
 91%|█████████ | 17309/19090 [8:27:14<48:05,  1.62s/it][gradnorm] step 17309: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 898.5

 91%|█████████ | 17310/19090 [8:27:16<53:49,  1.81s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1286.0924072265625, 'learning_rate': 1.3140112066858835e-06, 'recon_loss': 0.0014290063147200272, 'recon_static': 0.0004604319554346148, 'recon_motion': 0.0009685743789304979, 'recon_norm': 1.237213781476021, 'recon_norm_std': 0.040115983504801986, 'target_norm': 1.6175435096025468, 'recon_cos': 0.6628422841429711, 'gradnorm/compressor.query': 0.35980141261864074, 'gradnorm/compressor.layer0': 10.05924831001303, 'gradnorm/compressor.layer1': 1273.2599400750814, 'gradnorm/compressor.layer2': 176.1724785189219, 'gradnorm/compressor.layer3': 39.68154790504807, 'gradnorm/compressor.layer4': 9.00724162056998, 'gradnorm/compressor.layer5': 4.181556469187626, 'gradnorm/compressor.layer6': 2.630370323304254, 'gradnorm/compressor.layer7': 3.155360483653067, 'gradnorm/decoder.layer0': 3.258511036143419, 'gradnorm/decoder.layer1': 1.806983049263586, 'gradnorm/decoder.layer2': 0.9109548904041445, 'gradnorm/max_param': 898.5010375976562, 'epoch': 9.07}

 91%|█████████ | 17310/19090 [8:27:16<53:49,  1.81s/it]
 91%|█████████ | 17311/19090 [8:27:17<49:15,  1.66s/it]
 91%|█████████ | 17312/19090 [8:27:18<45:57,  1.55s/it]
 91%|█████████ | 17313/19090 [8:27:21<53:16,  1.80s/it]
 91%|█████████ | 17314/19090 [8:27:22<50:07,  1.69s/it]
 91%|█████████ | 17315/19090 [8:27:24<46:25,  1.57s/it]
 91%|█████████ | 17316/19090 [8:27:26<54:13,  1.83s/it]
 91%|█████████ | 17317/19090 [8:27:28<54:45,  1.85s/it]
 91%|█████████ | 17318/19090 [8:27:29<49:39,  1.68s/it]
 91%|█████████ | 17319/19090 [8:27:32<55:35,  1.88s/it][gradnorm] step 17319: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1028.5

 91%|█████████ | 17320/19090 [8:27:33<50:55,  1.73s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1470.3721923828125, 'learning_rate': 1.2994255731871963e-06, 'recon_loss': 0.001430588393122889, 'recon_static': 0.00045761192886857317, 'recon_motion': 0.000972976500634104, 'recon_norm': 1.2310772508382797, 'recon_norm_std': 0.04097598502412438, 'target_norm': 1.6205118715763092, 'recon_cos': 0.6640120401978493, 'gradnorm/compressor.query': 0.46986249198106766, 'gradnorm/compressor.layer0': 14.171880383474722, 'gradnorm/compressor.layer1': 1457.35256544187, 'gradnorm/compressor.layer2': 189.47274705086218, 'gradnorm/compressor.layer3': 43.22602523726302, 'gradnorm/compressor.layer4': 9.611303318970732, 'gradnorm/compressor.layer5': 4.543366117146897, 'gradnorm/compressor.layer6': 2.8285152945790273, 'gradnorm/compressor.layer7': 3.3537930266655467, 'gradnorm/decoder.layer0': 3.338340733393445, 'gradnorm/decoder.layer1': 1.931379248482259, 'gradnorm/decoder.layer2': 0.9534857020513828, 'gradnorm/max_param': 1028.5350341796875, 'epoch': 9.07}

 91%|█████████ | 17320/19090 [8:27:33<50:55,  1.73s/it]
 91%|█████████ | 17321/19090 [8:27:34<46:59,  1.59s/it]
 91%|█████████ | 17322/19090 [8:27:37<54:58,  1.87s/it]
 91%|█████████ | 17323/19090 [8:27:38<49:46,  1.69s/it]
 91%|█████████ | 17324/19090 [8:27:39<46:09,  1.57s/it]
 91%|█████████ | 17325/19090 [8:27:41<50:34,  1.72s/it]
 91%|█████████ | 17326/19090 [8:27:43<52:50,  1.80s/it]
 91%|█████████ | 17327/19090 [8:27:45<48:16,  1.64s/it]
 91%|█████████ | 17328/19090 [8:27:47<50:58,  1.74s/it]
 91%|█████████ | 17329/19090 [8:27:48<51:11,  1.74s/it][gradnorm] step 17329: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 637.7

 91%|█████████ | 17330/19090 [8:27:50<47:07,  1.61s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 912.6683959960938, 'learning_rate': 1.2849191828184226e-06, 'recon_loss': 0.0014369396230904385, 'recon_static': 0.0004556864143523853, 'recon_motion': 0.0009812532342039048, 'recon_norm': 1.238269755244255, 'recon_norm_std': 0.0419373894110322, 'target_norm': 1.6235659629106522, 'recon_cos': 0.6641799882054329, 'gradnorm/compressor.query': 0.35760695454797603, 'gradnorm/compressor.layer0': 11.143478306374918, 'gradnorm/compressor.layer1': 903.7395573393919, 'gradnorm/compressor.layer2': 122.06438410604258, 'gradnorm/compressor.layer3': 32.329665562616405, 'gradnorm/compressor.layer4': 10.012217496965544, 'gradnorm/compressor.layer5': 4.167427120750007, 'gradnorm/compressor.layer6': 2.653901801275433, 'gradnorm/compressor.layer7': 3.2560242792484924, 'gradnorm/decoder.layer0': 3.0710091225262826, 'gradnorm/decoder.layer1': 1.919132229386563, 'gradnorm/decoder.layer2': 0.9643607532984162, 'gradnorm/max_param': 637.65966796875, 'epoch': 9.08}

 91%|█████████ | 17330/19090 [8:27:50<47:07,  1.61s/it]
 91%|█████████ | 17331/19090 [8:27:52<50:56,  1.74s/it]
 91%|█████████ | 17332/19090 [8:27:54<52:12,  1.78s/it]
 91%|█████████ | 17333/19090 [8:27:55<47:53,  1.64s/it]
 91%|█████████ | 17334/19090 [8:27:57<50:31,  1.73s/it]
 91%|█████████ | 17335/19090 [8:27:59<51:26,  1.76s/it]
 91%|█████████ | 17336/19090 [8:28:00<47:17,  1.62s/it]
 91%|█████████ | 17337/19090 [8:28:02<55:39,  1.91s/it]
 91%|█████████ | 17338/19090 [8:28:04<52:41,  1.80s/it]
 91%|█████████ | 17339/19090 [8:28:05<48:09,  1.65s/it][gradnorm] step 17339: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 560.5

 91%|█████████ | 17340/19090 [8:28:07<52:01,  1.78s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 804.1737060546875, 'learning_rate': 1.2704920840818312e-06, 'recon_loss': 0.0014276960340794177, 'recon_static': 0.0004520914328168146, 'recon_motion': 0.0009756045794347302, 'recon_norm': 1.2323189496994018, 'recon_norm_std': 0.04174736998975277, 'target_norm': 1.620465636253357, 'recon_cos': 0.664671567082405, 'gradnorm/compressor.query': 0.291137132304225, 'gradnorm/compressor.layer0': 7.73002314042234, 'gradnorm/compressor.layer1': 794.1599728644097, 'gradnorm/compressor.layer2': 122.1094379798302, 'gradnorm/compressor.layer3': 30.179982767776412, 'gradnorm/compressor.layer4': 8.854953530446513, 'gradnorm/compressor.layer5': 3.909404000129747, 'gradnorm/compressor.layer6': 2.5944240567452077, 'gradnorm/compressor.layer7': 3.147598041101298, 'gradnorm/decoder.layer0': 2.9995940251633884, 'gradnorm/decoder.layer1': 1.8524794742409783, 'gradnorm/decoder.layer2': 1.0254070171680822, 'gradnorm/max_param': 560.4904174804688, 'epoch': 9.08}

 91%|█████████ | 17340/19090 [8:28:07<52:01,  1.78s/it]
 91%|█████████ | 17341/19090 [8:28:09<48:41,  1.67s/it]
 91%|█████████ | 17342/19090 [8:28:10<45:29,  1.56s/it]
 91%|█████████ | 17343/19090 [8:28:13<52:31,  1.80s/it]
 91%|█████████ | 17344/19090 [8:28:14<52:22,  1.80s/it]
 91%|█████████ | 17345/19090 [8:28:16<47:59,  1.65s/it]
 91%|█████████ | 17346/19090 [8:28:18<55:23,  1.91s/it]
 91%|█████████ | 17347/19090 [8:28:20<51:17,  1.77s/it]
 91%|█████████ | 17348/19090 [8:28:21<47:10,  1.62s/it]
 91%|█████████ | 17349/19090 [8:28:23<53:28,  1.84s/it][gradnorm] step 17349: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 825.3

 91%|█████████ | 17350/19090 [8:28:25<55:17,  1.91s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1178.162841796875, 'learning_rate': 1.2561443252145893e-06, 'recon_loss': 0.001423570711631328, 'recon_static': 0.00045050728658679873, 'recon_motion': 0.000973063388664741, 'recon_norm': 1.2335907220840454, 'recon_norm_std': 0.04116930207237601, 'target_norm': 1.6202069073915482, 'recon_cos': 0.6658383473753929, 'gradnorm/compressor.query': 0.2887189989250952, 'gradnorm/compressor.layer0': 7.934717606805919, 'gradnorm/compressor.layer1': 1169.51576517805, 'gradnorm/compressor.layer2': 136.78095370655595, 'gradnorm/compressor.layer3': 37.18722754627084, 'gradnorm/compressor.layer4': 9.603777984455233, 'gradnorm/compressor.layer5': 4.37661124940774, 'gradnorm/compressor.layer6': 2.9480668603672817, 'gradnorm/compressor.layer7': 3.3433347069790402, 'gradnorm/decoder.layer0': 3.1726918037937994, 'gradnorm/decoder.layer1': 1.8452513449642922, 'gradnorm/decoder.layer2': 0.9827482921204129, 'gradnorm/max_param': 825.2860717773438, 'epoch': 9.09}

 91%|█████████ | 17350/19090 [8:28:25<55:17,  1.91s/it]
 91%|█████████ | 17351/19090 [8:28:27<49:55,  1.72s/it]
 91%|█████████ | 17352/19090 [8:28:28<50:50,  1.76s/it]
 91%|█████████ | 17353/19090 [8:28:30<51:03,  1.76s/it]
 91%|█████████ | 17354/19090 [8:28:31<46:58,  1.62s/it]
 91%|█████████ | 17355/19090 [8:28:34<51:50,  1.79s/it]
 91%|█████████ | 17356/19090 [8:28:35<49:23,  1.71s/it]
 91%|█████████ | 17357/19090 [8:28:36<45:45,  1.58s/it]
 91%|█████████ | 17358/19090 [8:28:39<52:45,  1.83s/it]
 91%|█████████ | 17359/19090 [8:28:40<51:05,  1.77s/it][gradnorm] step 17359: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 582.2

 91%|█████████ | 17360/19090 [8:28:42<46:56,  1.63s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 835.54736328125, 'learning_rate': 1.2418759541885817e-06, 'recon_loss': 0.0014289382001152263, 'recon_static': 0.0004474531851883512, 'recon_motion': 0.0009814850316615775, 'recon_norm': 1.230291846394539, 'recon_norm_std': 0.04115575449541211, 'target_norm': 1.6219397991895677, 'recon_cos': 0.6650676861405372, 'gradnorm/compressor.query': 0.36558310032267494, 'gradnorm/compressor.layer0': 10.137334774754615, 'gradnorm/compressor.layer1': 825.1224909066532, 'gradnorm/compressor.layer2': 125.44971835734626, 'gradnorm/compressor.layer3': 36.461686601743345, 'gradnorm/compressor.layer4': 9.461378999169094, 'gradnorm/compressor.layer5': 4.477296352984275, 'gradnorm/compressor.layer6': 3.0478548479536123, 'gradnorm/compressor.layer7': 3.26499633578552, 'gradnorm/decoder.layer0': 2.9605448520370117, 'gradnorm/decoder.layer1': 1.7803783626469736, 'gradnorm/decoder.layer2': 0.9213805408199912, 'gradnorm/max_param': 582.2280883789062, 'epoch': 9.09}

 91%|█████████ | 17360/19090 [8:28:42<46:56,  1.63s/it]
 91%|█████████ | 17361/19090 [8:28:44<53:29,  1.86s/it]
 91%|█████████ | 17362/19090 [8:28:46<51:59,  1.81s/it]
 91%|█████████ | 17363/19090 [8:28:47<47:32,  1.65s/it]
 91%|█████████ | 17364/19090 [8:28:49<52:38,  1.83s/it]
 91%|█████████ | 17365/19090 [8:28:51<52:54,  1.84s/it]
 91%|█████████ | 17366/19090 [8:28:53<48:08,  1.68s/it]
 91%|█████████ | 17367/19090 [8:28:55<50:46,  1.77s/it]
 91%|█████████ | 17368/19090 [8:28:57<53:01,  1.85s/it]
 91%|█████████ | 17369/19090 [8:28:58<48:09,  1.68s/it][gradnorm] step 17369: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 882.2

 91%|█████████ | 17370/19090 [8:29:00<49:17,  1.72s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1262.300537109375, 'learning_rate': 1.2276870187102662e-06, 'recon_loss': 0.0014228775486117228, 'recon_static': 0.0004532665196165908, 'recon_motion': 0.0009696110559161752, 'recon_norm': 1.2316366881132126, 'recon_norm_std': 0.04084974080324173, 'target_norm': 1.620708990097046, 'recon_cos': 0.6663966953754425, 'gradnorm/compressor.query': 0.3272248250116671, 'gradnorm/compressor.layer0': 8.746685599252324, 'gradnorm/compressor.layer1': 1250.0153499057521, 'gradnorm/compressor.layer2': 170.7553416990227, 'gradnorm/compressor.layer3': 38.74598170949876, 'gradnorm/compressor.layer4': 8.799616926698606, 'gradnorm/compressor.layer5': 4.649912162689478, 'gradnorm/compressor.layer6': 2.6749727907312146, 'gradnorm/compressor.layer7': 3.1628626679556695, 'gradnorm/decoder.layer0': 2.9224629346130313, 'gradnorm/decoder.layer1': 1.8614749750447614, 'gradnorm/decoder.layer2': 0.9218482967323938, 'gradnorm/max_param': 882.1575927734375, 'epoch': 9.1}

 91%|█████████ | 17370/19090 [8:29:00<49:17,  1.72s/it]
 91%|█████████ | 17371/19090 [8:29:02<53:32,  1.87s/it]
 91%|█████████ | 17372/19090 [8:29:03<48:32,  1.70s/it]
 91%|█████████ | 17373/19090 [8:29:05<46:54,  1.64s/it]
 91%|█████████ | 17374/19090 [8:29:07<50:14,  1.76s/it]
 91%|█████████ | 17375/19090 [8:29:08<46:09,  1.62s/it]
 91%|█████████ | 17376/19090 [8:29:10<50:36,  1.77s/it]
 91%|█████████ | 17377/19090 [8:29:12<50:44,  1.78s/it]
 91%|█████████ | 17378/19090 [8:29:13<46:29,  1.63s/it]
 91%|█████████ | 17379/19090 [8:29:15<50:24,  1.77s/it][gradnorm] step 17379: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 755.1

 91%|█████████ | 17380/19090 [8:29:17<53:44,  1.89s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1078.7635498046875, 'learning_rate': 1.2135775662205028e-06, 'recon_loss': 0.001430826450814493, 'recon_static': 0.0004504381438891869, 'recon_motion': 0.0009803883614949882, 'recon_norm': 1.2264415919780731, 'recon_norm_std': 0.042382167931646106, 'target_norm': 1.623357245326042, 'recon_cos': 0.6646479427814483, 'gradnorm/compressor.query': 0.34620875219623604, 'gradnorm/compressor.layer0': 8.981941331391917, 'gradnorm/compressor.layer1': 1069.873298573247, 'gradnorm/compressor.layer2': 131.64396230932886, 'gradnorm/compressor.layer3': 39.304006559645096, 'gradnorm/compressor.layer4': 9.460369367102519, 'gradnorm/compressor.layer5': 4.796162207513041, 'gradnorm/compressor.layer6': 2.9150169793447454, 'gradnorm/compressor.layer7': 3.239320374311062, 'gradnorm/decoder.layer0': 3.160836239756515, 'gradnorm/decoder.layer1': 1.994797855165077, 'gradnorm/decoder.layer2': 0.9807792567047046, 'gradnorm/max_param': 755.1211547851562, 'epoch': 9.1}

 91%|█████████ | 17380/19090 [8:29:17<53:44,  1.89s/it]
 91%|█████████ | 17381/19090 [8:29:19<48:45,  1.71s/it]
 91%|█████████ | 17382/19090 [8:29:21<52:25,  1.84s/it]
 91%|█████████ | 17383/19090 [8:29:23<51:51,  1.82s/it]
 91%|█████████ | 17384/19090 [8:29:24<47:16,  1.66s/it]
 91%|█████████ | 17385/19090 [8:29:26<54:16,  1.91s/it]
 91%|█████████ | 17386/19090 [8:29:28<51:58,  1.83s/it]
 91%|█████████ | 17387/19090 [8:29:29<47:25,  1.67s/it]
 91%|█████████ | 17388/19090 [8:29:32<51:06,  1.80s/it]
 91%|█████████ | 17389/19090 [8:29:33<50:39,  1.79s/it][gradnorm] step 17389: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 917.4

 91%|█████████ | 17390/19090 [8:29:35<46:28,  1.64s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1309.59033203125, 'learning_rate': 1.199547643894397e-06, 'recon_loss': 0.0014338054723339155, 'recon_static': 0.0004422504636750091, 'recon_motion': 0.000991555025393609, 'recon_norm': 1.2306252330541612, 'recon_norm_std': 0.04140594201162458, 'target_norm': 1.623984482884407, 'recon_cos': 0.664687667787075, 'gradnorm/compressor.query': 0.6033282189609073, 'gradnorm/compressor.layer0': 15.717737002180899, 'gradnorm/compressor.layer1': 1300.0291342889204, 'gradnorm/compressor.layer2': 150.97203926762398, 'gradnorm/compressor.layer3': 41.9687441827993, 'gradnorm/compressor.layer4': 9.493512397891825, 'gradnorm/compressor.layer5': 4.802258778619812, 'gradnorm/compressor.layer6': 2.784762703804007, 'gradnorm/compressor.layer7': 3.3149724406251844, 'gradnorm/decoder.layer0': 3.4257330896522116, 'gradnorm/decoder.layer1': 2.2280909595858587, 'gradnorm/decoder.layer2': 1.0926382136264363, 'gradnorm/max_param': 917.4124755859375, 'epoch': 9.11}

 91%|█████████ | 17390/19090 [8:29:35<46:28,  1.64s/it]
 91%|█████████ | 17391/19090 [8:29:37<52:18,  1.85s/it]
 91%|█████████ | 17392/19090 [8:29:39<52:17,  1.85s/it]
 91%|█████████ | 17393/19090 [8:29:40<47:32,  1.68s/it]
 91%|█████████ | 17394/19090 [8:29:43<56:09,  1.99s/it]
 91%|█████████ | 17395/19090 [8:29:44<50:37,  1.79s/it]
 91%|█████████ | 17396/19090 [8:29:45<46:22,  1.64s/it]
 91%|█████████ | 17397/19090 [8:29:48<57:52,  2.05s/it]
 91%|█████████ | 17398/19090 [8:29:50<51:24,  1.82s/it]
 91%|█████████ | 17399/19090 [8:29:51<46:52,  1.66s/it][gradnorm] step 17399: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 927.0

 91%|█████████ | 17400/19090 [8:29:54<56:15,  2.00s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1328.743896484375, 'learning_rate': 1.185597298641153e-06, 'recon_loss': 0.0014348315715324133, 'recon_static': 0.00045185274284449407, 'recon_motion': 0.0009829788250499405, 'recon_norm': 1.2327277809381485, 'recon_norm_std': 0.041165520902723074, 'target_norm': 1.622338244318962, 'recon_cos': 0.6637219965457917, 'gradnorm/compressor.query': 0.3923342502405563, 'gradnorm/compressor.layer0': 13.955061040271207, 'gradnorm/compressor.layer1': 1313.5085172925221, 'gradnorm/compressor.layer2': 193.60661024678367, 'gradnorm/compressor.layer3': 49.04786555227837, 'gradnorm/compressor.layer4': 10.435110949450598, 'gradnorm/compressor.layer5': 4.94934592501365, 'gradnorm/compressor.layer6': 2.943002781723113, 'gradnorm/compressor.layer7': 3.48431588024397, 'gradnorm/decoder.layer0': 3.4089186827186393, 'gradnorm/decoder.layer1': 2.106961713583354, 'gradnorm/decoder.layer2': 1.108806948226087, 'gradnorm/max_param': 927.0244750976562, 'epoch': 9.11}

 91%|█████████ | 17400/19090 [8:29:54<56:15,  2.00s/it]
 91%|█████████ | 17401/19090 [8:29:55<50:16,  1.79s/it]
 91%|█████████ | 17402/19090 [8:29:56<46:04,  1.64s/it]
 91%|█████████ | 17403/19090 [8:29:59<56:22,  2.01s/it]
 91%|█████████ | 17404/19090 [8:30:00<50:20,  1.79s/it]
 91%|█████████ | 17405/19090 [8:30:02<46:05,  1.64s/it]
 91%|█████████ | 17406/19090 [8:30:04<54:59,  1.96s/it]
 91%|█████████ | 17407/19090 [8:30:06<49:18,  1.76s/it]
 91%|█████████ | 17408/19090 [8:30:07<45:18,  1.62s/it]
 91%|█████████ | 17409/19090 [8:30:09<51:31,  1.84s/it][gradnorm] step 17409: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 698.7

 91%|█████████ | 17410/19090 [8:30:11<47:54,  1.71s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1001.8908081054688, 'learning_rate': 1.1717265771039044e-06, 'recon_loss': 0.0014349952078191563, 'recon_static': 0.00046509638632414864, 'recon_motion': 0.0009698988709715195, 'recon_norm': 1.2406730979681015, 'recon_norm_std': 0.04166972069069743, 'target_norm': 1.619477963447571, 'recon_cos': 0.6624413952231407, 'gradnorm/compressor.query': 0.4281664194829239, 'gradnorm/compressor.layer0': 13.215550389398338, 'gradnorm/compressor.layer1': 989.9891413545909, 'gradnorm/compressor.layer2': 147.50034427925922, 'gradnorm/compressor.layer3': 40.333974113313424, 'gradnorm/compressor.layer4': 9.529007644783162, 'gradnorm/compressor.layer5': 4.6453759120766485, 'gradnorm/compressor.layer6': 3.089781093806097, 'gradnorm/compressor.layer7': 3.4278302214608942, 'gradnorm/decoder.layer0': 3.12637832762587, 'gradnorm/decoder.layer1': 1.9951919937487081, 'gradnorm/decoder.layer2': 1.043671274928159, 'gradnorm/max_param': 698.7454223632812, 'epoch': 9.12}

 91%|█████████ | 17410/19090 [8:30:11<47:54,  1.71s/it]
 91%|█████████ | 17411/19090 [8:30:12<44:20,  1.58s/it]
 91%|█████████ | 17412/19090 [8:30:15<53:36,  1.92s/it]
 91%|█████████ | 17413/19090 [8:30:16<50:35,  1.81s/it]
 91%|█████████ | 17414/19090 [8:30:18<46:16,  1.66s/it]
 91%|█████████ | 17415/19090 [8:30:20<52:11,  1.87s/it]
 91%|█████████ | 17416/19090 [8:30:21<48:48,  1.75s/it]
 91%|█████████ | 17417/19090 [8:30:23<44:58,  1.61s/it]
 91%|█████████ | 17418/19090 [8:30:25<51:00,  1.83s/it]
 91%|█████████ | 17419/19090 [8:30:27<50:57,  1.83s/it][gradnorm] step 17419: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 740.0

 91%|█████████▏| 17420/19090 [8:30:28<46:24,  1.67s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1062.5726318359375, 'learning_rate': 1.157935525659548e-06, 'recon_loss': 0.0014339052024297417, 'recon_static': 0.0004501956595049705, 'recon_motion': 0.000983709499996621, 'recon_norm': 1.2357632249593735, 'recon_norm_std': 0.041211523674428464, 'target_norm': 1.6204780757427215, 'recon_cos': 0.6634385615587235, 'gradnorm/compressor.query': 0.578511186437174, 'gradnorm/compressor.layer0': 15.120986279810545, 'gradnorm/compressor.layer1': 1048.53218834712, 'gradnorm/compressor.layer2': 166.34348288067977, 'gradnorm/compressor.layer3': 40.33595490145857, 'gradnorm/compressor.layer4': 8.107743938962624, 'gradnorm/compressor.layer5': 3.9057286028978804, 'gradnorm/compressor.layer6': 2.672950037576712, 'gradnorm/compressor.layer7': 3.448181341947602, 'gradnorm/decoder.layer0': 3.0854341893253396, 'gradnorm/decoder.layer1': 1.9789745706058823, 'gradnorm/decoder.layer2': 1.039376118304396, 'gradnorm/max_param': 740.0162353515625, 'epoch': 9.13}

 91%|█████████▏| 17420/19090 [8:30:28<46:24,  1.67s/it]
 91%|█████████▏| 17421/19090 [8:30:31<52:01,  1.87s/it]
 91%|█████████▏| 17422/19090 [8:30:32<50:02,  1.80s/it]
 91%|█████████▏| 17423/19090 [8:30:34<45:44,  1.65s/it]
 91%|█████████▏| 17424/19090 [8:30:36<53:08,  1.91s/it]
 91%|█████████▏| 17425/19090 [8:30:37<48:48,  1.76s/it]
 91%|█████████▏| 17426/19090 [8:30:39<44:50,  1.62s/it]
 91%|█████████▏| 17427/19090 [8:30:42<56:19,  2.03s/it]
 91%|█████████▏| 17428/19090 [8:30:43<50:04,  1.81s/it]
 91%|█████████▏| 17429/19090 [8:30:44<45:46,  1.65s/it][gradnorm] step 17429: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 839.4

 91%|█████████▏| 17430/19090 [8:30:47<54:30,  1.97s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1201.8165283203125, 'learning_rate': 1.1442241904186275e-06, 'recon_loss': 0.0014562770840711892, 'recon_static': 0.00046259780719992705, 'recon_motion': 0.0009936793008819222, 'recon_norm': 1.2391444891691208, 'recon_norm_std': 0.04121829615905881, 'target_norm': 1.6248029798269272, 'recon_cos': 0.659533117711544, 'gradnorm/compressor.query': 0.47755277649079614, 'gradnorm/compressor.layer0': 13.73061266371948, 'gradnorm/compressor.layer1': 1189.815269275025, 'gradnorm/compressor.layer2': 162.5088158666323, 'gradnorm/compressor.layer3': 44.24002911144216, 'gradnorm/compressor.layer4': 9.512356802906458, 'gradnorm/compressor.layer5': 4.717029384150315, 'gradnorm/compressor.layer6': 2.9910964769563098, 'gradnorm/compressor.layer7': 3.3502102059986387, 'gradnorm/decoder.layer0': 3.1344060103307703, 'gradnorm/decoder.layer1': 1.9510006146882153, 'gradnorm/decoder.layer2': 1.0165293848394175, 'gradnorm/max_param': 839.3555297851562, 'epoch': 9.13}

 91%|█████████▏| 17430/19090 [8:30:47<54:30,  1.97s/it]
 91%|█████████▏| 17431/19090 [8:30:48<48:47,  1.76s/it]
 91%|█████████▏| 17432/19090 [8:30:50<44:46,  1.62s/it]
 91%|█████████▏| 17433/19090 [8:30:52<54:37,  1.98s/it]
 91%|█████████▏| 17434/19090 [8:30:54<48:49,  1.77s/it]
 91%|█████████▏| 17435/19090 [8:30:55<44:47,  1.62s/it]
 91%|█████████▏| 17436/19090 [8:30:58<56:56,  2.07s/it]
 91%|█████████▏| 17437/19090 [8:30:59<50:24,  1.83s/it]
 91%|█████████▏| 17438/19090 [8:31:01<45:52,  1.67s/it]
 91%|█████████▏| 17439/19090 [8:31:03<54:09,  1.97s/it][gradnorm] step 17439: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 867.3

 91%|█████████▏| 17440/19090 [8:31:05<48:30,  1.76s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 1242.343505859375, 'learning_rate': 1.130592617225143e-06, 'recon_loss': 0.0014446031651459633, 'recon_static': 0.0004699679353507236, 'recon_motion': 0.0009746352559886872, 'recon_norm': 1.2389551818370819, 'recon_norm_std': 0.041356604918837546, 'target_norm': 1.620014163851738, 'recon_cos': 0.6598220631480217, 'gradnorm/compressor.query': 0.363169991468523, 'gradnorm/compressor.layer0': 9.247676248183158, 'gradnorm/compressor.layer1': 1228.9470086216086, 'gradnorm/compressor.layer2': 176.29491338202303, 'gradnorm/compressor.layer3': 42.34683020392804, 'gradnorm/compressor.layer4': 9.635111859973355, 'gradnorm/compressor.layer5': 4.464644673766131, 'gradnorm/compressor.layer6': 3.0489828008189783, 'gradnorm/compressor.layer7': 3.4126156462472053, 'gradnorm/decoder.layer0': 3.0394555158891285, 'gradnorm/decoder.layer1': 1.9437453610282076, 'gradnorm/decoder.layer2': 1.031315830325899, 'gradnorm/max_param': 867.25634765625, 'epoch': 9.14}

 91%|█████████▏| 17440/19090 [8:31:05<48:30,  1.76s/it]
 91%|█████████▏| 17441/19090 [8:31:06<44:32,  1.62s/it]
 91%|█████████▏| 17442/19090 [8:31:09<53:54,  1.96s/it]
 91%|█████████▏| 17443/19090 [8:31:10<48:19,  1.76s/it]
 91%|█████████▏| 17444/19090 [8:31:11<44:22,  1.62s/it]
 91%|█████████▏| 17445/19090 [8:31:14<53:26,  1.95s/it]
 91%|█████████▏| 17446/19090 [8:31:15<47:56,  1.75s/it]
 91%|█████████▏| 17447/19090 [8:31:16<44:06,  1.61s/it]
 91%|█████████▏| 17448/19090 [8:31:19<53:58,  1.97s/it]
 91%|█████████▏| 17449/19090 [8:31:21<48:20,  1.77s/it][gradnorm] step 17449: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1112.3

 91%|█████████▏| 17450/19090 [8:31:22<44:29,  1.63s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1589.022705078125, 'learning_rate': 1.1170408516563952e-06, 'recon_loss': 0.0014443901047343389, 'recon_static': 0.0004670944639656227, 'recon_motion': 0.0009772956545930356, 'recon_norm': 1.2488160222768783, 'recon_norm_std': 0.04100281000137329, 'target_norm': 1.6196413695812226, 'recon_cos': 0.6611580774188042, 'gradnorm/compressor.query': 0.5540123712762746, 'gradnorm/compressor.layer0': 16.57606681969331, 'gradnorm/compressor.layer1': 1576.4555179262115, 'gradnorm/compressor.layer2': 193.91757873836502, 'gradnorm/compressor.layer3': 41.72861394390769, 'gradnorm/compressor.layer4': 10.0799435889467, 'gradnorm/compressor.layer5': 4.8756308514222875, 'gradnorm/compressor.layer6': 3.096064924885136, 'gradnorm/compressor.layer7': 3.3310170390993963, 'gradnorm/decoder.layer0': 3.1366211530387966, 'gradnorm/decoder.layer1': 1.93595073655401, 'gradnorm/decoder.layer2': 1.0148849725610347, 'gradnorm/max_param': 1112.337158203125, 'epoch': 9.14}

 91%|█████████▏| 17450/19090 [8:31:22<44:29,  1.63s/it]
 91%|█████████▏| 17451/19090 [8:31:25<52:47,  1.93s/it]
 91%|█████████▏| 17452/19090 [8:31:26<47:28,  1.74s/it]
 91%|█████████▏| 17453/19090 [8:31:27<43:47,  1.60s/it]
 91%|█████████▏| 17454/19090 [8:31:30<52:48,  1.94s/it]
 91%|█████████▏| 17455/19090 [8:31:31<47:34,  1.75s/it]
 91%|█████████▏| 17456/19090 [8:31:32<43:53,  1.61s/it]
 91%|█████████▏| 17457/19090 [8:31:34<46:58,  1.73s/it]
 91%|█████████▏| 17458/19090 [8:31:36<43:30,  1.60s/it]
 91%|█████████▏| 17459/19090 [8:31:37<41:02,  1.51s/it][gradnorm] step 17459: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 953.6

 91%|█████████▏| 17460/19090 [8:31:39<44:43,  1.65s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1366.40234375, 'learning_rate': 1.1035689390228726e-06, 'recon_loss': 0.0014440270257182418, 'recon_static': 0.00045526258982135913, 'recon_motion': 0.000988764404610265, 'recon_norm': 1.2364019453525543, 'recon_norm_std': 0.041150576714426276, 'target_norm': 1.6208341479301454, 'recon_cos': 0.6606895789504051, 'gradnorm/compressor.query': 0.4795031264663617, 'gradnorm/compressor.layer0': 9.805955003977893, 'gradnorm/compressor.layer1': 1351.1705110754897, 'gradnorm/compressor.layer2': 197.84572193378355, 'gradnorm/compressor.layer3': 44.272288425841936, 'gradnorm/compressor.layer4': 10.988332976793464, 'gradnorm/compressor.layer5': 5.748209135388183, 'gradnorm/compressor.layer6': 3.3835135563798255, 'gradnorm/compressor.layer7': 3.380461900288444, 'gradnorm/decoder.layer0': 3.5579862945018204, 'gradnorm/decoder.layer1': 1.9581654929885033, 'gradnorm/decoder.layer2': 1.1738315271210837, 'gradnorm/max_param': 953.6215209960938, 'epoch': 9.15}

 91%|█████████▏| 17460/19090 [8:31:39<44:43,  1.65s/it]
 91%|█████████▏| 17461/19090 [8:31:40<41:50,  1.54s/it]
 91%|█████████▏| 17462/19090 [8:31:42<39:51,  1.47s/it]
 91%|█████████▏| 17463/19090 [8:31:44<45:48,  1.69s/it]
 91%|█████████▏| 17464/19090 [8:31:45<42:33,  1.57s/it]
 91%|█████████▏| 17465/19090 [8:31:46<40:18,  1.49s/it]
 91%|█████████▏| 17466/19090 [8:31:48<41:31,  1.53s/it]
 91%|█████████▏| 17467/19090 [8:31:49<39:33,  1.46s/it]
 92%|█████████▏| 17468/19090 [8:31:51<38:10,  1.41s/it]
 92%|█████████▏| 17469/19090 [8:31:53<42:51,  1.59s/it][gradnorm] step 17469: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 922.8

 92%|█████████▏| 17470/19090 [8:31:54<40:26,  1.50s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1319.370361328125, 'learning_rate': 1.0901769243680637e-06, 'recon_loss': 0.0014360317116370424, 'recon_static': 0.0004515562046435662, 'recon_motion': 0.0009844754764344542, 'recon_norm': 1.2362812489271164, 'recon_norm_std': 0.04173208987340331, 'target_norm': 1.6209275275468826, 'recon_cos': 0.6632610112428665, 'gradnorm/compressor.query': 0.4616729558706824, 'gradnorm/compressor.layer0': 16.02856184820943, 'gradnorm/compressor.layer1': 1307.7054331920472, 'gradnorm/compressor.layer2': 168.15399959003355, 'gradnorm/compressor.layer3': 44.041423372278174, 'gradnorm/compressor.layer4': 10.200715462518579, 'gradnorm/compressor.layer5': 5.02450298679205, 'gradnorm/compressor.layer6': 3.3988473228558638, 'gradnorm/compressor.layer7': 3.5528997264297306, 'gradnorm/decoder.layer0': 3.582699579188749, 'gradnorm/decoder.layer1': 2.0804265406556546, 'gradnorm/decoder.layer2': 1.1302295571486427, 'gradnorm/max_param': 922.7549438476562, 'epoch': 9.15}

 92%|█████████▏| 17470/19090 [8:31:54<40:26,  1.50s/it]
 92%|█████████▏| 17471/19090 [8:31:55<38:44,  1.44s/it]
 92%|█████████▏| 17472/19090 [8:31:57<43:59,  1.63s/it]
 92%|█████████▏| 17473/19090 [8:31:59<41:12,  1.53s/it]
 92%|█████████▏| 17474/19090 [8:32:00<39:15,  1.46s/it]
 92%|█████████▏| 17475/19090 [8:32:02<44:28,  1.65s/it]
 92%|█████████▏| 17476/19090 [8:32:03<41:29,  1.54s/it]
 92%|█████████▏| 17477/19090 [8:32:05<39:25,  1.47s/it]
 92%|█████████▏| 17478/19090 [8:32:07<44:53,  1.67s/it]
 92%|█████████▏| 17479/19090 [8:32:08<41:53,  1.56s/it][gradnorm] step 17479: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 833.0

 92%|█████████▏| 17480/19090 [8:32:09<39:40,  1.48s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1191.5894775390625, 'learning_rate': 1.0768648524683067e-06, 'recon_loss': 0.001436464794096537, 'recon_static': 0.00045326951949391513, 'recon_motion': 0.0009831952440435999, 'recon_norm': 1.239745408296585, 'recon_norm_std': 0.04179842742159963, 'target_norm': 1.6215332716703414, 'recon_cos': 0.663395780324936, 'gradnorm/compressor.query': 0.4008343168822902, 'gradnorm/compressor.layer0': 10.193277604637053, 'gradnorm/compressor.layer1': 1180.3335333709706, 'gradnorm/compressor.layer2': 156.4639722408165, 'gradnorm/compressor.layer3': 43.91764059319942, 'gradnorm/compressor.layer4': 10.618245928892648, 'gradnorm/compressor.layer5': 5.226180400480751, 'gradnorm/compressor.layer6': 3.5570740454009853, 'gradnorm/compressor.layer7': 3.6533371497456835, 'gradnorm/decoder.layer0': 3.489881603342251, 'gradnorm/decoder.layer1': 2.172847063724275, 'gradnorm/decoder.layer2': 1.2614460936923009, 'gradnorm/max_param': 833.017578125, 'epoch': 9.16}

 92%|█████████▏| 17480/19090 [8:32:09<39:40,  1.48s/it]
 92%|█████████▏| 17481/19090 [8:32:11<43:40,  1.63s/it]
 92%|█████████▏| 17482/19090 [8:32:13<40:57,  1.53s/it]
 92%|█████████▏| 17483/19090 [8:32:14<38:58,  1.46s/it]
 92%|█████████▏| 17484/19090 [8:32:16<41:58,  1.57s/it]
 92%|█████████▏| 17485/19090 [8:32:17<39:42,  1.48s/it]
 92%|█████████▏| 17486/19090 [8:32:18<38:13,  1.43s/it]
 92%|█████████▏| 17487/19090 [8:32:20<41:00,  1.53s/it]
 92%|█████████▏| 17488/19090 [8:32:21<38:58,  1.46s/it]
 92%|█████████▏| 17489/19090 [8:32:23<37:34,  1.41s/it][gradnorm] step 17489: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 834.1

 92%|█████████▏| 17490/19090 [8:32:25<42:02,  1.58s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1190.8173828125, 'learning_rate': 1.0636327678326645e-06, 'recon_loss': 0.0014433862554142252, 'recon_static': 0.0004531107289949432, 'recon_motion': 0.0009902755729854107, 'recon_norm': 1.2264341801404952, 'recon_norm_std': 0.0425034954212606, 'target_norm': 1.6221834748983384, 'recon_cos': 0.6604181855916977, 'gradnorm/compressor.query': 0.38983920215102125, 'gradnorm/compressor.layer0': 11.114546761754987, 'gradnorm/compressor.layer1': 1181.9763639768776, 'gradnorm/compressor.layer2': 137.05106398939412, 'gradnorm/compressor.layer3': 43.56661201674038, 'gradnorm/compressor.layer4': 10.084398407815836, 'gradnorm/compressor.layer5': 4.63346212377326, 'gradnorm/compressor.layer6': 3.8926050746997345, 'gradnorm/compressor.layer7': 3.8245810784208825, 'gradnorm/decoder.layer0': 3.6092855599291105, 'gradnorm/decoder.layer1': 2.316774284369713, 'gradnorm/decoder.layer2': 1.3514274573803973, 'gradnorm/max_param': 834.1383056640625, 'epoch': 9.16}

 92%|█████████▏| 17490/19090 [8:32:25<42:02,  1.58s/it]
 92%|█████████▏| 17491/19090 [8:32:26<39:41,  1.49s/it]
 92%|█████████▏| 17492/19090 [8:32:27<38:00,  1.43s/it]
 92%|█████████▏| 17493/19090 [8:32:29<44:25,  1.67s/it]
 92%|█████████▏| 17494/19090 [8:32:31<41:18,  1.55s/it]
 92%|█████████▏| 17495/19090 [8:32:32<39:12,  1.47s/it]
 92%|█████████▏| 17496/19090 [8:32:34<45:03,  1.70s/it]
 92%|█████████▏| 17497/19090 [8:32:35<41:45,  1.57s/it]
 92%|█████████▏| 17498/19090 [8:32:37<39:29,  1.49s/it]
 92%|█████████▏| 17499/19090 [8:32:39<44:22,  1.67s/it][gradnorm] step 17499: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 778.9

 92%|█████████▏| 17500/19090 [8:32:40<41:27,  1.56s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1114.3978271484375, 'learning_rate': 1.0504807147027617e-06, 'recon_loss': 0.0014375503174960612, 'recon_static': 0.0004394412382680457, 'recon_motion': 0.0009981090828659945, 'recon_norm': 1.2335550248622895, 'recon_norm_std': 0.04117138097062707, 'target_norm': 1.6228334575891494, 'recon_cos': 0.6633955061435699, 'gradnorm/compressor.query': 0.2670722551151031, 'gradnorm/compressor.layer0': 8.718017547358727, 'gradnorm/compressor.layer1': 1103.6469091154108, 'gradnorm/compressor.layer2': 149.0400336843259, 'gradnorm/compressor.layer3': 37.35493589007432, 'gradnorm/compressor.layer4': 10.071130309476152, 'gradnorm/compressor.layer5': 4.3344182832587865, 'gradnorm/compressor.layer6': 3.4427205064859403, 'gradnorm/compressor.layer7': 3.3774698897910715, 'gradnorm/decoder.layer0': 3.4881213109855236, 'gradnorm/decoder.layer1': 2.13387554351351, 'gradnorm/decoder.layer2': 1.1302975327880036, 'gradnorm/max_param': 778.9129638671875, 'epoch': 9.17}

 92%|█████████▏| 17500/19090 [8:32:40<41:27,  1.56s/it]
 92%|█████████▏| 17501/19090 [8:32:45<1:04:41,  2.44s/it]
 92%|█████████▏| 17502/19090 [8:32:46<55:24,  2.09s/it]  
 92%|█████████▏| 17503/19090 [8:32:47<49:01,  1.85s/it]
 92%|█████████▏| 17504/19090 [8:32:49<44:33,  1.69s/it]
 92%|█████████▏| 17505/19090 [8:32:50<41:23,  1.57s/it]
 92%|█████████▏| 17506/19090 [8:32:51<39:06,  1.48s/it]
 92%|█████████▏| 17507/19090 [8:32:52<37:35,  1.43s/it]
 92%|█████████▏| 17508/19090 [8:32:54<36:28,  1.38s/it]
 92%|█████████▏| 17509/19090 [8:32:55<35:40,  1.35s/it][gradnorm] step 17509: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 766.0

 92%|█████████▏| 17510/19090 [8:32:56<35:09,  1.34s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1095.847900390625, 'learning_rate': 1.0374087370526199e-06, 'recon_loss': 0.0014207128202542662, 'recon_static': 0.0004425478138728067, 'recon_motion': 0.0009781650631339289, 'recon_norm': 1.2223963409662246, 'recon_norm_std': 0.042725106980651614, 'target_norm': 1.6202813595533372, 'recon_cos': 0.6656173527240753, 'gradnorm/compressor.query': 0.33777317585763494, 'gradnorm/compressor.layer0': 8.70152369343452, 'gradnorm/compressor.layer1': 1085.3999262944512, 'gradnorm/compressor.layer2': 144.4025947654681, 'gradnorm/compressor.layer3': 41.400957798589936, 'gradnorm/compressor.layer4': 9.163401444081266, 'gradnorm/compressor.layer5': 5.0383550102902435, 'gradnorm/compressor.layer6': 3.0581929028221198, 'gradnorm/compressor.layer7': 3.364272127826945, 'gradnorm/decoder.layer0': 3.414056214909475, 'gradnorm/decoder.layer1': 2.152001660087541, 'gradnorm/decoder.layer2': 1.1183979972435314, 'gradnorm/max_param': 766.0003662109375, 'epoch': 9.17}

 92%|█████████▏| 17510/19090 [8:32:56<35:09,  1.34s/it]
 92%|█████████▏| 17511/19090 [8:32:58<34:51,  1.32s/it]
 92%|█████████▏| 17512/19090 [8:32:59<34:31,  1.31s/it]
 92%|█████████▏| 17513/19090 [8:33:00<34:17,  1.30s/it]
 92%|█████████▏| 17514/19090 [8:33:01<34:04,  1.30s/it]
 92%|█████████▏| 17515/19090 [8:33:03<33:59,  1.29s/it]
 92%|█████████▏| 17516/19090 [8:33:04<33:52,  1.29s/it]
 92%|█████████▏| 17517/19090 [8:33:05<33:46,  1.29s/it]
 92%|█████████▏| 17518/19090 [8:33:07<37:50,  1.44s/it]
 92%|█████████▏| 17519/19090 [8:33:08<37:14,  1.42s/it][gradnorm] step 17519: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 897.7

 92%|█████████▏| 17520/19090 [8:33:10<39:46,  1.52s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1284.72412109375, 'learning_rate': 1.024416878588552e-06, 'recon_loss': 0.0014338314038468524, 'recon_static': 0.00044360943866195157, 'recon_motion': 0.0009902219841023906, 'recon_norm': 1.2335597157478333, 'recon_norm_std': 0.0414370215497911, 'target_norm': 1.6221796989440918, 'recon_cos': 0.6643860831856727, 'gradnorm/compressor.query': 0.4777684509151851, 'gradnorm/compressor.layer0': 7.3175844953563525, 'gradnorm/compressor.layer1': 1272.0444174634783, 'gradnorm/compressor.layer2': 174.9384489970687, 'gradnorm/compressor.layer3': 40.1702598980071, 'gradnorm/compressor.layer4': 9.217150119383007, 'gradnorm/compressor.layer5': 4.906616197271213, 'gradnorm/compressor.layer6': 3.1669565034703346, 'gradnorm/compressor.layer7': 3.576853963659091, 'gradnorm/decoder.layer0': 3.316162100388008, 'gradnorm/decoder.layer1': 2.0772963858115228, 'gradnorm/decoder.layer2': 1.0815219162383882, 'gradnorm/max_param': 897.6669311523438, 'epoch': 9.18}

 92%|█████████▏| 17520/19090 [8:33:10<39:46,  1.52s/it]
 92%|█████████▏| 17521/19090 [8:33:12<41:49,  1.60s/it]
 92%|█████████▏| 17522/19090 [8:33:13<40:50,  1.56s/it]
 92%|█████████▏| 17523/19090 [8:33:15<42:46,  1.64s/it]
 92%|█████████▏| 17524/19090 [8:33:17<44:09,  1.69s/it]
 92%|█████████▏| 17525/19090 [8:33:18<41:41,  1.60s/it]
 92%|█████████▏| 17526/19090 [8:33:20<43:15,  1.66s/it]
 92%|█████████▏| 17527/19090 [8:33:22<46:48,  1.80s/it]
 92%|█████████▏| 17528/19090 [8:33:24<42:47,  1.64s/it]
 92%|█████████▏| 17529/19090 [8:33:25<41:45,  1.60s/it][gradnorm] step 17529: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1002.3

 92%|█████████▏| 17530/19090 [8:33:28<48:38,  1.87s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1435.2352294921875, 'learning_rate': 1.0115051827489702e-06, 'recon_loss': 0.001407929405104369, 'recon_static': 0.0004465489160793368, 'recon_motion': 0.0009613805043045432, 'recon_norm': 1.2352988392114639, 'recon_norm_std': 0.042197801917791364, 'target_norm': 1.61435064971447, 'recon_cos': 0.6671200066804885, 'gradnorm/compressor.query': 0.6073740115427538, 'gradnorm/compressor.layer0': 23.87198448111104, 'gradnorm/compressor.layer1': 1420.6943724812877, 'gradnorm/compressor.layer2': 196.5064506733812, 'gradnorm/compressor.layer3': 46.53526363650205, 'gradnorm/compressor.layer4': 10.277354514081212, 'gradnorm/compressor.layer5': 5.202108931207931, 'gradnorm/compressor.layer6': 3.288807766217467, 'gradnorm/compressor.layer7': 3.8404056297083775, 'gradnorm/decoder.layer0': 3.5951473007478714, 'gradnorm/decoder.layer1': 2.1144505225757007, 'gradnorm/decoder.layer2': 1.1931288975791356, 'gradnorm/max_param': 1002.3217163085938, 'epoch': 9.18}

 92%|█████████▏| 17530/19090 [8:33:28<48:38,  1.87s/it]
 92%|█████████▏| 17531/19090 [8:33:29<44:03,  1.70s/it]
 92%|█████████▏| 17532/19090 [8:33:30<41:56,  1.62s/it]
 92%|█████████▏| 17533/19090 [8:33:32<45:29,  1.75s/it]
 92%|█████████▏| 17534/19090 [8:33:34<42:33,  1.64s/it]
 92%|█████████▏| 17535/19090 [8:33:35<39:45,  1.53s/it]
 92%|█████████▏| 17536/19090 [8:33:37<45:24,  1.75s/it]
 92%|█████████▏| 17537/19090 [8:33:39<43:43,  1.69s/it]
 92%|█████████▏| 17538/19090 [8:33:40<41:51,  1.62s/it]
 92%|█████████▏| 17539/19090 [8:33:43<46:42,  1.81s/it][gradnorm] step 17539: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 805.2

 92%|█████████▏| 17540/19090 [8:33:44<44:30,  1.72s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1153.010498046875, 'learning_rate': 9.986736927042773e-07, 'recon_loss': 0.0014282099000411108, 'recon_static': 0.0004484382610826287, 'recon_motion': 0.0009797716629691422, 'recon_norm': 1.2281634360551834, 'recon_norm_std': 0.04142113160341978, 'target_norm': 1.620989939570427, 'recon_cos': 0.6646070718765259, 'gradnorm/compressor.query': 0.40209797880285664, 'gradnorm/compressor.layer0': 7.846308682209963, 'gradnorm/compressor.layer1': 1141.1375355319797, 'gradnorm/compressor.layer2': 157.13802650854439, 'gradnorm/compressor.layer3': 47.65853990400094, 'gradnorm/compressor.layer4': 11.56145872633983, 'gradnorm/compressor.layer5': 5.398784974062867, 'gradnorm/compressor.layer6': 3.774812751824961, 'gradnorm/compressor.layer7': 3.923619220616485, 'gradnorm/decoder.layer0': 3.749616984834449, 'gradnorm/decoder.layer1': 2.1531635253294352, 'gradnorm/decoder.layer2': 1.277094825497934, 'gradnorm/max_param': 805.2405395507812, 'epoch': 9.19}

 92%|█████████▏| 17540/19090 [8:33:44<44:30,  1.72s/it]
 92%|█████████▏| 17541/19090 [8:33:45<41:05,  1.59s/it]
 92%|█████████▏| 17542/19090 [8:33:48<47:31,  1.84s/it]
 92%|█████████▏| 17543/19090 [8:33:50<46:56,  1.82s/it]
 92%|█████████▏| 17544/19090 [8:33:51<42:46,  1.66s/it]
 92%|█████████▏| 17545/19090 [8:33:53<47:37,  1.85s/it]
 92%|█████████▏| 17546/19090 [8:33:55<46:17,  1.80s/it]
 92%|█████████▏| 17547/19090 [8:33:56<42:18,  1.64s/it]
 92%|█████████▏| 17548/19090 [8:33:58<46:29,  1.81s/it]
 92%|█████████▏| 17549/19090 [8:34:01<48:56,  1.91s/it][gradnorm] step 17549: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1227.0

 92%|█████████▏| 17550/19090 [8:34:02<44:08,  1.72s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1752.4713134765625, 'learning_rate': 9.85922451356694e-07, 'recon_loss': 0.0014221605961211026, 'recon_static': 0.0004428866785019636, 'recon_motion': 0.0009792739685508423, 'recon_norm': 1.2324265539646149, 'recon_norm_std': 0.042614151071757075, 'target_norm': 1.6202948451042176, 'recon_cos': 0.666328065097332, 'gradnorm/compressor.query': 0.5563478469513508, 'gradnorm/compressor.layer0': 19.3161617622293, 'gradnorm/compressor.layer1': 1739.0420495203673, 'gradnorm/compressor.layer2': 208.06498759402072, 'gradnorm/compressor.layer3': 54.92253289079652, 'gradnorm/compressor.layer4': 11.166962083543266, 'gradnorm/compressor.layer5': 5.919379601833748, 'gradnorm/compressor.layer6': 3.674911736706476, 'gradnorm/compressor.layer7': 3.7966431597974695, 'gradnorm/decoder.layer0': 3.694515602044799, 'gradnorm/decoder.layer1': 2.1738271379083, 'gradnorm/decoder.layer2': 1.286782531454531, 'gradnorm/max_param': 1227.0206298828125, 'epoch': 9.19}

 92%|█████████▏| 17550/19090 [8:34:02<44:08,  1.72s/it]
 92%|█████████▏| 17551/19090 [8:34:04<46:23,  1.81s/it]
 92%|█████████▏| 17552/19090 [8:34:06<48:08,  1.88s/it]
 92%|█████████▏| 17553/19090 [8:34:07<43:32,  1.70s/it]
 92%|█████████▏| 17554/19090 [8:34:09<45:28,  1.78s/it]
 92%|█████████▏| 17555/19090 [8:34:11<47:04,  1.84s/it]
 92%|█████████▏| 17556/19090 [8:34:12<43:11,  1.69s/it]
 92%|█████████▏| 17557/19090 [8:34:15<47:10,  1.85s/it]
 92%|█████████▏| 17558/19090 [8:34:16<46:16,  1.81s/it]
 92%|█████████▏| 17559/19090 [8:34:18<42:31,  1.67s/it][gradnorm] step 17559: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 967.2

 92%|█████████▏| 17560/19090 [8:34:20<47:22,  1.86s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1384.881103515625, 'learning_rate': 9.732515013401418e-07, 'recon_loss': 0.0014199150988133623, 'recon_static': 0.0004402659200422931, 'recon_motion': 0.000979649176588282, 'recon_norm': 1.228192102909088, 'recon_norm_std': 0.041615523584187034, 'target_norm': 1.6183611363172532, 'recon_cos': 0.6654930740594864, 'gradnorm/compressor.query': 0.5014160786274945, 'gradnorm/compressor.layer0': 16.94372219082004, 'gradnorm/compressor.layer1': 1370.4010361934202, 'gradnorm/compressor.layer2': 194.56163668295454, 'gradnorm/compressor.layer3': 39.92958585024236, 'gradnorm/compressor.layer4': 9.358370260526428, 'gradnorm/compressor.layer5': 5.765568430553277, 'gradnorm/compressor.layer6': 3.220512426131164, 'gradnorm/compressor.layer7': 3.5934168122125083, 'gradnorm/decoder.layer0': 3.321399753052932, 'gradnorm/decoder.layer1': 2.0219978307001116, 'gradnorm/decoder.layer2': 1.1851927277012628, 'gradnorm/max_param': 967.2146606445312, 'epoch': 9.2}

 92%|█████████▏| 17560/19090 [8:34:20<47:22,  1.86s/it]
 92%|█████████▏| 17561/19090 [8:34:22<45:49,  1.80s/it]
 92%|█████████▏| 17562/19090 [8:34:23<42:01,  1.65s/it]
 92%|█████████▏| 17563/19090 [8:34:25<44:35,  1.75s/it]
 92%|█████████▏| 17564/19090 [8:34:27<48:05,  1.89s/it]
 92%|█████████▏| 17565/19090 [8:34:28<43:26,  1.71s/it]
 92%|█████████▏| 17566/19090 [8:34:30<42:11,  1.66s/it]
 92%|█████████▏| 17567/19090 [8:34:32<45:14,  1.78s/it]
 92%|█████████▏| 17568/19090 [8:34:33<41:24,  1.63s/it]
 92%|█████████▏| 17569/19090 [8:34:35<39:54,  1.57s/it][gradnorm] step 17569: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1016.1

 92%|█████████▏| 17570/19090 [8:34:37<44:56,  1.77s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1455.0184326171875, 'learning_rate': 9.60660885020076e-07, 'recon_loss': 0.0014430131355766206, 'recon_static': 0.00045690947808907367, 'recon_motion': 0.0009861036538495681, 'recon_norm': 1.2410588592290879, 'recon_norm_std': 0.041433404199779034, 'target_norm': 1.6210647612810134, 'recon_cos': 0.6617687121033669, 'gradnorm/compressor.query': 0.30995684779064553, 'gradnorm/compressor.layer0': 7.607514140165779, 'gradnorm/compressor.layer1': 1439.737128784279, 'gradnorm/compressor.layer2': 203.7065617039618, 'gradnorm/compressor.layer3': 50.007628812543444, 'gradnorm/compressor.layer4': 10.241652276132736, 'gradnorm/compressor.layer5': 5.724318831120631, 'gradnorm/compressor.layer6': 3.330144278576785, 'gradnorm/compressor.layer7': 3.540712893843676, 'gradnorm/decoder.layer0': 3.674015411519063, 'gradnorm/decoder.layer1': 2.092090917271028, 'gradnorm/decoder.layer2': 1.1891751771090822, 'gradnorm/max_param': 1016.1168823242188, 'epoch': 9.2}

 92%|█████████▏| 17570/19090 [8:34:37<44:56,  1.77s/it]
 92%|█████████▏| 17571/19090 [8:34:38<41:13,  1.63s/it]
 92%|█████████▏| 17572/19090 [8:34:40<41:23,  1.64s/it]
 92%|█████████▏| 17573/19090 [8:34:42<45:52,  1.81s/it]
 92%|█████████▏| 17574/19090 [8:34:43<41:52,  1.66s/it]
 92%|█████████▏| 17575/19090 [8:34:45<39:02,  1.55s/it]
 92%|█████████▏| 17576/19090 [8:34:47<47:15,  1.87s/it]
 92%|█████████▏| 17577/19090 [8:34:49<42:45,  1.70s/it]
 92%|█████████▏| 17578/19090 [8:34:50<39:40,  1.57s/it]
 92%|█████████▏| 17579/19090 [8:34:53<47:42,  1.89s/it][gradnorm] step 17579: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1010.9

 92%|█████████▏| 17580/19090 [8:34:54<43:07,  1.71s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1444.5091552734375, 'learning_rate': 9.481506444933618e-07, 'recon_loss': 0.0014214741240721196, 'recon_static': 0.0004516006913036108, 'recon_motion': 0.0009698734400444664, 'recon_norm': 1.2303555250167846, 'recon_norm_std': 0.04221376646310091, 'target_norm': 1.616783019900322, 'recon_cos': 0.6641634613275528, 'gradnorm/compressor.query': 0.36808654616697345, 'gradnorm/compressor.layer0': 9.347480769090442, 'gradnorm/compressor.layer1': 1432.3192294317632, 'gradnorm/compressor.layer2': 180.31274371733963, 'gradnorm/compressor.layer3': 47.768065545273025, 'gradnorm/compressor.layer4': 10.994890933881482, 'gradnorm/compressor.layer5': 5.165220653034219, 'gradnorm/compressor.layer6': 3.121304272174866, 'gradnorm/compressor.layer7': 3.315543593864352, 'gradnorm/decoder.layer0': 3.5132438225053595, 'gradnorm/decoder.layer1': 2.0152452687817353, 'gradnorm/decoder.layer2': 1.2202375577023836, 'gradnorm/max_param': 1010.855224609375, 'epoch': 9.21}

 92%|█████████▏| 17580/19090 [8:34:54<43:07,  1.71s/it]
 92%|█████████▏| 17581/19090 [8:34:55<39:53,  1.59s/it]
 92%|█████████▏| 17582/19090 [8:34:58<48:37,  1.93s/it]
 92%|█████████▏| 17583/19090 [8:34:59<43:41,  1.74s/it]
 92%|█████████▏| 17584/19090 [8:35:01<40:19,  1.61s/it]
 92%|█████████▏| 17585/19090 [8:35:03<46:46,  1.86s/it]
 92%|█████████▏| 17586/19090 [8:35:04<42:21,  1.69s/it]
 92%|█████████▏| 17587/19090 [8:35:06<41:16,  1.65s/it]
 92%|█████████▏| 17588/19090 [8:35:08<48:05,  1.92s/it]
 92%|█████████▏| 17589/19090 [8:35:10<43:16,  1.73s/it][gradnorm] step 17589: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1022.1

 92%|█████████▏| 17590/19090 [8:35:11<42:43,  1.71s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1465.852294921875, 'learning_rate': 9.357208215881236e-07, 'recon_loss': 0.0014138671103864908, 'recon_static': 0.0004404436287586577, 'recon_motion': 0.0009734234787174501, 'recon_norm': 1.2288814783096313, 'recon_norm_std': 0.042224032990634444, 'target_norm': 1.6196214079856872, 'recon_cos': 0.668002949655056, 'gradnorm/compressor.query': 0.47182023381110066, 'gradnorm/compressor.layer0': 10.913825172459907, 'gradnorm/compressor.layer1': 1448.4104218072998, 'gradnorm/compressor.layer2': 218.92711579713009, 'gradnorm/compressor.layer3': 50.949226985978406, 'gradnorm/compressor.layer4': 10.712839882799154, 'gradnorm/compressor.layer5': 5.39584955436399, 'gradnorm/compressor.layer6': 3.4635768496607464, 'gradnorm/compressor.layer7': 3.3702934791618895, 'gradnorm/decoder.layer0': 3.6709459793409853, 'gradnorm/decoder.layer1': 2.0707367084559714, 'gradnorm/decoder.layer2': 1.2173295179379, 'gradnorm/max_param': 1022.1342163085938, 'epoch': 9.21}

 92%|█████████▏| 17590/19090 [8:35:11<42:43,  1.71s/it]
 92%|█████████▏| 17591/19090 [8:35:14<46:44,  1.87s/it]
 92%|█████████▏| 17592/19090 [8:35:15<42:19,  1.70s/it]
 92%|█████████▏| 17593/19090 [8:35:17<43:58,  1.76s/it]
 92%|█████████▏| 17594/19090 [8:35:19<46:15,  1.86s/it]
 92%|█████████▏| 17595/19090 [8:35:20<42:26,  1.70s/it]
 92%|█████████▏| 17596/19090 [8:35:22<46:38,  1.87s/it]
 92%|█████████▏| 17597/19090 [8:35:24<44:51,  1.80s/it]
 92%|█████████▏| 17598/19090 [8:35:26<41:59,  1.69s/it]
 92%|█████████▏| 17599/19090 [8:35:28<46:18,  1.86s/it][gradnorm] step 17599: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 954.5

 92%|█████████▏| 17600/19090 [8:35:29<44:34,  1.79s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1366.5074462890625, 'learning_rate': 9.233714578636144e-07, 'recon_loss': 0.001415911238291301, 'recon_static': 0.00043881784658879043, 'recon_motion': 0.000977093365509063, 'recon_norm': 1.2258138120174409, 'recon_norm_std': 0.04237524643540382, 'target_norm': 1.6195627391338348, 'recon_cos': 0.6670540615916252, 'gradnorm/compressor.query': 0.3741019940859243, 'gradnorm/compressor.layer0': 9.474406652064888, 'gradnorm/compressor.layer1': 1352.553095076123, 'gradnorm/compressor.layer2': 188.8345216386559, 'gradnorm/compressor.layer3': 44.95619748580625, 'gradnorm/compressor.layer4': 9.96800453962342, 'gradnorm/compressor.layer5': 5.6388154654063, 'gradnorm/compressor.layer6': 3.527203697558771, 'gradnorm/compressor.layer7': 3.2962897626484264, 'gradnorm/decoder.layer0': 3.5727891074377154, 'gradnorm/decoder.layer1': 2.0466120893219935, 'gradnorm/decoder.layer2': 1.228231618587227, 'gradnorm/max_param': 954.524658203125, 'epoch': 9.22}

 92%|█████████▏| 17600/19090 [8:35:29<44:34,  1.79s/it]
 92%|█████████▏| 17601/19090 [8:35:31<42:47,  1.72s/it]
 92%|█████████▏| 17602/19090 [8:35:33<44:25,  1.79s/it]
 92%|█████████▏| 17603/19090 [8:35:34<42:25,  1.71s/it]
 92%|█████████▏| 17604/19090 [8:35:36<41:48,  1.69s/it]
 92%|█████████▏| 17605/19090 [8:35:38<43:45,  1.77s/it]
 92%|█████████▏| 17606/19090 [8:35:40<42:49,  1.73s/it]
 92%|█████████▏| 17607/19090 [8:35:41<40:44,  1.65s/it]
 92%|█████████▏| 17608/19090 [8:35:43<42:15,  1.71s/it]
 92%|█████████▏| 17609/19090 [8:35:45<43:25,  1.76s/it][gradnorm] step 17609: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 881.5

 92%|█████████▏| 17610/19090 [8:35:47<43:01,  1.74s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1264.9268798828125, 'learning_rate': 9.11102594610061e-07, 'recon_loss': 0.0014194401126587763, 'recon_static': 0.00043606163962977006, 'recon_motion': 0.0009833785239607095, 'recon_norm': 1.2292714953422545, 'recon_norm_std': 0.04196293130517006, 'target_norm': 1.6209985822439195, 'recon_cos': 0.6672281548380852, 'gradnorm/compressor.query': 0.43757885835297017, 'gradnorm/compressor.layer0': 12.947700349326034, 'gradnorm/compressor.layer1': 1249.051340017695, 'gradnorm/compressor.layer2': 191.73498555949564, 'gradnorm/compressor.layer3': 52.8197942447952, 'gradnorm/compressor.layer4': 11.024180451465082, 'gradnorm/compressor.layer5': 5.415434516246675, 'gradnorm/compressor.layer6': 3.2274238231128836, 'gradnorm/compressor.layer7': 3.40244942116685, 'gradnorm/decoder.layer0': 3.4662016709566505, 'gradnorm/decoder.layer1': 1.99013763881456, 'gradnorm/decoder.layer2': 1.214253321984946, 'gradnorm/max_param': 881.5360717773438, 'epoch': 9.22}

 92%|█████████▏| 17610/19090 [8:35:47<43:01,  1.74s/it]
 92%|█████████▏| 17611/19090 [8:35:48<43:32,  1.77s/it]
 92%|█████████▏| 17612/19090 [8:35:50<41:31,  1.69s/it]
 92%|█████████▏| 17613/19090 [8:35:52<42:38,  1.73s/it]
 92%|█████████▏| 17614/19090 [8:35:53<41:43,  1.70s/it]
 92%|█████████▏| 17615/19090 [8:35:55<40:22,  1.64s/it]
 92%|█████████▏| 17616/19090 [8:35:57<40:31,  1.65s/it]
 92%|█████████▏| 17617/19090 [8:35:58<40:49,  1.66s/it]
 92%|█████████▏| 17618/19090 [8:36:00<43:15,  1.76s/it]
 92%|█████████▏| 17619/19090 [8:36:02<41:40,  1.70s/it][gradnorm] step 17619: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 849.6

 92%|█████████▏| 17620/19090 [8:36:03<39:02,  1.59s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1215.6510009765625, 'learning_rate': 8.989142728485472e-07, 'recon_loss': 0.0014178181532770394, 'recon_static': 0.0004422576857905369, 'recon_motion': 0.0009755604492966086, 'recon_norm': 1.224937242269516, 'recon_norm_std': 0.042669299710541966, 'target_norm': 1.6211862802505492, 'recon_cos': 0.6673209890723228, 'gradnorm/compressor.query': 0.37879840063162284, 'gradnorm/compressor.layer0': 10.059713605743944, 'gradnorm/compressor.layer1': 1203.7591640357302, 'gradnorm/compressor.layer2': 163.7645806962696, 'gradnorm/compressor.layer3': 41.09724319753698, 'gradnorm/compressor.layer4': 9.709981317315558, 'gradnorm/compressor.layer5': 5.451955238074571, 'gradnorm/compressor.layer6': 3.0843852714514415, 'gradnorm/compressor.layer7': 3.3741295716393194, 'gradnorm/decoder.layer0': 3.439717357911742, 'gradnorm/decoder.layer1': 1.9708195057465199, 'gradnorm/decoder.layer2': 1.1887760834986139, 'gradnorm/max_param': 849.5771484375, 'epoch': 9.23}

 92%|█████████▏| 17620/19090 [8:36:03<39:02,  1.59s/it]
 92%|█████████▏| 17621/19090 [8:36:05<44:31,  1.82s/it]
 92%|█████████▏| 17622/19090 [8:36:07<40:45,  1.67s/it]
 92%|█████████▏| 17623/19090 [8:36:08<37:55,  1.55s/it]
 92%|█████████▏| 17624/19090 [8:36:11<44:47,  1.83s/it]
 92%|█████████▏| 17625/19090 [8:36:12<41:48,  1.71s/it]
 92%|█████████▏| 17626/19090 [8:36:13<38:40,  1.59s/it]
 92%|█████████▏| 17627/19090 [8:36:15<42:42,  1.75s/it]
 92%|█████████▏| 17628/19090 [8:36:17<43:23,  1.78s/it]
 92%|█████████▏| 17629/19090 [8:36:19<39:44,  1.63s/it][gradnorm] step 17629: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1094.9

 92%|█████████▏| 17630/19090 [8:36:21<42:13,  1.74s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1565.993896484375, 'learning_rate': 8.868065333308584e-07, 'recon_loss': 0.001420764156500809, 'recon_static': 0.0004452143955859356, 'recon_motion': 0.0009755497856531292, 'recon_norm': 1.2352949380874634, 'recon_norm_std': 0.04252618923783302, 'target_norm': 1.6204898685216904, 'recon_cos': 0.6669091239571572, 'gradnorm/compressor.query': 0.37712433242612325, 'gradnorm/compressor.layer0': 9.177465655315316, 'gradnorm/compressor.layer1': 1551.5125684312068, 'gradnorm/compressor.layer2': 205.79188706372756, 'gradnorm/compressor.layer3': 50.30681014022232, 'gradnorm/compressor.layer4': 10.447992042480056, 'gradnorm/compressor.layer5': 5.441593076589028, 'gradnorm/compressor.layer6': 3.3894033736318567, 'gradnorm/compressor.layer7': 3.3939846287897986, 'gradnorm/decoder.layer0': 3.5040186563108384, 'gradnorm/decoder.layer1': 2.1082240602826094, 'gradnorm/decoder.layer2': 1.287313905247992, 'gradnorm/max_param': 1094.9383544921875, 'epoch': 9.24}

 92%|█████████▏| 17630/19090 [8:36:21<42:13,  1.74s/it]
 92%|█████████▏| 17631/19090 [8:36:23<46:36,  1.92s/it]
 92%|█████████▏| 17632/19090 [8:36:24<41:57,  1.73s/it]
 92%|█████████▏| 17633/19090 [8:36:26<41:05,  1.69s/it]
 92%|█████████▏| 17634/19090 [8:36:28<46:01,  1.90s/it]
 92%|█████████▏| 17635/19090 [8:36:29<41:39,  1.72s/it]
 92%|█████████▏| 17636/19090 [8:36:31<39:21,  1.62s/it]
 92%|█████████▏| 17637/19090 [8:36:34<48:10,  1.99s/it]
 92%|█████████▏| 17638/19090 [8:36:35<43:00,  1.78s/it]
 92%|█████████▏| 17639/19090 [8:36:36<39:24,  1.63s/it][gradnorm] step 17639: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1023.6

 92%|█████████▏| 17640/19090 [8:36:39<49:48,  2.06s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1463.9478759765625, 'learning_rate': 8.747794165393536e-07, 'recon_loss': 0.001430360763333738, 'recon_static': 0.0004462336197320838, 'recon_motion': 0.0009841271094046533, 'recon_norm': 1.2224690228700639, 'recon_norm_std': 0.04266013437882066, 'target_norm': 1.6216721713542939, 'recon_cos': 0.6635899856686592, 'gradnorm/compressor.query': 0.5459661561184771, 'gradnorm/compressor.layer0': 16.182507072646377, 'gradnorm/compressor.layer1': 1450.4565660508417, 'gradnorm/compressor.layer2': 192.50657456305225, 'gradnorm/compressor.layer3': 42.629223924696014, 'gradnorm/compressor.layer4': 10.362141925805076, 'gradnorm/compressor.layer5': 5.605317159590055, 'gradnorm/compressor.layer6': 3.471180519669681, 'gradnorm/compressor.layer7': 3.424855645472051, 'gradnorm/decoder.layer0': 3.4278311952137988, 'gradnorm/decoder.layer1': 2.2101614840443173, 'gradnorm/decoder.layer2': 1.3085065670293223, 'gradnorm/max_param': 1023.6080322265625, 'epoch': 9.24}

 92%|█████████▏| 17640/19090 [8:36:39<49:48,  2.06s/it]
 92%|█████████▏| 17641/19090 [8:36:41<44:08,  1.83s/it]
 92%|█████████▏| 17642/19090 [8:36:42<40:10,  1.66s/it]
 92%|█████████▏| 17643/19090 [8:36:45<48:59,  2.03s/it]
 92%|█████████▏| 17644/19090 [8:36:46<43:33,  1.81s/it]
 92%|█████████▏| 17645/19090 [8:36:47<39:48,  1.65s/it]
 92%|█████████▏| 17646/19090 [8:36:50<49:03,  2.04s/it]
 92%|█████████▏| 17647/19090 [8:36:52<43:35,  1.81s/it]
 92%|█████████▏| 17648/19090 [8:36:53<39:44,  1.65s/it]
 92%|█████████▏| 17649/19090 [8:36:56<49:03,  2.04s/it][gradnorm] step 17649: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 695.6

 92%|█████████▏| 17650/19090 [8:36:57<43:34,  1.82s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 997.2770385742188, 'learning_rate': 8.628329626868297e-07, 'recon_loss': 0.0014261798176448792, 'recon_static': 0.0004438710384420119, 'recon_motion': 0.0009823087777476758, 'recon_norm': 1.233384332060814, 'recon_norm_std': 0.04149332772940397, 'target_norm': 1.6189290791749955, 'recon_cos': 0.6645187944173813, 'gradnorm/compressor.query': 0.680459559627097, 'gradnorm/compressor.layer0': 28.29729344665758, 'gradnorm/compressor.layer1': 985.5905272982285, 'gradnorm/compressor.layer2': 142.9686919398789, 'gradnorm/compressor.layer3': 41.81344589737253, 'gradnorm/compressor.layer4': 10.447747413935247, 'gradnorm/compressor.layer5': 5.53562107974574, 'gradnorm/compressor.layer6': 3.5410328615824285, 'gradnorm/compressor.layer7': 3.602863862764015, 'gradnorm/decoder.layer0': 3.3888757649846784, 'gradnorm/decoder.layer1': 2.186740852508654, 'gradnorm/decoder.layer2': 1.3073095596527708, 'gradnorm/max_param': 695.589111328125, 'epoch': 9.25}

 92%|█████████▏| 17650/19090 [8:36:57<43:34,  1.82s/it]
 92%|█████████▏| 17651/19090 [8:36:58<39:43,  1.66s/it]
 92%|█████████▏| 17652/19090 [8:37:01<47:53,  2.00s/it]
 92%|█████████▏| 17653/19090 [8:37:02<42:41,  1.78s/it]
 92%|█████████▏| 17654/19090 [8:37:04<39:03,  1.63s/it]
 92%|█████████▏| 17655/19090 [8:37:06<46:08,  1.93s/it]
 92%|█████████▏| 17656/19090 [8:37:08<41:28,  1.74s/it]
 92%|█████████▏| 17657/19090 [8:37:09<38:10,  1.60s/it]
 92%|█████████▏| 17658/19090 [8:37:11<44:46,  1.88s/it]
 93%|█████████▎| 17659/19090 [8:37:13<40:31,  1.70s/it][gradnorm] step 17659: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 865.6

 93%|█████████▎| 17660/19090 [8:37:14<37:30,  1.57s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 1233.245849609375, 'learning_rate': 8.509672117163858e-07, 'recon_loss': 0.0014859756658552215, 'recon_static': 0.000494916320167249, 'recon_motion': 0.0009910593245876953, 'recon_norm': 1.2492045164108276, 'recon_norm_std': 0.04168695844709873, 'target_norm': 1.6246574491262435, 'recon_cos': 0.6519903123378754, 'gradnorm/compressor.query': 0.7474997489188803, 'gradnorm/compressor.layer0': 29.863296323164935, 'gradnorm/compressor.layer1': 1226.6769644042395, 'gradnorm/compressor.layer2': 117.60806201107134, 'gradnorm/compressor.layer3': 35.157708303087674, 'gradnorm/compressor.layer4': 11.374796100507693, 'gradnorm/compressor.layer5': 5.385792033569197, 'gradnorm/compressor.layer6': 3.306521941396728, 'gradnorm/compressor.layer7': 3.557726958615648, 'gradnorm/decoder.layer0': 3.3721497827661313, 'gradnorm/decoder.layer1': 1.9913058255997507, 'gradnorm/decoder.layer2': 1.2111924426060967, 'gradnorm/max_param': 865.5687255859375, 'epoch': 9.25}

 93%|█████████▎| 17660/19090 [8:37:14<37:30,  1.57s/it]
 93%|█████████▎| 17661/19090 [8:37:17<46:55,  1.97s/it]
 93%|█████████▎| 17662/19090 [8:37:18<41:59,  1.76s/it]
 93%|█████████▎| 17663/19090 [8:37:19<38:34,  1.62s/it]
 93%|█████████▎| 17664/19090 [8:37:22<47:12,  1.99s/it]
 93%|█████████▎| 17665/19090 [8:37:24<42:17,  1.78s/it]
 93%|█████████▎| 17666/19090 [8:37:25<38:44,  1.63s/it]
 93%|█████████▎| 17667/19090 [8:37:28<46:07,  1.94s/it]
 93%|█████████▎| 17668/19090 [8:37:29<41:27,  1.75s/it]
 93%|█████████▎| 17669/19090 [8:37:30<38:12,  1.61s/it][gradnorm] step 17669: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 882.4

 93%|█████████▎| 17670/19090 [8:37:33<45:02,  1.90s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 1261.582763671875, 'learning_rate': 8.391822033012919e-07, 'recon_loss': 0.0015175603330135345, 'recon_static': 0.0005319522708305158, 'recon_motion': 0.0009856080272584223, 'recon_norm': 1.2677929937839507, 'recon_norm_std': 0.04080157531425357, 'target_norm': 1.622547224164009, 'recon_cos': 0.6433960556983948, 'gradnorm/compressor.query': 0.43932310088160764, 'gradnorm/compressor.layer0': 16.488302997646166, 'gradnorm/compressor.layer1': 1250.2138317103997, 'gradnorm/compressor.layer2': 161.26744697280043, 'gradnorm/compressor.layer3': 46.09695963401491, 'gradnorm/compressor.layer4': 9.489773918328934, 'gradnorm/compressor.layer5': 5.030873349704462, 'gradnorm/compressor.layer6': 3.2162218257353494, 'gradnorm/compressor.layer7': 3.3598476343576404, 'gradnorm/decoder.layer0': 3.2674832692261564, 'gradnorm/decoder.layer1': 1.8498103869413505, 'gradnorm/decoder.layer2': 1.0802105068870362, 'gradnorm/max_param': 882.3782348632812, 'epoch': 9.26}

 93%|█████████▎| 17670/19090 [8:37:33<45:02,  1.90s/it]
 93%|█████████▎| 17671/19090 [8:37:34<40:39,  1.72s/it]
 93%|█████████▎| 17672/19090 [8:37:35<37:33,  1.59s/it]
 93%|█████████▎| 17673/19090 [8:37:38<44:30,  1.88s/it]
 93%|█████████▎| 17674/19090 [8:37:39<40:15,  1.71s/it]
 93%|█████████▎| 17675/19090 [8:37:40<37:20,  1.58s/it]
 93%|█████████▎| 17676/19090 [8:37:43<44:25,  1.89s/it]
 93%|█████████▎| 17677/19090 [8:37:44<40:09,  1.71s/it]
 93%|█████████▎| 17678/19090 [8:37:46<37:11,  1.58s/it]
 93%|█████████▎| 17679/19090 [8:37:49<46:22,  1.97s/it][gradnorm] step 17679: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 863.2

 93%|█████████▎| 17680/19090 [8:37:50<41:32,  1.77s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 1237.04296875, 'learning_rate': 8.274779768448482e-07, 'recon_loss': 0.0015172675193753093, 'recon_static': 0.0005335017922334373, 'recon_motion': 0.0009837657737080008, 'recon_norm': 1.2712500602006913, 'recon_norm_std': 0.041444329731166366, 'target_norm': 1.6234231799840928, 'recon_cos': 0.6449312418699265, 'gradnorm/compressor.query': 0.38528076774448794, 'gradnorm/compressor.layer0': 10.68433141306187, 'gradnorm/compressor.layer1': 1223.081763415676, 'gradnorm/compressor.layer2': 179.2927511780663, 'gradnorm/compressor.layer3': 44.08119616649201, 'gradnorm/compressor.layer4': 9.055269290115225, 'gradnorm/compressor.layer5': 4.811108871138367, 'gradnorm/compressor.layer6': 3.1115569313722045, 'gradnorm/compressor.layer7': 3.3724191475826997, 'gradnorm/decoder.layer0': 3.4224754050620136, 'gradnorm/decoder.layer1': 1.8926486699958631, 'gradnorm/decoder.layer2': 1.1869779744406594, 'gradnorm/max_param': 863.2110595703125, 'epoch': 9.26}

 93%|█████████▎| 17680/19090 [8:37:50<41:32,  1.77s/it]
 93%|█████████▎| 17681/19090 [8:37:51<38:09,  1.63s/it]
 93%|█████████▎| 17682/19090 [8:37:54<46:12,  1.97s/it]
 93%|█████████▎| 17683/19090 [8:37:55<41:23,  1.76s/it]
 93%|█████████▎| 17684/19090 [8:37:56<38:01,  1.62s/it]
 93%|█████████▎| 17685/19090 [8:37:59<46:39,  1.99s/it]
 93%|█████████▎| 17686/19090 [8:38:01<41:40,  1.78s/it]
 93%|█████████▎| 17687/19090 [8:38:02<38:09,  1.63s/it]
 93%|█████████▎| 17688/19090 [8:38:05<46:28,  1.99s/it]
 93%|█████████▎| 17689/19090 [8:38:06<41:32,  1.78s/it][gradnorm] step 17689: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1011.6

 93%|█████████▎| 17690/19090 [8:38:07<38:03,  1.63s/it]
                                                       
{'loss': 0.0015, 'grad_norm': 1445.6341552734375, 'learning_rate': 8.158545714802678e-07, 'recon_loss': 0.0014672638528281823, 'recon_static': 0.00047925149847287684, 'recon_motion': 0.000988012325251475, 'recon_norm': 1.242911270260811, 'recon_norm_std': 0.04135341309010983, 'target_norm': 1.623026430606842, 'recon_cos': 0.6554558858275413, 'gradnorm/compressor.query': 0.781534905936051, 'gradnorm/compressor.layer0': 14.033331293995422, 'gradnorm/compressor.layer1': 1433.4179432391657, 'gradnorm/compressor.layer2': 180.8711126050813, 'gradnorm/compressor.layer3': 45.911315347669195, 'gradnorm/compressor.layer4': 9.26159806898917, 'gradnorm/compressor.layer5': 5.130554724551353, 'gradnorm/compressor.layer6': 3.4024848777328436, 'gradnorm/compressor.layer7': 3.376906951182665, 'gradnorm/decoder.layer0': 3.349465307579657, 'gradnorm/decoder.layer1': 1.9536182848754968, 'gradnorm/decoder.layer2': 1.1235824767305653, 'gradnorm/max_param': 1011.5716552734375, 'epoch': 9.27}

 93%|█████████▎| 17690/19090 [8:38:07<38:03,  1.63s/it]
 93%|█████████▎| 17691/19090 [8:38:10<46:00,  1.97s/it]
 93%|█████████▎| 17692/19090 [8:38:11<41:11,  1.77s/it]
 93%|█████████▎| 17693/19090 [8:38:13<37:46,  1.62s/it]
 93%|█████████▎| 17694/19090 [8:38:15<46:14,  1.99s/it]
 93%|█████████▎| 17695/19090 [8:38:17<41:19,  1.78s/it]
 93%|█████████▎| 17696/19090 [8:38:18<37:51,  1.63s/it]
 93%|█████████▎| 17697/19090 [8:38:21<43:55,  1.89s/it]
 93%|█████████▎| 17698/19090 [8:38:22<39:39,  1.71s/it]
 93%|█████████▎| 17699/19090 [8:38:23<36:40,  1.58s/it][gradnorm] step 17699: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 906.3

 93%|█████████▎| 17700/19090 [8:38:26<45:31,  1.97s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1296.6246337890625, 'learning_rate': 8.043120260705334e-07, 'recon_loss': 0.0014250113483285532, 'recon_static': 0.0004509452861384489, 'recon_motion': 0.0009740660942043177, 'recon_norm': 1.2317363917827606, 'recon_norm_std': 0.04131538895890117, 'target_norm': 1.6186711490154266, 'recon_cos': 0.6646495118737221, 'gradnorm/compressor.query': 0.42273696509466474, 'gradnorm/compressor.layer0': 14.69078098742568, 'gradnorm/compressor.layer1': 1284.2549396439945, 'gradnorm/compressor.layer2': 171.1199266505073, 'gradnorm/compressor.layer3': 47.74649754006832, 'gradnorm/compressor.layer4': 9.09917652470455, 'gradnorm/compressor.layer5': 5.036516738232107, 'gradnorm/compressor.layer6': 3.24416605315328, 'gradnorm/compressor.layer7': 3.43538448086558, 'gradnorm/decoder.layer0': 3.3108231240752097, 'gradnorm/decoder.layer1': 1.9820911751763681, 'gradnorm/decoder.layer2': 1.1957211263473637, 'gradnorm/max_param': 906.3298950195312, 'epoch': 9.27}

 93%|█████████▎| 17700/19090 [8:38:26<45:31,  1.97s/it]
 93%|█████████▎| 17701/19090 [8:38:27<40:46,  1.76s/it]
 93%|█████████▎| 17702/19090 [8:38:29<37:31,  1.62s/it]
 93%|█████████▎| 17703/19090 [8:38:31<44:24,  1.92s/it]
 93%|█████████▎| 17704/19090 [8:38:32<39:59,  1.73s/it]
 93%|█████████▎| 17705/19090 [8:38:34<36:54,  1.60s/it]
 93%|█████████▎| 17706/19090 [8:38:37<46:03,  2.00s/it]
 93%|█████████▎| 17707/19090 [8:38:38<41:05,  1.78s/it]
 93%|█████████▎| 17708/19090 [8:38:39<37:41,  1.64s/it]
 93%|█████████▎| 17709/19090 [8:38:42<44:33,  1.94s/it][gradnorm] step 17709: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 760.6

 93%|█████████▎| 17710/19090 [8:38:43<40:10,  1.75s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1088.4193115234375, 'learning_rate': 7.928503792082681e-07, 'recon_loss': 0.0014259741525165736, 'recon_static': 0.0004407529268064536, 'recon_motion': 0.000985221228620503, 'recon_norm': 1.2315528124570847, 'recon_norm_std': 0.04309983672574162, 'target_norm': 1.6223230421543122, 'recon_cos': 0.6663506418466568, 'gradnorm/compressor.query': 0.32256558647741296, 'gradnorm/compressor.layer0': 7.395854881864148, 'gradnorm/compressor.layer1': 1077.773051249659, 'gradnorm/compressor.layer2': 141.86539064373665, 'gradnorm/compressor.layer3': 52.155770200758944, 'gradnorm/compressor.layer4': 9.74312799314575, 'gradnorm/compressor.layer5': 5.237025258983617, 'gradnorm/compressor.layer6': 3.390573808837352, 'gradnorm/compressor.layer7': 3.518412658891027, 'gradnorm/decoder.layer0': 3.132397707450962, 'gradnorm/decoder.layer1': 1.9097392557328565, 'gradnorm/decoder.layer2': 1.1799280287728575, 'gradnorm/max_param': 760.593505859375, 'epoch': 9.28}

 93%|█████████▎| 17710/19090 [8:38:43<40:10,  1.75s/it]
 93%|█████████▎| 17711/19090 [8:38:45<37:34,  1.63s/it]
 93%|█████████▎| 17712/19090 [8:38:47<44:43,  1.95s/it]
 93%|█████████▎| 17713/19090 [8:38:49<40:09,  1.75s/it]
 93%|█████████▎| 17714/19090 [8:38:50<36:56,  1.61s/it]
 93%|█████████▎| 17715/19090 [8:38:52<42:22,  1.85s/it]
 93%|█████████▎| 17716/19090 [8:38:54<38:30,  1.68s/it]
 93%|█████████▎| 17717/19090 [8:38:55<35:46,  1.56s/it]
 93%|█████████▎| 17718/19090 [8:38:58<46:06,  2.02s/it]
 93%|█████████▎| 17719/19090 [8:38:59<41:05,  1.80s/it][gradnorm] step 17719: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 846.2

 93%|█████████▎| 17720/19090 [8:39:00<37:35,  1.65s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1211.5125732421875, 'learning_rate': 7.814696692156203e-07, 'recon_loss': 0.0014057685155421495, 'recon_static': 0.00043581076242844575, 'recon_motion': 0.0009699577465653419, 'recon_norm': 1.2285964429378509, 'recon_norm_std': 0.04238914921879768, 'target_norm': 1.6177619457244874, 'recon_cos': 0.6691824197769165, 'gradnorm/compressor.query': 0.3869047777120007, 'gradnorm/compressor.layer0': 9.539521977186418, 'gradnorm/compressor.layer1': 1198.9096609002697, 'gradnorm/compressor.layer2': 168.06058826543776, 'gradnorm/compressor.layer3': 43.45336097562679, 'gradnorm/compressor.layer4': 9.391488592518144, 'gradnorm/compressor.layer5': 5.260015108500202, 'gradnorm/compressor.layer6': 3.387711077092896, 'gradnorm/compressor.layer7': 3.490706912917796, 'gradnorm/decoder.layer0': 3.191581506481106, 'gradnorm/decoder.layer1': 1.938788447044963, 'gradnorm/decoder.layer2': 1.1224055573764289, 'gradnorm/max_param': 846.2156372070312, 'epoch': 9.28}

 93%|█████████▎| 17720/19090 [8:39:00<37:35,  1.65s/it]
 93%|█████████▎| 17721/19090 [8:39:03<42:40,  1.87s/it]
 93%|█████████▎| 17722/19090 [8:39:04<38:38,  1.69s/it]
 93%|█████████▎| 17723/19090 [8:39:05<35:50,  1.57s/it]
 93%|█████████▎| 17724/19090 [8:39:08<45:20,  1.99s/it]
 93%|█████████▎| 17725/19090 [8:39:10<40:29,  1.78s/it]
 93%|█████████▎| 17726/19090 [8:39:11<37:34,  1.65s/it]
 93%|█████████▎| 17727/19090 [8:39:14<43:27,  1.91s/it]
 93%|█████████▎| 17728/19090 [8:39:15<39:11,  1.73s/it]
 93%|█████████▎| 17729/19090 [8:39:17<38:44,  1.71s/it][gradnorm] step 17729: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 824.6

 93%|█████████▎| 17730/19090 [8:39:19<45:25,  2.00s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1181.8724365234375, 'learning_rate': 7.701699341441104e-07, 'recon_loss': 0.0014001484523760155, 'recon_static': 0.0004332911856181454, 'recon_motion': 0.0009668572689406574, 'recon_norm': 1.2289652347564697, 'recon_norm_std': 0.04250016557052731, 'target_norm': 1.6164317011833191, 'recon_cos': 0.6703790068626404, 'gradnorm/compressor.query': 0.48548265678492375, 'gradnorm/compressor.layer0': 14.11786302513537, 'gradnorm/compressor.layer1': 1168.4292340574161, 'gradnorm/compressor.layer2': 169.17315579200206, 'gradnorm/compressor.layer3': 51.329092932273795, 'gradnorm/compressor.layer4': 8.69110822873886, 'gradnorm/compressor.layer5': 4.970116193552923, 'gradnorm/compressor.layer6': 3.453193353174945, 'gradnorm/compressor.layer7': 3.591824753919129, 'gradnorm/decoder.layer0': 3.330370651328758, 'gradnorm/decoder.layer1': 2.020092294114183, 'gradnorm/decoder.layer2': 1.2189892265087836, 'gradnorm/max_param': 824.6021118164062, 'epoch': 9.29}

 93%|█████████▎| 17730/19090 [8:39:19<45:25,  2.00s/it]
 93%|█████████▎| 17731/19090 [8:39:21<40:31,  1.79s/it]
 93%|█████████▎| 17732/19090 [8:39:22<37:04,  1.64s/it]
 93%|█████████▎| 17733/19090 [8:39:25<44:46,  1.98s/it]
 93%|█████████▎| 17734/19090 [8:39:26<40:03,  1.77s/it]
 93%|█████████▎| 17735/19090 [8:39:27<36:43,  1.63s/it]
 93%|█████████▎| 17736/19090 [8:39:30<44:18,  1.96s/it]
 93%|█████████▎| 17737/19090 [8:39:31<39:41,  1.76s/it]
 93%|█████████▎| 17738/19090 [8:39:32<36:29,  1.62s/it]
 93%|█████████▎| 17739/19090 [8:39:35<42:43,  1.90s/it][gradnorm] step 17739: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 823.1

 93%|█████████▎| 17740/19090 [8:39:36<38:35,  1.72s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1177.2044677734375, 'learning_rate': 7.589512117745334e-07, 'recon_loss': 0.0014084015128901228, 'recon_static': 0.00043208177157794125, 'recon_motion': 0.0009763197434949688, 'recon_norm': 1.2253131687641143, 'recon_norm_std': 0.04270974462851882, 'target_norm': 1.619828286767006, 'recon_cos': 0.66915353089571, 'gradnorm/compressor.query': 0.443298133956145, 'gradnorm/compressor.layer0': 12.937488205761532, 'gradnorm/compressor.layer1': 1166.5421123988624, 'gradnorm/compressor.layer2': 151.56623706732975, 'gradnorm/compressor.layer3': 41.29632962891875, 'gradnorm/compressor.layer4': 8.722678553906773, 'gradnorm/compressor.layer5': 5.047532360621233, 'gradnorm/compressor.layer6': 3.585814895694774, 'gradnorm/compressor.layer7': 3.568003235298024, 'gradnorm/decoder.layer0': 3.426966811095933, 'gradnorm/decoder.layer1': 2.0325879898826003, 'gradnorm/decoder.layer2': 1.3017584077414477, 'gradnorm/max_param': 823.0952758789062, 'epoch': 9.29}

 93%|█████████▎| 17740/19090 [8:39:36<38:35,  1.72s/it]
 93%|█████████▎| 17741/19090 [8:39:38<35:40,  1.59s/it]
 93%|█████████▎| 17742/19090 [8:39:41<45:00,  2.00s/it]
 93%|█████████▎| 17743/19090 [8:39:42<40:11,  1.79s/it]
 93%|█████████▎| 17744/19090 [8:39:43<36:45,  1.64s/it]
 93%|█████████▎| 17745/19090 [8:39:46<44:14,  1.97s/it]
 93%|█████████▎| 17746/19090 [8:39:47<39:44,  1.77s/it]
 93%|█████████▎| 17747/19090 [8:39:48<36:24,  1.63s/it]
 93%|█████████▎| 17748/19090 [8:39:51<42:41,  1.91s/it]
 93%|█████████▎| 17749/19090 [8:39:52<38:28,  1.72s/it][gradnorm] step 17749: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 830.2

 93%|█████████▎| 17750/19090 [8:39:54<35:31,  1.59s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1187.56689453125, 'learning_rate': 7.478135396168068e-07, 'recon_loss': 0.0014121859858278185, 'recon_static': 0.00044568542070919646, 'recon_motion': 0.0009665005825809203, 'recon_norm': 1.2285473495721817, 'recon_norm_std': 0.042661278322339055, 'target_norm': 1.6178805589675904, 'recon_cos': 0.6672440856695175, 'gradnorm/compressor.query': 0.4799815087680667, 'gradnorm/compressor.layer0': 10.450236880889314, 'gradnorm/compressor.layer1': 1176.4394374552394, 'gradnorm/compressor.layer2': 153.60308656770215, 'gradnorm/compressor.layer3': 49.530089201697635, 'gradnorm/compressor.layer4': 9.359002871749638, 'gradnorm/compressor.layer5': 5.23822699624682, 'gradnorm/compressor.layer6': 3.122891439514535, 'gradnorm/compressor.layer7': 3.1504169612003654, 'gradnorm/decoder.layer0': 3.006041483478002, 'gradnorm/decoder.layer1': 1.8623736908345607, 'gradnorm/decoder.layer2': 1.20968982039012, 'gradnorm/max_param': 830.1719970703125, 'epoch': 9.3}

 93%|█████████▎| 17750/19090 [8:39:54<35:31,  1.59s/it]
 93%|█████████▎| 17751/19090 [8:39:56<43:51,  1.97s/it]
 93%|█████████▎| 17752/19090 [8:39:58<39:17,  1.76s/it]
 93%|█████████▎| 17753/19090 [8:39:59<36:02,  1.62s/it]
 93%|█████████▎| 17754/19090 [8:40:02<42:59,  1.93s/it]
 93%|█████████▎| 17755/19090 [8:40:03<38:36,  1.73s/it]
 93%|█████████▎| 17756/19090 [8:40:04<35:32,  1.60s/it]
 93%|█████████▎| 17757/19090 [8:40:07<44:46,  2.02s/it]
 93%|█████████▎| 17758/19090 [8:40:09<39:51,  1.80s/it]
 93%|█████████▎| 17759/19090 [8:40:10<36:24,  1.64s/it][gradnorm] step 17759: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 888.6

 93%|█████████▎| 17760/19090 [8:40:12<43:05,  1.94s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1274.26806640625, 'learning_rate': 7.367569549098591e-07, 'recon_loss': 0.0014320104499347508, 'recon_static': 0.00043943889832007697, 'recon_motion': 0.0009925715465215034, 'recon_norm': 1.2292689561843873, 'recon_norm_std': 0.04151771878823638, 'target_norm': 1.6229405224323272, 'recon_cos': 0.664675185084343, 'gradnorm/compressor.query': 0.49240010831507863, 'gradnorm/compressor.layer0': 16.441645226494405, 'gradnorm/compressor.layer1': 1259.1082161593577, 'gradnorm/compressor.layer2': 189.28056638889265, 'gradnorm/compressor.layer3': 46.13506093695174, 'gradnorm/compressor.layer4': 10.561192538809093, 'gradnorm/compressor.layer5': 5.195267159400366, 'gradnorm/compressor.layer6': 3.5862217875634417, 'gradnorm/compressor.layer7': 3.3832016648352266, 'gradnorm/decoder.layer0': 3.3080964958941963, 'gradnorm/decoder.layer1': 1.977010683655031, 'gradnorm/decoder.layer2': 1.2654015143605928, 'gradnorm/max_param': 888.5592041015625, 'epoch': 9.3}

 93%|█████████▎| 17760/19090 [8:40:12<43:05,  1.94s/it]
 93%|█████████▎| 17761/19090 [8:40:14<38:55,  1.76s/it]
 93%|█████████▎| 17762/19090 [8:40:15<35:44,  1.62s/it]
 93%|█████████▎| 17763/19090 [8:40:18<42:41,  1.93s/it]
 93%|█████████▎| 17764/19090 [8:40:19<38:21,  1.74s/it]
 93%|█████████▎| 17765/19090 [8:40:20<35:19,  1.60s/it]
 93%|█████████▎| 17766/19090 [8:40:23<41:43,  1.89s/it]
 93%|█████████▎| 17767/19090 [8:40:24<37:39,  1.71s/it]
 93%|█████████▎| 17768/19090 [8:40:25<34:48,  1.58s/it]
 93%|█████████▎| 17769/19090 [8:40:28<42:27,  1.93s/it][gradnorm] step 17769: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 842.6

 93%|█████████▎| 17770/19090 [8:40:29<38:10,  1.74s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1207.249755859375, 'learning_rate': 7.257814946215025e-07, 'recon_loss': 0.0014283474069088698, 'recon_static': 0.00044914528261870147, 'recon_motion': 0.0009792021752218715, 'recon_norm': 1.2298603534698487, 'recon_norm_std': 0.04155312329530716, 'target_norm': 1.6210304081439972, 'recon_cos': 0.66455197930336, 'gradnorm/compressor.query': 0.5397540922005372, 'gradnorm/compressor.layer0': 16.317354028164882, 'gradnorm/compressor.layer1': 1194.0218276899295, 'gradnorm/compressor.layer2': 169.21124463965154, 'gradnorm/compressor.layer3': 51.71415434642265, 'gradnorm/compressor.layer4': 10.61035151755925, 'gradnorm/compressor.layer5': 5.550892409326359, 'gradnorm/compressor.layer6': 3.7831616574182587, 'gradnorm/compressor.layer7': 3.7339908929028267, 'gradnorm/decoder.layer0': 3.5753149467401872, 'gradnorm/decoder.layer1': 2.044639239959468, 'gradnorm/decoder.layer2': 1.3089152382465419, 'gradnorm/max_param': 842.5531616210938, 'epoch': 9.31}

 93%|█████████▎| 17770/19090 [8:40:29<38:10,  1.74s/it]
 93%|█████████▎| 17771/19090 [8:40:31<35:11,  1.60s/it]
 93%|█████████▎| 17772/19090 [8:40:34<45:34,  2.07s/it]
 93%|█████████▎| 17773/19090 [8:40:35<40:18,  1.84s/it]
 93%|█████████▎| 17774/19090 [8:40:36<36:37,  1.67s/it]
 93%|█████████▎| 17775/19090 [8:40:40<45:46,  2.09s/it]
 93%|█████████▎| 17776/19090 [8:40:41<40:26,  1.85s/it]
 93%|█████████▎| 17777/19090 [8:40:42<36:43,  1.68s/it]
 93%|█████████▎| 17778/19090 [8:40:45<41:21,  1.89s/it]
 93%|█████████▎| 17779/19090 [8:40:46<37:21,  1.71s/it][gradnorm] step 17779: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 887.8

 93%|█████████▎| 17780/19090 [8:40:47<34:33,  1.58s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1269.124267578125, 'learning_rate': 7.148871954483105e-07, 'recon_loss': 0.0014139071368845179, 'recon_static': 0.0004449146828847006, 'recon_motion': 0.0009689924627309665, 'recon_norm': 1.232374307513237, 'recon_norm_std': 0.04212161209434271, 'target_norm': 1.6178147047758102, 'recon_cos': 0.6671854868531227, 'gradnorm/compressor.query': 0.4135856581798152, 'gradnorm/compressor.layer0': 11.223086228384156, 'gradnorm/compressor.layer1': 1257.9120597243673, 'gradnorm/compressor.layer2': 160.37563936934126, 'gradnorm/compressor.layer3': 48.078224020860006, 'gradnorm/compressor.layer4': 9.944703759968677, 'gradnorm/compressor.layer5': 5.804242575268421, 'gradnorm/compressor.layer6': 3.3978001490171743, 'gradnorm/compressor.layer7': 3.4689519325108558, 'gradnorm/decoder.layer0': 3.739116133015496, 'gradnorm/decoder.layer1': 2.065725636303132, 'gradnorm/decoder.layer2': 1.2304229243388825, 'gradnorm/max_param': 887.7697143554688, 'epoch': 9.31}

 93%|█████████▎| 17780/19090 [8:40:47<34:33,  1.58s/it]
 93%|█████████▎| 17781/19090 [8:40:50<41:43,  1.91s/it]
 93%|█████████▎| 17782/19090 [8:40:51<37:36,  1.73s/it]
 93%|█████████▎| 17783/19090 [8:40:52<34:42,  1.59s/it]
 93%|█████████▎| 17784/19090 [8:40:55<40:17,  1.85s/it]
 93%|█████████▎| 17785/19090 [8:40:56<36:34,  1.68s/it]
 93%|█████████▎| 17786/19090 [8:40:58<37:06,  1.71s/it]
 93%|█████████▎| 17787/19090 [8:41:00<42:14,  1.95s/it]
 93%|█████████▎| 17788/19090 [8:41:02<37:53,  1.75s/it]
 93%|█████████▎| 17789/19090 [8:41:03<37:01,  1.71s/it][gradnorm] step 17789: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 440.6

 93%|█████████▎| 17790/19090 [8:41:06<40:47,  1.88s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 637.7957153320312, 'learning_rate': 7.040740938154877e-07, 'recon_loss': 0.0014194747753208503, 'recon_static': 0.000444693278404884, 'recon_motion': 0.000974781445984263, 'recon_norm': 1.2302075505256653, 'recon_norm_std': 0.042414021026343104, 'target_norm': 1.6205327093601227, 'recon_cos': 0.6668840423226357, 'gradnorm/compressor.query': 1.3667900597685054, 'gradnorm/compressor.layer0': 22.14614047833025, 'gradnorm/compressor.layer1': 626.1663117734777, 'gradnorm/compressor.layer2': 113.50013766511475, 'gradnorm/compressor.layer3': 34.22658300758345, 'gradnorm/compressor.layer4': 9.115739277517957, 'gradnorm/compressor.layer5': 5.35608574382816, 'gradnorm/compressor.layer6': 3.158182289182197, 'gradnorm/compressor.layer7': 3.5083146107612775, 'gradnorm/decoder.layer0': 3.6817702698078696, 'gradnorm/decoder.layer1': 1.969687405198446, 'gradnorm/decoder.layer2': 1.2649491471785792, 'gradnorm/max_param': 440.5704040527344, 'epoch': 9.32}

 93%|█████████▎| 17790/19090 [8:41:06<40:47,  1.88s/it]
 93%|█████████▎| 17791/19090 [8:41:07<36:52,  1.70s/it]
 93%|█████████▎| 17792/19090 [8:41:08<36:37,  1.69s/it]
 93%|█████████▎| 17793/19090 [8:41:11<41:18,  1.91s/it]
 93%|█████████▎| 17794/19090 [8:41:12<37:21,  1.73s/it]
 93%|█████████▎| 17795/19090 [8:41:14<37:29,  1.74s/it]
 93%|█████████▎| 17796/19090 [8:41:16<38:59,  1.81s/it]
 93%|█████████▎| 17797/19090 [8:41:17<35:47,  1.66s/it]
 93%|█████████▎| 17798/19090 [8:41:20<39:53,  1.85s/it]
 93%|█████████▎| 17799/19090 [8:41:21<36:17,  1.69s/it][gradnorm] step 17799: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 763.9

 93%|█████████▎| 17800/19090 [8:41:22<35:45,  1.66s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1099.06640625, 'learning_rate': 6.933422258767613e-07, 'recon_loss': 0.001432803031639196, 'recon_static': 0.00044603193309740164, 'recon_motion': 0.0009867710832622834, 'recon_norm': 1.2298004984855653, 'recon_norm_std': 0.04254967262968421, 'target_norm': 1.6227671772241592, 'recon_cos': 0.6641774296760559, 'gradnorm/compressor.query': 0.3469426299657189, 'gradnorm/compressor.layer0': 9.382415280283602, 'gradnorm/compressor.layer1': 1082.4582555461434, 'gradnorm/compressor.layer2': 184.6699124451517, 'gradnorm/compressor.layer3': 43.52938797166232, 'gradnorm/compressor.layer4': 8.818921069804924, 'gradnorm/compressor.layer5': 4.915187009589096, 'gradnorm/compressor.layer6': 3.2114526191600374, 'gradnorm/compressor.layer7': 3.703827996464277, 'gradnorm/decoder.layer0': 3.7279683466291984, 'gradnorm/decoder.layer1': 1.8809157825540173, 'gradnorm/decoder.layer2': 1.3196792438238885, 'gradnorm/max_param': 763.8678588867188, 'epoch': 9.32}

 93%|█████████▎| 17800/19090 [8:41:22<35:45,  1.66s/it]
 93%|█████████▎| 17801/19090 [8:41:25<40:33,  1.89s/it]
 93%|█████████▎| 17802/19090 [8:41:26<37:05,  1.73s/it]
 93%|█████████▎| 17803/19090 [8:41:28<35:59,  1.68s/it]
 93%|█████████▎| 17804/19090 [8:41:30<42:03,  1.96s/it]
 93%|█████████▎| 17805/19090 [8:41:32<37:48,  1.77s/it]
 93%|█████████▎| 17806/19090 [8:41:33<35:38,  1.67s/it]
 93%|█████████▎| 17807/19090 [8:41:36<40:37,  1.90s/it]
 93%|█████████▎| 17808/19090 [8:41:37<36:54,  1.73s/it]
 93%|█████████▎| 17809/19090 [8:41:38<34:09,  1.60s/it][gradnorm] step 17809: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 866.8

 93%|█████████▎| 17810/19090 [8:41:41<41:59,  1.97s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1241.6641845703125, 'learning_rate': 6.826916275142481e-07, 'recon_loss': 0.0014461349899647757, 'recon_static': 0.00045646382859558797, 'recon_motion': 0.0009896711824694648, 'recon_norm': 1.2426673352718354, 'recon_norm_std': 0.04285065969452262, 'target_norm': 1.6232190400362014, 'recon_cos': 0.6621148183941841, 'gradnorm/compressor.query': 0.38043553170136335, 'gradnorm/compressor.layer0': 11.496641912687142, 'gradnorm/compressor.layer1': 1228.261627260251, 'gradnorm/compressor.layer2': 176.27935814886268, 'gradnorm/compressor.layer3': 41.63239247888146, 'gradnorm/compressor.layer4': 9.834103268583668, 'gradnorm/compressor.layer5': 5.445636143090143, 'gradnorm/compressor.layer6': 2.997766298963801, 'gradnorm/compressor.layer7': 3.265438676839188, 'gradnorm/decoder.layer0': 3.49920522657823, 'gradnorm/decoder.layer1': 1.873459628295021, 'gradnorm/decoder.layer2': 1.29290891203469, 'gradnorm/max_param': 866.8396606445312, 'epoch': 9.33}

 93%|█████████▎| 17810/19090 [8:41:41<41:59,  1.97s/it]
 93%|█████████▎| 17811/19090 [8:41:42<38:14,  1.79s/it]
 93%|█████████▎| 17812/19090 [8:41:44<35:03,  1.65s/it]
 93%|█████████▎| 17813/19090 [8:41:46<41:43,  1.96s/it]
 93%|█████████▎| 17814/19090 [8:41:48<37:26,  1.76s/it]
 93%|█████████▎| 17815/19090 [8:41:49<34:29,  1.62s/it]
 93%|█████████▎| 17816/19090 [8:41:52<41:06,  1.94s/it]
 93%|█████████▎| 17817/19090 [8:41:53<36:55,  1.74s/it]
 93%|█████████▎| 17818/19090 [8:41:54<33:59,  1.60s/it]
 93%|█████████▎| 17819/19090 [8:41:57<39:03,  1.84s/it][gradnorm] step 17819: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 870.5

 93%|█████████▎| 17820/19090 [8:41:58<35:28,  1.68s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1246.9935302734375, 'learning_rate': 6.721223343383403e-07, 'recon_loss': 0.0014530866872519254, 'recon_static': 0.0004689332286943682, 'recon_motion': 0.0009841535094892607, 'recon_norm': 1.2477009683847426, 'recon_norm_std': 0.041850689332932234, 'target_norm': 1.6211168825626374, 'recon_cos': 0.6591096177697182, 'gradnorm/compressor.query': 0.4831928933616924, 'gradnorm/compressor.layer0': 18.354393439853084, 'gradnorm/compressor.layer1': 1233.5020774202205, 'gradnorm/compressor.layer2': 177.65148685347387, 'gradnorm/compressor.layer3': 37.23830563076145, 'gradnorm/compressor.layer4': 10.893944036142175, 'gradnorm/compressor.layer5': 5.0396885209488484, 'gradnorm/compressor.layer6': 3.0123925154256024, 'gradnorm/compressor.layer7': 3.491797781753294, 'gradnorm/decoder.layer0': 3.37873831094575, 'gradnorm/decoder.layer1': 1.729761121056659, 'gradnorm/decoder.layer2': 1.2411006275204, 'gradnorm/max_param': 870.4954223632812, 'epoch': 9.33}

 93%|█████████▎| 17820/19090 [8:41:58<35:28,  1.68s/it]
 93%|█████████▎| 17821/19090 [8:41:59<34:07,  1.61s/it]
 93%|█████████▎| 17822/19090 [8:42:02<41:26,  1.96s/it]
 93%|█████████▎| 17823/19090 [8:42:03<37:06,  1.76s/it]
 93%|█████████▎| 17824/19090 [8:42:05<34:47,  1.65s/it]
 93%|█████████▎| 17825/19090 [8:42:08<41:27,  1.97s/it]
 93%|█████████▎| 17826/19090 [8:42:09<37:06,  1.76s/it]
 93%|█████████▎| 17827/19090 [8:42:10<34:03,  1.62s/it]
 93%|█████████▎| 17828/19090 [8:42:13<41:00,  1.95s/it]
 93%|█████████▎| 17829/19090 [8:42:14<36:45,  1.75s/it][gradnorm] step 17829: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 873.5

 93%|█████████▎| 17830/19090 [8:42:15<33:50,  1.61s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1252.0523681640625, 'learning_rate': 6.616343816875842e-07, 'recon_loss': 0.0014449650392634795, 'recon_static': 0.0004695892464951612, 'recon_motion': 0.000975375778216403, 'recon_norm': 1.2319775134325028, 'recon_norm_std': 0.042249477654695514, 'target_norm': 1.6209338754415512, 'recon_cos': 0.6597216859459877, 'gradnorm/compressor.query': 0.4305946484459534, 'gradnorm/compressor.layer0': 11.819068011156537, 'gradnorm/compressor.layer1': 1237.657010241529, 'gradnorm/compressor.layer2': 181.8103837106121, 'gradnorm/compressor.layer3': 49.67758352302877, 'gradnorm/compressor.layer4': 10.420141713648597, 'gradnorm/compressor.layer5': 5.358010165533466, 'gradnorm/compressor.layer6': 3.0761235115245773, 'gradnorm/compressor.layer7': 3.547959888588471, 'gradnorm/decoder.layer0': 3.5171464806651254, 'gradnorm/decoder.layer1': 1.9646808412638361, 'gradnorm/decoder.layer2': 1.3836782299337569, 'gradnorm/max_param': 873.5438842773438, 'epoch': 9.34}

 93%|█████████▎| 17830/19090 [8:42:15<33:50,  1.61s/it]
 93%|█████████▎| 17831/19090 [8:42:19<43:04,  2.05s/it]
 93%|█████████▎| 17832/19090 [8:42:20<38:12,  1.82s/it]
 93%|█████████▎| 17833/19090 [8:42:21<34:48,  1.66s/it]
 93%|█████████▎| 17834/19090 [8:42:23<38:17,  1.83s/it]
 93%|█████████▎| 17835/19090 [8:42:25<34:49,  1.67s/it]
 93%|█████████▎| 17836/19090 [8:42:26<32:24,  1.55s/it]
 93%|█████████▎| 17837/19090 [8:42:28<37:59,  1.82s/it]
 93%|█████████▎| 17838/19090 [8:42:30<34:40,  1.66s/it]
 93%|█████████▎| 17839/19090 [8:42:31<32:31,  1.56s/it][gradnorm] step 17839: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1172.7

 93%|█████████▎| 17840/19090 [8:42:34<39:19,  1.89s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1679.0067138671875, 'learning_rate': 6.512278046285652e-07, 'recon_loss': 0.0014350717421621085, 'recon_static': 0.00045082272408762946, 'recon_motion': 0.0009842489496804773, 'recon_norm': 1.2421576708555222, 'recon_norm_std': 0.04135877145454288, 'target_norm': 1.6205693691968919, 'recon_cos': 0.6634629547595978, 'gradnorm/compressor.query': 0.846119097808121, 'gradnorm/compressor.layer0': 22.012802560385172, 'gradnorm/compressor.layer1': 1661.8439306986681, 'gradnorm/compressor.layer2': 231.85559217970138, 'gradnorm/compressor.layer3': 54.19364962941922, 'gradnorm/compressor.layer4': 9.734451024231621, 'gradnorm/compressor.layer5': 5.036869299241989, 'gradnorm/compressor.layer6': 3.0153974536176307, 'gradnorm/compressor.layer7': 3.3776465035605985, 'gradnorm/decoder.layer0': 3.4768491476806322, 'gradnorm/decoder.layer1': 1.9943814034211238, 'gradnorm/decoder.layer2': 1.3803824500853823, 'gradnorm/max_param': 1172.7119140625, 'epoch': 9.35}

 93%|█████████▎| 17840/19090 [8:42:34<39:19,  1.89s/it]
 93%|█████████▎| 17841/19090 [8:42:35<35:32,  1.71s/it]
 93%|█████████▎| 17842/19090 [8:42:36<32:52,  1.58s/it]
 93%|█████████▎| 17843/19090 [8:42:39<40:45,  1.96s/it]
 93%|█████████▎| 17844/19090 [8:42:40<36:36,  1.76s/it]
 93%|█████████▎| 17845/19090 [8:42:42<34:43,  1.67s/it]
 93%|█████████▎| 17846/19090 [8:42:45<42:21,  2.04s/it]
 93%|█████████▎| 17847/19090 [8:42:46<37:36,  1.82s/it]
 93%|█████████▎| 17848/19090 [8:42:47<34:20,  1.66s/it]
 93%|█████████▎| 17849/19090 [8:42:50<38:26,  1.86s/it][gradnorm] step 17849: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 994.6

 94%|█████████▎| 17850/19090 [8:42:51<34:54,  1.69s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1427.827392578125, 'learning_rate': 6.409026379557898e-07, 'recon_loss': 0.0014179553661961108, 'recon_static': 0.0004362949024653062, 'recon_motion': 0.000981660420075059, 'recon_norm': 1.225810381770134, 'recon_norm_std': 0.041622867714613675, 'target_norm': 1.6199762225151062, 'recon_cos': 0.667081481218338, 'gradnorm/compressor.query': 0.9615848810655276, 'gradnorm/compressor.layer0': 37.38116098102174, 'gradnorm/compressor.layer1': 1409.962543828736, 'gradnorm/compressor.layer2': 217.30388967699128, 'gradnorm/compressor.layer3': 43.98638982951857, 'gradnorm/compressor.layer4': 9.041666432276662, 'gradnorm/compressor.layer5': 5.01082868990209, 'gradnorm/compressor.layer6': 2.9655638613091657, 'gradnorm/compressor.layer7': 3.0649872624940038, 'gradnorm/decoder.layer0': 3.449271973904548, 'gradnorm/decoder.layer1': 1.9793216912306344, 'gradnorm/decoder.layer2': 1.3575087265661487, 'gradnorm/max_param': 994.6221923828125, 'epoch': 9.35}

 94%|█████████▎| 17850/19090 [8:42:51<34:54,  1.69s/it]
 94%|█████████▎| 17851/19090 [8:42:52<32:55,  1.59s/it]
 94%|█████████▎| 17852/19090 [8:42:55<42:01,  2.04s/it]
 94%|█████████▎| 17853/19090 [8:42:57<37:21,  1.81s/it]
 94%|█████████▎| 17854/19090 [8:42:58<34:07,  1.66s/it]
 94%|█████████▎| 17855/19090 [8:43:00<38:13,  1.86s/it]
 94%|█████████▎| 17856/19090 [8:43:02<34:40,  1.69s/it]
 94%|█████████▎| 17857/19090 [8:43:03<32:11,  1.57s/it]
 94%|█████████▎| 17858/19090 [8:43:05<37:49,  1.84s/it]
 94%|█████████▎| 17859/19090 [8:43:07<34:26,  1.68s/it][gradnorm] step 17859: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 932.3

 94%|█████████▎| 17860/19090 [8:43:08<32:06,  1.57s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1332.136474609375, 'learning_rate': 6.306589161915594e-07, 'recon_loss': 0.0014207169559085741, 'recon_static': 0.00043389162456151096, 'recon_motion': 0.000986825350264553, 'recon_norm': 1.2260026246309281, 'recon_norm_std': 0.042710440792143346, 'target_norm': 1.6215736091136932, 'recon_cos': 0.6668348923325539, 'gradnorm/compressor.query': 0.2926240800894414, 'gradnorm/compressor.layer0': 7.195366311467483, 'gradnorm/compressor.layer1': 1321.0644855570072, 'gradnorm/compressor.layer2': 165.76199522960624, 'gradnorm/compressor.layer3': 41.20090443627649, 'gradnorm/compressor.layer4': 9.01427895267398, 'gradnorm/compressor.layer5': 5.565036324025184, 'gradnorm/compressor.layer6': 2.837805602971914, 'gradnorm/compressor.layer7': 3.373678725833508, 'gradnorm/decoder.layer0': 3.4612724193812645, 'gradnorm/decoder.layer1': 1.9929907282772963, 'gradnorm/decoder.layer2': 1.3180976664429958, 'gradnorm/max_param': 932.3330688476562, 'epoch': 9.36}

 94%|█████████▎| 17860/19090 [8:43:08<32:06,  1.57s/it]
 94%|█████████▎| 17861/19090 [8:43:11<39:41,  1.94s/it]
 94%|█████████▎| 17862/19090 [8:43:12<35:42,  1.74s/it]
 94%|█████████▎| 17863/19090 [8:43:13<32:57,  1.61s/it]
 94%|█████████▎| 17864/19090 [8:43:16<40:01,  1.96s/it]
 94%|█████████▎| 17865/19090 [8:43:17<35:57,  1.76s/it]
 94%|█████████▎| 17866/19090 [8:43:19<33:08,  1.62s/it]
 94%|█████████▎| 17867/19090 [8:43:22<41:01,  2.01s/it]
 94%|█████████▎| 17868/19090 [8:43:23<36:39,  1.80s/it]
 94%|█████████▎| 17869/19090 [8:43:24<33:35,  1.65s/it][gradnorm] step 17869: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 767.0

 94%|█████████▎| 17870/19090 [8:43:27<41:34,  2.04s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1097.214111328125, 'learning_rate': 6.204966735858769e-07, 'recon_loss': 0.0014147975394735112, 'recon_static': 0.0004364977852674201, 'recon_motion': 0.0009782998022274114, 'recon_norm': 1.2269108831882476, 'recon_norm_std': 0.042961942963302135, 'target_norm': 1.6206905514001846, 'recon_cos': 0.6679665610194206, 'gradnorm/compressor.query': 0.3359970771834143, 'gradnorm/compressor.layer0': 7.403577877828152, 'gradnorm/compressor.layer1': 1086.7395042051246, 'gradnorm/compressor.layer2': 145.32174029760654, 'gradnorm/compressor.layer3': 39.56679324799632, 'gradnorm/compressor.layer4': 8.510953689372108, 'gradnorm/compressor.layer5': 5.279213055429676, 'gradnorm/compressor.layer6': 2.875957992830542, 'gradnorm/compressor.layer7': 3.3554846659983415, 'gradnorm/decoder.layer0': 3.4219154111662022, 'gradnorm/decoder.layer1': 2.0279672946361864, 'gradnorm/decoder.layer2': 1.2926572672276118, 'gradnorm/max_param': 766.9910888671875, 'epoch': 9.36}

 94%|█████████▎| 17870/19090 [8:43:27<41:34,  2.04s/it]
 94%|█████████▎| 17871/19090 [8:43:28<37:01,  1.82s/it]
 94%|█████████▎| 17872/19090 [8:43:30<33:47,  1.66s/it]
 94%|█████████▎| 17873/19090 [8:43:32<39:08,  1.93s/it]
 94%|█████████▎| 17874/19090 [8:43:34<35:13,  1.74s/it]
 94%|█████████▎| 17875/19090 [8:43:35<32:28,  1.60s/it]
 94%|█████████▎| 17876/19090 [8:43:38<40:04,  1.98s/it]
 94%|█████████▎| 17877/19090 [8:43:39<35:50,  1.77s/it]
 94%|█████████▎| 17878/19090 [8:43:40<32:51,  1.63s/it]
 94%|█████████▎| 17879/19090 [8:43:43<36:28,  1.81s/it][gradnorm] step 17879: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 842.2

 94%|█████████▎| 17880/19090 [8:43:44<33:18,  1.65s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1205.926025390625, 'learning_rate': 6.104159441163076e-07, 'recon_loss': 0.0014175655582221226, 'recon_static': 0.00044032988516846673, 'recon_motion': 0.000977235748723615, 'recon_norm': 1.2258390218019486, 'recon_norm_std': 0.042908634338527915, 'target_norm': 1.6195256561040878, 'recon_cos': 0.6665032774209976, 'gradnorm/compressor.query': 0.3606126867800334, 'gradnorm/compressor.layer0': 11.335548483590378, 'gradnorm/compressor.layer1': 1193.448417402277, 'gradnorm/compressor.layer2': 168.4831107077205, 'gradnorm/compressor.layer3': 35.621299149876606, 'gradnorm/compressor.layer4': 9.50208861831711, 'gradnorm/compressor.layer5': 5.225237622858125, 'gradnorm/compressor.layer6': 2.783342495825148, 'gradnorm/compressor.layer7': 3.403157081137526, 'gradnorm/decoder.layer0': 3.4867328141783767, 'gradnorm/decoder.layer1': 1.9849118084858155, 'gradnorm/decoder.layer2': 1.2454502752233403, 'gradnorm/max_param': 842.231201171875, 'epoch': 9.37}

 94%|█████████▎| 17880/19090 [8:43:44<33:18,  1.65s/it]
 94%|█████████▎| 17881/19090 [8:43:45<31:05,  1.54s/it]
 94%|█████████▎| 17882/19090 [8:43:48<38:30,  1.91s/it]
 94%|█████████▎| 17883/19090 [8:43:49<34:43,  1.73s/it]
 94%|█████████▎| 17884/19090 [8:43:50<32:01,  1.59s/it]
 94%|█████████▎| 17885/19090 [8:43:53<38:56,  1.94s/it]
 94%|█████████▎| 17886/19090 [8:43:55<34:56,  1.74s/it]
 94%|█████████▎| 17887/19090 [8:43:56<32:13,  1.61s/it]
 94%|█████████▎| 17888/19090 [8:43:58<38:08,  1.90s/it]
 94%|█████████▎| 17889/19090 [8:44:00<34:24,  1.72s/it][gradnorm] step 17889: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 825.9

 94%|█████████▎| 17890/19090 [8:44:01<31:50,  1.59s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1178.7257080078125, 'learning_rate': 6.004167614878819e-07, 'recon_loss': 0.001416066990350373, 'recon_static': 0.000437029712338699, 'recon_motion': 0.0009790373092982917, 'recon_norm': 1.2325684130191803, 'recon_norm_std': 0.042469021398574114, 'target_norm': 1.6202274680137634, 'recon_cos': 0.6682989791035652, 'gradnorm/compressor.query': 0.6194395432779642, 'gradnorm/compressor.layer0': 23.27034664982722, 'gradnorm/compressor.layer1': 1170.6028361489648, 'gradnorm/compressor.layer2': 130.26542914852735, 'gradnorm/compressor.layer3': 37.66730419958485, 'gradnorm/compressor.layer4': 9.59344223473164, 'gradnorm/compressor.layer5': 4.810841460351882, 'gradnorm/compressor.layer6': 2.460838872961657, 'gradnorm/compressor.layer7': 3.689536744478225, 'gradnorm/decoder.layer0': 3.595710286097568, 'gradnorm/decoder.layer1': 2.0049985173093874, 'gradnorm/decoder.layer2': 1.2096932694719555, 'gradnorm/max_param': 825.8834838867188, 'epoch': 9.37}

 94%|█████████▎| 17890/19090 [8:44:01<31:50,  1.59s/it]
 94%|█████████▎| 17891/19090 [8:44:04<39:14,  1.96s/it]
 94%|█████████▎| 17892/19090 [8:44:05<35:12,  1.76s/it]
 94%|█████████▎| 17893/19090 [8:44:06<32:24,  1.62s/it]
 94%|█████████▎| 17894/19090 [8:44:09<37:25,  1.88s/it]
 94%|█████████▎| 17895/19090 [8:44:10<34:07,  1.71s/it]
 94%|█████████▎| 17896/19090 [8:44:11<31:35,  1.59s/it]
 94%|█████████▍| 17897/19090 [8:44:14<37:10,  1.87s/it]
 94%|█████████▍| 17898/19090 [8:44:15<33:44,  1.70s/it]
 94%|█████████▍| 17899/19090 [8:44:17<31:20,  1.58s/it][gradnorm] step 17899: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 829.3

 94%|█████████▍| 17900/19090 [8:44:19<36:23,  1.83s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1190.1199951171875, 'learning_rate': 5.904991591329761e-07, 'recon_loss': 0.0014211235771654173, 'recon_static': 0.00043052925320807844, 'recon_motion': 0.0009905943734338507, 'recon_norm': 1.2228301852941512, 'recon_norm_std': 0.04252847088500857, 'target_norm': 1.6219740837812424, 'recon_cos': 0.6668945476412773, 'gradnorm/compressor.query': 0.42882930711669115, 'gradnorm/compressor.layer0': 8.44940792018353, 'gradnorm/compressor.layer1': 1175.1083460685657, 'gradnorm/compressor.layer2': 181.85474543574605, 'gradnorm/compressor.layer3': 46.689947225018365, 'gradnorm/compressor.layer4': 10.388301916647675, 'gradnorm/compressor.layer5': 5.311215054564724, 'gradnorm/compressor.layer6': 3.1245444919484706, 'gradnorm/compressor.layer7': 3.890296278705312, 'gradnorm/decoder.layer0': 4.021851695600096, 'gradnorm/decoder.layer1': 2.1105620258528117, 'gradnorm/decoder.layer2': 1.280635919184605, 'gradnorm/max_param': 829.3300170898438, 'epoch': 9.38}

 94%|█████████▍| 17900/19090 [8:44:19<36:23,  1.83s/it]
 94%|█████████▍| 17901/19090 [8:44:20<33:14,  1.68s/it]
 94%|█████████▍| 17902/19090 [8:44:22<30:59,  1.57s/it]
 94%|█████████▍| 17903/19090 [8:44:24<36:18,  1.84s/it]
 94%|█████████▍| 17904/19090 [8:44:26<33:53,  1.71s/it]
 94%|█████████▍| 17905/19090 [8:44:27<31:24,  1.59s/it]
 94%|█████████▍| 17906/19090 [8:44:30<39:14,  1.99s/it]
 94%|█████████▍| 17907/19090 [8:44:31<35:09,  1.78s/it]
 94%|█████████▍| 17908/19090 [8:44:32<32:14,  1.64s/it]
 94%|█████████▍| 17909/19090 [8:44:35<37:58,  1.93s/it][gradnorm] step 17909: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 982.2

 94%|█████████▍| 17910/19090 [8:44:36<34:13,  1.74s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1407.4407958984375, 'learning_rate': 5.806631702112014e-07, 'recon_loss': 0.0014214623108273372, 'recon_static': 0.00043146918105776424, 'recon_motion': 0.0009899931348627434, 'recon_norm': 1.2246494174003602, 'recon_norm_std': 0.04286073949187994, 'target_norm': 1.6220786303281785, 'recon_cos': 0.6666957840323449, 'gradnorm/compressor.query': 0.36774637423555545, 'gradnorm/compressor.layer0': 11.72533343721479, 'gradnorm/compressor.layer1': 1391.6807284718718, 'gradnorm/compressor.layer2': 203.02148323822777, 'gradnorm/compressor.layer3': 50.63862904891013, 'gradnorm/compressor.layer4': 10.69442863518233, 'gradnorm/compressor.layer5': 5.790963255981591, 'gradnorm/compressor.layer6': 3.3039350802899423, 'gradnorm/compressor.layer7': 4.01793868209207, 'gradnorm/decoder.layer0': 3.802280549808965, 'gradnorm/decoder.layer1': 1.9270109077109268, 'gradnorm/decoder.layer2': 1.2289808693243447, 'gradnorm/max_param': 982.2212524414062, 'epoch': 9.38}

 94%|█████████▍| 17910/19090 [8:44:36<34:13,  1.74s/it]
 94%|█████████▍| 17911/19090 [8:44:38<31:40,  1.61s/it]
 94%|█████████▍| 17912/19090 [8:44:40<37:05,  1.89s/it]
 94%|█████████▍| 17913/19090 [8:44:41<33:31,  1.71s/it]
 94%|█████████▍| 17914/19090 [8:44:43<31:01,  1.58s/it]
 94%|█████████▍| 17915/19090 [8:44:45<37:37,  1.92s/it]
 94%|█████████▍| 17916/19090 [8:44:47<33:53,  1.73s/it]
 94%|█████████▍| 17917/19090 [8:44:48<31:15,  1.60s/it]
 94%|█████████▍| 17918/19090 [8:44:51<38:26,  1.97s/it]
 94%|█████████▍| 17919/19090 [8:44:52<34:27,  1.77s/it][gradnorm] step 17919: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 827.0

 94%|█████████▍| 17920/19090 [8:44:53<31:38,  1.62s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1185.34521484375, 'learning_rate': 5.709088276092955e-07, 'recon_loss': 0.001406923253671266, 'recon_static': 0.00043385299577494154, 'recon_motion': 0.0009730702091474086, 'recon_norm': 1.2186937421560287, 'recon_norm_std': 0.04303770698606968, 'target_norm': 1.6190021067857743, 'recon_cos': 0.6687911212444305, 'gradnorm/compressor.query': 0.40232163896265516, 'gradnorm/compressor.layer0': 9.227727520911923, 'gradnorm/compressor.layer1': 1171.997066975852, 'gradnorm/compressor.layer2': 171.7062637635127, 'gradnorm/compressor.layer3': 41.301635236903, 'gradnorm/compressor.layer4': 10.732939911576315, 'gradnorm/compressor.layer5': 5.749342839378845, 'gradnorm/compressor.layer6': 3.3723176257320784, 'gradnorm/compressor.layer7': 3.745756036613655, 'gradnorm/decoder.layer0': 3.649696933235385, 'gradnorm/decoder.layer1': 1.9167069831699042, 'gradnorm/decoder.layer2': 1.2010321310987648, 'gradnorm/max_param': 827.0134887695312, 'epoch': 9.39}

 94%|█████████▍| 17920/19090 [8:44:53<31:38,  1.62s/it]
 94%|█████████▍| 17921/19090 [8:44:56<37:37,  1.93s/it]
 94%|█████████▍| 17922/19090 [8:44:57<34:18,  1.76s/it]
 94%|█████████▍| 17923/19090 [8:44:59<31:32,  1.62s/it]
 94%|█████████▍| 17924/19090 [8:45:01<36:25,  1.87s/it]
 94%|█████████▍| 17925/19090 [8:45:03<33:29,  1.72s/it]
 94%|█████████▍| 17926/19090 [8:45:04<30:57,  1.60s/it]
 94%|█████████▍| 17927/19090 [8:45:07<37:39,  1.94s/it]
 94%|█████████▍| 17928/19090 [8:45:08<33:51,  1.75s/it]
 94%|█████████▍| 17929/19090 [8:45:09<31:12,  1.61s/it][gradnorm] step 17929: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 841.7

 94%|█████████▍| 17930/19090 [8:45:12<37:55,  1.96s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1204.7716064453125, 'learning_rate': 5.612361639410119e-07, 'recon_loss': 0.001405563170555979, 'recon_static': 0.0004304123875044752, 'recon_motion': 0.0009751507226610556, 'recon_norm': 1.2231154561042785, 'recon_norm_std': 0.042310684360563754, 'target_norm': 1.618989259004593, 'recon_cos': 0.6697625041007995, 'gradnorm/compressor.query': 0.32806715001179054, 'gradnorm/compressor.layer0': 10.211627869737518, 'gradnorm/compressor.layer1': 1192.6809087094503, 'gradnorm/compressor.layer2': 163.74224695386343, 'gradnorm/compressor.layer3': 43.475368497755426, 'gradnorm/compressor.layer4': 10.081249894049325, 'gradnorm/compressor.layer5': 6.112625389139205, 'gradnorm/compressor.layer6': 3.351698083513756, 'gradnorm/compressor.layer7': 3.4377893499491825, 'gradnorm/decoder.layer0': 3.64298522147051, 'gradnorm/decoder.layer1': 1.9690795501518088, 'gradnorm/decoder.layer2': 1.1906807112988584, 'gradnorm/max_param': 841.7047729492188, 'epoch': 9.39}

 94%|█████████▍| 17930/19090 [8:45:12<37:55,  1.96s/it]
 94%|█████████▍| 17931/19090 [8:45:13<34:01,  1.76s/it]
 94%|█████████▍| 17932/19090 [8:45:15<31:15,  1.62s/it]
 94%|█████████▍| 17933/19090 [8:45:17<35:31,  1.84s/it]
 94%|█████████▍| 17934/19090 [8:45:18<32:28,  1.69s/it]
 94%|█████████▍| 17935/19090 [8:45:20<30:09,  1.57s/it]
 94%|█████████▍| 17936/19090 [8:45:22<33:31,  1.74s/it]
 94%|█████████▍| 17937/19090 [8:45:23<33:06,  1.72s/it]
 94%|█████████▍| 17938/19090 [8:45:25<30:35,  1.59s/it]
 94%|█████████▍| 17939/19090 [8:45:27<36:06,  1.88s/it][gradnorm] step 17939: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 752.3

 94%|█████████▍| 17940/19090 [8:45:29<33:31,  1.75s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1079.7283935546875, 'learning_rate': 5.516452115470028e-07, 'recon_loss': 0.0014221272809663787, 'recon_static': 0.00042986420303350317, 'recon_motion': 0.000992263061925769, 'recon_norm': 1.2244422376155852, 'recon_norm_std': 0.04195055523887277, 'target_norm': 1.6232316315174102, 'recon_cos': 0.6675117462873459, 'gradnorm/compressor.query': 0.3091729559198698, 'gradnorm/compressor.layer0': 8.955286640497699, 'gradnorm/compressor.layer1': 1065.982762524798, 'gradnorm/compressor.layer2': 165.6570075204336, 'gradnorm/compressor.layer3': 42.69866538184332, 'gradnorm/compressor.layer4': 8.935740477752372, 'gradnorm/compressor.layer5': 5.70858611872095, 'gradnorm/compressor.layer6': 2.9357910054736225, 'gradnorm/compressor.layer7': 3.21360812538379, 'gradnorm/decoder.layer0': 3.4739575860262426, 'gradnorm/decoder.layer1': 1.9257044283572953, 'gradnorm/decoder.layer2': 1.097080554883785, 'gradnorm/max_param': 752.33154296875, 'epoch': 9.4}

 94%|█████████▍| 17940/19090 [8:45:29<33:31,  1.75s/it]
 94%|█████████▍| 17941/19090 [8:45:30<30:50,  1.61s/it]
 94%|█████████▍| 17942/19090 [8:45:32<35:24,  1.85s/it]
 94%|█████████▍| 17943/19090 [8:45:34<33:37,  1.76s/it]
 94%|█████████▍| 17944/19090 [8:45:35<30:56,  1.62s/it]
 94%|█████████▍| 17945/19090 [8:45:38<37:24,  1.96s/it]
 94%|█████████▍| 17946/19090 [8:45:39<33:30,  1.76s/it]
 94%|█████████▍| 17947/19090 [8:45:41<30:49,  1.62s/it]
 94%|█████████▍| 17948/19090 [8:45:43<36:46,  1.93s/it]
 94%|█████████▍| 17949/19090 [8:45:44<33:03,  1.74s/it][gradnorm] step 17949: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 707.3

 94%|█████████▍| 17950/19090 [8:45:46<30:27,  1.60s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1010.4179077148438, 'learning_rate': 5.421360024947225e-07, 'recon_loss': 0.001415217894827947, 'recon_static': 0.0004279081171262078, 'recon_motion': 0.0009873097369563766, 'recon_norm': 1.2253825187683105, 'recon_norm_std': 0.04273184956982732, 'target_norm': 1.6220586389303207, 'recon_cos': 0.6686465129256248, 'gradnorm/compressor.query': 0.367886385436549, 'gradnorm/compressor.layer0': 9.055738992009777, 'gradnorm/compressor.layer1': 1002.393323501309, 'gradnorm/compressor.layer2': 121.48735225451455, 'gradnorm/compressor.layer3': 34.164845884855644, 'gradnorm/compressor.layer4': 9.012201303725256, 'gradnorm/compressor.layer5': 5.425249176631433, 'gradnorm/compressor.layer6': 2.6448465043871257, 'gradnorm/compressor.layer7': 3.2028114723652794, 'gradnorm/decoder.layer0': 3.3366874191395923, 'gradnorm/decoder.layer1': 1.8358610746019557, 'gradnorm/decoder.layer2': 1.031631225866669, 'gradnorm/max_param': 707.3469848632812, 'epoch': 9.4}

 94%|█████████▍| 17950/19090 [8:45:46<30:27,  1.60s/it]
 94%|█████████▍| 17951/19090 [8:45:49<37:57,  2.00s/it]
 94%|█████████▍| 17952/19090 [8:45:50<33:50,  1.78s/it]
 94%|█████████▍| 17953/19090 [8:45:51<30:59,  1.64s/it]
 94%|█████████▍| 17954/19090 [8:45:54<36:41,  1.94s/it]
 94%|█████████▍| 17955/19090 [8:45:55<32:58,  1.74s/it]
 94%|█████████▍| 17956/19090 [8:45:56<30:26,  1.61s/it]
 94%|█████████▍| 17957/19090 [8:45:59<38:04,  2.02s/it]
 94%|█████████▍| 17958/19090 [8:46:01<33:56,  1.80s/it]
 94%|█████████▍| 17959/19090 [8:46:02<31:00,  1.65s/it][gradnorm] step 17959: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 801.8

 94%|█████████▍| 17960/19090 [8:46:05<37:17,  1.98s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1148.722412109375, 'learning_rate': 5.32708568578319e-07, 'recon_loss': 0.001410278127877973, 'recon_static': 0.00042619596861186436, 'recon_motion': 0.000984082103241235, 'recon_norm': 1.2227890014648437, 'recon_norm_std': 0.043771062325686214, 'target_norm': 1.6211957603693008, 'recon_cos': 0.6692350640892982, 'gradnorm/compressor.query': 0.3189638519038102, 'gradnorm/compressor.layer0': 10.499454665781865, 'gradnorm/compressor.layer1': 1136.0877496918977, 'gradnorm/compressor.layer2': 163.11178706066892, 'gradnorm/compressor.layer3': 44.92202309758196, 'gradnorm/compressor.layer4': 8.598397180988515, 'gradnorm/compressor.layer5': 5.191449236207472, 'gradnorm/compressor.layer6': 2.6699024079735625, 'gradnorm/compressor.layer7': 3.1618670346547506, 'gradnorm/decoder.layer0': 3.3689332760017456, 'gradnorm/decoder.layer1': 1.9076432936455954, 'gradnorm/decoder.layer2': 1.0967934906934016, 'gradnorm/max_param': 801.7794189453125, 'epoch': 9.41}

 94%|█████████▍| 17960/19090 [8:46:05<37:17,  1.98s/it]
 94%|█████████▍| 17961/19090 [8:46:06<33:30,  1.78s/it]
 94%|█████████▍| 17962/19090 [8:46:07<30:41,  1.63s/it]
 94%|█████████▍| 17963/19090 [8:46:10<35:37,  1.90s/it]
 94%|█████████▍| 17964/19090 [8:46:11<32:10,  1.71s/it]
 94%|█████████▍| 17965/19090 [8:46:12<29:46,  1.59s/it]
 94%|█████████▍| 17966/19090 [8:46:15<36:06,  1.93s/it]
 94%|█████████▍| 17967/19090 [8:46:17<32:30,  1.74s/it]
 94%|█████████▍| 17968/19090 [8:46:18<29:59,  1.60s/it]
 94%|█████████▍| 17969/19090 [8:46:21<36:12,  1.94s/it][gradnorm] step 17969: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 733.6

 94%|█████████▍| 17970/19090 [8:46:22<32:32,  1.74s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1054.843017578125, 'learning_rate': 5.233629413185198e-07, 'recon_loss': 0.00141639641369693, 'recon_static': 0.0004336924401286524, 'recon_motion': 0.0009827039801166392, 'recon_norm': 1.2224327117204665, 'recon_norm_std': 0.04401864856481552, 'target_norm': 1.6220003515481949, 'recon_cos': 0.6678668394684791, 'gradnorm/compressor.query': 0.403705809864687, 'gradnorm/compressor.layer0': 11.417853152211617, 'gradnorm/compressor.layer1': 1039.6692983829041, 'gradnorm/compressor.layer2': 171.92310690392958, 'gradnorm/compressor.layer3': 43.891683248634216, 'gradnorm/compressor.layer4': 10.2783295446069, 'gradnorm/compressor.layer5': 4.869349212401934, 'gradnorm/compressor.layer6': 2.7258598056099816, 'gradnorm/compressor.layer7': 3.2621277695453417, 'gradnorm/decoder.layer0': 3.6844758386377525, 'gradnorm/decoder.layer1': 2.106332010998708, 'gradnorm/decoder.layer2': 1.1607043066748608, 'gradnorm/max_param': 733.6298217773438, 'epoch': 9.41}

 94%|█████████▍| 17970/19090 [8:46:22<32:32,  1.74s/it]
 94%|█████████▍| 17971/19090 [8:46:23<29:59,  1.61s/it]
 94%|█████████▍| 17972/19090 [8:46:26<36:24,  1.95s/it]
 94%|█████████▍| 17973/19090 [8:46:27<32:39,  1.75s/it]
 94%|█████████▍| 17974/19090 [8:46:28<30:03,  1.62s/it]
 94%|█████████▍| 17975/19090 [8:46:31<37:09,  2.00s/it]
 94%|█████████▍| 17976/19090 [8:46:33<33:08,  1.78s/it]
 94%|█████████▍| 17977/19090 [8:46:34<30:20,  1.64s/it]
 94%|█████████▍| 17978/19090 [8:46:37<37:05,  2.00s/it]
 94%|█████████▍| 17979/19090 [8:46:38<33:04,  1.79s/it][gradnorm] step 17979: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 659.1

 94%|█████████▍| 17980/19090 [8:46:39<30:18,  1.64s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 942.8574829101562, 'learning_rate': 5.140991519625326e-07, 'recon_loss': 0.0014152544812532143, 'recon_static': 0.00043916154027101584, 'recon_motion': 0.0009760929591720924, 'recon_norm': 1.235319060087204, 'recon_norm_std': 0.042764703184366225, 'target_norm': 1.6201523840427399, 'recon_cos': 0.6682218179106713, 'gradnorm/compressor.query': 0.3980355947167084, 'gradnorm/compressor.layer0': 14.920282760272487, 'gradnorm/compressor.layer1': 934.0700589356239, 'gradnorm/compressor.layer2': 122.64044484156726, 'gradnorm/compressor.layer3': 32.927658782407, 'gradnorm/compressor.layer4': 9.188374380426758, 'gradnorm/compressor.layer5': 4.720625548855762, 'gradnorm/compressor.layer6': 2.8034758009147676, 'gradnorm/compressor.layer7': 3.2414420500327785, 'gradnorm/decoder.layer0': 3.7436888675442965, 'gradnorm/decoder.layer1': 2.247662919942528, 'gradnorm/decoder.layer2': 1.2112951427350047, 'gradnorm/max_param': 659.0947265625, 'epoch': 9.42}

 94%|█████████▍| 17980/19090 [8:46:39<30:18,  1.64s/it]
 94%|█████████▍| 17981/19090 [8:46:42<36:38,  1.98s/it]
 94%|█████████▍| 17982/19090 [8:46:43<32:45,  1.77s/it]
 94%|█████████▍| 17983/19090 [8:46:45<30:04,  1.63s/it]
 94%|█████████▍| 17984/19090 [8:46:48<37:01,  2.01s/it]
 94%|█████████▍| 17985/19090 [8:46:49<32:59,  1.79s/it]
 94%|█████████▍| 17986/19090 [8:46:50<30:12,  1.64s/it]
 94%|█████████▍| 17987/19090 [8:46:53<36:39,  1.99s/it]
 94%|█████████▍| 17988/19090 [8:46:54<32:43,  1.78s/it]
 94%|█████████▍| 17989/19090 [8:46:56<30:00,  1.64s/it][gradnorm] step 17989: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 779.5

 94%|█████████▍| 17990/19090 [8:46:58<36:21,  1.98s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1112.945068359375, 'learning_rate': 5.04917231483934e-07, 'recon_loss': 0.0014206812571501358, 'recon_static': 0.00044678756239591166, 'recon_motion': 0.0009738937296788208, 'recon_norm': 1.2378157705068589, 'recon_norm_std': 0.04385727979242802, 'target_norm': 1.6187954068183898, 'recon_cos': 0.6664059028029442, 'gradnorm/compressor.query': 0.31680387072946936, 'gradnorm/compressor.layer0': 10.08148033413442, 'gradnorm/compressor.layer1': 1104.5280100567843, 'gradnorm/compressor.layer2': 130.07097671800193, 'gradnorm/compressor.layer3': 38.95482660242502, 'gradnorm/compressor.layer4': 8.38676741878636, 'gradnorm/compressor.layer5': 4.695433959405528, 'gradnorm/compressor.layer6': 2.6473984382031706, 'gradnorm/compressor.layer7': 3.0233047022474984, 'gradnorm/decoder.layer0': 3.4613747761816436, 'gradnorm/decoder.layer1': 2.2258775159344197, 'gradnorm/decoder.layer2': 1.121125277298557, 'gradnorm/max_param': 779.5358276367188, 'epoch': 9.42}

 94%|█████████▍| 17990/19090 [8:46:58<36:21,  1.98s/it]
 94%|█████████▍| 17991/19090 [8:47:00<32:29,  1.77s/it]
 94%|█████████▍| 17992/19090 [8:47:01<29:47,  1.63s/it]
 94%|█████████▍| 17993/19090 [8:47:03<31:57,  1.75s/it]
 94%|█████████▍| 17994/19090 [8:47:04<29:25,  1.61s/it]
 94%|█████████▍| 17995/19090 [8:47:06<27:39,  1.52s/it]
 94%|█████████▍| 17996/19090 [8:47:07<28:45,  1.58s/it]
 94%|█████████▍| 17997/19090 [8:47:09<27:09,  1.49s/it]
 94%|█████████▍| 17998/19090 [8:47:10<26:03,  1.43s/it]
 94%|█████████▍| 17999/19090 [8:47:12<31:25,  1.73s/it][gradnorm] step 17999: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 644.3

 94%|█████████▍| 18000/19090 [8:47:14<29:02,  1.60s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 923.130126953125, 'learning_rate': 4.958172105825803e-07, 'recon_loss': 0.0014352442289236934, 'recon_static': 0.0004529672376520466, 'recon_motion': 0.000982277002185583, 'recon_norm': 1.2345232874155045, 'recon_norm_std': 0.0423297961242497, 'target_norm': 1.6218725621700287, 'recon_cos': 0.6632333517074585, 'gradnorm/compressor.query': 0.46915615287957113, 'gradnorm/compressor.layer0': 13.029471808945948, 'gradnorm/compressor.layer1': 913.289726757068, 'gradnorm/compressor.layer2': 129.4289968839093, 'gradnorm/compressor.layer3': 31.97067251978268, 'gradnorm/compressor.layer4': 8.557457100183697, 'gradnorm/compressor.layer5': 4.501623920311633, 'gradnorm/compressor.layer6': 2.706648718378225, 'gradnorm/compressor.layer7': 3.044925790017127, 'gradnorm/decoder.layer0': 3.323207894281994, 'gradnorm/decoder.layer1': 2.174816198365642, 'gradnorm/decoder.layer2': 1.1437586236847506, 'gradnorm/max_param': 644.2506103515625, 'epoch': 9.43}

 94%|█████████▍| 18000/19090 [8:47:14<29:02,  1.60s/it]
 94%|█████████▍| 18001/19090 [8:47:18<42:00,  2.31s/it]
 94%|█████████▍| 18002/19090 [8:47:19<36:29,  2.01s/it]
 94%|█████████▍| 18003/19090 [8:47:20<32:35,  1.80s/it]
 94%|█████████▍| 18004/19090 [8:47:21<29:49,  1.65s/it]
 94%|█████████▍| 18005/19090 [8:47:23<27:54,  1.54s/it]
 94%|█████████▍| 18006/19090 [8:47:24<26:32,  1.47s/it]
 94%|█████████▍| 18007/19090 [8:47:25<25:35,  1.42s/it]
 94%|█████████▍| 18008/19090 [8:47:27<24:52,  1.38s/it]
 94%|█████████▍| 18009/19090 [8:47:28<24:23,  1.35s/it][gradnorm] step 18009: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 921.3

 94%|█████████▍| 18010/19090 [8:47:29<24:03,  1.34s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1313.81494140625, 'learning_rate': 4.867991196844918e-07, 'recon_loss': 0.001430312223965302, 'recon_static': 0.00044157273005112073, 'recon_motion': 0.000988739558670204, 'recon_norm': 1.228732407093048, 'recon_norm_std': 0.041234986204653976, 'target_norm': 1.620240616798401, 'recon_cos': 0.6633080586791038, 'gradnorm/compressor.query': 0.40961858840525006, 'gradnorm/compressor.layer0': 9.31712850973905, 'gradnorm/compressor.layer1': 1305.4487734108911, 'gradnorm/compressor.layer2': 140.5439697101053, 'gradnorm/compressor.layer3': 43.70222698621533, 'gradnorm/compressor.layer4': 9.694559785824108, 'gradnorm/compressor.layer5': 5.3782987895684125, 'gradnorm/compressor.layer6': 2.9182297106053636, 'gradnorm/compressor.layer7': 3.3873228526512293, 'gradnorm/decoder.layer0': 3.766154904891614, 'gradnorm/decoder.layer1': 2.20178549877807, 'gradnorm/decoder.layer2': 1.3255125044265612, 'gradnorm/max_param': 921.2772827148438, 'epoch': 9.43}

 94%|█████████▍| 18010/19090 [8:47:29<24:03,  1.34s/it]
 94%|█████████▍| 18011/19090 [8:47:31<23:48,  1.32s/it]
 94%|█████████▍| 18012/19090 [8:47:32<23:34,  1.31s/it]
 94%|█████████▍| 18013/19090 [8:47:33<23:26,  1.31s/it]
 94%|█████████▍| 18014/19090 [8:47:34<23:20,  1.30s/it]
 94%|█████████▍| 18015/19090 [8:47:36<23:17,  1.30s/it]
 94%|█████████▍| 18016/19090 [8:47:37<23:14,  1.30s/it]
 94%|█████████▍| 18017/19090 [8:47:38<23:11,  1.30s/it]
 94%|█████████▍| 18018/19090 [8:47:40<23:09,  1.30s/it]
 94%|█████████▍| 18019/19090 [8:47:41<23:09,  1.30s/it][gradnorm] step 18019: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 894.0

 94%|█████████▍| 18020/19090 [8:47:42<23:23,  1.31s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1277.49853515625, 'learning_rate': 4.778629889417491e-07, 'recon_loss': 0.0014337820612126961, 'recon_static': 0.0004525247844867408, 'recon_motion': 0.0009812572738155723, 'recon_norm': 1.241762164235115, 'recon_norm_std': 0.041954856179654595, 'target_norm': 1.619752225279808, 'recon_cos': 0.6633972242474556, 'gradnorm/compressor.query': 0.4423120538909547, 'gradnorm/compressor.layer0': 12.051176933333757, 'gradnorm/compressor.layer1': 1266.7787395595176, 'gradnorm/compressor.layer2': 158.70594245923812, 'gradnorm/compressor.layer3': 42.06512343230435, 'gradnorm/compressor.layer4': 10.18100948342366, 'gradnorm/compressor.layer5': 5.184506207369976, 'gradnorm/compressor.layer6': 2.968352763051509, 'gradnorm/compressor.layer7': 3.3678994543512046, 'gradnorm/decoder.layer0': 3.8032463211104965, 'gradnorm/decoder.layer1': 2.1227311756780596, 'gradnorm/decoder.layer2': 1.3346770290927745, 'gradnorm/max_param': 893.9619750976562, 'epoch': 9.44}

 94%|█████████▍| 18020/19090 [8:47:42<23:23,  1.31s/it]
 94%|█████████▍| 18021/19090 [8:47:44<24:33,  1.38s/it]
 94%|█████████▍| 18022/19090 [8:47:45<24:57,  1.40s/it]
 94%|█████████▍| 18023/19090 [8:47:47<24:48,  1.40s/it]
 94%|█████████▍| 18024/19090 [8:47:48<27:16,  1.54s/it]
 94%|█████████▍| 18025/19090 [8:47:50<27:48,  1.57s/it]
 94%|█████████▍| 18026/19090 [8:47:52<27:20,  1.54s/it]
 94%|█████████▍| 18027/19090 [8:47:54<31:44,  1.79s/it]
 94%|█████████▍| 18028/19090 [8:47:55<29:01,  1.64s/it]
 94%|█████████▍| 18029/19090 [8:47:57<29:23,  1.66s/it][gradnorm] step 18029: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 999.6

 94%|█████████▍| 18030/19090 [8:47:59<32:38,  1.85s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1427.7840576171875, 'learning_rate': 4.690088482324051e-07, 'recon_loss': 0.0014406452159164474, 'recon_static': 0.0004508143036218826, 'recon_motion': 0.0009898308708216064, 'recon_norm': 1.2240211427211762, 'recon_norm_std': 0.041754096187651156, 'target_norm': 1.621415662765503, 'recon_cos': 0.6604270428419113, 'gradnorm/compressor.query': 0.5792018809922246, 'gradnorm/compressor.layer0': 22.133677704364974, 'gradnorm/compressor.layer1': 1416.5937579277977, 'gradnorm/compressor.layer2': 170.555005102826, 'gradnorm/compressor.layer3': 45.42075162009514, 'gradnorm/compressor.layer4': 10.788655532354381, 'gradnorm/compressor.layer5': 5.3457917891464195, 'gradnorm/compressor.layer6': 3.0211177472803223, 'gradnorm/compressor.layer7': 3.443285010319721, 'gradnorm/decoder.layer0': 3.8776163682132925, 'gradnorm/decoder.layer1': 2.092802032403513, 'gradnorm/decoder.layer2': 1.4209755368633676, 'gradnorm/max_param': 999.6309204101562, 'epoch': 9.44}

 94%|█████████▍| 18030/19090 [8:47:59<32:38,  1.85s/it]
 94%|█████████▍| 18031/19090 [8:48:01<29:50,  1.69s/it]
 94%|█████████▍| 18032/19090 [8:48:02<29:33,  1.68s/it]
 94%|█████████▍| 18033/19090 [8:48:05<35:06,  1.99s/it]
 94%|█████████▍| 18034/19090 [8:48:06<31:20,  1.78s/it]
 94%|█████████▍| 18035/19090 [8:48:08<28:42,  1.63s/it]
 94%|█████████▍| 18036/19090 [8:48:10<34:46,  1.98s/it]
 94%|█████████▍| 18037/19090 [8:48:12<32:49,  1.87s/it]
 94%|█████████▍| 18038/19090 [8:48:13<29:44,  1.70s/it]
 94%|█████████▍| 18039/19090 [8:48:15<32:31,  1.86s/it][gradnorm] step 18039: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 970.7

 94%|█████████▍| 18040/19090 [8:48:17<32:07,  1.84s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1390.3663330078125, 'learning_rate': 4.602367271603791e-07, 'recon_loss': 0.0014333202212583273, 'recon_static': 0.0004425137551152147, 'recon_motion': 0.0009908065243507736, 'recon_norm': 1.2325703859329225, 'recon_norm_std': 0.04217771394178271, 'target_norm': 1.6215196698904037, 'recon_cos': 0.6637252807617188, 'gradnorm/compressor.query': 0.5258277577750974, 'gradnorm/compressor.layer0': 10.33636893527477, 'gradnorm/compressor.layer1': 1375.3838100690293, 'gradnorm/compressor.layer2': 196.314131639192, 'gradnorm/compressor.layer3': 50.7354793727969, 'gradnorm/compressor.layer4': 11.879484273588872, 'gradnorm/compressor.layer5': 5.906687563880023, 'gradnorm/compressor.layer6': 2.9709653920430936, 'gradnorm/compressor.layer7': 3.2761843862006996, 'gradnorm/decoder.layer0': 3.955550825340024, 'gradnorm/decoder.layer1': 2.0450236560227433, 'gradnorm/decoder.layer2': 1.4216019085218148, 'gradnorm/max_param': 970.717529296875, 'epoch': 9.45}

 94%|█████████▍| 18040/19090 [8:48:17<32:07,  1.84s/it]
 95%|█████████▍| 18041/19090 [8:48:19<29:14,  1.67s/it]
 95%|█████████▍| 18042/19090 [8:48:21<31:26,  1.80s/it]
 95%|█████████▍| 18043/19090 [8:48:22<31:41,  1.82s/it]
 95%|█████████▍| 18044/19090 [8:48:24<28:57,  1.66s/it]
 95%|█████████▍| 18045/19090 [8:48:26<33:12,  1.91s/it]
 95%|█████████▍| 18046/19090 [8:48:28<31:44,  1.82s/it]
 95%|█████████▍| 18047/19090 [8:48:29<28:57,  1.67s/it]
 95%|█████████▍| 18048/19090 [8:48:32<32:52,  1.89s/it]
 95%|█████████▍| 18049/19090 [8:48:33<31:16,  1.80s/it][gradnorm] step 18049: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 875.5

 95%|█████████▍| 18050/19090 [8:48:34<28:35,  1.65s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1252.1805419921875, 'learning_rate': 4.515466550553515e-07, 'recon_loss': 0.0014247042359784245, 'recon_static': 0.00043697028959286397, 'recon_motion': 0.0009877339776721782, 'recon_norm': 1.2289853751659394, 'recon_norm_std': 0.041774496994912626, 'target_norm': 1.6212614983320237, 'recon_cos': 0.665806595981121, 'gradnorm/compressor.query': 0.3816399034475731, 'gradnorm/compressor.layer0': 9.558005878811466, 'gradnorm/compressor.layer1': 1240.527710291068, 'gradnorm/compressor.layer2': 163.96828247026923, 'gradnorm/compressor.layer3': 43.30080926504507, 'gradnorm/compressor.layer4': 11.06112541667431, 'gradnorm/compressor.layer5': 5.6057814454377075, 'gradnorm/compressor.layer6': 3.142759897535854, 'gradnorm/compressor.layer7': 3.30481995662074, 'gradnorm/decoder.layer0': 3.732244290728932, 'gradnorm/decoder.layer1': 2.0399126056575816, 'gradnorm/decoder.layer2': 1.384379388186559, 'gradnorm/max_param': 875.4799194335938, 'epoch': 9.46}

 95%|█████████▍| 18050/19090 [8:48:34<28:35,  1.65s/it]
 95%|█████████▍| 18051/19090 [8:48:37<31:41,  1.83s/it]
 95%|█████████▍| 18052/19090 [8:48:38<31:07,  1.80s/it]
 95%|█████████▍| 18053/19090 [8:48:40<28:25,  1.64s/it]
 95%|█████████▍| 18054/19090 [8:48:42<29:58,  1.74s/it]
 95%|█████████▍| 18055/19090 [8:48:43<29:23,  1.70s/it]
 95%|█████████▍| 18056/19090 [8:48:45<28:13,  1.64s/it]
 95%|█████████▍| 18057/19090 [8:48:47<31:44,  1.84s/it]
 95%|█████████▍| 18058/19090 [8:48:49<30:31,  1.77s/it]
 95%|█████████▍| 18059/19090 [8:48:50<28:18,  1.65s/it][gradnorm] step 18059: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 885.3

 95%|█████████▍| 18060/19090 [8:48:52<31:46,  1.85s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1265.16064453125, 'learning_rate': 4.429386609726721e-07, 'recon_loss': 0.0014049300545593724, 'recon_static': 0.0004302462017221842, 'recon_motion': 0.0009746839190484024, 'recon_norm': 1.2230250775814056, 'recon_norm_std': 0.042741897702217105, 'target_norm': 1.6188305556774139, 'recon_cos': 0.6695837214589119, 'gradnorm/compressor.query': 0.4082694296397697, 'gradnorm/compressor.layer0': 8.634980991786584, 'gradnorm/compressor.layer1': 1254.441160437587, 'gradnorm/compressor.layer2': 158.0537821340413, 'gradnorm/compressor.layer3': 41.90229258672415, 'gradnorm/compressor.layer4': 11.073365818123783, 'gradnorm/compressor.layer5': 6.030787316793752, 'gradnorm/compressor.layer6': 2.9167721502212256, 'gradnorm/compressor.layer7': 3.3115962973037516, 'gradnorm/decoder.layer0': 3.5867760702084293, 'gradnorm/decoder.layer1': 2.070808783014044, 'gradnorm/decoder.layer2': 1.283849823217673, 'gradnorm/max_param': 885.2769165039062, 'epoch': 9.46}

 95%|█████████▍| 18060/19090 [8:48:52<31:46,  1.85s/it]
 95%|█████████▍| 18061/19090 [8:48:54<29:59,  1.75s/it]
 95%|█████████▍| 18062/19090 [8:48:55<28:11,  1.65s/it]
 95%|█████████▍| 18063/19090 [8:48:58<32:25,  1.89s/it]
 95%|█████████▍| 18064/19090 [8:49:00<31:29,  1.84s/it]
 95%|█████████▍| 18065/19090 [8:49:01<28:38,  1.68s/it]
 95%|█████████▍| 18066/19090 [8:49:03<32:37,  1.91s/it]
 95%|█████████▍| 18067/19090 [8:49:05<32:19,  1.90s/it]
 95%|█████████▍| 18068/19090 [8:49:06<29:11,  1.71s/it]
 95%|█████████▍| 18069/19090 [8:49:09<32:42,  1.92s/it][gradnorm] step 18069: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 796.6

 95%|█████████▍| 18070/19090 [8:49:11<32:51,  1.93s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1137.80126953125, 'learning_rate': 4.3441277369326294e-07, 'recon_loss': 0.0014022542076418177, 'recon_static': 0.00042595663617248645, 'recon_motion': 0.000976297544548288, 'recon_norm': 1.2197464644908904, 'recon_norm_std': 0.04284841334447265, 'target_norm': 1.6207783430814744, 'recon_cos': 0.6714136347174644, 'gradnorm/compressor.query': 0.3178467396959435, 'gradnorm/compressor.layer0': 10.081548821879961, 'gradnorm/compressor.layer1': 1128.6573328517384, 'gradnorm/compressor.layer2': 137.8865518166112, 'gradnorm/compressor.layer3': 37.802476396126316, 'gradnorm/compressor.layer4': 10.81475947591867, 'gradnorm/compressor.layer5': 5.39055317264194, 'gradnorm/compressor.layer6': 2.6396266697851543, 'gradnorm/compressor.layer7': 3.1850534661564773, 'gradnorm/decoder.layer0': 3.451955605912102, 'gradnorm/decoder.layer1': 2.0191869682060752, 'gradnorm/decoder.layer2': 1.2326186517239213, 'gradnorm/max_param': 796.6126708984375, 'epoch': 9.47}

 95%|█████████▍| 18070/19090 [8:49:11<32:51,  1.93s/it]
 95%|█████████▍| 18071/19090 [8:49:12<29:32,  1.74s/it]
 95%|█████████▍| 18072/19090 [8:49:14<32:17,  1.90s/it]
 95%|█████████▍| 18073/19090 [8:49:16<30:57,  1.83s/it]
 95%|█████████▍| 18074/19090 [8:49:17<28:10,  1.66s/it]
 95%|█████████▍| 18075/19090 [8:49:19<30:34,  1.81s/it]
 95%|█████████▍| 18076/19090 [8:49:21<31:48,  1.88s/it]
 95%|█████████▍| 18077/19090 [8:49:23<28:45,  1.70s/it]
 95%|█████████▍| 18078/19090 [8:49:25<31:48,  1.89s/it]
 95%|█████████▍| 18079/19090 [8:49:27<31:30,  1.87s/it][gradnorm] step 18079: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 857.6

 95%|█████████▍| 18080/19090 [8:49:28<28:31,  1.69s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1224.4306640625, 'learning_rate': 4.259690217235268e-07, 'recon_loss': 0.0013987794023705646, 'recon_static': 0.0004249604891811032, 'recon_motion': 0.0009738189008203335, 'recon_norm': 1.22086401283741, 'recon_norm_std': 0.04417690308764577, 'target_norm': 1.6207615643739701, 'recon_cos': 0.6726066991686821, 'gradnorm/compressor.query': 0.6698282471115904, 'gradnorm/compressor.layer0': 22.897941716741293, 'gradnorm/compressor.layer1': 1215.2104241652967, 'gradnorm/compressor.layer2': 142.6272526662594, 'gradnorm/compressor.layer3': 38.149049471591525, 'gradnorm/compressor.layer4': 10.545696548746422, 'gradnorm/compressor.layer5': 5.256900974745137, 'gradnorm/compressor.layer6': 2.751432999136372, 'gradnorm/compressor.layer7': 2.9618678581439863, 'gradnorm/decoder.layer0': 3.3432692422805848, 'gradnorm/decoder.layer1': 1.9715019852630427, 'gradnorm/decoder.layer2': 1.163468074719596, 'gradnorm/max_param': 857.6309204101562, 'epoch': 9.47}

 95%|█████████▍| 18080/19090 [8:49:28<28:31,  1.69s/it]
 95%|█████████▍| 18081/19090 [8:49:30<30:34,  1.82s/it]
 95%|█████████▍| 18082/19090 [8:49:32<30:41,  1.83s/it]
 95%|█████████▍| 18083/19090 [8:49:33<27:57,  1.67s/it]
 95%|█████████▍| 18084/19090 [8:49:36<30:12,  1.80s/it]
 95%|█████████▍| 18085/19090 [8:49:37<28:49,  1.72s/it]
 95%|█████████▍| 18086/19090 [8:49:38<26:42,  1.60s/it]
 95%|█████████▍| 18087/19090 [8:49:41<30:43,  1.84s/it]
 95%|█████████▍| 18088/19090 [8:49:42<29:14,  1.75s/it]
 95%|█████████▍| 18089/19090 [8:49:44<29:10,  1.75s/it][gradnorm] step 18089: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 897.6

 95%|█████████▍| 18090/19090 [8:49:46<31:33,  1.89s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1284.830322265625, 'learning_rate': 4.1760743329523624e-07, 'recon_loss': 0.0013996650493936613, 'recon_static': 0.00042828874866245313, 'recon_motion': 0.0009713762759929523, 'recon_norm': 1.2161870419979095, 'recon_norm_std': 0.04255697261542082, 'target_norm': 1.6166375249624252, 'recon_cos': 0.6694748967885971, 'gradnorm/compressor.query': 0.44555291161217786, 'gradnorm/compressor.layer0': 18.861980874493685, 'gradnorm/compressor.layer1': 1271.8674754077172, 'gradnorm/compressor.layer2': 175.44176940832818, 'gradnorm/compressor.layer3': 42.927660751748256, 'gradnorm/compressor.layer4': 10.206089027382784, 'gradnorm/compressor.layer5': 5.304133750092204, 'gradnorm/compressor.layer6': 2.5975785093219153, 'gradnorm/compressor.layer7': 2.988157900589422, 'gradnorm/decoder.layer0': 3.2357643571081387, 'gradnorm/decoder.layer1': 1.9239762767153985, 'gradnorm/decoder.layer2': 1.1145329983953027, 'gradnorm/max_param': 897.6259765625, 'epoch': 9.48}

 95%|█████████▍| 18090/19090 [8:49:46<31:33,  1.89s/it]
 95%|█████████▍| 18091/19090 [8:49:48<28:53,  1.74s/it]
 95%|█████████▍| 18092/19090 [8:49:49<28:51,  1.73s/it]
 95%|█████████▍| 18093/19090 [8:49:52<31:04,  1.87s/it]
 95%|█████████▍| 18094/19090 [8:49:53<30:36,  1.84s/it]
 95%|█████████▍| 18095/19090 [8:49:55<27:51,  1.68s/it]
 95%|█████████▍| 18096/19090 [8:49:57<30:40,  1.85s/it]
 95%|█████████▍| 18097/19090 [8:49:59<32:05,  1.94s/it]
 95%|█████████▍| 18098/19090 [8:50:00<28:49,  1.74s/it]
 95%|█████████▍| 18099/19090 [8:50:02<28:59,  1.76s/it][gradnorm] step 18099: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 759.7

 95%|█████████▍| 18100/19090 [8:50:05<32:24,  1.96s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1084.5616455078125, 'learning_rate': 4.093280363654528e-07, 'recon_loss': 0.0014088876981986687, 'recon_static': 0.0004351637093350291, 'recon_motion': 0.0009737239772221074, 'recon_norm': 1.2283444941043853, 'recon_norm_std': 0.04299738602712751, 'target_norm': 1.6179769545793534, 'recon_cos': 0.6687162458896637, 'gradnorm/compressor.query': 0.3339249184316124, 'gradnorm/compressor.layer0': 12.52605195339148, 'gradnorm/compressor.layer1': 1076.4023991983668, 'gradnorm/compressor.layer2': 126.4526645317824, 'gradnorm/compressor.layer3': 36.630492335621156, 'gradnorm/compressor.layer4': 9.263486773068994, 'gradnorm/compressor.layer5': 5.1281412313833235, 'gradnorm/compressor.layer6': 2.5348299871613196, 'gradnorm/compressor.layer7': 2.9223141263749266, 'gradnorm/decoder.layer0': 3.235450455259283, 'gradnorm/decoder.layer1': 1.9434334132213962, 'gradnorm/decoder.layer2': 1.175829661440416, 'gradnorm/max_param': 759.6925659179688, 'epoch': 9.48}

 95%|█████████▍| 18100/19090 [8:50:05<32:24,  1.96s/it]
 95%|█████████▍| 18101/19090 [8:50:06<28:59,  1.76s/it]
 95%|█████████▍| 18102/19090 [8:50:08<28:18,  1.72s/it]
 95%|█████████▍| 18103/19090 [8:50:10<31:53,  1.94s/it]
 95%|█████████▍| 18104/19090 [8:50:11<28:36,  1.74s/it]
 95%|█████████▍| 18105/19090 [8:50:13<28:12,  1.72s/it]
 95%|█████████▍| 18106/19090 [8:50:15<32:12,  1.96s/it]
 95%|█████████▍| 18107/19090 [8:50:17<28:51,  1.76s/it]
 95%|█████████▍| 18108/19090 [8:50:18<27:21,  1.67s/it]
 95%|█████████▍| 18109/19090 [8:50:21<30:43,  1.88s/it][gradnorm] step 18109: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 767.5

 95%|█████████▍| 18110/19090 [8:50:22<27:49,  1.70s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1097.41748046875, 'learning_rate': 4.0113085861643583e-07, 'recon_loss': 0.0014234070549719036, 'recon_static': 0.00044180010008858516, 'recon_motion': 0.0009816069112275727, 'recon_norm': 1.239842489361763, 'recon_norm_std': 0.04183138208463788, 'target_norm': 1.620388498902321, 'recon_cos': 0.6670271813869476, 'gradnorm/compressor.query': 0.3410663317967899, 'gradnorm/compressor.layer0': 8.988217588594631, 'gradnorm/compressor.layer1': 1087.463332715177, 'gradnorm/compressor.layer2': 142.0355438019829, 'gradnorm/compressor.layer3': 36.542252005597916, 'gradnorm/compressor.layer4': 9.665177548470439, 'gradnorm/compressor.layer5': 5.455029174279119, 'gradnorm/compressor.layer6': 2.7542927788636486, 'gradnorm/compressor.layer7': 3.039584941135859, 'gradnorm/decoder.layer0': 3.4783093125141917, 'gradnorm/decoder.layer1': 2.063166077305151, 'gradnorm/decoder.layer2': 1.322557335459833, 'gradnorm/max_param': 767.5006103515625, 'epoch': 9.49}

 95%|█████████▍| 18110/19090 [8:50:22<27:49,  1.70s/it]
 95%|█████████▍| 18111/19090 [8:50:24<28:03,  1.72s/it]
 95%|█████████▍| 18112/19090 [8:50:25<28:38,  1.76s/it]
 95%|█████████▍| 18113/19090 [8:50:27<26:22,  1.62s/it]
 95%|█████████▍| 18114/19090 [8:50:29<28:22,  1.74s/it]
 95%|█████████▍| 18115/19090 [8:50:31<28:59,  1.78s/it]
 95%|█████████▍| 18116/19090 [8:50:32<26:36,  1.64s/it]
 95%|█████████▍| 18117/19090 [8:50:34<30:07,  1.86s/it]
 95%|█████████▍| 18118/19090 [8:50:36<28:22,  1.75s/it]
 95%|█████████▍| 18119/19090 [8:50:37<26:45,  1.65s/it][gradnorm] step 18119: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 836.5

 95%|█████████▍| 18120/19090 [8:50:40<29:58,  1.85s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1194.8602294921875, 'learning_rate': 3.930159274555312e-07, 'recon_loss': 0.0014208670618245378, 'recon_static': 0.000443156070105033, 'recon_motion': 0.000977711024461314, 'recon_norm': 1.2255005955696106, 'recon_norm_std': 0.04241366032510996, 'target_norm': 1.619694349169731, 'recon_cos': 0.665046489238739, 'gradnorm/compressor.query': 0.37898973154887927, 'gradnorm/compressor.layer0': 7.720386902991635, 'gradnorm/compressor.layer1': 1185.2053092186181, 'gradnorm/compressor.layer2': 144.98576169317454, 'gradnorm/compressor.layer3': 41.58543921714441, 'gradnorm/compressor.layer4': 10.248486290786625, 'gradnorm/compressor.layer5': 5.307579184977796, 'gradnorm/compressor.layer6': 2.9285583761635445, 'gradnorm/compressor.layer7': 3.187819072236832, 'gradnorm/decoder.layer0': 3.4078670478103685, 'gradnorm/decoder.layer1': 1.9928741468513758, 'gradnorm/decoder.layer2': 1.3306040347316939, 'gradnorm/max_param': 836.45654296875, 'epoch': 9.49}

 95%|█████████▍| 18120/19090 [8:50:40<29:58,  1.85s/it]
 95%|█████████▍| 18121/19090 [8:50:41<28:14,  1.75s/it]
 95%|█████████▍| 18122/19090 [8:50:42<26:02,  1.61s/it]
 95%|█████████▍| 18123/19090 [8:50:45<29:58,  1.86s/it]
 95%|█████████▍| 18124/19090 [8:50:46<28:05,  1.74s/it]
 95%|█████████▍| 18125/19090 [8:50:48<25:51,  1.61s/it]
 95%|█████████▍| 18126/19090 [8:50:50<31:28,  1.96s/it]
 95%|█████████▍| 18127/19090 [8:50:52<28:46,  1.79s/it]
 95%|█████████▍| 18128/19090 [8:50:53<26:38,  1.66s/it]
 95%|█████████▍| 18129/19090 [8:50:56<31:39,  1.98s/it][gradnorm] step 18129: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 855.0

 95%|█████████▍| 18130/19090 [8:50:57<28:46,  1.80s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1221.292724609375, 'learning_rate': 3.8498327001510736e-07, 'recon_loss': 0.0014169814065098762, 'recon_static': 0.00043116346569149755, 'recon_motion': 0.0009858179022558034, 'recon_norm': 1.226080420613289, 'recon_norm_std': 0.04200153388082981, 'target_norm': 1.6211475491523744, 'recon_cos': 0.6677280783653259, 'gradnorm/compressor.query': 0.43353460097463187, 'gradnorm/compressor.layer0': 13.73729271216022, 'gradnorm/compressor.layer1': 1211.706802407249, 'gradnorm/compressor.layer2': 146.81764172562848, 'gradnorm/compressor.layer3': 37.70605821578112, 'gradnorm/compressor.layer4': 9.688115451090184, 'gradnorm/compressor.layer5': 5.3276060864086485, 'gradnorm/compressor.layer6': 2.891522737231924, 'gradnorm/compressor.layer7': 3.0234116348619087, 'gradnorm/decoder.layer0': 3.3361024480773045, 'gradnorm/decoder.layer1': 2.0035796317565704, 'gradnorm/decoder.layer2': 1.2275703808560792, 'gradnorm/max_param': 855.0416870117188, 'epoch': 9.5}

 95%|█████████▍| 18130/19090 [8:50:57<28:46,  1.80s/it]
 95%|█████████▍| 18131/19090 [8:50:59<27:02,  1.69s/it]
 95%|█████████▍| 18132/19090 [8:51:01<30:55,  1.94s/it]
 95%|█████████▍| 18133/19090 [8:51:03<29:27,  1.85s/it]
 95%|█████████▍| 18134/19090 [8:51:04<26:47,  1.68s/it]
 95%|█████████▍| 18135/19090 [8:51:07<30:20,  1.91s/it]
 95%|█████████▌| 18136/19090 [8:51:08<28:27,  1.79s/it]
 95%|█████████▌| 18137/19090 [8:51:09<26:02,  1.64s/it]
 95%|█████████▌| 18138/19090 [8:51:12<29:51,  1.88s/it]
 95%|█████████▌| 18139/19090 [8:51:13<28:38,  1.81s/it][gradnorm] step 18139: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 796.9

 95%|█████████▌| 18140/19090 [8:51:15<26:08,  1.65s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1140.997802734375, 'learning_rate': 3.7703291315243916e-07, 'recon_loss': 0.0014076744293561206, 'recon_static': 0.0004261383175617084, 'recon_motion': 0.0009815361467190088, 'recon_norm': 1.225132092833519, 'recon_norm_std': 0.042842441145330666, 'target_norm': 1.6211072951555252, 'recon_cos': 0.6704877078533172, 'gradnorm/compressor.query': 0.28131228657942997, 'gradnorm/compressor.layer0': 6.921254976286984, 'gradnorm/compressor.layer1': 1129.0411307831084, 'gradnorm/compressor.layer2': 158.74164594160538, 'gradnorm/compressor.layer3': 41.5401387802546, 'gradnorm/compressor.layer4': 10.286128315842674, 'gradnorm/compressor.layer5': 5.444911597228714, 'gradnorm/compressor.layer6': 2.6482195469118945, 'gradnorm/compressor.layer7': 3.126301913861841, 'gradnorm/decoder.layer0': 3.453717487305049, 'gradnorm/decoder.layer1': 2.014773403328586, 'gradnorm/decoder.layer2': 1.2349619375378977, 'gradnorm/max_param': 796.8733520507812, 'epoch': 9.5}

 95%|█████████▌| 18140/19090 [8:51:15<26:08,  1.65s/it]
 95%|█████████▌| 18141/19090 [8:51:17<29:32,  1.87s/it]
 95%|█████████▌| 18142/19090 [8:51:18<27:05,  1.72s/it]
 95%|█████████▌| 18143/19090 [8:51:20<25:02,  1.59s/it]
 95%|█████████▌| 18144/19090 [8:51:22<29:31,  1.87s/it]
 95%|█████████▌| 18145/19090 [8:51:24<30:09,  1.91s/it]
 95%|█████████▌| 18146/19090 [8:51:26<27:09,  1.73s/it]
 95%|█████████▌| 18147/19090 [8:51:28<29:41,  1.89s/it]
 95%|█████████▌| 18148/19090 [8:51:29<28:32,  1.82s/it]
 95%|█████████▌| 18149/19090 [8:51:31<26:04,  1.66s/it][gradnorm] step 18149: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 943.3

 95%|█████████▌| 18150/19090 [8:51:33<28:50,  1.84s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1346.3055419921875, 'learning_rate': 3.6916488344962974e-07, 'recon_loss': 0.0013897241879021749, 'recon_static': 0.00042022784691653214, 'recon_motion': 0.0009694964275695384, 'recon_norm': 1.2190510421991347, 'recon_norm_std': 0.04253688994795084, 'target_norm': 1.616743692755699, 'recon_cos': 0.6724103882908821, 'gradnorm/compressor.query': 0.4171799676908095, 'gradnorm/compressor.layer0': 11.995903587511004, 'gradnorm/compressor.layer1': 1336.5687412176, 'gradnorm/compressor.layer2': 152.62121847280935, 'gradnorm/compressor.layer3': 50.129086296966854, 'gradnorm/compressor.layer4': 10.419302970431776, 'gradnorm/compressor.layer5': 5.513679312506054, 'gradnorm/compressor.layer6': 2.680229021079772, 'gradnorm/compressor.layer7': 3.11490449982496, 'gradnorm/decoder.layer0': 3.3454661020919914, 'gradnorm/decoder.layer1': 1.9907093623084449, 'gradnorm/decoder.layer2': 1.274116990115043, 'gradnorm/max_param': 943.3087768554688, 'epoch': 9.51}

 95%|█████████▌| 18150/19090 [8:51:33<28:50,  1.84s/it]
 95%|█████████▌| 18151/19090 [8:51:35<28:31,  1.82s/it]
 95%|█████████▌| 18152/19090 [8:51:36<26:00,  1.66s/it]
 95%|█████████▌| 18153/19090 [8:51:38<28:22,  1.82s/it]
 95%|█████████▌| 18154/19090 [8:51:40<27:51,  1.79s/it]
 95%|█████████▌| 18155/19090 [8:51:41<25:30,  1.64s/it]
 95%|█████████▌| 18156/19090 [8:51:44<29:00,  1.86s/it]
 95%|█████████▌| 18157/19090 [8:51:45<28:13,  1.81s/it]
 95%|█████████▌| 18158/19090 [8:51:47<25:44,  1.66s/it]
 95%|█████████▌| 18159/19090 [8:51:49<28:45,  1.85s/it][gradnorm] step 18159: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 731.5

 95%|█████████▌| 18160/19090 [8:51:51<27:56,  1.80s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1050.75439453125, 'learning_rate': 3.613792072135275e-07, 'recon_loss': 0.0013997493486385792, 'recon_static': 0.00042208413069602104, 'recon_motion': 0.000977665175742004, 'recon_norm': 1.217289000749588, 'recon_norm_std': 0.04184828121215105, 'target_norm': 1.6184653520584107, 'recon_cos': 0.670628798007965, 'gradnorm/compressor.query': 0.376949250987015, 'gradnorm/compressor.layer0': 7.98900492405116, 'gradnorm/compressor.layer1': 1036.4766639920072, 'gradnorm/compressor.layer2': 167.73022890194838, 'gradnorm/compressor.layer3': 38.087920621446706, 'gradnorm/compressor.layer4': 9.712519276721338, 'gradnorm/compressor.layer5': 5.3036520485303225, 'gradnorm/compressor.layer6': 2.470140379495725, 'gradnorm/compressor.layer7': 2.9785193071186256, 'gradnorm/decoder.layer0': 3.2138695613976345, 'gradnorm/decoder.layer1': 1.9908864878087544, 'gradnorm/decoder.layer2': 1.1799730372177115, 'gradnorm/max_param': 731.4824829101562, 'epoch': 9.51}

 95%|█████████▌| 18160/19090 [8:51:51<27:56,  1.80s/it]
 95%|█████████▌| 18161/19090 [8:51:52<25:32,  1.65s/it]
 95%|█████████▌| 18162/19090 [8:51:54<27:32,  1.78s/it]
 95%|█████████▌| 18163/19090 [8:51:56<29:39,  1.92s/it]
 95%|█████████▌| 18164/19090 [8:51:58<26:41,  1.73s/it]
 95%|█████████▌| 18165/19090 [8:52:00<27:48,  1.80s/it]
 95%|█████████▌| 18166/19090 [8:52:02<30:02,  1.95s/it]
 95%|█████████▌| 18167/19090 [8:52:03<26:57,  1.75s/it]
 95%|█████████▌| 18168/19090 [8:52:05<25:09,  1.64s/it]
 95%|█████████▌| 18169/19090 [8:52:07<29:31,  1.92s/it][gradnorm] step 18169: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 804.1

 95%|█████████▌| 18170/19090 [8:52:08<26:35,  1.73s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1150.8558349609375, 'learning_rate': 3.536759104756232e-07, 'recon_loss': 0.0014101163833402097, 'recon_static': 0.0004214088534354232, 'recon_motion': 0.0009887075080769136, 'recon_norm': 1.2214605301618575, 'recon_norm_std': 0.04273149315267801, 'target_norm': 1.622442689538002, 'recon_cos': 0.6703029200434685, 'gradnorm/compressor.query': 0.39888510152278517, 'gradnorm/compressor.layer0': 9.657828883990511, 'gradnorm/compressor.layer1': 1139.3720309890005, 'gradnorm/compressor.layer2': 156.76189897022172, 'gradnorm/compressor.layer3': 38.64988800466875, 'gradnorm/compressor.layer4': 8.962881474377003, 'gradnorm/compressor.layer5': 5.3450360451634635, 'gradnorm/compressor.layer6': 2.338453793874843, 'gradnorm/compressor.layer7': 2.9937794088619456, 'gradnorm/decoder.layer0': 3.240743807432954, 'gradnorm/decoder.layer1': 1.9937918990465107, 'gradnorm/decoder.layer2': 1.1735530791271473, 'gradnorm/max_param': 804.0942993164062, 'epoch': 9.52}

 95%|█████████▌| 18170/19090 [8:52:08<26:35,  1.73s/it]
 95%|█████████▌| 18171/19090 [8:52:10<25:52,  1.69s/it]
 95%|█████████▌| 18172/19090 [8:52:12<28:45,  1.88s/it]
 95%|█████████▌| 18173/19090 [8:52:14<26:01,  1.70s/it]
 95%|█████████▌| 18174/19090 [8:52:15<26:23,  1.73s/it]
 95%|█████████▌| 18175/19090 [8:52:18<29:53,  1.96s/it]
 95%|█████████▌| 18176/19090 [8:52:19<26:46,  1.76s/it]
 95%|█████████▌| 18177/19090 [8:52:20<24:45,  1.63s/it]
 95%|█████████▌| 18178/19090 [8:52:23<28:31,  1.88s/it]
 95%|█████████▌| 18179/19090 [8:52:24<25:47,  1.70s/it][gradnorm] step 18179: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 771.1

 95%|█████████▌| 18180/19090 [8:52:26<25:30,  1.68s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1102.31982421875, 'learning_rate': 3.4605501899197256e-07, 'recon_loss': 0.0014193870214512573, 'recon_static': 0.0004198918933980167, 'recon_motion': 0.0009994951527914965, 'recon_norm': 1.2184629619121552, 'recon_norm_std': 0.0425909467972815, 'target_norm': 1.6243101745843886, 'recon_cos': 0.6686765149235725, 'gradnorm/compressor.query': 0.43876393847591205, 'gradnorm/compressor.layer0': 11.286103026332698, 'gradnorm/compressor.layer1': 1092.7767384054257, 'gradnorm/compressor.layer2': 138.4120814700617, 'gradnorm/compressor.layer3': 39.12477223875399, 'gradnorm/compressor.layer4': 8.629142319371535, 'gradnorm/compressor.layer5': 5.221729273335023, 'gradnorm/compressor.layer6': 2.272998570808871, 'gradnorm/compressor.layer7': 3.05478865121849, 'gradnorm/decoder.layer0': 3.2298019440577708, 'gradnorm/decoder.layer1': 2.0141915835616353, 'gradnorm/decoder.layer2': 1.1779039555014346, 'gradnorm/max_param': 771.1156005859375, 'epoch': 9.52}

 95%|█████████▌| 18180/19090 [8:52:26<25:30,  1.68s/it]
 95%|█████████▌| 18181/19090 [8:52:28<28:37,  1.89s/it]
 95%|█████████▌| 18182/19090 [8:52:30<25:51,  1.71s/it]
 95%|█████████▌| 18183/19090 [8:52:31<25:08,  1.66s/it]
 95%|█████████▌| 18184/19090 [8:52:34<28:30,  1.89s/it]
 95%|█████████▌| 18185/19090 [8:52:35<25:45,  1.71s/it]
 95%|█████████▌| 18186/19090 [8:52:36<25:42,  1.71s/it]
 95%|█████████▌| 18187/19090 [8:52:39<27:53,  1.85s/it]
 95%|█████████▌| 18188/19090 [8:52:40<25:17,  1.68s/it]
 95%|█████████▌| 18189/19090 [8:52:42<26:26,  1.76s/it][gradnorm] step 18189: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1105.1

 95%|█████████▌| 18190/19090 [8:52:44<27:44,  1.85s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1575.7332763671875, 'learning_rate': 3.385165582431099e-07, 'recon_loss': 0.001413110105204396, 'recon_static': 0.00042750076681841165, 'recon_motion': 0.000985609357303474, 'recon_norm': 1.2139410585165025, 'recon_norm_std': 0.0436006280593574, 'target_norm': 1.6244821786880492, 'recon_cos': 0.6697673633694649, 'gradnorm/compressor.query': 0.8118833255708019, 'gradnorm/compressor.layer0': 27.290979212676096, 'gradnorm/compressor.layer1': 1566.251895449771, 'gradnorm/compressor.layer2': 165.27154502136779, 'gradnorm/compressor.layer3': 39.703112701690344, 'gradnorm/compressor.layer4': 9.782135652171283, 'gradnorm/compressor.layer5': 5.233505677870031, 'gradnorm/compressor.layer6': 2.4150981965250695, 'gradnorm/compressor.layer7': 2.9196347711067, 'gradnorm/decoder.layer0': 3.353921695711872, 'gradnorm/decoder.layer1': 2.0242428152817356, 'gradnorm/decoder.layer2': 1.114688933649363, 'gradnorm/max_param': 1105.078369140625, 'epoch': 9.53}

 95%|█████████▌| 18190/19090 [8:52:44<27:44,  1.85s/it]
 95%|█████████▌| 18191/19090 [8:52:45<25:14,  1.68s/it]
 95%|█████████▌| 18192/19090 [8:52:47<25:41,  1.72s/it]
 95%|█████████▌| 18193/19090 [8:52:49<28:38,  1.92s/it]
 95%|█████████▌| 18194/19090 [8:52:51<25:48,  1.73s/it]
 95%|█████████▌| 18195/19090 [8:52:52<25:34,  1.71s/it]
 95%|█████████▌| 18196/19090 [8:52:54<27:03,  1.82s/it]
 95%|█████████▌| 18197/19090 [8:52:56<24:39,  1.66s/it]
 95%|█████████▌| 18198/19090 [8:52:58<25:02,  1.68s/it]
 95%|█████████▌| 18199/19090 [8:53:00<27:07,  1.83s/it][gradnorm] step 18199: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 792.1

 95%|█████████▌| 18200/19090 [8:53:01<24:43,  1.67s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1134.2054443359375, 'learning_rate': 3.3106055343396504e-07, 'recon_loss': 0.001407635904615745, 'recon_static': 0.0004201191186439246, 'recon_motion': 0.0009875168150756508, 'recon_norm': 1.2200067728757857, 'recon_norm_std': 0.04251375449821353, 'target_norm': 1.6216535419225693, 'recon_cos': 0.6702368289232254, 'gradnorm/compressor.query': 0.43187201671425635, 'gradnorm/compressor.layer0': 13.251379553034988, 'gradnorm/compressor.layer1': 1122.2941347971127, 'gradnorm/compressor.layer2': 158.16024187315534, 'gradnorm/compressor.layer3': 39.37956167772424, 'gradnorm/compressor.layer4': 8.928906130375578, 'gradnorm/compressor.layer5': 5.220617816654772, 'gradnorm/compressor.layer6': 2.432238843018865, 'gradnorm/compressor.layer7': 2.9212182270716753, 'gradnorm/decoder.layer0': 3.1982657441880558, 'gradnorm/decoder.layer1': 1.9465657682049968, 'gradnorm/decoder.layer2': 1.0633609873564869, 'gradnorm/max_param': 792.0961303710938, 'epoch': 9.53}

 95%|█████████▌| 18200/19090 [8:53:01<24:43,  1.67s/it]
 95%|█████████▌| 18201/19090 [8:53:03<25:42,  1.73s/it]
 95%|█████████▌| 18202/19090 [8:53:05<27:38,  1.87s/it]
 95%|█████████▌| 18203/19090 [8:53:06<25:21,  1.72s/it]
 95%|█████████▌| 18204/19090 [8:53:08<26:13,  1.78s/it]
 95%|█████████▌| 18205/19090 [8:53:10<26:22,  1.79s/it]
 95%|█████████▌| 18206/19090 [8:53:12<25:28,  1.73s/it]
 95%|█████████▌| 18207/19090 [8:53:14<26:13,  1.78s/it]
 95%|█████████▌| 18208/19090 [8:53:15<25:53,  1.76s/it]
 95%|█████████▌| 18209/19090 [8:53:17<24:27,  1.67s/it][gradnorm] step 18209: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 820.4

 95%|█████████▌| 18210/19090 [8:53:19<27:10,  1.85s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1171.4813232421875, 'learning_rate': 3.236870294937661e-07, 'recon_loss': 0.0014055491628823802, 'recon_static': 0.0004214590408082586, 'recon_motion': 0.0009840901140705683, 'recon_norm': 1.2263902127742767, 'recon_norm_std': 0.04241068921983242, 'target_norm': 1.619917231798172, 'recon_cos': 0.6706340491771698, 'gradnorm/compressor.query': 0.38336900237584515, 'gradnorm/compressor.layer0': 13.083163185390102, 'gradnorm/compressor.layer1': 1162.5863408796786, 'gradnorm/compressor.layer2': 137.93840620454478, 'gradnorm/compressor.layer3': 38.00231123470897, 'gradnorm/compressor.layer4': 7.8954157119125785, 'gradnorm/compressor.layer5': 5.133681596774438, 'gradnorm/compressor.layer6': 2.4624541922244383, 'gradnorm/compressor.layer7': 2.9876243916555567, 'gradnorm/decoder.layer0': 3.200040346129703, 'gradnorm/decoder.layer1': 1.9340590408449585, 'gradnorm/decoder.layer2': 1.0787808178934262, 'gradnorm/max_param': 820.38037109375, 'epoch': 9.54}

 95%|█████████▌| 18210/19090 [8:53:19<27:10,  1.85s/it]
 95%|█████████▌| 18211/19090 [8:53:21<26:17,  1.79s/it]
 95%|█████████▌| 18212/19090 [8:53:22<24:48,  1.70s/it]
 95%|█████████▌| 18213/19090 [8:53:24<26:15,  1.80s/it]
 95%|█████████▌| 18214/19090 [8:53:26<26:59,  1.85s/it]
 95%|█████████▌| 18215/19090 [8:53:27<24:30,  1.68s/it]
 95%|█████████▌| 18216/19090 [8:53:30<28:17,  1.94s/it]
 95%|█████████▌| 18217/19090 [8:53:31<25:49,  1.77s/it]
 95%|█████████▌| 18218/19090 [8:53:33<24:06,  1.66s/it]
 95%|█████████▌| 18219/19090 [8:53:35<27:09,  1.87s/it][gradnorm] step 18219: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 959.2

 95%|█████████▌| 18220/19090 [8:53:37<26:21,  1.82s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1371.412841796875, 'learning_rate': 3.163960110759756e-07, 'recon_loss': 0.0014067464944673702, 'recon_static': 0.00042302381407353094, 'recon_motion': 0.0009837226898525842, 'recon_norm': 1.217179012298584, 'recon_norm_std': 0.04294965360313654, 'target_norm': 1.6204458057880402, 'recon_cos': 0.6696277976036071, 'gradnorm/compressor.query': 0.4208456656880555, 'gradnorm/compressor.layer0': 9.73938349282502, 'gradnorm/compressor.layer1': 1359.1440321025582, 'gradnorm/compressor.layer2': 176.49350975142116, 'gradnorm/compressor.layer3': 45.98535188394914, 'gradnorm/compressor.layer4': 8.95798646713703, 'gradnorm/compressor.layer5': 5.407745330931273, 'gradnorm/compressor.layer6': 2.6745531368457653, 'gradnorm/compressor.layer7': 3.075244931026344, 'gradnorm/decoder.layer0': 3.2085886436702276, 'gradnorm/decoder.layer1': 1.8459790220380858, 'gradnorm/decoder.layer2': 1.0949565227360074, 'gradnorm/max_param': 959.1785278320312, 'epoch': 9.54}

 95%|█████████▌| 18220/19090 [8:53:37<26:21,  1.82s/it]
 95%|█████████▌| 18221/19090 [8:53:38<24:03,  1.66s/it]
 95%|█████████▌| 18222/19090 [8:53:41<27:16,  1.89s/it]
 95%|█████████▌| 18223/19090 [8:53:43<28:02,  1.94s/it]
 95%|█████████▌| 18224/19090 [8:53:44<25:11,  1.75s/it]
 95%|█████████▌| 18225/19090 [8:53:46<24:30,  1.70s/it]
 95%|█████████▌| 18226/19090 [8:53:48<28:48,  2.00s/it]
 95%|█████████▌| 18227/19090 [8:53:50<25:41,  1.79s/it]
 95%|█████████▌| 18228/19090 [8:53:51<23:30,  1.64s/it]
 95%|█████████▌| 18229/19090 [8:53:54<28:57,  2.02s/it][gradnorm] step 18229: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 700.4

 95%|█████████▌| 18230/19090 [8:53:55<25:47,  1.80s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1005.8809814453125, 'learning_rate': 3.0918752255819084e-07, 'recon_loss': 0.001406835616216995, 'recon_static': 0.00042363134925835764, 'recon_motion': 0.0009832042851485312, 'recon_norm': 1.2253321081399917, 'recon_norm_std': 0.042433468624949455, 'target_norm': 1.6207210302352906, 'recon_cos': 0.67048449665308, 'gradnorm/compressor.query': 0.4283308981975895, 'gradnorm/compressor.layer0': 10.987298741846338, 'gradnorm/compressor.layer1': 992.4775186370722, 'gradnorm/compressor.layer2': 159.40895471632075, 'gradnorm/compressor.layer3': 33.25193104120045, 'gradnorm/compressor.layer4': 9.348458523083497, 'gradnorm/compressor.layer5': 5.429776373485786, 'gradnorm/compressor.layer6': 2.473747988676436, 'gradnorm/compressor.layer7': 3.0106096216861884, 'gradnorm/decoder.layer0': 3.220967242050939, 'gradnorm/decoder.layer1': 1.855949901636365, 'gradnorm/decoder.layer2': 1.0869968105188654, 'gradnorm/max_param': 700.3572998046875, 'epoch': 9.55}

 95%|█████████▌| 18230/19090 [8:53:55<25:47,  1.80s/it]
 96%|█████████▌| 18231/19090 [8:53:56<23:33,  1.65s/it]
 96%|█████████▌| 18232/19090 [8:53:59<27:39,  1.93s/it]
 96%|█████████▌| 18233/19090 [8:54:00<24:52,  1.74s/it]
 96%|█████████▌| 18234/19090 [8:54:02<23:29,  1.65s/it]
 96%|█████████▌| 18235/19090 [8:54:04<26:36,  1.87s/it]
 96%|█████████▌| 18236/19090 [8:54:05<24:12,  1.70s/it]
 96%|█████████▌| 18237/19090 [8:54:07<25:53,  1.82s/it]
 96%|█████████▌| 18238/19090 [8:54:09<25:42,  1.81s/it]
 96%|█████████▌| 18239/19090 [8:54:11<24:13,  1.71s/it][gradnorm] step 18239: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 938.2

 96%|█████████▌| 18240/19090 [8:54:13<25:02,  1.77s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1338.3656005859375, 'learning_rate': 3.020615880420713e-07, 'recon_loss': 0.0014087976684095337, 'recon_static': 0.00042668152673286386, 'recon_motion': 0.0009821161496802234, 'recon_norm': 1.2168329536914826, 'recon_norm_std': 0.04296825006604195, 'target_norm': 1.6208135068416596, 'recon_cos': 0.6693944618105888, 'gradnorm/compressor.query': 0.5354220155035632, 'gradnorm/compressor.layer0': 16.74739714938703, 'gradnorm/compressor.layer1': 1329.521671504455, 'gradnorm/compressor.layer2': 149.14607196051125, 'gradnorm/compressor.layer3': 30.689206450772314, 'gradnorm/compressor.layer4': 8.413014239483374, 'gradnorm/compressor.layer5': 5.212030710734413, 'gradnorm/compressor.layer6': 2.5612856150618204, 'gradnorm/compressor.layer7': 2.9560921108673406, 'gradnorm/decoder.layer0': 3.148976017259482, 'gradnorm/decoder.layer1': 1.84050555788842, 'gradnorm/decoder.layer2': 1.056486897735303, 'gradnorm/max_param': 938.1792602539062, 'epoch': 9.55}

 96%|█████████▌| 18240/19090 [8:54:13<25:02,  1.77s/it]
 96%|█████████▌| 18241/19090 [8:54:15<26:21,  1.86s/it]
 96%|█████████▌| 18242/19090 [8:54:16<24:23,  1.73s/it]
 96%|█████████▌| 18243/19090 [8:54:18<24:53,  1.76s/it]
 96%|█████████▌| 18244/19090 [8:54:20<25:59,  1.84s/it]
 96%|█████████▌| 18245/19090 [8:54:21<24:18,  1.73s/it]
 96%|█████████▌| 18246/19090 [8:54:24<26:12,  1.86s/it]
 96%|█████████▌| 18247/19090 [8:54:25<23:45,  1.69s/it]
 96%|█████████▌| 18248/19090 [8:54:27<24:03,  1.71s/it]
 96%|█████████▌| 18249/19090 [8:54:29<26:06,  1.86s/it][gradnorm] step 18249: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 773.0

 96%|█████████▌| 18250/19090 [8:54:30<23:39,  1.69s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1106.1856689453125, 'learning_rate': 2.950182313532612e-07, 'recon_loss': 0.0014117234968580306, 'recon_static': 0.00042227118683513254, 'recon_motion': 0.0009894522998365573, 'recon_norm': 1.2218469649553299, 'recon_norm_std': 0.04297140166163445, 'target_norm': 1.621207395195961, 'recon_cos': 0.6690942943096161, 'gradnorm/compressor.query': 0.4135160979558855, 'gradnorm/compressor.layer0': 11.140304839040011, 'gradnorm/compressor.layer1': 1095.4169297577978, 'gradnorm/compressor.layer2': 150.3109797220416, 'gradnorm/compressor.layer3': 29.207120847545465, 'gradnorm/compressor.layer4': 8.905805395974951, 'gradnorm/compressor.layer5': 5.153721414801805, 'gradnorm/compressor.layer6': 2.6538463712668783, 'gradnorm/compressor.layer7': 3.068592473044441, 'gradnorm/decoder.layer0': 3.2603679486779713, 'gradnorm/decoder.layer1': 1.8723753679100574, 'gradnorm/decoder.layer2': 1.0814012146099332, 'gradnorm/max_param': 773.0213623046875, 'epoch': 9.56}

 96%|█████████▌| 18250/19090 [8:54:30<23:39,  1.69s/it]
 96%|█████████▌| 18251/19090 [8:54:32<23:17,  1.67s/it]
 96%|█████████▌| 18252/19090 [8:54:34<25:58,  1.86s/it]
 96%|█████████▌| 18253/19090 [8:54:35<23:32,  1.69s/it]
 96%|█████████▌| 18254/19090 [8:54:37<23:31,  1.69s/it]
 96%|█████████▌| 18255/19090 [8:54:39<26:27,  1.90s/it]
 96%|█████████▌| 18256/19090 [8:54:41<23:52,  1.72s/it]
 96%|█████████▌| 18257/19090 [8:54:42<23:52,  1.72s/it]
 96%|█████████▌| 18258/19090 [8:54:45<26:14,  1.89s/it]
 96%|█████████▌| 18259/19090 [8:54:46<23:43,  1.71s/it][gradnorm] step 18259: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 898.2

 96%|█████████▌| 18260/19090 [8:54:48<23:46,  1.72s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1283.1539306640625, 'learning_rate': 2.8805747604129794e-07, 'recon_loss': 0.0014070159086259082, 'recon_static': 0.00042302878937334755, 'recon_motion': 0.0009839871170697733, 'recon_norm': 1.2239757537841798, 'recon_norm_std': 0.04248037114739418, 'target_norm': 1.6213707149028778, 'recon_cos': 0.6705038368701934, 'gradnorm/compressor.query': 0.3801987812288085, 'gradnorm/compressor.layer0': 10.29312922372534, 'gradnorm/compressor.layer1': 1272.6831891716022, 'gradnorm/compressor.layer2': 157.8517711038587, 'gradnorm/compressor.layer3': 39.68389039933664, 'gradnorm/compressor.layer4': 10.046758054820938, 'gradnorm/compressor.layer5': 5.405085090863745, 'gradnorm/compressor.layer6': 2.742245317594043, 'gradnorm/compressor.layer7': 3.1707319633598297, 'gradnorm/decoder.layer0': 3.323057373001594, 'gradnorm/decoder.layer1': 1.8817955848156347, 'gradnorm/decoder.layer2': 1.0884221694309113, 'gradnorm/max_param': 898.2168579101562, 'epoch': 9.57}

 96%|█████████▌| 18260/19090 [8:54:48<23:46,  1.72s/it]
 96%|█████████▌| 18261/19090 [8:54:50<25:41,  1.86s/it]
 96%|█████████▌| 18262/19090 [8:54:51<23:18,  1.69s/it]
 96%|█████████▌| 18263/19090 [8:54:53<23:45,  1.72s/it]
 96%|█████████▌| 18264/19090 [8:54:55<25:58,  1.89s/it]
 96%|█████████▌| 18265/19090 [8:54:57<23:29,  1.71s/it]
 96%|█████████▌| 18266/19090 [8:54:58<23:11,  1.69s/it]
 96%|█████████▌| 18267/19090 [8:55:01<26:38,  1.94s/it]
 96%|█████████▌| 18268/19090 [8:55:02<23:57,  1.75s/it]
 96%|█████████▌| 18269/19090 [8:55:04<23:19,  1.70s/it][gradnorm] step 18269: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 901.1

 96%|█████████▌| 18270/19090 [8:55:06<26:42,  1.95s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1284.9659423828125, 'learning_rate': 2.8117934537953973e-07, 'recon_loss': 0.0014120466250460595, 'recon_static': 0.0004211517829389777, 'recon_motion': 0.0009908948602969758, 'recon_norm': 1.2205082476139069, 'recon_norm_std': 0.04280576379969716, 'target_norm': 1.622682160139084, 'recon_cos': 0.6695769637823105, 'gradnorm/compressor.query': 0.42588650863499383, 'gradnorm/compressor.layer0': 8.251736544864928, 'gradnorm/compressor.layer1': 1276.7258417530365, 'gradnorm/compressor.layer2': 140.46113369229582, 'gradnorm/compressor.layer3': 33.97279256648804, 'gradnorm/compressor.layer4': 9.869424525589555, 'gradnorm/compressor.layer5': 5.22575593435526, 'gradnorm/compressor.layer6': 2.551150612135783, 'gradnorm/compressor.layer7': 3.1121333936624134, 'gradnorm/decoder.layer0': 3.3536484280452927, 'gradnorm/decoder.layer1': 1.8572080791988625, 'gradnorm/decoder.layer2': 1.1029300722797386, 'gradnorm/max_param': 901.0709838867188, 'epoch': 9.57}

 96%|█████████▌| 18270/19090 [8:55:06<26:42,  1.95s/it]
 96%|█████████▌| 18271/19090 [8:55:08<23:58,  1.76s/it]
 96%|█████████▌| 18272/19090 [8:55:09<22:36,  1.66s/it]
 96%|█████████▌| 18273/19090 [8:55:11<25:39,  1.88s/it]
 96%|█████████▌| 18274/19090 [8:55:13<23:11,  1.71s/it]
 96%|█████████▌| 18275/19090 [8:55:14<22:31,  1.66s/it]
 96%|█████████▌| 18276/19090 [8:55:17<25:19,  1.87s/it]
 96%|█████████▌| 18277/19090 [8:55:18<22:59,  1.70s/it]
 96%|█████████▌| 18278/19090 [8:55:20<23:48,  1.76s/it]
 96%|█████████▌| 18279/19090 [8:55:21<23:01,  1.70s/it][gradnorm] step 18279: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 713.1

 96%|█████████▌| 18280/19090 [8:55:23<23:41,  1.76s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1020.5359497070312, 'learning_rate': 2.743838623650991e-07, 'recon_loss': 0.0014100017229793592, 'recon_static': 0.0004217642344883643, 'recon_motion': 0.0009882375336019323, 'recon_norm': 1.2187907338142394, 'recon_norm_std': 0.04326957259327173, 'target_norm': 1.6210919827222825, 'recon_cos': 0.6693680509924889, 'gradnorm/compressor.query': 0.5300799556017465, 'gradnorm/compressor.layer0': 14.675775011643603, 'gradnorm/compressor.layer1': 1010.792078025941, 'gradnorm/compressor.layer2': 134.42108221462138, 'gradnorm/compressor.layer3': 36.97493575414888, 'gradnorm/compressor.layer4': 9.068434041866308, 'gradnorm/compressor.layer5': 5.21658499575578, 'gradnorm/compressor.layer6': 2.487959668804668, 'gradnorm/compressor.layer7': 3.2008407263226397, 'gradnorm/decoder.layer0': 3.2936708328606272, 'gradnorm/decoder.layer1': 1.8372197644637946, 'gradnorm/decoder.layer2': 1.0834962037728086, 'gradnorm/max_param': 713.0673828125, 'epoch': 9.58}

 96%|█████████▌| 18280/19090 [8:55:23<23:41,  1.76s/it]
 96%|█████████▌| 18281/19090 [8:55:25<23:58,  1.78s/it]
 96%|█████████▌| 18282/19090 [8:55:26<22:22,  1.66s/it]
 96%|█████████▌| 18283/19090 [8:55:28<22:46,  1.69s/it]
 96%|█████████▌| 18284/19090 [8:55:30<24:47,  1.85s/it]
 96%|█████████▌| 18285/19090 [8:55:32<22:31,  1.68s/it]
 96%|█████████▌| 18286/19090 [8:55:34<23:36,  1.76s/it]
 96%|█████████▌| 18287/19090 [8:55:36<24:51,  1.86s/it]
 96%|█████████▌| 18288/19090 [8:55:37<23:03,  1.72s/it]
 96%|█████████▌| 18289/19090 [8:55:39<23:46,  1.78s/it][gradnorm] step 18289: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 770.1

 96%|█████████▌| 18290/19090 [8:55:41<24:02,  1.80s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1102.78564453125, 'learning_rate': 2.6767104971874013e-07, 'recon_loss': 0.0014097787672653795, 'recon_static': 0.00043009288710891267, 'recon_motion': 0.0009796858983463607, 'recon_norm': 1.226012110710144, 'recon_norm_std': 0.044007453974336384, 'target_norm': 1.6213883399963378, 'recon_cos': 0.6699529647827148, 'gradnorm/compressor.query': 0.4439579624130917, 'gradnorm/compressor.layer0': 12.418747763784713, 'gradnorm/compressor.layer1': 1091.2988820666867, 'gradnorm/compressor.layer2': 153.04937545813442, 'gradnorm/compressor.layer3': 38.4515154833818, 'gradnorm/compressor.layer4': 9.66492336821037, 'gradnorm/compressor.layer5': 4.633239005978802, 'gradnorm/compressor.layer6': 2.5454322145151314, 'gradnorm/compressor.layer7': 3.041178224267735, 'gradnorm/decoder.layer0': 3.274505368218861, 'gradnorm/decoder.layer1': 1.8333105244807146, 'gradnorm/decoder.layer2': 1.0837717208533575, 'gradnorm/max_param': 770.1087646484375, 'epoch': 9.58}

 96%|█████████▌| 18290/19090 [8:55:41<24:02,  1.80s/it]
 96%|█████████▌| 18291/19090 [8:55:42<22:09,  1.66s/it]
 96%|█████████▌| 18292/19090 [8:55:44<24:31,  1.84s/it]
 96%|█████████▌| 18293/19090 [8:55:46<23:16,  1.75s/it]
 96%|█████████▌| 18294/19090 [8:55:47<21:24,  1.61s/it]
 96%|█████████▌| 18295/19090 [8:55:50<26:15,  1.98s/it]
 96%|█████████▌| 18296/19090 [8:55:51<23:30,  1.78s/it]
 96%|█████████▌| 18297/19090 [8:55:53<22:06,  1.67s/it]
 96%|█████████▌| 18298/19090 [8:55:56<27:04,  2.05s/it]
 96%|█████████▌| 18299/19090 [8:55:57<24:02,  1.82s/it][gradnorm] step 18299: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 565.8

 96%|█████████▌| 18300/19090 [8:55:58<21:54,  1.66s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 813.5530395507812, 'learning_rate': 2.610409298848315e-07, 'recon_loss': 0.0014058919216040522, 'recon_static': 0.00043289041277603244, 'recon_motion': 0.0009730015255627223, 'recon_norm': 1.2263966232538224, 'recon_norm_std': 0.0436142785474658, 'target_norm': 1.6195569902658462, 'recon_cos': 0.6702746585011482, 'gradnorm/compressor.query': 0.4320684358345498, 'gradnorm/compressor.layer0': 10.289294217801737, 'gradnorm/compressor.layer1': 801.946421215283, 'gradnorm/compressor.layer2': 130.03313880507537, 'gradnorm/compressor.layer3': 39.89335912829431, 'gradnorm/compressor.layer4': 9.516504507024969, 'gradnorm/compressor.layer5': 4.7308048094269814, 'gradnorm/compressor.layer6': 2.3509964879576404, 'gradnorm/compressor.layer7': 3.147631217760432, 'gradnorm/decoder.layer0': 3.277029611220341, 'gradnorm/decoder.layer1': 1.9527691326188148, 'gradnorm/decoder.layer2': 1.180020721059646, 'gradnorm/max_param': 565.8036499023438, 'epoch': 9.59}

 96%|█████████▌| 18300/19090 [8:55:58<21:54,  1.66s/it]
 96%|█████████▌| 18301/19090 [8:56:01<25:59,  1.98s/it]
 96%|█████████▌| 18302/19090 [8:56:02<23:15,  1.77s/it]
 96%|█████████▌| 18303/19090 [8:56:04<21:19,  1.63s/it]
 96%|█████████▌| 18304/19090 [8:56:07<26:10,  2.00s/it]
 96%|█████████▌| 18305/19090 [8:56:08<23:22,  1.79s/it]
 96%|█████████▌| 18306/19090 [8:56:09<21:24,  1.64s/it]
 96%|█████████▌| 18307/19090 [8:56:12<25:59,  1.99s/it]
 96%|█████████▌| 18308/19090 [8:56:13<23:13,  1.78s/it]
 96%|█████████▌| 18309/19090 [8:56:15<21:15,  1.63s/it][gradnorm] step 18309: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 618.5

 96%|█████████▌| 18310/19090 [8:56:18<26:42,  2.05s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 890.4022216796875, 'learning_rate': 2.5449352503124914e-07, 'recon_loss': 0.0014015846652910113, 'recon_static': 0.0004263439011992887, 'recon_motion': 0.0009752407670021058, 'recon_norm': 1.2237321525812148, 'recon_norm_std': 0.043830903433263305, 'target_norm': 1.618220689892769, 'recon_cos': 0.670472726225853, 'gradnorm/compressor.query': 0.7585569441536022, 'gradnorm/compressor.layer0': 18.93050403856952, 'gradnorm/compressor.layer1': 876.7140711201115, 'gradnorm/compressor.layer2': 150.45768715821734, 'gradnorm/compressor.layer3': 32.11040334423857, 'gradnorm/compressor.layer4': 10.396897337338869, 'gradnorm/compressor.layer5': 4.562526546035252, 'gradnorm/compressor.layer6': 2.6552791840917958, 'gradnorm/compressor.layer7': 3.112635144589703, 'gradnorm/decoder.layer0': 3.1900250774103385, 'gradnorm/decoder.layer1': 1.9404180149136212, 'gradnorm/decoder.layer2': 1.1515423072897542, 'gradnorm/max_param': 618.5365600585938, 'epoch': 9.59}

 96%|█████████▌| 18310/19090 [8:56:18<26:42,  2.05s/it]
 96%|█████████▌| 18311/19090 [8:56:19<23:43,  1.83s/it]
 96%|█████████▌| 18312/19090 [8:56:20<21:37,  1.67s/it]
 96%|█████████▌| 18313/19090 [8:56:23<25:20,  1.96s/it]
 96%|█████████▌| 18314/19090 [8:56:24<22:43,  1.76s/it]
 96%|█████████▌| 18315/19090 [8:56:25<20:53,  1.62s/it]
 96%|█████████▌| 18316/19090 [8:56:28<26:07,  2.03s/it]
 96%|█████████▌| 18317/19090 [8:56:30<23:14,  1.80s/it]
 96%|█████████▌| 18318/19090 [8:56:31<21:12,  1.65s/it]
 96%|█████████▌| 18319/19090 [8:56:34<25:31,  1.99s/it][gradnorm] step 18319: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1148.6

 96%|█████████▌| 18320/19090 [8:56:35<22:48,  1.78s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1638.0550537109375, 'learning_rate': 2.480288570493122e-07, 'recon_loss': 0.0013999411166878417, 'recon_static': 0.0004280139444745146, 'recon_motion': 0.0009719272013171576, 'recon_norm': 1.2250092178583145, 'recon_norm_std': 0.043583654798567295, 'target_norm': 1.6194172739982604, 'recon_cos': 0.6714451417326928, 'gradnorm/compressor.query': 0.4796262969197586, 'gradnorm/compressor.layer0': 11.602875688516363, 'gradnorm/compressor.layer1': 1627.3202665732397, 'gradnorm/compressor.layer2': 181.03295193831426, 'gradnorm/compressor.layer3': 44.729016717601084, 'gradnorm/compressor.layer4': 9.287222536442542, 'gradnorm/compressor.layer5': 5.082914470654726, 'gradnorm/compressor.layer6': 2.504807427593709, 'gradnorm/compressor.layer7': 3.23656592131424, 'gradnorm/decoder.layer0': 3.2896348178950534, 'gradnorm/decoder.layer1': 1.9746912968486319, 'gradnorm/decoder.layer2': 1.1867876425881143, 'gradnorm/max_param': 1148.5758056640625, 'epoch': 9.6}

 96%|█████████▌| 18320/19090 [8:56:35<22:48,  1.78s/it]
 96%|█████████▌| 18321/19090 [8:56:36<20:53,  1.63s/it]
 96%|█████████▌| 18322/19090 [8:56:39<24:11,  1.89s/it]
 96%|█████████▌| 18323/19090 [8:56:40<21:50,  1.71s/it]
 96%|█████████▌| 18324/19090 [8:56:41<20:11,  1.58s/it]
 96%|█████████▌| 18325/19090 [8:56:44<25:49,  2.03s/it]
 96%|█████████▌| 18326/19090 [8:56:46<22:58,  1.80s/it]
 96%|█████████▌| 18327/19090 [8:56:47<20:58,  1.65s/it]
 96%|█████████▌| 18328/19090 [8:56:50<24:18,  1.91s/it]
 96%|█████████▌| 18329/19090 [8:56:51<21:52,  1.72s/it][gradnorm] step 18329: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 766.0

 96%|█████████▌| 18330/19090 [8:56:52<20:11,  1.59s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1099.8316650390625, 'learning_rate': 2.4164694755370865e-07, 'recon_loss': 0.00140158673748374, 'recon_static': 0.00042092482690350155, 'recon_motion': 0.000980661938956473, 'recon_norm': 1.2210813730955123, 'recon_norm_std': 0.04284183029085398, 'target_norm': 1.6210316658020019, 'recon_cos': 0.6716269254684448, 'gradnorm/compressor.query': 0.3201495081544035, 'gradnorm/compressor.layer0': 7.1732203067445095, 'gradnorm/compressor.layer1': 1085.3752807209128, 'gradnorm/compressor.layer2': 171.34974285069703, 'gradnorm/compressor.layer3': 44.99417415630398, 'gradnorm/compressor.layer4': 9.563873790717109, 'gradnorm/compressor.layer5': 5.359575718271253, 'gradnorm/compressor.layer6': 2.7218851888577595, 'gradnorm/compressor.layer7': 3.21226885962266, 'gradnorm/decoder.layer0': 3.2006314131544564, 'gradnorm/decoder.layer1': 1.9396442115590922, 'gradnorm/decoder.layer2': 1.1836421944048845, 'gradnorm/max_param': 766.0454711914062, 'epoch': 9.6}

 96%|█████████▌| 18330/19090 [8:56:52<20:11,  1.59s/it]
 96%|█████████▌| 18331/19090 [8:56:54<23:02,  1.82s/it]
 96%|█████████▌| 18332/19090 [8:56:56<20:59,  1.66s/it]
 96%|█████████▌| 18333/19090 [8:56:57<19:37,  1.56s/it]
 96%|█████████▌| 18334/19090 [8:56:59<21:48,  1.73s/it]
 96%|█████████▌| 18335/19090 [8:57:00<20:05,  1.60s/it]
 96%|█████████▌| 18336/19090 [8:57:02<18:53,  1.50s/it]
 96%|█████████▌| 18337/19090 [8:57:04<22:29,  1.79s/it]
 96%|█████████▌| 18338/19090 [8:57:06<20:33,  1.64s/it]
 96%|█████████▌| 18339/19090 [8:57:07<19:13,  1.54s/it][gradnorm] step 18339: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 931.3

 96%|█████████▌| 18340/19090 [8:57:10<24:16,  1.94s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1330.7274169921875, 'learning_rate': 2.3534781788242531e-07, 'recon_loss': 0.0014012240862939507, 'recon_static': 0.00042136954070883805, 'recon_motion': 0.0009798545550438575, 'recon_norm': 1.2221404045820237, 'recon_norm_std': 0.04373272005468607, 'target_norm': 1.6201682329177856, 'recon_cos': 0.6715891465544701, 'gradnorm/compressor.query': 0.3654461007278025, 'gradnorm/compressor.layer0': 7.2529475041226235, 'gradnorm/compressor.layer1': 1319.4853163260286, 'gradnorm/compressor.layer2': 163.46591531187778, 'gradnorm/compressor.layer3': 53.49672339557349, 'gradnorm/compressor.layer4': 9.848037803124912, 'gradnorm/compressor.layer5': 5.371014559075299, 'gradnorm/compressor.layer6': 2.65310806251035, 'gradnorm/compressor.layer7': 3.137588627387975, 'gradnorm/decoder.layer0': 3.1493371478011083, 'gradnorm/decoder.layer1': 1.9573913364772542, 'gradnorm/decoder.layer2': 1.1924823341420643, 'gradnorm/max_param': 931.3375854492188, 'epoch': 9.61}

 96%|█████████▌| 18340/19090 [8:57:10<24:16,  1.94s/it]
 96%|█████████▌| 18341/19090 [8:57:11<21:47,  1.75s/it]
 96%|█████████▌| 18342/19090 [8:57:12<20:03,  1.61s/it]
 96%|█████████▌| 18343/19090 [8:57:15<23:28,  1.89s/it]
 96%|█████████▌| 18344/19090 [8:57:16<21:12,  1.71s/it]
 96%|█████████▌| 18345/19090 [8:57:17<19:37,  1.58s/it]
 96%|█████████▌| 18346/19090 [8:57:20<23:46,  1.92s/it]
 96%|█████████▌| 18347/19090 [8:57:21<21:24,  1.73s/it]
 96%|█████████▌| 18348/19090 [8:57:23<19:44,  1.60s/it]
 96%|█████████▌| 18349/19090 [8:57:25<23:44,  1.92s/it][gradnorm] step 18349: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1266.0

 96%|█████████▌| 18350/19090 [8:57:27<21:22,  1.73s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1806.048828125, 'learning_rate': 2.291314890966678e-07, 'recon_loss': 0.0014136606943793595, 'recon_static': 0.0004190553205262404, 'recon_motion': 0.0009946053847670554, 'recon_norm': 1.2197284817695617, 'recon_norm_std': 0.0425287994556129, 'target_norm': 1.6230355948209763, 'recon_cos': 0.6696813732385636, 'gradnorm/compressor.query': 0.7232337087885742, 'gradnorm/compressor.layer0': 24.282919022432207, 'gradnorm/compressor.layer1': 1793.9487172157403, 'gradnorm/compressor.layer2': 200.5078902749216, 'gradnorm/compressor.layer3': 50.66036091431594, 'gradnorm/compressor.layer4': 11.363341005128207, 'gradnorm/compressor.layer5': 5.932203962589264, 'gradnorm/compressor.layer6': 2.7767783231904084, 'gradnorm/compressor.layer7': 3.2914014459951146, 'gradnorm/decoder.layer0': 3.4215855584903654, 'gradnorm/decoder.layer1': 2.0180251151677964, 'gradnorm/decoder.layer2': 1.2703883637582403, 'gradnorm/max_param': 1265.954345703125, 'epoch': 9.61}

 96%|█████████▌| 18350/19090 [8:57:27<21:22,  1.73s/it]
 96%|█████████▌| 18351/19090 [8:57:28<19:41,  1.60s/it]
 96%|█████████▌| 18352/19090 [8:57:30<23:12,  1.89s/it]
 96%|█████████▌| 18353/19090 [8:57:32<20:58,  1.71s/it]
 96%|█████████▌| 18354/19090 [8:57:33<19:23,  1.58s/it]
 96%|█████████▌| 18355/19090 [8:57:36<22:51,  1.87s/it]
 96%|█████████▌| 18356/19090 [8:57:37<20:43,  1.69s/it]
 96%|█████████▌| 18357/19090 [8:57:38<19:13,  1.57s/it]
 96%|█████████▌| 18358/19090 [8:57:41<23:51,  1.96s/it]
 96%|█████████▌| 18359/19090 [8:57:42<21:22,  1.75s/it][gradnorm] step 18359: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 899.3

 96%|█████████▌| 18360/19090 [8:57:44<19:38,  1.61s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1286.0303955078125, 'learning_rate': 2.2299798198079923e-07, 'recon_loss': 0.0014051533507881686, 'recon_static': 0.0004156530943873804, 'recon_motion': 0.000989500254218001, 'recon_norm': 1.219348132610321, 'recon_norm_std': 0.04282995369285345, 'target_norm': 1.6215243130922317, 'recon_cos': 0.670909421145916, 'gradnorm/compressor.query': 0.4373758514550821, 'gradnorm/compressor.layer0': 12.428294086337207, 'gradnorm/compressor.layer1': 1274.382301352306, 'gradnorm/compressor.layer2': 167.0554443422153, 'gradnorm/compressor.layer3': 39.9597407873107, 'gradnorm/compressor.layer4': 10.225686614992995, 'gradnorm/compressor.layer5': 4.9748883028766215, 'gradnorm/compressor.layer6': 2.866932205351317, 'gradnorm/compressor.layer7': 3.3799257008438954, 'gradnorm/decoder.layer0': 3.2742056662266923, 'gradnorm/decoder.layer1': 2.017346497171083, 'gradnorm/decoder.layer2': 1.2335131086743851, 'gradnorm/max_param': 899.3125, 'epoch': 9.62}

 96%|█████████▌| 18360/19090 [8:57:44<19:38,  1.61s/it]
 96%|█████████▌| 18361/19090 [8:57:46<23:53,  1.97s/it]
 96%|█████████▌| 18362/19090 [8:57:48<21:22,  1.76s/it]
 96%|█████████▌| 18363/19090 [8:57:49<19:36,  1.62s/it]
 96%|█████████▌| 18364/19090 [8:57:52<24:18,  2.01s/it]
 96%|█████████▌| 18365/19090 [8:57:53<21:39,  1.79s/it]
 96%|█████████▌| 18366/19090 [8:57:54<19:47,  1.64s/it]
 96%|█████████▌| 18367/19090 [8:57:57<24:14,  2.01s/it]
 96%|█████████▌| 18368/19090 [8:57:59<21:35,  1.79s/it]
 96%|█████████▌| 18369/19090 [8:58:00<19:43,  1.64s/it][gradnorm] step 18369: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 912.4

 96%|█████████▌| 18370/19090 [8:58:03<24:52,  2.07s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1304.7738037109375, 'learning_rate': 2.169473170422709e-07, 'recon_loss': 0.001404771054512821, 'recon_static': 0.0004186115227639675, 'recon_motion': 0.0009861595739494079, 'recon_norm': 1.2192373365163802, 'recon_norm_std': 0.04315126752480865, 'target_norm': 1.6220911979675292, 'recon_cos': 0.6713266283273697, 'gradnorm/compressor.query': 0.41474749926640375, 'gradnorm/compressor.layer0': 8.96413821433947, 'gradnorm/compressor.layer1': 1292.7077202523392, 'gradnorm/compressor.layer2': 171.33133903164068, 'gradnorm/compressor.layer3': 41.704535850994375, 'gradnorm/compressor.layer4': 9.945442912047552, 'gradnorm/compressor.layer5': 5.732065177360579, 'gradnorm/compressor.layer6': 2.8295424476320656, 'gradnorm/compressor.layer7': 3.275917443051483, 'gradnorm/decoder.layer0': 3.3213523762250317, 'gradnorm/decoder.layer1': 2.0281633840728652, 'gradnorm/decoder.layer2': 1.2959488928602, 'gradnorm/max_param': 912.37841796875, 'epoch': 9.62}

 96%|█████████▌| 18370/19090 [8:58:03<24:52,  2.07s/it]
 96%|█████████▌| 18371/19090 [8:58:04<22:02,  1.84s/it]
 96%|█████████▌| 18372/19090 [8:58:06<20:01,  1.67s/it]
 96%|█████████▌| 18373/19090 [8:58:08<22:28,  1.88s/it]
 96%|█████████▌| 18374/19090 [8:58:09<20:20,  1.70s/it]
 96%|█████████▋| 18375/19090 [8:58:10<18:48,  1.58s/it]
 96%|█████████▋| 18376/19090 [8:58:13<23:52,  2.01s/it]
 96%|█████████▋| 18377/19090 [8:58:15<21:17,  1.79s/it]
 96%|█████████▋| 18378/19090 [8:58:16<19:27,  1.64s/it]
 96%|█████████▋| 18379/19090 [8:58:19<22:18,  1.88s/it][gradnorm] step 18379: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 890.1

 96%|█████████▋| 18380/19090 [8:58:20<20:10,  1.70s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1274.6888427734375, 'learning_rate': 2.1097951451154452e-07, 'recon_loss': 0.0014097024337388574, 'recon_static': 0.0004176044538326096, 'recon_motion': 0.0009920979355229065, 'recon_norm': 1.2187687754631042, 'recon_norm_std': 0.04224078524857759, 'target_norm': 1.6226702511310578, 'recon_cos': 0.6704959750175477, 'gradnorm/compressor.query': 0.43643281409861734, 'gradnorm/compressor.layer0': 11.294852132382381, 'gradnorm/compressor.layer1': 1261.1843144441657, 'gradnorm/compressor.layer2': 177.91772479681444, 'gradnorm/compressor.layer3': 47.97177565783473, 'gradnorm/compressor.layer4': 9.366484576680588, 'gradnorm/compressor.layer5': 6.081953162131177, 'gradnorm/compressor.layer6': 2.946120093244809, 'gradnorm/compressor.layer7': 3.4047015108868006, 'gradnorm/decoder.layer0': 3.3112427826828315, 'gradnorm/decoder.layer1': 2.0834086086661117, 'gradnorm/decoder.layer2': 1.3520240216209096, 'gradnorm/max_param': 890.1345825195312, 'epoch': 9.63}

 96%|█████████▋| 18380/19090 [8:58:20<20:10,  1.70s/it]
 96%|█████████▋| 18381/19090 [8:58:21<18:39,  1.58s/it]
 96%|█████████▋| 18382/19090 [8:58:24<22:37,  1.92s/it]
 96%|█████████▋| 18383/19090 [8:58:25<20:21,  1.73s/it]
 96%|█████████▋| 18384/19090 [8:58:26<18:45,  1.59s/it]
 96%|█████████▋| 18385/19090 [8:58:29<22:25,  1.91s/it]
 96%|█████████▋| 18386/19090 [8:58:30<20:12,  1.72s/it]
 96%|█████████▋| 18387/19090 [8:58:32<18:38,  1.59s/it]
 96%|█████████▋| 18388/19090 [8:58:35<23:30,  2.01s/it]
 96%|█████████▋| 18389/19090 [8:58:36<20:56,  1.79s/it][gradnorm] step 18389: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 816.9

 96%|█████████▋| 18390/19090 [8:58:37<19:09,  1.64s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1168.4454345703125, 'learning_rate': 2.050945943420368e-07, 'recon_loss': 0.0014017878216691314, 'recon_static': 0.0004181970267381985, 'recon_motion': 0.0009835908233071678, 'recon_norm': 1.219329047203064, 'recon_norm_std': 0.04345841007307172, 'target_norm': 1.6204441100358964, 'recon_cos': 0.6710708051919937, 'gradnorm/compressor.query': 0.5848928166688442, 'gradnorm/compressor.layer0': 19.089115160922287, 'gradnorm/compressor.layer1': 1157.9826747408615, 'gradnorm/compressor.layer2': 148.09768999295702, 'gradnorm/compressor.layer3': 43.32446088681096, 'gradnorm/compressor.layer4': 10.184376243551435, 'gradnorm/compressor.layer5': 5.213273515642678, 'gradnorm/compressor.layer6': 2.8391134170392016, 'gradnorm/compressor.layer7': 3.310289761171593, 'gradnorm/decoder.layer0': 3.216967339043658, 'gradnorm/decoder.layer1': 2.0238817824492203, 'gradnorm/decoder.layer2': 1.3025607289308223, 'gradnorm/max_param': 816.9474487304688, 'epoch': 9.63}

 96%|█████████▋| 18390/19090 [8:58:37<19:09,  1.64s/it]
 96%|█████████▋| 18391/19090 [8:58:40<22:35,  1.94s/it]
 96%|█████████▋| 18392/19090 [8:58:41<20:17,  1.74s/it]
 96%|█████████▋| 18393/19090 [8:58:42<18:39,  1.61s/it]
 96%|█████████▋| 18394/19090 [8:58:45<21:10,  1.83s/it]
 96%|█████████▋| 18395/19090 [8:58:46<19:16,  1.66s/it]
 96%|█████████▋| 18396/19090 [8:58:47<17:55,  1.55s/it]
 96%|█████████▋| 18397/19090 [8:58:50<20:35,  1.78s/it]
 96%|█████████▋| 18398/19090 [8:58:51<18:50,  1.63s/it]
 96%|█████████▋| 18399/19090 [8:58:52<17:37,  1.53s/it][gradnorm] step 18399: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1004.0

 96%|█████████▋| 18400/19090 [8:58:55<22:13,  1.93s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1433.111083984375, 'learning_rate': 1.9929257621003895e-07, 'recon_loss': 0.001400422482402064, 'recon_static': 0.00042446192019269803, 'recon_motion': 0.0009759605411090889, 'recon_norm': 1.2238065242767333, 'recon_norm_std': 0.042837774567306044, 'target_norm': 1.6213791221380234, 'recon_cos': 0.6723579838871956, 'gradnorm/compressor.query': 0.3743296632078689, 'gradnorm/compressor.layer0': 9.470103554245064, 'gradnorm/compressor.layer1': 1422.5709121165103, 'gradnorm/compressor.layer2': 167.25895669216044, 'gradnorm/compressor.layer3': 43.20652111425861, 'gradnorm/compressor.layer4': 10.03832074107922, 'gradnorm/compressor.layer5': 5.519391813359501, 'gradnorm/compressor.layer6': 2.851822013019977, 'gradnorm/compressor.layer7': 3.2865611371114323, 'gradnorm/decoder.layer0': 3.3293573667855245, 'gradnorm/decoder.layer1': 2.084670349883138, 'gradnorm/decoder.layer2': 1.3159508525865773, 'gradnorm/max_param': 1004.0176391601562, 'epoch': 9.64}

 96%|█████████▋| 18400/19090 [8:58:55<22:13,  1.93s/it]
 96%|█████████▋| 18401/19090 [8:58:56<19:58,  1.74s/it]
 96%|█████████▋| 18402/19090 [8:58:58<18:24,  1.61s/it]
 96%|█████████▋| 18403/19090 [8:59:00<22:15,  1.94s/it]
 96%|█████████▋| 18404/19090 [8:59:02<19:58,  1.75s/it]
 96%|█████████▋| 18405/19090 [8:59:03<18:22,  1.61s/it]
 96%|█████████▋| 18406/19090 [8:59:06<21:51,  1.92s/it]
 96%|█████████▋| 18407/19090 [8:59:07<19:42,  1.73s/it]
 96%|█████████▋| 18408/19090 [8:59:08<18:09,  1.60s/it]
 96%|█████████▋| 18409/19090 [8:59:11<23:11,  2.04s/it][gradnorm] step 18409: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 961.4

 96%|█████████▋| 18410/19090 [8:59:13<20:35,  1.82s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1374.6094970703125, 'learning_rate': 1.9357347951466376e-07, 'recon_loss': 0.0014022067683981732, 'recon_static': 0.00041749091615201906, 'recon_motion': 0.0009847158406046219, 'recon_norm': 1.2209269016981126, 'recon_norm_std': 0.04228703044354916, 'target_norm': 1.6208430349826812, 'recon_cos': 0.6714527472853661, 'gradnorm/compressor.query': 0.45170292832208875, 'gradnorm/compressor.layer0': 9.684352043348678, 'gradnorm/compressor.layer1': 1362.1965717179, 'gradnorm/compressor.layer2': 178.2605460151657, 'gradnorm/compressor.layer3': 43.95734904604974, 'gradnorm/compressor.layer4': 9.953998425545471, 'gradnorm/compressor.layer5': 5.665548027176381, 'gradnorm/compressor.layer6': 2.951656238688984, 'gradnorm/compressor.layer7': 3.258699803991865, 'gradnorm/decoder.layer0': 3.4063037553097786, 'gradnorm/decoder.layer1': 2.1277453695916257, 'gradnorm/decoder.layer2': 1.365681412707931, 'gradnorm/max_param': 961.4070434570312, 'epoch': 9.64}

 96%|█████████▋| 18410/19090 [8:59:13<20:35,  1.82s/it]
 96%|█████████▋| 18411/19090 [8:59:14<18:47,  1.66s/it]
 96%|█████████▋| 18412/19090 [8:59:16<20:57,  1.86s/it]
 96%|█████████▋| 18413/19090 [8:59:17<19:00,  1.68s/it]
 96%|█████████▋| 18414/19090 [8:59:19<17:37,  1.56s/it]
 96%|█████████▋| 18415/19090 [8:59:22<22:13,  1.98s/it]
 96%|█████████▋| 18416/19090 [8:59:23<19:53,  1.77s/it]
 96%|█████████▋| 18417/19090 [8:59:24<18:15,  1.63s/it]
 96%|█████████▋| 18418/19090 [8:59:27<21:22,  1.91s/it]
 96%|█████████▋| 18419/19090 [8:59:28<19:18,  1.73s/it][gradnorm] step 18419: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1019.2

 96%|█████████▋| 18420/19090 [8:59:29<17:51,  1.60s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1457.6563720703125, 'learning_rate': 1.8793732337777093e-07, 'recon_loss': 0.0014019996480783448, 'recon_static': 0.0004151983972406015, 'recon_motion': 0.0009868012173683382, 'recon_norm': 1.2171858161687852, 'recon_norm_std': 0.04270154805853963, 'target_norm': 1.6197735249996186, 'recon_cos': 0.670472939312458, 'gradnorm/compressor.query': 0.444971707092914, 'gradnorm/compressor.layer0': 10.687580666042436, 'gradnorm/compressor.layer1': 1443.9208773336577, 'gradnorm/compressor.layer2': 193.09502194515528, 'gradnorm/compressor.layer3': 47.605480128670585, 'gradnorm/compressor.layer4': 10.795170827541396, 'gradnorm/compressor.layer5': 5.653866766348592, 'gradnorm/compressor.layer6': 3.0847136183112087, 'gradnorm/compressor.layer7': 3.4287875652306705, 'gradnorm/decoder.layer0': 3.4698784169820267, 'gradnorm/decoder.layer1': 2.1556308244342817, 'gradnorm/decoder.layer2': 1.3732296644281916, 'gradnorm/max_param': 1019.1563110351562, 'epoch': 9.65}

 96%|█████████▋| 18420/19090 [8:59:29<17:51,  1.60s/it]
 96%|█████████▋| 18421/19090 [8:59:32<21:51,  1.96s/it]
 97%|█████████▋| 18422/19090 [8:59:33<19:35,  1.76s/it]
 97%|█████████▋| 18423/19090 [8:59:35<18:00,  1.62s/it]
 97%|█████████▋| 18424/19090 [8:59:37<20:51,  1.88s/it]
 97%|█████████▋| 18425/19090 [8:59:39<18:52,  1.70s/it]
 97%|█████████▋| 18426/19090 [8:59:40<17:28,  1.58s/it]
 97%|█████████▋| 18427/19090 [8:59:42<20:42,  1.87s/it]
 97%|█████████▋| 18428/19090 [8:59:44<18:44,  1.70s/it]
 97%|█████████▋| 18429/19090 [8:59:45<17:22,  1.58s/it][gradnorm] step 18429: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 961.9

 97%|█████████▋| 18430/19090 [8:59:48<21:37,  1.97s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1376.0340576171875, 'learning_rate': 1.8238412664391414e-07, 'recon_loss': 0.0014001284871483222, 'recon_static': 0.00041743920664885084, 'recon_motion': 0.0009826892812270672, 'recon_norm': 1.2211478054523468, 'recon_norm_std': 0.04265520256012678, 'target_norm': 1.6200626820325852, 'recon_cos': 0.6718044087290764, 'gradnorm/compressor.query': 0.5071508351797971, 'gradnorm/compressor.layer0': 13.607756456029097, 'gradnorm/compressor.layer1': 1363.0551162737331, 'gradnorm/compressor.layer2': 180.6874243341246, 'gradnorm/compressor.layer3': 50.154580384983284, 'gradnorm/compressor.layer4': 11.442519790468944, 'gradnorm/compressor.layer5': 5.765376553441979, 'gradnorm/compressor.layer6': 2.9015722879584676, 'gradnorm/compressor.layer7': 3.3898234320825327, 'gradnorm/decoder.layer0': 3.3530959953767243, 'gradnorm/decoder.layer1': 2.144538921071492, 'gradnorm/decoder.layer2': 1.3305447245939903, 'gradnorm/max_param': 961.884033203125, 'epoch': 9.65}

 97%|█████████▋| 18430/19090 [8:59:48<21:37,  1.97s/it]
 97%|█████████▋| 18431/19090 [8:59:49<19:21,  1.76s/it]
 97%|█████████▋| 18432/19090 [8:59:50<17:45,  1.62s/it]
 97%|█████████▋| 18433/19090 [8:59:53<20:54,  1.91s/it]
 97%|█████████▋| 18434/19090 [8:59:54<18:50,  1.72s/it]
 97%|█████████▋| 18435/19090 [8:59:56<17:23,  1.59s/it]
 97%|█████████▋| 18436/19090 [8:59:59<21:53,  2.01s/it]
 97%|█████████▋| 18437/19090 [9:00:00<19:29,  1.79s/it]
 97%|█████████▋| 18438/19090 [9:00:01<17:49,  1.64s/it]
 97%|█████████▋| 18439/19090 [9:00:04<21:57,  2.02s/it][gradnorm] step 18439: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 953.6

 97%|█████████▋| 18440/19090 [9:00:05<19:31,  1.80s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1363.14990234375, 'learning_rate': 1.7691390788026618e-07, 'recon_loss': 0.001396011668839492, 'recon_static': 0.0004150469474552665, 'recon_motion': 0.0009809647905058227, 'recon_norm': 1.217628502845764, 'recon_norm_std': 0.04193841628730297, 'target_norm': 1.6183156102895737, 'recon_cos': 0.671743443608284, 'gradnorm/compressor.query': 0.4575214365590747, 'gradnorm/compressor.layer0': 9.343445431647776, 'gradnorm/compressor.layer1': 1351.1161127009034, 'gradnorm/compressor.layer2': 175.48095182419374, 'gradnorm/compressor.layer3': 40.190604224977136, 'gradnorm/compressor.layer4': 9.796146055417022, 'gradnorm/compressor.layer5': 5.722086615740547, 'gradnorm/compressor.layer6': 2.9187943645103474, 'gradnorm/compressor.layer7': 3.377232837289734, 'gradnorm/decoder.layer0': 3.4526605768448952, 'gradnorm/decoder.layer1': 2.1324691775599964, 'gradnorm/decoder.layer2': 1.268588376203578, 'gradnorm/max_param': 953.5806274414062, 'epoch': 9.66}

 97%|█████████▋| 18440/19090 [9:00:05<19:31,  1.80s/it]
 97%|█████████▋| 18441/19090 [9:00:07<17:49,  1.65s/it]
 97%|█████████▋| 18442/19090 [9:00:09<21:18,  1.97s/it]
 97%|█████████▋| 18443/19090 [9:00:11<19:02,  1.77s/it]
 97%|█████████▋| 18444/19090 [9:00:12<17:28,  1.62s/it]
 97%|█████████▋| 18445/19090 [9:00:15<20:34,  1.91s/it]
 97%|█████████▋| 18446/19090 [9:00:16<18:34,  1.73s/it]
 97%|█████████▋| 18447/19090 [9:00:17<17:08,  1.60s/it]
 97%|█████████▋| 18448/19090 [9:00:20<19:47,  1.85s/it]
 97%|█████████▋| 18449/19090 [9:00:21<17:57,  1.68s/it][gradnorm] step 18449: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1095.6

 97%|█████████▋| 18450/19090 [9:00:22<16:41,  1.57s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1565.810302734375, 'learning_rate': 1.715266853765607e-07, 'recon_loss': 0.0013936554663814605, 'recon_static': 0.00042035393562400714, 'recon_motion': 0.000973301574413199, 'recon_norm': 1.2180485427379608, 'recon_norm_std': 0.04308904223144054, 'target_norm': 1.6198568493127823, 'recon_cos': 0.6729028701782227, 'gradnorm/compressor.query': 0.4355275102766692, 'gradnorm/compressor.layer0': 8.764918660963191, 'gradnorm/compressor.layer1': 1552.3055433773338, 'gradnorm/compressor.layer2': 198.37368068244336, 'gradnorm/compressor.layer3': 49.68453951479147, 'gradnorm/compressor.layer4': 11.87579214565976, 'gradnorm/compressor.layer5': 5.58110911158394, 'gradnorm/compressor.layer6': 2.945604385766821, 'gradnorm/compressor.layer7': 3.3485896657598686, 'gradnorm/decoder.layer0': 3.5955636133649347, 'gradnorm/decoder.layer1': 2.1472375466459397, 'gradnorm/decoder.layer2': 1.2926479529399597, 'gradnorm/max_param': 1095.6029052734375, 'epoch': 9.66}

 97%|█████████▋| 18450/19090 [9:00:22<16:41,  1.57s/it]
 97%|█████████▋| 18451/19090 [9:00:25<19:51,  1.87s/it]
 97%|█████████▋| 18452/19090 [9:00:26<18:06,  1.70s/it]
 97%|█████████▋| 18453/19090 [9:00:27<16:53,  1.59s/it]
 97%|█████████▋| 18454/19090 [9:00:30<19:52,  1.88s/it]
 97%|█████████▋| 18455/19090 [9:00:31<18:05,  1.71s/it]
 97%|█████████▋| 18456/19090 [9:00:33<16:45,  1.59s/it]
 97%|█████████▋| 18457/19090 [9:00:35<20:27,  1.94s/it]
 97%|█████████▋| 18458/19090 [9:00:37<18:22,  1.74s/it]
 97%|█████████▋| 18459/19090 [9:00:38<16:55,  1.61s/it][gradnorm] step 18459: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 826.3

 97%|█████████▋| 18460/19090 [9:00:41<20:46,  1.98s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1182.4263916015625, 'learning_rate': 1.6622247714503936e-07, 'recon_loss': 0.0014006124925799667, 'recon_static': 0.0004183401135378517, 'recon_motion': 0.0009822724226978608, 'recon_norm': 1.2194959789514541, 'recon_norm_std': 0.043346654437482354, 'target_norm': 1.6202570915222168, 'recon_cos': 0.6715390145778656, 'gradnorm/compressor.query': 0.6275109634733493, 'gradnorm/compressor.layer0': 19.051088413173733, 'gradnorm/compressor.layer1': 1170.9717652445768, 'gradnorm/compressor.layer2': 157.34633500855685, 'gradnorm/compressor.layer3': 40.76578220791841, 'gradnorm/compressor.layer4': 10.224067447732397, 'gradnorm/compressor.layer5': 5.581196428236211, 'gradnorm/compressor.layer6': 2.923400806908288, 'gradnorm/compressor.layer7': 3.375213475010446, 'gradnorm/decoder.layer0': 3.5172237575700853, 'gradnorm/decoder.layer1': 2.106929802918809, 'gradnorm/decoder.layer2': 1.2564534492294437, 'gradnorm/max_param': 826.285888671875, 'epoch': 9.67}

 97%|█████████▋| 18460/19090 [9:00:41<20:46,  1.98s/it]
 97%|█████████▋| 18461/19090 [9:00:42<18:35,  1.77s/it]
 97%|█████████▋| 18462/19090 [9:00:43<17:02,  1.63s/it]
 97%|█████████▋| 18463/19090 [9:00:46<20:40,  1.98s/it]
 97%|█████████▋| 18464/19090 [9:00:47<18:34,  1.78s/it]
 97%|█████████▋| 18465/19090 [9:00:49<17:01,  1.63s/it]
 97%|█████████▋| 18466/19090 [9:00:51<19:55,  1.92s/it]
 97%|█████████▋| 18467/19090 [9:00:53<17:55,  1.73s/it]
 97%|█████████▋| 18468/19090 [9:00:54<16:33,  1.60s/it]
 97%|█████████▋| 18469/19090 [9:00:56<19:29,  1.88s/it][gradnorm] step 18469: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 977.0

 97%|█████████▋| 18470/19090 [9:00:58<17:38,  1.71s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1395.52490234375, 'learning_rate': 1.6100130092037703e-07, 'recon_loss': 0.0013960367301478981, 'recon_static': 0.00042106632390641606, 'recon_motion': 0.0009749703749548644, 'recon_norm': 1.2195706993341446, 'recon_norm_std': 0.043626993522048, 'target_norm': 1.619350764155388, 'recon_cos': 0.6724183529615402, 'gradnorm/compressor.query': 0.519943045422301, 'gradnorm/compressor.layer0': 17.230215032279595, 'gradnorm/compressor.layer1': 1384.4096214632432, 'gradnorm/compressor.layer2': 168.12458178386646, 'gradnorm/compressor.layer3': 46.63994712656469, 'gradnorm/compressor.layer4': 9.76878933108485, 'gradnorm/compressor.layer5': 5.429027226321996, 'gradnorm/compressor.layer6': 2.9205805721287503, 'gradnorm/compressor.layer7': 3.2445737582902687, 'gradnorm/decoder.layer0': 3.4372638968100016, 'gradnorm/decoder.layer1': 2.0530570055242325, 'gradnorm/decoder.layer2': 1.2385364353763522, 'gradnorm/max_param': 977.0040893554688, 'epoch': 9.68}

 97%|█████████▋| 18470/19090 [9:00:58<17:38,  1.71s/it]
 97%|█████████▋| 18471/19090 [9:00:59<16:19,  1.58s/it]
 97%|█████████▋| 18472/19090 [9:01:01<18:59,  1.84s/it]
 97%|█████████▋| 18473/19090 [9:01:03<17:15,  1.68s/it]
 97%|█████████▋| 18474/19090 [9:01:04<16:02,  1.56s/it]
 97%|█████████▋| 18475/19090 [9:01:07<20:25,  1.99s/it]
 97%|█████████▋| 18476/19090 [9:01:08<18:13,  1.78s/it]
 97%|█████████▋| 18477/19090 [9:01:10<16:40,  1.63s/it]
 97%|█████████▋| 18478/19090 [9:01:12<18:51,  1.85s/it]
 97%|█████████▋| 18479/19090 [9:01:13<17:07,  1.68s/it][gradnorm] step 18479: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 947.0

 97%|█████████▋| 18480/19090 [9:01:15<15:54,  1.57s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1354.998046875, 'learning_rate': 1.5586317415963725e-07, 'recon_loss': 0.0014036854554433376, 'recon_static': 0.0004161500524787698, 'recon_motion': 0.0009875354066025467, 'recon_norm': 1.2186976492404937, 'recon_norm_std': 0.04283210253342986, 'target_norm': 1.6207679778337478, 'recon_cos': 0.6708198577165604, 'gradnorm/compressor.query': 0.3922673983705721, 'gradnorm/compressor.layer0': 9.404504759838874, 'gradnorm/compressor.layer1': 1341.6712059964618, 'gradnorm/compressor.layer2': 183.69697364137494, 'gradnorm/compressor.layer3': 43.89531669922202, 'gradnorm/compressor.layer4': 10.422024704345517, 'gradnorm/compressor.layer5': 5.730698407335693, 'gradnorm/compressor.layer6': 2.8366284750292343, 'gradnorm/compressor.layer7': 3.285665538378843, 'gradnorm/decoder.layer0': 3.480732883145243, 'gradnorm/decoder.layer1': 2.0328197575878684, 'gradnorm/decoder.layer2': 1.2256489650128262, 'gradnorm/max_param': 946.9534912109375, 'epoch': 9.68}

 97%|█████████▋| 18480/19090 [9:01:15<15:54,  1.57s/it]
 97%|█████████▋| 18481/19090 [9:01:17<19:04,  1.88s/it]
 97%|█████████▋| 18482/19090 [9:01:18<17:16,  1.70s/it]
 97%|█████████▋| 18483/19090 [9:01:20<15:58,  1.58s/it]
 97%|█████████▋| 18484/19090 [9:01:22<19:17,  1.91s/it]
 97%|█████████▋| 18485/19090 [9:01:24<18:01,  1.79s/it]
 97%|█████████▋| 18486/19090 [9:01:25<16:29,  1.64s/it]
 97%|█████████▋| 18487/19090 [9:01:28<18:39,  1.86s/it]
 97%|█████████▋| 18488/19090 [9:01:29<17:44,  1.77s/it]
 97%|█████████▋| 18489/19090 [9:01:30<16:15,  1.62s/it][gradnorm] step 18489: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 967.0

 97%|█████████▋| 18490/19090 [9:01:33<18:46,  1.88s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1382.4140625, 'learning_rate': 1.508081140421974e-07, 'recon_loss': 0.0013968326151371003, 'recon_static': 0.00041895581598510033, 'recon_motion': 0.0009778767969692125, 'recon_norm': 1.2231140047311784, 'recon_norm_std': 0.042819577362388374, 'target_norm': 1.6192279636859894, 'recon_cos': 0.6723997235298157, 'gradnorm/compressor.query': 0.42085218064748325, 'gradnorm/compressor.layer0': 11.270792755975267, 'gradnorm/compressor.layer1': 1370.036450245029, 'gradnorm/compressor.layer2': 178.44564889448833, 'gradnorm/compressor.layer3': 43.867794825493135, 'gradnorm/compressor.layer4': 10.112026238395353, 'gradnorm/compressor.layer5': 5.961191234462948, 'gradnorm/compressor.layer6': 2.8095138804019357, 'gradnorm/compressor.layer7': 3.201082350553771, 'gradnorm/decoder.layer0': 3.563203976325843, 'gradnorm/decoder.layer1': 2.0313194849894134, 'gradnorm/decoder.layer2': 1.200476815939998, 'gradnorm/max_param': 966.96484375, 'epoch': 9.69}

 97%|█████████▋| 18490/19090 [9:01:33<18:46,  1.88s/it]
 97%|█████████▋| 18491/19090 [9:01:34<17:10,  1.72s/it]
 97%|█████████▋| 18492/19090 [9:01:36<15:50,  1.59s/it]
 97%|█████████▋| 18493/19090 [9:01:38<19:15,  1.94s/it]
 97%|█████████▋| 18494/19090 [9:01:40<17:21,  1.75s/it]
 97%|█████████▋| 18495/19090 [9:01:41<15:57,  1.61s/it]
 97%|█████████▋| 18496/19090 [9:01:44<19:09,  1.93s/it]
 97%|█████████▋| 18497/19090 [9:01:45<17:37,  1.78s/it]
 97%|█████████▋| 18498/19090 [9:01:46<16:08,  1.64s/it]
 97%|█████████▋| 18499/19090 [9:01:49<18:29,  1.88s/it][gradnorm] step 18499: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 339.7

 97%|█████████▋| 18500/19090 [9:01:51<19:00,  1.93s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 500.42877197265625, 'learning_rate': 1.458361374697126e-07, 'recon_loss': 0.0014007941324962303, 'recon_static': 0.0004214111133478582, 'recon_motion': 0.0009793830366106703, 'recon_norm': 1.2278793275356292, 'recon_norm_std': 0.041604493465274575, 'target_norm': 1.6180617392063141, 'recon_cos': 0.6707726463675499, 'gradnorm/compressor.query': 1.287406921725817, 'gradnorm/compressor.layer0': 30.831902328138543, 'gradnorm/compressor.layer1': 484.6277888751325, 'gradnorm/compressor.layer2': 114.53098141169052, 'gradnorm/compressor.layer3': 36.57943818392462, 'gradnorm/compressor.layer4': 9.601155259238949, 'gradnorm/compressor.layer5': 5.479726359085733, 'gradnorm/compressor.layer6': 2.7850568627899484, 'gradnorm/compressor.layer7': 3.12488219992814, 'gradnorm/decoder.layer0': 3.4807982283461247, 'gradnorm/decoder.layer1': 1.979430639498564, 'gradnorm/decoder.layer2': 1.1875600799621013, 'gradnorm/max_param': 339.71246337890625, 'epoch': 9.69}

 97%|█████████▋| 18500/19090 [9:01:51<19:00,  1.93s/it]
 97%|█████████▋| 18501/19090 [9:01:54<23:47,  2.42s/it]
 97%|█████████▋| 18502/19090 [9:01:56<20:23,  2.08s/it]
 97%|█████████▋| 18503/19090 [9:01:57<18:02,  1.84s/it]
 97%|█████████▋| 18504/19090 [9:01:58<16:23,  1.68s/it]
 97%|█████████▋| 18505/19090 [9:02:00<15:13,  1.56s/it]
 97%|█████████▋| 18506/19090 [9:02:01<14:27,  1.49s/it]
 97%|█████████▋| 18507/19090 [9:02:02<13:51,  1.43s/it]
 97%|█████████▋| 18508/19090 [9:02:04<14:34,  1.50s/it]
 97%|█████████▋| 18509/19090 [9:02:06<16:43,  1.73s/it][gradnorm] step 18509: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 775.1

 97%|█████████▋| 18510/19090 [9:02:07<15:25,  1.60s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1107.056884765625, 'learning_rate': 1.4094726106603505e-07, 'recon_loss': 0.0014180552301695571, 'recon_static': 0.00043386676043155603, 'recon_motion': 0.0009841885039350018, 'recon_norm': 1.2228909462690354, 'recon_norm_std': 0.04203474540263415, 'target_norm': 1.6191497921943665, 'recon_cos': 0.6658835157752037, 'gradnorm/compressor.query': 0.3853208341107131, 'gradnorm/compressor.layer0': 16.092419734160263, 'gradnorm/compressor.layer1': 1098.1831586761837, 'gradnorm/compressor.layer2': 130.9587766126234, 'gradnorm/compressor.layer3': 44.68102280418528, 'gradnorm/compressor.layer4': 9.881242541431602, 'gradnorm/compressor.layer5': 5.4035593735239225, 'gradnorm/compressor.layer6': 2.8493357704174787, 'gradnorm/compressor.layer7': 3.273843307651079, 'gradnorm/decoder.layer0': 3.414828356426926, 'gradnorm/decoder.layer1': 1.989340928840881, 'gradnorm/decoder.layer2': 1.231897934961899, 'gradnorm/max_param': 775.1309814453125, 'epoch': 9.7}

 97%|█████████▋| 18510/19090 [9:02:07<15:25,  1.60s/it]
 97%|█████████▋| 18511/19090 [9:02:09<15:41,  1.63s/it]
 97%|█████████▋| 18512/19090 [9:02:11<17:27,  1.81s/it]
 97%|█████████▋| 18513/19090 [9:02:13<15:54,  1.65s/it]
 97%|█████████▋| 18514/19090 [9:02:14<15:45,  1.64s/it]
 97%|█████████▋| 18515/19090 [9:02:16<17:08,  1.79s/it]
 97%|█████████▋| 18516/19090 [9:02:18<15:40,  1.64s/it]
 97%|█████████▋| 18517/19090 [9:02:20<17:12,  1.80s/it]
 97%|█████████▋| 18518/19090 [9:02:22<17:14,  1.81s/it]
 97%|█████████▋| 18519/19090 [9:02:23<15:43,  1.65s/it][gradnorm] step 18519: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 954.5

 97%|█████████▋| 18520/19090 [9:02:25<17:18,  1.82s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1364.55859375, 'learning_rate': 1.3614150117718105e-07, 'recon_loss': 0.0014182057813741267, 'recon_static': 0.0004364385546068661, 'recon_motion': 0.0009817672049393877, 'recon_norm': 1.2238898873329163, 'recon_norm_std': 0.04223877405747771, 'target_norm': 1.6202670872211455, 'recon_cos': 0.6662898987531662, 'gradnorm/compressor.query': 0.38212106503153626, 'gradnorm/compressor.layer0': 10.960680653396585, 'gradnorm/compressor.layer1': 1352.437983790754, 'gradnorm/compressor.layer2': 175.30197160899246, 'gradnorm/compressor.layer3': 43.64619940677825, 'gradnorm/compressor.layer4': 10.269143069284866, 'gradnorm/compressor.layer5': 5.746364605275044, 'gradnorm/compressor.layer6': 2.9554142228213274, 'gradnorm/compressor.layer7': 3.1583898863011135, 'gradnorm/decoder.layer0': 3.566852654453439, 'gradnorm/decoder.layer1': 1.9590395675292998, 'gradnorm/decoder.layer2': 1.2270975100222408, 'gradnorm/max_param': 954.4716186523438, 'epoch': 9.7}

 97%|█████████▋| 18520/19090 [9:02:25<17:18,  1.82s/it]
 97%|█████████▋| 18521/19090 [9:02:27<17:12,  1.82s/it]
 97%|█████████▋| 18522/19090 [9:02:28<15:42,  1.66s/it]
 97%|█████████▋| 18523/19090 [9:02:30<17:03,  1.80s/it]
 97%|█████████▋| 18524/19090 [9:02:32<16:42,  1.77s/it]
 97%|█████████▋| 18525/19090 [9:02:33<15:18,  1.63s/it]
 97%|█████████▋| 18526/19090 [9:02:36<17:17,  1.84s/it]
 97%|█████████▋| 18527/19090 [9:02:37<17:14,  1.84s/it]
 97%|█████████▋| 18528/19090 [9:02:39<15:39,  1.67s/it]
 97%|█████████▋| 18529/19090 [9:02:41<17:09,  1.83s/it][gradnorm] step 18529: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 916.1

 97%|█████████▋| 18530/19090 [9:02:42<15:59,  1.71s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1311.2059326171875, 'learning_rate': 1.314188738712585e-07, 'recon_loss': 0.0014316334913019092, 'recon_static': 0.00043540968181332573, 'recon_motion': 0.0009962238153093494, 'recon_norm': 1.2239775359630585, 'recon_norm_std': 0.041605105716735126, 'target_norm': 1.6252056777477264, 'recon_cos': 0.6656681641936302, 'gradnorm/compressor.query': 0.3941867570750619, 'gradnorm/compressor.layer0': 9.882854959147922, 'gradnorm/compressor.layer1': 1297.9868065585258, 'gradnorm/compressor.layer2': 178.89286794929808, 'gradnorm/compressor.layer3': 46.98189634299977, 'gradnorm/compressor.layer4': 10.904069305401181, 'gradnorm/compressor.layer5': 5.304196319537146, 'gradnorm/compressor.layer6': 2.745169298131325, 'gradnorm/compressor.layer7': 3.25053650755919, 'gradnorm/decoder.layer0': 3.6057521588826127, 'gradnorm/decoder.layer1': 1.9668721522724881, 'gradnorm/decoder.layer2': 1.2281319363230738, 'gradnorm/max_param': 916.1112060546875, 'epoch': 9.71}

 97%|█████████▋| 18530/19090 [9:02:42<15:59,  1.71s/it]
 97%|█████████▋| 18531/19090 [9:02:44<14:46,  1.59s/it]
 97%|█████████▋| 18532/19090 [9:02:46<17:57,  1.93s/it]
 97%|█████████▋| 18533/19090 [9:02:48<16:19,  1.76s/it]
 97%|█████████▋| 18534/19090 [9:02:49<14:59,  1.62s/it]
 97%|█████████▋| 18535/19090 [9:02:52<17:32,  1.90s/it]
 97%|█████████▋| 18536/19090 [9:02:53<17:07,  1.85s/it]
 97%|█████████▋| 18537/19090 [9:02:55<15:32,  1.69s/it]
 97%|█████████▋| 18538/19090 [9:02:57<17:04,  1.86s/it]
 97%|█████████▋| 18539/19090 [9:02:59<16:28,  1.79s/it][gradnorm] step 18539: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 948.7

 97%|█████████▋| 18540/19090 [9:03:00<15:05,  1.65s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1356.121826171875, 'learning_rate': 1.267793949384255e-07, 'recon_loss': 0.0014152845105854794, 'recon_static': 0.0004380831320304424, 'recon_motion': 0.0009772013887413777, 'recon_norm': 1.2319487780332565, 'recon_norm_std': 0.042854031547904015, 'target_norm': 1.620783331990242, 'recon_cos': 0.6684192150831223, 'gradnorm/compressor.query': 0.3707539659212122, 'gradnorm/compressor.layer0': 9.405210114342044, 'gradnorm/compressor.layer1': 1344.1199723238994, 'gradnorm/compressor.layer2': 173.0549497269581, 'gradnorm/compressor.layer3': 47.04068634570503, 'gradnorm/compressor.layer4': 9.780876761923421, 'gradnorm/compressor.layer5': 5.32065360478433, 'gradnorm/compressor.layer6': 2.5879802645147474, 'gradnorm/compressor.layer7': 3.122695989024254, 'gradnorm/decoder.layer0': 3.5578667476976364, 'gradnorm/decoder.layer1': 1.9341118628963556, 'gradnorm/decoder.layer2': 1.210340537316026, 'gradnorm/max_param': 948.6757202148438, 'epoch': 9.71}

 97%|█████████▋| 18540/19090 [9:03:00<15:05,  1.65s/it]
 97%|█████████▋| 18541/19090 [9:03:02<17:31,  1.92s/it]
 97%|█████████▋| 18542/19090 [9:03:04<16:00,  1.75s/it]
 97%|█████████▋| 18543/19090 [9:03:05<14:42,  1.61s/it]
 97%|█████████▋| 18544/19090 [9:03:08<18:18,  2.01s/it]
 97%|█████████▋| 18545/19090 [9:03:09<16:19,  1.80s/it]
 97%|█████████▋| 18546/19090 [9:03:11<14:55,  1.65s/it]
 97%|█████████▋| 18547/19090 [9:03:13<17:57,  1.98s/it]
 97%|█████████▋| 18548/19090 [9:03:15<16:04,  1.78s/it]
 97%|█████████▋| 18549/19090 [9:03:16<14:45,  1.64s/it][gradnorm] step 18549: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 981.3

 97%|█████████▋| 18550/19090 [9:03:19<17:47,  1.98s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1403.0296630859375, 'learning_rate': 1.2222307989082915e-07, 'recon_loss': 0.0013987388170789926, 'recon_static': 0.00042553957682684996, 'recon_motion': 0.0009731992409797386, 'recon_norm': 1.2264240622520446, 'recon_norm_std': 0.042503305710852146, 'target_norm': 1.6179109424352647, 'recon_cos': 0.671610289812088, 'gradnorm/compressor.query': 0.48587933097127994, 'gradnorm/compressor.layer0': 13.710911903466135, 'gradnorm/compressor.layer1': 1390.4305088712633, 'gradnorm/compressor.layer2': 180.23147009470904, 'gradnorm/compressor.layer3': 48.36788848728824, 'gradnorm/compressor.layer4': 10.938013729564638, 'gradnorm/compressor.layer5': 5.483756092938564, 'gradnorm/compressor.layer6': 2.5341827921765665, 'gradnorm/compressor.layer7': 3.1434524486325484, 'gradnorm/decoder.layer0': 3.539956760896742, 'gradnorm/decoder.layer1': 1.943462242586036, 'gradnorm/decoder.layer2': 1.2188565623328822, 'gradnorm/max_param': 981.2653198242188, 'epoch': 9.72}

 97%|█████████▋| 18550/19090 [9:03:19<17:47,  1.98s/it]
 97%|█████████▋| 18551/19090 [9:03:20<15:56,  1.78s/it]
 97%|█████████▋| 18552/19090 [9:03:21<14:38,  1.63s/it]
 97%|█████████▋| 18553/19090 [9:03:24<17:50,  1.99s/it]
 97%|█████████▋| 18554/19090 [9:03:26<15:56,  1.78s/it]
 97%|█████████▋| 18555/19090 [9:03:27<14:36,  1.64s/it]
 97%|█████████▋| 18556/19090 [9:03:30<18:40,  2.10s/it]
 97%|█████████▋| 18557/19090 [9:03:31<16:30,  1.86s/it]
 97%|█████████▋| 18558/19090 [9:03:33<14:58,  1.69s/it]
 97%|█████████▋| 18559/19090 [9:03:35<17:13,  1.95s/it][gradnorm] step 18559: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1014.9

 97%|█████████▋| 18560/19090 [9:03:36<15:29,  1.75s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1450.275390625, 'learning_rate': 1.1774994396255844e-07, 'recon_loss': 0.00139277633279562, 'recon_static': 0.00042227404774166646, 'recon_motion': 0.0009705022646812722, 'recon_norm': 1.2212367236614228, 'recon_norm_std': 0.04181272620335221, 'target_norm': 1.617498755455017, 'recon_cos': 0.6723494350910186, 'gradnorm/compressor.query': 0.40917384250444755, 'gradnorm/compressor.layer0': 10.560402927225436, 'gradnorm/compressor.layer1': 1437.888078050583, 'gradnorm/compressor.layer2': 182.26657803544236, 'gradnorm/compressor.layer3': 47.78035837542844, 'gradnorm/compressor.layer4': 9.799196591823987, 'gradnorm/compressor.layer5': 5.541721066468434, 'gradnorm/compressor.layer6': 2.651861357607736, 'gradnorm/compressor.layer7': 3.1659476986563506, 'gradnorm/decoder.layer0': 3.4606386490522185, 'gradnorm/decoder.layer1': 1.9147872486906772, 'gradnorm/decoder.layer2': 1.1920820977050037, 'gradnorm/max_param': 1014.9036865234375, 'epoch': 9.72}

 97%|█████████▋| 18560/19090 [9:03:36<15:29,  1.75s/it]
 97%|█████████▋| 18561/19090 [9:03:38<14:15,  1.62s/it]
 97%|█████████▋| 18562/19090 [9:03:40<16:18,  1.85s/it]
 97%|█████████▋| 18563/19090 [9:03:41<14:47,  1.68s/it]
 97%|█████████▋| 18564/19090 [9:03:43<13:44,  1.57s/it]
 97%|█████████▋| 18565/19090 [9:03:46<16:56,  1.94s/it]
 97%|█████████▋| 18566/19090 [9:03:47<15:14,  1.75s/it]
 97%|█████████▋| 18567/19090 [9:03:48<14:01,  1.61s/it]
 97%|█████████▋| 18568/19090 [9:03:51<16:37,  1.91s/it]
 97%|█████████▋| 18569/19090 [9:03:52<15:17,  1.76s/it][gradnorm] step 18569: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1041.3

 97%|█████████▋| 18570/19090 [9:03:53<14:02,  1.62s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1485.930908203125, 'learning_rate': 1.133600021095943e-07, 'recon_loss': 0.0014015443506650628, 'recon_static': 0.0004146998056967277, 'recon_motion': 0.000986844526778441, 'recon_norm': 1.2189965158700944, 'recon_norm_std': 0.04205905133858323, 'target_norm': 1.619621494412422, 'recon_cos': 0.670836764574051, 'gradnorm/compressor.query': 0.47672542538590645, 'gradnorm/compressor.layer0': 11.368787473309599, 'gradnorm/compressor.layer1': 1475.3768162744052, 'gradnorm/compressor.layer2': 169.580208944396, 'gradnorm/compressor.layer3': 46.72130271036035, 'gradnorm/compressor.layer4': 10.795819596684643, 'gradnorm/compressor.layer5': 5.647204674796026, 'gradnorm/compressor.layer6': 2.793569049479573, 'gradnorm/compressor.layer7': 3.180161199502594, 'gradnorm/decoder.layer0': 3.562045821300511, 'gradnorm/decoder.layer1': 1.9293770482454342, 'gradnorm/decoder.layer2': 1.187926065404866, 'gradnorm/max_param': 1041.3250732421875, 'epoch': 9.73}

 97%|█████████▋| 18570/19090 [9:03:53<14:02,  1.62s/it]
 97%|█████████▋| 18571/19090 [9:03:56<17:10,  1.99s/it]
 97%|█████████▋| 18572/19090 [9:03:58<15:21,  1.78s/it]
 97%|█████████▋| 18573/19090 [9:03:59<14:04,  1.63s/it]
 97%|█████████▋| 18574/19090 [9:04:02<17:11,  2.00s/it]
 97%|█████████▋| 18575/19090 [9:04:03<15:20,  1.79s/it]
 97%|█████████▋| 18576/19090 [9:04:04<14:02,  1.64s/it]
 97%|█████████▋| 18577/19090 [9:04:07<16:53,  1.98s/it]
 97%|█████████▋| 18578/19090 [9:04:08<15:06,  1.77s/it]
 97%|█████████▋| 18579/19090 [9:04:10<13:52,  1.63s/it][gradnorm] step 18579: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 815.1

 97%|█████████▋| 18580/19090 [9:04:12<16:30,  1.94s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1165.959228515625, 'learning_rate': 1.09053269009754e-07, 'recon_loss': 0.0014013278210768475, 'recon_static': 0.00041896878756233493, 'recon_motion': 0.0009823590517044067, 'recon_norm': 1.219794237613678, 'recon_norm_std': 0.0432672206312418, 'target_norm': 1.6216519236564637, 'recon_cos': 0.6718819856643676, 'gradnorm/compressor.query': 0.3526547207881576, 'gradnorm/compressor.layer0': 10.671471785025426, 'gradnorm/compressor.layer1': 1154.8843556824206, 'gradnorm/compressor.layer2': 153.6228132472518, 'gradnorm/compressor.layer3': 42.85297043607727, 'gradnorm/compressor.layer4': 9.487482356883094, 'gradnorm/compressor.layer5': 5.252587906738694, 'gradnorm/compressor.layer6': 2.693787605040861, 'gradnorm/compressor.layer7': 3.28028870079621, 'gradnorm/decoder.layer0': 3.446147929863777, 'gradnorm/decoder.layer1': 1.9276722875798216, 'gradnorm/decoder.layer2': 1.1677352518772686, 'gradnorm/max_param': 815.067626953125, 'epoch': 9.73}

 97%|█████████▋| 18580/19090 [9:04:12<16:30,  1.94s/it]
 97%|█████████▋| 18581/19090 [9:04:14<14:48,  1.75s/it]
 97%|█████████▋| 18582/19090 [9:04:15<13:37,  1.61s/it]
 97%|█████████▋| 18583/19090 [9:04:18<16:41,  1.98s/it]
 97%|█████████▋| 18584/19090 [9:04:19<14:55,  1.77s/it]
 97%|█████████▋| 18585/19090 [9:04:20<13:40,  1.62s/it]
 97%|█████████▋| 18586/19090 [9:04:23<17:08,  2.04s/it]
 97%|█████████▋| 18587/19090 [9:04:25<15:13,  1.82s/it]
 97%|█████████▋| 18588/19090 [9:04:26<13:53,  1.66s/it]
 97%|█████████▋| 18589/19090 [9:04:29<16:42,  2.00s/it][gradnorm] step 18589: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 968.2

 97%|█████████▋| 18590/19090 [9:04:30<14:54,  1.79s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1384.07763671875, 'learning_rate': 1.048297590626468e-07, 'recon_loss': 0.001397003815509379, 'recon_static': 0.0004159419753705151, 'recon_motion': 0.0009810618401388637, 'recon_norm': 1.2191629528999328, 'recon_norm_std': 0.04223603643476963, 'target_norm': 1.6205700427293777, 'recon_cos': 0.6725130334496499, 'gradnorm/compressor.query': 0.369793229256494, 'gradnorm/compressor.layer0': 7.3585448232162864, 'gradnorm/compressor.layer1': 1371.8451534338706, 'gradnorm/compressor.layer2': 177.2399944284585, 'gradnorm/compressor.layer3': 45.55288071468423, 'gradnorm/compressor.layer4': 10.235290393222547, 'gradnorm/compressor.layer5': 5.370489135003844, 'gradnorm/compressor.layer6': 2.7305555425337564, 'gradnorm/compressor.layer7': 3.139686710011673, 'gradnorm/decoder.layer0': 3.452037795477658, 'gradnorm/decoder.layer1': 1.9191385031171482, 'gradnorm/decoder.layer2': 1.1677175908851776, 'gradnorm/max_param': 968.232421875, 'epoch': 9.74}

 97%|█████████▋| 18590/19090 [9:04:30<14:54,  1.79s/it]
 97%|█████████▋| 18591/19090 [9:04:31<13:37,  1.64s/it]
 97%|█████████▋| 18592/19090 [9:04:34<17:16,  2.08s/it]
 97%|█████████▋| 18593/19090 [9:04:36<15:16,  1.84s/it]
 97%|█████████▋| 18594/19090 [9:04:37<13:51,  1.68s/it]
 97%|█████████▋| 18595/19090 [9:04:40<16:02,  1.94s/it]
 97%|█████████▋| 18596/19090 [9:04:41<14:24,  1.75s/it]
 97%|█████████▋| 18597/19090 [9:04:42<13:14,  1.61s/it]
 97%|█████████▋| 18598/19090 [9:04:45<16:27,  2.01s/it]
 97%|█████████▋| 18599/19090 [9:04:46<14:39,  1.79s/it][gradnorm] step 18599: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1077.7

 97%|█████████▋| 18600/19090 [9:04:48<13:23,  1.64s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1539.6103515625, 'learning_rate': 1.0068948638962405e-07, 'recon_loss': 0.001390663962229155, 'recon_static': 0.0004117283766390756, 'recon_motion': 0.0009789356059627607, 'recon_norm': 1.2176672518253326, 'recon_norm_std': 0.04177925381809473, 'target_norm': 1.617496246099472, 'recon_cos': 0.6726455315947533, 'gradnorm/compressor.query': 0.3783245734357386, 'gradnorm/compressor.layer0': 11.095959424572433, 'gradnorm/compressor.layer1': 1527.118774031673, 'gradnorm/compressor.layer2': 189.9231649740231, 'gradnorm/compressor.layer3': 44.07594680677958, 'gradnorm/compressor.layer4': 10.289276792804502, 'gradnorm/compressor.layer5': 5.509140049451698, 'gradnorm/compressor.layer6': 2.855503817863497, 'gradnorm/compressor.layer7': 3.221826212945593, 'gradnorm/decoder.layer0': 3.4454340145627596, 'gradnorm/decoder.layer1': 1.9303871770503223, 'gradnorm/decoder.layer2': 1.1624538391956918, 'gradnorm/max_param': 1077.7259521484375, 'epoch': 9.74}

 97%|█████████▋| 18600/19090 [9:04:48<13:23,  1.64s/it]
 97%|█████████▋| 18601/19090 [9:04:50<15:59,  1.96s/it]
 97%|█████████▋| 18602/19090 [9:04:52<14:18,  1.76s/it]
 97%|█████████▋| 18603/19090 [9:04:53<13:08,  1.62s/it]
 97%|█████████▋| 18604/19090 [9:04:55<13:38,  1.68s/it]
 97%|█████████▋| 18605/19090 [9:04:56<12:39,  1.57s/it]
 97%|█████████▋| 18606/19090 [9:04:57<11:58,  1.48s/it]
 97%|█████████▋| 18607/19090 [9:04:59<13:32,  1.68s/it]
 97%|█████████▋| 18608/19090 [9:05:01<12:34,  1.57s/it]
 97%|█████████▋| 18609/19090 [9:05:02<11:53,  1.48s/it][gradnorm] step 18609: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1063.4

 97%|█████████▋| 18610/19090 [9:05:05<14:12,  1.78s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1519.314453125, 'learning_rate': 9.66324648337319e-08, 'recon_loss': 0.0013958966068457813, 'recon_static': 0.0004126460444240365, 'recon_motion': 0.00098325053404551, 'recon_norm': 1.2163643747568131, 'recon_norm_std': 0.04182560201734305, 'target_norm': 1.619208201766014, 'recon_cos': 0.6721614599227905, 'gradnorm/compressor.query': 0.3071886256246812, 'gradnorm/compressor.layer0': 7.174473642853666, 'gradnorm/compressor.layer1': 1506.5745086121694, 'gradnorm/compressor.layer2': 189.91984328921504, 'gradnorm/compressor.layer3': 47.60589552841643, 'gradnorm/compressor.layer4': 9.92187269315918, 'gradnorm/compressor.layer5': 5.287472822692142, 'gradnorm/compressor.layer6': 2.791518724611382, 'gradnorm/compressor.layer7': 3.2358775311507766, 'gradnorm/decoder.layer0': 3.468799350146905, 'gradnorm/decoder.layer1': 1.9637999900122454, 'gradnorm/decoder.layer2': 1.1564950425585596, 'gradnorm/max_param': 1063.3758544921875, 'epoch': 9.75}

 97%|█████████▋| 18610/19090 [9:05:05<14:12,  1.78s/it]
 97%|█████████▋| 18611/19090 [9:05:06<13:01,  1.63s/it]
 97%|█████████▋| 18612/19090 [9:05:07<12:13,  1.53s/it]
 98%|█████████▊| 18613/19090 [9:05:10<14:16,  1.80s/it]
 98%|█████████▊| 18614/19090 [9:05:12<15:11,  1.91s/it]
 98%|█████████▊| 18615/19090 [9:05:13<13:40,  1.73s/it]
 98%|█████████▊| 18616/19090 [9:05:14<12:53,  1.63s/it]
 98%|█████████▊| 18617/19090 [9:05:16<12:03,  1.53s/it]
 98%|█████████▊| 18618/19090 [9:05:17<11:27,  1.46s/it]
 98%|█████████▊| 18619/19090 [9:05:20<14:14,  1.81s/it][gradnorm] step 18619: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 970.5

 98%|█████████▊| 18620/19090 [9:05:21<12:58,  1.66s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1386.4453125, 'learning_rate': 9.265870795966969e-08, 'recon_loss': 0.001398692064685747, 'recon_static': 0.0004166739272477571, 'recon_motion': 0.0009820181265240535, 'recon_norm': 1.2189565658569337, 'recon_norm_std': 0.04209631560370326, 'target_norm': 1.6202682673931121, 'recon_cos': 0.671979084610939, 'gradnorm/compressor.query': 0.31541550312096384, 'gradnorm/compressor.layer0': 7.5794653896737145, 'gradnorm/compressor.layer1': 1375.146264947842, 'gradnorm/compressor.layer2': 169.93856886386916, 'gradnorm/compressor.layer3': 45.7247251307348, 'gradnorm/compressor.layer4': 10.473296587333426, 'gradnorm/compressor.layer5': 5.570608165967618, 'gradnorm/compressor.layer6': 2.759701349437279, 'gradnorm/compressor.layer7': 3.2098660131519035, 'gradnorm/decoder.layer0': 3.486369977427358, 'gradnorm/decoder.layer1': 1.9459467700292086, 'gradnorm/decoder.layer2': 1.1509117409865848, 'gradnorm/max_param': 970.5431518554688, 'epoch': 9.75}

 98%|█████████▊| 18620/19090 [9:05:21<12:58,  1.66s/it]
 98%|█████████▊| 18621/19090 [9:05:22<12:05,  1.55s/it]
 98%|█████████▊| 18622/19090 [9:05:24<13:31,  1.73s/it]
 98%|█████████▊| 18623/19090 [9:05:26<12:26,  1.60s/it]
 98%|█████████▊| 18624/19090 [9:05:27<11:41,  1.51s/it]
 98%|█████████▊| 18625/19090 [9:05:29<13:28,  1.74s/it]
 98%|█████████▊| 18626/19090 [9:05:31<12:24,  1.60s/it]
 98%|█████████▊| 18627/19090 [9:05:32<11:38,  1.51s/it]
 98%|█████████▊| 18628/19090 [9:05:34<14:02,  1.82s/it]
 98%|█████████▊| 18629/19090 [9:05:36<12:49,  1.67s/it][gradnorm] step 18629: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 946.9

 98%|█████████▊| 18630/19090 [9:05:37<11:56,  1.56s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1351.7286376953125, 'learning_rate': 8.87682290537345e-08, 'recon_loss': 0.0014154543954646216, 'recon_static': 0.0004203173528367188, 'recon_motion': 0.0009951370229828171, 'recon_norm': 1.219898873567581, 'recon_norm_std': 0.043515426199883224, 'target_norm': 1.6258059352636338, 'recon_cos': 0.6702783420681954, 'gradnorm/compressor.query': 0.38586043464868885, 'gradnorm/compressor.layer0': 11.51492601444595, 'gradnorm/compressor.layer1': 1341.9368465020998, 'gradnorm/compressor.layer2': 156.03721135717916, 'gradnorm/compressor.layer3': 41.604911103494594, 'gradnorm/compressor.layer4': 9.851087372687159, 'gradnorm/compressor.layer5': 5.577673629322029, 'gradnorm/compressor.layer6': 2.880302349485325, 'gradnorm/compressor.layer7': 3.2747315100184196, 'gradnorm/decoder.layer0': 3.4802925587593445, 'gradnorm/decoder.layer1': 1.9879662164853211, 'gradnorm/decoder.layer2': 1.1521883067662866, 'gradnorm/max_param': 946.9481811523438, 'epoch': 9.76}

 98%|█████████▊| 18630/19090 [9:05:37<11:56,  1.56s/it]
 98%|█████████▊| 18631/19090 [9:05:40<14:37,  1.91s/it]
 98%|█████████▊| 18632/19090 [9:05:41<13:24,  1.76s/it]
 98%|█████████▊| 18633/19090 [9:05:42<12:22,  1.63s/it]
 98%|█████████▊| 18634/19090 [9:05:45<14:22,  1.89s/it]
 98%|█████████▊| 18635/19090 [9:05:47<14:08,  1.87s/it]
 98%|█████████▊| 18636/19090 [9:05:48<12:48,  1.69s/it]
 98%|█████████▊| 18637/19090 [9:05:50<14:27,  1.91s/it]
 98%|█████████▊| 18638/19090 [9:05:52<13:49,  1.83s/it]
 98%|█████████▊| 18639/19090 [9:05:53<12:33,  1.67s/it][gradnorm] step 18639: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 990.2

 98%|█████████▊| 18640/19090 [9:05:56<14:20,  1.91s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1412.86328125, 'learning_rate': 8.4961041123785e-08, 'recon_loss': 0.0014059504901524633, 'recon_static': 0.00041401882481295615, 'recon_motion': 0.0009919315969455055, 'recon_norm': 1.2184817284345626, 'recon_norm_std': 0.04208917254582047, 'target_norm': 1.6222213596105575, 'recon_cos': 0.6710690319538116, 'gradnorm/compressor.query': 0.45939869138388445, 'gradnorm/compressor.layer0': 12.166719689101251, 'gradnorm/compressor.layer1': 1402.9977280808405, 'gradnorm/compressor.layer2': 160.08084505375714, 'gradnorm/compressor.layer3': 42.835003494969655, 'gradnorm/compressor.layer4': 10.151788758757808, 'gradnorm/compressor.layer5': 5.635177898695939, 'gradnorm/compressor.layer6': 2.9620557299000456, 'gradnorm/compressor.layer7': 3.189262501240506, 'gradnorm/decoder.layer0': 3.4822733061672824, 'gradnorm/decoder.layer1': 1.9861819833353396, 'gradnorm/decoder.layer2': 1.1592951189268585, 'gradnorm/max_param': 990.1856689453125, 'epoch': 9.76}

 98%|█████████▊| 18640/19090 [9:05:56<14:20,  1.91s/it]
 98%|█████████▊| 18641/19090 [9:05:58<13:54,  1.86s/it]
 98%|█████████▊| 18642/19090 [9:05:59<12:37,  1.69s/it]
 98%|█████████▊| 18643/19090 [9:06:01<13:29,  1.81s/it]
 98%|█████████▊| 18644/19090 [9:06:03<13:31,  1.82s/it]
 98%|█████████▊| 18645/19090 [9:06:04<12:19,  1.66s/it]
 98%|█████████▊| 18646/19090 [9:06:06<13:46,  1.86s/it]
 98%|█████████▊| 18647/19090 [9:06:08<13:31,  1.83s/it]
 98%|█████████▊| 18648/19090 [9:06:10<12:18,  1.67s/it]
 98%|█████████▊| 18649/19090 [9:06:12<14:18,  1.95s/it][gradnorm] step 18649: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1100.5

 98%|█████████▊| 18650/19090 [9:06:14<13:36,  1.86s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1571.404296875, 'learning_rate': 8.123715689919709e-08, 'recon_loss': 0.0014023236057255418, 'recon_static': 0.00041402630740776657, 'recon_motion': 0.000988297261937987, 'recon_norm': 1.2169702082872391, 'recon_norm_std': 0.04215322006493807, 'target_norm': 1.621164807677269, 'recon_cos': 0.6713627457618714, 'gradnorm/compressor.query': 0.41225156093924187, 'gradnorm/compressor.layer0': 10.891378061974857, 'gradnorm/compressor.layer1': 1559.3535519567074, 'gradnorm/compressor.layer2': 186.38827710923775, 'gradnorm/compressor.layer3': 51.815487323410366, 'gradnorm/compressor.layer4': 10.6874282566808, 'gradnorm/compressor.layer5': 5.682324412595647, 'gradnorm/compressor.layer6': 3.003687182153365, 'gradnorm/compressor.layer7': 3.2427242707962596, 'gradnorm/decoder.layer0': 3.5077423564465198, 'gradnorm/decoder.layer1': 2.0103870317188846, 'gradnorm/decoder.layer2': 1.1708594435815707, 'gradnorm/max_param': 1100.5311279296875, 'epoch': 9.77}

 98%|█████████▊| 18650/19090 [9:06:14<13:36,  1.86s/it]
 98%|█████████▊| 18651/19090 [9:06:15<12:20,  1.69s/it]
 98%|█████████▊| 18652/19090 [9:06:17<13:48,  1.89s/it]
 98%|█████████▊| 18653/19090 [9:06:19<12:57,  1.78s/it]
 98%|█████████▊| 18654/19090 [9:06:20<11:51,  1.63s/it]
 98%|█████████▊| 18655/19090 [9:06:23<13:58,  1.93s/it]
 98%|█████████▊| 18656/19090 [9:06:24<12:53,  1.78s/it]
 98%|█████████▊| 18657/19090 [9:06:26<11:47,  1.63s/it]
 98%|█████████▊| 18658/19090 [9:06:28<13:53,  1.93s/it]
 98%|█████████▊| 18659/19090 [9:06:30<13:12,  1.84s/it][gradnorm] step 18659: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1042.1

 98%|█████████▊| 18660/19090 [9:06:31<12:01,  1.68s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1487.53955078125, 'learning_rate': 7.759658883081666e-08, 'recon_loss': 0.0013946237828349694, 'recon_static': 0.00041834552539512517, 'recon_motion': 0.0009762782632606104, 'recon_norm': 1.2183372676372528, 'recon_norm_std': 0.04211113750934601, 'target_norm': 1.6203303694725038, 'recon_cos': 0.6733115509152412, 'gradnorm/compressor.query': 0.457514890079306, 'gradnorm/compressor.layer0': 10.52029129495301, 'gradnorm/compressor.layer1': 1476.6544111605804, 'gradnorm/compressor.layer2': 173.73537483411374, 'gradnorm/compressor.layer3': 42.56367730796486, 'gradnorm/compressor.layer4': 9.719549115559097, 'gradnorm/compressor.layer5': 5.452567260551213, 'gradnorm/compressor.layer6': 2.7988666284206047, 'gradnorm/compressor.layer7': 3.1397752513755486, 'gradnorm/decoder.layer0': 3.4255226237716894, 'gradnorm/decoder.layer1': 1.9883156044620753, 'gradnorm/decoder.layer2': 1.1745844735289677, 'gradnorm/max_param': 1042.1092529296875, 'epoch': 9.77}

 98%|█████████▊| 18660/19090 [9:06:31<12:01,  1.68s/it]
 98%|█████████▊| 18661/19090 [9:06:34<13:34,  1.90s/it]
 98%|█████████▊| 18662/19090 [9:06:35<12:37,  1.77s/it]
 98%|█████████▊| 18663/19090 [9:06:36<11:34,  1.63s/it]
 98%|█████████▊| 18664/19090 [9:06:39<14:06,  1.99s/it]
 98%|█████████▊| 18665/19090 [9:06:41<12:46,  1.80s/it]
 98%|█████████▊| 18666/19090 [9:06:42<11:40,  1.65s/it]
 98%|█████████▊| 18667/19090 [9:06:44<13:32,  1.92s/it]
 98%|█████████▊| 18668/19090 [9:06:46<12:43,  1.81s/it]
 98%|█████████▊| 18669/19090 [9:06:47<11:34,  1.65s/it][gradnorm] step 18669: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1120.1

 98%|█████████▊| 18670/19090 [9:06:50<13:09,  1.88s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1602.0888671875, 'learning_rate': 7.403934909092359e-08, 'recon_loss': 0.0013983706943690778, 'recon_static': 0.00041700681613292544, 'recon_motion': 0.0009813638622290454, 'recon_norm': 1.2181854009628297, 'recon_norm_std': 0.041793265659362075, 'target_norm': 1.6205040395259858, 'recon_cos': 0.6721320658922195, 'gradnorm/compressor.query': 0.38593121496480653, 'gradnorm/compressor.layer0': 7.2569788524860455, 'gradnorm/compressor.layer1': 1587.0018903580424, 'gradnorm/compressor.layer2': 214.511581855153, 'gradnorm/compressor.layer3': 43.27152276748748, 'gradnorm/compressor.layer4': 10.24443349646891, 'gradnorm/compressor.layer5': 5.712546063327172, 'gradnorm/compressor.layer6': 2.876714609777264, 'gradnorm/compressor.layer7': 3.1838473166370176, 'gradnorm/decoder.layer0': 3.4949469193123712, 'gradnorm/decoder.layer1': 1.9910696446314853, 'gradnorm/decoder.layer2': 1.180200326369964, 'gradnorm/max_param': 1120.1337890625, 'epoch': 9.78}

 98%|█████████▊| 18670/19090 [9:06:50<13:09,  1.88s/it]
 98%|█████████▊| 18671/19090 [9:06:51<12:35,  1.80s/it]
 98%|█████████▊| 18672/19090 [9:06:53<11:29,  1.65s/it]
 98%|█████████▊| 18673/19090 [9:06:55<12:28,  1.80s/it]
 98%|█████████▊| 18674/19090 [9:06:57<12:43,  1.83s/it]
 98%|█████████▊| 18675/19090 [9:06:58<11:33,  1.67s/it]
 98%|█████████▊| 18676/19090 [9:07:00<13:10,  1.91s/it]
 98%|█████████▊| 18677/19090 [9:07:02<12:35,  1.83s/it]
 98%|█████████▊| 18678/19090 [9:07:03<11:26,  1.67s/it]
 98%|█████████▊| 18679/19090 [9:07:06<13:15,  1.93s/it][gradnorm] step 18679: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 907.3

 98%|█████████▊| 18680/19090 [9:07:07<11:59,  1.76s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1297.1534423828125, 'learning_rate': 7.056544957318722e-08, 'recon_loss': 0.0013899042562115938, 'recon_static': 0.00041953181353164836, 'recon_motion': 0.0009703724237624556, 'recon_norm': 1.2181081295013427, 'recon_norm_std': 0.04270865703001618, 'target_norm': 1.6182495892047881, 'recon_cos': 0.6733493372797966, 'gradnorm/compressor.query': 0.37392101429261615, 'gradnorm/compressor.layer0': 9.72336716464799, 'gradnorm/compressor.layer1': 1285.8330957009935, 'gradnorm/compressor.layer2': 166.34718728836384, 'gradnorm/compressor.layer3': 36.17731128155463, 'gradnorm/compressor.layer4': 10.058719659406133, 'gradnorm/compressor.layer5': 5.463033598582674, 'gradnorm/compressor.layer6': 2.727859931644428, 'gradnorm/compressor.layer7': 3.188850864563249, 'gradnorm/decoder.layer0': 3.406652304313628, 'gradnorm/decoder.layer1': 1.9838878604360792, 'gradnorm/decoder.layer2': 1.1433276204299183, 'gradnorm/max_param': 907.3402709960938, 'epoch': 9.79}

 98%|█████████▊| 18680/19090 [9:07:07<11:59,  1.76s/it]
 98%|█████████▊| 18681/19090 [9:07:08<11:00,  1.62s/it]
 98%|█████████▊| 18682/19090 [9:07:11<13:17,  1.95s/it]
 98%|█████████▊| 18683/19090 [9:07:12<11:53,  1.75s/it]
 98%|█████████▊| 18684/19090 [9:07:14<10:54,  1.61s/it]
 98%|█████████▊| 18685/19090 [9:07:17<13:54,  2.06s/it]
 98%|█████████▊| 18686/19090 [9:07:18<12:18,  1.83s/it]
 98%|█████████▊| 18687/19090 [9:07:19<11:11,  1.67s/it]
 98%|█████████▊| 18688/19090 [9:07:22<12:24,  1.85s/it]
 98%|█████████▊| 18689/19090 [9:07:23<11:15,  1.68s/it][gradnorm] step 18689: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 931.6

 98%|█████████▊| 18690/19090 [9:07:24<10:26,  1.57s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1330.8343505859375, 'learning_rate': 6.717490189263043e-08, 'recon_loss': 0.0014021552953636274, 'recon_static': 0.0004191865707980469, 'recon_motion': 0.0009829687696765177, 'recon_norm': 1.21831114590168, 'recon_norm_std': 0.04292094632983208, 'target_norm': 1.6221111923456193, 'recon_cos': 0.6720791414380074, 'gradnorm/compressor.query': 0.3505181377169312, 'gradnorm/compressor.layer0': 9.013925164033312, 'gradnorm/compressor.layer1': 1320.238898836116, 'gradnorm/compressor.layer2': 160.97549308823378, 'gradnorm/compressor.layer3': 43.87960017506355, 'gradnorm/compressor.layer4': 10.159412067853784, 'gradnorm/compressor.layer5': 5.540657276590984, 'gradnorm/compressor.layer6': 2.8333312950874925, 'gradnorm/compressor.layer7': 3.279705883190872, 'gradnorm/decoder.layer0': 3.388215786805283, 'gradnorm/decoder.layer1': 1.9755688005157626, 'gradnorm/decoder.layer2': 1.1359872988105002, 'gradnorm/max_param': 931.6241455078125, 'epoch': 9.79}

 98%|█████████▊| 18690/19090 [9:07:24<10:26,  1.57s/it]
 98%|█████████▊| 18691/19090 [9:07:27<12:06,  1.82s/it]
 98%|█████████▊| 18692/19090 [9:07:28<11:09,  1.68s/it]
 98%|█████████▊| 18693/19090 [9:07:29<10:20,  1.56s/it]
 98%|█████████▊| 18694/19090 [9:07:32<11:37,  1.76s/it]
 98%|█████████▊| 18695/19090 [9:07:33<10:55,  1.66s/it]
 98%|█████████▊| 18696/19090 [9:07:34<10:10,  1.55s/it]
 98%|█████████▊| 18697/19090 [9:07:36<11:14,  1.72s/it]
 98%|█████████▊| 18698/19090 [9:07:38<11:27,  1.75s/it]
 98%|█████████▊| 18699/19090 [9:07:40<10:31,  1.61s/it][gradnorm] step 18699: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 924.0

 98%|█████████▊| 18700/19090 [9:07:41<10:47,  1.66s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1321.840087890625, 'learning_rate': 6.386771738558506e-08, 'recon_loss': 0.001391009142389521, 'recon_static': 0.00041755504717002625, 'recon_motion': 0.0009734541003126651, 'recon_norm': 1.2181152313947679, 'recon_norm_std': 0.0425351670011878, 'target_norm': 1.6184330999851226, 'recon_cos': 0.6731541469693184, 'gradnorm/compressor.query': 0.34180366509368726, 'gradnorm/compressor.layer0': 10.491871548551698, 'gradnorm/compressor.layer1': 1309.3267353873134, 'gradnorm/compressor.layer2': 176.02291998309482, 'gradnorm/compressor.layer3': 40.91833998190787, 'gradnorm/compressor.layer4': 9.489716837027997, 'gradnorm/compressor.layer5': 5.451084575001863, 'gradnorm/compressor.layer6': 2.964571612386974, 'gradnorm/compressor.layer7': 3.3009967801304563, 'gradnorm/decoder.layer0': 3.40645843708053, 'gradnorm/decoder.layer1': 1.9579516776164583, 'gradnorm/decoder.layer2': 1.1249548585100215, 'gradnorm/max_param': 924.0120849609375, 'epoch': 9.8}

 98%|█████████▊| 18700/19090 [9:07:41<10:47,  1.66s/it]
 98%|█████████▊| 18701/19090 [9:07:43<11:45,  1.81s/it]
 98%|█████████▊| 18702/19090 [9:07:45<10:42,  1.66s/it]
 98%|█████████▊| 18703/19090 [9:07:46<10:09,  1.57s/it]
 98%|█████████▊| 18704/19090 [9:07:48<11:01,  1.71s/it]
 98%|█████████▊| 18705/19090 [9:07:49<10:10,  1.59s/it]
 98%|█████████▊| 18706/19090 [9:07:51<09:47,  1.53s/it]
 98%|█████████▊| 18707/19090 [9:07:53<11:31,  1.81s/it]
 98%|█████████▊| 18708/19090 [9:07:55<10:30,  1.65s/it]
 98%|█████████▊| 18709/19090 [9:07:56<09:55,  1.56s/it][gradnorm] step 18709: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 948.7

 98%|█████████▊| 18710/19090 [9:07:58<11:40,  1.84s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1354.3477783203125, 'learning_rate': 6.064390710966151e-08, 'recon_loss': 0.0013956967653939501, 'recon_static': 0.0004155688780883793, 'recon_motion': 0.0009801278763916343, 'recon_norm': 1.2174586176872253, 'recon_norm_std': 0.042009332310408355, 'target_norm': 1.6200536876916884, 'recon_cos': 0.6724968656897545, 'gradnorm/compressor.query': 0.3846652873154729, 'gradnorm/compressor.layer0': 12.061681531898989, 'gradnorm/compressor.layer1': 1344.3447288549169, 'gradnorm/compressor.layer2': 159.5690118138544, 'gradnorm/compressor.layer3': 35.13815680249268, 'gradnorm/compressor.layer4': 9.565198330615205, 'gradnorm/compressor.layer5': 5.225095443218915, 'gradnorm/compressor.layer6': 2.8097775316244666, 'gradnorm/compressor.layer7': 3.179283622566909, 'gradnorm/decoder.layer0': 3.3126050824465896, 'gradnorm/decoder.layer1': 1.9639028188393368, 'gradnorm/decoder.layer2': 1.128787235731524, 'gradnorm/max_param': 948.7194213867188, 'epoch': 9.8}

 98%|█████████▊| 18710/19090 [9:07:58<11:40,  1.84s/it]
 98%|█████████▊| 18711/19090 [9:08:00<10:36,  1.68s/it]
 98%|█████████▊| 18712/19090 [9:08:01<09:50,  1.56s/it]
 98%|█████████▊| 18713/19090 [9:08:04<11:42,  1.86s/it]
 98%|█████████▊| 18714/19090 [9:08:05<10:39,  1.70s/it]
 98%|█████████▊| 18715/19090 [9:08:06<09:51,  1.58s/it]
 98%|█████████▊| 18716/19090 [9:08:09<11:45,  1.89s/it]
 98%|█████████▊| 18717/19090 [9:08:10<10:37,  1.71s/it]
 98%|█████████▊| 18718/19090 [9:08:11<09:48,  1.58s/it]
 98%|█████████▊| 18719/19090 [9:08:14<12:11,  1.97s/it][gradnorm] step 18719: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 911.7

 98%|█████████▊| 18720/19090 [9:08:16<10:54,  1.77s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1303.5728759765625, 'learning_rate': 5.750348184370147e-08, 'recon_loss': 0.0013954204681795091, 'recon_static': 0.0004134220049309079, 'recon_motion': 0.0009819985250942409, 'recon_norm': 1.2172000348567962, 'recon_norm_std': 0.04198259050026536, 'target_norm': 1.6188486099243165, 'recon_cos': 0.6722254574298858, 'gradnorm/compressor.query': 0.3810711589716005, 'gradnorm/compressor.layer0': 10.489368687807193, 'gradnorm/compressor.layer1': 1291.9577198964369, 'gradnorm/compressor.layer2': 168.72469138918288, 'gradnorm/compressor.layer3': 37.64516985586016, 'gradnorm/compressor.layer4': 9.573027357278528, 'gradnorm/compressor.layer5': 5.096400220919859, 'gradnorm/compressor.layer6': 2.834826711643304, 'gradnorm/compressor.layer7': 3.1923054393728543, 'gradnorm/decoder.layer0': 3.3618966507874943, 'gradnorm/decoder.layer1': 1.9703966550831935, 'gradnorm/decoder.layer2': 1.111820427004011, 'gradnorm/max_param': 911.7496337890625, 'epoch': 9.81}

 98%|█████████▊| 18720/19090 [9:08:16<10:54,  1.77s/it]
 98%|█████████▊| 18721/19090 [9:08:17<10:00,  1.63s/it]
 98%|█████████▊| 18722/19090 [9:08:19<11:24,  1.86s/it]
 98%|█████████▊| 18723/19090 [9:08:21<10:18,  1.69s/it]
 98%|█████████▊| 18724/19090 [9:08:22<09:34,  1.57s/it]
 98%|█████████▊| 18725/19090 [9:08:25<11:53,  1.95s/it]
 98%|█████████▊| 18726/19090 [9:08:26<10:39,  1.76s/it]
 98%|█████████▊| 18727/19090 [9:08:27<09:47,  1.62s/it]
 98%|█████████▊| 18728/19090 [9:08:30<11:59,  1.99s/it]
 98%|█████████▊| 18729/19090 [9:08:31<10:42,  1.78s/it][gradnorm] step 18729: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 855.1

 98%|█████████▊| 18730/19090 [9:08:33<09:47,  1.63s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1224.283203125, 'learning_rate': 5.444645208775023e-08, 'recon_loss': 0.0014043888892047106, 'recon_static': 0.0004127254207560327, 'recon_motion': 0.0009916634939145298, 'recon_norm': 1.216841670870781, 'recon_norm_std': 0.04142902297899127, 'target_norm': 1.6209350407123566, 'recon_cos': 0.670819365978241, 'gradnorm/compressor.query': 0.4242585980157207, 'gradnorm/compressor.layer0': 7.173555065164637, 'gradnorm/compressor.layer1': 1211.5342545714504, 'gradnorm/compressor.layer2': 171.39219689428688, 'gradnorm/compressor.layer3': 38.498029385951206, 'gradnorm/compressor.layer4': 9.169046358962992, 'gradnorm/compressor.layer5': 5.075793860973059, 'gradnorm/compressor.layer6': 2.844237904036992, 'gradnorm/compressor.layer7': 3.1831616067419874, 'gradnorm/decoder.layer0': 3.43527551981405, 'gradnorm/decoder.layer1': 1.954762863060264, 'gradnorm/decoder.layer2': 1.115139451137569, 'gradnorm/max_param': 855.0901489257812, 'epoch': 9.81}

 98%|█████████▊| 18730/19090 [9:08:33<09:47,  1.63s/it]
 98%|█████████▊| 18731/19090 [9:08:35<11:22,  1.90s/it]
 98%|█████████▊| 18732/19090 [9:08:37<10:14,  1.72s/it]
 98%|█████████▊| 18733/19090 [9:08:38<09:27,  1.59s/it]
 98%|█████████▊| 18734/19090 [9:08:40<11:15,  1.90s/it]
 98%|█████████▊| 18735/19090 [9:08:42<10:08,  1.71s/it]
 98%|█████████▊| 18736/19090 [9:08:43<09:55,  1.68s/it]
 98%|█████████▊| 18737/19090 [9:08:46<10:58,  1.86s/it]
 98%|█████████▊| 18738/19090 [9:08:47<09:54,  1.69s/it]
 98%|█████████▊| 18739/19090 [9:08:49<10:32,  1.80s/it][gradnorm] step 18739: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 855.8

 98%|█████████▊| 18740/19090 [9:08:51<10:41,  1.83s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1224.248291015625, 'learning_rate': 5.147282806302334e-08, 'recon_loss': 0.0013956168753793464, 'recon_static': 0.0004135331662837416, 'recon_motion': 0.0009820837323786691, 'recon_norm': 1.217140045762062, 'recon_norm_std': 0.04216649262234569, 'target_norm': 1.619279220700264, 'recon_cos': 0.6723025515675545, 'gradnorm/compressor.query': 0.29703917481246805, 'gradnorm/compressor.layer0': 6.629491471110536, 'gradnorm/compressor.layer1': 1212.5591222699204, 'gradnorm/compressor.layer2': 163.41641833441952, 'gradnorm/compressor.layer3': 39.60451153573763, 'gradnorm/compressor.layer4': 10.026714499880276, 'gradnorm/compressor.layer5': 5.562016498254519, 'gradnorm/compressor.layer6': 2.800508493844766, 'gradnorm/compressor.layer7': 3.1640741041936127, 'gradnorm/decoder.layer0': 3.4392785759673363, 'gradnorm/decoder.layer1': 1.9669488812132745, 'gradnorm/decoder.layer2': 1.1464403914116863, 'gradnorm/max_param': 855.8175048828125, 'epoch': 9.82}

 98%|█████████▊| 18740/19090 [9:08:51<10:41,  1.83s/it]
 98%|█████████▊| 18741/19090 [9:08:52<09:42,  1.67s/it]
 98%|█████████▊| 18742/19090 [9:08:54<10:05,  1.74s/it]
 98%|█████████▊| 18743/19090 [9:08:56<10:30,  1.82s/it]
 98%|█████████▊| 18744/19090 [9:08:57<09:33,  1.66s/it]
 98%|█████████▊| 18745/19090 [9:09:00<10:38,  1.85s/it]
 98%|█████████▊| 18746/19090 [9:09:01<10:22,  1.81s/it]
 98%|█████████▊| 18747/19090 [9:09:03<09:26,  1.65s/it]
 98%|█████████▊| 18748/19090 [9:09:05<10:22,  1.82s/it]
 98%|█████████▊| 18749/19090 [9:09:07<10:24,  1.83s/it][gradnorm] step 18749: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1038.8

 98%|█████████▊| 18750/19090 [9:09:08<09:27,  1.67s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1482.28857421875, 'learning_rate': 4.858261971186218e-08, 'recon_loss': 0.0014055050007300452, 'recon_static': 0.00041818738536676393, 'recon_motion': 0.0009873175833490678, 'recon_norm': 1.217657521367073, 'recon_norm_std': 0.04284648280590773, 'target_norm': 1.6222001701593398, 'recon_cos': 0.6710376650094986, 'gradnorm/compressor.query': 0.35396347341160134, 'gradnorm/compressor.layer0': 6.466672407062513, 'gradnorm/compressor.layer1': 1471.8439115612769, 'gradnorm/compressor.layer2': 170.69865626301808, 'gradnorm/compressor.layer3': 38.98855657683819, 'gradnorm/compressor.layer4': 9.673151230924084, 'gradnorm/compressor.layer5': 5.060869495939571, 'gradnorm/compressor.layer6': 2.7471502450433163, 'gradnorm/compressor.layer7': 3.256646475935955, 'gradnorm/decoder.layer0': 3.506581249671377, 'gradnorm/decoder.layer1': 1.9713997946883925, 'gradnorm/decoder.layer2': 1.1578520424457261, 'gradnorm/max_param': 1038.7799072265625, 'epoch': 9.82}

 98%|█████████▊| 18750/19090 [9:09:08<09:27,  1.67s/it]
 98%|█████████▊| 18751/19090 [9:09:10<10:38,  1.88s/it]
 98%|█████████▊| 18752/19090 [9:09:12<10:35,  1.88s/it]
 98%|█████████▊| 18753/19090 [9:09:14<09:33,  1.70s/it]
 98%|█████████▊| 18754/19090 [9:09:16<10:02,  1.79s/it]
 98%|█████████▊| 18755/19090 [9:09:18<10:19,  1.85s/it]
 98%|█████████▊| 18756/19090 [9:09:19<09:21,  1.68s/it]
 98%|█████████▊| 18757/19090 [9:09:21<10:21,  1.87s/it]
 98%|█████████▊| 18758/19090 [9:09:23<10:20,  1.87s/it]
 98%|█████████▊| 18759/19090 [9:09:24<09:20,  1.69s/it][gradnorm] step 18759: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 928.0

 98%|█████████▊| 18760/19090 [9:09:27<10:26,  1.90s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1327.237548828125, 'learning_rate': 4.577583669771457e-08, 'recon_loss': 0.0013977416907437145, 'recon_static': 0.00041986093201558107, 'recon_motion': 0.000977880772552453, 'recon_norm': 1.2174887835979462, 'recon_norm_std': 0.04353027921169996, 'target_norm': 1.6211438596248626, 'recon_cos': 0.672668094933033, 'gradnorm/compressor.query': 0.4654059976751122, 'gradnorm/compressor.layer0': 11.524322661978136, 'gradnorm/compressor.layer1': 1315.1298414985495, 'gradnorm/compressor.layer2': 173.88399734089967, 'gradnorm/compressor.layer3': 38.12423635249547, 'gradnorm/compressor.layer4': 10.445852261437237, 'gradnorm/compressor.layer5': 5.180609665566207, 'gradnorm/compressor.layer6': 2.7267336204192834, 'gradnorm/compressor.layer7': 3.1754711792721695, 'gradnorm/decoder.layer0': 3.4666781724836686, 'gradnorm/decoder.layer1': 1.9844571794929744, 'gradnorm/decoder.layer2': 1.1707946374999108, 'gradnorm/max_param': 928.0360107421875, 'epoch': 9.83}

 98%|█████████▊| 18760/19090 [9:09:27<10:26,  1.90s/it]
 98%|█████████▊| 18761/19090 [9:09:28<09:33,  1.74s/it]
 98%|█████████▊| 18762/19090 [9:09:29<08:47,  1.61s/it]
 98%|█████████▊| 18763/19090 [9:09:32<10:32,  1.93s/it]
 98%|█████████▊| 18764/19090 [9:09:33<09:27,  1.74s/it]
 98%|█████████▊| 18765/19090 [9:09:35<08:40,  1.60s/it]
 98%|█████████▊| 18766/19090 [9:09:37<10:37,  1.97s/it]
 98%|█████████▊| 18767/19090 [9:09:39<09:30,  1.77s/it]
 98%|█████████▊| 18768/19090 [9:09:40<08:41,  1.62s/it]
 98%|█████████▊| 18769/19090 [9:09:43<10:15,  1.92s/it][gradnorm] step 18769: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 939.2

 98%|█████████▊| 18770/19090 [9:09:44<09:12,  1.73s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1341.127685546875, 'learning_rate': 4.30524884050848e-08, 'recon_loss': 0.0014028858538949862, 'recon_static': 0.0004155728718615137, 'recon_motion': 0.000987312977667898, 'recon_norm': 1.2175064623355865, 'recon_norm_std': 0.042142273485660554, 'target_norm': 1.6218623667955399, 'recon_cos': 0.6715796083211899, 'gradnorm/compressor.query': 0.3500945278650434, 'gradnorm/compressor.layer0': 7.42649450192742, 'gradnorm/compressor.layer1': 1330.7104305595565, 'gradnorm/compressor.layer2': 161.25148376255333, 'gradnorm/compressor.layer3': 40.21924928060334, 'gradnorm/compressor.layer4': 9.67395963373412, 'gradnorm/compressor.layer5': 5.451583163396369, 'gradnorm/compressor.layer6': 2.7159283783504575, 'gradnorm/compressor.layer7': 3.2455305030857375, 'gradnorm/decoder.layer0': 3.4316649417091645, 'gradnorm/decoder.layer1': 2.016035289553932, 'gradnorm/decoder.layer2': 1.187524745081023, 'gradnorm/max_param': 939.178955078125, 'epoch': 9.83}

 98%|█████████▊| 18770/19090 [9:09:44<09:12,  1.73s/it]
 98%|█████████▊| 18771/19090 [9:09:45<08:28,  1.59s/it]
 98%|█████████▊| 18772/19090 [9:09:48<10:25,  1.97s/it]
 98%|█████████▊| 18773/19090 [9:09:49<09:18,  1.76s/it]
 98%|█████████▊| 18774/19090 [9:09:51<08:30,  1.62s/it]
 98%|█████████▊| 18775/19090 [9:09:53<10:15,  1.95s/it]
 98%|█████████▊| 18776/19090 [9:09:55<09:10,  1.75s/it]
 98%|█████████▊| 18777/19090 [9:09:56<08:24,  1.61s/it]
 98%|█████████▊| 18778/19090 [9:09:58<09:33,  1.84s/it]
 98%|█████████▊| 18779/19090 [9:10:00<08:39,  1.67s/it][gradnorm] step 18779: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 930.5

 98%|█████████▊| 18780/19090 [9:10:01<08:01,  1.55s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1330.1651611328125, 'learning_rate': 4.041258393952252e-08, 'recon_loss': 0.001386982755502686, 'recon_static': 0.0004156010691076517, 'recon_motion': 0.0009713816689327359, 'recon_norm': 1.2184561491012573, 'recon_norm_std': 0.041799092292785646, 'target_norm': 1.6173220306634903, 'recon_cos': 0.673675125837326, 'gradnorm/compressor.query': 0.44250031988487964, 'gradnorm/compressor.layer0': 12.903832604076335, 'gradnorm/compressor.layer1': 1318.5205250582942, 'gradnorm/compressor.layer2': 169.2121276405004, 'gradnorm/compressor.layer3': 43.29378625981952, 'gradnorm/compressor.layer4': 10.102600941161993, 'gradnorm/compressor.layer5': 5.606053636188845, 'gradnorm/compressor.layer6': 2.6944298224909953, 'gradnorm/compressor.layer7': 3.310908673076785, 'gradnorm/decoder.layer0': 3.4905015942403654, 'gradnorm/decoder.layer1': 2.0063960797613807, 'gradnorm/decoder.layer2': 1.1917323432288158, 'gradnorm/max_param': 930.4929809570312, 'epoch': 9.84}

 98%|█████████▊| 18780/19090 [9:10:01<08:01,  1.55s/it]
 98%|█████████▊| 18781/19090 [9:10:04<10:03,  1.95s/it]
 98%|█████████▊| 18782/19090 [9:10:05<08:59,  1.75s/it]
 98%|█████████▊| 18783/19090 [9:10:06<08:14,  1.61s/it]
 98%|█████████▊| 18784/19090 [9:10:09<10:05,  1.98s/it]
 98%|█████████▊| 18785/19090 [9:10:10<08:59,  1.77s/it]
 98%|█████████▊| 18786/19090 [9:10:12<08:12,  1.62s/it]
 98%|█████████▊| 18787/19090 [9:10:14<09:47,  1.94s/it]
 98%|█████████▊| 18788/19090 [9:10:16<08:46,  1.74s/it]
 98%|█████████▊| 18789/19090 [9:10:17<08:02,  1.60s/it][gradnorm] step 18789: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 951.6

 98%|█████████▊| 18790/19090 [9:10:20<09:57,  1.99s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1359.538818359375, 'learning_rate': 3.785613212757277e-08, 'recon_loss': 0.001392076868796721, 'recon_static': 0.0004202053933113348, 'recon_motion': 0.0009718714718474075, 'recon_norm': 1.2182653218507766, 'recon_norm_std': 0.04400576977059245, 'target_norm': 1.6191063493490219, 'recon_cos': 0.67314293384552, 'gradnorm/compressor.query': 0.3957884558116963, 'gradnorm/compressor.layer0': 12.790303873205879, 'gradnorm/compressor.layer1': 1348.414578310395, 'gradnorm/compressor.layer2': 166.28777178997558, 'gradnorm/compressor.layer3': 46.231130018770905, 'gradnorm/compressor.layer4': 10.108122899904329, 'gradnorm/compressor.layer5': 5.743170289514776, 'gradnorm/compressor.layer6': 2.806328253727578, 'gradnorm/compressor.layer7': 3.255323598219688, 'gradnorm/decoder.layer0': 3.4604258966447228, 'gradnorm/decoder.layer1': 2.0163846026222623, 'gradnorm/decoder.layer2': 1.1944210243639752, 'gradnorm/max_param': 951.6405639648438, 'epoch': 9.84}

 98%|█████████▊| 18790/19090 [9:10:20<09:57,  1.99s/it]
 98%|█████████▊| 18791/19090 [9:10:21<08:51,  1.78s/it]
 98%|█████████▊| 18792/19090 [9:10:22<08:05,  1.63s/it]
 98%|█████████▊| 18793/19090 [9:10:25<09:59,  2.02s/it]
 98%|█████████▊| 18794/19090 [9:10:27<08:52,  1.80s/it]
 98%|█████████▊| 18795/19090 [9:10:28<08:05,  1.64s/it]
 98%|█████████▊| 18796/19090 [9:10:30<09:23,  1.92s/it]
 98%|█████████▊| 18797/19090 [9:10:32<08:26,  1.73s/it]
 98%|█████████▊| 18798/19090 [9:10:33<07:45,  1.59s/it]
 98%|█████████▊| 18799/19090 [9:10:36<09:24,  1.94s/it][gradnorm] step 18799: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 994.2

 98%|█████████▊| 18800/19090 [9:10:37<08:26,  1.74s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1427.0494384765625, 'learning_rate': 3.5383141516764917e-08, 'recon_loss': 0.0014031159225851297, 'recon_static': 0.000417293358623283, 'recon_motion': 0.0009858225661446341, 'recon_norm': 1.21824691593647, 'recon_norm_std': 0.04248418919742107, 'target_norm': 1.6214689373970033, 'recon_cos': 0.6712928012013435, 'gradnorm/compressor.query': 0.3278805071615115, 'gradnorm/compressor.layer0': 7.762961347203388, 'gradnorm/compressor.layer1': 1408.7131982770659, 'gradnorm/compressor.layer2': 222.53656651772985, 'gradnorm/compressor.layer3': 47.36309681020789, 'gradnorm/compressor.layer4': 10.184078835660937, 'gradnorm/compressor.layer5': 5.6540600662846074, 'gradnorm/compressor.layer6': 2.7733060563867373, 'gradnorm/compressor.layer7': 3.279460746119605, 'gradnorm/decoder.layer0': 3.4323669950162037, 'gradnorm/decoder.layer1': 1.999987363775389, 'gradnorm/decoder.layer2': 1.1955078563743151, 'gradnorm/max_param': 994.2169799804688, 'epoch': 9.85}

 98%|█████████▊| 18800/19090 [9:10:37<08:26,  1.74s/it]
 98%|█████████▊| 18801/19090 [9:10:38<07:44,  1.61s/it]
 98%|█████████▊| 18802/19090 [9:10:41<09:05,  1.89s/it]
 98%|█████████▊| 18803/19090 [9:10:42<08:11,  1.71s/it]
 99%|█████████▊| 18804/19090 [9:10:43<07:32,  1.58s/it]
 99%|█████████▊| 18805/19090 [9:10:46<08:41,  1.83s/it]
 99%|█████████▊| 18806/19090 [9:10:47<07:52,  1.67s/it]
 99%|█████████▊| 18807/19090 [9:10:48<07:18,  1.55s/it]
 99%|█████████▊| 18808/19090 [9:10:51<08:22,  1.78s/it]
 99%|█████████▊| 18809/19090 [9:10:52<07:38,  1.63s/it][gradnorm] step 18809: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 820.1

 99%|█████████▊| 18810/19090 [9:10:53<07:07,  1.53s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1172.94677734375, 'learning_rate': 3.299362037556542e-08, 'recon_loss': 0.001393721203203313, 'recon_static': 0.00041554154959158043, 'recon_motion': 0.0009781796732568182, 'recon_norm': 1.217826932668686, 'recon_norm_std': 0.042839807271957395, 'target_norm': 1.6189927220344544, 'recon_cos': 0.6726021721959115, 'gradnorm/compressor.query': 0.5581197261240748, 'gradnorm/compressor.layer0': 15.104950029075216, 'gradnorm/compressor.layer1': 1162.262179974897, 'gradnorm/compressor.layer2': 151.65142365594528, 'gradnorm/compressor.layer3': 39.59104451342579, 'gradnorm/compressor.layer4': 9.552410312881484, 'gradnorm/compressor.layer5': 5.510140001626493, 'gradnorm/compressor.layer6': 2.679468440674895, 'gradnorm/compressor.layer7': 3.263157546621411, 'gradnorm/decoder.layer0': 3.4220448630338995, 'gradnorm/decoder.layer1': 1.9847910887855267, 'gradnorm/decoder.layer2': 1.1907854307527008, 'gradnorm/max_param': 820.1395874023438, 'epoch': 9.85}

 99%|█████████▊| 18810/19090 [9:10:53<07:07,  1.53s/it]
 99%|█████████▊| 18811/19090 [9:10:56<08:20,  1.79s/it]
 99%|█████████▊| 18812/19090 [9:10:57<07:35,  1.64s/it]
 99%|█████████▊| 18813/19090 [9:10:58<07:04,  1.53s/it]
 99%|█████████▊| 18814/19090 [9:11:01<08:29,  1.85s/it]
 99%|█████████▊| 18815/19090 [9:11:02<07:42,  1.68s/it]
 99%|█████████▊| 18816/19090 [9:11:03<07:09,  1.57s/it]
 99%|█████████▊| 18817/19090 [9:11:06<08:10,  1.80s/it]
 99%|█████████▊| 18818/19090 [9:11:07<07:30,  1.66s/it]
 99%|█████████▊| 18819/19090 [9:11:08<06:59,  1.55s/it][gradnorm] step 18819: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1043.5

 99%|█████████▊| 18820/19090 [9:11:10<07:38,  1.70s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1491.2103271484375, 'learning_rate': 3.068757669336675e-08, 'recon_loss': 0.0013982852455228568, 'recon_static': 0.0004200651732389815, 'recon_motion': 0.0009782200708286838, 'recon_norm': 1.2181389331817627, 'recon_norm_std': 0.04293230641633272, 'target_norm': 1.6206505060195924, 'recon_cos': 0.6721069037914276, 'gradnorm/compressor.query': 0.41178333347775153, 'gradnorm/compressor.layer0': 11.253914385193172, 'gradnorm/compressor.layer1': 1478.7308747706595, 'gradnorm/compressor.layer2': 186.6808878976367, 'gradnorm/compressor.layer3': 43.68746088159908, 'gradnorm/compressor.layer4': 10.67220277967066, 'gradnorm/compressor.layer5': 5.360485618000576, 'gradnorm/compressor.layer6': 2.751514537941321, 'gradnorm/compressor.layer7': 3.2799975230044804, 'gradnorm/decoder.layer0': 3.403721422067759, 'gradnorm/decoder.layer1': 1.992365271920478, 'gradnorm/decoder.layer2': 1.1956072175956385, 'gradnorm/max_param': 1043.4632568359375, 'epoch': 9.86}

 99%|█████████▊| 18820/19090 [9:11:10<07:38,  1.70s/it]
 99%|█████████▊| 18821/19090 [9:11:12<07:17,  1.63s/it]
 99%|█████████▊| 18822/19090 [9:11:13<06:48,  1.52s/it]
 99%|█████████▊| 18823/19090 [9:11:16<08:07,  1.83s/it]
 99%|█████████▊| 18824/19090 [9:11:17<07:38,  1.72s/it]
 99%|█████████▊| 18825/19090 [9:11:19<07:01,  1.59s/it]
 99%|█████████▊| 18826/19090 [9:11:21<08:08,  1.85s/it]
 99%|█████████▊| 18827/19090 [9:11:23<07:50,  1.79s/it]
 99%|█████████▊| 18828/19090 [9:11:24<07:09,  1.64s/it]
 99%|█████████▊| 18829/19090 [9:11:26<07:57,  1.83s/it][gradnorm] step 18829: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1115.4

 99%|█████████▊| 18830/19090 [9:11:28<08:03,  1.86s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1595.1793212890625, 'learning_rate': 2.846501818045133e-08, 'recon_loss': 0.0014038394205272199, 'recon_static': 0.00040998797994689084, 'recon_motion': 0.000993851434031967, 'recon_norm': 1.2162604868412017, 'recon_norm_std': 0.04141861675307155, 'target_norm': 1.6203699231147766, 'recon_cos': 0.6705399304628372, 'gradnorm/compressor.query': 0.5497948491518868, 'gradnorm/compressor.layer0': 18.663550707193956, 'gradnorm/compressor.layer1': 1580.6841082265614, 'gradnorm/compressor.layer2': 209.02338946574855, 'gradnorm/compressor.layer3': 42.798995646163384, 'gradnorm/compressor.layer4': 9.898623253239327, 'gradnorm/compressor.layer5': 5.559306395923844, 'gradnorm/compressor.layer6': 2.7062442841150123, 'gradnorm/compressor.layer7': 3.2080801991928327, 'gradnorm/decoder.layer0': 3.4214580403784454, 'gradnorm/decoder.layer1': 1.9934254708041526, 'gradnorm/decoder.layer2': 1.1982083615080854, 'gradnorm/max_param': 1115.4366455078125, 'epoch': 9.86}

 99%|█████████▊| 18830/19090 [9:11:28<08:03,  1.86s/it]
 99%|█████████▊| 18831/19090 [9:11:29<07:16,  1.69s/it]
 99%|█████████▊| 18832/19090 [9:11:31<07:28,  1.74s/it]
 99%|█████████▊| 18833/19090 [9:11:33<07:33,  1.77s/it]
 99%|█████████▊| 18834/19090 [9:11:34<06:55,  1.62s/it]
 99%|█████████▊| 18835/19090 [9:11:36<07:04,  1.66s/it]
 99%|█████████▊| 18836/19090 [9:11:38<07:50,  1.85s/it]
 99%|█████████▊| 18837/19090 [9:11:40<07:05,  1.68s/it]
 99%|█████████▊| 18838/19090 [9:11:41<06:57,  1.66s/it]
 99%|█████████▊| 18839/19090 [9:11:43<07:32,  1.80s/it][gradnorm] step 18839: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 826.5

 99%|█████████▊| 18840/19090 [9:11:45<06:51,  1.65s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1180.461669921875, 'learning_rate': 2.6325952267966524e-08, 'recon_loss': 0.001402043568668887, 'recon_static': 0.0004134082060772926, 'recon_motion': 0.000988635422254447, 'recon_norm': 1.216834056377411, 'recon_norm_std': 0.0422817550599575, 'target_norm': 1.6201041132211684, 'recon_cos': 0.6709951370954513, 'gradnorm/compressor.query': 0.3428878810492042, 'gradnorm/compressor.layer0': 8.558567007032613, 'gradnorm/compressor.layer1': 1171.0253519885896, 'gradnorm/compressor.layer2': 142.76598668143964, 'gradnorm/compressor.layer3': 39.60778319040101, 'gradnorm/compressor.layer4': 9.896877343503037, 'gradnorm/compressor.layer5': 5.600019979441288, 'gradnorm/compressor.layer6': 2.7917386420541463, 'gradnorm/compressor.layer7': 3.2393983908766844, 'gradnorm/decoder.layer0': 3.450552970688587, 'gradnorm/decoder.layer1': 2.010621832885714, 'gradnorm/decoder.layer2': 1.2027737488702415, 'gradnorm/max_param': 826.470458984375, 'epoch': 9.87}

 99%|█████████▊| 18840/19090 [9:11:45<06:51,  1.65s/it]
 99%|█████████▊| 18841/19090 [9:11:46<06:52,  1.66s/it]
 99%|█████████▊| 18842/19090 [9:11:48<07:14,  1.75s/it]
 99%|█████████▊| 18843/19090 [9:11:50<06:39,  1.62s/it]
 99%|█████████▊| 18844/19090 [9:11:52<07:07,  1.74s/it]
 99%|█████████▊| 18845/19090 [9:11:53<06:47,  1.66s/it]
 99%|█████████▊| 18846/19090 [9:11:54<06:17,  1.55s/it]
 99%|█████████▊| 18847/19090 [9:11:57<06:59,  1.73s/it]
 99%|█████████▊| 18848/19090 [9:11:58<06:50,  1.70s/it]
 99%|█████████▊| 18849/19090 [9:12:00<06:19,  1.58s/it][gradnorm] step 18849: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 965.8

 99%|█████████▊| 18850/19090 [9:12:02<07:07,  1.78s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1377.89404296875, 'learning_rate': 2.427038610790522e-08, 'recon_loss': 0.0014072256191866473, 'recon_static': 0.00041450904682278634, 'recon_motion': 0.0009927165665430948, 'recon_norm': 1.2166088670492172, 'recon_norm_std': 0.04288644604384899, 'target_norm': 1.6224584132432938, 'recon_cos': 0.6706778064370156, 'gradnorm/compressor.query': 0.36777246832656585, 'gradnorm/compressor.layer0': 8.424676489349327, 'gradnorm/compressor.layer1': 1368.5188617625993, 'gradnorm/compressor.layer2': 153.31086933498878, 'gradnorm/compressor.layer3': 44.734205815258136, 'gradnorm/compressor.layer4': 10.37171277283965, 'gradnorm/compressor.layer5': 5.352280795487576, 'gradnorm/compressor.layer6': 2.8063968135772153, 'gradnorm/compressor.layer7': 3.2559895709665096, 'gradnorm/decoder.layer0': 3.464000440152605, 'gradnorm/decoder.layer1': 1.9788134276393112, 'gradnorm/decoder.layer2': 1.189549634804677, 'gradnorm/max_param': 965.7890625, 'epoch': 9.87}

 99%|█████████▊| 18850/19090 [9:12:02<07:07,  1.78s/it]
 99%|█████████▊| 18851/19090 [9:12:03<06:40,  1.68s/it]
 99%|█████████▉| 18852/19090 [9:12:05<06:10,  1.56s/it]
 99%|█████████▉| 18853/19090 [9:12:06<06:30,  1.65s/it]
 99%|█████████▉| 18854/19090 [9:12:08<06:26,  1.64s/it]
 99%|█████████▉| 18855/19090 [9:12:09<05:59,  1.53s/it]
 99%|█████████▉| 18856/19090 [9:12:11<06:25,  1.65s/it]
 99%|█████████▉| 18857/19090 [9:12:13<06:35,  1.70s/it]
 99%|█████████▉| 18858/19090 [9:12:14<06:04,  1.57s/it]
 99%|█████████▉| 18859/19090 [9:12:16<05:58,  1.55s/it][gradnorm] step 18859: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 779.9

 99%|█████████▉| 18860/19090 [9:12:17<06:05,  1.59s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1114.1923828125, 'learning_rate': 2.2298326573072515e-08, 'recon_loss': 0.0013882578117772938, 'recon_static': 0.0004132274436415173, 'recon_motion': 0.0009750303812325001, 'recon_norm': 1.216953733563423, 'recon_norm_std': 0.04254338955506683, 'target_norm': 1.6172627657651901, 'recon_cos': 0.6729830026626586, 'gradnorm/compressor.query': 0.4691112874408235, 'gradnorm/compressor.layer0': 10.524374508790038, 'gradnorm/compressor.layer1': 1105.2286528135253, 'gradnorm/compressor.layer2': 133.55850777842645, 'gradnorm/compressor.layer3': 42.301410162757136, 'gradnorm/compressor.layer4': 9.574153804986787, 'gradnorm/compressor.layer5': 5.3986604371252005, 'gradnorm/compressor.layer6': 2.8768238419271466, 'gradnorm/compressor.layer7': 3.233959374006892, 'gradnorm/decoder.layer0': 3.385389936292709, 'gradnorm/decoder.layer1': 1.9944565958888454, 'gradnorm/decoder.layer2': 1.1943645333510688, 'gradnorm/max_param': 779.9343872070312, 'epoch': 9.88}

 99%|█████████▉| 18860/19090 [9:12:17<06:05,  1.59s/it]
 99%|█████████▉| 18861/19090 [9:12:19<05:43,  1.50s/it]
 99%|█████████▉| 18862/19090 [9:12:20<05:49,  1.53s/it]
 99%|█████████▉| 18863/19090 [9:12:22<05:49,  1.54s/it]
 99%|█████████▉| 18864/19090 [9:12:23<05:30,  1.46s/it]
 99%|█████████▉| 18865/19090 [9:12:25<05:46,  1.54s/it]
 99%|█████████▉| 18866/19090 [9:12:27<06:01,  1.61s/it]
 99%|█████████▉| 18867/19090 [9:12:28<05:37,  1.51s/it]
 99%|█████████▉| 18868/19090 [9:12:30<06:03,  1.64s/it]
 99%|█████████▉| 18869/19090 [9:12:32<06:05,  1.65s/it][gradnorm] step 18869: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1004.7

 99%|█████████▉| 18870/19090 [9:12:33<05:39,  1.54s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1433.7476806640625, 'learning_rate': 2.040978025707463e-08, 'recon_loss': 0.0013921001402195543, 'recon_static': 0.00041939345392165704, 'recon_motion': 0.0009727067605126649, 'recon_norm': 1.21788909137249, 'recon_norm_std': 0.043021156825125216, 'target_norm': 1.6185077160596848, 'recon_cos': 0.6729097217321396, 'gradnorm/compressor.query': 0.3859505971624482, 'gradnorm/compressor.layer0': 7.731363592647094, 'gradnorm/compressor.layer1': 1423.7545873499407, 'gradnorm/compressor.layer2': 162.88965901991753, 'gradnorm/compressor.layer3': 42.55568215860691, 'gradnorm/compressor.layer4': 9.253915731184463, 'gradnorm/compressor.layer5': 5.448434118297721, 'gradnorm/compressor.layer6': 2.857865899422937, 'gradnorm/compressor.layer7': 3.1953507027056123, 'gradnorm/decoder.layer0': 3.4459764874453755, 'gradnorm/decoder.layer1': 1.9757880702868271, 'gradnorm/decoder.layer2': 1.185794157228634, 'gradnorm/max_param': 1004.744873046875, 'epoch': 9.88}

 99%|█████████▉| 18870/19090 [9:12:33<05:39,  1.54s/it]
 99%|█████████▉| 18871/19090 [9:12:35<06:28,  1.77s/it]
 99%|█████████▉| 18872/19090 [9:12:37<05:57,  1.64s/it]
 99%|█████████▉| 18873/19090 [9:12:38<05:33,  1.54s/it]
 99%|█████████▉| 18874/19090 [9:12:40<06:24,  1.78s/it]
 99%|█████████▉| 18875/19090 [9:12:41<05:50,  1.63s/it]
 99%|█████████▉| 18876/19090 [9:12:43<05:26,  1.53s/it]
 99%|█████████▉| 18877/19090 [9:12:45<06:16,  1.77s/it]
 99%|█████████▉| 18878/19090 [9:12:47<05:56,  1.68s/it]
 99%|█████████▉| 18879/19090 [9:12:48<05:29,  1.56s/it][gradnorm] step 18879: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1053.9

 99%|█████████▉| 18880/19090 [9:12:50<06:19,  1.81s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1507.045654296875, 'learning_rate': 1.8604753474280036e-08, 'recon_loss': 0.0013981476513436064, 'recon_static': 0.00041622017015470193, 'recon_motion': 0.0009819275117479264, 'recon_norm': 1.2178600698709487, 'recon_norm_std': 0.042465387284755705, 'target_norm': 1.6199190616607666, 'recon_cos': 0.6720697447657585, 'gradnorm/compressor.query': 0.3863393290161161, 'gradnorm/compressor.layer0': 11.198369207183278, 'gradnorm/compressor.layer1': 1493.2570944080594, 'gradnorm/compressor.layer2': 197.63090005424252, 'gradnorm/compressor.layer3': 44.74280750290696, 'gradnorm/compressor.layer4': 10.824680061105408, 'gradnorm/compressor.layer5': 5.643232509000029, 'gradnorm/compressor.layer6': 2.803796143069187, 'gradnorm/compressor.layer7': 3.2248574986874723, 'gradnorm/decoder.layer0': 3.3940715939507347, 'gradnorm/decoder.layer1': 1.966757113860752, 'gradnorm/decoder.layer2': 1.1892252987310554, 'gradnorm/max_param': 1053.9102783203125, 'epoch': 9.89}

 99%|█████████▉| 18880/19090 [9:12:50<06:19,  1.81s/it]
 99%|█████████▉| 18881/19090 [9:12:52<05:46,  1.66s/it]
 99%|█████████▉| 18882/19090 [9:12:53<05:21,  1.55s/it]
 99%|█████████▉| 18883/19090 [9:12:55<06:18,  1.83s/it]
 99%|█████████▉| 18884/19090 [9:12:57<05:45,  1.68s/it]
 99%|█████████▉| 18885/19090 [9:12:58<05:19,  1.56s/it]
 99%|█████████▉| 18886/19090 [9:13:01<06:31,  1.92s/it]
 99%|█████████▉| 18887/19090 [9:13:02<06:04,  1.80s/it]
 99%|█████████▉| 18888/19090 [9:13:03<05:31,  1.64s/it]
 99%|█████████▉| 18889/19090 [9:13:06<06:16,  1.87s/it][gradnorm] step 18889: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 967.4

 99%|█████████▉| 18890/19090 [9:13:08<06:17,  1.89s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1381.0703125, 'learning_rate': 1.6883252259822235e-08, 'recon_loss': 0.001397439491120167, 'recon_static': 0.000419681837229291, 'recon_motion': 0.0009777576473425143, 'recon_norm': 1.2180467307567597, 'recon_norm_std': 0.04332617139443755, 'target_norm': 1.620122829079628, 'recon_cos': 0.6722355902194976, 'gradnorm/compressor.query': 0.33852514324457056, 'gradnorm/compressor.layer0': 8.955166089739722, 'gradnorm/compressor.layer1': 1370.5666711254873, 'gradnorm/compressor.layer2': 163.4423636943617, 'gradnorm/compressor.layer3': 43.94430623847929, 'gradnorm/compressor.layer4': 10.527582543094125, 'gradnorm/compressor.layer5': 5.600795124008569, 'gradnorm/compressor.layer6': 2.8643110851038047, 'gradnorm/compressor.layer7': 3.2290719459341632, 'gradnorm/decoder.layer0': 3.3888059737966043, 'gradnorm/decoder.layer1': 1.965286235394231, 'gradnorm/decoder.layer2': 1.1741089445283692, 'gradnorm/max_param': 967.357177734375, 'epoch': 9.9}

 99%|█████████▉| 18890/19090 [9:13:08<06:17,  1.89s/it]
 99%|█████████▉| 18891/19090 [9:13:09<05:39,  1.71s/it]
 99%|█████████▉| 18892/19090 [9:13:11<06:17,  1.91s/it]
 99%|█████████▉| 18893/19090 [9:13:13<05:57,  1.81s/it]
 99%|█████████▉| 18894/19090 [9:13:14<05:25,  1.66s/it]
 99%|█████████▉| 18895/19090 [9:13:17<06:01,  1.85s/it]
 99%|█████████▉| 18896/19090 [9:13:19<06:03,  1.88s/it]
 99%|█████████▉| 18897/19090 [9:13:20<05:28,  1.70s/it]
 99%|█████████▉| 18898/19090 [9:13:22<05:53,  1.84s/it]
 99%|█████████▉| 18899/19090 [9:13:24<05:58,  1.88s/it][gradnorm] step 18899: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1101.8

 99%|█████████▉| 18900/19090 [9:13:25<05:23,  1.70s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1573.0953369140625, 'learning_rate': 1.5245282369558135e-08, 'recon_loss': 0.001402039299136959, 'recon_static': 0.0004158850853855256, 'recon_motion': 0.000986154202837497, 'recon_norm': 1.217264461517334, 'recon_norm_std': 0.042040863633155824, 'target_norm': 1.6213477075099945, 'recon_cos': 0.6715509623289109, 'gradnorm/compressor.query': 0.8075691033547951, 'gradnorm/compressor.layer0': 17.768554445975884, 'gradnorm/compressor.layer1': 1561.1498166415677, 'gradnorm/compressor.layer2': 186.13319533669969, 'gradnorm/compressor.layer3': 48.0282865465108, 'gradnorm/compressor.layer4': 10.377338191907342, 'gradnorm/compressor.layer5': 5.414469561886619, 'gradnorm/compressor.layer6': 2.7322839207211604, 'gradnorm/compressor.layer7': 3.207158819117014, 'gradnorm/decoder.layer0': 3.376606735370455, 'gradnorm/decoder.layer1': 1.9684830136388634, 'gradnorm/decoder.layer2': 1.1745656976168632, 'gradnorm/max_param': 1101.7620849609375, 'epoch': 9.9}

 99%|█████████▉| 18900/19090 [9:13:25<05:23,  1.70s/it]
 99%|█████████▉| 18901/19090 [9:13:27<05:28,  1.74s/it]
 99%|█████████▉| 18902/19090 [9:13:29<05:37,  1.80s/it]
 99%|█████████▉| 18903/19090 [9:13:30<05:07,  1.64s/it]
 99%|█████████▉| 18904/19090 [9:13:33<05:42,  1.84s/it]
 99%|█████████▉| 18905/19090 [9:13:35<05:43,  1.85s/it]
 99%|█████████▉| 18906/19090 [9:13:36<05:09,  1.68s/it]
 99%|█████████▉| 18907/19090 [9:13:38<05:41,  1.87s/it]
 99%|█████████▉| 18908/19090 [9:13:40<05:32,  1.83s/it]
 99%|█████████▉| 18909/19090 [9:13:41<05:00,  1.66s/it][gradnorm] step 18909: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 962.4

 99%|█████████▉| 18910/19090 [9:13:43<05:33,  1.85s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1374.78759765625, 'learning_rate': 1.3690849280056928e-08, 'recon_loss': 0.0014018924644915386, 'recon_static': 0.00041676607143017466, 'recon_motion': 0.000985126402520109, 'recon_norm': 1.2170828819274901, 'recon_norm_std': 0.04167673941701651, 'target_norm': 1.6204241037368774, 'recon_cos': 0.6712213352322578, 'gradnorm/compressor.query': 0.39903564061024205, 'gradnorm/compressor.layer0': 9.733265135365034, 'gradnorm/compressor.layer1': 1363.5430869613178, 'gradnorm/compressor.layer2': 169.77679510536768, 'gradnorm/compressor.layer3': 41.479385587432915, 'gradnorm/compressor.layer4': 9.485924181035783, 'gradnorm/compressor.layer5': 5.147215355167839, 'gradnorm/compressor.layer6': 2.824302070623238, 'gradnorm/compressor.layer7': 3.171925511451256, 'gradnorm/decoder.layer0': 3.423958897201326, 'gradnorm/decoder.layer1': 1.9698860281686348, 'gradnorm/decoder.layer2': 1.1837530249261927, 'gradnorm/max_param': 962.3846435546875, 'epoch': 9.91}

 99%|█████████▉| 18910/19090 [9:13:43<05:33,  1.85s/it]
 99%|█████████▉| 18911/19090 [9:13:45<05:21,  1.80s/it]
 99%|█████████▉| 18912/19090 [9:13:46<04:52,  1.64s/it]
 99%|█████████▉| 18913/19090 [9:13:49<05:24,  1.84s/it]
 99%|█████████▉| 18914/19090 [9:13:50<05:24,  1.85s/it]
 99%|█████████▉| 18915/19090 [9:13:52<04:53,  1.68s/it]
 99%|█████████▉| 18916/19090 [9:13:54<05:11,  1.79s/it]
 99%|█████████▉| 18917/19090 [9:13:56<05:37,  1.95s/it]
 99%|█████████▉| 18918/19090 [9:13:57<05:01,  1.75s/it]
 99%|█████████▉| 18919/19090 [9:13:59<05:01,  1.76s/it][gradnorm] step 18919: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1070.2

 99%|█████████▉| 18920/19090 [9:14:01<05:24,  1.91s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1527.8603515625, 'learning_rate': 1.2219958188580682e-08, 'recon_loss': 0.0013964743353426456, 'recon_static': 0.0004173752538918052, 'recon_motion': 0.0009790990588953719, 'recon_norm': 1.218177255988121, 'recon_norm_std': 0.0430349656380713, 'target_norm': 1.620012539625168, 'recon_cos': 0.6724866017699241, 'gradnorm/compressor.query': 0.4305171932371465, 'gradnorm/compressor.layer0': 10.02372064584491, 'gradnorm/compressor.layer1': 1516.4030631728492, 'gradnorm/compressor.layer2': 181.7638083647292, 'gradnorm/compressor.layer3': 39.58746620420092, 'gradnorm/compressor.layer4': 10.404899074169453, 'gradnorm/compressor.layer5': 5.46985880918841, 'gradnorm/compressor.layer6': 2.726544136925538, 'gradnorm/compressor.layer7': 3.2506978679543788, 'gradnorm/decoder.layer0': 3.4303514537422264, 'gradnorm/decoder.layer1': 1.963549815463722, 'gradnorm/decoder.layer2': 1.1843308775239565, 'gradnorm/max_param': 1070.201171875, 'epoch': 9.91}

 99%|█████████▉| 18920/19090 [9:14:02<05:24,  1.91s/it]
 99%|█████████▉| 18921/19090 [9:14:03<04:51,  1.72s/it]
 99%|█████████▉| 18922/19090 [9:14:05<05:04,  1.81s/it]
 99%|█████████▉| 18923/19090 [9:14:07<05:04,  1.82s/it]
 99%|█████████▉| 18924/19090 [9:14:08<04:35,  1.66s/it]
 99%|█████████▉| 18925/19090 [9:14:10<04:55,  1.79s/it]
 99%|█████████▉| 18926/19090 [9:14:12<05:11,  1.90s/it]
 99%|█████████▉| 18927/19090 [9:14:13<04:39,  1.71s/it]
 99%|█████████▉| 18928/19090 [9:14:16<05:00,  1.85s/it]
 99%|█████████▉| 18929/19090 [9:14:17<04:38,  1.73s/it][gradnorm] step 18929: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 826.6

 99%|█████████▉| 18930/19090 [9:14:18<04:15,  1.60s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1183.073486328125, 'learning_rate': 1.0832614013073228e-08, 'recon_loss': 0.001396520368871279, 'recon_static': 0.00042004219722002745, 'recon_motion': 0.0009764781105332076, 'recon_norm': 1.2182438015937804, 'recon_norm_std': 0.043104514665901664, 'target_norm': 1.6206108182668686, 'recon_cos': 0.6725381940603257, 'gradnorm/compressor.query': 0.34919947031211446, 'gradnorm/compressor.layer0': 7.506474307200299, 'gradnorm/compressor.layer1': 1171.1977843216746, 'gradnorm/compressor.layer2': 160.20414832526964, 'gradnorm/compressor.layer3': 45.32039532508377, 'gradnorm/compressor.layer4': 10.785016184349455, 'gradnorm/compressor.layer5': 5.571488394084209, 'gradnorm/compressor.layer6': 2.7982026257716734, 'gradnorm/compressor.layer7': 3.248789782043861, 'gradnorm/decoder.layer0': 3.459794176407468, 'gradnorm/decoder.layer1': 1.9713714948234433, 'gradnorm/decoder.layer2': 1.177790196046399, 'gradnorm/max_param': 826.5943603515625, 'epoch': 9.92}

 99%|█████████▉| 18930/19090 [9:14:18<04:15,  1.60s/it]
 99%|█████████▉| 18931/19090 [9:14:21<04:54,  1.85s/it]
 99%|█████████▉| 18932/19090 [9:14:23<04:46,  1.81s/it]
 99%|█████████▉| 18933/19090 [9:14:24<04:20,  1.66s/it]
 99%|█████████▉| 18934/19090 [9:14:26<04:48,  1.85s/it]
 99%|█████████▉| 18935/19090 [9:14:28<04:46,  1.85s/it]
 99%|█████████▉| 18936/19090 [9:14:29<04:18,  1.68s/it]
 99%|█████████▉| 18937/19090 [9:14:31<04:40,  1.83s/it]
 99%|█████████▉| 18938/19090 [9:14:34<04:53,  1.93s/it]
 99%|█████████▉| 18939/19090 [9:14:35<04:21,  1.73s/it][gradnorm] step 18939: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 923.0

 99%|█████████▉| 18940/19090 [9:14:37<04:22,  1.75s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1321.9866943359375, 'learning_rate': 9.528821392132404e-09, 'recon_loss': 0.0014022933726664632, 'recon_static': 0.00042080322964466177, 'recon_motion': 0.0009814901859499514, 'recon_norm': 1.218305665254593, 'recon_norm_std': 0.04323938293382525, 'target_norm': 1.6217082917690278, 'recon_cos': 0.6717040106654167, 'gradnorm/compressor.query': 0.5324150941435146, 'gradnorm/compressor.layer0': 14.025878421793399, 'gradnorm/compressor.layer1': 1308.0332086762935, 'gradnorm/compressor.layer2': 185.9019924079621, 'gradnorm/compressor.layer3': 42.03351962970669, 'gradnorm/compressor.layer4': 10.471617712982816, 'gradnorm/compressor.layer5': 5.393962324611303, 'gradnorm/compressor.layer6': 2.7788374649127263, 'gradnorm/compressor.layer7': 3.2646599762827044, 'gradnorm/decoder.layer0': 3.422847243889929, 'gradnorm/decoder.layer1': 1.9747919890559797, 'gradnorm/decoder.layer2': 1.1789845778707955, 'gradnorm/max_param': 922.9578247070312, 'epoch': 9.92}

 99%|█████████▉| 18940/19090 [9:14:37<04:22,  1.75s/it]
 99%|█████████▉| 18941/19090 [9:14:39<04:33,  1.84s/it]
 99%|█████████▉| 18942/19090 [9:14:40<04:07,  1.67s/it]
 99%|█████████▉| 18943/19090 [9:14:42<04:18,  1.76s/it]
 99%|█████████▉| 18944/19090 [9:14:44<04:28,  1.84s/it]
 99%|█████████▉| 18945/19090 [9:14:45<04:02,  1.67s/it]
 99%|█████████▉| 18946/19090 [9:14:47<04:06,  1.71s/it]
 99%|█████████▉| 18947/19090 [9:14:49<04:25,  1.85s/it]
 99%|█████████▉| 18948/19090 [9:14:51<03:59,  1.68s/it]
 99%|█████████▉| 18949/19090 [9:14:52<04:02,  1.72s/it][gradnorm] step 18949: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 986.4

 99%|█████████▉| 18950/19090 [9:14:55<04:20,  1.86s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1411.679443359375, 'learning_rate': 8.308584685007282e-09, 'recon_loss': 0.00140476307424251, 'recon_static': 0.0004126509818888735, 'recon_motion': 0.000992112078529317, 'recon_norm': 1.217075663805008, 'recon_norm_std': 0.04239345099776983, 'target_norm': 1.621772763133049, 'recon_cos': 0.6710523024201394, 'gradnorm/compressor.query': 0.3453806003029528, 'gradnorm/compressor.layer0': 9.169111261096104, 'gradnorm/compressor.layer1': 1397.8423552031895, 'gradnorm/compressor.layer2': 190.3564604400124, 'gradnorm/compressor.layer3': 48.57779078768918, 'gradnorm/compressor.layer4': 11.243444270440442, 'gradnorm/compressor.layer5': 5.86121846000791, 'gradnorm/compressor.layer6': 2.774991398016167, 'gradnorm/compressor.layer7': 3.276573554417377, 'gradnorm/decoder.layer0': 3.3941605237091492, 'gradnorm/decoder.layer1': 1.975699194746186, 'gradnorm/decoder.layer2': 1.1795030664434025, 'gradnorm/max_param': 986.4479370117188, 'epoch': 9.93}

 99%|█████████▉| 18950/19090 [9:14:55<04:20,  1.86s/it]
 99%|█████████▉| 18951/19090 [9:14:56<03:54,  1.69s/it]
 99%|█████████▉| 18952/19090 [9:14:58<04:05,  1.78s/it]
 99%|█████████▉| 18953/19090 [9:15:00<04:22,  1.92s/it]
 99%|█████████▉| 18954/19090 [9:15:01<03:55,  1.73s/it]
 99%|█████████▉| 18955/19090 [9:15:03<04:05,  1.82s/it]
 99%|█████████▉| 18956/19090 [9:15:05<04:02,  1.81s/it]
 99%|█████████▉| 18957/19090 [9:15:06<03:39,  1.65s/it]
 99%|█████████▉| 18958/19090 [9:15:09<03:56,  1.79s/it]
 99%|█████████▉| 18959/19090 [9:15:10<03:59,  1.83s/it][gradnorm] step 18959: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1050.4

 99%|█████████▉| 18960/19090 [9:15:12<03:36,  1.66s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1499.870361328125, 'learning_rate': 7.171907971567638e-09, 'recon_loss': 0.001402268305537291, 'recon_static': 0.0004193675580609124, 'recon_motion': 0.000982900736562442, 'recon_norm': 1.2181051254272461, 'recon_norm_std': 0.043614133913069966, 'target_norm': 1.622235941886902, 'recon_cos': 0.6717708125710488, 'gradnorm/compressor.query': 0.47181344357091853, 'gradnorm/compressor.layer0': 17.258969835854106, 'gradnorm/compressor.layer1': 1488.5312391750467, 'gradnorm/compressor.layer2': 177.66796080575418, 'gradnorm/compressor.layer3': 42.981887196272865, 'gradnorm/compressor.layer4': 10.458457875111812, 'gradnorm/compressor.layer5': 5.521439640172955, 'gradnorm/compressor.layer6': 2.823348506164615, 'gradnorm/compressor.layer7': 3.1684901023687284, 'gradnorm/decoder.layer0': 3.4211472386240156, 'gradnorm/decoder.layer1': 1.9702597989953277, 'gradnorm/decoder.layer2': 1.1822317111011764, 'gradnorm/max_param': 1050.447998046875, 'epoch': 9.93}

 99%|█████████▉| 18960/19090 [9:15:12<03:36,  1.66s/it]
 99%|█████████▉| 18961/19090 [9:15:14<03:53,  1.81s/it]
 99%|█████████▉| 18962/19090 [9:15:16<03:55,  1.84s/it]
 99%|█████████▉| 18963/19090 [9:15:17<03:32,  1.67s/it]
 99%|█████████▉| 18964/19090 [9:15:19<03:52,  1.84s/it]
 99%|█████████▉| 18965/19090 [9:15:21<03:53,  1.87s/it]
 99%|█████████▉| 18966/19090 [9:15:23<03:30,  1.69s/it]
 99%|█████████▉| 18967/19090 [9:15:25<03:40,  1.79s/it]
 99%|█████████▉| 18968/19090 [9:15:27<04:02,  1.99s/it]
 99%|█████████▉| 18969/19090 [9:15:28<03:35,  1.78s/it][gradnorm] step 18969: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 998.8

 99%|█████████▉| 18970/19090 [9:15:30<03:24,  1.71s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1424.34375, 'learning_rate': 6.1187950523039496e-09, 'recon_loss': 0.0014062081085285173, 'recon_static': 0.0004166865772276651, 'recon_motion': 0.0009895215189317241, 'recon_norm': 1.2176768362522126, 'recon_norm_std': 0.042691546771675346, 'target_norm': 1.6218890428543091, 'recon_cos': 0.670793955028057, 'gradnorm/compressor.query': 0.4501382456459085, 'gradnorm/compressor.layer0': 16.139344924433697, 'gradnorm/compressor.layer1': 1415.3644407006982, 'gradnorm/compressor.layer2': 153.75680244098146, 'gradnorm/compressor.layer3': 37.989659558864666, 'gradnorm/compressor.layer4': 9.521068300203945, 'gradnorm/compressor.layer5': 5.39872403081401, 'gradnorm/compressor.layer6': 2.6406696022374994, 'gradnorm/compressor.layer7': 3.188055401453497, 'gradnorm/decoder.layer0': 3.3934685521892294, 'gradnorm/decoder.layer1': 1.9550397108879096, 'gradnorm/decoder.layer2': 1.179371166132046, 'gradnorm/max_param': 998.8190307617188, 'epoch': 9.94}

 99%|█████████▉| 18970/19090 [9:15:30<03:24,  1.71s/it]
 99%|█████████▉| 18971/19090 [9:15:33<03:58,  2.01s/it]
 99%|█████████▉| 18972/19090 [9:15:34<03:31,  1.79s/it]
 99%|█████████▉| 18973/19090 [9:15:35<03:15,  1.67s/it]
 99%|█████████▉| 18974/19090 [9:15:38<03:46,  1.95s/it]
 99%|█████████▉| 18975/19090 [9:15:39<03:21,  1.75s/it]
 99%|█████████▉| 18976/19090 [9:15:41<03:15,  1.71s/it]
 99%|█████████▉| 18977/19090 [9:15:43<03:39,  1.94s/it]
 99%|█████████▉| 18978/19090 [9:15:44<03:15,  1.74s/it]
 99%|█████████▉| 18979/19090 [9:15:46<03:09,  1.71s/it][gradnorm] step 18979: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1031.6

 99%|█████████▉| 18980/19090 [9:15:49<03:33,  1.94s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1472.9119873046875, 'learning_rate': 5.149249448316296e-09, 'recon_loss': 0.0013880839193006977, 'recon_static': 0.00041642762589617633, 'recon_motion': 0.0009716562941321172, 'recon_norm': 1.2181306570768355, 'recon_norm_std': 0.042622745595872404, 'target_norm': 1.6173243552446366, 'recon_cos': 0.6734951585531235, 'gradnorm/compressor.query': 0.382663464242603, 'gradnorm/compressor.layer0': 10.125558343256698, 'gradnorm/compressor.layer1': 1461.7519454408125, 'gradnorm/compressor.layer2': 174.51949098703847, 'gradnorm/compressor.layer3': 44.86101856097479, 'gradnorm/compressor.layer4': 10.815064583637131, 'gradnorm/compressor.layer5': 5.250734731941781, 'gradnorm/compressor.layer6': 2.672357705220407, 'gradnorm/compressor.layer7': 3.207335444902502, 'gradnorm/decoder.layer0': 3.3881661073393157, 'gradnorm/decoder.layer1': 1.9634375574117091, 'gradnorm/decoder.layer2': 1.1795524368273251, 'gradnorm/max_param': 1031.6309814453125, 'epoch': 9.94}

 99%|█████████▉| 18980/19090 [9:15:49<03:33,  1.94s/it]
 99%|█████████▉| 18981/19090 [9:15:50<03:10,  1.74s/it]
 99%|█████████▉| 18982/19090 [9:15:51<02:58,  1.65s/it]
 99%|█████████▉| 18983/19090 [9:15:54<03:21,  1.89s/it]
 99%|█████████▉| 18984/19090 [9:15:55<03:00,  1.71s/it]
 99%|█████████▉| 18985/19090 [9:15:57<03:00,  1.72s/it]
 99%|█████████▉| 18986/19090 [9:15:59<03:17,  1.90s/it]
 99%|█████████▉| 18987/19090 [9:16:00<02:56,  1.71s/it]
 99%|█████████▉| 18988/19090 [9:16:02<02:58,  1.75s/it]
 99%|█████████▉| 18989/19090 [9:16:04<03:10,  1.88s/it][gradnorm] step 18989: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1009.9

 99%|█████████▉| 18990/19090 [9:16:06<02:50,  1.70s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1442.216796875, 'learning_rate': 4.263274401289374e-09, 'recon_loss': 0.001401608917512931, 'recon_static': 0.00041217162506654856, 'recon_motion': 0.0009894373011775314, 'recon_norm': 1.21741826236248, 'recon_norm_std': 0.042463039979338646, 'target_norm': 1.6211416333913804, 'recon_cos': 0.6711769849061966, 'gradnorm/compressor.query': 0.3704990239124079, 'gradnorm/compressor.layer0': 8.407825875320388, 'gradnorm/compressor.layer1': 1430.9623946840811, 'gradnorm/compressor.layer2': 174.6810206759023, 'gradnorm/compressor.layer3': 39.949988572139326, 'gradnorm/compressor.layer4': 9.599020637783306, 'gradnorm/compressor.layer5': 5.460815253683582, 'gradnorm/compressor.layer6': 2.6910196367140586, 'gradnorm/compressor.layer7': 3.1976797573214397, 'gradnorm/decoder.layer0': 3.339486657399417, 'gradnorm/decoder.layer1': 1.9668146339076904, 'gradnorm/decoder.layer2': 1.1807598772445527, 'gradnorm/max_param': 1009.91064453125, 'epoch': 9.95}

 99%|█████████▉| 18990/19090 [9:16:06<02:50,  1.70s/it]
 99%|█████████▉| 18991/19090 [9:16:08<02:56,  1.78s/it]
 99%|█████████▉| 18992/19090 [9:16:10<03:04,  1.88s/it]
 99%|█████████▉| 18993/19090 [9:16:11<02:44,  1.70s/it]
 99%|█████████▉| 18994/19090 [9:16:13<02:53,  1.80s/it]
100%|█████████▉| 18995/19090 [9:16:15<02:54,  1.83s/it]
100%|█████████▉| 18996/19090 [9:16:16<02:37,  1.67s/it]
100%|█████████▉| 18997/19090 [9:16:18<02:49,  1.83s/it]
100%|█████████▉| 18998/19090 [9:16:20<02:39,  1.73s/it]
100%|█████████▉| 18999/19090 [9:16:21<02:25,  1.60s/it][gradnorm] step 18999: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 978.7

100%|█████████▉| 19000/19090 [9:16:24<02:52,  1.92s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1397.428466796875, 'learning_rate': 3.4608728734869534e-09, 'recon_loss': 0.001393243920756504, 'recon_static': 0.0004187410340819042, 'recon_motion': 0.0009745028437464498, 'recon_norm': 1.218057543039322, 'recon_norm_std': 0.04282641839236021, 'target_norm': 1.6188159137964249, 'recon_cos': 0.6726863086223602, 'gradnorm/compressor.query': 0.33495038296722757, 'gradnorm/compressor.layer0': 7.925887376461118, 'gradnorm/compressor.layer1': 1386.7504281592994, 'gradnorm/compressor.layer2': 166.62022334676243, 'gradnorm/compressor.layer3': 41.69679021763096, 'gradnorm/compressor.layer4': 10.175742637282786, 'gradnorm/compressor.layer5': 5.196148384941492, 'gradnorm/compressor.layer6': 2.830948237154426, 'gradnorm/compressor.layer7': 3.300897250874016, 'gradnorm/decoder.layer0': 3.369030087634166, 'gradnorm/decoder.layer1': 1.9687085828511783, 'gradnorm/decoder.layer2': 1.1783947967099766, 'gradnorm/max_param': 978.6932373046875, 'epoch': 9.95}

100%|█████████▉| 19000/19090 [9:16:24<02:52,  1.92s/it]
100%|█████████▉| 19001/19090 [9:16:28<03:53,  2.62s/it]
100%|█████████▉| 19002/19090 [9:16:29<03:15,  2.22s/it]
100%|█████████▉| 19003/19090 [9:16:31<02:48,  1.94s/it]
100%|█████████▉| 19004/19090 [9:16:32<02:29,  1.74s/it]
100%|█████████▉| 19005/19090 [9:16:33<02:16,  1.61s/it]
100%|█████████▉| 19006/19090 [9:16:35<02:06,  1.51s/it]
100%|█████████▉| 19007/19090 [9:16:36<01:59,  1.44s/it]
100%|█████████▉| 19008/19090 [9:16:37<01:54,  1.40s/it]
100%|█████████▉| 19009/19090 [9:16:38<01:50,  1.36s/it][gradnorm] step 19009: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1038.1

100%|█████████▉| 19010/19090 [9:16:40<01:47,  1.34s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1484.60791015625, 'learning_rate': 2.7420475477490936e-09, 'recon_loss': 0.0014115033118287101, 'recon_static': 0.00041621368800406346, 'recon_motion': 0.0009952896391041577, 'recon_norm': 1.2172326862812042, 'recon_norm_std': 0.041952573042362926, 'target_norm': 1.6234993934631348, 'recon_cos': 0.6701883196830749, 'gradnorm/compressor.query': 0.3862405382620886, 'gradnorm/compressor.layer0': 10.725096133497075, 'gradnorm/compressor.layer1': 1471.011896620826, 'gradnorm/compressor.layer2': 193.97061028993542, 'gradnorm/compressor.layer3': 47.57143579824189, 'gradnorm/compressor.layer4': 10.72327098647675, 'gradnorm/compressor.layer5': 5.612341986048193, 'gradnorm/compressor.layer6': 2.7889859912967756, 'gradnorm/compressor.layer7': 3.2026305770756975, 'gradnorm/decoder.layer0': 3.4341643361675596, 'gradnorm/decoder.layer1': 1.9668151187902931, 'gradnorm/decoder.layer2': 1.1795035212465437, 'gradnorm/max_param': 1038.121337890625, 'epoch': 9.96}

100%|█████████▉| 19010/19090 [9:16:40<01:47,  1.34s/it]
100%|█████████▉| 19011/19090 [9:16:41<01:44,  1.32s/it]
100%|█████████▉| 19012/19090 [9:16:42<01:44,  1.33s/it]
100%|█████████▉| 19013/19090 [9:16:44<01:55,  1.50s/it]
100%|█████████▉| 19014/19090 [9:16:46<01:48,  1.43s/it]
100%|█████████▉| 19015/19090 [9:16:48<01:58,  1.58s/it]
100%|█████████▉| 19016/19090 [9:16:50<02:14,  1.82s/it]
100%|█████████▉| 19017/19090 [9:16:51<02:00,  1.66s/it]
100%|█████████▉| 19018/19090 [9:16:53<02:06,  1.75s/it]
100%|█████████▉| 19019/19090 [9:16:55<02:06,  1.78s/it][gradnorm] step 19019: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 819.9

100%|█████████▉| 19020/19090 [9:16:56<01:54,  1.63s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1174.2640380859375, 'learning_rate': 2.1068008274782725e-09, 'recon_loss': 0.0013999217393575237, 'recon_static': 0.00041228826448787005, 'recon_motion': 0.0009876334923319518, 'recon_norm': 1.2171144306659698, 'recon_norm_std': 0.042373277712613344, 'target_norm': 1.6198467582464218, 'recon_cos': 0.6712465554475784, 'gradnorm/compressor.query': 0.37826795003925623, 'gradnorm/compressor.layer0': 8.820113119840206, 'gradnorm/compressor.layer1': 1161.760463692925, 'gradnorm/compressor.layer2': 165.670842712289, 'gradnorm/compressor.layer3': 38.921580187562725, 'gradnorm/compressor.layer4': 10.1951983799887, 'gradnorm/compressor.layer5': 5.482951009266774, 'gradnorm/compressor.layer6': 2.7610256075225936, 'gradnorm/compressor.layer7': 3.214932001323251, 'gradnorm/decoder.layer0': 3.4208067180110815, 'gradnorm/decoder.layer1': 1.9731896614062143, 'gradnorm/decoder.layer2': 1.1794335805700367, 'gradnorm/max_param': 819.8824462890625, 'epoch': 9.96}

100%|█████████▉| 19020/19090 [9:16:56<01:54,  1.63s/it]
100%|█████████▉| 19021/19090 [9:16:59<02:10,  1.90s/it]
100%|█████████▉| 19022/19090 [9:17:00<02:04,  1.83s/it]
100%|█████████▉| 19023/19090 [9:17:02<01:51,  1.66s/it]
100%|█████████▉| 19024/19090 [9:17:04<02:04,  1.89s/it]
100%|█████████▉| 19025/19090 [9:17:06<02:01,  1.88s/it]
100%|█████████▉| 19026/19090 [9:17:07<01:48,  1.70s/it]
100%|█████████▉| 19027/19090 [9:17:10<01:58,  1.88s/it]
100%|█████████▉| 19028/19090 [9:17:11<01:51,  1.79s/it]
100%|█████████▉| 19029/19090 [9:17:12<01:40,  1.64s/it][gradnorm] step 19029: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 830.2

100%|█████████▉| 19030/19090 [9:17:15<01:54,  1.91s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1188.1751708984375, 'learning_rate': 1.5551348366255047e-09, 'recon_loss': 0.0013941669574705883, 'recon_static': 0.00041509232542011885, 'recon_motion': 0.0009790746029466391, 'recon_norm': 1.217815062403679, 'recon_norm_std': 0.04294046750292182, 'target_norm': 1.619430872797966, 'recon_cos': 0.67256710678339, 'gradnorm/compressor.query': 0.39645799069866716, 'gradnorm/compressor.layer0': 9.495608870925015, 'gradnorm/compressor.layer1': 1176.3681396569698, 'gradnorm/compressor.layer2': 161.08290761281907, 'gradnorm/compressor.layer3': 41.37743543389244, 'gradnorm/compressor.layer4': 10.245133151691999, 'gradnorm/compressor.layer5': 5.330694121776853, 'gradnorm/compressor.layer6': 2.7278726922083885, 'gradnorm/compressor.layer7': 3.2017463508418955, 'gradnorm/decoder.layer0': 3.4025902642379116, 'gradnorm/decoder.layer1': 1.9733459475582347, 'gradnorm/decoder.layer2': 1.1795688089446479, 'gradnorm/max_param': 830.1882934570312, 'epoch': 9.97}

100%|█████████▉| 19030/19090 [9:17:15<01:54,  1.91s/it]
100%|█████████▉| 19031/19090 [9:17:17<01:46,  1.81s/it]
100%|█████████▉| 19032/19090 [9:17:18<01:35,  1.65s/it]
100%|█████████▉| 19033/19090 [9:17:21<01:51,  1.96s/it]
100%|█████████▉| 19034/19090 [9:17:22<01:42,  1.83s/it]
100%|█████████▉| 19035/19090 [9:17:23<01:31,  1.66s/it]
100%|█████████▉| 19036/19090 [9:17:26<01:44,  1.94s/it]
100%|█████████▉| 19037/19090 [9:17:27<01:34,  1.78s/it]
100%|█████████▉| 19038/19090 [9:17:29<01:24,  1.63s/it]
100%|█████████▉| 19039/19090 [9:17:32<01:44,  2.05s/it][gradnorm] step 19039: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1102.9

100%|█████████▉| 19040/19090 [9:17:33<01:31,  1.83s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1576.9605712890625, 'learning_rate': 1.087051419693119e-09, 'recon_loss': 0.0013996749970829115, 'recon_static': 0.00041928002756321805, 'recon_motion': 0.0009803949695196934, 'recon_norm': 1.2177575021982192, 'recon_norm_std': 0.04254167722538114, 'target_norm': 1.620427393913269, 'recon_cos': 0.6720516815781593, 'gradnorm/compressor.query': 0.4690734383199095, 'gradnorm/compressor.layer0': 7.079033597519134, 'gradnorm/compressor.layer1': 1562.5777580651786, 'gradnorm/compressor.layer2': 205.9146705859007, 'gradnorm/compressor.layer3': 50.39856768798097, 'gradnorm/compressor.layer4': 10.014165001206377, 'gradnorm/compressor.layer5': 5.351870963680799, 'gradnorm/compressor.layer6': 2.8123343524895597, 'gradnorm/compressor.layer7': 3.2304595984037543, 'gradnorm/decoder.layer0': 3.4155458767478786, 'gradnorm/decoder.layer1': 1.9616804805108266, 'gradnorm/decoder.layer2': 1.1798547285173133, 'gradnorm/max_param': 1102.8970947265625, 'epoch': 9.97}

100%|█████████▉| 19040/19090 [9:17:33<01:31,  1.83s/it]
100%|█████████▉| 19041/19090 [9:17:34<01:21,  1.67s/it]
100%|█████████▉| 19042/19090 [9:17:36<01:25,  1.77s/it]
100%|█████████▉| 19043/19090 [9:17:38<01:23,  1.79s/it]
100%|█████████▉| 19044/19090 [9:17:39<01:15,  1.64s/it]
100%|█████████▉| 19045/19090 [9:17:42<01:21,  1.80s/it]
100%|█████████▉| 19046/19090 [9:17:43<01:18,  1.77s/it]
100%|█████████▉| 19047/19090 [9:17:45<01:09,  1.63s/it]
100%|█████████▉| 19048/19090 [9:17:47<01:17,  1.85s/it]
100%|█████████▉| 19049/19090 [9:17:49<01:16,  1.86s/it][gradnorm] step 19049: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1017.4

100%|█████████▉| 19050/19090 [9:17:50<01:07,  1.69s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1456.111572265625, 'learning_rate': 7.025521417264313e-10, 'recon_loss': 0.0013947795698186383, 'recon_static': 0.0004123912425711751, 'recon_motion': 0.0009823883112403565, 'recon_norm': 1.2175590634346007, 'recon_norm_std': 0.042470278311520816, 'target_norm': 1.6186196684837342, 'recon_cos': 0.6720430076122283, 'gradnorm/compressor.query': 0.3907775962320168, 'gradnorm/compressor.layer0': 15.247654969065376, 'gradnorm/compressor.layer1': 1441.6798361633557, 'gradnorm/compressor.layer2': 198.52432155783833, 'gradnorm/compressor.layer3': 44.8956268423809, 'gradnorm/compressor.layer4': 9.919408680919522, 'gradnorm/compressor.layer5': 5.205718622647005, 'gradnorm/compressor.layer6': 2.7152919481649724, 'gradnorm/compressor.layer7': 3.167070781286582, 'gradnorm/decoder.layer0': 3.4058430279832637, 'gradnorm/decoder.layer1': 1.9620929368743287, 'gradnorm/decoder.layer2': 1.1788581307451764, 'gradnorm/max_param': 1017.3702392578125, 'epoch': 9.98}

100%|█████████▉| 19050/19090 [9:17:50<01:07,  1.69s/it]
100%|█████████▉| 19051/19090 [9:17:52<01:11,  1.82s/it]
100%|█████████▉| 19052/19090 [9:17:54<01:07,  1.79s/it]
100%|█████████▉| 19053/19090 [9:17:55<01:00,  1.64s/it]
100%|█████████▉| 19054/19090 [9:17:57<01:04,  1.79s/it]
100%|█████████▉| 19055/19090 [9:17:59<01:03,  1.82s/it]
100%|█████████▉| 19056/19090 [9:18:01<00:56,  1.66s/it]
100%|█████████▉| 19057/19090 [9:18:02<00:54,  1.65s/it]
100%|█████████▉| 19058/19090 [9:18:04<00:54,  1.69s/it]
100%|█████████▉| 19059/19090 [9:18:05<00:48,  1.57s/it][gradnorm] step 19059: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1017.3

100%|█████████▉| 19060/19090 [9:18:07<00:47,  1.59s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1453.482666015625, 'learning_rate': 4.0163828829709036e-10, 'recon_loss': 0.0014016572677064687, 'recon_static': 0.0004160441814747173, 'recon_motion': 0.0009856131509877741, 'recon_norm': 1.2173679232597352, 'recon_norm_std': 0.04271439090371132, 'target_norm': 1.6212292581796646, 'recon_cos': 0.6714999064803123, 'gradnorm/compressor.query': 0.3816735393031719, 'gradnorm/compressor.layer0': 8.257315426737714, 'gradnorm/compressor.layer1': 1441.2819380676356, 'gradnorm/compressor.layer2': 181.72040250753352, 'gradnorm/compressor.layer3': 45.416800757160615, 'gradnorm/compressor.layer4': 10.009609945435717, 'gradnorm/compressor.layer5': 5.4492625026245305, 'gradnorm/compressor.layer6': 2.8552830501347355, 'gradnorm/compressor.layer7': 3.218953098444864, 'gradnorm/decoder.layer0': 3.391935877382834, 'gradnorm/decoder.layer1': 1.9827427796956367, 'gradnorm/decoder.layer2': 1.180063857112573, 'gradnorm/max_param': 1017.2646484375, 'epoch': 9.98}

100%|█████████▉| 19060/19090 [9:18:07<00:47,  1.59s/it]
100%|█████████▉| 19061/19090 [9:18:09<00:49,  1.72s/it]
100%|█████████▉| 19062/19090 [9:18:10<00:44,  1.59s/it]
100%|█████████▉| 19063/19090 [9:18:12<00:41,  1.54s/it]
100%|█████████▉| 19064/19090 [9:18:14<00:46,  1.78s/it]
100%|█████████▉| 19065/19090 [9:18:15<00:40,  1.63s/it]
100%|█████████▉| 19066/19090 [9:18:17<00:37,  1.57s/it]
100%|█████████▉| 19067/19090 [9:18:19<00:40,  1.74s/it]
100%|█████████▉| 19068/19090 [9:18:20<00:36,  1.66s/it]
100%|█████████▉| 19069/19090 [9:18:22<00:33,  1.60s/it][gradnorm] step 19069: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 1063.3

100%|█████████▉| 19070/19090 [9:18:24<00:36,  1.82s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1520.1507568359375, 'learning_rate': 1.8431086551973232e-10, 'recon_loss': 0.001396019119420089, 'recon_static': 0.0004149633379711304, 'recon_motion': 0.0009810557879973204, 'recon_norm': 1.2175392776727676, 'recon_norm_std': 0.04208761313930154, 'target_norm': 1.6193251252174377, 'recon_cos': 0.6721511945128441, 'gradnorm/compressor.query': 0.38224739092029353, 'gradnorm/compressor.layer0': 10.41069050575612, 'gradnorm/compressor.layer1': 1506.5398268880913, 'gradnorm/compressor.layer2': 197.41991058084795, 'gradnorm/compressor.layer3': 43.959154070142205, 'gradnorm/compressor.layer4': 10.725824184805537, 'gradnorm/compressor.layer5': 5.493047655016585, 'gradnorm/compressor.layer6': 2.837871638088153, 'gradnorm/compressor.layer7': 3.267382865037883, 'gradnorm/decoder.layer0': 3.4253122841690784, 'gradnorm/decoder.layer1': 1.9708062590123194, 'gradnorm/decoder.layer2': 1.1797116510811678, 'gradnorm/max_param': 1063.28564453125, 'epoch': 9.99}

100%|█████████▉| 19070/19090 [9:18:24<00:36,  1.82s/it]
100%|█████████▉| 19071/19090 [9:18:26<00:32,  1.72s/it]
100%|█████████▉| 19072/19090 [9:18:27<00:30,  1.70s/it]
100%|█████████▉| 19073/19090 [9:18:29<00:31,  1.84s/it]
100%|█████████▉| 19074/19090 [9:18:31<00:27,  1.72s/it]
100%|█████████▉| 19075/19090 [9:18:33<00:26,  1.79s/it]
100%|█████████▉| 19076/19090 [9:18:35<00:25,  1.80s/it]
100%|█████████▉| 19077/19090 [9:18:36<00:22,  1.76s/it]
100%|█████████▉| 19078/19090 [9:18:38<00:21,  1.76s/it]
100%|█████████▉| 19079/19090 [9:18:40<00:20,  1.86s/it][gradnorm] step 19079: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 980.7

100%|█████████▉| 19080/19090 [9:18:41<00:17,  1.71s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1401.9212646484375, 'learning_rate': 5.057060002700009e-11, 'recon_loss': 0.0013942130870418623, 'recon_static': 0.0004129895431105979, 'recon_motion': 0.0009812235264689661, 'recon_norm': 1.2173381507396699, 'recon_norm_std': 0.042347037233412266, 'target_norm': 1.6182693004608155, 'recon_cos': 0.6722304582595825, 'gradnorm/compressor.query': 0.36766398741511663, 'gradnorm/compressor.layer0': 12.243036276016639, 'gradnorm/compressor.layer1': 1389.700147513844, 'gradnorm/compressor.layer2': 178.56897520636107, 'gradnorm/compressor.layer3': 43.71434512521676, 'gradnorm/compressor.layer4': 10.242540948444697, 'gradnorm/compressor.layer5': 5.395673516262061, 'gradnorm/compressor.layer6': 2.6948196623981913, 'gradnorm/compressor.layer7': 3.2412174106405685, 'gradnorm/decoder.layer0': 3.404037877186993, 'gradnorm/decoder.layer1': 1.972370391378078, 'gradnorm/decoder.layer2': 1.181990945132665, 'gradnorm/max_param': 980.7415161132812, 'epoch': 9.99}

100%|█████████▉| 19080/19090 [9:18:41<00:17,  1.71s/it]
100%|█████████▉| 19081/19090 [9:18:43<00:15,  1.75s/it]
100%|█████████▉| 19082/19090 [9:18:45<00:14,  1.83s/it]
100%|█████████▉| 19083/19090 [9:18:47<00:12,  1.73s/it]
100%|█████████▉| 19084/19090 [9:18:49<00:11,  1.84s/it]
100%|█████████▉| 19085/19090 [9:18:51<00:09,  1.81s/it]
100%|█████████▉| 19086/19090 [9:18:52<00:06,  1.75s/it]
100%|█████████▉| 19087/19090 [9:18:54<00:05,  1.79s/it]
100%|█████████▉| 19088/19090 [9:18:55<00:03,  1.64s/it]
100%|█████████▉| 19089/19090 [9:18:57<00:01,  1.53s/it][gradnorm] step 19089: largest grad ‖compressor.layers.1.layer_norm1.weight‖ = 784.4

100%|██████████| 19090/19090 [9:18:58<00:00,  1.60s/it]
                                                       
{'loss': 0.0014, 'grad_norm': 1123.489013671875, 'learning_rate': 4.17938983421351e-13, 'recon_loss': 0.0013997289672261104, 'recon_static': 0.0004156430724833626, 'recon_motion': 0.0009840858721872792, 'recon_norm': 1.2170117259025575, 'recon_norm_std': 0.041822701971977946, 'target_norm': 1.619986069202423, 'recon_cos': 0.6715702548623085, 'gradnorm/compressor.query': 0.5812689634531648, 'gradnorm/compressor.layer0': 24.206573299787163, 'gradnorm/compressor.layer1': 1112.226202262831, 'gradnorm/compressor.layer2': 151.2817438279484, 'gradnorm/compressor.layer3': 39.20865937170719, 'gradnorm/compressor.layer4': 10.241156135500484, 'gradnorm/compressor.layer5': 5.6049174924931995, 'gradnorm/compressor.layer6': 2.7267420143996097, 'gradnorm/compressor.layer7': 3.1994412709262154, 'gradnorm/decoder.layer0': 3.3565507239048236, 'gradnorm/decoder.layer1': 1.9691548007000041, 'gradnorm/decoder.layer2': 1.1801183557960737, 'gradnorm/max_param': 784.449951171875, 'epoch': 10.0}

100%|██████████| 19090/19090 [9:18:59<00:00,  1.60s/it]
                                                       
{'train_runtime': 33540.852, 'train_samples_per_second': 291.389, 'train_steps_per_second': 0.569, 'train_loss': 0.005531415344016395, 'epoch': 10.0}

100%|██████████| 19090/19090 [9:19:00<00:00,  1.60s/it]
100%|██████████| 19090/19090 [9:19:00<00:00,  1.76s/it]
Compressor weights saved to work_dirs/dts_compressor_pretrain/compressor_pretrained.pt