diff --git a/results/lr_table.json b/results/lr_table.json
new file mode 100644
index 0000000000000000000000000000000000000000..3c45c9a5fcd66edbe71acfef1db23b710d7264ea
--- /dev/null
+++ b/results/lr_table.json
@@ -0,0 +1,7 @@
+{
+ "S1": 6e-4,
+ "S2": 6e-4,
+ "S3": 6e-4,
+ "S4": 6e-4,
+ "S5": 6e-4
+}
diff --git a/results/t2a/t2a_19M_C1_rand15.json b/results/t2a/t2a_19M_C1_rand15.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4f3818fb49ff3480657e9080a316b614410781b
--- /dev/null
+++ b/results/t2a/t2a_19M_C1_rand15.json
@@ -0,0 +1,225 @@
+{
+ "run_name": "t2a_19M_C1_rand15",
+ "output_dir": "/home/ubuntu/workspace/data-augmentations-for-pretraining/runs/t2a/t2a_19M_C1_rand15",
+ "params_total": 145833984,
+ "params_non_embedding": 68178432,
+ "hidden_size": 512,
+ "num_layers": 20,
+ "num_heads": 4,
+ "intermediate_size": 1536,
+ "unique_tokens_m": 19,
+ "training_files_no": 1,
+ "epochs_requested": 395,
+ "learning_rate": 0.0006,
+ "weight_decay": 0.033,
+ "seed": 42,
+ "precision": "bf16",
+ "global_batch_sequences": 512,
+ "augmentation": {
+ "l2r_percent": 100.0,
+ "max_next_i": 1,
+ "next_i_weighting": "uniform",
+ "mask_percent": 0.0,
+ "random_token_percent": 15.0,
+ "psm_percent": 0.0,
+ "spm_percent": 0.0
+ },
+ "n_train_blocks": 9536,
+ "n_eval_blocks": 2441,
+ "status": "stopped_early",
+ "min_eval_loss": 4.512935638427734,
+ "best_epoch": 41.698,
+ "best_step": 792,
+ "best_checkpoint": "/home/ubuntu/workspace/data-augmentations-for-pretraining/runs/t2a/t2a_19M_C1_rand15/checkpoint-792",
+ "epochs_completed": 117.483,
+ "stopped_early": true,
+ "wall_time_s": 1653.1,
+ "history": [
+ {
+ "epoch": 3.805,
+ "step": 72,
+ "eval_loss": 7.48493,
+ "elapsed_s": 125.9
+ },
+ {
+ "epoch": 7.591,
+ "step": 144,
+ "eval_loss": 6.80025,
+ "elapsed_s": 177.6
+ },
+ {
+ "epoch": 11.376,
+ "step": 216,
+ "eval_loss": 6.3035,
+ "elapsed_s": 233.2
+ },
+ {
+ "epoch": 15.161,
+ "step": 288,
+ "eval_loss": 5.88859,
+ "elapsed_s": 285.1
+ },
+ {
+ "epoch": 18.966,
+ "step": 360,
+ "eval_loss": 5.50239,
+ "elapsed_s": 337.2
+ },
+ {
+ "epoch": 22.752,
+ "step": 432,
+ "eval_loss": 5.24248,
+ "elapsed_s": 389.1
+ },
+ {
+ "epoch": 26.537,
+ "step": 504,
+ "eval_loss": 4.97219,
+ "elapsed_s": 441.1
+ },
+ {
+ "epoch": 30.322,
+ "step": 576,
+ "eval_loss": 4.79071,
+ "elapsed_s": 493.0
+ },
+ {
+ "epoch": 34.107,
+ "step": 648,
+ "eval_loss": 4.62867,
+ "elapsed_s": 545.0
+ },
+ {
+ "epoch": 37.913,
+ "step": 720,
+ "eval_loss": 4.54408,
+ "elapsed_s": 597.2
+ },
+ {
+ "epoch": 41.698,
+ "step": 792,
+ "eval_loss": 4.51294,
+ "elapsed_s": 649.2
+ },
+ {
+ "epoch": 45.483,
+ "step": 864,
+ "eval_loss": 4.52861,
+ "elapsed_s": 701.1
+ },
+ {
+ "epoch": 49.268,
+ "step": 936,
+ "eval_loss": 4.56255,
+ "elapsed_s": 751.0
+ },
+ {
+ "epoch": 53.054,
+ "step": 1008,
+ "eval_loss": 4.6425,
+ "elapsed_s": 800.9
+ },
+ {
+ "epoch": 56.859,
+ "step": 1080,
+ "eval_loss": 4.6819,
+ "elapsed_s": 851.0
+ },
+ {
+ "epoch": 60.644,
+ "step": 1152,
+ "eval_loss": 4.78252,
+ "elapsed_s": 901.2
+ },
+ {
+ "epoch": 64.43,
+ "step": 1224,
+ "eval_loss": 4.88789,
+ "elapsed_s": 951.1
+ },
+ {
+ "epoch": 68.215,
+ "step": 1296,
+ "eval_loss": 4.99185,
+ "elapsed_s": 1003.9
+ },
+ {
+ "epoch": 72.0,
+ "step": 1368,
+ "eval_loss": 5.04577,
+ "elapsed_s": 1053.7
+ },
+ {
+ "epoch": 75.805,
+ "step": 1440,
+ "eval_loss": 5.17304,
+ "elapsed_s": 1103.9
+ },
+ {
+ "epoch": 79.591,
+ "step": 1512,
+ "eval_loss": 5.29197,
+ "elapsed_s": 1153.7
+ },
+ {
+ "epoch": 83.376,
+ "step": 1584,
+ "eval_loss": 5.39826,
+ "elapsed_s": 1203.6
+ },
+ {
+ "epoch": 87.161,
+ "step": 1656,
+ "eval_loss": 5.5085,
+ "elapsed_s": 1253.5
+ },
+ {
+ "epoch": 90.966,
+ "step": 1728,
+ "eval_loss": 5.55253,
+ "elapsed_s": 1303.5
+ },
+ {
+ "epoch": 94.752,
+ "step": 1800,
+ "eval_loss": 5.64962,
+ "elapsed_s": 1353.5
+ },
+ {
+ "epoch": 98.537,
+ "step": 1872,
+ "eval_loss": 5.75072,
+ "elapsed_s": 1403.3
+ },
+ {
+ "epoch": 102.322,
+ "step": 1944,
+ "eval_loss": 5.85908,
+ "elapsed_s": 1453.2
+ },
+ {
+ "epoch": 106.107,
+ "step": 2016,
+ "eval_loss": 5.92747,
+ "elapsed_s": 1503.1
+ },
+ {
+ "epoch": 109.913,
+ "step": 2088,
+ "eval_loss": 5.98685,
+ "elapsed_s": 1553.1
+ },
+ {
+ "epoch": 113.698,
+ "step": 2160,
+ "eval_loss": 6.09918,
+ "elapsed_s": 1602.9
+ },
+ {
+ "epoch": 117.483,
+ "step": 2232,
+ "eval_loss": 6.2247,
+ "elapsed_s": 1653.1
+ }
+ ]
+}
\ No newline at end of file
diff --git a/results/t2a/t2a_19M_C2_r2l50_i5exp.json b/results/t2a/t2a_19M_C2_r2l50_i5exp.json
new file mode 100644
index 0000000000000000000000000000000000000000..229d5fdea071efe2124119d9ff84b8dfbdb558e9
--- /dev/null
+++ b/results/t2a/t2a_19M_C2_r2l50_i5exp.json
@@ -0,0 +1,243 @@
+{
+ "run_name": "t2a_19M_C2_r2l50_i5exp",
+ "output_dir": "/home/ubuntu/workspace/data-augmentations-for-pretraining/runs/t2a/t2a_19M_C2_r2l50_i5exp",
+ "params_total": 145836544,
+ "params_non_embedding": 68178432,
+ "hidden_size": 512,
+ "num_layers": 20,
+ "num_heads": 4,
+ "intermediate_size": 1536,
+ "unique_tokens_m": 19,
+ "training_files_no": 1,
+ "epochs_requested": 395,
+ "learning_rate": 0.0006,
+ "weight_decay": 0.033,
+ "seed": 42,
+ "precision": "bf16",
+ "global_batch_sequences": 512,
+ "augmentation": {
+ "l2r_percent": 50.0,
+ "max_next_i": 5,
+ "next_i_weighting": "exp",
+ "mask_percent": 0.0,
+ "random_token_percent": 0.0,
+ "psm_percent": 0.0,
+ "spm_percent": 0.0
+ },
+ "n_train_blocks": 9536,
+ "n_eval_blocks": 2441,
+ "status": "stopped_early",
+ "min_eval_loss": 4.377115726470947,
+ "best_epoch": 53.054,
+ "best_step": 1008,
+ "best_checkpoint": "/home/ubuntu/workspace/data-augmentations-for-pretraining/runs/t2a/t2a_19M_C2_r2l50_i5exp/checkpoint-1008",
+ "epochs_completed": 128.859,
+ "stopped_early": true,
+ "wall_time_s": 1727.4,
+ "history": [
+ {
+ "epoch": 3.805,
+ "step": 72,
+ "eval_loss": 7.51965,
+ "elapsed_s": 117.5
+ },
+ {
+ "epoch": 7.591,
+ "step": 144,
+ "eval_loss": 7.17176,
+ "elapsed_s": 166.9
+ },
+ {
+ "epoch": 11.376,
+ "step": 216,
+ "eval_loss": 6.62926,
+ "elapsed_s": 216.9
+ },
+ {
+ "epoch": 15.161,
+ "step": 288,
+ "eval_loss": 6.04692,
+ "elapsed_s": 268.7
+ },
+ {
+ "epoch": 18.966,
+ "step": 360,
+ "eval_loss": 5.69713,
+ "elapsed_s": 318.6
+ },
+ {
+ "epoch": 22.752,
+ "step": 432,
+ "eval_loss": 5.37029,
+ "elapsed_s": 368.2
+ },
+ {
+ "epoch": 26.537,
+ "step": 504,
+ "eval_loss": 5.19568,
+ "elapsed_s": 418.0
+ },
+ {
+ "epoch": 30.322,
+ "step": 576,
+ "eval_loss": 4.89342,
+ "elapsed_s": 467.7
+ },
+ {
+ "epoch": 34.107,
+ "step": 648,
+ "eval_loss": 4.7053,
+ "elapsed_s": 517.5
+ },
+ {
+ "epoch": 37.913,
+ "step": 720,
+ "eval_loss": 4.56396,
+ "elapsed_s": 567.4
+ },
+ {
+ "epoch": 41.698,
+ "step": 792,
+ "eval_loss": 4.48094,
+ "elapsed_s": 617.1
+ },
+ {
+ "epoch": 45.483,
+ "step": 864,
+ "eval_loss": 4.42799,
+ "elapsed_s": 666.8
+ },
+ {
+ "epoch": 49.268,
+ "step": 936,
+ "eval_loss": 4.47173,
+ "elapsed_s": 716.6
+ },
+ {
+ "epoch": 53.054,
+ "step": 1008,
+ "eval_loss": 4.37712,
+ "elapsed_s": 764.3
+ },
+ {
+ "epoch": 56.859,
+ "step": 1080,
+ "eval_loss": 4.38566,
+ "elapsed_s": 814.3
+ },
+ {
+ "epoch": 60.644,
+ "step": 1152,
+ "eval_loss": 4.39558,
+ "elapsed_s": 862.1
+ },
+ {
+ "epoch": 64.43,
+ "step": 1224,
+ "eval_loss": 4.42843,
+ "elapsed_s": 909.9
+ },
+ {
+ "epoch": 68.215,
+ "step": 1296,
+ "eval_loss": 4.47586,
+ "elapsed_s": 958.1
+ },
+ {
+ "epoch": 72.0,
+ "step": 1368,
+ "eval_loss": 4.48618,
+ "elapsed_s": 1008.4
+ },
+ {
+ "epoch": 75.805,
+ "step": 1440,
+ "eval_loss": 4.53592,
+ "elapsed_s": 1056.4
+ },
+ {
+ "epoch": 79.591,
+ "step": 1512,
+ "eval_loss": 4.59939,
+ "elapsed_s": 1104.3
+ },
+ {
+ "epoch": 83.376,
+ "step": 1584,
+ "eval_loss": 4.65905,
+ "elapsed_s": 1152.1
+ },
+ {
+ "epoch": 87.161,
+ "step": 1656,
+ "eval_loss": 4.71088,
+ "elapsed_s": 1200.0
+ },
+ {
+ "epoch": 90.966,
+ "step": 1728,
+ "eval_loss": 4.76819,
+ "elapsed_s": 1247.9
+ },
+ {
+ "epoch": 94.752,
+ "step": 1800,
+ "eval_loss": 4.82391,
+ "elapsed_s": 1295.7
+ },
+ {
+ "epoch": 98.537,
+ "step": 1872,
+ "eval_loss": 4.88805,
+ "elapsed_s": 1343.5
+ },
+ {
+ "epoch": 102.322,
+ "step": 1944,
+ "eval_loss": 4.96657,
+ "elapsed_s": 1391.2
+ },
+ {
+ "epoch": 106.107,
+ "step": 2016,
+ "eval_loss": 5.00135,
+ "elapsed_s": 1438.9
+ },
+ {
+ "epoch": 109.913,
+ "step": 2088,
+ "eval_loss": 5.0964,
+ "elapsed_s": 1486.8
+ },
+ {
+ "epoch": 113.698,
+ "step": 2160,
+ "eval_loss": 5.15928,
+ "elapsed_s": 1534.4
+ },
+ {
+ "epoch": 117.483,
+ "step": 2232,
+ "eval_loss": 5.21496,
+ "elapsed_s": 1582.2
+ },
+ {
+ "epoch": 121.268,
+ "step": 2304,
+ "eval_loss": 5.34861,
+ "elapsed_s": 1629.9
+ },
+ {
+ "epoch": 125.054,
+ "step": 2376,
+ "eval_loss": 5.39634,
+ "elapsed_s": 1678.4
+ },
+ {
+ "epoch": 128.859,
+ "step": 2448,
+ "eval_loss": 5.39407,
+ "elapsed_s": 1727.4
+ }
+ ]
+}
\ No newline at end of file
diff --git a/results/t2a/t2a_300M_C1_rand15.json b/results/t2a/t2a_300M_C1_rand15.json
new file mode 100644
index 0000000000000000000000000000000000000000..13cde6a062dc4d32ed6b3a7e6d7e12230a287c80
--- /dev/null
+++ b/results/t2a/t2a_300M_C1_rand15.json
@@ -0,0 +1,1785 @@
+{
+ "run_name": "t2a_300M_C1_rand15",
+ "output_dir": "/home/ubuntu/workspace/data-augmentations-for-pretraining/runs/t2a/t2a_300M_C1_rand15",
+ "params_total": 145833984,
+ "params_non_embedding": 68178432,
+ "hidden_size": 512,
+ "num_layers": 20,
+ "num_heads": 4,
+ "intermediate_size": 1536,
+ "unique_tokens_m": 300,
+ "training_files_no": 10,
+ "epochs_requested": 100,
+ "learning_rate": 0.0006,
+ "weight_decay": 0.033,
+ "seed": 42,
+ "precision": "bf16",
+ "global_batch_sequences": 512,
+ "augmentation": {
+ "l2r_percent": 100.0,
+ "max_next_i": 1,
+ "next_i_weighting": "uniform",
+ "mask_percent": 0.0,
+ "random_token_percent": 15.0,
+ "psm_percent": 0.0,
+ "spm_percent": 0.0
+ },
+ "n_train_blocks": 146474,
+ "n_eval_blocks": 2441,
+ "status": "stopped_early",
+ "min_eval_loss": 3.381537437438965,
+ "best_epoch": 67.989,
+ "best_step": 19512,
+ "best_checkpoint": "/home/ubuntu/workspace/data-augmentations-for-pretraining/runs/t2a/t2a_300M_C1_rand15/checkpoint-19512",
+ "epochs_completed": 73.003,
+ "stopped_early": true,
+ "wall_time_s": 51585.0,
+ "history": [
+ {
+ "epoch": 0.252,
+ "step": 72,
+ "eval_loss": 7.47781,
+ "elapsed_s": 209.0
+ },
+ {
+ "epoch": 0.503,
+ "step": 144,
+ "eval_loss": 6.63767,
+ "elapsed_s": 321.8
+ },
+ {
+ "epoch": 0.755,
+ "step": 216,
+ "eval_loss": 6.33,
+ "elapsed_s": 446.5
+ },
+ {
+ "epoch": 1.003,
+ "step": 288,
+ "eval_loss": 5.84163,
+ "elapsed_s": 587.5
+ },
+ {
+ "epoch": 1.255,
+ "step": 360,
+ "eval_loss": 5.47762,
+ "elapsed_s": 747.3
+ },
+ {
+ "epoch": 1.507,
+ "step": 432,
+ "eval_loss": 5.15379,
+ "elapsed_s": 908.4
+ },
+ {
+ "epoch": 1.758,
+ "step": 504,
+ "eval_loss": 4.85598,
+ "elapsed_s": 1068.6
+ },
+ {
+ "epoch": 2.007,
+ "step": 576,
+ "eval_loss": 4.66071,
+ "elapsed_s": 1226.9
+ },
+ {
+ "epoch": 2.259,
+ "step": 648,
+ "eval_loss": 4.43924,
+ "elapsed_s": 1387.6
+ },
+ {
+ "epoch": 2.51,
+ "step": 720,
+ "eval_loss": 4.29804,
+ "elapsed_s": 1547.6
+ },
+ {
+ "epoch": 2.762,
+ "step": 792,
+ "eval_loss": 4.20955,
+ "elapsed_s": 1707.2
+ },
+ {
+ "epoch": 3.01,
+ "step": 864,
+ "eval_loss": 4.14516,
+ "elapsed_s": 1865.9
+ },
+ {
+ "epoch": 3.262,
+ "step": 936,
+ "eval_loss": 4.06406,
+ "elapsed_s": 2025.2
+ },
+ {
+ "epoch": 3.514,
+ "step": 1008,
+ "eval_loss": 4.00761,
+ "elapsed_s": 2185.1
+ },
+ {
+ "epoch": 3.765,
+ "step": 1080,
+ "eval_loss": 3.9798,
+ "elapsed_s": 2346.0
+ },
+ {
+ "epoch": 4.014,
+ "step": 1152,
+ "eval_loss": 3.9535,
+ "elapsed_s": 2505.5
+ },
+ {
+ "epoch": 4.266,
+ "step": 1224,
+ "eval_loss": 3.8875,
+ "elapsed_s": 2665.9
+ },
+ {
+ "epoch": 4.517,
+ "step": 1296,
+ "eval_loss": 3.85769,
+ "elapsed_s": 2827.9
+ },
+ {
+ "epoch": 4.769,
+ "step": 1368,
+ "eval_loss": 3.83513,
+ "elapsed_s": 2989.4
+ },
+ {
+ "epoch": 5.017,
+ "step": 1440,
+ "eval_loss": 3.82523,
+ "elapsed_s": 3150.4
+ },
+ {
+ "epoch": 5.269,
+ "step": 1512,
+ "eval_loss": 3.79088,
+ "elapsed_s": 3311.6
+ },
+ {
+ "epoch": 5.521,
+ "step": 1584,
+ "eval_loss": 3.77118,
+ "elapsed_s": 3474.0
+ },
+ {
+ "epoch": 5.772,
+ "step": 1656,
+ "eval_loss": 3.74885,
+ "elapsed_s": 3636.3
+ },
+ {
+ "epoch": 6.021,
+ "step": 1728,
+ "eval_loss": 3.7574,
+ "elapsed_s": 3797.2
+ },
+ {
+ "epoch": 6.273,
+ "step": 1800,
+ "eval_loss": 3.72736,
+ "elapsed_s": 3956.8
+ },
+ {
+ "epoch": 6.524,
+ "step": 1872,
+ "eval_loss": 3.70726,
+ "elapsed_s": 4118.7
+ },
+ {
+ "epoch": 6.776,
+ "step": 1944,
+ "eval_loss": 3.68956,
+ "elapsed_s": 4281.4
+ },
+ {
+ "epoch": 7.024,
+ "step": 2016,
+ "eval_loss": 3.69514,
+ "elapsed_s": 4441.5
+ },
+ {
+ "epoch": 7.276,
+ "step": 2088,
+ "eval_loss": 3.67099,
+ "elapsed_s": 4547.9
+ },
+ {
+ "epoch": 7.528,
+ "step": 2160,
+ "eval_loss": 3.66101,
+ "elapsed_s": 4650.3
+ },
+ {
+ "epoch": 7.779,
+ "step": 2232,
+ "eval_loss": 3.64873,
+ "elapsed_s": 4753.5
+ },
+ {
+ "epoch": 8.028,
+ "step": 2304,
+ "eval_loss": 3.66089,
+ "elapsed_s": 4854.4
+ },
+ {
+ "epoch": 8.28,
+ "step": 2376,
+ "eval_loss": 3.63726,
+ "elapsed_s": 4954.4
+ },
+ {
+ "epoch": 8.531,
+ "step": 2448,
+ "eval_loss": 3.62707,
+ "elapsed_s": 5056.3
+ },
+ {
+ "epoch": 8.783,
+ "step": 2520,
+ "eval_loss": 3.61535,
+ "elapsed_s": 5158.3
+ },
+ {
+ "epoch": 9.031,
+ "step": 2592,
+ "eval_loss": 3.62973,
+ "elapsed_s": 5259.2
+ },
+ {
+ "epoch": 9.283,
+ "step": 2664,
+ "eval_loss": 3.6058,
+ "elapsed_s": 5358.9
+ },
+ {
+ "epoch": 9.535,
+ "step": 2736,
+ "eval_loss": 3.59815,
+ "elapsed_s": 5460.8
+ },
+ {
+ "epoch": 9.786,
+ "step": 2808,
+ "eval_loss": 3.58669,
+ "elapsed_s": 5562.7
+ },
+ {
+ "epoch": 10.035,
+ "step": 2880,
+ "eval_loss": 3.60731,
+ "elapsed_s": 5663.6
+ },
+ {
+ "epoch": 10.287,
+ "step": 2952,
+ "eval_loss": 3.58192,
+ "elapsed_s": 5763.4
+ },
+ {
+ "epoch": 10.538,
+ "step": 3024,
+ "eval_loss": 3.57458,
+ "elapsed_s": 5865.8
+ },
+ {
+ "epoch": 10.79,
+ "step": 3096,
+ "eval_loss": 3.56469,
+ "elapsed_s": 5967.6
+ },
+ {
+ "epoch": 11.038,
+ "step": 3168,
+ "eval_loss": 3.58305,
+ "elapsed_s": 6068.5
+ },
+ {
+ "epoch": 11.29,
+ "step": 3240,
+ "eval_loss": 3.55963,
+ "elapsed_s": 6169.4
+ },
+ {
+ "epoch": 11.542,
+ "step": 3312,
+ "eval_loss": 3.55849,
+ "elapsed_s": 6272.0
+ },
+ {
+ "epoch": 11.793,
+ "step": 3384,
+ "eval_loss": 3.54895,
+ "elapsed_s": 6373.8
+ },
+ {
+ "epoch": 12.042,
+ "step": 3456,
+ "eval_loss": 3.56101,
+ "elapsed_s": 6475.1
+ },
+ {
+ "epoch": 12.294,
+ "step": 3528,
+ "eval_loss": 3.54503,
+ "elapsed_s": 6574.8
+ },
+ {
+ "epoch": 12.545,
+ "step": 3600,
+ "eval_loss": 3.5415,
+ "elapsed_s": 6676.1
+ },
+ {
+ "epoch": 12.797,
+ "step": 3672,
+ "eval_loss": 3.5335,
+ "elapsed_s": 6777.7
+ },
+ {
+ "epoch": 13.045,
+ "step": 3744,
+ "eval_loss": 3.54538,
+ "elapsed_s": 6878.2
+ },
+ {
+ "epoch": 13.297,
+ "step": 3816,
+ "eval_loss": 3.53243,
+ "elapsed_s": 6977.4
+ },
+ {
+ "epoch": 13.549,
+ "step": 3888,
+ "eval_loss": 3.52492,
+ "elapsed_s": 7078.8
+ },
+ {
+ "epoch": 13.8,
+ "step": 3960,
+ "eval_loss": 3.52274,
+ "elapsed_s": 7180.0
+ },
+ {
+ "epoch": 14.049,
+ "step": 4032,
+ "eval_loss": 3.53394,
+ "elapsed_s": 7280.8
+ },
+ {
+ "epoch": 14.301,
+ "step": 4104,
+ "eval_loss": 3.5174,
+ "elapsed_s": 7380.1
+ },
+ {
+ "epoch": 14.552,
+ "step": 4176,
+ "eval_loss": 3.51582,
+ "elapsed_s": 7481.5
+ },
+ {
+ "epoch": 14.804,
+ "step": 4248,
+ "eval_loss": 3.51114,
+ "elapsed_s": 7582.9
+ },
+ {
+ "epoch": 15.052,
+ "step": 4320,
+ "eval_loss": 3.52365,
+ "elapsed_s": 7684.1
+ },
+ {
+ "epoch": 15.304,
+ "step": 4392,
+ "eval_loss": 3.50919,
+ "elapsed_s": 7783.3
+ },
+ {
+ "epoch": 15.556,
+ "step": 4464,
+ "eval_loss": 3.50248,
+ "elapsed_s": 7884.6
+ },
+ {
+ "epoch": 15.807,
+ "step": 4536,
+ "eval_loss": 3.4977,
+ "elapsed_s": 7986.0
+ },
+ {
+ "epoch": 16.056,
+ "step": 4608,
+ "eval_loss": 3.50976,
+ "elapsed_s": 8086.4
+ },
+ {
+ "epoch": 16.308,
+ "step": 4680,
+ "eval_loss": 3.50117,
+ "elapsed_s": 8185.3
+ },
+ {
+ "epoch": 16.559,
+ "step": 4752,
+ "eval_loss": 3.49185,
+ "elapsed_s": 8284.4
+ },
+ {
+ "epoch": 16.811,
+ "step": 4824,
+ "eval_loss": 3.49268,
+ "elapsed_s": 8385.7
+ },
+ {
+ "epoch": 17.059,
+ "step": 4896,
+ "eval_loss": 3.4979,
+ "elapsed_s": 8483.7
+ },
+ {
+ "epoch": 17.311,
+ "step": 4968,
+ "eval_loss": 3.48732,
+ "elapsed_s": 8582.9
+ },
+ {
+ "epoch": 17.563,
+ "step": 5040,
+ "eval_loss": 3.48823,
+ "elapsed_s": 8684.6
+ },
+ {
+ "epoch": 17.814,
+ "step": 5112,
+ "eval_loss": 3.48224,
+ "elapsed_s": 8783.7
+ },
+ {
+ "epoch": 18.063,
+ "step": 5184,
+ "eval_loss": 3.49706,
+ "elapsed_s": 8884.7
+ },
+ {
+ "epoch": 18.315,
+ "step": 5256,
+ "eval_loss": 3.48522,
+ "elapsed_s": 8983.9
+ },
+ {
+ "epoch": 18.566,
+ "step": 5328,
+ "eval_loss": 3.48158,
+ "elapsed_s": 9084.0
+ },
+ {
+ "epoch": 18.818,
+ "step": 5400,
+ "eval_loss": 3.47448,
+ "elapsed_s": 9185.5
+ },
+ {
+ "epoch": 19.066,
+ "step": 5472,
+ "eval_loss": 3.48555,
+ "elapsed_s": 9285.8
+ },
+ {
+ "epoch": 19.318,
+ "step": 5544,
+ "eval_loss": 3.47503,
+ "elapsed_s": 9385.1
+ },
+ {
+ "epoch": 19.57,
+ "step": 5616,
+ "eval_loss": 3.47199,
+ "elapsed_s": 9484.5
+ },
+ {
+ "epoch": 19.821,
+ "step": 5688,
+ "eval_loss": 3.46709,
+ "elapsed_s": 9586.0
+ },
+ {
+ "epoch": 20.07,
+ "step": 5760,
+ "eval_loss": 3.4801,
+ "elapsed_s": 9686.3
+ },
+ {
+ "epoch": 20.322,
+ "step": 5832,
+ "eval_loss": 3.46762,
+ "elapsed_s": 9785.4
+ },
+ {
+ "epoch": 20.573,
+ "step": 5904,
+ "eval_loss": 3.46753,
+ "elapsed_s": 9884.8
+ },
+ {
+ "epoch": 20.825,
+ "step": 5976,
+ "eval_loss": 3.4618,
+ "elapsed_s": 9984.0
+ },
+ {
+ "epoch": 21.073,
+ "step": 6048,
+ "eval_loss": 3.47328,
+ "elapsed_s": 10084.6
+ },
+ {
+ "epoch": 21.325,
+ "step": 6120,
+ "eval_loss": 3.46523,
+ "elapsed_s": 10184.2
+ },
+ {
+ "epoch": 21.577,
+ "step": 6192,
+ "eval_loss": 3.45943,
+ "elapsed_s": 10283.7
+ },
+ {
+ "epoch": 21.828,
+ "step": 6264,
+ "eval_loss": 3.45733,
+ "elapsed_s": 10385.1
+ },
+ {
+ "epoch": 22.077,
+ "step": 6336,
+ "eval_loss": 3.46784,
+ "elapsed_s": 10485.6
+ },
+ {
+ "epoch": 22.329,
+ "step": 6408,
+ "eval_loss": 3.45975,
+ "elapsed_s": 10606.1
+ },
+ {
+ "epoch": 22.58,
+ "step": 6480,
+ "eval_loss": 3.46007,
+ "elapsed_s": 10763.2
+ },
+ {
+ "epoch": 22.832,
+ "step": 6552,
+ "eval_loss": 3.45163,
+ "elapsed_s": 10919.6
+ },
+ {
+ "epoch": 23.08,
+ "step": 6624,
+ "eval_loss": 3.46103,
+ "elapsed_s": 11077.1
+ },
+ {
+ "epoch": 23.332,
+ "step": 6696,
+ "eval_loss": 3.45698,
+ "elapsed_s": 11234.1
+ },
+ {
+ "epoch": 23.584,
+ "step": 6768,
+ "eval_loss": 3.45056,
+ "elapsed_s": 11390.3
+ },
+ {
+ "epoch": 23.835,
+ "step": 6840,
+ "eval_loss": 3.44924,
+ "elapsed_s": 11549.5
+ },
+ {
+ "epoch": 24.084,
+ "step": 6912,
+ "eval_loss": 3.45593,
+ "elapsed_s": 11706.1
+ },
+ {
+ "epoch": 24.336,
+ "step": 6984,
+ "eval_loss": 3.45231,
+ "elapsed_s": 11863.3
+ },
+ {
+ "epoch": 24.587,
+ "step": 7056,
+ "eval_loss": 3.4468,
+ "elapsed_s": 12019.2
+ },
+ {
+ "epoch": 24.839,
+ "step": 7128,
+ "eval_loss": 3.44177,
+ "elapsed_s": 12178.4
+ },
+ {
+ "epoch": 25.087,
+ "step": 7200,
+ "eval_loss": 3.45587,
+ "elapsed_s": 12599.6
+ },
+ {
+ "epoch": 25.339,
+ "step": 7272,
+ "eval_loss": 3.44759,
+ "elapsed_s": 12804.3
+ },
+ {
+ "epoch": 25.591,
+ "step": 7344,
+ "eval_loss": 3.44228,
+ "elapsed_s": 13009.2
+ },
+ {
+ "epoch": 25.842,
+ "step": 7416,
+ "eval_loss": 3.43997,
+ "elapsed_s": 13214.5
+ },
+ {
+ "epoch": 26.091,
+ "step": 7488,
+ "eval_loss": 3.45062,
+ "elapsed_s": 13419.3
+ },
+ {
+ "epoch": 26.343,
+ "step": 7560,
+ "eval_loss": 3.4448,
+ "elapsed_s": 13623.1
+ },
+ {
+ "epoch": 26.594,
+ "step": 7632,
+ "eval_loss": 3.43839,
+ "elapsed_s": 13827.2
+ },
+ {
+ "epoch": 26.846,
+ "step": 7704,
+ "eval_loss": 3.43419,
+ "elapsed_s": 14033.9
+ },
+ {
+ "epoch": 27.094,
+ "step": 7776,
+ "eval_loss": 3.4501,
+ "elapsed_s": 14238.5
+ },
+ {
+ "epoch": 27.346,
+ "step": 7848,
+ "eval_loss": 3.43994,
+ "elapsed_s": 14442.4
+ },
+ {
+ "epoch": 27.598,
+ "step": 7920,
+ "eval_loss": 3.4365,
+ "elapsed_s": 14646.4
+ },
+ {
+ "epoch": 27.849,
+ "step": 7992,
+ "eval_loss": 3.43381,
+ "elapsed_s": 14850.3
+ },
+ {
+ "epoch": 28.098,
+ "step": 8064,
+ "eval_loss": 3.44481,
+ "elapsed_s": 15054.9
+ },
+ {
+ "epoch": 28.35,
+ "step": 8136,
+ "eval_loss": 3.43708,
+ "elapsed_s": 15259.3
+ },
+ {
+ "epoch": 28.601,
+ "step": 8208,
+ "eval_loss": 3.43279,
+ "elapsed_s": 15463.3
+ },
+ {
+ "epoch": 28.853,
+ "step": 8280,
+ "eval_loss": 3.42868,
+ "elapsed_s": 15669.8
+ },
+ {
+ "epoch": 29.101,
+ "step": 8352,
+ "eval_loss": 3.44357,
+ "elapsed_s": 15874.7
+ },
+ {
+ "epoch": 29.353,
+ "step": 8424,
+ "eval_loss": 3.43377,
+ "elapsed_s": 16080.1
+ },
+ {
+ "epoch": 29.605,
+ "step": 8496,
+ "eval_loss": 3.43162,
+ "elapsed_s": 16284.1
+ },
+ {
+ "epoch": 29.856,
+ "step": 8568,
+ "eval_loss": 3.42739,
+ "elapsed_s": 16488.4
+ },
+ {
+ "epoch": 30.105,
+ "step": 8640,
+ "eval_loss": 3.44035,
+ "elapsed_s": 16692.5
+ },
+ {
+ "epoch": 30.357,
+ "step": 8712,
+ "eval_loss": 3.43171,
+ "elapsed_s": 16897.2
+ },
+ {
+ "epoch": 30.608,
+ "step": 8784,
+ "eval_loss": 3.42564,
+ "elapsed_s": 17101.2
+ },
+ {
+ "epoch": 30.86,
+ "step": 8856,
+ "eval_loss": 3.42279,
+ "elapsed_s": 17307.4
+ },
+ {
+ "epoch": 31.108,
+ "step": 8928,
+ "eval_loss": 3.43693,
+ "elapsed_s": 17511.1
+ },
+ {
+ "epoch": 31.36,
+ "step": 9000,
+ "eval_loss": 3.42999,
+ "elapsed_s": 17715.9
+ },
+ {
+ "epoch": 31.612,
+ "step": 9072,
+ "eval_loss": 3.42445,
+ "elapsed_s": 17920.8
+ },
+ {
+ "epoch": 31.863,
+ "step": 9144,
+ "eval_loss": 3.41898,
+ "elapsed_s": 18124.7
+ },
+ {
+ "epoch": 32.112,
+ "step": 9216,
+ "eval_loss": 3.43611,
+ "elapsed_s": 18329.0
+ },
+ {
+ "epoch": 32.364,
+ "step": 9288,
+ "eval_loss": 3.43069,
+ "elapsed_s": 18533.7
+ },
+ {
+ "epoch": 32.615,
+ "step": 9360,
+ "eval_loss": 3.42467,
+ "elapsed_s": 18738.3
+ },
+ {
+ "epoch": 32.867,
+ "step": 9432,
+ "eval_loss": 3.41992,
+ "elapsed_s": 18942.2
+ },
+ {
+ "epoch": 33.115,
+ "step": 9504,
+ "eval_loss": 3.43255,
+ "elapsed_s": 19143.9
+ },
+ {
+ "epoch": 33.367,
+ "step": 9576,
+ "eval_loss": 3.42445,
+ "elapsed_s": 19347.8
+ },
+ {
+ "epoch": 33.619,
+ "step": 9648,
+ "eval_loss": 3.42137,
+ "elapsed_s": 19552.1
+ },
+ {
+ "epoch": 33.87,
+ "step": 9720,
+ "eval_loss": 3.41663,
+ "elapsed_s": 19757.1
+ },
+ {
+ "epoch": 34.119,
+ "step": 9792,
+ "eval_loss": 3.43098,
+ "elapsed_s": 19961.2
+ },
+ {
+ "epoch": 34.371,
+ "step": 9864,
+ "eval_loss": 3.42123,
+ "elapsed_s": 20165.1
+ },
+ {
+ "epoch": 34.622,
+ "step": 9936,
+ "eval_loss": 3.41772,
+ "elapsed_s": 20369.5
+ },
+ {
+ "epoch": 34.874,
+ "step": 10008,
+ "eval_loss": 3.41466,
+ "elapsed_s": 20573.8
+ },
+ {
+ "epoch": 35.122,
+ "step": 10080,
+ "eval_loss": 3.42782,
+ "elapsed_s": 20777.9
+ },
+ {
+ "epoch": 35.374,
+ "step": 10152,
+ "eval_loss": 3.41813,
+ "elapsed_s": 20982.3
+ },
+ {
+ "epoch": 35.626,
+ "step": 10224,
+ "eval_loss": 3.41362,
+ "elapsed_s": 21186.6
+ },
+ {
+ "epoch": 35.877,
+ "step": 10296,
+ "eval_loss": 3.41372,
+ "elapsed_s": 21393.2
+ },
+ {
+ "epoch": 36.126,
+ "step": 10368,
+ "eval_loss": 3.4253,
+ "elapsed_s": 21595.3
+ },
+ {
+ "epoch": 36.377,
+ "step": 10440,
+ "eval_loss": 3.41996,
+ "elapsed_s": 21799.3
+ },
+ {
+ "epoch": 36.629,
+ "step": 10512,
+ "eval_loss": 3.41363,
+ "elapsed_s": 22005.0
+ },
+ {
+ "epoch": 36.881,
+ "step": 10584,
+ "eval_loss": 3.41426,
+ "elapsed_s": 22209.4
+ },
+ {
+ "epoch": 37.129,
+ "step": 10656,
+ "eval_loss": 3.4261,
+ "elapsed_s": 22411.4
+ },
+ {
+ "epoch": 37.381,
+ "step": 10728,
+ "eval_loss": 3.41384,
+ "elapsed_s": 22616.1
+ },
+ {
+ "epoch": 37.633,
+ "step": 10800,
+ "eval_loss": 3.4152,
+ "elapsed_s": 22821.1
+ },
+ {
+ "epoch": 37.884,
+ "step": 10872,
+ "eval_loss": 3.40925,
+ "elapsed_s": 23027.6
+ },
+ {
+ "epoch": 38.133,
+ "step": 10944,
+ "eval_loss": 3.41945,
+ "elapsed_s": 23231.8
+ },
+ {
+ "epoch": 38.384,
+ "step": 11016,
+ "eval_loss": 3.41429,
+ "elapsed_s": 23437.0
+ },
+ {
+ "epoch": 38.636,
+ "step": 11088,
+ "eval_loss": 3.40793,
+ "elapsed_s": 23641.1
+ },
+ {
+ "epoch": 38.888,
+ "step": 11160,
+ "eval_loss": 3.40817,
+ "elapsed_s": 23847.7
+ },
+ {
+ "epoch": 39.136,
+ "step": 11232,
+ "eval_loss": 3.41789,
+ "elapsed_s": 24049.5
+ },
+ {
+ "epoch": 39.388,
+ "step": 11304,
+ "eval_loss": 3.41123,
+ "elapsed_s": 24253.3
+ },
+ {
+ "epoch": 39.64,
+ "step": 11376,
+ "eval_loss": 3.40937,
+ "elapsed_s": 24457.3
+ },
+ {
+ "epoch": 39.891,
+ "step": 11448,
+ "eval_loss": 3.4062,
+ "elapsed_s": 24661.8
+ },
+ {
+ "epoch": 40.14,
+ "step": 11520,
+ "eval_loss": 3.42085,
+ "elapsed_s": 24866.7
+ },
+ {
+ "epoch": 40.391,
+ "step": 11592,
+ "eval_loss": 3.41282,
+ "elapsed_s": 25071.0
+ },
+ {
+ "epoch": 40.643,
+ "step": 11664,
+ "eval_loss": 3.4083,
+ "elapsed_s": 25275.2
+ },
+ {
+ "epoch": 40.895,
+ "step": 11736,
+ "eval_loss": 3.40527,
+ "elapsed_s": 25479.2
+ },
+ {
+ "epoch": 41.143,
+ "step": 11808,
+ "eval_loss": 3.4167,
+ "elapsed_s": 25683.3
+ },
+ {
+ "epoch": 41.395,
+ "step": 11880,
+ "eval_loss": 3.41097,
+ "elapsed_s": 25888.4
+ },
+ {
+ "epoch": 41.647,
+ "step": 11952,
+ "eval_loss": 3.40369,
+ "elapsed_s": 26092.4
+ },
+ {
+ "epoch": 41.898,
+ "step": 12024,
+ "eval_loss": 3.40498,
+ "elapsed_s": 26298.7
+ },
+ {
+ "epoch": 42.147,
+ "step": 12096,
+ "eval_loss": 3.41666,
+ "elapsed_s": 26500.9
+ },
+ {
+ "epoch": 42.398,
+ "step": 12168,
+ "eval_loss": 3.41194,
+ "elapsed_s": 26705.9
+ },
+ {
+ "epoch": 42.65,
+ "step": 12240,
+ "eval_loss": 3.40639,
+ "elapsed_s": 26909.7
+ },
+ {
+ "epoch": 42.902,
+ "step": 12312,
+ "eval_loss": 3.40208,
+ "elapsed_s": 27113.5
+ },
+ {
+ "epoch": 43.15,
+ "step": 12384,
+ "eval_loss": 3.41187,
+ "elapsed_s": 27317.7
+ },
+ {
+ "epoch": 43.402,
+ "step": 12456,
+ "eval_loss": 3.41038,
+ "elapsed_s": 27521.5
+ },
+ {
+ "epoch": 43.654,
+ "step": 12528,
+ "eval_loss": 3.40919,
+ "elapsed_s": 27725.7
+ },
+ {
+ "epoch": 43.905,
+ "step": 12600,
+ "eval_loss": 3.4016,
+ "elapsed_s": 27929.5
+ },
+ {
+ "epoch": 44.154,
+ "step": 12672,
+ "eval_loss": 3.41388,
+ "elapsed_s": 28133.5
+ },
+ {
+ "epoch": 44.405,
+ "step": 12744,
+ "eval_loss": 3.40858,
+ "elapsed_s": 28337.7
+ },
+ {
+ "epoch": 44.657,
+ "step": 12816,
+ "eval_loss": 3.40521,
+ "elapsed_s": 28542.1
+ },
+ {
+ "epoch": 44.909,
+ "step": 12888,
+ "eval_loss": 3.40003,
+ "elapsed_s": 28746.0
+ },
+ {
+ "epoch": 45.157,
+ "step": 12960,
+ "eval_loss": 3.4107,
+ "elapsed_s": 28949.8
+ },
+ {
+ "epoch": 45.409,
+ "step": 13032,
+ "eval_loss": 3.40672,
+ "elapsed_s": 29154.2
+ },
+ {
+ "epoch": 45.661,
+ "step": 13104,
+ "eval_loss": 3.40362,
+ "elapsed_s": 29358.1
+ },
+ {
+ "epoch": 45.912,
+ "step": 13176,
+ "eval_loss": 3.39857,
+ "elapsed_s": 29562.4
+ },
+ {
+ "epoch": 46.161,
+ "step": 13248,
+ "eval_loss": 3.4102,
+ "elapsed_s": 29766.4
+ },
+ {
+ "epoch": 46.412,
+ "step": 13320,
+ "eval_loss": 3.40725,
+ "elapsed_s": 29970.8
+ },
+ {
+ "epoch": 46.664,
+ "step": 13392,
+ "eval_loss": 3.4055,
+ "elapsed_s": 30175.6
+ },
+ {
+ "epoch": 46.916,
+ "step": 13464,
+ "eval_loss": 3.39912,
+ "elapsed_s": 30379.9
+ },
+ {
+ "epoch": 47.164,
+ "step": 13536,
+ "eval_loss": 3.41161,
+ "elapsed_s": 30581.6
+ },
+ {
+ "epoch": 47.416,
+ "step": 13608,
+ "eval_loss": 3.40272,
+ "elapsed_s": 30786.1
+ },
+ {
+ "epoch": 47.668,
+ "step": 13680,
+ "eval_loss": 3.39932,
+ "elapsed_s": 30990.5
+ },
+ {
+ "epoch": 47.919,
+ "step": 13752,
+ "eval_loss": 3.39348,
+ "elapsed_s": 31195.0
+ },
+ {
+ "epoch": 48.168,
+ "step": 13824,
+ "eval_loss": 3.4065,
+ "elapsed_s": 31399.5
+ },
+ {
+ "epoch": 48.419,
+ "step": 13896,
+ "eval_loss": 3.4034,
+ "elapsed_s": 31603.4
+ },
+ {
+ "epoch": 48.671,
+ "step": 13968,
+ "eval_loss": 3.39744,
+ "elapsed_s": 31808.9
+ },
+ {
+ "epoch": 48.923,
+ "step": 14040,
+ "eval_loss": 3.39534,
+ "elapsed_s": 32012.8
+ },
+ {
+ "epoch": 49.171,
+ "step": 14112,
+ "eval_loss": 3.40833,
+ "elapsed_s": 32214.7
+ },
+ {
+ "epoch": 49.423,
+ "step": 14184,
+ "eval_loss": 3.40238,
+ "elapsed_s": 32418.6
+ },
+ {
+ "epoch": 49.675,
+ "step": 14256,
+ "eval_loss": 3.3964,
+ "elapsed_s": 32622.9
+ },
+ {
+ "epoch": 49.926,
+ "step": 14328,
+ "eval_loss": 3.39217,
+ "elapsed_s": 32826.7
+ },
+ {
+ "epoch": 50.175,
+ "step": 14400,
+ "eval_loss": 3.40406,
+ "elapsed_s": 33030.9
+ },
+ {
+ "epoch": 50.426,
+ "step": 14472,
+ "eval_loss": 3.40228,
+ "elapsed_s": 33235.1
+ },
+ {
+ "epoch": 50.678,
+ "step": 14544,
+ "eval_loss": 3.39827,
+ "elapsed_s": 33439.4
+ },
+ {
+ "epoch": 50.93,
+ "step": 14616,
+ "eval_loss": 3.39445,
+ "elapsed_s": 33643.1
+ },
+ {
+ "epoch": 51.178,
+ "step": 14688,
+ "eval_loss": 3.4077,
+ "elapsed_s": 33844.7
+ },
+ {
+ "epoch": 51.43,
+ "step": 14760,
+ "eval_loss": 3.40413,
+ "elapsed_s": 34048.6
+ },
+ {
+ "epoch": 51.682,
+ "step": 14832,
+ "eval_loss": 3.39621,
+ "elapsed_s": 34252.9
+ },
+ {
+ "epoch": 51.933,
+ "step": 14904,
+ "eval_loss": 3.3936,
+ "elapsed_s": 34456.7
+ },
+ {
+ "epoch": 52.182,
+ "step": 14976,
+ "eval_loss": 3.40605,
+ "elapsed_s": 34658.2
+ },
+ {
+ "epoch": 52.433,
+ "step": 15048,
+ "eval_loss": 3.40076,
+ "elapsed_s": 34862.4
+ },
+ {
+ "epoch": 52.685,
+ "step": 15120,
+ "eval_loss": 3.39669,
+ "elapsed_s": 35066.4
+ },
+ {
+ "epoch": 52.937,
+ "step": 15192,
+ "eval_loss": 3.39423,
+ "elapsed_s": 35270.7
+ },
+ {
+ "epoch": 53.185,
+ "step": 15264,
+ "eval_loss": 3.40449,
+ "elapsed_s": 35472.4
+ },
+ {
+ "epoch": 53.437,
+ "step": 15336,
+ "eval_loss": 3.40282,
+ "elapsed_s": 35677.0
+ },
+ {
+ "epoch": 53.689,
+ "step": 15408,
+ "eval_loss": 3.3966,
+ "elapsed_s": 35881.0
+ },
+ {
+ "epoch": 53.94,
+ "step": 15480,
+ "eval_loss": 3.39523,
+ "elapsed_s": 36085.2
+ },
+ {
+ "epoch": 54.189,
+ "step": 15552,
+ "eval_loss": 3.40414,
+ "elapsed_s": 36286.8
+ },
+ {
+ "epoch": 54.44,
+ "step": 15624,
+ "eval_loss": 3.39815,
+ "elapsed_s": 36491.0
+ },
+ {
+ "epoch": 54.692,
+ "step": 15696,
+ "eval_loss": 3.39718,
+ "elapsed_s": 36695.4
+ },
+ {
+ "epoch": 54.944,
+ "step": 15768,
+ "eval_loss": 3.3904,
+ "elapsed_s": 36899.8
+ },
+ {
+ "epoch": 55.192,
+ "step": 15840,
+ "eval_loss": 3.40362,
+ "elapsed_s": 37103.8
+ },
+ {
+ "epoch": 55.444,
+ "step": 15912,
+ "eval_loss": 3.39939,
+ "elapsed_s": 37308.0
+ },
+ {
+ "epoch": 55.696,
+ "step": 15984,
+ "eval_loss": 3.39295,
+ "elapsed_s": 37512.5
+ },
+ {
+ "epoch": 55.947,
+ "step": 16056,
+ "eval_loss": 3.39076,
+ "elapsed_s": 37716.7
+ },
+ {
+ "epoch": 56.196,
+ "step": 16128,
+ "eval_loss": 3.4013,
+ "elapsed_s": 37918.8
+ },
+ {
+ "epoch": 56.447,
+ "step": 16200,
+ "eval_loss": 3.39857,
+ "elapsed_s": 38122.8
+ },
+ {
+ "epoch": 56.699,
+ "step": 16272,
+ "eval_loss": 3.39402,
+ "elapsed_s": 38327.4
+ },
+ {
+ "epoch": 56.951,
+ "step": 16344,
+ "eval_loss": 3.38902,
+ "elapsed_s": 38531.4
+ },
+ {
+ "epoch": 57.199,
+ "step": 16416,
+ "eval_loss": 3.40536,
+ "elapsed_s": 38735.8
+ },
+ {
+ "epoch": 57.451,
+ "step": 16488,
+ "eval_loss": 3.39564,
+ "elapsed_s": 38939.8
+ },
+ {
+ "epoch": 57.703,
+ "step": 16560,
+ "eval_loss": 3.395,
+ "elapsed_s": 39144.3
+ },
+ {
+ "epoch": 57.954,
+ "step": 16632,
+ "eval_loss": 3.38877,
+ "elapsed_s": 39348.7
+ },
+ {
+ "epoch": 58.203,
+ "step": 16704,
+ "eval_loss": 3.40047,
+ "elapsed_s": 39553.3
+ },
+ {
+ "epoch": 58.454,
+ "step": 16776,
+ "eval_loss": 3.39547,
+ "elapsed_s": 39757.5
+ },
+ {
+ "epoch": 58.706,
+ "step": 16848,
+ "eval_loss": 3.39209,
+ "elapsed_s": 39961.9
+ },
+ {
+ "epoch": 58.958,
+ "step": 16920,
+ "eval_loss": 3.38893,
+ "elapsed_s": 40166.4
+ },
+ {
+ "epoch": 59.206,
+ "step": 16992,
+ "eval_loss": 3.3971,
+ "elapsed_s": 40368.1
+ },
+ {
+ "epoch": 59.458,
+ "step": 17064,
+ "eval_loss": 3.39985,
+ "elapsed_s": 40572.7
+ },
+ {
+ "epoch": 59.71,
+ "step": 17136,
+ "eval_loss": 3.39435,
+ "elapsed_s": 40777.0
+ },
+ {
+ "epoch": 59.961,
+ "step": 17208,
+ "eval_loss": 3.38534,
+ "elapsed_s": 40981.4
+ },
+ {
+ "epoch": 60.21,
+ "step": 17280,
+ "eval_loss": 3.39778,
+ "elapsed_s": 41185.5
+ },
+ {
+ "epoch": 60.461,
+ "step": 17352,
+ "eval_loss": 3.40129,
+ "elapsed_s": 41389.9
+ },
+ {
+ "epoch": 60.713,
+ "step": 17424,
+ "eval_loss": 3.39185,
+ "elapsed_s": 41593.9
+ },
+ {
+ "epoch": 60.965,
+ "step": 17496,
+ "eval_loss": 3.39002,
+ "elapsed_s": 41798.3
+ },
+ {
+ "epoch": 61.213,
+ "step": 17568,
+ "eval_loss": 3.3996,
+ "elapsed_s": 41999.9
+ },
+ {
+ "epoch": 61.465,
+ "step": 17640,
+ "eval_loss": 3.39641,
+ "elapsed_s": 42204.3
+ },
+ {
+ "epoch": 61.717,
+ "step": 17712,
+ "eval_loss": 3.38737,
+ "elapsed_s": 42408.3
+ },
+ {
+ "epoch": 61.968,
+ "step": 17784,
+ "eval_loss": 3.3878,
+ "elapsed_s": 42612.3
+ },
+ {
+ "epoch": 62.217,
+ "step": 17856,
+ "eval_loss": 3.39947,
+ "elapsed_s": 42814.3
+ },
+ {
+ "epoch": 62.468,
+ "step": 17928,
+ "eval_loss": 3.39234,
+ "elapsed_s": 43018.3
+ },
+ {
+ "epoch": 62.72,
+ "step": 18000,
+ "eval_loss": 3.39504,
+ "elapsed_s": 43222.9
+ },
+ {
+ "epoch": 62.972,
+ "step": 18072,
+ "eval_loss": 3.38606,
+ "elapsed_s": 43426.9
+ },
+ {
+ "epoch": 63.22,
+ "step": 18144,
+ "eval_loss": 3.3981,
+ "elapsed_s": 43628.9
+ },
+ {
+ "epoch": 63.472,
+ "step": 18216,
+ "eval_loss": 3.39377,
+ "elapsed_s": 43832.9
+ },
+ {
+ "epoch": 63.724,
+ "step": 18288,
+ "eval_loss": 3.38904,
+ "elapsed_s": 44037.3
+ },
+ {
+ "epoch": 63.975,
+ "step": 18360,
+ "eval_loss": 3.38702,
+ "elapsed_s": 44241.3
+ },
+ {
+ "epoch": 64.224,
+ "step": 18432,
+ "eval_loss": 3.40148,
+ "elapsed_s": 44443.4
+ },
+ {
+ "epoch": 64.475,
+ "step": 18504,
+ "eval_loss": 3.39358,
+ "elapsed_s": 44647.4
+ },
+ {
+ "epoch": 64.727,
+ "step": 18576,
+ "eval_loss": 3.38772,
+ "elapsed_s": 44852.0
+ },
+ {
+ "epoch": 64.979,
+ "step": 18648,
+ "eval_loss": 3.38474,
+ "elapsed_s": 45056.5
+ },
+ {
+ "epoch": 65.227,
+ "step": 18720,
+ "eval_loss": 3.39815,
+ "elapsed_s": 45260.8
+ },
+ {
+ "epoch": 65.479,
+ "step": 18792,
+ "eval_loss": 3.39347,
+ "elapsed_s": 45465.6
+ },
+ {
+ "epoch": 65.731,
+ "step": 18864,
+ "eval_loss": 3.38993,
+ "elapsed_s": 45670.3
+ },
+ {
+ "epoch": 65.982,
+ "step": 18936,
+ "eval_loss": 3.38564,
+ "elapsed_s": 45874.9
+ },
+ {
+ "epoch": 66.231,
+ "step": 19008,
+ "eval_loss": 3.39986,
+ "elapsed_s": 46076.8
+ },
+ {
+ "epoch": 66.482,
+ "step": 19080,
+ "eval_loss": 3.39348,
+ "elapsed_s": 46281.3
+ },
+ {
+ "epoch": 66.734,
+ "step": 19152,
+ "eval_loss": 3.38795,
+ "elapsed_s": 46485.3
+ },
+ {
+ "epoch": 66.986,
+ "step": 19224,
+ "eval_loss": 3.38298,
+ "elapsed_s": 46689.8
+ },
+ {
+ "epoch": 67.234,
+ "step": 19296,
+ "eval_loss": 3.39488,
+ "elapsed_s": 46894.7
+ },
+ {
+ "epoch": 67.486,
+ "step": 19368,
+ "eval_loss": 3.38974,
+ "elapsed_s": 47098.8
+ },
+ {
+ "epoch": 67.738,
+ "step": 19440,
+ "eval_loss": 3.38659,
+ "elapsed_s": 47303.9
+ },
+ {
+ "epoch": 67.989,
+ "step": 19512,
+ "eval_loss": 3.38154,
+ "elapsed_s": 47508.4
+ },
+ {
+ "epoch": 68.238,
+ "step": 19584,
+ "eval_loss": 3.40078,
+ "elapsed_s": 47712.4
+ },
+ {
+ "epoch": 68.489,
+ "step": 19656,
+ "eval_loss": 3.39331,
+ "elapsed_s": 47916.5
+ },
+ {
+ "epoch": 68.741,
+ "step": 19728,
+ "eval_loss": 3.38707,
+ "elapsed_s": 48121.0
+ },
+ {
+ "epoch": 68.993,
+ "step": 19800,
+ "eval_loss": 3.38532,
+ "elapsed_s": 48325.5
+ },
+ {
+ "epoch": 69.241,
+ "step": 19872,
+ "eval_loss": 3.39663,
+ "elapsed_s": 48527.3
+ },
+ {
+ "epoch": 69.493,
+ "step": 19944,
+ "eval_loss": 3.39187,
+ "elapsed_s": 48731.6
+ },
+ {
+ "epoch": 69.745,
+ "step": 20016,
+ "eval_loss": 3.39039,
+ "elapsed_s": 48936.2
+ },
+ {
+ "epoch": 69.996,
+ "step": 20088,
+ "eval_loss": 3.38487,
+ "elapsed_s": 49140.4
+ },
+ {
+ "epoch": 70.245,
+ "step": 20160,
+ "eval_loss": 3.39715,
+ "elapsed_s": 49342.7
+ },
+ {
+ "epoch": 70.496,
+ "step": 20232,
+ "eval_loss": 3.39364,
+ "elapsed_s": 49546.9
+ },
+ {
+ "epoch": 70.748,
+ "step": 20304,
+ "eval_loss": 3.38664,
+ "elapsed_s": 49751.2
+ },
+ {
+ "epoch": 71.0,
+ "step": 20376,
+ "eval_loss": 3.38608,
+ "elapsed_s": 49956.0
+ },
+ {
+ "epoch": 71.248,
+ "step": 20448,
+ "eval_loss": 3.39906,
+ "elapsed_s": 50158.3
+ },
+ {
+ "epoch": 71.5,
+ "step": 20520,
+ "eval_loss": 3.39788,
+ "elapsed_s": 50362.7
+ },
+ {
+ "epoch": 71.752,
+ "step": 20592,
+ "eval_loss": 3.38562,
+ "elapsed_s": 50567.0
+ },
+ {
+ "epoch": 72.0,
+ "step": 20664,
+ "eval_loss": 3.39249,
+ "elapsed_s": 50769.4
+ },
+ {
+ "epoch": 72.252,
+ "step": 20736,
+ "eval_loss": 3.39337,
+ "elapsed_s": 50974.0
+ },
+ {
+ "epoch": 72.503,
+ "step": 20808,
+ "eval_loss": 3.39345,
+ "elapsed_s": 51178.4
+ },
+ {
+ "epoch": 72.755,
+ "step": 20880,
+ "eval_loss": 3.38645,
+ "elapsed_s": 51382.8
+ },
+ {
+ "epoch": 73.003,
+ "step": 20952,
+ "eval_loss": 3.41251,
+ "elapsed_s": 51585.0
+ }
+ ]
+}
\ No newline at end of file
diff --git a/results/t2a/t2a_300M_C3_rand5_r2l50_i5exp.json b/results/t2a/t2a_300M_C3_rand5_r2l50_i5exp.json
new file mode 100644
index 0000000000000000000000000000000000000000..6a2d64a22c6bba6597fa365a751587115e2c8471
--- /dev/null
+++ b/results/t2a/t2a_300M_C3_rand5_r2l50_i5exp.json
@@ -0,0 +1,2433 @@
+{
+ "run_name": "t2a_300M_C3_rand5_r2l50_i5exp",
+ "output_dir": "/home/ubuntu/workspace/data-augmentations-for-pretraining/runs/t2a/t2a_300M_C3_rand5_r2l50_i5exp",
+ "params_total": 145836544,
+ "params_non_embedding": 68178432,
+ "hidden_size": 512,
+ "num_layers": 20,
+ "num_heads": 4,
+ "intermediate_size": 1536,
+ "unique_tokens_m": 300,
+ "training_files_no": 10,
+ "epochs_requested": 100,
+ "learning_rate": 0.0006,
+ "weight_decay": 0.033,
+ "seed": 42,
+ "precision": "bf16",
+ "global_batch_sequences": 512,
+ "augmentation": {
+ "l2r_percent": 50.0,
+ "max_next_i": 5,
+ "next_i_weighting": "exp",
+ "mask_percent": 0.0,
+ "random_token_percent": 5.0,
+ "psm_percent": 0.0,
+ "spm_percent": 0.0
+ },
+ "n_train_blocks": 146474,
+ "n_eval_blocks": 2441,
+ "status": "completed",
+ "min_eval_loss": 3.4511053562164307,
+ "best_epoch": 97.842,
+ "best_step": 28080,
+ "best_checkpoint": "/home/ubuntu/workspace/data-augmentations-for-pretraining/runs/t2a/t2a_300M_C3_rand5_r2l50_i5exp/checkpoint-28080",
+ "epochs_completed": 100.0,
+ "stopped_early": false,
+ "wall_time_s": 122736.4,
+ "history": [
+ {
+ "epoch": 0.252,
+ "step": 72,
+ "eval_loss": 7.51969,
+ "elapsed_s": 180.3
+ },
+ {
+ "epoch": 0.503,
+ "step": 144,
+ "eval_loss": 7.17048,
+ "elapsed_s": 291.7
+ },
+ {
+ "epoch": 0.755,
+ "step": 216,
+ "eval_loss": 6.59654,
+ "elapsed_s": 404.8
+ },
+ {
+ "epoch": 1.003,
+ "step": 288,
+ "eval_loss": 6.04802,
+ "elapsed_s": 514.5
+ },
+ {
+ "epoch": 1.255,
+ "step": 360,
+ "eval_loss": 5.72007,
+ "elapsed_s": 625.9
+ },
+ {
+ "epoch": 1.507,
+ "step": 432,
+ "eval_loss": 5.36466,
+ "elapsed_s": 736.8
+ },
+ {
+ "epoch": 1.758,
+ "step": 504,
+ "eval_loss": 5.1004,
+ "elapsed_s": 848.1
+ },
+ {
+ "epoch": 2.007,
+ "step": 576,
+ "eval_loss": 4.89382,
+ "elapsed_s": 958.3
+ },
+ {
+ "epoch": 2.259,
+ "step": 648,
+ "eval_loss": 4.66662,
+ "elapsed_s": 1069.9
+ },
+ {
+ "epoch": 2.51,
+ "step": 720,
+ "eval_loss": 4.54406,
+ "elapsed_s": 1181.6
+ },
+ {
+ "epoch": 2.762,
+ "step": 792,
+ "eval_loss": 4.45819,
+ "elapsed_s": 1292.9
+ },
+ {
+ "epoch": 3.01,
+ "step": 864,
+ "eval_loss": 4.39381,
+ "elapsed_s": 1403.3
+ },
+ {
+ "epoch": 3.262,
+ "step": 936,
+ "eval_loss": 4.30605,
+ "elapsed_s": 1514.5
+ },
+ {
+ "epoch": 3.514,
+ "step": 1008,
+ "eval_loss": 4.24958,
+ "elapsed_s": 1626.1
+ },
+ {
+ "epoch": 3.765,
+ "step": 1080,
+ "eval_loss": 4.20871,
+ "elapsed_s": 1737.8
+ },
+ {
+ "epoch": 4.014,
+ "step": 1152,
+ "eval_loss": 4.18517,
+ "elapsed_s": 1848.9
+ },
+ {
+ "epoch": 4.266,
+ "step": 1224,
+ "eval_loss": 4.12573,
+ "elapsed_s": 1960.7
+ },
+ {
+ "epoch": 4.517,
+ "step": 1296,
+ "eval_loss": 4.09374,
+ "elapsed_s": 2073.1
+ },
+ {
+ "epoch": 4.769,
+ "step": 1368,
+ "eval_loss": 4.0638,
+ "elapsed_s": 2184.4
+ },
+ {
+ "epoch": 5.017,
+ "step": 1440,
+ "eval_loss": 4.05752,
+ "elapsed_s": 2251.9
+ },
+ {
+ "epoch": 5.269,
+ "step": 1512,
+ "eval_loss": 4.01479,
+ "elapsed_s": 2302.4
+ },
+ {
+ "epoch": 5.521,
+ "step": 1584,
+ "eval_loss": 3.99538,
+ "elapsed_s": 2352.9
+ },
+ {
+ "epoch": 5.772,
+ "step": 1656,
+ "eval_loss": 3.97248,
+ "elapsed_s": 2403.3
+ },
+ {
+ "epoch": 6.021,
+ "step": 1728,
+ "eval_loss": 3.97645,
+ "elapsed_s": 2453.3
+ },
+ {
+ "epoch": 6.273,
+ "step": 1800,
+ "eval_loss": 3.94189,
+ "elapsed_s": 2501.7
+ },
+ {
+ "epoch": 6.524,
+ "step": 1872,
+ "eval_loss": 3.92661,
+ "elapsed_s": 2552.2
+ },
+ {
+ "epoch": 6.776,
+ "step": 1944,
+ "eval_loss": 3.90724,
+ "elapsed_s": 2602.6
+ },
+ {
+ "epoch": 7.024,
+ "step": 2016,
+ "eval_loss": 3.90509,
+ "elapsed_s": 2652.7
+ },
+ {
+ "epoch": 7.276,
+ "step": 2088,
+ "eval_loss": 3.88133,
+ "elapsed_s": 2703.5
+ },
+ {
+ "epoch": 7.528,
+ "step": 2160,
+ "eval_loss": 3.86836,
+ "elapsed_s": 2754.3
+ },
+ {
+ "epoch": 7.779,
+ "step": 2232,
+ "eval_loss": 3.85506,
+ "elapsed_s": 2805.4
+ },
+ {
+ "epoch": 8.028,
+ "step": 2304,
+ "eval_loss": 3.86648,
+ "elapsed_s": 2856.2
+ },
+ {
+ "epoch": 8.28,
+ "step": 2376,
+ "eval_loss": 3.83863,
+ "elapsed_s": 2905.2
+ },
+ {
+ "epoch": 8.531,
+ "step": 2448,
+ "eval_loss": 3.82509,
+ "elapsed_s": 2955.7
+ },
+ {
+ "epoch": 8.783,
+ "step": 2520,
+ "eval_loss": 3.81414,
+ "elapsed_s": 3006.1
+ },
+ {
+ "epoch": 9.031,
+ "step": 2592,
+ "eval_loss": 6.67552,
+ "elapsed_s": 3056.1
+ },
+ {
+ "epoch": 9.283,
+ "step": 2664,
+ "eval_loss": 3.90493,
+ "elapsed_s": 3104.4
+ },
+ {
+ "epoch": 9.535,
+ "step": 2736,
+ "eval_loss": 3.84185,
+ "elapsed_s": 3152.8
+ },
+ {
+ "epoch": 9.786,
+ "step": 2808,
+ "eval_loss": 3.8162,
+ "elapsed_s": 3201.1
+ },
+ {
+ "epoch": 10.035,
+ "step": 2880,
+ "eval_loss": 3.8884,
+ "elapsed_s": 3249.0
+ },
+ {
+ "epoch": 10.287,
+ "step": 2952,
+ "eval_loss": 3.79475,
+ "elapsed_s": 3297.4
+ },
+ {
+ "epoch": 10.538,
+ "step": 3024,
+ "eval_loss": 3.78375,
+ "elapsed_s": 3348.0
+ },
+ {
+ "epoch": 10.79,
+ "step": 3096,
+ "eval_loss": 3.77443,
+ "elapsed_s": 3398.8
+ },
+ {
+ "epoch": 11.038,
+ "step": 3168,
+ "eval_loss": 3.77884,
+ "elapsed_s": 3448.7
+ },
+ {
+ "epoch": 11.29,
+ "step": 3240,
+ "eval_loss": 3.75877,
+ "elapsed_s": 3498.1
+ },
+ {
+ "epoch": 11.542,
+ "step": 3312,
+ "eval_loss": 3.75273,
+ "elapsed_s": 3548.6
+ },
+ {
+ "epoch": 11.793,
+ "step": 3384,
+ "eval_loss": 3.74983,
+ "elapsed_s": 3600.6
+ },
+ {
+ "epoch": 12.042,
+ "step": 3456,
+ "eval_loss": 3.7504,
+ "elapsed_s": 3650.6
+ },
+ {
+ "epoch": 12.294,
+ "step": 3528,
+ "eval_loss": 3.73684,
+ "elapsed_s": 3699.0
+ },
+ {
+ "epoch": 12.545,
+ "step": 3600,
+ "eval_loss": 3.72968,
+ "elapsed_s": 3749.5
+ },
+ {
+ "epoch": 12.797,
+ "step": 3672,
+ "eval_loss": 3.72447,
+ "elapsed_s": 3799.9
+ },
+ {
+ "epoch": 13.045,
+ "step": 3744,
+ "eval_loss": 3.73333,
+ "elapsed_s": 3849.9
+ },
+ {
+ "epoch": 13.297,
+ "step": 3816,
+ "eval_loss": 3.71978,
+ "elapsed_s": 3898.3
+ },
+ {
+ "epoch": 13.549,
+ "step": 3888,
+ "eval_loss": 3.71028,
+ "elapsed_s": 3948.8
+ },
+ {
+ "epoch": 13.8,
+ "step": 3960,
+ "eval_loss": 3.71057,
+ "elapsed_s": 3999.3
+ },
+ {
+ "epoch": 14.049,
+ "step": 4032,
+ "eval_loss": 3.71514,
+ "elapsed_s": 4047.1
+ },
+ {
+ "epoch": 14.301,
+ "step": 4104,
+ "eval_loss": 3.69998,
+ "elapsed_s": 4095.4
+ },
+ {
+ "epoch": 14.552,
+ "step": 4176,
+ "eval_loss": 3.6978,
+ "elapsed_s": 4146.6
+ },
+ {
+ "epoch": 14.804,
+ "step": 4248,
+ "eval_loss": 3.69385,
+ "elapsed_s": 4197.4
+ },
+ {
+ "epoch": 15.052,
+ "step": 4320,
+ "eval_loss": 3.70108,
+ "elapsed_s": 4247.8
+ },
+ {
+ "epoch": 15.304,
+ "step": 4392,
+ "eval_loss": 3.68508,
+ "elapsed_s": 4296.5
+ },
+ {
+ "epoch": 15.556,
+ "step": 4464,
+ "eval_loss": 3.68391,
+ "elapsed_s": 4348.0
+ },
+ {
+ "epoch": 15.807,
+ "step": 4536,
+ "eval_loss": 3.67802,
+ "elapsed_s": 4398.5
+ },
+ {
+ "epoch": 16.056,
+ "step": 4608,
+ "eval_loss": 3.68553,
+ "elapsed_s": 4448.5
+ },
+ {
+ "epoch": 16.308,
+ "step": 4680,
+ "eval_loss": 3.67264,
+ "elapsed_s": 4496.8
+ },
+ {
+ "epoch": 16.559,
+ "step": 4752,
+ "eval_loss": 3.66847,
+ "elapsed_s": 4547.3
+ },
+ {
+ "epoch": 16.811,
+ "step": 4824,
+ "eval_loss": 3.66607,
+ "elapsed_s": 4597.8
+ },
+ {
+ "epoch": 17.059,
+ "step": 4896,
+ "eval_loss": 3.6719,
+ "elapsed_s": 4647.8
+ },
+ {
+ "epoch": 17.311,
+ "step": 4968,
+ "eval_loss": 3.65935,
+ "elapsed_s": 4696.1
+ },
+ {
+ "epoch": 17.563,
+ "step": 5040,
+ "eval_loss": 3.65483,
+ "elapsed_s": 4746.6
+ },
+ {
+ "epoch": 17.814,
+ "step": 5112,
+ "eval_loss": 3.6532,
+ "elapsed_s": 4797.0
+ },
+ {
+ "epoch": 18.063,
+ "step": 5184,
+ "eval_loss": 3.65884,
+ "elapsed_s": 4847.7
+ },
+ {
+ "epoch": 18.315,
+ "step": 5256,
+ "eval_loss": 3.65168,
+ "elapsed_s": 4896.4
+ },
+ {
+ "epoch": 18.566,
+ "step": 5328,
+ "eval_loss": 3.6469,
+ "elapsed_s": 4947.3
+ },
+ {
+ "epoch": 18.818,
+ "step": 5400,
+ "eval_loss": 3.64219,
+ "elapsed_s": 4997.8
+ },
+ {
+ "epoch": 19.066,
+ "step": 5472,
+ "eval_loss": 3.65132,
+ "elapsed_s": 5049.2
+ },
+ {
+ "epoch": 19.318,
+ "step": 5544,
+ "eval_loss": 3.64344,
+ "elapsed_s": 5097.6
+ },
+ {
+ "epoch": 19.57,
+ "step": 5616,
+ "eval_loss": 3.63814,
+ "elapsed_s": 5145.9
+ },
+ {
+ "epoch": 19.821,
+ "step": 5688,
+ "eval_loss": 3.63341,
+ "elapsed_s": 5196.4
+ },
+ {
+ "epoch": 20.07,
+ "step": 5760,
+ "eval_loss": 3.64032,
+ "elapsed_s": 5246.6
+ },
+ {
+ "epoch": 20.322,
+ "step": 5832,
+ "eval_loss": 3.63137,
+ "elapsed_s": 5294.9
+ },
+ {
+ "epoch": 20.573,
+ "step": 5904,
+ "eval_loss": 3.62756,
+ "elapsed_s": 5345.4
+ },
+ {
+ "epoch": 20.825,
+ "step": 5976,
+ "eval_loss": 3.62243,
+ "elapsed_s": 5395.8
+ },
+ {
+ "epoch": 21.073,
+ "step": 6048,
+ "eval_loss": 3.63047,
+ "elapsed_s": 5445.8
+ },
+ {
+ "epoch": 21.325,
+ "step": 6120,
+ "eval_loss": 3.6236,
+ "elapsed_s": 5494.4
+ },
+ {
+ "epoch": 21.577,
+ "step": 6192,
+ "eval_loss": 3.61664,
+ "elapsed_s": 5542.7
+ },
+ {
+ "epoch": 21.828,
+ "step": 6264,
+ "eval_loss": 3.61628,
+ "elapsed_s": 5593.6
+ },
+ {
+ "epoch": 22.077,
+ "step": 6336,
+ "eval_loss": 3.62122,
+ "elapsed_s": 5644.0
+ },
+ {
+ "epoch": 22.329,
+ "step": 6408,
+ "eval_loss": 3.61457,
+ "elapsed_s": 5692.6
+ },
+ {
+ "epoch": 22.58,
+ "step": 6480,
+ "eval_loss": 3.61361,
+ "elapsed_s": 5743.8
+ },
+ {
+ "epoch": 22.832,
+ "step": 6552,
+ "eval_loss": 3.60379,
+ "elapsed_s": 5794.9
+ },
+ {
+ "epoch": 23.08,
+ "step": 6624,
+ "eval_loss": 3.61525,
+ "elapsed_s": 5844.9
+ },
+ {
+ "epoch": 23.332,
+ "step": 6696,
+ "eval_loss": 3.60656,
+ "elapsed_s": 5893.2
+ },
+ {
+ "epoch": 23.584,
+ "step": 6768,
+ "eval_loss": 3.60627,
+ "elapsed_s": 5941.6
+ },
+ {
+ "epoch": 23.835,
+ "step": 6840,
+ "eval_loss": 3.59835,
+ "elapsed_s": 5989.9
+ },
+ {
+ "epoch": 24.084,
+ "step": 6912,
+ "eval_loss": 3.60698,
+ "elapsed_s": 6040.0
+ },
+ {
+ "epoch": 24.336,
+ "step": 6984,
+ "eval_loss": 3.59738,
+ "elapsed_s": 6088.3
+ },
+ {
+ "epoch": 24.587,
+ "step": 7056,
+ "eval_loss": 3.59733,
+ "elapsed_s": 6138.8
+ },
+ {
+ "epoch": 24.839,
+ "step": 7128,
+ "eval_loss": 3.59534,
+ "elapsed_s": 6189.3
+ },
+ {
+ "epoch": 25.087,
+ "step": 7200,
+ "eval_loss": 3.60307,
+ "elapsed_s": 6731.8
+ },
+ {
+ "epoch": 25.339,
+ "step": 7272,
+ "eval_loss": 3.59451,
+ "elapsed_s": 7120.7
+ },
+ {
+ "epoch": 25.591,
+ "step": 7344,
+ "eval_loss": 3.59171,
+ "elapsed_s": 7511.6
+ },
+ {
+ "epoch": 25.842,
+ "step": 7416,
+ "eval_loss": 3.58608,
+ "elapsed_s": 7903.2
+ },
+ {
+ "epoch": 26.091,
+ "step": 7488,
+ "eval_loss": 3.59436,
+ "elapsed_s": 8289.8
+ },
+ {
+ "epoch": 26.343,
+ "step": 7560,
+ "eval_loss": 3.58613,
+ "elapsed_s": 8678.8
+ },
+ {
+ "epoch": 26.594,
+ "step": 7632,
+ "eval_loss": 3.58231,
+ "elapsed_s": 9067.7
+ },
+ {
+ "epoch": 26.846,
+ "step": 7704,
+ "eval_loss": 3.58196,
+ "elapsed_s": 9459.9
+ },
+ {
+ "epoch": 27.094,
+ "step": 7776,
+ "eval_loss": 3.59193,
+ "elapsed_s": 9848.0
+ },
+ {
+ "epoch": 27.346,
+ "step": 7848,
+ "eval_loss": 3.57938,
+ "elapsed_s": 10237.7
+ },
+ {
+ "epoch": 27.598,
+ "step": 7920,
+ "eval_loss": 3.57778,
+ "elapsed_s": 10630.1
+ },
+ {
+ "epoch": 27.849,
+ "step": 7992,
+ "eval_loss": 3.57668,
+ "elapsed_s": 11022.2
+ },
+ {
+ "epoch": 28.098,
+ "step": 8064,
+ "eval_loss": 3.58371,
+ "elapsed_s": 11423.8
+ },
+ {
+ "epoch": 28.35,
+ "step": 8136,
+ "eval_loss": 3.57704,
+ "elapsed_s": 11813.7
+ },
+ {
+ "epoch": 28.601,
+ "step": 8208,
+ "eval_loss": 3.57326,
+ "elapsed_s": 12203.4
+ },
+ {
+ "epoch": 28.853,
+ "step": 8280,
+ "eval_loss": 3.57204,
+ "elapsed_s": 12595.1
+ },
+ {
+ "epoch": 29.101,
+ "step": 8352,
+ "eval_loss": 3.57795,
+ "elapsed_s": 12982.0
+ },
+ {
+ "epoch": 29.353,
+ "step": 8424,
+ "eval_loss": 3.57024,
+ "elapsed_s": 13370.9
+ },
+ {
+ "epoch": 29.605,
+ "step": 8496,
+ "eval_loss": 3.57179,
+ "elapsed_s": 13764.0
+ },
+ {
+ "epoch": 29.856,
+ "step": 8568,
+ "eval_loss": 3.56631,
+ "elapsed_s": 14154.1
+ },
+ {
+ "epoch": 30.105,
+ "step": 8640,
+ "eval_loss": 3.57341,
+ "elapsed_s": 14541.2
+ },
+ {
+ "epoch": 30.357,
+ "step": 8712,
+ "eval_loss": 3.56351,
+ "elapsed_s": 14930.3
+ },
+ {
+ "epoch": 30.608,
+ "step": 8784,
+ "eval_loss": 3.56147,
+ "elapsed_s": 15321.7
+ },
+ {
+ "epoch": 30.86,
+ "step": 8856,
+ "eval_loss": 3.564,
+ "elapsed_s": 15713.7
+ },
+ {
+ "epoch": 31.108,
+ "step": 8928,
+ "eval_loss": 3.56728,
+ "elapsed_s": 16098.2
+ },
+ {
+ "epoch": 31.36,
+ "step": 9000,
+ "eval_loss": 3.5604,
+ "elapsed_s": 16487.8
+ },
+ {
+ "epoch": 31.612,
+ "step": 9072,
+ "eval_loss": 3.56232,
+ "elapsed_s": 16879.6
+ },
+ {
+ "epoch": 31.863,
+ "step": 9144,
+ "eval_loss": 3.55983,
+ "elapsed_s": 17269.0
+ },
+ {
+ "epoch": 32.112,
+ "step": 9216,
+ "eval_loss": 3.56572,
+ "elapsed_s": 17655.8
+ },
+ {
+ "epoch": 32.364,
+ "step": 9288,
+ "eval_loss": 3.55957,
+ "elapsed_s": 18045.8
+ },
+ {
+ "epoch": 32.615,
+ "step": 9360,
+ "eval_loss": 3.55477,
+ "elapsed_s": 18437.3
+ },
+ {
+ "epoch": 32.867,
+ "step": 9432,
+ "eval_loss": 3.55578,
+ "elapsed_s": 18828.6
+ },
+ {
+ "epoch": 33.115,
+ "step": 9504,
+ "eval_loss": 3.56018,
+ "elapsed_s": 19213.0
+ },
+ {
+ "epoch": 33.367,
+ "step": 9576,
+ "eval_loss": 3.55524,
+ "elapsed_s": 19602.4
+ },
+ {
+ "epoch": 33.619,
+ "step": 9648,
+ "eval_loss": 3.55347,
+ "elapsed_s": 19992.4
+ },
+ {
+ "epoch": 33.87,
+ "step": 9720,
+ "eval_loss": 3.54808,
+ "elapsed_s": 20384.1
+ },
+ {
+ "epoch": 34.119,
+ "step": 9792,
+ "eval_loss": 3.55563,
+ "elapsed_s": 20771.1
+ },
+ {
+ "epoch": 34.371,
+ "step": 9864,
+ "eval_loss": 3.55029,
+ "elapsed_s": 21160.5
+ },
+ {
+ "epoch": 34.622,
+ "step": 9936,
+ "eval_loss": 3.54849,
+ "elapsed_s": 21549.8
+ },
+ {
+ "epoch": 34.874,
+ "step": 10008,
+ "eval_loss": 3.5451,
+ "elapsed_s": 21939.5
+ },
+ {
+ "epoch": 35.122,
+ "step": 10080,
+ "eval_loss": 3.55206,
+ "elapsed_s": 22326.6
+ },
+ {
+ "epoch": 35.374,
+ "step": 10152,
+ "eval_loss": 3.54608,
+ "elapsed_s": 22715.5
+ },
+ {
+ "epoch": 35.626,
+ "step": 10224,
+ "eval_loss": 3.5456,
+ "elapsed_s": 23104.0
+ },
+ {
+ "epoch": 35.877,
+ "step": 10296,
+ "eval_loss": 3.54011,
+ "elapsed_s": 23492.8
+ },
+ {
+ "epoch": 36.126,
+ "step": 10368,
+ "eval_loss": 3.54651,
+ "elapsed_s": 23879.0
+ },
+ {
+ "epoch": 36.377,
+ "step": 10440,
+ "eval_loss": 3.54329,
+ "elapsed_s": 24266.6
+ },
+ {
+ "epoch": 36.629,
+ "step": 10512,
+ "eval_loss": 3.54259,
+ "elapsed_s": 24654.4
+ },
+ {
+ "epoch": 36.881,
+ "step": 10584,
+ "eval_loss": 3.53793,
+ "elapsed_s": 25042.0
+ },
+ {
+ "epoch": 37.129,
+ "step": 10656,
+ "eval_loss": 3.54122,
+ "elapsed_s": 25427.2
+ },
+ {
+ "epoch": 37.381,
+ "step": 10728,
+ "eval_loss": 3.53876,
+ "elapsed_s": 25815.3
+ },
+ {
+ "epoch": 37.633,
+ "step": 10800,
+ "eval_loss": 3.53613,
+ "elapsed_s": 26202.8
+ },
+ {
+ "epoch": 37.884,
+ "step": 10872,
+ "eval_loss": 3.53545,
+ "elapsed_s": 26593.0
+ },
+ {
+ "epoch": 38.133,
+ "step": 10944,
+ "eval_loss": 3.54046,
+ "elapsed_s": 26978.3
+ },
+ {
+ "epoch": 38.384,
+ "step": 11016,
+ "eval_loss": 3.53802,
+ "elapsed_s": 27366.0
+ },
+ {
+ "epoch": 38.636,
+ "step": 11088,
+ "eval_loss": 3.535,
+ "elapsed_s": 27754.1
+ },
+ {
+ "epoch": 38.888,
+ "step": 11160,
+ "eval_loss": 3.53246,
+ "elapsed_s": 28144.1
+ },
+ {
+ "epoch": 39.136,
+ "step": 11232,
+ "eval_loss": 3.53649,
+ "elapsed_s": 28529.3
+ },
+ {
+ "epoch": 39.388,
+ "step": 11304,
+ "eval_loss": 3.53324,
+ "elapsed_s": 28916.8
+ },
+ {
+ "epoch": 39.64,
+ "step": 11376,
+ "eval_loss": 3.53043,
+ "elapsed_s": 29304.2
+ },
+ {
+ "epoch": 39.891,
+ "step": 11448,
+ "eval_loss": 3.52849,
+ "elapsed_s": 29694.1
+ },
+ {
+ "epoch": 40.14,
+ "step": 11520,
+ "eval_loss": 3.53379,
+ "elapsed_s": 30079.8
+ },
+ {
+ "epoch": 40.391,
+ "step": 11592,
+ "eval_loss": 3.52913,
+ "elapsed_s": 30467.3
+ },
+ {
+ "epoch": 40.643,
+ "step": 11664,
+ "eval_loss": 3.52448,
+ "elapsed_s": 30854.9
+ },
+ {
+ "epoch": 40.895,
+ "step": 11736,
+ "eval_loss": 3.52893,
+ "elapsed_s": 31244.8
+ },
+ {
+ "epoch": 41.143,
+ "step": 11808,
+ "eval_loss": 3.52947,
+ "elapsed_s": 31627.8
+ },
+ {
+ "epoch": 41.395,
+ "step": 11880,
+ "eval_loss": 3.52704,
+ "elapsed_s": 32016.0
+ },
+ {
+ "epoch": 41.647,
+ "step": 11952,
+ "eval_loss": 3.52387,
+ "elapsed_s": 32403.6
+ },
+ {
+ "epoch": 41.898,
+ "step": 12024,
+ "eval_loss": 3.52453,
+ "elapsed_s": 32793.6
+ },
+ {
+ "epoch": 42.147,
+ "step": 12096,
+ "eval_loss": 3.52956,
+ "elapsed_s": 33176.6
+ },
+ {
+ "epoch": 42.398,
+ "step": 12168,
+ "eval_loss": 3.52508,
+ "elapsed_s": 33564.3
+ },
+ {
+ "epoch": 42.65,
+ "step": 12240,
+ "eval_loss": 3.52467,
+ "elapsed_s": 33952.3
+ },
+ {
+ "epoch": 42.902,
+ "step": 12312,
+ "eval_loss": 3.52072,
+ "elapsed_s": 34339.9
+ },
+ {
+ "epoch": 43.15,
+ "step": 12384,
+ "eval_loss": 3.52664,
+ "elapsed_s": 34725.2
+ },
+ {
+ "epoch": 43.402,
+ "step": 12456,
+ "eval_loss": 3.52103,
+ "elapsed_s": 35112.6
+ },
+ {
+ "epoch": 43.654,
+ "step": 12528,
+ "eval_loss": 3.52203,
+ "elapsed_s": 35500.2
+ },
+ {
+ "epoch": 43.905,
+ "step": 12600,
+ "eval_loss": 3.51634,
+ "elapsed_s": 35888.3
+ },
+ {
+ "epoch": 44.154,
+ "step": 12672,
+ "eval_loss": 3.52388,
+ "elapsed_s": 36273.5
+ },
+ {
+ "epoch": 44.405,
+ "step": 12744,
+ "eval_loss": 3.52112,
+ "elapsed_s": 36661.4
+ },
+ {
+ "epoch": 44.657,
+ "step": 12816,
+ "eval_loss": 3.5174,
+ "elapsed_s": 37048.9
+ },
+ {
+ "epoch": 44.909,
+ "step": 12888,
+ "eval_loss": 3.51659,
+ "elapsed_s": 37436.6
+ },
+ {
+ "epoch": 45.157,
+ "step": 12960,
+ "eval_loss": 3.52008,
+ "elapsed_s": 37820.2
+ },
+ {
+ "epoch": 45.409,
+ "step": 13032,
+ "eval_loss": 3.51681,
+ "elapsed_s": 38207.9
+ },
+ {
+ "epoch": 45.661,
+ "step": 13104,
+ "eval_loss": 3.5159,
+ "elapsed_s": 38595.5
+ },
+ {
+ "epoch": 45.912,
+ "step": 13176,
+ "eval_loss": 3.51148,
+ "elapsed_s": 38985.5
+ },
+ {
+ "epoch": 46.161,
+ "step": 13248,
+ "eval_loss": 3.51896,
+ "elapsed_s": 39370.9
+ },
+ {
+ "epoch": 46.412,
+ "step": 13320,
+ "eval_loss": 3.51379,
+ "elapsed_s": 39758.9
+ },
+ {
+ "epoch": 46.664,
+ "step": 13392,
+ "eval_loss": 3.51426,
+ "elapsed_s": 40146.4
+ },
+ {
+ "epoch": 46.916,
+ "step": 13464,
+ "eval_loss": 3.51021,
+ "elapsed_s": 40534.1
+ },
+ {
+ "epoch": 47.164,
+ "step": 13536,
+ "eval_loss": 3.51518,
+ "elapsed_s": 40921.0
+ },
+ {
+ "epoch": 47.416,
+ "step": 13608,
+ "eval_loss": 3.51122,
+ "elapsed_s": 41310.7
+ },
+ {
+ "epoch": 47.668,
+ "step": 13680,
+ "eval_loss": 3.51003,
+ "elapsed_s": 41699.6
+ },
+ {
+ "epoch": 47.919,
+ "step": 13752,
+ "eval_loss": 3.50739,
+ "elapsed_s": 42091.6
+ },
+ {
+ "epoch": 48.168,
+ "step": 13824,
+ "eval_loss": 3.51306,
+ "elapsed_s": 42477.9
+ },
+ {
+ "epoch": 48.419,
+ "step": 13896,
+ "eval_loss": 3.50942,
+ "elapsed_s": 42866.6
+ },
+ {
+ "epoch": 48.671,
+ "step": 13968,
+ "eval_loss": 3.50699,
+ "elapsed_s": 43255.4
+ },
+ {
+ "epoch": 48.923,
+ "step": 14040,
+ "eval_loss": 3.50626,
+ "elapsed_s": 43646.5
+ },
+ {
+ "epoch": 49.171,
+ "step": 14112,
+ "eval_loss": 3.5114,
+ "elapsed_s": 44033.5
+ },
+ {
+ "epoch": 49.423,
+ "step": 14184,
+ "eval_loss": 3.50836,
+ "elapsed_s": 44422.3
+ },
+ {
+ "epoch": 49.675,
+ "step": 14256,
+ "eval_loss": 3.50505,
+ "elapsed_s": 44811.1
+ },
+ {
+ "epoch": 49.926,
+ "step": 14328,
+ "eval_loss": 3.50282,
+ "elapsed_s": 45202.1
+ },
+ {
+ "epoch": 50.175,
+ "step": 14400,
+ "eval_loss": 3.50959,
+ "elapsed_s": 45588.3
+ },
+ {
+ "epoch": 50.426,
+ "step": 14472,
+ "eval_loss": 3.5086,
+ "elapsed_s": 45977.4
+ },
+ {
+ "epoch": 50.678,
+ "step": 14544,
+ "eval_loss": 3.50372,
+ "elapsed_s": 46366.3
+ },
+ {
+ "epoch": 50.93,
+ "step": 14616,
+ "eval_loss": 3.50344,
+ "elapsed_s": 46755.0
+ },
+ {
+ "epoch": 51.178,
+ "step": 14688,
+ "eval_loss": 3.50644,
+ "elapsed_s": 47139.1
+ },
+ {
+ "epoch": 51.43,
+ "step": 14760,
+ "eval_loss": 3.50503,
+ "elapsed_s": 47527.8
+ },
+ {
+ "epoch": 51.682,
+ "step": 14832,
+ "eval_loss": 3.50296,
+ "elapsed_s": 47917.0
+ },
+ {
+ "epoch": 51.933,
+ "step": 14904,
+ "eval_loss": 3.50271,
+ "elapsed_s": 48305.5
+ },
+ {
+ "epoch": 52.182,
+ "step": 14976,
+ "eval_loss": 3.50374,
+ "elapsed_s": 48691.8
+ },
+ {
+ "epoch": 52.433,
+ "step": 15048,
+ "eval_loss": 3.50375,
+ "elapsed_s": 49080.4
+ },
+ {
+ "epoch": 52.685,
+ "step": 15120,
+ "eval_loss": 3.4994,
+ "elapsed_s": 49469.2
+ },
+ {
+ "epoch": 52.937,
+ "step": 15192,
+ "eval_loss": 3.4981,
+ "elapsed_s": 49860.6
+ },
+ {
+ "epoch": 53.185,
+ "step": 15264,
+ "eval_loss": 3.50164,
+ "elapsed_s": 50247.1
+ },
+ {
+ "epoch": 53.437,
+ "step": 15336,
+ "eval_loss": 3.5008,
+ "elapsed_s": 50636.2
+ },
+ {
+ "epoch": 53.689,
+ "step": 15408,
+ "eval_loss": 3.49973,
+ "elapsed_s": 51025.1
+ },
+ {
+ "epoch": 53.94,
+ "step": 15480,
+ "eval_loss": 3.49597,
+ "elapsed_s": 51414.2
+ },
+ {
+ "epoch": 54.189,
+ "step": 15552,
+ "eval_loss": 3.50117,
+ "elapsed_s": 51801.2
+ },
+ {
+ "epoch": 54.44,
+ "step": 15624,
+ "eval_loss": 3.49882,
+ "elapsed_s": 52190.2
+ },
+ {
+ "epoch": 54.692,
+ "step": 15696,
+ "eval_loss": 3.49636,
+ "elapsed_s": 52579.0
+ },
+ {
+ "epoch": 54.944,
+ "step": 15768,
+ "eval_loss": 3.49504,
+ "elapsed_s": 52967.9
+ },
+ {
+ "epoch": 55.192,
+ "step": 15840,
+ "eval_loss": 3.498,
+ "elapsed_s": 53354.5
+ },
+ {
+ "epoch": 55.444,
+ "step": 15912,
+ "eval_loss": 3.49579,
+ "elapsed_s": 53743.5
+ },
+ {
+ "epoch": 55.696,
+ "step": 15984,
+ "eval_loss": 3.49525,
+ "elapsed_s": 54132.9
+ },
+ {
+ "epoch": 55.947,
+ "step": 16056,
+ "eval_loss": 3.49321,
+ "elapsed_s": 54521.8
+ },
+ {
+ "epoch": 56.196,
+ "step": 16128,
+ "eval_loss": 3.49531,
+ "elapsed_s": 54908.2
+ },
+ {
+ "epoch": 56.447,
+ "step": 16200,
+ "eval_loss": 3.49431,
+ "elapsed_s": 55297.3
+ },
+ {
+ "epoch": 56.699,
+ "step": 16272,
+ "eval_loss": 3.49227,
+ "elapsed_s": 55686.2
+ },
+ {
+ "epoch": 56.951,
+ "step": 16344,
+ "eval_loss": 3.49298,
+ "elapsed_s": 56077.9
+ },
+ {
+ "epoch": 57.199,
+ "step": 16416,
+ "eval_loss": 3.49562,
+ "elapsed_s": 56462.4
+ },
+ {
+ "epoch": 57.451,
+ "step": 16488,
+ "eval_loss": 3.49362,
+ "elapsed_s": 56851.3
+ },
+ {
+ "epoch": 57.703,
+ "step": 16560,
+ "eval_loss": 3.49118,
+ "elapsed_s": 57240.3
+ },
+ {
+ "epoch": 57.954,
+ "step": 16632,
+ "eval_loss": 3.4903,
+ "elapsed_s": 57631.4
+ },
+ {
+ "epoch": 58.203,
+ "step": 16704,
+ "eval_loss": 3.49353,
+ "elapsed_s": 58018.4
+ },
+ {
+ "epoch": 58.454,
+ "step": 16776,
+ "eval_loss": 3.49317,
+ "elapsed_s": 58407.3
+ },
+ {
+ "epoch": 58.706,
+ "step": 16848,
+ "eval_loss": 3.49039,
+ "elapsed_s": 58796.3
+ },
+ {
+ "epoch": 58.958,
+ "step": 16920,
+ "eval_loss": 3.48705,
+ "elapsed_s": 59185.1
+ },
+ {
+ "epoch": 59.206,
+ "step": 16992,
+ "eval_loss": 3.48937,
+ "elapsed_s": 59571.7
+ },
+ {
+ "epoch": 59.458,
+ "step": 17064,
+ "eval_loss": 3.49072,
+ "elapsed_s": 59961.4
+ },
+ {
+ "epoch": 59.71,
+ "step": 17136,
+ "eval_loss": 3.48841,
+ "elapsed_s": 60350.5
+ },
+ {
+ "epoch": 59.961,
+ "step": 17208,
+ "eval_loss": 3.48576,
+ "elapsed_s": 60739.4
+ },
+ {
+ "epoch": 60.21,
+ "step": 17280,
+ "eval_loss": 3.48876,
+ "elapsed_s": 61125.9
+ },
+ {
+ "epoch": 60.461,
+ "step": 17352,
+ "eval_loss": 3.48727,
+ "elapsed_s": 61515.0
+ },
+ {
+ "epoch": 60.713,
+ "step": 17424,
+ "eval_loss": 3.48815,
+ "elapsed_s": 61904.3
+ },
+ {
+ "epoch": 60.965,
+ "step": 17496,
+ "eval_loss": 3.48469,
+ "elapsed_s": 62293.5
+ },
+ {
+ "epoch": 61.213,
+ "step": 17568,
+ "eval_loss": 3.49021,
+ "elapsed_s": 62680.5
+ },
+ {
+ "epoch": 61.465,
+ "step": 17640,
+ "eval_loss": 3.48764,
+ "elapsed_s": 63069.5
+ },
+ {
+ "epoch": 61.717,
+ "step": 17712,
+ "eval_loss": 3.4857,
+ "elapsed_s": 63458.5
+ },
+ {
+ "epoch": 61.968,
+ "step": 17784,
+ "eval_loss": 3.48401,
+ "elapsed_s": 63847.9
+ },
+ {
+ "epoch": 62.217,
+ "step": 17856,
+ "eval_loss": 3.48801,
+ "elapsed_s": 64234.6
+ },
+ {
+ "epoch": 62.468,
+ "step": 17928,
+ "eval_loss": 3.48568,
+ "elapsed_s": 64623.4
+ },
+ {
+ "epoch": 62.72,
+ "step": 18000,
+ "eval_loss": 3.48372,
+ "elapsed_s": 65012.3
+ },
+ {
+ "epoch": 62.972,
+ "step": 18072,
+ "eval_loss": 3.48394,
+ "elapsed_s": 65403.4
+ },
+ {
+ "epoch": 63.22,
+ "step": 18144,
+ "eval_loss": 3.48875,
+ "elapsed_s": 65787.7
+ },
+ {
+ "epoch": 63.472,
+ "step": 18216,
+ "eval_loss": 3.48375,
+ "elapsed_s": 66177.0
+ },
+ {
+ "epoch": 63.724,
+ "step": 18288,
+ "eval_loss": 3.48224,
+ "elapsed_s": 66566.2
+ },
+ {
+ "epoch": 63.975,
+ "step": 18360,
+ "eval_loss": 3.48099,
+ "elapsed_s": 66957.5
+ },
+ {
+ "epoch": 64.224,
+ "step": 18432,
+ "eval_loss": 3.48419,
+ "elapsed_s": 67343.9
+ },
+ {
+ "epoch": 64.475,
+ "step": 18504,
+ "eval_loss": 3.48345,
+ "elapsed_s": 67732.6
+ },
+ {
+ "epoch": 64.727,
+ "step": 18576,
+ "eval_loss": 3.48304,
+ "elapsed_s": 68121.9
+ },
+ {
+ "epoch": 64.979,
+ "step": 18648,
+ "eval_loss": 3.48054,
+ "elapsed_s": 68510.7
+ },
+ {
+ "epoch": 65.227,
+ "step": 18720,
+ "eval_loss": 3.48483,
+ "elapsed_s": 68897.3
+ },
+ {
+ "epoch": 65.479,
+ "step": 18792,
+ "eval_loss": 3.4828,
+ "elapsed_s": 69286.3
+ },
+ {
+ "epoch": 65.731,
+ "step": 18864,
+ "eval_loss": 3.47941,
+ "elapsed_s": 69675.0
+ },
+ {
+ "epoch": 65.982,
+ "step": 18936,
+ "eval_loss": 3.47778,
+ "elapsed_s": 70066.6
+ },
+ {
+ "epoch": 66.231,
+ "step": 19008,
+ "eval_loss": 3.48055,
+ "elapsed_s": 70453.3
+ },
+ {
+ "epoch": 66.482,
+ "step": 19080,
+ "eval_loss": 3.48087,
+ "elapsed_s": 70842.5
+ },
+ {
+ "epoch": 66.734,
+ "step": 19152,
+ "eval_loss": 3.47915,
+ "elapsed_s": 71231.7
+ },
+ {
+ "epoch": 66.986,
+ "step": 19224,
+ "eval_loss": 3.48032,
+ "elapsed_s": 71621.6
+ },
+ {
+ "epoch": 67.234,
+ "step": 19296,
+ "eval_loss": 3.4822,
+ "elapsed_s": 72006.8
+ },
+ {
+ "epoch": 67.486,
+ "step": 19368,
+ "eval_loss": 3.4791,
+ "elapsed_s": 72396.3
+ },
+ {
+ "epoch": 67.738,
+ "step": 19440,
+ "eval_loss": 3.47912,
+ "elapsed_s": 72785.7
+ },
+ {
+ "epoch": 67.989,
+ "step": 19512,
+ "eval_loss": 3.47593,
+ "elapsed_s": 73176.2
+ },
+ {
+ "epoch": 68.238,
+ "step": 19584,
+ "eval_loss": 3.48069,
+ "elapsed_s": 73564.1
+ },
+ {
+ "epoch": 68.489,
+ "step": 19656,
+ "eval_loss": 3.47962,
+ "elapsed_s": 73955.2
+ },
+ {
+ "epoch": 68.741,
+ "step": 19728,
+ "eval_loss": 3.47807,
+ "elapsed_s": 74345.6
+ },
+ {
+ "epoch": 68.993,
+ "step": 19800,
+ "eval_loss": 3.47457,
+ "elapsed_s": 74736.0
+ },
+ {
+ "epoch": 69.241,
+ "step": 19872,
+ "eval_loss": 3.47805,
+ "elapsed_s": 75123.7
+ },
+ {
+ "epoch": 69.493,
+ "step": 19944,
+ "eval_loss": 3.47605,
+ "elapsed_s": 75513.4
+ },
+ {
+ "epoch": 69.745,
+ "step": 20016,
+ "eval_loss": 3.47705,
+ "elapsed_s": 75903.2
+ },
+ {
+ "epoch": 69.996,
+ "step": 20088,
+ "eval_loss": 3.47458,
+ "elapsed_s": 76292.3
+ },
+ {
+ "epoch": 70.245,
+ "step": 20160,
+ "eval_loss": 3.48024,
+ "elapsed_s": 76676.9
+ },
+ {
+ "epoch": 70.496,
+ "step": 20232,
+ "eval_loss": 3.47494,
+ "elapsed_s": 77066.0
+ },
+ {
+ "epoch": 70.748,
+ "step": 20304,
+ "eval_loss": 3.47602,
+ "elapsed_s": 77455.0
+ },
+ {
+ "epoch": 71.0,
+ "step": 20376,
+ "eval_loss": 3.47144,
+ "elapsed_s": 77844.0
+ },
+ {
+ "epoch": 71.248,
+ "step": 20448,
+ "eval_loss": 3.47709,
+ "elapsed_s": 78231.1
+ },
+ {
+ "epoch": 71.5,
+ "step": 20520,
+ "eval_loss": 3.47558,
+ "elapsed_s": 78620.2
+ },
+ {
+ "epoch": 71.752,
+ "step": 20592,
+ "eval_loss": 3.47113,
+ "elapsed_s": 79009.3
+ },
+ {
+ "epoch": 72.0,
+ "step": 20664,
+ "eval_loss": 3.47885,
+ "elapsed_s": 79395.9
+ },
+ {
+ "epoch": 72.252,
+ "step": 20736,
+ "eval_loss": 3.47689,
+ "elapsed_s": 79785.5
+ },
+ {
+ "epoch": 72.503,
+ "step": 20808,
+ "eval_loss": 3.47502,
+ "elapsed_s": 80175.1
+ },
+ {
+ "epoch": 72.755,
+ "step": 20880,
+ "eval_loss": 3.47205,
+ "elapsed_s": 80563.9
+ },
+ {
+ "epoch": 73.003,
+ "step": 20952,
+ "eval_loss": 3.48866,
+ "elapsed_s": 80948.4
+ },
+ {
+ "epoch": 73.255,
+ "step": 21024,
+ "eval_loss": 3.47697,
+ "elapsed_s": 81337.4
+ },
+ {
+ "epoch": 73.507,
+ "step": 21096,
+ "eval_loss": 3.47673,
+ "elapsed_s": 81726.1
+ },
+ {
+ "epoch": 73.758,
+ "step": 21168,
+ "eval_loss": 3.47205,
+ "elapsed_s": 82115.5
+ },
+ {
+ "epoch": 74.007,
+ "step": 21240,
+ "eval_loss": 3.48928,
+ "elapsed_s": 82499.7
+ },
+ {
+ "epoch": 74.259,
+ "step": 21312,
+ "eval_loss": 3.47716,
+ "elapsed_s": 82888.8
+ },
+ {
+ "epoch": 74.51,
+ "step": 21384,
+ "eval_loss": 3.47378,
+ "elapsed_s": 83277.8
+ },
+ {
+ "epoch": 74.762,
+ "step": 21456,
+ "eval_loss": 3.47079,
+ "elapsed_s": 83666.7
+ },
+ {
+ "epoch": 75.01,
+ "step": 21528,
+ "eval_loss": 3.48633,
+ "elapsed_s": 84053.6
+ },
+ {
+ "epoch": 75.262,
+ "step": 21600,
+ "eval_loss": 3.47297,
+ "elapsed_s": 84442.3
+ },
+ {
+ "epoch": 75.514,
+ "step": 21672,
+ "eval_loss": 3.47189,
+ "elapsed_s": 84831.1
+ },
+ {
+ "epoch": 75.765,
+ "step": 21744,
+ "eval_loss": 3.47008,
+ "elapsed_s": 85220.0
+ },
+ {
+ "epoch": 76.014,
+ "step": 21816,
+ "eval_loss": 3.48892,
+ "elapsed_s": 85606.5
+ },
+ {
+ "epoch": 76.266,
+ "step": 21888,
+ "eval_loss": 3.47142,
+ "elapsed_s": 85995.7
+ },
+ {
+ "epoch": 76.517,
+ "step": 21960,
+ "eval_loss": 3.4702,
+ "elapsed_s": 86384.5
+ },
+ {
+ "epoch": 76.769,
+ "step": 22032,
+ "eval_loss": 3.46946,
+ "elapsed_s": 86773.2
+ },
+ {
+ "epoch": 77.017,
+ "step": 22104,
+ "eval_loss": 3.48544,
+ "elapsed_s": 87159.3
+ },
+ {
+ "epoch": 77.269,
+ "step": 22176,
+ "eval_loss": 3.47133,
+ "elapsed_s": 87548.0
+ },
+ {
+ "epoch": 77.521,
+ "step": 22248,
+ "eval_loss": 3.46839,
+ "elapsed_s": 87936.7
+ },
+ {
+ "epoch": 77.772,
+ "step": 22320,
+ "eval_loss": 3.46851,
+ "elapsed_s": 88328.1
+ },
+ {
+ "epoch": 78.021,
+ "step": 22392,
+ "eval_loss": 3.486,
+ "elapsed_s": 88712.4
+ },
+ {
+ "epoch": 78.273,
+ "step": 22464,
+ "eval_loss": 3.46979,
+ "elapsed_s": 89101.1
+ },
+ {
+ "epoch": 78.524,
+ "step": 22536,
+ "eval_loss": 3.46774,
+ "elapsed_s": 89489.9
+ },
+ {
+ "epoch": 78.776,
+ "step": 22608,
+ "eval_loss": 3.46715,
+ "elapsed_s": 89881.0
+ },
+ {
+ "epoch": 79.024,
+ "step": 22680,
+ "eval_loss": 3.48532,
+ "elapsed_s": 90268.1
+ },
+ {
+ "epoch": 79.276,
+ "step": 22752,
+ "eval_loss": 3.46925,
+ "elapsed_s": 90657.1
+ },
+ {
+ "epoch": 79.528,
+ "step": 22824,
+ "eval_loss": 3.46606,
+ "elapsed_s": 91046.0
+ },
+ {
+ "epoch": 79.779,
+ "step": 22896,
+ "eval_loss": 3.46757,
+ "elapsed_s": 91437.1
+ },
+ {
+ "epoch": 80.028,
+ "step": 22968,
+ "eval_loss": 3.48999,
+ "elapsed_s": 91821.4
+ },
+ {
+ "epoch": 80.28,
+ "step": 23040,
+ "eval_loss": 3.46791,
+ "elapsed_s": 92211.0
+ },
+ {
+ "epoch": 80.531,
+ "step": 23112,
+ "eval_loss": 3.46956,
+ "elapsed_s": 92599.7
+ },
+ {
+ "epoch": 80.783,
+ "step": 23184,
+ "eval_loss": 3.4646,
+ "elapsed_s": 92988.5
+ },
+ {
+ "epoch": 81.031,
+ "step": 23256,
+ "eval_loss": 3.4815,
+ "elapsed_s": 93375.0
+ },
+ {
+ "epoch": 81.283,
+ "step": 23328,
+ "eval_loss": 3.46536,
+ "elapsed_s": 93763.8
+ },
+ {
+ "epoch": 81.535,
+ "step": 23400,
+ "eval_loss": 3.46569,
+ "elapsed_s": 94153.0
+ },
+ {
+ "epoch": 81.786,
+ "step": 23472,
+ "eval_loss": 3.46344,
+ "elapsed_s": 94541.7
+ },
+ {
+ "epoch": 82.035,
+ "step": 23544,
+ "eval_loss": 3.48111,
+ "elapsed_s": 94928.0
+ },
+ {
+ "epoch": 82.287,
+ "step": 23616,
+ "eval_loss": 3.46746,
+ "elapsed_s": 95316.9
+ },
+ {
+ "epoch": 82.538,
+ "step": 23688,
+ "eval_loss": 3.46583,
+ "elapsed_s": 95705.7
+ },
+ {
+ "epoch": 82.79,
+ "step": 23760,
+ "eval_loss": 3.46213,
+ "elapsed_s": 96095.4
+ },
+ {
+ "epoch": 83.038,
+ "step": 23832,
+ "eval_loss": 3.47903,
+ "elapsed_s": 96482.3
+ },
+ {
+ "epoch": 83.29,
+ "step": 23904,
+ "eval_loss": 3.46636,
+ "elapsed_s": 96871.4
+ },
+ {
+ "epoch": 83.542,
+ "step": 23976,
+ "eval_loss": 3.46661,
+ "elapsed_s": 97260.3
+ },
+ {
+ "epoch": 83.793,
+ "step": 24048,
+ "eval_loss": 3.46158,
+ "elapsed_s": 97649.1
+ },
+ {
+ "epoch": 84.042,
+ "step": 24120,
+ "eval_loss": 3.47778,
+ "elapsed_s": 98036.2
+ },
+ {
+ "epoch": 84.294,
+ "step": 24192,
+ "eval_loss": 3.46433,
+ "elapsed_s": 98425.2
+ },
+ {
+ "epoch": 84.545,
+ "step": 24264,
+ "eval_loss": 3.46564,
+ "elapsed_s": 98814.1
+ },
+ {
+ "epoch": 84.797,
+ "step": 24336,
+ "eval_loss": 3.45994,
+ "elapsed_s": 99203.1
+ },
+ {
+ "epoch": 85.045,
+ "step": 24408,
+ "eval_loss": 3.4792,
+ "elapsed_s": 99589.7
+ },
+ {
+ "epoch": 85.297,
+ "step": 24480,
+ "eval_loss": 3.46525,
+ "elapsed_s": 99978.4
+ },
+ {
+ "epoch": 85.549,
+ "step": 24552,
+ "eval_loss": 3.46195,
+ "elapsed_s": 100367.8
+ },
+ {
+ "epoch": 85.8,
+ "step": 24624,
+ "eval_loss": 3.4632,
+ "elapsed_s": 100756.8
+ },
+ {
+ "epoch": 86.049,
+ "step": 24696,
+ "eval_loss": 3.47633,
+ "elapsed_s": 101140.9
+ },
+ {
+ "epoch": 86.301,
+ "step": 24768,
+ "eval_loss": 3.46357,
+ "elapsed_s": 101529.8
+ },
+ {
+ "epoch": 86.552,
+ "step": 24840,
+ "eval_loss": 3.46218,
+ "elapsed_s": 101918.5
+ },
+ {
+ "epoch": 86.804,
+ "step": 24912,
+ "eval_loss": 3.46054,
+ "elapsed_s": 102307.8
+ },
+ {
+ "epoch": 87.052,
+ "step": 24984,
+ "eval_loss": 3.47164,
+ "elapsed_s": 102692.0
+ },
+ {
+ "epoch": 87.304,
+ "step": 25056,
+ "eval_loss": 3.46179,
+ "elapsed_s": 103081.2
+ },
+ {
+ "epoch": 87.556,
+ "step": 25128,
+ "eval_loss": 3.45948,
+ "elapsed_s": 103470.3
+ },
+ {
+ "epoch": 87.807,
+ "step": 25200,
+ "eval_loss": 3.45815,
+ "elapsed_s": 103861.4
+ },
+ {
+ "epoch": 88.056,
+ "step": 25272,
+ "eval_loss": 3.47303,
+ "elapsed_s": 104248.5
+ },
+ {
+ "epoch": 88.308,
+ "step": 25344,
+ "eval_loss": 3.46486,
+ "elapsed_s": 104637.5
+ },
+ {
+ "epoch": 88.559,
+ "step": 25416,
+ "eval_loss": 3.46072,
+ "elapsed_s": 105026.8
+ },
+ {
+ "epoch": 88.811,
+ "step": 25488,
+ "eval_loss": 3.45918,
+ "elapsed_s": 105415.8
+ },
+ {
+ "epoch": 89.059,
+ "step": 25560,
+ "eval_loss": 3.4714,
+ "elapsed_s": 105800.2
+ },
+ {
+ "epoch": 89.311,
+ "step": 25632,
+ "eval_loss": 3.4611,
+ "elapsed_s": 106189.6
+ },
+ {
+ "epoch": 89.563,
+ "step": 25704,
+ "eval_loss": 3.46092,
+ "elapsed_s": 106578.4
+ },
+ {
+ "epoch": 89.814,
+ "step": 25776,
+ "eval_loss": 3.45895,
+ "elapsed_s": 106967.3
+ },
+ {
+ "epoch": 90.063,
+ "step": 25848,
+ "eval_loss": 3.46962,
+ "elapsed_s": 107351.4
+ },
+ {
+ "epoch": 90.315,
+ "step": 25920,
+ "eval_loss": 3.46026,
+ "elapsed_s": 107740.2
+ },
+ {
+ "epoch": 90.566,
+ "step": 25992,
+ "eval_loss": 3.45862,
+ "elapsed_s": 108129.3
+ },
+ {
+ "epoch": 90.818,
+ "step": 26064,
+ "eval_loss": 3.45601,
+ "elapsed_s": 108517.9
+ },
+ {
+ "epoch": 91.066,
+ "step": 26136,
+ "eval_loss": 3.47264,
+ "elapsed_s": 108904.2
+ },
+ {
+ "epoch": 91.318,
+ "step": 26208,
+ "eval_loss": 3.45937,
+ "elapsed_s": 109293.0
+ },
+ {
+ "epoch": 91.57,
+ "step": 26280,
+ "eval_loss": 3.45704,
+ "elapsed_s": 109681.7
+ },
+ {
+ "epoch": 91.821,
+ "step": 26352,
+ "eval_loss": 3.45522,
+ "elapsed_s": 110071.1
+ },
+ {
+ "epoch": 92.07,
+ "step": 26424,
+ "eval_loss": 3.467,
+ "elapsed_s": 110457.4
+ },
+ {
+ "epoch": 92.322,
+ "step": 26496,
+ "eval_loss": 3.45968,
+ "elapsed_s": 110846.0
+ },
+ {
+ "epoch": 92.573,
+ "step": 26568,
+ "eval_loss": 3.45643,
+ "elapsed_s": 111234.7
+ },
+ {
+ "epoch": 92.825,
+ "step": 26640,
+ "eval_loss": 3.45593,
+ "elapsed_s": 111623.4
+ },
+ {
+ "epoch": 93.073,
+ "step": 26712,
+ "eval_loss": 3.46404,
+ "elapsed_s": 112007.4
+ },
+ {
+ "epoch": 93.325,
+ "step": 26784,
+ "eval_loss": 3.45695,
+ "elapsed_s": 112397.2
+ },
+ {
+ "epoch": 93.577,
+ "step": 26856,
+ "eval_loss": 3.45681,
+ "elapsed_s": 112785.9
+ },
+ {
+ "epoch": 93.828,
+ "step": 26928,
+ "eval_loss": 3.45389,
+ "elapsed_s": 113174.6
+ },
+ {
+ "epoch": 94.077,
+ "step": 27000,
+ "eval_loss": 3.46589,
+ "elapsed_s": 113561.1
+ },
+ {
+ "epoch": 94.329,
+ "step": 27072,
+ "eval_loss": 3.4585,
+ "elapsed_s": 113949.9
+ },
+ {
+ "epoch": 94.58,
+ "step": 27144,
+ "eval_loss": 3.45663,
+ "elapsed_s": 114339.2
+ },
+ {
+ "epoch": 94.832,
+ "step": 27216,
+ "eval_loss": 3.45637,
+ "elapsed_s": 114727.9
+ },
+ {
+ "epoch": 95.08,
+ "step": 27288,
+ "eval_loss": 3.46537,
+ "elapsed_s": 115111.9
+ },
+ {
+ "epoch": 95.332,
+ "step": 27360,
+ "eval_loss": 3.45726,
+ "elapsed_s": 115501.2
+ },
+ {
+ "epoch": 95.584,
+ "step": 27432,
+ "eval_loss": 3.45674,
+ "elapsed_s": 115890.2
+ },
+ {
+ "epoch": 95.835,
+ "step": 27504,
+ "eval_loss": 3.45358,
+ "elapsed_s": 116279.6
+ },
+ {
+ "epoch": 96.084,
+ "step": 27576,
+ "eval_loss": 3.4619,
+ "elapsed_s": 116666.4
+ },
+ {
+ "epoch": 96.336,
+ "step": 27648,
+ "eval_loss": 3.45672,
+ "elapsed_s": 117055.6
+ },
+ {
+ "epoch": 96.587,
+ "step": 27720,
+ "eval_loss": 3.45462,
+ "elapsed_s": 117444.5
+ },
+ {
+ "epoch": 96.839,
+ "step": 27792,
+ "eval_loss": 3.45335,
+ "elapsed_s": 117833.6
+ },
+ {
+ "epoch": 97.087,
+ "step": 27864,
+ "eval_loss": 3.46172,
+ "elapsed_s": 118220.8
+ },
+ {
+ "epoch": 97.339,
+ "step": 27936,
+ "eval_loss": 3.4556,
+ "elapsed_s": 118609.6
+ },
+ {
+ "epoch": 97.591,
+ "step": 28008,
+ "eval_loss": 3.45384,
+ "elapsed_s": 118998.8
+ },
+ {
+ "epoch": 97.842,
+ "step": 28080,
+ "eval_loss": 3.45111,
+ "elapsed_s": 119387.8
+ },
+ {
+ "epoch": 98.091,
+ "step": 28152,
+ "eval_loss": 3.46318,
+ "elapsed_s": 119774.8
+ },
+ {
+ "epoch": 98.343,
+ "step": 28224,
+ "eval_loss": 3.45553,
+ "elapsed_s": 120164.9
+ },
+ {
+ "epoch": 98.594,
+ "step": 28296,
+ "eval_loss": 3.4537,
+ "elapsed_s": 120554.2
+ },
+ {
+ "epoch": 98.846,
+ "step": 28368,
+ "eval_loss": 3.45245,
+ "elapsed_s": 120943.3
+ },
+ {
+ "epoch": 99.094,
+ "step": 28440,
+ "eval_loss": 3.46077,
+ "elapsed_s": 121327.8
+ },
+ {
+ "epoch": 99.346,
+ "step": 28512,
+ "eval_loss": 3.45517,
+ "elapsed_s": 121717.1
+ },
+ {
+ "epoch": 99.598,
+ "step": 28584,
+ "eval_loss": 3.45381,
+ "elapsed_s": 122106.5
+ },
+ {
+ "epoch": 99.849,
+ "step": 28656,
+ "eval_loss": 3.45124,
+ "elapsed_s": 122496.0
+ },
+ {
+ "epoch": 100.0,
+ "step": 28700,
+ "eval_loss": 3.46162,
+ "elapsed_s": 122736.4
+ }
+ ]
+}
\ No newline at end of file
diff --git a/results/t2a/t2a_37M_C1_rand15.json b/results/t2a/t2a_37M_C1_rand15.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce16dc8eee98efc415459c59f768ffc4df151185
--- /dev/null
+++ b/results/t2a/t2a_37M_C1_rand15.json
@@ -0,0 +1,261 @@
+{
+ "run_name": "t2a_37M_C1_rand15",
+ "output_dir": "/home/ubuntu/workspace/data-augmentations-for-pretraining/runs/t2a/t2a_37M_C1_rand15",
+ "params_total": 145833984,
+ "params_non_embedding": 68178432,
+ "hidden_size": 512,
+ "num_layers": 20,
+ "num_heads": 4,
+ "intermediate_size": 1536,
+ "unique_tokens_m": 37,
+ "training_files_no": 2,
+ "epochs_requested": 203,
+ "learning_rate": 0.0006,
+ "weight_decay": 0.033,
+ "seed": 42,
+ "precision": "bf16",
+ "global_batch_sequences": 512,
+ "augmentation": {
+ "l2r_percent": 100.0,
+ "max_next_i": 1,
+ "next_i_weighting": "uniform",
+ "mask_percent": 0.0,
+ "random_token_percent": 15.0,
+ "psm_percent": 0.0,
+ "spm_percent": 0.0
+ },
+ "n_train_blocks": 18063,
+ "n_eval_blocks": 2441,
+ "status": "stopped_early",
+ "min_eval_loss": 4.151440620422363,
+ "best_epoch": 34.0,
+ "best_step": 1224,
+ "best_checkpoint": "/home/ubuntu/workspace/data-augmentations-for-pretraining/runs/t2a/t2a_37M_C1_rand15/checkpoint-1224",
+ "epochs_completed": 74.0,
+ "stopped_early": true,
+ "wall_time_s": 1955.2,
+ "history": [
+ {
+ "epoch": 2.0,
+ "step": 72,
+ "eval_loss": 7.5206,
+ "elapsed_s": 119.4
+ },
+ {
+ "epoch": 4.0,
+ "step": 144,
+ "eval_loss": 6.78172,
+ "elapsed_s": 171.0
+ },
+ {
+ "epoch": 6.0,
+ "step": 216,
+ "eval_loss": 6.37518,
+ "elapsed_s": 226.8
+ },
+ {
+ "epoch": 8.0,
+ "step": 288,
+ "eval_loss": 6.04518,
+ "elapsed_s": 278.6
+ },
+ {
+ "epoch": 10.0,
+ "step": 360,
+ "eval_loss": 5.70815,
+ "elapsed_s": 330.4
+ },
+ {
+ "epoch": 12.0,
+ "step": 432,
+ "eval_loss": 5.40467,
+ "elapsed_s": 382.3
+ },
+ {
+ "epoch": 14.0,
+ "step": 504,
+ "eval_loss": 5.1175,
+ "elapsed_s": 434.2
+ },
+ {
+ "epoch": 16.0,
+ "step": 576,
+ "eval_loss": 4.80114,
+ "elapsed_s": 486.1
+ },
+ {
+ "epoch": 18.0,
+ "step": 648,
+ "eval_loss": 4.59389,
+ "elapsed_s": 538.0
+ },
+ {
+ "epoch": 20.0,
+ "step": 720,
+ "eval_loss": 4.47436,
+ "elapsed_s": 589.8
+ },
+ {
+ "epoch": 22.0,
+ "step": 792,
+ "eval_loss": 4.34286,
+ "elapsed_s": 641.7
+ },
+ {
+ "epoch": 24.0,
+ "step": 864,
+ "eval_loss": 4.26268,
+ "elapsed_s": 693.6
+ },
+ {
+ "epoch": 26.0,
+ "step": 936,
+ "eval_loss": 4.21533,
+ "elapsed_s": 745.4
+ },
+ {
+ "epoch": 28.0,
+ "step": 1008,
+ "eval_loss": 4.18341,
+ "elapsed_s": 797.4
+ },
+ {
+ "epoch": 30.0,
+ "step": 1080,
+ "eval_loss": 4.16125,
+ "elapsed_s": 849.2
+ },
+ {
+ "epoch": 32.0,
+ "step": 1152,
+ "eval_loss": 4.16378,
+ "elapsed_s": 901.5
+ },
+ {
+ "epoch": 34.0,
+ "step": 1224,
+ "eval_loss": 4.15144,
+ "elapsed_s": 952.0
+ },
+ {
+ "epoch": 36.0,
+ "step": 1296,
+ "eval_loss": 4.15504,
+ "elapsed_s": 1005.6
+ },
+ {
+ "epoch": 38.0,
+ "step": 1368,
+ "eval_loss": 4.17152,
+ "elapsed_s": 1055.4
+ },
+ {
+ "epoch": 40.0,
+ "step": 1440,
+ "eval_loss": 4.17673,
+ "elapsed_s": 1105.2
+ },
+ {
+ "epoch": 42.0,
+ "step": 1512,
+ "eval_loss": 4.19451,
+ "elapsed_s": 1155.0
+ },
+ {
+ "epoch": 44.0,
+ "step": 1584,
+ "eval_loss": 4.21416,
+ "elapsed_s": 1204.9
+ },
+ {
+ "epoch": 46.0,
+ "step": 1656,
+ "eval_loss": 4.24918,
+ "elapsed_s": 1254.7
+ },
+ {
+ "epoch": 48.0,
+ "step": 1728,
+ "eval_loss": 4.26682,
+ "elapsed_s": 1304.5
+ },
+ {
+ "epoch": 50.0,
+ "step": 1800,
+ "eval_loss": 4.29837,
+ "elapsed_s": 1354.4
+ },
+ {
+ "epoch": 52.0,
+ "step": 1872,
+ "eval_loss": 4.3181,
+ "elapsed_s": 1404.2
+ },
+ {
+ "epoch": 54.0,
+ "step": 1944,
+ "eval_loss": 4.35438,
+ "elapsed_s": 1454.1
+ },
+ {
+ "epoch": 56.0,
+ "step": 2016,
+ "eval_loss": 4.38436,
+ "elapsed_s": 1503.9
+ },
+ {
+ "epoch": 58.0,
+ "step": 2088,
+ "eval_loss": 4.4114,
+ "elapsed_s": 1553.7
+ },
+ {
+ "epoch": 60.0,
+ "step": 2160,
+ "eval_loss": 4.44982,
+ "elapsed_s": 1603.9
+ },
+ {
+ "epoch": 62.0,
+ "step": 2232,
+ "eval_loss": 4.46788,
+ "elapsed_s": 1653.7
+ },
+ {
+ "epoch": 64.0,
+ "step": 2304,
+ "eval_loss": 4.50051,
+ "elapsed_s": 1703.9
+ },
+ {
+ "epoch": 66.0,
+ "step": 2376,
+ "eval_loss": 4.53415,
+ "elapsed_s": 1755.2
+ },
+ {
+ "epoch": 68.0,
+ "step": 2448,
+ "eval_loss": 4.56415,
+ "elapsed_s": 1805.7
+ },
+ {
+ "epoch": 70.0,
+ "step": 2520,
+ "eval_loss": 4.59697,
+ "elapsed_s": 1855.5
+ },
+ {
+ "epoch": 72.0,
+ "step": 2592,
+ "eval_loss": 4.63622,
+ "elapsed_s": 1905.3
+ },
+ {
+ "epoch": 74.0,
+ "step": 2664,
+ "eval_loss": 4.66356,
+ "elapsed_s": 1955.2
+ }
+ ]
+}
\ No newline at end of file
diff --git a/results/t2a/t2a_37M_C2_r2l50_i5exp.json b/results/t2a/t2a_37M_C2_r2l50_i5exp.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d035e4daa9bca52f97040ff55290939a39733d3
--- /dev/null
+++ b/results/t2a/t2a_37M_C2_r2l50_i5exp.json
@@ -0,0 +1,297 @@
+{
+ "run_name": "t2a_37M_C2_r2l50_i5exp",
+ "output_dir": "/home/ubuntu/workspace/data-augmentations-for-pretraining/runs/t2a/t2a_37M_C2_r2l50_i5exp",
+ "params_total": 145836544,
+ "params_non_embedding": 68178432,
+ "hidden_size": 512,
+ "num_layers": 20,
+ "num_heads": 4,
+ "intermediate_size": 1536,
+ "unique_tokens_m": 37,
+ "training_files_no": 2,
+ "epochs_requested": 203,
+ "learning_rate": 0.0006,
+ "weight_decay": 0.033,
+ "seed": 42,
+ "precision": "bf16",
+ "global_batch_sequences": 512,
+ "augmentation": {
+ "l2r_percent": 50.0,
+ "max_next_i": 5,
+ "next_i_weighting": "exp",
+ "mask_percent": 0.0,
+ "random_token_percent": 0.0,
+ "psm_percent": 0.0,
+ "spm_percent": 0.0
+ },
+ "n_train_blocks": 18063,
+ "n_eval_blocks": 2441,
+ "status": "stopped_early",
+ "min_eval_loss": 4.079798698425293,
+ "best_epoch": 46.0,
+ "best_step": 1656,
+ "best_checkpoint": "/home/ubuntu/workspace/data-augmentations-for-pretraining/runs/t2a/t2a_37M_C2_r2l50_i5exp/checkpoint-1656",
+ "epochs_completed": 86.0,
+ "stopped_early": true,
+ "wall_time_s": 2172.2,
+ "history": [
+ {
+ "epoch": 2.0,
+ "step": 72,
+ "eval_loss": 7.52672,
+ "elapsed_s": 116.7
+ },
+ {
+ "epoch": 4.0,
+ "step": 144,
+ "eval_loss": 7.14,
+ "elapsed_s": 166.2
+ },
+ {
+ "epoch": 6.0,
+ "step": 216,
+ "eval_loss": 6.48305,
+ "elapsed_s": 217.4
+ },
+ {
+ "epoch": 8.0,
+ "step": 288,
+ "eval_loss": 5.96312,
+ "elapsed_s": 268.4
+ },
+ {
+ "epoch": 10.0,
+ "step": 360,
+ "eval_loss": 5.58084,
+ "elapsed_s": 318.0
+ },
+ {
+ "epoch": 12.0,
+ "step": 432,
+ "eval_loss": 5.29584,
+ "elapsed_s": 367.7
+ },
+ {
+ "epoch": 14.0,
+ "step": 504,
+ "eval_loss": 5.02825,
+ "elapsed_s": 417.4
+ },
+ {
+ "epoch": 16.0,
+ "step": 576,
+ "eval_loss": 4.78162,
+ "elapsed_s": 467.0
+ },
+ {
+ "epoch": 18.0,
+ "step": 648,
+ "eval_loss": 4.59714,
+ "elapsed_s": 516.7
+ },
+ {
+ "epoch": 20.0,
+ "step": 720,
+ "eval_loss": 4.46527,
+ "elapsed_s": 566.3
+ },
+ {
+ "epoch": 22.0,
+ "step": 792,
+ "eval_loss": 4.37638,
+ "elapsed_s": 616.0
+ },
+ {
+ "epoch": 24.0,
+ "step": 864,
+ "eval_loss": 4.32115,
+ "elapsed_s": 665.6
+ },
+ {
+ "epoch": 26.0,
+ "step": 936,
+ "eval_loss": 4.25375,
+ "elapsed_s": 715.2
+ },
+ {
+ "epoch": 28.0,
+ "step": 1008,
+ "eval_loss": 4.20693,
+ "elapsed_s": 765.0
+ },
+ {
+ "epoch": 30.0,
+ "step": 1080,
+ "eval_loss": 4.58834,
+ "elapsed_s": 814.7
+ },
+ {
+ "epoch": 32.0,
+ "step": 1152,
+ "eval_loss": 4.17829,
+ "elapsed_s": 862.4
+ },
+ {
+ "epoch": 34.0,
+ "step": 1224,
+ "eval_loss": 4.13763,
+ "elapsed_s": 912.1
+ },
+ {
+ "epoch": 36.0,
+ "step": 1296,
+ "eval_loss": 4.11672,
+ "elapsed_s": 962.9
+ },
+ {
+ "epoch": 38.0,
+ "step": 1368,
+ "eval_loss": 4.10486,
+ "elapsed_s": 1014.0
+ },
+ {
+ "epoch": 40.0,
+ "step": 1440,
+ "eval_loss": 4.09814,
+ "elapsed_s": 1063.8
+ },
+ {
+ "epoch": 42.0,
+ "step": 1512,
+ "eval_loss": 4.08577,
+ "elapsed_s": 1113.4
+ },
+ {
+ "epoch": 44.0,
+ "step": 1584,
+ "eval_loss": 4.084,
+ "elapsed_s": 1163.2
+ },
+ {
+ "epoch": 46.0,
+ "step": 1656,
+ "eval_loss": 4.0798,
+ "elapsed_s": 1212.9
+ },
+ {
+ "epoch": 48.0,
+ "step": 1728,
+ "eval_loss": 4.08528,
+ "elapsed_s": 1262.6
+ },
+ {
+ "epoch": 50.0,
+ "step": 1800,
+ "eval_loss": 4.0851,
+ "elapsed_s": 1310.4
+ },
+ {
+ "epoch": 52.0,
+ "step": 1872,
+ "eval_loss": 4.09052,
+ "elapsed_s": 1358.1
+ },
+ {
+ "epoch": 54.0,
+ "step": 1944,
+ "eval_loss": 4.1025,
+ "elapsed_s": 1405.8
+ },
+ {
+ "epoch": 56.0,
+ "step": 2016,
+ "eval_loss": 4.10952,
+ "elapsed_s": 1453.5
+ },
+ {
+ "epoch": 58.0,
+ "step": 2088,
+ "eval_loss": 4.11233,
+ "elapsed_s": 1501.3
+ },
+ {
+ "epoch": 60.0,
+ "step": 2160,
+ "eval_loss": 4.13134,
+ "elapsed_s": 1549.0
+ },
+ {
+ "epoch": 62.0,
+ "step": 2232,
+ "eval_loss": 4.1438,
+ "elapsed_s": 1597.2
+ },
+ {
+ "epoch": 64.0,
+ "step": 2304,
+ "eval_loss": 4.14942,
+ "elapsed_s": 1645.0
+ },
+ {
+ "epoch": 66.0,
+ "step": 2376,
+ "eval_loss": 4.16111,
+ "elapsed_s": 1693.4
+ },
+ {
+ "epoch": 68.0,
+ "step": 2448,
+ "eval_loss": 4.17894,
+ "elapsed_s": 1742.2
+ },
+ {
+ "epoch": 70.0,
+ "step": 2520,
+ "eval_loss": 4.19261,
+ "elapsed_s": 1790.6
+ },
+ {
+ "epoch": 72.0,
+ "step": 2592,
+ "eval_loss": 4.20516,
+ "elapsed_s": 1838.3
+ },
+ {
+ "epoch": 74.0,
+ "step": 2664,
+ "eval_loss": 4.22512,
+ "elapsed_s": 1886.0
+ },
+ {
+ "epoch": 76.0,
+ "step": 2736,
+ "eval_loss": 4.24635,
+ "elapsed_s": 1933.6
+ },
+ {
+ "epoch": 78.0,
+ "step": 2808,
+ "eval_loss": 4.24745,
+ "elapsed_s": 1981.4
+ },
+ {
+ "epoch": 80.0,
+ "step": 2880,
+ "eval_loss": 4.28013,
+ "elapsed_s": 2029.1
+ },
+ {
+ "epoch": 82.0,
+ "step": 2952,
+ "eval_loss": 4.28783,
+ "elapsed_s": 2076.8
+ },
+ {
+ "epoch": 84.0,
+ "step": 3024,
+ "eval_loss": 4.30111,
+ "elapsed_s": 2124.5
+ },
+ {
+ "epoch": 86.0,
+ "step": 3096,
+ "eval_loss": 4.30909,
+ "elapsed_s": 2172.2
+ }
+ ]
+}
\ No newline at end of file
diff --git a/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/chat_template.jinja b/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+ {{- '<|im_start|>system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/config.json b/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8ca584a0751672f280da7c54ff3916cdf24d6f4
--- /dev/null
+++ b/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 512,
+ "initializer_range": 0.02,
+ "intermediate_size": 1536,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 4,
+ "num_hidden_layers": 20,
+ "num_key_value_heads": 4,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151671
+}
diff --git a/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/generation_config.json b/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/tokenizer_config.json b/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..839a3743177a3ed77e1ffbae67520e8500ca0532
--- /dev/null
+++ b/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/tokenizer_config.json
@@ -0,0 +1,19 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/trainer_state.json b/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff7bafed86cc4f6c78634ede61461aae5e6a09d8
--- /dev/null
+++ b/runs/seeds/t1a_S2_C0_baseline_s43/checkpoint-1296/trainer_state.json
@@ -0,0 +1,1088 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 18.0,
+ "eval_steps": 72,
+ "global_step": 1296,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013982958269608914,
+ "grad_norm": 2.375232696533203,
+ "learning_rate": 0.0,
+ "loss": 12.019524574279785,
+ "step": 1
+ },
+ {
+ "epoch": 0.13982958269608914,
+ "grad_norm": 1.7634280920028687,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.67848120795356,
+ "step": 10
+ },
+ {
+ "epoch": 0.2796591653921783,
+ "grad_norm": 1.667712926864624,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 10.944271087646484,
+ "step": 20
+ },
+ {
+ "epoch": 0.4194887480882674,
+ "grad_norm": 1.629617691040039,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 10.180948638916016,
+ "step": 30
+ },
+ {
+ "epoch": 0.5593183307843566,
+ "grad_norm": 1.4207147359848022,
+ "learning_rate": 0.000234,
+ "loss": 9.143335723876953,
+ "step": 40
+ },
+ {
+ "epoch": 0.6991479134804457,
+ "grad_norm": 0.7671460509300232,
+ "learning_rate": 0.000294,
+ "loss": 8.190089416503906,
+ "step": 50
+ },
+ {
+ "epoch": 0.8389774961765348,
+ "grad_norm": 0.3558426797389984,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 7.699327087402343,
+ "step": 60
+ },
+ {
+ "epoch": 0.978807078872624,
+ "grad_norm": 0.2832612991333008,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 7.554665374755859,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.476141452789307,
+ "eval_runtime": 48.7625,
+ "eval_samples_per_second": 50.059,
+ "eval_steps_per_second": 6.275,
+ "step": 72
+ },
+ {
+ "epoch": 1.1118636661568713,
+ "grad_norm": 0.6864347457885742,
+ "learning_rate": 0.000474,
+ "loss": 7.422309112548828,
+ "step": 80
+ },
+ {
+ "epoch": 1.2516932488529604,
+ "grad_norm": 0.4724583923816681,
+ "learning_rate": 0.000534,
+ "loss": 7.2266487121582035,
+ "step": 90
+ },
+ {
+ "epoch": 1.3915228315490495,
+ "grad_norm": 0.6818152666091919,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 7.08424072265625,
+ "step": 100
+ },
+ {
+ "epoch": 1.5313524142451387,
+ "grad_norm": 0.39162811636924744,
+ "learning_rate": 0.0006,
+ "loss": 6.959481048583984,
+ "step": 110
+ },
+ {
+ "epoch": 1.671181996941228,
+ "grad_norm": 0.3845437467098236,
+ "learning_rate": 0.0006,
+ "loss": 6.806480407714844,
+ "step": 120
+ },
+ {
+ "epoch": 1.811011579637317,
+ "grad_norm": 0.7419171929359436,
+ "learning_rate": 0.0006,
+ "loss": 6.664456939697265,
+ "step": 130
+ },
+ {
+ "epoch": 1.9508411623334063,
+ "grad_norm": 0.5003949999809265,
+ "learning_rate": 0.0006,
+ "loss": 6.579576110839843,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 6.517765998840332,
+ "eval_runtime": 24.7192,
+ "eval_samples_per_second": 98.749,
+ "eval_steps_per_second": 12.379,
+ "step": 144
+ },
+ {
+ "epoch": 2.0838977496176536,
+ "grad_norm": 0.642443835735321,
+ "learning_rate": 0.0006,
+ "loss": 6.481057739257812,
+ "step": 150
+ },
+ {
+ "epoch": 2.2237273323137425,
+ "grad_norm": 0.6833203434944153,
+ "learning_rate": 0.0006,
+ "loss": 6.3787578582763675,
+ "step": 160
+ },
+ {
+ "epoch": 2.363556915009832,
+ "grad_norm": 0.4382517635822296,
+ "learning_rate": 0.0006,
+ "loss": 6.3045402526855465,
+ "step": 170
+ },
+ {
+ "epoch": 2.503386497705921,
+ "grad_norm": 0.8139774203300476,
+ "learning_rate": 0.0006,
+ "loss": 6.222132873535156,
+ "step": 180
+ },
+ {
+ "epoch": 2.64321608040201,
+ "grad_norm": 1.0031579732894897,
+ "learning_rate": 0.0006,
+ "loss": 6.147603607177734,
+ "step": 190
+ },
+ {
+ "epoch": 2.783045663098099,
+ "grad_norm": 0.4443194270133972,
+ "learning_rate": 0.0006,
+ "loss": 6.073279190063476,
+ "step": 200
+ },
+ {
+ "epoch": 2.9228752457941884,
+ "grad_norm": 0.7461162805557251,
+ "learning_rate": 0.0006,
+ "loss": 6.0187641143798825,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 5.968244552612305,
+ "eval_runtime": 24.6891,
+ "eval_samples_per_second": 98.869,
+ "eval_steps_per_second": 12.394,
+ "step": 216
+ },
+ {
+ "epoch": 3.0559318330784357,
+ "grad_norm": 0.5114433169364929,
+ "learning_rate": 0.0006,
+ "loss": 5.940036773681641,
+ "step": 220
+ },
+ {
+ "epoch": 3.1957614157745247,
+ "grad_norm": 0.5054575204849243,
+ "learning_rate": 0.0006,
+ "loss": 5.8752197265625,
+ "step": 230
+ },
+ {
+ "epoch": 3.335590998470614,
+ "grad_norm": 0.4514678120613098,
+ "learning_rate": 0.0006,
+ "loss": 5.802175521850586,
+ "step": 240
+ },
+ {
+ "epoch": 3.475420581166703,
+ "grad_norm": 0.6185819506645203,
+ "learning_rate": 0.0006,
+ "loss": 5.782912063598633,
+ "step": 250
+ },
+ {
+ "epoch": 3.6152501638627923,
+ "grad_norm": 0.6814908981323242,
+ "learning_rate": 0.0006,
+ "loss": 5.707101440429687,
+ "step": 260
+ },
+ {
+ "epoch": 3.755079746558881,
+ "grad_norm": 0.3964880704879761,
+ "learning_rate": 0.0006,
+ "loss": 5.641536712646484,
+ "step": 270
+ },
+ {
+ "epoch": 3.8949093292549706,
+ "grad_norm": 0.8126519918441772,
+ "learning_rate": 0.0006,
+ "loss": 5.608086776733399,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 5.57916259765625,
+ "eval_runtime": 24.698,
+ "eval_samples_per_second": 98.834,
+ "eval_steps_per_second": 12.39,
+ "step": 288
+ },
+ {
+ "epoch": 4.027965916539218,
+ "grad_norm": 0.3748660981655121,
+ "learning_rate": 0.0006,
+ "loss": 5.567253875732422,
+ "step": 290
+ },
+ {
+ "epoch": 4.167795499235307,
+ "grad_norm": 0.2246149182319641,
+ "learning_rate": 0.0006,
+ "loss": 5.454840850830078,
+ "step": 300
+ },
+ {
+ "epoch": 4.307625081931396,
+ "grad_norm": 0.863617479801178,
+ "learning_rate": 0.0006,
+ "loss": 5.407225799560547,
+ "step": 310
+ },
+ {
+ "epoch": 4.447454664627485,
+ "grad_norm": 0.5173997282981873,
+ "learning_rate": 0.0006,
+ "loss": 5.418251419067383,
+ "step": 320
+ },
+ {
+ "epoch": 4.587284247323574,
+ "grad_norm": 0.2761935293674469,
+ "learning_rate": 0.0006,
+ "loss": 5.336494064331054,
+ "step": 330
+ },
+ {
+ "epoch": 4.727113830019664,
+ "grad_norm": 0.3900691568851471,
+ "learning_rate": 0.0006,
+ "loss": 5.250099182128906,
+ "step": 340
+ },
+ {
+ "epoch": 4.866943412715752,
+ "grad_norm": 1.011061429977417,
+ "learning_rate": 0.0006,
+ "loss": 5.256734848022461,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.32449254393577576,
+ "learning_rate": 0.0006,
+ "loss": 5.193204498291015,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 5.218024730682373,
+ "eval_runtime": 24.7064,
+ "eval_samples_per_second": 98.8,
+ "eval_steps_per_second": 12.385,
+ "step": 360
+ },
+ {
+ "epoch": 5.139829582696089,
+ "grad_norm": 0.4058181345462799,
+ "learning_rate": 0.0006,
+ "loss": 5.101683807373047,
+ "step": 370
+ },
+ {
+ "epoch": 5.279659165392179,
+ "grad_norm": 0.4626823961734772,
+ "learning_rate": 0.0006,
+ "loss": 5.042427825927734,
+ "step": 380
+ },
+ {
+ "epoch": 5.419488748088267,
+ "grad_norm": 0.6983886957168579,
+ "learning_rate": 0.0006,
+ "loss": 5.01214828491211,
+ "step": 390
+ },
+ {
+ "epoch": 5.5593183307843566,
+ "grad_norm": 0.5250049233436584,
+ "learning_rate": 0.0006,
+ "loss": 4.970829010009766,
+ "step": 400
+ },
+ {
+ "epoch": 5.699147913480446,
+ "grad_norm": 0.2453128844499588,
+ "learning_rate": 0.0006,
+ "loss": 4.912003707885742,
+ "step": 410
+ },
+ {
+ "epoch": 5.838977496176534,
+ "grad_norm": 0.40166300535202026,
+ "learning_rate": 0.0006,
+ "loss": 4.8406837463378904,
+ "step": 420
+ },
+ {
+ "epoch": 5.978807078872624,
+ "grad_norm": 0.7126420140266418,
+ "learning_rate": 0.0006,
+ "loss": 4.814286804199218,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 4.916359901428223,
+ "eval_runtime": 24.7188,
+ "eval_samples_per_second": 98.751,
+ "eval_steps_per_second": 12.379,
+ "step": 432
+ },
+ {
+ "epoch": 6.1118636661568715,
+ "grad_norm": 0.5457494258880615,
+ "learning_rate": 0.0006,
+ "loss": 4.790187072753906,
+ "step": 440
+ },
+ {
+ "epoch": 6.251693248852961,
+ "grad_norm": 0.3876442313194275,
+ "learning_rate": 0.0006,
+ "loss": 4.717232131958008,
+ "step": 450
+ },
+ {
+ "epoch": 6.391522831549049,
+ "grad_norm": 0.2710684835910797,
+ "learning_rate": 0.0006,
+ "loss": 4.655804443359375,
+ "step": 460
+ },
+ {
+ "epoch": 6.531352414245139,
+ "grad_norm": 0.746433436870575,
+ "learning_rate": 0.0006,
+ "loss": 4.613198471069336,
+ "step": 470
+ },
+ {
+ "epoch": 6.671181996941228,
+ "grad_norm": 0.3904099762439728,
+ "learning_rate": 0.0006,
+ "loss": 4.601712799072265,
+ "step": 480
+ },
+ {
+ "epoch": 6.811011579637317,
+ "grad_norm": 0.30078476667404175,
+ "learning_rate": 0.0006,
+ "loss": 4.543019866943359,
+ "step": 490
+ },
+ {
+ "epoch": 6.950841162333406,
+ "grad_norm": 0.3962862193584442,
+ "learning_rate": 0.0006,
+ "loss": 4.50672607421875,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 4.624927043914795,
+ "eval_runtime": 24.6964,
+ "eval_samples_per_second": 98.84,
+ "eval_steps_per_second": 12.39,
+ "step": 504
+ },
+ {
+ "epoch": 7.083897749617654,
+ "grad_norm": 0.6839303374290466,
+ "learning_rate": 0.0006,
+ "loss": 4.478728485107422,
+ "step": 510
+ },
+ {
+ "epoch": 7.223727332313743,
+ "grad_norm": 0.2952784597873688,
+ "learning_rate": 0.0006,
+ "loss": 4.462466812133789,
+ "step": 520
+ },
+ {
+ "epoch": 7.3635569150098314,
+ "grad_norm": 0.27230408787727356,
+ "learning_rate": 0.0006,
+ "loss": 4.410188293457031,
+ "step": 530
+ },
+ {
+ "epoch": 7.503386497705921,
+ "grad_norm": 0.2467435598373413,
+ "learning_rate": 0.0006,
+ "loss": 4.3608451843261715,
+ "step": 540
+ },
+ {
+ "epoch": 7.64321608040201,
+ "grad_norm": 0.305284321308136,
+ "learning_rate": 0.0006,
+ "loss": 4.338183975219726,
+ "step": 550
+ },
+ {
+ "epoch": 7.7830456630980995,
+ "grad_norm": 0.34834012389183044,
+ "learning_rate": 0.0006,
+ "loss": 4.3151611328125,
+ "step": 560
+ },
+ {
+ "epoch": 7.922875245794188,
+ "grad_norm": 0.5147038698196411,
+ "learning_rate": 0.0006,
+ "loss": 4.308943176269532,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.424359321594238,
+ "eval_runtime": 24.6813,
+ "eval_samples_per_second": 98.901,
+ "eval_steps_per_second": 12.398,
+ "step": 576
+ },
+ {
+ "epoch": 8.055931833078436,
+ "grad_norm": 0.5936110019683838,
+ "learning_rate": 0.0006,
+ "loss": 4.267201232910156,
+ "step": 580
+ },
+ {
+ "epoch": 8.195761415774525,
+ "grad_norm": 0.5244451761245728,
+ "learning_rate": 0.0006,
+ "loss": 4.256524276733399,
+ "step": 590
+ },
+ {
+ "epoch": 8.335590998470614,
+ "grad_norm": 0.2731720805168152,
+ "learning_rate": 0.0006,
+ "loss": 4.217560958862305,
+ "step": 600
+ },
+ {
+ "epoch": 8.475420581166704,
+ "grad_norm": 0.20850954949855804,
+ "learning_rate": 0.0006,
+ "loss": 4.187934494018554,
+ "step": 610
+ },
+ {
+ "epoch": 8.615250163862791,
+ "grad_norm": 0.41143301129341125,
+ "learning_rate": 0.0006,
+ "loss": 4.159894180297852,
+ "step": 620
+ },
+ {
+ "epoch": 8.75507974655888,
+ "grad_norm": 0.4033993184566498,
+ "learning_rate": 0.0006,
+ "loss": 4.162160110473633,
+ "step": 630
+ },
+ {
+ "epoch": 8.89490932925497,
+ "grad_norm": 0.3558467626571655,
+ "learning_rate": 0.0006,
+ "loss": 4.138478851318359,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.279713153839111,
+ "eval_runtime": 24.7176,
+ "eval_samples_per_second": 98.756,
+ "eval_steps_per_second": 12.38,
+ "step": 648
+ },
+ {
+ "epoch": 9.027965916539218,
+ "grad_norm": 0.3516102731227875,
+ "learning_rate": 0.0006,
+ "loss": 4.09784164428711,
+ "step": 650
+ },
+ {
+ "epoch": 9.167795499235307,
+ "grad_norm": 0.319232314825058,
+ "learning_rate": 0.0006,
+ "loss": 4.041664123535156,
+ "step": 660
+ },
+ {
+ "epoch": 9.307625081931397,
+ "grad_norm": 0.5206624865531921,
+ "learning_rate": 0.0006,
+ "loss": 4.048720169067383,
+ "step": 670
+ },
+ {
+ "epoch": 9.447454664627486,
+ "grad_norm": 0.30111628770828247,
+ "learning_rate": 0.0006,
+ "loss": 4.0331066131591795,
+ "step": 680
+ },
+ {
+ "epoch": 9.587284247323574,
+ "grad_norm": 0.2599193751811981,
+ "learning_rate": 0.0006,
+ "loss": 4.033816528320313,
+ "step": 690
+ },
+ {
+ "epoch": 9.727113830019663,
+ "grad_norm": 0.22139449417591095,
+ "learning_rate": 0.0006,
+ "loss": 4.011519241333008,
+ "step": 700
+ },
+ {
+ "epoch": 9.866943412715752,
+ "grad_norm": 0.3313088119029999,
+ "learning_rate": 0.0006,
+ "loss": 3.998866653442383,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.3502883315086365,
+ "learning_rate": 0.0006,
+ "loss": 3.9937400817871094,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.189582347869873,
+ "eval_runtime": 24.7323,
+ "eval_samples_per_second": 98.697,
+ "eval_steps_per_second": 12.372,
+ "step": 720
+ },
+ {
+ "epoch": 10.13982958269609,
+ "grad_norm": 0.6041761040687561,
+ "learning_rate": 0.0006,
+ "loss": 3.936572265625,
+ "step": 730
+ },
+ {
+ "epoch": 10.279659165392179,
+ "grad_norm": 0.5447992086410522,
+ "learning_rate": 0.0006,
+ "loss": 3.939463424682617,
+ "step": 740
+ },
+ {
+ "epoch": 10.419488748088268,
+ "grad_norm": 0.3177001476287842,
+ "learning_rate": 0.0006,
+ "loss": 3.9437026977539062,
+ "step": 750
+ },
+ {
+ "epoch": 10.559318330784357,
+ "grad_norm": 0.22915370762348175,
+ "learning_rate": 0.0006,
+ "loss": 3.9254680633544923,
+ "step": 760
+ },
+ {
+ "epoch": 10.699147913480445,
+ "grad_norm": 0.20068734884262085,
+ "learning_rate": 0.0006,
+ "loss": 3.9024749755859376,
+ "step": 770
+ },
+ {
+ "epoch": 10.838977496176534,
+ "grad_norm": 0.2246491014957428,
+ "learning_rate": 0.0006,
+ "loss": 3.883301544189453,
+ "step": 780
+ },
+ {
+ "epoch": 10.978807078872624,
+ "grad_norm": 0.23191338777542114,
+ "learning_rate": 0.0006,
+ "loss": 3.8875961303710938,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.113893508911133,
+ "eval_runtime": 24.6973,
+ "eval_samples_per_second": 98.837,
+ "eval_steps_per_second": 12.39,
+ "step": 792
+ },
+ {
+ "epoch": 11.111863666156871,
+ "grad_norm": 0.5284678339958191,
+ "learning_rate": 0.0006,
+ "loss": 3.8266395568847655,
+ "step": 800
+ },
+ {
+ "epoch": 11.25169324885296,
+ "grad_norm": 0.6822974681854248,
+ "learning_rate": 0.0006,
+ "loss": 3.839751434326172,
+ "step": 810
+ },
+ {
+ "epoch": 11.39152283154905,
+ "grad_norm": 0.30802544951438904,
+ "learning_rate": 0.0006,
+ "loss": 3.858293914794922,
+ "step": 820
+ },
+ {
+ "epoch": 11.531352414245138,
+ "grad_norm": 0.22074365615844727,
+ "learning_rate": 0.0006,
+ "loss": 3.821986770629883,
+ "step": 830
+ },
+ {
+ "epoch": 11.671181996941227,
+ "grad_norm": 0.21211892366409302,
+ "learning_rate": 0.0006,
+ "loss": 3.8076541900634764,
+ "step": 840
+ },
+ {
+ "epoch": 11.811011579637317,
+ "grad_norm": 0.24717390537261963,
+ "learning_rate": 0.0006,
+ "loss": 3.8004505157470705,
+ "step": 850
+ },
+ {
+ "epoch": 11.950841162333406,
+ "grad_norm": 0.2313273549079895,
+ "learning_rate": 0.0006,
+ "loss": 3.787952423095703,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.0590410232543945,
+ "eval_runtime": 24.7106,
+ "eval_samples_per_second": 98.783,
+ "eval_steps_per_second": 12.383,
+ "step": 864
+ },
+ {
+ "epoch": 12.083897749617654,
+ "grad_norm": 0.41993117332458496,
+ "learning_rate": 0.0006,
+ "loss": 3.728977584838867,
+ "step": 870
+ },
+ {
+ "epoch": 12.223727332313743,
+ "grad_norm": 0.4242228865623474,
+ "learning_rate": 0.0006,
+ "loss": 3.7197185516357423,
+ "step": 880
+ },
+ {
+ "epoch": 12.363556915009832,
+ "grad_norm": 0.28596243262290955,
+ "learning_rate": 0.0006,
+ "loss": 3.7217967987060545,
+ "step": 890
+ },
+ {
+ "epoch": 12.503386497705922,
+ "grad_norm": 0.2717858552932739,
+ "learning_rate": 0.0006,
+ "loss": 3.705760955810547,
+ "step": 900
+ },
+ {
+ "epoch": 12.64321608040201,
+ "grad_norm": 0.2528406083583832,
+ "learning_rate": 0.0006,
+ "loss": 3.705844497680664,
+ "step": 910
+ },
+ {
+ "epoch": 12.783045663098099,
+ "grad_norm": 0.22674719989299774,
+ "learning_rate": 0.0006,
+ "loss": 3.7115554809570312,
+ "step": 920
+ },
+ {
+ "epoch": 12.922875245794188,
+ "grad_norm": 0.2811884582042694,
+ "learning_rate": 0.0006,
+ "loss": 3.707436370849609,
+ "step": 930
+ },
+ {
+ "epoch": 13.0,
+ "eval_loss": 4.025057315826416,
+ "eval_runtime": 24.7235,
+ "eval_samples_per_second": 98.732,
+ "eval_steps_per_second": 12.377,
+ "step": 936
+ },
+ {
+ "epoch": 13.055931833078436,
+ "grad_norm": 0.4466950297355652,
+ "learning_rate": 0.0006,
+ "loss": 3.659811019897461,
+ "step": 940
+ },
+ {
+ "epoch": 13.195761415774525,
+ "grad_norm": 0.6004473567008972,
+ "learning_rate": 0.0006,
+ "loss": 3.6371246337890626,
+ "step": 950
+ },
+ {
+ "epoch": 13.335590998470614,
+ "grad_norm": 0.39167705178260803,
+ "learning_rate": 0.0006,
+ "loss": 3.6598812103271485,
+ "step": 960
+ },
+ {
+ "epoch": 13.475420581166704,
+ "grad_norm": 0.2607665956020355,
+ "learning_rate": 0.0006,
+ "loss": 3.6511695861816404,
+ "step": 970
+ },
+ {
+ "epoch": 13.615250163862791,
+ "grad_norm": 0.2687254250049591,
+ "learning_rate": 0.0006,
+ "loss": 3.6365684509277343,
+ "step": 980
+ },
+ {
+ "epoch": 13.75507974655888,
+ "grad_norm": 0.23303253948688507,
+ "learning_rate": 0.0006,
+ "loss": 3.632395553588867,
+ "step": 990
+ },
+ {
+ "epoch": 13.89490932925497,
+ "grad_norm": 0.24283629655838013,
+ "learning_rate": 0.0006,
+ "loss": 3.6287868499755858,
+ "step": 1000
+ },
+ {
+ "epoch": 14.0,
+ "eval_loss": 3.995600461959839,
+ "eval_runtime": 24.7632,
+ "eval_samples_per_second": 98.574,
+ "eval_steps_per_second": 12.357,
+ "step": 1008
+ },
+ {
+ "epoch": 14.027965916539218,
+ "grad_norm": 0.3419814109802246,
+ "learning_rate": 0.0006,
+ "loss": 3.609127426147461,
+ "step": 1010
+ },
+ {
+ "epoch": 14.167795499235307,
+ "grad_norm": 0.5565268397331238,
+ "learning_rate": 0.0006,
+ "loss": 3.5457992553710938,
+ "step": 1020
+ },
+ {
+ "epoch": 14.307625081931397,
+ "grad_norm": 0.4364655613899231,
+ "learning_rate": 0.0006,
+ "loss": 3.5700809478759767,
+ "step": 1030
+ },
+ {
+ "epoch": 14.447454664627486,
+ "grad_norm": 0.3026326894760132,
+ "learning_rate": 0.0006,
+ "loss": 3.573344421386719,
+ "step": 1040
+ },
+ {
+ "epoch": 14.587284247323574,
+ "grad_norm": 0.25068891048431396,
+ "learning_rate": 0.0006,
+ "loss": 3.5619869232177734,
+ "step": 1050
+ },
+ {
+ "epoch": 14.727113830019663,
+ "grad_norm": 0.23275867104530334,
+ "learning_rate": 0.0006,
+ "loss": 3.575175094604492,
+ "step": 1060
+ },
+ {
+ "epoch": 14.866943412715752,
+ "grad_norm": 0.24496883153915405,
+ "learning_rate": 0.0006,
+ "loss": 3.5660015106201173,
+ "step": 1070
+ },
+ {
+ "epoch": 15.0,
+ "grad_norm": 0.3224358856678009,
+ "learning_rate": 0.0006,
+ "loss": 3.5606307983398438,
+ "step": 1080
+ },
+ {
+ "epoch": 15.0,
+ "eval_loss": 3.9805526733398438,
+ "eval_runtime": 24.7042,
+ "eval_samples_per_second": 98.809,
+ "eval_steps_per_second": 12.387,
+ "step": 1080
+ },
+ {
+ "epoch": 15.13982958269609,
+ "grad_norm": 0.43684500455856323,
+ "learning_rate": 0.0006,
+ "loss": 3.4655994415283202,
+ "step": 1090
+ },
+ {
+ "epoch": 15.279659165392179,
+ "grad_norm": 0.41879406571388245,
+ "learning_rate": 0.0006,
+ "loss": 3.47398681640625,
+ "step": 1100
+ },
+ {
+ "epoch": 15.419488748088268,
+ "grad_norm": 0.2702624797821045,
+ "learning_rate": 0.0006,
+ "loss": 3.4843406677246094,
+ "step": 1110
+ },
+ {
+ "epoch": 15.559318330784357,
+ "grad_norm": 0.24965833127498627,
+ "learning_rate": 0.0006,
+ "loss": 3.485365295410156,
+ "step": 1120
+ },
+ {
+ "epoch": 15.699147913480445,
+ "grad_norm": 0.2912072539329529,
+ "learning_rate": 0.0006,
+ "loss": 3.4926563262939454,
+ "step": 1130
+ },
+ {
+ "epoch": 15.838977496176534,
+ "grad_norm": 0.28987565636634827,
+ "learning_rate": 0.0006,
+ "loss": 3.5066951751708983,
+ "step": 1140
+ },
+ {
+ "epoch": 15.978807078872624,
+ "grad_norm": 0.2562166154384613,
+ "learning_rate": 0.0006,
+ "loss": 3.4969253540039062,
+ "step": 1150
+ },
+ {
+ "epoch": 16.0,
+ "eval_loss": 3.9704155921936035,
+ "eval_runtime": 24.7214,
+ "eval_samples_per_second": 98.74,
+ "eval_steps_per_second": 12.378,
+ "step": 1152
+ },
+ {
+ "epoch": 16.11186366615687,
+ "grad_norm": 0.5590243339538574,
+ "learning_rate": 0.0006,
+ "loss": 3.425371551513672,
+ "step": 1160
+ },
+ {
+ "epoch": 16.25169324885296,
+ "grad_norm": 0.6208043694496155,
+ "learning_rate": 0.0006,
+ "loss": 3.4234569549560545,
+ "step": 1170
+ },
+ {
+ "epoch": 16.39152283154905,
+ "grad_norm": 0.5976361632347107,
+ "learning_rate": 0.0006,
+ "loss": 3.466999053955078,
+ "step": 1180
+ },
+ {
+ "epoch": 16.531352414245138,
+ "grad_norm": 0.3120851516723633,
+ "learning_rate": 0.0006,
+ "loss": 3.4879119873046873,
+ "step": 1190
+ },
+ {
+ "epoch": 16.67118199694123,
+ "grad_norm": 0.9178828597068787,
+ "learning_rate": 0.0006,
+ "loss": 3.7457725524902346,
+ "step": 1200
+ },
+ {
+ "epoch": 16.811011579637317,
+ "grad_norm": 0.5263794660568237,
+ "learning_rate": 0.0006,
+ "loss": 3.6625492095947267,
+ "step": 1210
+ },
+ {
+ "epoch": 16.950841162333408,
+ "grad_norm": 0.896612823009491,
+ "learning_rate": 0.0006,
+ "loss": 3.5959259033203126,
+ "step": 1220
+ },
+ {
+ "epoch": 17.0,
+ "eval_loss": 4.032678127288818,
+ "eval_runtime": 24.7339,
+ "eval_samples_per_second": 98.691,
+ "eval_steps_per_second": 12.372,
+ "step": 1224
+ },
+ {
+ "epoch": 17.083897749617652,
+ "grad_norm": 0.3626440763473511,
+ "learning_rate": 0.0006,
+ "loss": 3.5578964233398436,
+ "step": 1230
+ },
+ {
+ "epoch": 17.223727332313743,
+ "grad_norm": 1.1442128419876099,
+ "learning_rate": 0.0006,
+ "loss": 3.4709918975830076,
+ "step": 1240
+ },
+ {
+ "epoch": 17.36355691500983,
+ "grad_norm": 0.6263262629508972,
+ "learning_rate": 0.0006,
+ "loss": 3.4783706665039062,
+ "step": 1250
+ },
+ {
+ "epoch": 17.50338649770592,
+ "grad_norm": 0.2557455003261566,
+ "learning_rate": 0.0006,
+ "loss": 3.4899002075195313,
+ "step": 1260
+ },
+ {
+ "epoch": 17.64321608040201,
+ "grad_norm": 0.21957996487617493,
+ "learning_rate": 0.0006,
+ "loss": 3.4568851470947264,
+ "step": 1270
+ },
+ {
+ "epoch": 17.7830456630981,
+ "grad_norm": 0.22016368806362152,
+ "learning_rate": 0.0006,
+ "loss": 3.4597808837890627,
+ "step": 1280
+ },
+ {
+ "epoch": 17.922875245794188,
+ "grad_norm": 0.2185015082359314,
+ "learning_rate": 0.0006,
+ "loss": 3.4343223571777344,
+ "step": 1290
+ },
+ {
+ "epoch": 18.0,
+ "eval_loss": 3.9495179653167725,
+ "eval_runtime": 24.7036,
+ "eval_samples_per_second": 98.811,
+ "eval_steps_per_second": 12.387,
+ "step": 1296
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 7200,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 100,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 5.521383256039096e+17,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/seeds/t1a_S2_C0_baseline_s43/train.log b/runs/seeds/t1a_S2_C0_baseline_s43/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..0f0141645098522456759e0f5978a13e8a590faf
--- /dev/null
+++ b/runs/seeds/t1a_S2_C0_baseline_s43/train.log
@@ -0,0 +1,6750 @@
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 100
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 8
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 64
+[transformers] Total optimization steps = 7,200
+[transformers] Number of trainable parameters = 145,833,984
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: setting up run wdti1pc2
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260916_172608-wdti1pc2
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t1a_S2_C0_baseline_s43
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/wdti1pc2
+
0%| | 0/7200 [00:00system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/seeds/t1a_S2_C0_baseline_s44/checkpoint-1296/config.json b/runs/seeds/t1a_S2_C0_baseline_s44/checkpoint-1296/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8ca584a0751672f280da7c54ff3916cdf24d6f4
--- /dev/null
+++ b/runs/seeds/t1a_S2_C0_baseline_s44/checkpoint-1296/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 512,
+ "initializer_range": 0.02,
+ "intermediate_size": 1536,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 4,
+ "num_hidden_layers": 20,
+ "num_key_value_heads": 4,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151671
+}
diff --git a/runs/seeds/t1a_S2_C0_baseline_s44/checkpoint-1296/generation_config.json b/runs/seeds/t1a_S2_C0_baseline_s44/checkpoint-1296/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/seeds/t1a_S2_C0_baseline_s44/checkpoint-1296/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/seeds/t1a_S2_C0_baseline_s44/checkpoint-1296/tokenizer_config.json b/runs/seeds/t1a_S2_C0_baseline_s44/checkpoint-1296/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..839a3743177a3ed77e1ffbae67520e8500ca0532
--- /dev/null
+++ b/runs/seeds/t1a_S2_C0_baseline_s44/checkpoint-1296/tokenizer_config.json
@@ -0,0 +1,19 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/seeds/t1a_S2_C0_baseline_s44/checkpoint-1296/trainer_state.json b/runs/seeds/t1a_S2_C0_baseline_s44/checkpoint-1296/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0464220bc8bb7b7a329a63bfedb48442dbe7113
--- /dev/null
+++ b/runs/seeds/t1a_S2_C0_baseline_s44/checkpoint-1296/trainer_state.json
@@ -0,0 +1,1088 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 18.0,
+ "eval_steps": 72,
+ "global_step": 1296,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013982958269608914,
+ "grad_norm": 1.8334938287734985,
+ "learning_rate": 0.0,
+ "loss": 12.028677940368652,
+ "step": 1
+ },
+ {
+ "epoch": 0.13982958269608914,
+ "grad_norm": 1.7492808103561401,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.741778055826822,
+ "step": 10
+ },
+ {
+ "epoch": 0.2796591653921783,
+ "grad_norm": 1.683693528175354,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 11.03465576171875,
+ "step": 20
+ },
+ {
+ "epoch": 0.4194887480882674,
+ "grad_norm": 1.643078327178955,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 10.271154022216797,
+ "step": 30
+ },
+ {
+ "epoch": 0.5593183307843566,
+ "grad_norm": 1.477729320526123,
+ "learning_rate": 0.000234,
+ "loss": 9.23305892944336,
+ "step": 40
+ },
+ {
+ "epoch": 0.6991479134804457,
+ "grad_norm": 1.0173728466033936,
+ "learning_rate": 0.000294,
+ "loss": 8.255239868164063,
+ "step": 50
+ },
+ {
+ "epoch": 0.8389774961765348,
+ "grad_norm": 0.32469049096107483,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 7.690232849121093,
+ "step": 60
+ },
+ {
+ "epoch": 0.978807078872624,
+ "grad_norm": 0.297742635011673,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 7.537940216064453,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.465608596801758,
+ "eval_runtime": 48.4696,
+ "eval_samples_per_second": 50.361,
+ "eval_steps_per_second": 6.313,
+ "step": 72
+ },
+ {
+ "epoch": 1.1118636661568713,
+ "grad_norm": 0.6259150505065918,
+ "learning_rate": 0.000474,
+ "loss": 7.364170837402344,
+ "step": 80
+ },
+ {
+ "epoch": 1.2516932488529604,
+ "grad_norm": 0.451576828956604,
+ "learning_rate": 0.000534,
+ "loss": 7.139842224121094,
+ "step": 90
+ },
+ {
+ "epoch": 1.3915228315490495,
+ "grad_norm": 0.7480049729347229,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 7.011451721191406,
+ "step": 100
+ },
+ {
+ "epoch": 1.5313524142451387,
+ "grad_norm": 0.2749083638191223,
+ "learning_rate": 0.0006,
+ "loss": 6.867728424072266,
+ "step": 110
+ },
+ {
+ "epoch": 1.671181996941228,
+ "grad_norm": 0.278017520904541,
+ "learning_rate": 0.0006,
+ "loss": 6.725141143798828,
+ "step": 120
+ },
+ {
+ "epoch": 1.811011579637317,
+ "grad_norm": 0.5721828937530518,
+ "learning_rate": 0.0006,
+ "loss": 6.6303955078125,
+ "step": 130
+ },
+ {
+ "epoch": 1.9508411623334063,
+ "grad_norm": 0.7588843107223511,
+ "learning_rate": 0.0006,
+ "loss": 6.543772888183594,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 6.4951677322387695,
+ "eval_runtime": 24.6893,
+ "eval_samples_per_second": 98.869,
+ "eval_steps_per_second": 12.394,
+ "step": 144
+ },
+ {
+ "epoch": 2.0838977496176536,
+ "grad_norm": 0.8217096924781799,
+ "learning_rate": 0.0006,
+ "loss": 6.437287139892578,
+ "step": 150
+ },
+ {
+ "epoch": 2.2237273323137425,
+ "grad_norm": 0.5574862957000732,
+ "learning_rate": 0.0006,
+ "loss": 6.343304061889649,
+ "step": 160
+ },
+ {
+ "epoch": 2.363556915009832,
+ "grad_norm": 0.4718858301639557,
+ "learning_rate": 0.0006,
+ "loss": 6.291109848022461,
+ "step": 170
+ },
+ {
+ "epoch": 2.503386497705921,
+ "grad_norm": 0.4388621151447296,
+ "learning_rate": 0.0006,
+ "loss": 6.204110717773437,
+ "step": 180
+ },
+ {
+ "epoch": 2.64321608040201,
+ "grad_norm": 0.734935998916626,
+ "learning_rate": 0.0006,
+ "loss": 6.147423553466797,
+ "step": 190
+ },
+ {
+ "epoch": 2.783045663098099,
+ "grad_norm": 0.6569491624832153,
+ "learning_rate": 0.0006,
+ "loss": 6.0961254119873045,
+ "step": 200
+ },
+ {
+ "epoch": 2.9228752457941884,
+ "grad_norm": 0.6080736517906189,
+ "learning_rate": 0.0006,
+ "loss": 6.012973403930664,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 6.053739547729492,
+ "eval_runtime": 24.7238,
+ "eval_samples_per_second": 98.731,
+ "eval_steps_per_second": 12.377,
+ "step": 216
+ },
+ {
+ "epoch": 3.0559318330784357,
+ "grad_norm": 0.8229308724403381,
+ "learning_rate": 0.0006,
+ "loss": 5.987596893310547,
+ "step": 220
+ },
+ {
+ "epoch": 3.1957614157745247,
+ "grad_norm": 0.5357822179794312,
+ "learning_rate": 0.0006,
+ "loss": 5.903062057495117,
+ "step": 230
+ },
+ {
+ "epoch": 3.335590998470614,
+ "grad_norm": 0.8974347710609436,
+ "learning_rate": 0.0006,
+ "loss": 5.848337173461914,
+ "step": 240
+ },
+ {
+ "epoch": 3.475420581166703,
+ "grad_norm": 0.9083970785140991,
+ "learning_rate": 0.0006,
+ "loss": 5.803395843505859,
+ "step": 250
+ },
+ {
+ "epoch": 3.6152501638627923,
+ "grad_norm": 0.2732692062854767,
+ "learning_rate": 0.0006,
+ "loss": 5.728768157958984,
+ "step": 260
+ },
+ {
+ "epoch": 3.755079746558881,
+ "grad_norm": 0.6739834547042847,
+ "learning_rate": 0.0006,
+ "loss": 5.677251815795898,
+ "step": 270
+ },
+ {
+ "epoch": 3.8949093292549706,
+ "grad_norm": 0.47453534603118896,
+ "learning_rate": 0.0006,
+ "loss": 5.678570175170899,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 5.6061882972717285,
+ "eval_runtime": 24.6354,
+ "eval_samples_per_second": 99.085,
+ "eval_steps_per_second": 12.421,
+ "step": 288
+ },
+ {
+ "epoch": 4.027965916539218,
+ "grad_norm": 0.492387592792511,
+ "learning_rate": 0.0006,
+ "loss": 5.571780014038086,
+ "step": 290
+ },
+ {
+ "epoch": 4.167795499235307,
+ "grad_norm": 0.7294889092445374,
+ "learning_rate": 0.0006,
+ "loss": 5.528846359252929,
+ "step": 300
+ },
+ {
+ "epoch": 4.307625081931396,
+ "grad_norm": 0.6767717599868774,
+ "learning_rate": 0.0006,
+ "loss": 5.4938301086425785,
+ "step": 310
+ },
+ {
+ "epoch": 4.447454664627485,
+ "grad_norm": 0.3465206027030945,
+ "learning_rate": 0.0006,
+ "loss": 5.418291091918945,
+ "step": 320
+ },
+ {
+ "epoch": 4.587284247323574,
+ "grad_norm": 0.600513756275177,
+ "learning_rate": 0.0006,
+ "loss": 5.369298934936523,
+ "step": 330
+ },
+ {
+ "epoch": 4.727113830019664,
+ "grad_norm": 0.6565630435943604,
+ "learning_rate": 0.0006,
+ "loss": 5.37110595703125,
+ "step": 340
+ },
+ {
+ "epoch": 4.866943412715752,
+ "grad_norm": 0.3675380349159241,
+ "learning_rate": 0.0006,
+ "loss": 5.31409912109375,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.32460615038871765,
+ "learning_rate": 0.0006,
+ "loss": 5.232719039916992,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 5.27567720413208,
+ "eval_runtime": 24.623,
+ "eval_samples_per_second": 99.135,
+ "eval_steps_per_second": 12.427,
+ "step": 360
+ },
+ {
+ "epoch": 5.139829582696089,
+ "grad_norm": 1.3546526432037354,
+ "learning_rate": 0.0006,
+ "loss": 5.197935485839844,
+ "step": 370
+ },
+ {
+ "epoch": 5.279659165392179,
+ "grad_norm": 0.5418855547904968,
+ "learning_rate": 0.0006,
+ "loss": 5.192986297607422,
+ "step": 380
+ },
+ {
+ "epoch": 5.419488748088267,
+ "grad_norm": 0.2906038165092468,
+ "learning_rate": 0.0006,
+ "loss": 5.116780471801758,
+ "step": 390
+ },
+ {
+ "epoch": 5.5593183307843566,
+ "grad_norm": 0.5237523317337036,
+ "learning_rate": 0.0006,
+ "loss": 5.065874099731445,
+ "step": 400
+ },
+ {
+ "epoch": 5.699147913480446,
+ "grad_norm": 0.6911041140556335,
+ "learning_rate": 0.0006,
+ "loss": 5.017612075805664,
+ "step": 410
+ },
+ {
+ "epoch": 5.838977496176534,
+ "grad_norm": 0.596603512763977,
+ "learning_rate": 0.0006,
+ "loss": 4.994011688232422,
+ "step": 420
+ },
+ {
+ "epoch": 5.978807078872624,
+ "grad_norm": 0.4822384715080261,
+ "learning_rate": 0.0006,
+ "loss": 4.941204833984375,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 4.992903709411621,
+ "eval_runtime": 24.629,
+ "eval_samples_per_second": 99.111,
+ "eval_steps_per_second": 12.424,
+ "step": 432
+ },
+ {
+ "epoch": 6.1118636661568715,
+ "grad_norm": 0.5521188378334045,
+ "learning_rate": 0.0006,
+ "loss": 4.863029861450196,
+ "step": 440
+ },
+ {
+ "epoch": 6.251693248852961,
+ "grad_norm": 0.8977644443511963,
+ "learning_rate": 0.0006,
+ "loss": 4.885480117797852,
+ "step": 450
+ },
+ {
+ "epoch": 6.391522831549049,
+ "grad_norm": 0.41182735562324524,
+ "learning_rate": 0.0006,
+ "loss": 4.869599533081055,
+ "step": 460
+ },
+ {
+ "epoch": 6.531352414245139,
+ "grad_norm": 0.28195977210998535,
+ "learning_rate": 0.0006,
+ "loss": 4.786893463134765,
+ "step": 470
+ },
+ {
+ "epoch": 6.671181996941228,
+ "grad_norm": 0.3179889917373657,
+ "learning_rate": 0.0006,
+ "loss": 4.732822418212891,
+ "step": 480
+ },
+ {
+ "epoch": 6.811011579637317,
+ "grad_norm": 0.5683456063270569,
+ "learning_rate": 0.0006,
+ "loss": 4.696435928344727,
+ "step": 490
+ },
+ {
+ "epoch": 6.950841162333406,
+ "grad_norm": 1.0174541473388672,
+ "learning_rate": 0.0006,
+ "loss": 4.730963516235351,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 4.801668643951416,
+ "eval_runtime": 24.6324,
+ "eval_samples_per_second": 99.097,
+ "eval_steps_per_second": 12.423,
+ "step": 504
+ },
+ {
+ "epoch": 7.083897749617654,
+ "grad_norm": 0.5913375020027161,
+ "learning_rate": 0.0006,
+ "loss": 4.675345993041992,
+ "step": 510
+ },
+ {
+ "epoch": 7.223727332313743,
+ "grad_norm": 0.313446044921875,
+ "learning_rate": 0.0006,
+ "loss": 4.596883392333984,
+ "step": 520
+ },
+ {
+ "epoch": 7.3635569150098314,
+ "grad_norm": 0.35517868399620056,
+ "learning_rate": 0.0006,
+ "loss": 4.55975570678711,
+ "step": 530
+ },
+ {
+ "epoch": 7.503386497705921,
+ "grad_norm": 0.7493647336959839,
+ "learning_rate": 0.0006,
+ "loss": 4.526339340209961,
+ "step": 540
+ },
+ {
+ "epoch": 7.64321608040201,
+ "grad_norm": 0.5046770572662354,
+ "learning_rate": 0.0006,
+ "loss": 4.523983764648437,
+ "step": 550
+ },
+ {
+ "epoch": 7.7830456630980995,
+ "grad_norm": 0.3778055012226105,
+ "learning_rate": 0.0006,
+ "loss": 4.490219497680664,
+ "step": 560
+ },
+ {
+ "epoch": 7.922875245794188,
+ "grad_norm": 0.34399765729904175,
+ "learning_rate": 0.0006,
+ "loss": 4.451963806152344,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.559064865112305,
+ "eval_runtime": 24.6431,
+ "eval_samples_per_second": 99.054,
+ "eval_steps_per_second": 12.417,
+ "step": 576
+ },
+ {
+ "epoch": 8.055931833078436,
+ "grad_norm": 0.8399431705474854,
+ "learning_rate": 0.0006,
+ "loss": 4.402481460571289,
+ "step": 580
+ },
+ {
+ "epoch": 8.195761415774525,
+ "grad_norm": 0.8644046187400818,
+ "learning_rate": 0.0006,
+ "loss": 4.428829956054687,
+ "step": 590
+ },
+ {
+ "epoch": 8.335590998470614,
+ "grad_norm": 0.394546777009964,
+ "learning_rate": 0.0006,
+ "loss": 4.410576629638672,
+ "step": 600
+ },
+ {
+ "epoch": 8.475420581166704,
+ "grad_norm": 0.20794852077960968,
+ "learning_rate": 0.0006,
+ "loss": 4.3493598937988285,
+ "step": 610
+ },
+ {
+ "epoch": 8.615250163862791,
+ "grad_norm": 0.2713848352432251,
+ "learning_rate": 0.0006,
+ "loss": 4.322542190551758,
+ "step": 620
+ },
+ {
+ "epoch": 8.75507974655888,
+ "grad_norm": 0.4981835186481476,
+ "learning_rate": 0.0006,
+ "loss": 4.294540786743164,
+ "step": 630
+ },
+ {
+ "epoch": 8.89490932925497,
+ "grad_norm": 0.699718177318573,
+ "learning_rate": 0.0006,
+ "loss": 4.30582275390625,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.429544925689697,
+ "eval_runtime": 24.6377,
+ "eval_samples_per_second": 99.076,
+ "eval_steps_per_second": 12.42,
+ "step": 648
+ },
+ {
+ "epoch": 9.027965916539218,
+ "grad_norm": 0.3532475531101227,
+ "learning_rate": 0.0006,
+ "loss": 4.287893295288086,
+ "step": 650
+ },
+ {
+ "epoch": 9.167795499235307,
+ "grad_norm": 0.44184809923171997,
+ "learning_rate": 0.0006,
+ "loss": 4.211767578125,
+ "step": 660
+ },
+ {
+ "epoch": 9.307625081931397,
+ "grad_norm": 0.41308942437171936,
+ "learning_rate": 0.0006,
+ "loss": 4.16943473815918,
+ "step": 670
+ },
+ {
+ "epoch": 9.447454664627486,
+ "grad_norm": 0.3993234634399414,
+ "learning_rate": 0.0006,
+ "loss": 4.175564193725586,
+ "step": 680
+ },
+ {
+ "epoch": 9.587284247323574,
+ "grad_norm": 0.2485351264476776,
+ "learning_rate": 0.0006,
+ "loss": 4.166370391845703,
+ "step": 690
+ },
+ {
+ "epoch": 9.727113830019663,
+ "grad_norm": 0.32261335849761963,
+ "learning_rate": 0.0006,
+ "loss": 4.14605827331543,
+ "step": 700
+ },
+ {
+ "epoch": 9.866943412715752,
+ "grad_norm": 0.4945806562900543,
+ "learning_rate": 0.0006,
+ "loss": 4.126834487915039,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.3762032091617584,
+ "learning_rate": 0.0006,
+ "loss": 4.136430740356445,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.296835422515869,
+ "eval_runtime": 24.6855,
+ "eval_samples_per_second": 98.884,
+ "eval_steps_per_second": 12.396,
+ "step": 720
+ },
+ {
+ "epoch": 10.13982958269609,
+ "grad_norm": 0.49442607164382935,
+ "learning_rate": 0.0006,
+ "loss": 4.057977294921875,
+ "step": 730
+ },
+ {
+ "epoch": 10.279659165392179,
+ "grad_norm": 0.5792047381401062,
+ "learning_rate": 0.0006,
+ "loss": 4.059200286865234,
+ "step": 740
+ },
+ {
+ "epoch": 10.419488748088268,
+ "grad_norm": 0.2959899604320526,
+ "learning_rate": 0.0006,
+ "loss": 4.053987503051758,
+ "step": 750
+ },
+ {
+ "epoch": 10.559318330784357,
+ "grad_norm": 0.29814913868904114,
+ "learning_rate": 0.0006,
+ "loss": 4.0261188507080075,
+ "step": 760
+ },
+ {
+ "epoch": 10.699147913480445,
+ "grad_norm": 0.31743693351745605,
+ "learning_rate": 0.0006,
+ "loss": 4.025604629516602,
+ "step": 770
+ },
+ {
+ "epoch": 10.838977496176534,
+ "grad_norm": 0.3086645007133484,
+ "learning_rate": 0.0006,
+ "loss": 4.0008594512939455,
+ "step": 780
+ },
+ {
+ "epoch": 10.978807078872624,
+ "grad_norm": 0.3680846691131592,
+ "learning_rate": 0.0006,
+ "loss": 3.9998542785644533,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.209330081939697,
+ "eval_runtime": 24.7168,
+ "eval_samples_per_second": 98.759,
+ "eval_steps_per_second": 12.38,
+ "step": 792
+ },
+ {
+ "epoch": 11.111863666156871,
+ "grad_norm": 0.7032161951065063,
+ "learning_rate": 0.0006,
+ "loss": 3.9327136993408205,
+ "step": 800
+ },
+ {
+ "epoch": 11.25169324885296,
+ "grad_norm": 1.1157039403915405,
+ "learning_rate": 0.0006,
+ "loss": 4.047645187377929,
+ "step": 810
+ },
+ {
+ "epoch": 11.39152283154905,
+ "grad_norm": 1.0475879907608032,
+ "learning_rate": 0.0006,
+ "loss": 4.165793991088867,
+ "step": 820
+ },
+ {
+ "epoch": 11.531352414245138,
+ "grad_norm": 0.3133474886417389,
+ "learning_rate": 0.0006,
+ "loss": 4.097693634033203,
+ "step": 830
+ },
+ {
+ "epoch": 11.671181996941227,
+ "grad_norm": 0.20984508097171783,
+ "learning_rate": 0.0006,
+ "loss": 4.026138687133789,
+ "step": 840
+ },
+ {
+ "epoch": 11.811011579637317,
+ "grad_norm": 0.16996830701828003,
+ "learning_rate": 0.0006,
+ "loss": 3.960520553588867,
+ "step": 850
+ },
+ {
+ "epoch": 11.950841162333406,
+ "grad_norm": 0.18071477115154266,
+ "learning_rate": 0.0006,
+ "loss": 3.938106918334961,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.157978057861328,
+ "eval_runtime": 24.7353,
+ "eval_samples_per_second": 98.685,
+ "eval_steps_per_second": 12.371,
+ "step": 864
+ },
+ {
+ "epoch": 12.083897749617654,
+ "grad_norm": 0.21923750638961792,
+ "learning_rate": 0.0006,
+ "loss": 3.872021484375,
+ "step": 870
+ },
+ {
+ "epoch": 12.223727332313743,
+ "grad_norm": 0.28477150201797485,
+ "learning_rate": 0.0006,
+ "loss": 3.8500225067138674,
+ "step": 880
+ },
+ {
+ "epoch": 12.363556915009832,
+ "grad_norm": 0.24383853375911713,
+ "learning_rate": 0.0006,
+ "loss": 3.838703918457031,
+ "step": 890
+ },
+ {
+ "epoch": 12.503386497705922,
+ "grad_norm": 0.24215921759605408,
+ "learning_rate": 0.0006,
+ "loss": 3.828591156005859,
+ "step": 900
+ },
+ {
+ "epoch": 12.64321608040201,
+ "grad_norm": 0.23639480769634247,
+ "learning_rate": 0.0006,
+ "loss": 3.8390918731689454,
+ "step": 910
+ },
+ {
+ "epoch": 12.783045663098099,
+ "grad_norm": 0.2590061128139496,
+ "learning_rate": 0.0006,
+ "loss": 3.829446029663086,
+ "step": 920
+ },
+ {
+ "epoch": 12.922875245794188,
+ "grad_norm": 0.2876424789428711,
+ "learning_rate": 0.0006,
+ "loss": 3.8259723663330076,
+ "step": 930
+ },
+ {
+ "epoch": 13.0,
+ "eval_loss": 4.091265678405762,
+ "eval_runtime": 24.6917,
+ "eval_samples_per_second": 98.859,
+ "eval_steps_per_second": 12.393,
+ "step": 936
+ },
+ {
+ "epoch": 13.055931833078436,
+ "grad_norm": 0.4924207925796509,
+ "learning_rate": 0.0006,
+ "loss": 3.780427169799805,
+ "step": 940
+ },
+ {
+ "epoch": 13.195761415774525,
+ "grad_norm": 0.49917492270469666,
+ "learning_rate": 0.0006,
+ "loss": 3.755952835083008,
+ "step": 950
+ },
+ {
+ "epoch": 13.335590998470614,
+ "grad_norm": 0.40393784642219543,
+ "learning_rate": 0.0006,
+ "loss": 3.7526317596435548,
+ "step": 960
+ },
+ {
+ "epoch": 13.475420581166704,
+ "grad_norm": 0.2594180405139923,
+ "learning_rate": 0.0006,
+ "loss": 3.759630584716797,
+ "step": 970
+ },
+ {
+ "epoch": 13.615250163862791,
+ "grad_norm": 0.221713587641716,
+ "learning_rate": 0.0006,
+ "loss": 3.7400432586669923,
+ "step": 980
+ },
+ {
+ "epoch": 13.75507974655888,
+ "grad_norm": 0.2620590031147003,
+ "learning_rate": 0.0006,
+ "loss": 3.7549789428710936,
+ "step": 990
+ },
+ {
+ "epoch": 13.89490932925497,
+ "grad_norm": 0.2159755527973175,
+ "learning_rate": 0.0006,
+ "loss": 3.740696334838867,
+ "step": 1000
+ },
+ {
+ "epoch": 14.0,
+ "eval_loss": 4.0555949211120605,
+ "eval_runtime": 24.6884,
+ "eval_samples_per_second": 98.872,
+ "eval_steps_per_second": 12.394,
+ "step": 1008
+ },
+ {
+ "epoch": 14.027965916539218,
+ "grad_norm": 0.4056479036808014,
+ "learning_rate": 0.0006,
+ "loss": 3.720285415649414,
+ "step": 1010
+ },
+ {
+ "epoch": 14.167795499235307,
+ "grad_norm": 0.46656379103660583,
+ "learning_rate": 0.0006,
+ "loss": 3.665673828125,
+ "step": 1020
+ },
+ {
+ "epoch": 14.307625081931397,
+ "grad_norm": 0.3969050347805023,
+ "learning_rate": 0.0006,
+ "loss": 3.662765884399414,
+ "step": 1030
+ },
+ {
+ "epoch": 14.447454664627486,
+ "grad_norm": 0.3152327537536621,
+ "learning_rate": 0.0006,
+ "loss": 3.669620895385742,
+ "step": 1040
+ },
+ {
+ "epoch": 14.587284247323574,
+ "grad_norm": 0.2888181507587433,
+ "learning_rate": 0.0006,
+ "loss": 3.663343048095703,
+ "step": 1050
+ },
+ {
+ "epoch": 14.727113830019663,
+ "grad_norm": 0.255076140165329,
+ "learning_rate": 0.0006,
+ "loss": 3.662271499633789,
+ "step": 1060
+ },
+ {
+ "epoch": 14.866943412715752,
+ "grad_norm": 0.2713582217693329,
+ "learning_rate": 0.0006,
+ "loss": 3.6715457916259764,
+ "step": 1070
+ },
+ {
+ "epoch": 15.0,
+ "grad_norm": 0.2856007218360901,
+ "learning_rate": 0.0006,
+ "loss": 3.6595436096191407,
+ "step": 1080
+ },
+ {
+ "epoch": 15.0,
+ "eval_loss": 4.026495933532715,
+ "eval_runtime": 24.696,
+ "eval_samples_per_second": 98.842,
+ "eval_steps_per_second": 12.391,
+ "step": 1080
+ },
+ {
+ "epoch": 15.13982958269609,
+ "grad_norm": 0.6364638209342957,
+ "learning_rate": 0.0006,
+ "loss": 3.5743358612060545,
+ "step": 1090
+ },
+ {
+ "epoch": 15.279659165392179,
+ "grad_norm": 0.544431746006012,
+ "learning_rate": 0.0006,
+ "loss": 3.612900161743164,
+ "step": 1100
+ },
+ {
+ "epoch": 15.419488748088268,
+ "grad_norm": 0.5008583068847656,
+ "learning_rate": 0.0006,
+ "loss": 3.631303405761719,
+ "step": 1110
+ },
+ {
+ "epoch": 15.559318330784357,
+ "grad_norm": 0.2894301414489746,
+ "learning_rate": 0.0006,
+ "loss": 3.614338684082031,
+ "step": 1120
+ },
+ {
+ "epoch": 15.699147913480445,
+ "grad_norm": 0.25058484077453613,
+ "learning_rate": 0.0006,
+ "loss": 3.5952014923095703,
+ "step": 1130
+ },
+ {
+ "epoch": 15.838977496176534,
+ "grad_norm": 0.22610515356063843,
+ "learning_rate": 0.0006,
+ "loss": 3.597136306762695,
+ "step": 1140
+ },
+ {
+ "epoch": 15.978807078872624,
+ "grad_norm": 0.2551708221435547,
+ "learning_rate": 0.0006,
+ "loss": 3.6054733276367186,
+ "step": 1150
+ },
+ {
+ "epoch": 16.0,
+ "eval_loss": 4.008790493011475,
+ "eval_runtime": 24.6865,
+ "eval_samples_per_second": 98.88,
+ "eval_steps_per_second": 12.395,
+ "step": 1152
+ },
+ {
+ "epoch": 16.11186366615687,
+ "grad_norm": 0.3378470242023468,
+ "learning_rate": 0.0006,
+ "loss": 3.5192214965820314,
+ "step": 1160
+ },
+ {
+ "epoch": 16.25169324885296,
+ "grad_norm": 0.4257342517375946,
+ "learning_rate": 0.0006,
+ "loss": 3.4944427490234373,
+ "step": 1170
+ },
+ {
+ "epoch": 16.39152283154905,
+ "grad_norm": 0.30550745129585266,
+ "learning_rate": 0.0006,
+ "loss": 3.5084255218505858,
+ "step": 1180
+ },
+ {
+ "epoch": 16.531352414245138,
+ "grad_norm": 0.28210559487342834,
+ "learning_rate": 0.0006,
+ "loss": 3.525606155395508,
+ "step": 1190
+ },
+ {
+ "epoch": 16.67118199694123,
+ "grad_norm": 0.27610498666763306,
+ "learning_rate": 0.0006,
+ "loss": 3.5220409393310548,
+ "step": 1200
+ },
+ {
+ "epoch": 16.811011579637317,
+ "grad_norm": 0.30206093192100525,
+ "learning_rate": 0.0006,
+ "loss": 3.537362289428711,
+ "step": 1210
+ },
+ {
+ "epoch": 16.950841162333408,
+ "grad_norm": 0.28662291169166565,
+ "learning_rate": 0.0006,
+ "loss": 3.537405014038086,
+ "step": 1220
+ },
+ {
+ "epoch": 17.0,
+ "eval_loss": 3.998685359954834,
+ "eval_runtime": 24.6826,
+ "eval_samples_per_second": 98.896,
+ "eval_steps_per_second": 12.397,
+ "step": 1224
+ },
+ {
+ "epoch": 17.083897749617652,
+ "grad_norm": 0.5890938639640808,
+ "learning_rate": 0.0006,
+ "loss": 3.471303176879883,
+ "step": 1230
+ },
+ {
+ "epoch": 17.223727332313743,
+ "grad_norm": 0.6294847130775452,
+ "learning_rate": 0.0006,
+ "loss": 3.449800491333008,
+ "step": 1240
+ },
+ {
+ "epoch": 17.36355691500983,
+ "grad_norm": 0.5015324950218201,
+ "learning_rate": 0.0006,
+ "loss": 3.508951187133789,
+ "step": 1250
+ },
+ {
+ "epoch": 17.50338649770592,
+ "grad_norm": 0.34270888566970825,
+ "learning_rate": 0.0006,
+ "loss": 3.4922630310058596,
+ "step": 1260
+ },
+ {
+ "epoch": 17.64321608040201,
+ "grad_norm": 0.26928767561912537,
+ "learning_rate": 0.0006,
+ "loss": 3.4804683685302735,
+ "step": 1270
+ },
+ {
+ "epoch": 17.7830456630981,
+ "grad_norm": 0.24288047850131989,
+ "learning_rate": 0.0006,
+ "loss": 3.489974594116211,
+ "step": 1280
+ },
+ {
+ "epoch": 17.922875245794188,
+ "grad_norm": 0.24163252115249634,
+ "learning_rate": 0.0006,
+ "loss": 3.4853389739990233,
+ "step": 1290
+ },
+ {
+ "epoch": 18.0,
+ "eval_loss": 3.993281602859497,
+ "eval_runtime": 24.6897,
+ "eval_samples_per_second": 98.867,
+ "eval_steps_per_second": 12.394,
+ "step": 1296
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 7200,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 100,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 5.521383256039096e+17,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/seeds/t1a_S2_C0_baseline_s44/run.lock b/runs/seeds/t1a_S2_C0_baseline_s44/run.lock
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/runs/seeds/t1a_S2_C0_baseline_s44/train.log b/runs/seeds/t1a_S2_C0_baseline_s44/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..1359656c654d1742f1e3d7cc03a01a256a0eb4bc
--- /dev/null
+++ b/runs/seeds/t1a_S2_C0_baseline_s44/train.log
@@ -0,0 +1,6762 @@
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 100
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 8
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 64
+[transformers] Total optimization steps = 7,200
+[transformers] Number of trainable parameters = 145,833,984
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: setting up run p4p21aci
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260917_173452-p4p21aci
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t1a_S2_C0_baseline_s44
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/p4p21aci
+
0%| | 0/7200 [00:00system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/seeds/t1a_S2_C1_rand15_s43/checkpoint-2016/config.json b/runs/seeds/t1a_S2_C1_rand15_s43/checkpoint-2016/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8ca584a0751672f280da7c54ff3916cdf24d6f4
--- /dev/null
+++ b/runs/seeds/t1a_S2_C1_rand15_s43/checkpoint-2016/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 512,
+ "initializer_range": 0.02,
+ "intermediate_size": 1536,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 4,
+ "num_hidden_layers": 20,
+ "num_key_value_heads": 4,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151671
+}
diff --git a/runs/seeds/t1a_S2_C1_rand15_s43/checkpoint-2016/generation_config.json b/runs/seeds/t1a_S2_C1_rand15_s43/checkpoint-2016/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/seeds/t1a_S2_C1_rand15_s43/checkpoint-2016/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/seeds/t1a_S2_C1_rand15_s43/checkpoint-2016/tokenizer_config.json b/runs/seeds/t1a_S2_C1_rand15_s43/checkpoint-2016/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..839a3743177a3ed77e1ffbae67520e8500ca0532
--- /dev/null
+++ b/runs/seeds/t1a_S2_C1_rand15_s43/checkpoint-2016/tokenizer_config.json
@@ -0,0 +1,19 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/seeds/t1a_S2_C1_rand15_s43/checkpoint-2016/trainer_state.json b/runs/seeds/t1a_S2_C1_rand15_s43/checkpoint-2016/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..af298da4cb40752271bce609e30609d642f34610
--- /dev/null
+++ b/runs/seeds/t1a_S2_C1_rand15_s43/checkpoint-2016/trainer_state.json
@@ -0,0 +1,1672 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 28.0,
+ "eval_steps": 72,
+ "global_step": 2016,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013982958269608914,
+ "grad_norm": 1.6578404903411865,
+ "learning_rate": 0.0,
+ "loss": 12.035407066345215,
+ "step": 1
+ },
+ {
+ "epoch": 0.13982958269608914,
+ "grad_norm": 1.7482141256332397,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.754433525933159,
+ "step": 10
+ },
+ {
+ "epoch": 0.2796591653921783,
+ "grad_norm": 1.6698530912399292,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 11.057215118408203,
+ "step": 20
+ },
+ {
+ "epoch": 0.4194887480882674,
+ "grad_norm": 1.647369623184204,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 10.309608459472656,
+ "step": 30
+ },
+ {
+ "epoch": 0.5593183307843566,
+ "grad_norm": 1.492473840713501,
+ "learning_rate": 0.000234,
+ "loss": 9.269853973388672,
+ "step": 40
+ },
+ {
+ "epoch": 0.6991479134804457,
+ "grad_norm": 0.8945454955101013,
+ "learning_rate": 0.000294,
+ "loss": 8.268359375,
+ "step": 50
+ },
+ {
+ "epoch": 0.8389774961765348,
+ "grad_norm": 0.2286079078912735,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 7.705179595947266,
+ "step": 60
+ },
+ {
+ "epoch": 0.978807078872624,
+ "grad_norm": 0.8780362010002136,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 7.5606849670410154,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.511532306671143,
+ "eval_runtime": 49.2326,
+ "eval_samples_per_second": 49.581,
+ "eval_steps_per_second": 6.215,
+ "step": 72
+ },
+ {
+ "epoch": 1.1118636661568713,
+ "grad_norm": 0.7163554430007935,
+ "learning_rate": 0.000474,
+ "loss": 7.4222900390625,
+ "step": 80
+ },
+ {
+ "epoch": 1.2516932488529604,
+ "grad_norm": 0.47508037090301514,
+ "learning_rate": 0.000534,
+ "loss": 7.255441284179687,
+ "step": 90
+ },
+ {
+ "epoch": 1.3915228315490495,
+ "grad_norm": 0.3592031002044678,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 7.138349914550782,
+ "step": 100
+ },
+ {
+ "epoch": 1.5313524142451387,
+ "grad_norm": 0.2927418649196625,
+ "learning_rate": 0.0006,
+ "loss": 6.995320892333984,
+ "step": 110
+ },
+ {
+ "epoch": 1.671181996941228,
+ "grad_norm": 0.4029527008533478,
+ "learning_rate": 0.0006,
+ "loss": 6.885779571533203,
+ "step": 120
+ },
+ {
+ "epoch": 1.811011579637317,
+ "grad_norm": 0.8407447934150696,
+ "learning_rate": 0.0006,
+ "loss": 6.772330474853516,
+ "step": 130
+ },
+ {
+ "epoch": 1.9508411623334063,
+ "grad_norm": 0.3003447651863098,
+ "learning_rate": 0.0006,
+ "loss": 6.6887054443359375,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 6.89017391204834,
+ "eval_runtime": 24.6668,
+ "eval_samples_per_second": 98.959,
+ "eval_steps_per_second": 12.405,
+ "step": 144
+ },
+ {
+ "epoch": 2.0838977496176536,
+ "grad_norm": 0.6871715784072876,
+ "learning_rate": 0.0006,
+ "loss": 6.61016845703125,
+ "step": 150
+ },
+ {
+ "epoch": 2.2237273323137425,
+ "grad_norm": 0.6107402443885803,
+ "learning_rate": 0.0006,
+ "loss": 6.575196075439453,
+ "step": 160
+ },
+ {
+ "epoch": 2.363556915009832,
+ "grad_norm": 0.42170074582099915,
+ "learning_rate": 0.0006,
+ "loss": 6.5037590026855465,
+ "step": 170
+ },
+ {
+ "epoch": 2.503386497705921,
+ "grad_norm": 0.3174009919166565,
+ "learning_rate": 0.0006,
+ "loss": 6.435280609130859,
+ "step": 180
+ },
+ {
+ "epoch": 2.64321608040201,
+ "grad_norm": 0.8898131251335144,
+ "learning_rate": 0.0006,
+ "loss": 6.389397430419922,
+ "step": 190
+ },
+ {
+ "epoch": 2.783045663098099,
+ "grad_norm": 0.4237094819545746,
+ "learning_rate": 0.0006,
+ "loss": 6.326219940185547,
+ "step": 200
+ },
+ {
+ "epoch": 2.9228752457941884,
+ "grad_norm": 0.6562232971191406,
+ "learning_rate": 0.0006,
+ "loss": 6.264762115478516,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 6.516037464141846,
+ "eval_runtime": 24.6602,
+ "eval_samples_per_second": 98.985,
+ "eval_steps_per_second": 12.409,
+ "step": 216
+ },
+ {
+ "epoch": 3.0559318330784357,
+ "grad_norm": 1.1316200494766235,
+ "learning_rate": 0.0006,
+ "loss": 6.220421600341797,
+ "step": 220
+ },
+ {
+ "epoch": 3.1957614157745247,
+ "grad_norm": 0.7726063132286072,
+ "learning_rate": 0.0006,
+ "loss": 6.176301574707031,
+ "step": 230
+ },
+ {
+ "epoch": 3.335590998470614,
+ "grad_norm": 0.6438485383987427,
+ "learning_rate": 0.0006,
+ "loss": 6.119086074829101,
+ "step": 240
+ },
+ {
+ "epoch": 3.475420581166703,
+ "grad_norm": 0.9242563843727112,
+ "learning_rate": 0.0006,
+ "loss": 6.083078384399414,
+ "step": 250
+ },
+ {
+ "epoch": 3.6152501638627923,
+ "grad_norm": 0.3441637456417084,
+ "learning_rate": 0.0006,
+ "loss": 6.025389099121094,
+ "step": 260
+ },
+ {
+ "epoch": 3.755079746558881,
+ "grad_norm": 0.6736602783203125,
+ "learning_rate": 0.0006,
+ "loss": 6.006442260742188,
+ "step": 270
+ },
+ {
+ "epoch": 3.8949093292549706,
+ "grad_norm": 0.3657565414905548,
+ "learning_rate": 0.0006,
+ "loss": 5.946316909790039,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 6.073910713195801,
+ "eval_runtime": 24.663,
+ "eval_samples_per_second": 98.974,
+ "eval_steps_per_second": 12.407,
+ "step": 288
+ },
+ {
+ "epoch": 4.027965916539218,
+ "grad_norm": 0.5377328991889954,
+ "learning_rate": 0.0006,
+ "loss": 5.90440673828125,
+ "step": 290
+ },
+ {
+ "epoch": 4.167795499235307,
+ "grad_norm": 0.6109813451766968,
+ "learning_rate": 0.0006,
+ "loss": 5.841012954711914,
+ "step": 300
+ },
+ {
+ "epoch": 4.307625081931396,
+ "grad_norm": 0.28243085741996765,
+ "learning_rate": 0.0006,
+ "loss": 5.792630004882812,
+ "step": 310
+ },
+ {
+ "epoch": 4.447454664627485,
+ "grad_norm": 1.0198432207107544,
+ "learning_rate": 0.0006,
+ "loss": 5.762587356567383,
+ "step": 320
+ },
+ {
+ "epoch": 4.587284247323574,
+ "grad_norm": 0.8833813667297363,
+ "learning_rate": 0.0006,
+ "loss": 5.768093109130859,
+ "step": 330
+ },
+ {
+ "epoch": 4.727113830019664,
+ "grad_norm": 0.36912551522254944,
+ "learning_rate": 0.0006,
+ "loss": 5.703379058837891,
+ "step": 340
+ },
+ {
+ "epoch": 4.866943412715752,
+ "grad_norm": 0.3014258146286011,
+ "learning_rate": 0.0006,
+ "loss": 5.659283447265625,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.6336300373077393,
+ "learning_rate": 0.0006,
+ "loss": 5.610838317871094,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 5.661688327789307,
+ "eval_runtime": 24.6642,
+ "eval_samples_per_second": 98.969,
+ "eval_steps_per_second": 12.407,
+ "step": 360
+ },
+ {
+ "epoch": 5.139829582696089,
+ "grad_norm": 0.4710099995136261,
+ "learning_rate": 0.0006,
+ "loss": 5.582360458374024,
+ "step": 370
+ },
+ {
+ "epoch": 5.279659165392179,
+ "grad_norm": 0.6385644674301147,
+ "learning_rate": 0.0006,
+ "loss": 5.561551284790039,
+ "step": 380
+ },
+ {
+ "epoch": 5.419488748088267,
+ "grad_norm": 0.3658435344696045,
+ "learning_rate": 0.0006,
+ "loss": 5.513004302978516,
+ "step": 390
+ },
+ {
+ "epoch": 5.5593183307843566,
+ "grad_norm": 0.27364179491996765,
+ "learning_rate": 0.0006,
+ "loss": 5.450923156738281,
+ "step": 400
+ },
+ {
+ "epoch": 5.699147913480446,
+ "grad_norm": 0.636008083820343,
+ "learning_rate": 0.0006,
+ "loss": 5.421584701538086,
+ "step": 410
+ },
+ {
+ "epoch": 5.838977496176534,
+ "grad_norm": 0.6311033964157104,
+ "learning_rate": 0.0006,
+ "loss": 5.437456130981445,
+ "step": 420
+ },
+ {
+ "epoch": 5.978807078872624,
+ "grad_norm": 0.3370312452316284,
+ "learning_rate": 0.0006,
+ "loss": 5.4034912109375,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 5.343160152435303,
+ "eval_runtime": 24.6667,
+ "eval_samples_per_second": 98.959,
+ "eval_steps_per_second": 12.405,
+ "step": 432
+ },
+ {
+ "epoch": 6.1118636661568715,
+ "grad_norm": 0.590539813041687,
+ "learning_rate": 0.0006,
+ "loss": 5.333251953125,
+ "step": 440
+ },
+ {
+ "epoch": 6.251693248852961,
+ "grad_norm": 0.7476915717124939,
+ "learning_rate": 0.0006,
+ "loss": 5.294948959350586,
+ "step": 450
+ },
+ {
+ "epoch": 6.391522831549049,
+ "grad_norm": 0.5279034972190857,
+ "learning_rate": 0.0006,
+ "loss": 5.2989250183105465,
+ "step": 460
+ },
+ {
+ "epoch": 6.531352414245139,
+ "grad_norm": 0.2214391827583313,
+ "learning_rate": 0.0006,
+ "loss": 5.245613479614258,
+ "step": 470
+ },
+ {
+ "epoch": 6.671181996941228,
+ "grad_norm": 0.3393130302429199,
+ "learning_rate": 0.0006,
+ "loss": 5.1989295959472654,
+ "step": 480
+ },
+ {
+ "epoch": 6.811011579637317,
+ "grad_norm": 0.5216792225837708,
+ "learning_rate": 0.0006,
+ "loss": 5.173765182495117,
+ "step": 490
+ },
+ {
+ "epoch": 6.950841162333406,
+ "grad_norm": 0.5537815690040588,
+ "learning_rate": 0.0006,
+ "loss": 5.1631732940673825,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 5.065269470214844,
+ "eval_runtime": 24.6657,
+ "eval_samples_per_second": 98.963,
+ "eval_steps_per_second": 12.406,
+ "step": 504
+ },
+ {
+ "epoch": 7.083897749617654,
+ "grad_norm": 0.6344583034515381,
+ "learning_rate": 0.0006,
+ "loss": 5.125476837158203,
+ "step": 510
+ },
+ {
+ "epoch": 7.223727332313743,
+ "grad_norm": 0.35630273818969727,
+ "learning_rate": 0.0006,
+ "loss": 5.086610412597656,
+ "step": 520
+ },
+ {
+ "epoch": 7.3635569150098314,
+ "grad_norm": 0.23968952894210815,
+ "learning_rate": 0.0006,
+ "loss": 5.038983535766602,
+ "step": 530
+ },
+ {
+ "epoch": 7.503386497705921,
+ "grad_norm": 0.30835428833961487,
+ "learning_rate": 0.0006,
+ "loss": 4.989981460571289,
+ "step": 540
+ },
+ {
+ "epoch": 7.64321608040201,
+ "grad_norm": 0.595817506313324,
+ "learning_rate": 0.0006,
+ "loss": 4.975848007202148,
+ "step": 550
+ },
+ {
+ "epoch": 7.7830456630980995,
+ "grad_norm": 0.9597760438919067,
+ "learning_rate": 0.0006,
+ "loss": 5.027736663818359,
+ "step": 560
+ },
+ {
+ "epoch": 7.922875245794188,
+ "grad_norm": 0.5566813349723816,
+ "learning_rate": 0.0006,
+ "loss": 5.021109008789063,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.765532970428467,
+ "eval_runtime": 24.6558,
+ "eval_samples_per_second": 99.003,
+ "eval_steps_per_second": 12.411,
+ "step": 576
+ },
+ {
+ "epoch": 8.055931833078436,
+ "grad_norm": 0.2652179002761841,
+ "learning_rate": 0.0006,
+ "loss": 4.936310958862305,
+ "step": 580
+ },
+ {
+ "epoch": 8.195761415774525,
+ "grad_norm": 0.21949534118175507,
+ "learning_rate": 0.0006,
+ "loss": 4.8749347686767575,
+ "step": 590
+ },
+ {
+ "epoch": 8.335590998470614,
+ "grad_norm": 0.2743476629257202,
+ "learning_rate": 0.0006,
+ "loss": 4.841225051879883,
+ "step": 600
+ },
+ {
+ "epoch": 8.475420581166704,
+ "grad_norm": 0.2697311043739319,
+ "learning_rate": 0.0006,
+ "loss": 4.817669677734375,
+ "step": 610
+ },
+ {
+ "epoch": 8.615250163862791,
+ "grad_norm": 0.34996268153190613,
+ "learning_rate": 0.0006,
+ "loss": 4.801321411132813,
+ "step": 620
+ },
+ {
+ "epoch": 8.75507974655888,
+ "grad_norm": 0.35798320174217224,
+ "learning_rate": 0.0006,
+ "loss": 4.7877647399902346,
+ "step": 630
+ },
+ {
+ "epoch": 8.89490932925497,
+ "grad_norm": 0.34078916907310486,
+ "learning_rate": 0.0006,
+ "loss": 4.764733505249024,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.520884037017822,
+ "eval_runtime": 24.6804,
+ "eval_samples_per_second": 98.904,
+ "eval_steps_per_second": 12.398,
+ "step": 648
+ },
+ {
+ "epoch": 9.027965916539218,
+ "grad_norm": 0.37346968054771423,
+ "learning_rate": 0.0006,
+ "loss": 4.720360565185547,
+ "step": 650
+ },
+ {
+ "epoch": 9.167795499235307,
+ "grad_norm": 0.29245972633361816,
+ "learning_rate": 0.0006,
+ "loss": 4.686754989624023,
+ "step": 660
+ },
+ {
+ "epoch": 9.307625081931397,
+ "grad_norm": 0.31028804183006287,
+ "learning_rate": 0.0006,
+ "loss": 4.675933074951172,
+ "step": 670
+ },
+ {
+ "epoch": 9.447454664627486,
+ "grad_norm": 0.39809322357177734,
+ "learning_rate": 0.0006,
+ "loss": 4.6483501434326175,
+ "step": 680
+ },
+ {
+ "epoch": 9.587284247323574,
+ "grad_norm": 0.5012796521186829,
+ "learning_rate": 0.0006,
+ "loss": 4.6853588104248045,
+ "step": 690
+ },
+ {
+ "epoch": 9.727113830019663,
+ "grad_norm": 0.3657071590423584,
+ "learning_rate": 0.0006,
+ "loss": 4.665429306030274,
+ "step": 700
+ },
+ {
+ "epoch": 9.866943412715752,
+ "grad_norm": 0.2372114062309265,
+ "learning_rate": 0.0006,
+ "loss": 4.630749893188477,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.22031980752944946,
+ "learning_rate": 0.0006,
+ "loss": 4.60394172668457,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.376460075378418,
+ "eval_runtime": 24.6651,
+ "eval_samples_per_second": 98.966,
+ "eval_steps_per_second": 12.406,
+ "step": 720
+ },
+ {
+ "epoch": 10.13982958269609,
+ "grad_norm": 0.3931390941143036,
+ "learning_rate": 0.0006,
+ "loss": 4.5581298828125,
+ "step": 730
+ },
+ {
+ "epoch": 10.279659165392179,
+ "grad_norm": 0.35578975081443787,
+ "learning_rate": 0.0006,
+ "loss": 4.531334304809571,
+ "step": 740
+ },
+ {
+ "epoch": 10.419488748088268,
+ "grad_norm": 0.483913779258728,
+ "learning_rate": 0.0006,
+ "loss": 4.5454059600830075,
+ "step": 750
+ },
+ {
+ "epoch": 10.559318330784357,
+ "grad_norm": 0.2588062286376953,
+ "learning_rate": 0.0006,
+ "loss": 4.553162384033203,
+ "step": 760
+ },
+ {
+ "epoch": 10.699147913480445,
+ "grad_norm": 0.21749193966388702,
+ "learning_rate": 0.0006,
+ "loss": 4.521955490112305,
+ "step": 770
+ },
+ {
+ "epoch": 10.838977496176534,
+ "grad_norm": 0.26840701699256897,
+ "learning_rate": 0.0006,
+ "loss": 4.499534225463867,
+ "step": 780
+ },
+ {
+ "epoch": 10.978807078872624,
+ "grad_norm": 0.2636498510837555,
+ "learning_rate": 0.0006,
+ "loss": 4.502541351318359,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.270487308502197,
+ "eval_runtime": 24.6651,
+ "eval_samples_per_second": 98.966,
+ "eval_steps_per_second": 12.406,
+ "step": 792
+ },
+ {
+ "epoch": 11.111863666156871,
+ "grad_norm": 0.44413068890571594,
+ "learning_rate": 0.0006,
+ "loss": 4.453483200073242,
+ "step": 800
+ },
+ {
+ "epoch": 11.25169324885296,
+ "grad_norm": 0.4273896813392639,
+ "learning_rate": 0.0006,
+ "loss": 4.455494689941406,
+ "step": 810
+ },
+ {
+ "epoch": 11.39152283154905,
+ "grad_norm": 0.4153364896774292,
+ "learning_rate": 0.0006,
+ "loss": 4.457523345947266,
+ "step": 820
+ },
+ {
+ "epoch": 11.531352414245138,
+ "grad_norm": 0.24993915855884552,
+ "learning_rate": 0.0006,
+ "loss": 4.441752243041992,
+ "step": 830
+ },
+ {
+ "epoch": 11.671181996941227,
+ "grad_norm": 0.19715352356433868,
+ "learning_rate": 0.0006,
+ "loss": 4.423892974853516,
+ "step": 840
+ },
+ {
+ "epoch": 11.811011579637317,
+ "grad_norm": 0.19220297038555145,
+ "learning_rate": 0.0006,
+ "loss": 4.412079238891602,
+ "step": 850
+ },
+ {
+ "epoch": 11.950841162333406,
+ "grad_norm": 0.22718091309070587,
+ "learning_rate": 0.0006,
+ "loss": 4.393754577636718,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.184975624084473,
+ "eval_runtime": 24.6892,
+ "eval_samples_per_second": 98.869,
+ "eval_steps_per_second": 12.394,
+ "step": 864
+ },
+ {
+ "epoch": 12.083897749617654,
+ "grad_norm": 0.32650327682495117,
+ "learning_rate": 0.0006,
+ "loss": 4.3578346252441404,
+ "step": 870
+ },
+ {
+ "epoch": 12.223727332313743,
+ "grad_norm": 0.3262386918067932,
+ "learning_rate": 0.0006,
+ "loss": 4.356068801879883,
+ "step": 880
+ },
+ {
+ "epoch": 12.363556915009832,
+ "grad_norm": 0.31983649730682373,
+ "learning_rate": 0.0006,
+ "loss": 4.357526779174805,
+ "step": 890
+ },
+ {
+ "epoch": 12.503386497705922,
+ "grad_norm": 0.2256602644920349,
+ "learning_rate": 0.0006,
+ "loss": 4.340785980224609,
+ "step": 900
+ },
+ {
+ "epoch": 12.64321608040201,
+ "grad_norm": 0.29326891899108887,
+ "learning_rate": 0.0006,
+ "loss": 4.333747863769531,
+ "step": 910
+ },
+ {
+ "epoch": 12.783045663098099,
+ "grad_norm": 0.3772833049297333,
+ "learning_rate": 0.0006,
+ "loss": 4.340696334838867,
+ "step": 920
+ },
+ {
+ "epoch": 12.922875245794188,
+ "grad_norm": 0.28702324628829956,
+ "learning_rate": 0.0006,
+ "loss": 4.332079696655273,
+ "step": 930
+ },
+ {
+ "epoch": 13.0,
+ "eval_loss": 4.1266560554504395,
+ "eval_runtime": 24.6574,
+ "eval_samples_per_second": 98.997,
+ "eval_steps_per_second": 12.41,
+ "step": 936
+ },
+ {
+ "epoch": 13.055931833078436,
+ "grad_norm": 0.2850724160671234,
+ "learning_rate": 0.0006,
+ "loss": 4.292006301879883,
+ "step": 940
+ },
+ {
+ "epoch": 13.195761415774525,
+ "grad_norm": 0.2522817552089691,
+ "learning_rate": 0.0006,
+ "loss": 4.276935577392578,
+ "step": 950
+ },
+ {
+ "epoch": 13.335590998470614,
+ "grad_norm": 0.27872705459594727,
+ "learning_rate": 0.0006,
+ "loss": 4.264604568481445,
+ "step": 960
+ },
+ {
+ "epoch": 13.475420581166704,
+ "grad_norm": 0.3768681287765503,
+ "learning_rate": 0.0006,
+ "loss": 4.271211242675781,
+ "step": 970
+ },
+ {
+ "epoch": 13.615250163862791,
+ "grad_norm": 0.2586802840232849,
+ "learning_rate": 0.0006,
+ "loss": 4.267250061035156,
+ "step": 980
+ },
+ {
+ "epoch": 13.75507974655888,
+ "grad_norm": 0.3383229374885559,
+ "learning_rate": 0.0006,
+ "loss": 4.261027908325195,
+ "step": 990
+ },
+ {
+ "epoch": 13.89490932925497,
+ "grad_norm": 0.27569469809532166,
+ "learning_rate": 0.0006,
+ "loss": 4.25892219543457,
+ "step": 1000
+ },
+ {
+ "epoch": 14.0,
+ "eval_loss": 4.0733866691589355,
+ "eval_runtime": 24.6791,
+ "eval_samples_per_second": 98.909,
+ "eval_steps_per_second": 12.399,
+ "step": 1008
+ },
+ {
+ "epoch": 14.027965916539218,
+ "grad_norm": 0.25540560483932495,
+ "learning_rate": 0.0006,
+ "loss": 4.243086242675782,
+ "step": 1010
+ },
+ {
+ "epoch": 14.167795499235307,
+ "grad_norm": 0.31541797518730164,
+ "learning_rate": 0.0006,
+ "loss": 4.202405166625977,
+ "step": 1020
+ },
+ {
+ "epoch": 14.307625081931397,
+ "grad_norm": 0.28216561675071716,
+ "learning_rate": 0.0006,
+ "loss": 4.203763580322265,
+ "step": 1030
+ },
+ {
+ "epoch": 14.447454664627486,
+ "grad_norm": 0.26779431104660034,
+ "learning_rate": 0.0006,
+ "loss": 4.212853622436524,
+ "step": 1040
+ },
+ {
+ "epoch": 14.587284247323574,
+ "grad_norm": 0.23704424500465393,
+ "learning_rate": 0.0006,
+ "loss": 4.200860595703125,
+ "step": 1050
+ },
+ {
+ "epoch": 14.727113830019663,
+ "grad_norm": 0.24000394344329834,
+ "learning_rate": 0.0006,
+ "loss": 4.214182281494141,
+ "step": 1060
+ },
+ {
+ "epoch": 14.866943412715752,
+ "grad_norm": 0.30822622776031494,
+ "learning_rate": 0.0006,
+ "loss": 4.201005554199218,
+ "step": 1070
+ },
+ {
+ "epoch": 15.0,
+ "grad_norm": 0.30915284156799316,
+ "learning_rate": 0.0006,
+ "loss": 4.193174743652344,
+ "step": 1080
+ },
+ {
+ "epoch": 15.0,
+ "eval_loss": 4.0307817459106445,
+ "eval_runtime": 24.6709,
+ "eval_samples_per_second": 98.942,
+ "eval_steps_per_second": 12.403,
+ "step": 1080
+ },
+ {
+ "epoch": 15.13982958269609,
+ "grad_norm": 0.3552546501159668,
+ "learning_rate": 0.0006,
+ "loss": 4.157976913452148,
+ "step": 1090
+ },
+ {
+ "epoch": 15.279659165392179,
+ "grad_norm": 0.29449737071990967,
+ "learning_rate": 0.0006,
+ "loss": 4.151531982421875,
+ "step": 1100
+ },
+ {
+ "epoch": 15.419488748088268,
+ "grad_norm": 0.24161840975284576,
+ "learning_rate": 0.0006,
+ "loss": 4.151103591918945,
+ "step": 1110
+ },
+ {
+ "epoch": 15.559318330784357,
+ "grad_norm": 0.23485687375068665,
+ "learning_rate": 0.0006,
+ "loss": 4.144001770019531,
+ "step": 1120
+ },
+ {
+ "epoch": 15.699147913480445,
+ "grad_norm": 0.2986936569213867,
+ "learning_rate": 0.0006,
+ "loss": 4.1452892303466795,
+ "step": 1130
+ },
+ {
+ "epoch": 15.838977496176534,
+ "grad_norm": 0.2680480182170868,
+ "learning_rate": 0.0006,
+ "loss": 4.153837203979492,
+ "step": 1140
+ },
+ {
+ "epoch": 15.978807078872624,
+ "grad_norm": 0.22058901190757751,
+ "learning_rate": 0.0006,
+ "loss": 4.134096908569336,
+ "step": 1150
+ },
+ {
+ "epoch": 16.0,
+ "eval_loss": 3.995894432067871,
+ "eval_runtime": 24.6657,
+ "eval_samples_per_second": 98.963,
+ "eval_steps_per_second": 12.406,
+ "step": 1152
+ },
+ {
+ "epoch": 16.11186366615687,
+ "grad_norm": 0.3240770697593689,
+ "learning_rate": 0.0006,
+ "loss": 4.104195785522461,
+ "step": 1160
+ },
+ {
+ "epoch": 16.25169324885296,
+ "grad_norm": 0.3424763083457947,
+ "learning_rate": 0.0006,
+ "loss": 4.092544174194336,
+ "step": 1170
+ },
+ {
+ "epoch": 16.39152283154905,
+ "grad_norm": 0.32821163535118103,
+ "learning_rate": 0.0006,
+ "loss": 4.104222106933594,
+ "step": 1180
+ },
+ {
+ "epoch": 16.531352414245138,
+ "grad_norm": 0.20734155178070068,
+ "learning_rate": 0.0006,
+ "loss": 4.110722351074219,
+ "step": 1190
+ },
+ {
+ "epoch": 16.67118199694123,
+ "grad_norm": 0.2697821855545044,
+ "learning_rate": 0.0006,
+ "loss": 4.084817504882812,
+ "step": 1200
+ },
+ {
+ "epoch": 16.811011579637317,
+ "grad_norm": 0.24700933694839478,
+ "learning_rate": 0.0006,
+ "loss": 4.093772506713867,
+ "step": 1210
+ },
+ {
+ "epoch": 16.950841162333408,
+ "grad_norm": 0.23114480078220367,
+ "learning_rate": 0.0006,
+ "loss": 4.099676132202148,
+ "step": 1220
+ },
+ {
+ "epoch": 17.0,
+ "eval_loss": 3.9669485092163086,
+ "eval_runtime": 24.6646,
+ "eval_samples_per_second": 98.968,
+ "eval_steps_per_second": 12.406,
+ "step": 1224
+ },
+ {
+ "epoch": 17.083897749617652,
+ "grad_norm": 0.23501476645469666,
+ "learning_rate": 0.0006,
+ "loss": 4.071145248413086,
+ "step": 1230
+ },
+ {
+ "epoch": 17.223727332313743,
+ "grad_norm": 0.327950119972229,
+ "learning_rate": 0.0006,
+ "loss": 4.041724395751953,
+ "step": 1240
+ },
+ {
+ "epoch": 17.36355691500983,
+ "grad_norm": 0.2777678966522217,
+ "learning_rate": 0.0006,
+ "loss": 4.0449260711669925,
+ "step": 1250
+ },
+ {
+ "epoch": 17.50338649770592,
+ "grad_norm": 0.25167393684387207,
+ "learning_rate": 0.0006,
+ "loss": 4.058339691162109,
+ "step": 1260
+ },
+ {
+ "epoch": 17.64321608040201,
+ "grad_norm": 0.3230398893356323,
+ "learning_rate": 0.0006,
+ "loss": 4.050300598144531,
+ "step": 1270
+ },
+ {
+ "epoch": 17.7830456630981,
+ "grad_norm": 0.27687814831733704,
+ "learning_rate": 0.0006,
+ "loss": 4.071380615234375,
+ "step": 1280
+ },
+ {
+ "epoch": 17.922875245794188,
+ "grad_norm": 0.23182205855846405,
+ "learning_rate": 0.0006,
+ "loss": 4.043432235717773,
+ "step": 1290
+ },
+ {
+ "epoch": 18.0,
+ "eval_loss": 3.9458584785461426,
+ "eval_runtime": 24.662,
+ "eval_samples_per_second": 98.978,
+ "eval_steps_per_second": 12.408,
+ "step": 1296
+ },
+ {
+ "epoch": 18.055931833078436,
+ "grad_norm": 0.285981684923172,
+ "learning_rate": 0.0006,
+ "loss": 4.0329429626464846,
+ "step": 1300
+ },
+ {
+ "epoch": 18.195761415774523,
+ "grad_norm": 0.35469233989715576,
+ "learning_rate": 0.0006,
+ "loss": 3.985848617553711,
+ "step": 1310
+ },
+ {
+ "epoch": 18.335590998470614,
+ "grad_norm": 0.24750180542469025,
+ "learning_rate": 0.0006,
+ "loss": 4.006353759765625,
+ "step": 1320
+ },
+ {
+ "epoch": 18.475420581166702,
+ "grad_norm": 0.2939675450325012,
+ "learning_rate": 0.0006,
+ "loss": 4.012543106079102,
+ "step": 1330
+ },
+ {
+ "epoch": 18.615250163862793,
+ "grad_norm": 0.22177353501319885,
+ "learning_rate": 0.0006,
+ "loss": 4.013228988647461,
+ "step": 1340
+ },
+ {
+ "epoch": 18.75507974655888,
+ "grad_norm": 0.2331022024154663,
+ "learning_rate": 0.0006,
+ "loss": 4.015605545043945,
+ "step": 1350
+ },
+ {
+ "epoch": 18.894909329254972,
+ "grad_norm": 0.23938514292240143,
+ "learning_rate": 0.0006,
+ "loss": 4.01312370300293,
+ "step": 1360
+ },
+ {
+ "epoch": 19.0,
+ "eval_loss": 3.9237921237945557,
+ "eval_runtime": 24.6633,
+ "eval_samples_per_second": 98.973,
+ "eval_steps_per_second": 12.407,
+ "step": 1368
+ },
+ {
+ "epoch": 19.027965916539216,
+ "grad_norm": 0.2715682089328766,
+ "learning_rate": 0.0006,
+ "loss": 4.0193336486816404,
+ "step": 1370
+ },
+ {
+ "epoch": 19.167795499235307,
+ "grad_norm": 0.27566561102867126,
+ "learning_rate": 0.0006,
+ "loss": 3.9624229431152345,
+ "step": 1380
+ },
+ {
+ "epoch": 19.307625081931395,
+ "grad_norm": 0.3039558231830597,
+ "learning_rate": 0.0006,
+ "loss": 3.9599342346191406,
+ "step": 1390
+ },
+ {
+ "epoch": 19.447454664627486,
+ "grad_norm": 0.31524384021759033,
+ "learning_rate": 0.0006,
+ "loss": 3.979692077636719,
+ "step": 1400
+ },
+ {
+ "epoch": 19.587284247323574,
+ "grad_norm": 0.2438536137342453,
+ "learning_rate": 0.0006,
+ "loss": 3.9778110504150392,
+ "step": 1410
+ },
+ {
+ "epoch": 19.727113830019665,
+ "grad_norm": 0.2365117073059082,
+ "learning_rate": 0.0006,
+ "loss": 3.976381301879883,
+ "step": 1420
+ },
+ {
+ "epoch": 19.866943412715752,
+ "grad_norm": 0.2351199835538864,
+ "learning_rate": 0.0006,
+ "loss": 3.9808120727539062,
+ "step": 1430
+ },
+ {
+ "epoch": 20.0,
+ "grad_norm": 0.26068729162216187,
+ "learning_rate": 0.0006,
+ "loss": 3.9801464080810547,
+ "step": 1440
+ },
+ {
+ "epoch": 20.0,
+ "eval_loss": 3.9103195667266846,
+ "eval_runtime": 24.6684,
+ "eval_samples_per_second": 98.953,
+ "eval_steps_per_second": 12.405,
+ "step": 1440
+ },
+ {
+ "epoch": 20.139829582696088,
+ "grad_norm": 0.328689843416214,
+ "learning_rate": 0.0006,
+ "loss": 3.9272014617919924,
+ "step": 1450
+ },
+ {
+ "epoch": 20.27965916539218,
+ "grad_norm": 0.3154630661010742,
+ "learning_rate": 0.0006,
+ "loss": 3.938679504394531,
+ "step": 1460
+ },
+ {
+ "epoch": 20.419488748088266,
+ "grad_norm": 0.2584904432296753,
+ "learning_rate": 0.0006,
+ "loss": 3.935191345214844,
+ "step": 1470
+ },
+ {
+ "epoch": 20.559318330784357,
+ "grad_norm": 0.24333415925502777,
+ "learning_rate": 0.0006,
+ "loss": 3.930644989013672,
+ "step": 1480
+ },
+ {
+ "epoch": 20.699147913480445,
+ "grad_norm": 0.23811334371566772,
+ "learning_rate": 0.0006,
+ "loss": 3.9496768951416015,
+ "step": 1490
+ },
+ {
+ "epoch": 20.838977496176536,
+ "grad_norm": 0.2687422037124634,
+ "learning_rate": 0.0006,
+ "loss": 3.9468662261962892,
+ "step": 1500
+ },
+ {
+ "epoch": 20.978807078872624,
+ "grad_norm": 0.2426464706659317,
+ "learning_rate": 0.0006,
+ "loss": 3.960001754760742,
+ "step": 1510
+ },
+ {
+ "epoch": 21.0,
+ "eval_loss": 3.8974549770355225,
+ "eval_runtime": 24.6796,
+ "eval_samples_per_second": 98.908,
+ "eval_steps_per_second": 12.399,
+ "step": 1512
+ },
+ {
+ "epoch": 21.11186366615687,
+ "grad_norm": 0.27562302350997925,
+ "learning_rate": 0.0006,
+ "loss": 3.8891429901123047,
+ "step": 1520
+ },
+ {
+ "epoch": 21.25169324885296,
+ "grad_norm": 0.2836039960384369,
+ "learning_rate": 0.0006,
+ "loss": 3.8858299255371094,
+ "step": 1530
+ },
+ {
+ "epoch": 21.39152283154905,
+ "grad_norm": 0.2728021442890167,
+ "learning_rate": 0.0006,
+ "loss": 3.8873390197753905,
+ "step": 1540
+ },
+ {
+ "epoch": 21.531352414245138,
+ "grad_norm": 0.27416184544563293,
+ "learning_rate": 0.0006,
+ "loss": 3.9172950744628907,
+ "step": 1550
+ },
+ {
+ "epoch": 21.67118199694123,
+ "grad_norm": 0.27684128284454346,
+ "learning_rate": 0.0006,
+ "loss": 3.9083099365234375,
+ "step": 1560
+ },
+ {
+ "epoch": 21.811011579637317,
+ "grad_norm": 0.227409228682518,
+ "learning_rate": 0.0006,
+ "loss": 3.924216461181641,
+ "step": 1570
+ },
+ {
+ "epoch": 21.950841162333408,
+ "grad_norm": 0.2332412451505661,
+ "learning_rate": 0.0006,
+ "loss": 3.9231121063232424,
+ "step": 1580
+ },
+ {
+ "epoch": 22.0,
+ "eval_loss": 3.888030529022217,
+ "eval_runtime": 24.6551,
+ "eval_samples_per_second": 99.006,
+ "eval_steps_per_second": 12.411,
+ "step": 1584
+ },
+ {
+ "epoch": 22.083897749617652,
+ "grad_norm": 0.3228440284729004,
+ "learning_rate": 0.0006,
+ "loss": 3.8889293670654297,
+ "step": 1590
+ },
+ {
+ "epoch": 22.223727332313743,
+ "grad_norm": 0.24223943054676056,
+ "learning_rate": 0.0006,
+ "loss": 3.8577880859375,
+ "step": 1600
+ },
+ {
+ "epoch": 22.36355691500983,
+ "grad_norm": 0.32957133650779724,
+ "learning_rate": 0.0006,
+ "loss": 3.8739051818847656,
+ "step": 1610
+ },
+ {
+ "epoch": 22.50338649770592,
+ "grad_norm": 0.2937435209751129,
+ "learning_rate": 0.0006,
+ "loss": 3.8871036529541017,
+ "step": 1620
+ },
+ {
+ "epoch": 22.64321608040201,
+ "grad_norm": 0.2406030148267746,
+ "learning_rate": 0.0006,
+ "loss": 3.878630828857422,
+ "step": 1630
+ },
+ {
+ "epoch": 22.7830456630981,
+ "grad_norm": 0.23464885354042053,
+ "learning_rate": 0.0006,
+ "loss": 3.8893058776855467,
+ "step": 1640
+ },
+ {
+ "epoch": 22.922875245794188,
+ "grad_norm": 0.24810010194778442,
+ "learning_rate": 0.0006,
+ "loss": 3.8855140686035154,
+ "step": 1650
+ },
+ {
+ "epoch": 23.0,
+ "eval_loss": 3.8753597736358643,
+ "eval_runtime": 24.6728,
+ "eval_samples_per_second": 98.935,
+ "eval_steps_per_second": 12.402,
+ "step": 1656
+ },
+ {
+ "epoch": 23.055931833078436,
+ "grad_norm": 0.3364110589027405,
+ "learning_rate": 0.0006,
+ "loss": 3.8633838653564454,
+ "step": 1660
+ },
+ {
+ "epoch": 23.195761415774523,
+ "grad_norm": 0.281914621591568,
+ "learning_rate": 0.0006,
+ "loss": 3.8339195251464844,
+ "step": 1670
+ },
+ {
+ "epoch": 23.335590998470614,
+ "grad_norm": 0.26887112855911255,
+ "learning_rate": 0.0006,
+ "loss": 3.8345130920410155,
+ "step": 1680
+ },
+ {
+ "epoch": 23.475420581166702,
+ "grad_norm": 0.2571367621421814,
+ "learning_rate": 0.0006,
+ "loss": 3.841265869140625,
+ "step": 1690
+ },
+ {
+ "epoch": 23.615250163862793,
+ "grad_norm": 0.27085667848587036,
+ "learning_rate": 0.0006,
+ "loss": 3.8569644927978515,
+ "step": 1700
+ },
+ {
+ "epoch": 23.75507974655888,
+ "grad_norm": 0.24878937005996704,
+ "learning_rate": 0.0006,
+ "loss": 3.8654529571533205,
+ "step": 1710
+ },
+ {
+ "epoch": 23.894909329254972,
+ "grad_norm": 0.2355813831090927,
+ "learning_rate": 0.0006,
+ "loss": 3.8678340911865234,
+ "step": 1720
+ },
+ {
+ "epoch": 24.0,
+ "eval_loss": 3.8702948093414307,
+ "eval_runtime": 24.6718,
+ "eval_samples_per_second": 98.939,
+ "eval_steps_per_second": 12.403,
+ "step": 1728
+ },
+ {
+ "epoch": 24.027965916539216,
+ "grad_norm": 0.2823176085948944,
+ "learning_rate": 0.0006,
+ "loss": 3.8550838470458983,
+ "step": 1730
+ },
+ {
+ "epoch": 24.167795499235307,
+ "grad_norm": 0.3232603371143341,
+ "learning_rate": 0.0006,
+ "loss": 3.7976524353027346,
+ "step": 1740
+ },
+ {
+ "epoch": 24.307625081931395,
+ "grad_norm": 0.26159271597862244,
+ "learning_rate": 0.0006,
+ "loss": 3.8153995513916015,
+ "step": 1750
+ },
+ {
+ "epoch": 24.447454664627486,
+ "grad_norm": 0.29702314734458923,
+ "learning_rate": 0.0006,
+ "loss": 3.826192855834961,
+ "step": 1760
+ },
+ {
+ "epoch": 24.587284247323574,
+ "grad_norm": 0.24167074263095856,
+ "learning_rate": 0.0006,
+ "loss": 3.8285758972167967,
+ "step": 1770
+ },
+ {
+ "epoch": 24.727113830019665,
+ "grad_norm": 0.2669106125831604,
+ "learning_rate": 0.0006,
+ "loss": 3.8389522552490236,
+ "step": 1780
+ },
+ {
+ "epoch": 24.866943412715752,
+ "grad_norm": 0.21751989424228668,
+ "learning_rate": 0.0006,
+ "loss": 3.832489776611328,
+ "step": 1790
+ },
+ {
+ "epoch": 25.0,
+ "grad_norm": 0.27419140934944153,
+ "learning_rate": 0.0006,
+ "loss": 3.8374176025390625,
+ "step": 1800
+ },
+ {
+ "epoch": 25.0,
+ "eval_loss": 3.8606648445129395,
+ "eval_runtime": 24.6646,
+ "eval_samples_per_second": 98.968,
+ "eval_steps_per_second": 12.406,
+ "step": 1800
+ },
+ {
+ "epoch": 25.139829582696088,
+ "grad_norm": 0.3125711977481842,
+ "learning_rate": 0.0006,
+ "loss": 3.7561229705810546,
+ "step": 1810
+ },
+ {
+ "epoch": 25.27965916539218,
+ "grad_norm": 0.3050541579723358,
+ "learning_rate": 0.0006,
+ "loss": 3.7991729736328126,
+ "step": 1820
+ },
+ {
+ "epoch": 25.419488748088266,
+ "grad_norm": 0.2465241700410843,
+ "learning_rate": 0.0006,
+ "loss": 3.800311279296875,
+ "step": 1830
+ },
+ {
+ "epoch": 25.559318330784357,
+ "grad_norm": 0.22653311491012573,
+ "learning_rate": 0.0006,
+ "loss": 3.790579986572266,
+ "step": 1840
+ },
+ {
+ "epoch": 25.699147913480445,
+ "grad_norm": 0.24378487467765808,
+ "learning_rate": 0.0006,
+ "loss": 3.808407211303711,
+ "step": 1850
+ },
+ {
+ "epoch": 25.838977496176536,
+ "grad_norm": 0.2568083703517914,
+ "learning_rate": 0.0006,
+ "loss": 3.8271217346191406,
+ "step": 1860
+ },
+ {
+ "epoch": 25.978807078872624,
+ "grad_norm": 0.2565271854400635,
+ "learning_rate": 0.0006,
+ "loss": 3.821619415283203,
+ "step": 1870
+ },
+ {
+ "epoch": 26.0,
+ "eval_loss": 3.8615264892578125,
+ "eval_runtime": 24.6731,
+ "eval_samples_per_second": 98.934,
+ "eval_steps_per_second": 12.402,
+ "step": 1872
+ },
+ {
+ "epoch": 26.11186366615687,
+ "grad_norm": 0.2573966085910797,
+ "learning_rate": 0.0006,
+ "loss": 3.7596084594726564,
+ "step": 1880
+ },
+ {
+ "epoch": 26.25169324885296,
+ "grad_norm": 0.3139076232910156,
+ "learning_rate": 0.0006,
+ "loss": 3.744648742675781,
+ "step": 1890
+ },
+ {
+ "epoch": 26.39152283154905,
+ "grad_norm": 0.26629361510276794,
+ "learning_rate": 0.0006,
+ "loss": 3.765020751953125,
+ "step": 1900
+ },
+ {
+ "epoch": 26.531352414245138,
+ "grad_norm": 0.2374720573425293,
+ "learning_rate": 0.0006,
+ "loss": 3.770278549194336,
+ "step": 1910
+ },
+ {
+ "epoch": 26.67118199694123,
+ "grad_norm": 0.23601582646369934,
+ "learning_rate": 0.0006,
+ "loss": 3.789570999145508,
+ "step": 1920
+ },
+ {
+ "epoch": 26.811011579637317,
+ "grad_norm": 0.2579778730869293,
+ "learning_rate": 0.0006,
+ "loss": 3.7972648620605467,
+ "step": 1930
+ },
+ {
+ "epoch": 26.950841162333408,
+ "grad_norm": 0.2833050787448883,
+ "learning_rate": 0.0006,
+ "loss": 3.808428955078125,
+ "step": 1940
+ },
+ {
+ "epoch": 27.0,
+ "eval_loss": 3.8560781478881836,
+ "eval_runtime": 24.6609,
+ "eval_samples_per_second": 98.982,
+ "eval_steps_per_second": 12.408,
+ "step": 1944
+ },
+ {
+ "epoch": 27.083897749617652,
+ "grad_norm": 0.2856670022010803,
+ "learning_rate": 0.0006,
+ "loss": 3.764959716796875,
+ "step": 1950
+ },
+ {
+ "epoch": 27.223727332313743,
+ "grad_norm": 0.313749223947525,
+ "learning_rate": 0.0006,
+ "loss": 3.733307647705078,
+ "step": 1960
+ },
+ {
+ "epoch": 27.36355691500983,
+ "grad_norm": 0.2764676511287689,
+ "learning_rate": 0.0006,
+ "loss": 3.741012954711914,
+ "step": 1970
+ },
+ {
+ "epoch": 27.50338649770592,
+ "grad_norm": 0.2547197639942169,
+ "learning_rate": 0.0006,
+ "loss": 3.75512580871582,
+ "step": 1980
+ },
+ {
+ "epoch": 27.64321608040201,
+ "grad_norm": 0.2636491060256958,
+ "learning_rate": 0.0006,
+ "loss": 3.7716468811035155,
+ "step": 1990
+ },
+ {
+ "epoch": 27.7830456630981,
+ "grad_norm": 0.22693867981433868,
+ "learning_rate": 0.0006,
+ "loss": 3.7627937316894533,
+ "step": 2000
+ },
+ {
+ "epoch": 27.922875245794188,
+ "grad_norm": 0.23370100557804108,
+ "learning_rate": 0.0006,
+ "loss": 3.775929260253906,
+ "step": 2010
+ },
+ {
+ "epoch": 28.0,
+ "eval_loss": 3.854705333709717,
+ "eval_runtime": 24.6764,
+ "eval_samples_per_second": 98.92,
+ "eval_steps_per_second": 12.401,
+ "step": 2016
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 7200,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 100,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 8.588818398283039e+17,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/seeds/t1a_S2_C1_rand15_s43/train.log b/runs/seeds/t1a_S2_C1_rand15_s43/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..ab4dcc1add88e1c24e42dda37482714863e9c789
--- /dev/null
+++ b/runs/seeds/t1a_S2_C1_rand15_s43/train.log
@@ -0,0 +1,8552 @@
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 100
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 8
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 64
+[transformers] Total optimization steps = 7,200
+[transformers] Number of trainable parameters = 145,833,984
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: setting up run 3q7qxc0i
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260916_214505-3q7qxc0i
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t1a_S2_C1_rand15_s43
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/3q7qxc0i
+
0%| | 0/7200 [00:00system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/seeds/t1a_S2_C1_rand15_s44/checkpoint-2088/config.json b/runs/seeds/t1a_S2_C1_rand15_s44/checkpoint-2088/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8ca584a0751672f280da7c54ff3916cdf24d6f4
--- /dev/null
+++ b/runs/seeds/t1a_S2_C1_rand15_s44/checkpoint-2088/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 512,
+ "initializer_range": 0.02,
+ "intermediate_size": 1536,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 4,
+ "num_hidden_layers": 20,
+ "num_key_value_heads": 4,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151671
+}
diff --git a/runs/seeds/t1a_S2_C1_rand15_s44/checkpoint-2088/generation_config.json b/runs/seeds/t1a_S2_C1_rand15_s44/checkpoint-2088/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/seeds/t1a_S2_C1_rand15_s44/checkpoint-2088/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/seeds/t1a_S2_C1_rand15_s44/checkpoint-2088/tokenizer_config.json b/runs/seeds/t1a_S2_C1_rand15_s44/checkpoint-2088/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..839a3743177a3ed77e1ffbae67520e8500ca0532
--- /dev/null
+++ b/runs/seeds/t1a_S2_C1_rand15_s44/checkpoint-2088/tokenizer_config.json
@@ -0,0 +1,19 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/seeds/t1a_S2_C1_rand15_s44/checkpoint-2088/trainer_state.json b/runs/seeds/t1a_S2_C1_rand15_s44/checkpoint-2088/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..760ffbf67e782c73633fdf39d5bfbc4eb879df1f
--- /dev/null
+++ b/runs/seeds/t1a_S2_C1_rand15_s44/checkpoint-2088/trainer_state.json
@@ -0,0 +1,1729 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 29.0,
+ "eval_steps": 72,
+ "global_step": 2088,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013982958269608914,
+ "grad_norm": 1.700635552406311,
+ "learning_rate": 0.0,
+ "loss": 12.026216506958008,
+ "step": 1
+ },
+ {
+ "epoch": 0.13982958269608914,
+ "grad_norm": 1.7452380657196045,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.738906860351562,
+ "step": 10
+ },
+ {
+ "epoch": 0.2796591653921783,
+ "grad_norm": 1.6758230924606323,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 11.035172271728516,
+ "step": 20
+ },
+ {
+ "epoch": 0.4194887480882674,
+ "grad_norm": 1.62517249584198,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 10.288231658935548,
+ "step": 30
+ },
+ {
+ "epoch": 0.5593183307843566,
+ "grad_norm": 1.4502497911453247,
+ "learning_rate": 0.000234,
+ "loss": 9.258869934082032,
+ "step": 40
+ },
+ {
+ "epoch": 0.6991479134804457,
+ "grad_norm": 0.8760762214660645,
+ "learning_rate": 0.000294,
+ "loss": 8.262522888183593,
+ "step": 50
+ },
+ {
+ "epoch": 0.8389774961765348,
+ "grad_norm": 0.23997892439365387,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 7.690213012695312,
+ "step": 60
+ },
+ {
+ "epoch": 0.978807078872624,
+ "grad_norm": 0.34019219875335693,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 7.56441650390625,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.4646406173706055,
+ "eval_runtime": 48.3032,
+ "eval_samples_per_second": 50.535,
+ "eval_steps_per_second": 6.335,
+ "step": 72
+ },
+ {
+ "epoch": 1.1118636661568713,
+ "grad_norm": 1.241033673286438,
+ "learning_rate": 0.000474,
+ "loss": 7.427163696289062,
+ "step": 80
+ },
+ {
+ "epoch": 1.2516932488529604,
+ "grad_norm": 0.3806003928184509,
+ "learning_rate": 0.000534,
+ "loss": 7.2731483459472654,
+ "step": 90
+ },
+ {
+ "epoch": 1.3915228315490495,
+ "grad_norm": 0.3991130292415619,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 7.142841339111328,
+ "step": 100
+ },
+ {
+ "epoch": 1.5313524142451387,
+ "grad_norm": 0.4181273579597473,
+ "learning_rate": 0.0006,
+ "loss": 7.027252197265625,
+ "step": 110
+ },
+ {
+ "epoch": 1.671181996941228,
+ "grad_norm": 0.4648735821247101,
+ "learning_rate": 0.0006,
+ "loss": 6.902782440185547,
+ "step": 120
+ },
+ {
+ "epoch": 1.811011579637317,
+ "grad_norm": 0.3165874481201172,
+ "learning_rate": 0.0006,
+ "loss": 6.794158172607422,
+ "step": 130
+ },
+ {
+ "epoch": 1.9508411623334063,
+ "grad_norm": 0.9041488170623779,
+ "learning_rate": 0.0006,
+ "loss": 6.721250915527344,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 6.778120040893555,
+ "eval_runtime": 24.6913,
+ "eval_samples_per_second": 98.861,
+ "eval_steps_per_second": 12.393,
+ "step": 144
+ },
+ {
+ "epoch": 2.0838977496176536,
+ "grad_norm": 0.5713947415351868,
+ "learning_rate": 0.0006,
+ "loss": 6.634746551513672,
+ "step": 150
+ },
+ {
+ "epoch": 2.2237273323137425,
+ "grad_norm": 1.454228162765503,
+ "learning_rate": 0.0006,
+ "loss": 6.556246185302735,
+ "step": 160
+ },
+ {
+ "epoch": 2.363556915009832,
+ "grad_norm": 0.47748813033103943,
+ "learning_rate": 0.0006,
+ "loss": 6.490283203125,
+ "step": 170
+ },
+ {
+ "epoch": 2.503386497705921,
+ "grad_norm": 0.5950282216072083,
+ "learning_rate": 0.0006,
+ "loss": 6.454790496826172,
+ "step": 180
+ },
+ {
+ "epoch": 2.64321608040201,
+ "grad_norm": 0.3154385983943939,
+ "learning_rate": 0.0006,
+ "loss": 6.375165939331055,
+ "step": 190
+ },
+ {
+ "epoch": 2.783045663098099,
+ "grad_norm": 0.28560134768486023,
+ "learning_rate": 0.0006,
+ "loss": 6.336666870117187,
+ "step": 200
+ },
+ {
+ "epoch": 2.9228752457941884,
+ "grad_norm": 0.4426203668117523,
+ "learning_rate": 0.0006,
+ "loss": 6.281334304809571,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 6.3986029624938965,
+ "eval_runtime": 24.6852,
+ "eval_samples_per_second": 98.885,
+ "eval_steps_per_second": 12.396,
+ "step": 216
+ },
+ {
+ "epoch": 3.0559318330784357,
+ "grad_norm": 0.7598302364349365,
+ "learning_rate": 0.0006,
+ "loss": 6.233018493652343,
+ "step": 220
+ },
+ {
+ "epoch": 3.1957614157745247,
+ "grad_norm": 0.6198500990867615,
+ "learning_rate": 0.0006,
+ "loss": 6.202322769165039,
+ "step": 230
+ },
+ {
+ "epoch": 3.335590998470614,
+ "grad_norm": 0.33425042033195496,
+ "learning_rate": 0.0006,
+ "loss": 6.143678283691406,
+ "step": 240
+ },
+ {
+ "epoch": 3.475420581166703,
+ "grad_norm": 0.6165902018547058,
+ "learning_rate": 0.0006,
+ "loss": 6.084502792358398,
+ "step": 250
+ },
+ {
+ "epoch": 3.6152501638627923,
+ "grad_norm": 0.6409001350402832,
+ "learning_rate": 0.0006,
+ "loss": 6.035546112060547,
+ "step": 260
+ },
+ {
+ "epoch": 3.755079746558881,
+ "grad_norm": 0.9433044791221619,
+ "learning_rate": 0.0006,
+ "loss": 6.008433532714844,
+ "step": 270
+ },
+ {
+ "epoch": 3.8949093292549706,
+ "grad_norm": 0.3646930754184723,
+ "learning_rate": 0.0006,
+ "loss": 5.979931259155274,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 6.057840347290039,
+ "eval_runtime": 24.6584,
+ "eval_samples_per_second": 98.992,
+ "eval_steps_per_second": 12.41,
+ "step": 288
+ },
+ {
+ "epoch": 4.027965916539218,
+ "grad_norm": 0.5168285965919495,
+ "learning_rate": 0.0006,
+ "loss": 5.912349700927734,
+ "step": 290
+ },
+ {
+ "epoch": 4.167795499235307,
+ "grad_norm": 0.4544762372970581,
+ "learning_rate": 0.0006,
+ "loss": 5.8961952209472654,
+ "step": 300
+ },
+ {
+ "epoch": 4.307625081931396,
+ "grad_norm": 0.5194211006164551,
+ "learning_rate": 0.0006,
+ "loss": 5.8183330535888675,
+ "step": 310
+ },
+ {
+ "epoch": 4.447454664627485,
+ "grad_norm": 0.7712510824203491,
+ "learning_rate": 0.0006,
+ "loss": 5.775445938110352,
+ "step": 320
+ },
+ {
+ "epoch": 4.587284247323574,
+ "grad_norm": 0.5032666921615601,
+ "learning_rate": 0.0006,
+ "loss": 5.743329238891602,
+ "step": 330
+ },
+ {
+ "epoch": 4.727113830019664,
+ "grad_norm": 0.6890515089035034,
+ "learning_rate": 0.0006,
+ "loss": 5.728188705444336,
+ "step": 340
+ },
+ {
+ "epoch": 4.866943412715752,
+ "grad_norm": 0.33983203768730164,
+ "learning_rate": 0.0006,
+ "loss": 5.688345336914063,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.5651666522026062,
+ "learning_rate": 0.0006,
+ "loss": 5.626438903808594,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 5.64415979385376,
+ "eval_runtime": 24.663,
+ "eval_samples_per_second": 98.974,
+ "eval_steps_per_second": 12.407,
+ "step": 360
+ },
+ {
+ "epoch": 5.139829582696089,
+ "grad_norm": 0.7637493014335632,
+ "learning_rate": 0.0006,
+ "loss": 5.60391845703125,
+ "step": 370
+ },
+ {
+ "epoch": 5.279659165392179,
+ "grad_norm": 0.5426981449127197,
+ "learning_rate": 0.0006,
+ "loss": 5.582627487182617,
+ "step": 380
+ },
+ {
+ "epoch": 5.419488748088267,
+ "grad_norm": 0.3129769265651703,
+ "learning_rate": 0.0006,
+ "loss": 5.5291492462158205,
+ "step": 390
+ },
+ {
+ "epoch": 5.5593183307843566,
+ "grad_norm": 0.458799809217453,
+ "learning_rate": 0.0006,
+ "loss": 5.4846244812011715,
+ "step": 400
+ },
+ {
+ "epoch": 5.699147913480446,
+ "grad_norm": 0.4885479211807251,
+ "learning_rate": 0.0006,
+ "loss": 5.459403991699219,
+ "step": 410
+ },
+ {
+ "epoch": 5.838977496176534,
+ "grad_norm": 0.29193568229675293,
+ "learning_rate": 0.0006,
+ "loss": 5.423981094360352,
+ "step": 420
+ },
+ {
+ "epoch": 5.978807078872624,
+ "grad_norm": 0.6042985320091248,
+ "learning_rate": 0.0006,
+ "loss": 5.385090637207031,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 5.310145378112793,
+ "eval_runtime": 24.7292,
+ "eval_samples_per_second": 98.709,
+ "eval_steps_per_second": 12.374,
+ "step": 432
+ },
+ {
+ "epoch": 6.1118636661568715,
+ "grad_norm": 0.85264652967453,
+ "learning_rate": 0.0006,
+ "loss": 5.366058731079102,
+ "step": 440
+ },
+ {
+ "epoch": 6.251693248852961,
+ "grad_norm": 0.39753609895706177,
+ "learning_rate": 0.0006,
+ "loss": 5.372171020507812,
+ "step": 450
+ },
+ {
+ "epoch": 6.391522831549049,
+ "grad_norm": 0.2149779051542282,
+ "learning_rate": 0.0006,
+ "loss": 5.296001052856445,
+ "step": 460
+ },
+ {
+ "epoch": 6.531352414245139,
+ "grad_norm": 0.4946272671222687,
+ "learning_rate": 0.0006,
+ "loss": 5.262713241577148,
+ "step": 470
+ },
+ {
+ "epoch": 6.671181996941228,
+ "grad_norm": 0.3141118288040161,
+ "learning_rate": 0.0006,
+ "loss": 5.2262218475341795,
+ "step": 480
+ },
+ {
+ "epoch": 6.811011579637317,
+ "grad_norm": 0.6372417211532593,
+ "learning_rate": 0.0006,
+ "loss": 5.205792236328125,
+ "step": 490
+ },
+ {
+ "epoch": 6.950841162333406,
+ "grad_norm": 0.4364970326423645,
+ "learning_rate": 0.0006,
+ "loss": 5.210205459594727,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 5.0295491218566895,
+ "eval_runtime": 24.7123,
+ "eval_samples_per_second": 98.777,
+ "eval_steps_per_second": 12.382,
+ "step": 504
+ },
+ {
+ "epoch": 7.083897749617654,
+ "grad_norm": 0.45730671286582947,
+ "learning_rate": 0.0006,
+ "loss": 5.137186431884766,
+ "step": 510
+ },
+ {
+ "epoch": 7.223727332313743,
+ "grad_norm": 0.3208220303058624,
+ "learning_rate": 0.0006,
+ "loss": 5.086276245117188,
+ "step": 520
+ },
+ {
+ "epoch": 7.3635569150098314,
+ "grad_norm": 0.4488894045352936,
+ "learning_rate": 0.0006,
+ "loss": 5.066110992431641,
+ "step": 530
+ },
+ {
+ "epoch": 7.503386497705921,
+ "grad_norm": 0.4613317847251892,
+ "learning_rate": 0.0006,
+ "loss": 5.040608978271484,
+ "step": 540
+ },
+ {
+ "epoch": 7.64321608040201,
+ "grad_norm": 0.3950221836566925,
+ "learning_rate": 0.0006,
+ "loss": 5.015333938598633,
+ "step": 550
+ },
+ {
+ "epoch": 7.7830456630980995,
+ "grad_norm": 0.8586593270301819,
+ "learning_rate": 0.0006,
+ "loss": 5.012929916381836,
+ "step": 560
+ },
+ {
+ "epoch": 7.922875245794188,
+ "grad_norm": 0.5539677739143372,
+ "learning_rate": 0.0006,
+ "loss": 5.07519645690918,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.821177959442139,
+ "eval_runtime": 24.713,
+ "eval_samples_per_second": 98.774,
+ "eval_steps_per_second": 12.382,
+ "step": 576
+ },
+ {
+ "epoch": 8.055931833078436,
+ "grad_norm": 0.27647900581359863,
+ "learning_rate": 0.0006,
+ "loss": 4.999440765380859,
+ "step": 580
+ },
+ {
+ "epoch": 8.195761415774525,
+ "grad_norm": 0.19495148956775665,
+ "learning_rate": 0.0006,
+ "loss": 4.912628936767578,
+ "step": 590
+ },
+ {
+ "epoch": 8.335590998470614,
+ "grad_norm": 0.16093184053897858,
+ "learning_rate": 0.0006,
+ "loss": 4.875960159301758,
+ "step": 600
+ },
+ {
+ "epoch": 8.475420581166704,
+ "grad_norm": 0.21979071199893951,
+ "learning_rate": 0.0006,
+ "loss": 4.840864562988282,
+ "step": 610
+ },
+ {
+ "epoch": 8.615250163862791,
+ "grad_norm": 0.4405856430530548,
+ "learning_rate": 0.0006,
+ "loss": 4.839099502563476,
+ "step": 620
+ },
+ {
+ "epoch": 8.75507974655888,
+ "grad_norm": 0.318869024515152,
+ "learning_rate": 0.0006,
+ "loss": 4.834414291381836,
+ "step": 630
+ },
+ {
+ "epoch": 8.89490932925497,
+ "grad_norm": 0.33685195446014404,
+ "learning_rate": 0.0006,
+ "loss": 4.795156860351563,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.544349193572998,
+ "eval_runtime": 24.7137,
+ "eval_samples_per_second": 98.771,
+ "eval_steps_per_second": 12.382,
+ "step": 648
+ },
+ {
+ "epoch": 9.027965916539218,
+ "grad_norm": 0.2542240023612976,
+ "learning_rate": 0.0006,
+ "loss": 4.775139999389649,
+ "step": 650
+ },
+ {
+ "epoch": 9.167795499235307,
+ "grad_norm": 0.27856338024139404,
+ "learning_rate": 0.0006,
+ "loss": 4.718971633911133,
+ "step": 660
+ },
+ {
+ "epoch": 9.307625081931397,
+ "grad_norm": 0.3762546479701996,
+ "learning_rate": 0.0006,
+ "loss": 4.688461685180664,
+ "step": 670
+ },
+ {
+ "epoch": 9.447454664627486,
+ "grad_norm": 0.36661285161972046,
+ "learning_rate": 0.0006,
+ "loss": 4.690666198730469,
+ "step": 680
+ },
+ {
+ "epoch": 9.587284247323574,
+ "grad_norm": 0.22779567539691925,
+ "learning_rate": 0.0006,
+ "loss": 4.680194091796875,
+ "step": 690
+ },
+ {
+ "epoch": 9.727113830019663,
+ "grad_norm": 0.44679802656173706,
+ "learning_rate": 0.0006,
+ "loss": 4.660523223876953,
+ "step": 700
+ },
+ {
+ "epoch": 9.866943412715752,
+ "grad_norm": 0.32901275157928467,
+ "learning_rate": 0.0006,
+ "loss": 4.645096206665039,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.3506550192832947,
+ "learning_rate": 0.0006,
+ "loss": 4.643287658691406,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.40696907043457,
+ "eval_runtime": 24.6836,
+ "eval_samples_per_second": 98.892,
+ "eval_steps_per_second": 12.397,
+ "step": 720
+ },
+ {
+ "epoch": 10.13982958269609,
+ "grad_norm": 0.4840763509273529,
+ "learning_rate": 0.0006,
+ "loss": 4.591173934936523,
+ "step": 730
+ },
+ {
+ "epoch": 10.279659165392179,
+ "grad_norm": 0.5585325956344604,
+ "learning_rate": 0.0006,
+ "loss": 4.601737976074219,
+ "step": 740
+ },
+ {
+ "epoch": 10.419488748088268,
+ "grad_norm": 0.24250474572181702,
+ "learning_rate": 0.0006,
+ "loss": 4.607014846801758,
+ "step": 750
+ },
+ {
+ "epoch": 10.559318330784357,
+ "grad_norm": 0.21373368799686432,
+ "learning_rate": 0.0006,
+ "loss": 4.562186431884766,
+ "step": 760
+ },
+ {
+ "epoch": 10.699147913480445,
+ "grad_norm": 0.17331542074680328,
+ "learning_rate": 0.0006,
+ "loss": 4.549794006347656,
+ "step": 770
+ },
+ {
+ "epoch": 10.838977496176534,
+ "grad_norm": 0.2518278956413269,
+ "learning_rate": 0.0006,
+ "loss": 4.5227313995361325,
+ "step": 780
+ },
+ {
+ "epoch": 10.978807078872624,
+ "grad_norm": 0.29905983805656433,
+ "learning_rate": 0.0006,
+ "loss": 4.515983200073242,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.289587020874023,
+ "eval_runtime": 24.6826,
+ "eval_samples_per_second": 98.895,
+ "eval_steps_per_second": 12.397,
+ "step": 792
+ },
+ {
+ "epoch": 11.111863666156871,
+ "grad_norm": 0.3133828341960907,
+ "learning_rate": 0.0006,
+ "loss": 4.4603534698486325,
+ "step": 800
+ },
+ {
+ "epoch": 11.25169324885296,
+ "grad_norm": 0.32280510663986206,
+ "learning_rate": 0.0006,
+ "loss": 4.473176574707031,
+ "step": 810
+ },
+ {
+ "epoch": 11.39152283154905,
+ "grad_norm": 0.3710092604160309,
+ "learning_rate": 0.0006,
+ "loss": 4.466022491455078,
+ "step": 820
+ },
+ {
+ "epoch": 11.531352414245138,
+ "grad_norm": 0.273171991109848,
+ "learning_rate": 0.0006,
+ "loss": 4.442111968994141,
+ "step": 830
+ },
+ {
+ "epoch": 11.671181996941227,
+ "grad_norm": 0.3018277585506439,
+ "learning_rate": 0.0006,
+ "loss": 4.460958862304688,
+ "step": 840
+ },
+ {
+ "epoch": 11.811011579637317,
+ "grad_norm": 0.2122471034526825,
+ "learning_rate": 0.0006,
+ "loss": 4.428183746337891,
+ "step": 850
+ },
+ {
+ "epoch": 11.950841162333406,
+ "grad_norm": 0.21472617983818054,
+ "learning_rate": 0.0006,
+ "loss": 4.424028015136718,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.201544761657715,
+ "eval_runtime": 24.6843,
+ "eval_samples_per_second": 98.889,
+ "eval_steps_per_second": 12.397,
+ "step": 864
+ },
+ {
+ "epoch": 12.083897749617654,
+ "grad_norm": 0.281538724899292,
+ "learning_rate": 0.0006,
+ "loss": 4.379949569702148,
+ "step": 870
+ },
+ {
+ "epoch": 12.223727332313743,
+ "grad_norm": 0.3857930302619934,
+ "learning_rate": 0.0006,
+ "loss": 4.375837707519532,
+ "step": 880
+ },
+ {
+ "epoch": 12.363556915009832,
+ "grad_norm": 0.3517296612262726,
+ "learning_rate": 0.0006,
+ "loss": 4.371627426147461,
+ "step": 890
+ },
+ {
+ "epoch": 12.503386497705922,
+ "grad_norm": 0.2682530879974365,
+ "learning_rate": 0.0006,
+ "loss": 4.359623718261719,
+ "step": 900
+ },
+ {
+ "epoch": 12.64321608040201,
+ "grad_norm": 0.2992182970046997,
+ "learning_rate": 0.0006,
+ "loss": 4.36429443359375,
+ "step": 910
+ },
+ {
+ "epoch": 12.783045663098099,
+ "grad_norm": 0.2901749014854431,
+ "learning_rate": 0.0006,
+ "loss": 4.35257568359375,
+ "step": 920
+ },
+ {
+ "epoch": 12.922875245794188,
+ "grad_norm": 0.27002108097076416,
+ "learning_rate": 0.0006,
+ "loss": 4.3425548553466795,
+ "step": 930
+ },
+ {
+ "epoch": 13.0,
+ "eval_loss": 4.138019561767578,
+ "eval_runtime": 24.6791,
+ "eval_samples_per_second": 98.91,
+ "eval_steps_per_second": 12.399,
+ "step": 936
+ },
+ {
+ "epoch": 13.055931833078436,
+ "grad_norm": 0.3134957551956177,
+ "learning_rate": 0.0006,
+ "loss": 4.308504867553711,
+ "step": 940
+ },
+ {
+ "epoch": 13.195761415774525,
+ "grad_norm": 0.3547867238521576,
+ "learning_rate": 0.0006,
+ "loss": 4.296566390991211,
+ "step": 950
+ },
+ {
+ "epoch": 13.335590998470614,
+ "grad_norm": 0.4088982045650482,
+ "learning_rate": 0.0006,
+ "loss": 4.295370101928711,
+ "step": 960
+ },
+ {
+ "epoch": 13.475420581166704,
+ "grad_norm": 0.2418454885482788,
+ "learning_rate": 0.0006,
+ "loss": 4.310224151611328,
+ "step": 970
+ },
+ {
+ "epoch": 13.615250163862791,
+ "grad_norm": 0.20925147831439972,
+ "learning_rate": 0.0006,
+ "loss": 4.279408645629883,
+ "step": 980
+ },
+ {
+ "epoch": 13.75507974655888,
+ "grad_norm": 0.21894854307174683,
+ "learning_rate": 0.0006,
+ "loss": 4.291513061523437,
+ "step": 990
+ },
+ {
+ "epoch": 13.89490932925497,
+ "grad_norm": 0.23122107982635498,
+ "learning_rate": 0.0006,
+ "loss": 4.270157241821289,
+ "step": 1000
+ },
+ {
+ "epoch": 14.0,
+ "eval_loss": 4.082798480987549,
+ "eval_runtime": 24.6872,
+ "eval_samples_per_second": 98.877,
+ "eval_steps_per_second": 12.395,
+ "step": 1008
+ },
+ {
+ "epoch": 14.027965916539218,
+ "grad_norm": 0.29842662811279297,
+ "learning_rate": 0.0006,
+ "loss": 4.259118270874024,
+ "step": 1010
+ },
+ {
+ "epoch": 14.167795499235307,
+ "grad_norm": 0.33632612228393555,
+ "learning_rate": 0.0006,
+ "loss": 4.235930252075195,
+ "step": 1020
+ },
+ {
+ "epoch": 14.307625081931397,
+ "grad_norm": 0.4343245327472687,
+ "learning_rate": 0.0006,
+ "loss": 4.23004150390625,
+ "step": 1030
+ },
+ {
+ "epoch": 14.447454664627486,
+ "grad_norm": 0.25136545300483704,
+ "learning_rate": 0.0006,
+ "loss": 4.238217163085937,
+ "step": 1040
+ },
+ {
+ "epoch": 14.587284247323574,
+ "grad_norm": 0.22935324907302856,
+ "learning_rate": 0.0006,
+ "loss": 4.222027206420899,
+ "step": 1050
+ },
+ {
+ "epoch": 14.727113830019663,
+ "grad_norm": 0.24239933490753174,
+ "learning_rate": 0.0006,
+ "loss": 4.2153465270996096,
+ "step": 1060
+ },
+ {
+ "epoch": 14.866943412715752,
+ "grad_norm": 0.24448852241039276,
+ "learning_rate": 0.0006,
+ "loss": 4.2178184509277346,
+ "step": 1070
+ },
+ {
+ "epoch": 15.0,
+ "grad_norm": 0.2605650722980499,
+ "learning_rate": 0.0006,
+ "loss": 4.201531600952149,
+ "step": 1080
+ },
+ {
+ "epoch": 15.0,
+ "eval_loss": 4.039035320281982,
+ "eval_runtime": 24.6807,
+ "eval_samples_per_second": 98.903,
+ "eval_steps_per_second": 12.398,
+ "step": 1080
+ },
+ {
+ "epoch": 15.13982958269609,
+ "grad_norm": 0.2722151577472687,
+ "learning_rate": 0.0006,
+ "loss": 4.157822799682617,
+ "step": 1090
+ },
+ {
+ "epoch": 15.279659165392179,
+ "grad_norm": 0.3256915807723999,
+ "learning_rate": 0.0006,
+ "loss": 4.161100769042969,
+ "step": 1100
+ },
+ {
+ "epoch": 15.419488748088268,
+ "grad_norm": 0.32064902782440186,
+ "learning_rate": 0.0006,
+ "loss": 4.1776878356933596,
+ "step": 1110
+ },
+ {
+ "epoch": 15.559318330784357,
+ "grad_norm": 0.25596535205841064,
+ "learning_rate": 0.0006,
+ "loss": 4.173097229003906,
+ "step": 1120
+ },
+ {
+ "epoch": 15.699147913480445,
+ "grad_norm": 0.23048138618469238,
+ "learning_rate": 0.0006,
+ "loss": 4.156645202636719,
+ "step": 1130
+ },
+ {
+ "epoch": 15.838977496176534,
+ "grad_norm": 0.21978062391281128,
+ "learning_rate": 0.0006,
+ "loss": 4.156274032592774,
+ "step": 1140
+ },
+ {
+ "epoch": 15.978807078872624,
+ "grad_norm": 0.2334022969007492,
+ "learning_rate": 0.0006,
+ "loss": 4.16171875,
+ "step": 1150
+ },
+ {
+ "epoch": 16.0,
+ "eval_loss": 4.000299453735352,
+ "eval_runtime": 24.6925,
+ "eval_samples_per_second": 98.856,
+ "eval_steps_per_second": 12.392,
+ "step": 1152
+ },
+ {
+ "epoch": 16.11186366615687,
+ "grad_norm": 0.2948233187198639,
+ "learning_rate": 0.0006,
+ "loss": 4.121419143676758,
+ "step": 1160
+ },
+ {
+ "epoch": 16.25169324885296,
+ "grad_norm": 0.2913409471511841,
+ "learning_rate": 0.0006,
+ "loss": 4.103119659423828,
+ "step": 1170
+ },
+ {
+ "epoch": 16.39152283154905,
+ "grad_norm": 0.272742360830307,
+ "learning_rate": 0.0006,
+ "loss": 4.107622528076172,
+ "step": 1180
+ },
+ {
+ "epoch": 16.531352414245138,
+ "grad_norm": 0.24283553659915924,
+ "learning_rate": 0.0006,
+ "loss": 4.116996765136719,
+ "step": 1190
+ },
+ {
+ "epoch": 16.67118199694123,
+ "grad_norm": 0.22873002290725708,
+ "learning_rate": 0.0006,
+ "loss": 4.107876586914062,
+ "step": 1200
+ },
+ {
+ "epoch": 16.811011579637317,
+ "grad_norm": 0.24282026290893555,
+ "learning_rate": 0.0006,
+ "loss": 4.116302490234375,
+ "step": 1210
+ },
+ {
+ "epoch": 16.950841162333408,
+ "grad_norm": 0.2679036259651184,
+ "learning_rate": 0.0006,
+ "loss": 4.106821060180664,
+ "step": 1220
+ },
+ {
+ "epoch": 17.0,
+ "eval_loss": 3.9705379009246826,
+ "eval_runtime": 24.6852,
+ "eval_samples_per_second": 98.885,
+ "eval_steps_per_second": 12.396,
+ "step": 1224
+ },
+ {
+ "epoch": 17.083897749617652,
+ "grad_norm": 0.35190531611442566,
+ "learning_rate": 0.0006,
+ "loss": 4.0787097930908205,
+ "step": 1230
+ },
+ {
+ "epoch": 17.223727332313743,
+ "grad_norm": 0.4045930504798889,
+ "learning_rate": 0.0006,
+ "loss": 4.050899887084961,
+ "step": 1240
+ },
+ {
+ "epoch": 17.36355691500983,
+ "grad_norm": 0.2644161283969879,
+ "learning_rate": 0.0006,
+ "loss": 4.081280899047852,
+ "step": 1250
+ },
+ {
+ "epoch": 17.50338649770592,
+ "grad_norm": 0.26449888944625854,
+ "learning_rate": 0.0006,
+ "loss": 4.067046356201172,
+ "step": 1260
+ },
+ {
+ "epoch": 17.64321608040201,
+ "grad_norm": 0.2188943475484848,
+ "learning_rate": 0.0006,
+ "loss": 4.063029861450195,
+ "step": 1270
+ },
+ {
+ "epoch": 17.7830456630981,
+ "grad_norm": 0.24780043959617615,
+ "learning_rate": 0.0006,
+ "loss": 4.075260162353516,
+ "step": 1280
+ },
+ {
+ "epoch": 17.922875245794188,
+ "grad_norm": 0.24471357464790344,
+ "learning_rate": 0.0006,
+ "loss": 4.070241546630859,
+ "step": 1290
+ },
+ {
+ "epoch": 18.0,
+ "eval_loss": 3.949449300765991,
+ "eval_runtime": 24.7154,
+ "eval_samples_per_second": 98.764,
+ "eval_steps_per_second": 12.381,
+ "step": 1296
+ },
+ {
+ "epoch": 18.055931833078436,
+ "grad_norm": 0.34041520953178406,
+ "learning_rate": 0.0006,
+ "loss": 4.055683898925781,
+ "step": 1300
+ },
+ {
+ "epoch": 18.195761415774523,
+ "grad_norm": 0.30039289593696594,
+ "learning_rate": 0.0006,
+ "loss": 4.016337585449219,
+ "step": 1310
+ },
+ {
+ "epoch": 18.335590998470614,
+ "grad_norm": 0.2928770184516907,
+ "learning_rate": 0.0006,
+ "loss": 4.025400161743164,
+ "step": 1320
+ },
+ {
+ "epoch": 18.475420581166702,
+ "grad_norm": 0.24031077325344086,
+ "learning_rate": 0.0006,
+ "loss": 4.030586624145508,
+ "step": 1330
+ },
+ {
+ "epoch": 18.615250163862793,
+ "grad_norm": 0.22677262127399445,
+ "learning_rate": 0.0006,
+ "loss": 4.018151092529297,
+ "step": 1340
+ },
+ {
+ "epoch": 18.75507974655888,
+ "grad_norm": 0.24206310510635376,
+ "learning_rate": 0.0006,
+ "loss": 4.020846557617188,
+ "step": 1350
+ },
+ {
+ "epoch": 18.894909329254972,
+ "grad_norm": 0.21817366778850555,
+ "learning_rate": 0.0006,
+ "loss": 4.032101440429687,
+ "step": 1360
+ },
+ {
+ "epoch": 19.0,
+ "eval_loss": 3.9266469478607178,
+ "eval_runtime": 24.7116,
+ "eval_samples_per_second": 98.779,
+ "eval_steps_per_second": 12.383,
+ "step": 1368
+ },
+ {
+ "epoch": 19.027965916539216,
+ "grad_norm": 0.30548733472824097,
+ "learning_rate": 0.0006,
+ "loss": 4.014656448364258,
+ "step": 1370
+ },
+ {
+ "epoch": 19.167795499235307,
+ "grad_norm": 0.37717387080192566,
+ "learning_rate": 0.0006,
+ "loss": 3.9822765350341798,
+ "step": 1380
+ },
+ {
+ "epoch": 19.307625081931395,
+ "grad_norm": 0.3577038645744324,
+ "learning_rate": 0.0006,
+ "loss": 3.9947589874267577,
+ "step": 1390
+ },
+ {
+ "epoch": 19.447454664627486,
+ "grad_norm": 0.271404892206192,
+ "learning_rate": 0.0006,
+ "loss": 3.986977767944336,
+ "step": 1400
+ },
+ {
+ "epoch": 19.587284247323574,
+ "grad_norm": 0.2469920963048935,
+ "learning_rate": 0.0006,
+ "loss": 3.984069061279297,
+ "step": 1410
+ },
+ {
+ "epoch": 19.727113830019665,
+ "grad_norm": 0.2517824172973633,
+ "learning_rate": 0.0006,
+ "loss": 3.9947254180908205,
+ "step": 1420
+ },
+ {
+ "epoch": 19.866943412715752,
+ "grad_norm": 0.24619552493095398,
+ "learning_rate": 0.0006,
+ "loss": 3.993204116821289,
+ "step": 1430
+ },
+ {
+ "epoch": 20.0,
+ "grad_norm": 0.293952077627182,
+ "learning_rate": 0.0006,
+ "loss": 3.995740509033203,
+ "step": 1440
+ },
+ {
+ "epoch": 20.0,
+ "eval_loss": 3.913203477859497,
+ "eval_runtime": 24.6942,
+ "eval_samples_per_second": 98.849,
+ "eval_steps_per_second": 12.392,
+ "step": 1440
+ },
+ {
+ "epoch": 20.139829582696088,
+ "grad_norm": 0.2794000506401062,
+ "learning_rate": 0.0006,
+ "loss": 3.9209800720214845,
+ "step": 1450
+ },
+ {
+ "epoch": 20.27965916539218,
+ "grad_norm": 0.2759738564491272,
+ "learning_rate": 0.0006,
+ "loss": 3.937378692626953,
+ "step": 1460
+ },
+ {
+ "epoch": 20.419488748088266,
+ "grad_norm": 0.26212143898010254,
+ "learning_rate": 0.0006,
+ "loss": 3.942386245727539,
+ "step": 1470
+ },
+ {
+ "epoch": 20.559318330784357,
+ "grad_norm": 0.31635782122612,
+ "learning_rate": 0.0006,
+ "loss": 3.9615810394287108,
+ "step": 1480
+ },
+ {
+ "epoch": 20.699147913480445,
+ "grad_norm": 0.24542030692100525,
+ "learning_rate": 0.0006,
+ "loss": 3.9448005676269533,
+ "step": 1490
+ },
+ {
+ "epoch": 20.838977496176536,
+ "grad_norm": 0.24720963835716248,
+ "learning_rate": 0.0006,
+ "loss": 3.9657512664794923,
+ "step": 1500
+ },
+ {
+ "epoch": 20.978807078872624,
+ "grad_norm": 0.25068414211273193,
+ "learning_rate": 0.0006,
+ "loss": 3.968650054931641,
+ "step": 1510
+ },
+ {
+ "epoch": 21.0,
+ "eval_loss": 3.898155927658081,
+ "eval_runtime": 24.6889,
+ "eval_samples_per_second": 98.87,
+ "eval_steps_per_second": 12.394,
+ "step": 1512
+ },
+ {
+ "epoch": 21.11186366615687,
+ "grad_norm": 0.25450006127357483,
+ "learning_rate": 0.0006,
+ "loss": 3.9160320281982424,
+ "step": 1520
+ },
+ {
+ "epoch": 21.25169324885296,
+ "grad_norm": 0.2801636755466461,
+ "learning_rate": 0.0006,
+ "loss": 3.9062171936035157,
+ "step": 1530
+ },
+ {
+ "epoch": 21.39152283154905,
+ "grad_norm": 0.2986571192741394,
+ "learning_rate": 0.0006,
+ "loss": 3.922093963623047,
+ "step": 1540
+ },
+ {
+ "epoch": 21.531352414245138,
+ "grad_norm": 0.26316359639167786,
+ "learning_rate": 0.0006,
+ "loss": 3.932251739501953,
+ "step": 1550
+ },
+ {
+ "epoch": 21.67118199694123,
+ "grad_norm": 0.24477893114089966,
+ "learning_rate": 0.0006,
+ "loss": 3.9160587310791017,
+ "step": 1560
+ },
+ {
+ "epoch": 21.811011579637317,
+ "grad_norm": 0.26223233342170715,
+ "learning_rate": 0.0006,
+ "loss": 3.9295070648193358,
+ "step": 1570
+ },
+ {
+ "epoch": 21.950841162333408,
+ "grad_norm": 0.2287198156118393,
+ "learning_rate": 0.0006,
+ "loss": 3.928273010253906,
+ "step": 1580
+ },
+ {
+ "epoch": 22.0,
+ "eval_loss": 3.8834431171417236,
+ "eval_runtime": 24.7508,
+ "eval_samples_per_second": 98.623,
+ "eval_steps_per_second": 12.363,
+ "step": 1584
+ },
+ {
+ "epoch": 22.083897749617652,
+ "grad_norm": 0.2716076970100403,
+ "learning_rate": 0.0006,
+ "loss": 3.8900310516357424,
+ "step": 1590
+ },
+ {
+ "epoch": 22.223727332313743,
+ "grad_norm": 0.3099433183670044,
+ "learning_rate": 0.0006,
+ "loss": 3.8721824645996095,
+ "step": 1600
+ },
+ {
+ "epoch": 22.36355691500983,
+ "grad_norm": 0.2886529564857483,
+ "learning_rate": 0.0006,
+ "loss": 3.881540298461914,
+ "step": 1610
+ },
+ {
+ "epoch": 22.50338649770592,
+ "grad_norm": 0.2604324221611023,
+ "learning_rate": 0.0006,
+ "loss": 3.87463264465332,
+ "step": 1620
+ },
+ {
+ "epoch": 22.64321608040201,
+ "grad_norm": 0.23295792937278748,
+ "learning_rate": 0.0006,
+ "loss": 3.8972091674804688,
+ "step": 1630
+ },
+ {
+ "epoch": 22.7830456630981,
+ "grad_norm": 0.25366052985191345,
+ "learning_rate": 0.0006,
+ "loss": 3.903220367431641,
+ "step": 1640
+ },
+ {
+ "epoch": 22.922875245794188,
+ "grad_norm": 0.23160940408706665,
+ "learning_rate": 0.0006,
+ "loss": 3.901224899291992,
+ "step": 1650
+ },
+ {
+ "epoch": 23.0,
+ "eval_loss": 3.880795478820801,
+ "eval_runtime": 24.7176,
+ "eval_samples_per_second": 98.755,
+ "eval_steps_per_second": 12.38,
+ "step": 1656
+ },
+ {
+ "epoch": 23.055931833078436,
+ "grad_norm": 0.3570951223373413,
+ "learning_rate": 0.0006,
+ "loss": 3.878731918334961,
+ "step": 1660
+ },
+ {
+ "epoch": 23.195761415774523,
+ "grad_norm": 0.27881109714508057,
+ "learning_rate": 0.0006,
+ "loss": 3.839316177368164,
+ "step": 1670
+ },
+ {
+ "epoch": 23.335590998470614,
+ "grad_norm": 0.279045432806015,
+ "learning_rate": 0.0006,
+ "loss": 3.8574909210205077,
+ "step": 1680
+ },
+ {
+ "epoch": 23.475420581166702,
+ "grad_norm": 0.27157124876976013,
+ "learning_rate": 0.0006,
+ "loss": 3.856222915649414,
+ "step": 1690
+ },
+ {
+ "epoch": 23.615250163862793,
+ "grad_norm": 0.2575245797634125,
+ "learning_rate": 0.0006,
+ "loss": 3.861238479614258,
+ "step": 1700
+ },
+ {
+ "epoch": 23.75507974655888,
+ "grad_norm": 0.29189425706863403,
+ "learning_rate": 0.0006,
+ "loss": 3.8741355895996095,
+ "step": 1710
+ },
+ {
+ "epoch": 23.894909329254972,
+ "grad_norm": 0.2326284497976303,
+ "learning_rate": 0.0006,
+ "loss": 3.8727760314941406,
+ "step": 1720
+ },
+ {
+ "epoch": 24.0,
+ "eval_loss": 3.871624231338501,
+ "eval_runtime": 24.7049,
+ "eval_samples_per_second": 98.806,
+ "eval_steps_per_second": 12.386,
+ "step": 1728
+ },
+ {
+ "epoch": 24.027965916539216,
+ "grad_norm": 0.2619258463382721,
+ "learning_rate": 0.0006,
+ "loss": 3.850482940673828,
+ "step": 1730
+ },
+ {
+ "epoch": 24.167795499235307,
+ "grad_norm": 0.318989634513855,
+ "learning_rate": 0.0006,
+ "loss": 3.797576141357422,
+ "step": 1740
+ },
+ {
+ "epoch": 24.307625081931395,
+ "grad_norm": 0.34048542380332947,
+ "learning_rate": 0.0006,
+ "loss": 3.835769271850586,
+ "step": 1750
+ },
+ {
+ "epoch": 24.447454664627486,
+ "grad_norm": 0.27635887265205383,
+ "learning_rate": 0.0006,
+ "loss": 3.8384330749511717,
+ "step": 1760
+ },
+ {
+ "epoch": 24.587284247323574,
+ "grad_norm": 0.2443767488002777,
+ "learning_rate": 0.0006,
+ "loss": 3.8239139556884765,
+ "step": 1770
+ },
+ {
+ "epoch": 24.727113830019665,
+ "grad_norm": 0.2602272629737854,
+ "learning_rate": 0.0006,
+ "loss": 3.8507083892822265,
+ "step": 1780
+ },
+ {
+ "epoch": 24.866943412715752,
+ "grad_norm": 0.25705334544181824,
+ "learning_rate": 0.0006,
+ "loss": 3.857419967651367,
+ "step": 1790
+ },
+ {
+ "epoch": 25.0,
+ "grad_norm": 0.2956896424293518,
+ "learning_rate": 0.0006,
+ "loss": 3.859505844116211,
+ "step": 1800
+ },
+ {
+ "epoch": 25.0,
+ "eval_loss": 3.8609087467193604,
+ "eval_runtime": 24.6911,
+ "eval_samples_per_second": 98.861,
+ "eval_steps_per_second": 12.393,
+ "step": 1800
+ },
+ {
+ "epoch": 25.139829582696088,
+ "grad_norm": 0.26035305857658386,
+ "learning_rate": 0.0006,
+ "loss": 3.7773101806640623,
+ "step": 1810
+ },
+ {
+ "epoch": 25.27965916539218,
+ "grad_norm": 0.2907235026359558,
+ "learning_rate": 0.0006,
+ "loss": 3.7813282012939453,
+ "step": 1820
+ },
+ {
+ "epoch": 25.419488748088266,
+ "grad_norm": 0.258512020111084,
+ "learning_rate": 0.0006,
+ "loss": 3.7959663391113283,
+ "step": 1830
+ },
+ {
+ "epoch": 25.559318330784357,
+ "grad_norm": 0.2634963393211365,
+ "learning_rate": 0.0006,
+ "loss": 3.8082351684570312,
+ "step": 1840
+ },
+ {
+ "epoch": 25.699147913480445,
+ "grad_norm": 0.23857305943965912,
+ "learning_rate": 0.0006,
+ "loss": 3.822832489013672,
+ "step": 1850
+ },
+ {
+ "epoch": 25.838977496176536,
+ "grad_norm": 0.25320932269096375,
+ "learning_rate": 0.0006,
+ "loss": 3.839424896240234,
+ "step": 1860
+ },
+ {
+ "epoch": 25.978807078872624,
+ "grad_norm": 0.254719078540802,
+ "learning_rate": 0.0006,
+ "loss": 3.8290809631347655,
+ "step": 1870
+ },
+ {
+ "epoch": 26.0,
+ "eval_loss": 3.860543966293335,
+ "eval_runtime": 24.6969,
+ "eval_samples_per_second": 98.838,
+ "eval_steps_per_second": 12.39,
+ "step": 1872
+ },
+ {
+ "epoch": 26.11186366615687,
+ "grad_norm": 0.26978418231010437,
+ "learning_rate": 0.0006,
+ "loss": 3.783213424682617,
+ "step": 1880
+ },
+ {
+ "epoch": 26.25169324885296,
+ "grad_norm": 0.3025876581668854,
+ "learning_rate": 0.0006,
+ "loss": 3.7678558349609377,
+ "step": 1890
+ },
+ {
+ "epoch": 26.39152283154905,
+ "grad_norm": 0.28634828329086304,
+ "learning_rate": 0.0006,
+ "loss": 3.772584915161133,
+ "step": 1900
+ },
+ {
+ "epoch": 26.531352414245138,
+ "grad_norm": 0.27327099442481995,
+ "learning_rate": 0.0006,
+ "loss": 3.792584228515625,
+ "step": 1910
+ },
+ {
+ "epoch": 26.67118199694123,
+ "grad_norm": 0.25156402587890625,
+ "learning_rate": 0.0006,
+ "loss": 3.7997390747070314,
+ "step": 1920
+ },
+ {
+ "epoch": 26.811011579637317,
+ "grad_norm": 0.24488316476345062,
+ "learning_rate": 0.0006,
+ "loss": 3.796276092529297,
+ "step": 1930
+ },
+ {
+ "epoch": 26.950841162333408,
+ "grad_norm": 0.2255452275276184,
+ "learning_rate": 0.0006,
+ "loss": 3.7969295501708986,
+ "step": 1940
+ },
+ {
+ "epoch": 27.0,
+ "eval_loss": 3.856274366378784,
+ "eval_runtime": 24.7412,
+ "eval_samples_per_second": 98.661,
+ "eval_steps_per_second": 12.368,
+ "step": 1944
+ },
+ {
+ "epoch": 27.083897749617652,
+ "grad_norm": 0.32721033692359924,
+ "learning_rate": 0.0006,
+ "loss": 3.7623157501220703,
+ "step": 1950
+ },
+ {
+ "epoch": 27.223727332313743,
+ "grad_norm": 0.35942360758781433,
+ "learning_rate": 0.0006,
+ "loss": 3.747374725341797,
+ "step": 1960
+ },
+ {
+ "epoch": 27.36355691500983,
+ "grad_norm": 0.28229889273643494,
+ "learning_rate": 0.0006,
+ "loss": 3.7484817504882812,
+ "step": 1970
+ },
+ {
+ "epoch": 27.50338649770592,
+ "grad_norm": 0.24066105484962463,
+ "learning_rate": 0.0006,
+ "loss": 3.7558967590332033,
+ "step": 1980
+ },
+ {
+ "epoch": 27.64321608040201,
+ "grad_norm": 0.23449255526065826,
+ "learning_rate": 0.0006,
+ "loss": 3.7748146057128906,
+ "step": 1990
+ },
+ {
+ "epoch": 27.7830456630981,
+ "grad_norm": 0.23389872908592224,
+ "learning_rate": 0.0006,
+ "loss": 3.777179718017578,
+ "step": 2000
+ },
+ {
+ "epoch": 27.922875245794188,
+ "grad_norm": 0.2445559799671173,
+ "learning_rate": 0.0006,
+ "loss": 3.787713623046875,
+ "step": 2010
+ },
+ {
+ "epoch": 28.0,
+ "eval_loss": 3.8526878356933594,
+ "eval_runtime": 24.7011,
+ "eval_samples_per_second": 98.821,
+ "eval_steps_per_second": 12.388,
+ "step": 2016
+ },
+ {
+ "epoch": 28.055931833078436,
+ "grad_norm": 0.34017670154571533,
+ "learning_rate": 0.0006,
+ "loss": 3.775929260253906,
+ "step": 2020
+ },
+ {
+ "epoch": 28.195761415774523,
+ "grad_norm": 0.3126900792121887,
+ "learning_rate": 0.0006,
+ "loss": 3.734564208984375,
+ "step": 2030
+ },
+ {
+ "epoch": 28.335590998470614,
+ "grad_norm": 0.24868397414684296,
+ "learning_rate": 0.0006,
+ "loss": 3.7313697814941404,
+ "step": 2040
+ },
+ {
+ "epoch": 28.475420581166702,
+ "grad_norm": 0.27567771077156067,
+ "learning_rate": 0.0006,
+ "loss": 3.7346027374267576,
+ "step": 2050
+ },
+ {
+ "epoch": 28.615250163862793,
+ "grad_norm": 0.29415446519851685,
+ "learning_rate": 0.0006,
+ "loss": 3.7519935607910155,
+ "step": 2060
+ },
+ {
+ "epoch": 28.75507974655888,
+ "grad_norm": 0.2452312707901001,
+ "learning_rate": 0.0006,
+ "loss": 3.7637069702148436,
+ "step": 2070
+ },
+ {
+ "epoch": 28.894909329254972,
+ "grad_norm": 0.24146981537342072,
+ "learning_rate": 0.0006,
+ "loss": 3.7546928405761717,
+ "step": 2080
+ },
+ {
+ "epoch": 29.0,
+ "eval_loss": 3.8505256175994873,
+ "eval_runtime": 24.6969,
+ "eval_samples_per_second": 98.838,
+ "eval_steps_per_second": 12.39,
+ "step": 2088
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 7200,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 100,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 8.895561912507433e+17,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/seeds/t1a_S2_C1_rand15_s44/run.lock b/runs/seeds/t1a_S2_C1_rand15_s44/run.lock
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/runs/seeds/t1a_S2_C1_rand15_s44/train.log b/runs/seeds/t1a_S2_C1_rand15_s44/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..c5a47dc4349202bca073f0a66b62956893e4193b
--- /dev/null
+++ b/runs/seeds/t1a_S2_C1_rand15_s44/train.log
@@ -0,0 +1,8744 @@
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 100
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 8
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 64
+[transformers] Total optimization steps = 7,200
+[transformers] Number of trainable parameters = 145,833,984
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: setting up run ku7gmqg8
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260917_215328-ku7gmqg8
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t1a_S2_C1_rand15_s44
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/ku7gmqg8
+
0%| | 0/7200 [00:00system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/checkpoint-3744/config.json b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/checkpoint-3744/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3474619f10931027097cc05d837296fa4f33a6c
--- /dev/null
+++ b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/checkpoint-3744/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 512,
+ "initializer_range": 0.02,
+ "intermediate_size": 1536,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 4,
+ "num_hidden_layers": 20,
+ "num_key_value_heads": 4,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151676
+}
diff --git a/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/checkpoint-3744/generation_config.json b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/checkpoint-3744/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/checkpoint-3744/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/checkpoint-3744/tokenizer_config.json b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/checkpoint-3744/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d43c97a447f4b25a8c36be5873f7b622935fa90
--- /dev/null
+++ b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/checkpoint-3744/tokenizer_config.json
@@ -0,0 +1,24 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>",
+ "<|next_1_pred|>",
+ "<|next_2_pred|>",
+ "<|next_3_pred|>",
+ "<|next_4_pred|>",
+ "<|next_5_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/checkpoint-3744/trainer_state.json b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/checkpoint-3744/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8adf37622886bd8e93169f91963ddea9918303d6
--- /dev/null
+++ b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/checkpoint-3744/trainer_state.json
@@ -0,0 +1,3075 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 52.0,
+ "eval_steps": 72,
+ "global_step": 3744,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013982958269608914,
+ "grad_norm": 2.876962423324585,
+ "learning_rate": 0.0,
+ "loss": 12.033732414245605,
+ "step": 1
+ },
+ {
+ "epoch": 0.13982958269608914,
+ "grad_norm": 1.7589133977890015,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.674965752495659,
+ "step": 10
+ },
+ {
+ "epoch": 0.2796591653921783,
+ "grad_norm": 1.6702193021774292,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 10.973360443115235,
+ "step": 20
+ },
+ {
+ "epoch": 0.4194887480882674,
+ "grad_norm": 1.645583987236023,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 10.207978820800781,
+ "step": 30
+ },
+ {
+ "epoch": 0.5593183307843566,
+ "grad_norm": 1.4699299335479736,
+ "learning_rate": 0.000234,
+ "loss": 9.164710235595702,
+ "step": 40
+ },
+ {
+ "epoch": 0.6991479134804457,
+ "grad_norm": 0.8350820541381836,
+ "learning_rate": 0.000294,
+ "loss": 8.197768402099609,
+ "step": 50
+ },
+ {
+ "epoch": 0.8389774961765348,
+ "grad_norm": 0.22249726951122284,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 7.696961212158203,
+ "step": 60
+ },
+ {
+ "epoch": 0.978807078872624,
+ "grad_norm": 0.17536215484142303,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 7.5758544921875,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.527614116668701,
+ "eval_runtime": 43.5349,
+ "eval_samples_per_second": 56.07,
+ "eval_steps_per_second": 7.029,
+ "step": 72
+ },
+ {
+ "epoch": 1.1118636661568713,
+ "grad_norm": 0.33962082862854004,
+ "learning_rate": 0.000474,
+ "loss": 7.511866760253906,
+ "step": 80
+ },
+ {
+ "epoch": 1.2516932488529604,
+ "grad_norm": 0.36346229910850525,
+ "learning_rate": 0.000534,
+ "loss": 7.423744964599609,
+ "step": 90
+ },
+ {
+ "epoch": 1.3915228315490495,
+ "grad_norm": 0.11035402864217758,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 7.388601684570313,
+ "step": 100
+ },
+ {
+ "epoch": 1.5313524142451387,
+ "grad_norm": 0.37662366032600403,
+ "learning_rate": 0.0006,
+ "loss": 7.367597961425782,
+ "step": 110
+ },
+ {
+ "epoch": 1.671181996941228,
+ "grad_norm": 0.13388605415821075,
+ "learning_rate": 0.0006,
+ "loss": 7.331645202636719,
+ "step": 120
+ },
+ {
+ "epoch": 1.811011579637317,
+ "grad_norm": 0.480042964220047,
+ "learning_rate": 0.0006,
+ "loss": 7.299599456787109,
+ "step": 130
+ },
+ {
+ "epoch": 1.9508411623334063,
+ "grad_norm": 0.24864225089550018,
+ "learning_rate": 0.0006,
+ "loss": 7.289546203613281,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 7.116514682769775,
+ "eval_runtime": 19.868,
+ "eval_samples_per_second": 122.861,
+ "eval_steps_per_second": 15.402,
+ "step": 144
+ },
+ {
+ "epoch": 2.0838977496176536,
+ "grad_norm": 0.16448751091957092,
+ "learning_rate": 0.0006,
+ "loss": 7.2175537109375,
+ "step": 150
+ },
+ {
+ "epoch": 2.2237273323137425,
+ "grad_norm": 0.5487401485443115,
+ "learning_rate": 0.0006,
+ "loss": 7.1661827087402346,
+ "step": 160
+ },
+ {
+ "epoch": 2.363556915009832,
+ "grad_norm": 0.28657466173171997,
+ "learning_rate": 0.0006,
+ "loss": 7.138227081298828,
+ "step": 170
+ },
+ {
+ "epoch": 2.503386497705921,
+ "grad_norm": 0.5549154877662659,
+ "learning_rate": 0.0006,
+ "loss": 7.058498382568359,
+ "step": 180
+ },
+ {
+ "epoch": 2.64321608040201,
+ "grad_norm": 0.3143436014652252,
+ "learning_rate": 0.0006,
+ "loss": 6.999155426025391,
+ "step": 190
+ },
+ {
+ "epoch": 2.783045663098099,
+ "grad_norm": 0.7382907867431641,
+ "learning_rate": 0.0006,
+ "loss": 6.882961273193359,
+ "step": 200
+ },
+ {
+ "epoch": 2.9228752457941884,
+ "grad_norm": 0.3659225404262543,
+ "learning_rate": 0.0006,
+ "loss": 6.836590576171875,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 6.464047431945801,
+ "eval_runtime": 19.9003,
+ "eval_samples_per_second": 122.662,
+ "eval_steps_per_second": 15.377,
+ "step": 216
+ },
+ {
+ "epoch": 3.0559318330784357,
+ "grad_norm": 0.38216081261634827,
+ "learning_rate": 0.0006,
+ "loss": 6.730919647216797,
+ "step": 220
+ },
+ {
+ "epoch": 3.1957614157745247,
+ "grad_norm": 1.1124485731124878,
+ "learning_rate": 0.0006,
+ "loss": 6.6904449462890625,
+ "step": 230
+ },
+ {
+ "epoch": 3.335590998470614,
+ "grad_norm": 0.26441943645477295,
+ "learning_rate": 0.0006,
+ "loss": 6.640740966796875,
+ "step": 240
+ },
+ {
+ "epoch": 3.475420581166703,
+ "grad_norm": 0.40204131603240967,
+ "learning_rate": 0.0006,
+ "loss": 6.554674530029297,
+ "step": 250
+ },
+ {
+ "epoch": 3.6152501638627923,
+ "grad_norm": 0.9979112148284912,
+ "learning_rate": 0.0006,
+ "loss": 6.492906188964843,
+ "step": 260
+ },
+ {
+ "epoch": 3.755079746558881,
+ "grad_norm": 0.49941176176071167,
+ "learning_rate": 0.0006,
+ "loss": 6.450276184082031,
+ "step": 270
+ },
+ {
+ "epoch": 3.8949093292549706,
+ "grad_norm": 0.4352740943431854,
+ "learning_rate": 0.0006,
+ "loss": 6.399027252197266,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 5.985569477081299,
+ "eval_runtime": 19.868,
+ "eval_samples_per_second": 122.861,
+ "eval_steps_per_second": 15.402,
+ "step": 288
+ },
+ {
+ "epoch": 4.027965916539218,
+ "grad_norm": 0.4469015598297119,
+ "learning_rate": 0.0006,
+ "loss": 6.338511657714844,
+ "step": 290
+ },
+ {
+ "epoch": 4.167795499235307,
+ "grad_norm": 0.5202305316925049,
+ "learning_rate": 0.0006,
+ "loss": 6.325920486450196,
+ "step": 300
+ },
+ {
+ "epoch": 4.307625081931396,
+ "grad_norm": 0.22929437458515167,
+ "learning_rate": 0.0006,
+ "loss": 6.275778198242188,
+ "step": 310
+ },
+ {
+ "epoch": 4.447454664627485,
+ "grad_norm": 0.26329585909843445,
+ "learning_rate": 0.0006,
+ "loss": 6.181620025634766,
+ "step": 320
+ },
+ {
+ "epoch": 4.587284247323574,
+ "grad_norm": 0.5213761925697327,
+ "learning_rate": 0.0006,
+ "loss": 6.125854110717773,
+ "step": 330
+ },
+ {
+ "epoch": 4.727113830019664,
+ "grad_norm": 0.8569974303245544,
+ "learning_rate": 0.0006,
+ "loss": 6.107173919677734,
+ "step": 340
+ },
+ {
+ "epoch": 4.866943412715752,
+ "grad_norm": 0.3612479567527771,
+ "learning_rate": 0.0006,
+ "loss": 6.116686248779297,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.19996359944343567,
+ "learning_rate": 0.0006,
+ "loss": 6.025409698486328,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 5.643312931060791,
+ "eval_runtime": 19.8656,
+ "eval_samples_per_second": 122.875,
+ "eval_steps_per_second": 15.403,
+ "step": 360
+ },
+ {
+ "epoch": 5.139829582696089,
+ "grad_norm": 0.29818713665008545,
+ "learning_rate": 0.0006,
+ "loss": 5.992686462402344,
+ "step": 370
+ },
+ {
+ "epoch": 5.279659165392179,
+ "grad_norm": 0.5892453193664551,
+ "learning_rate": 0.0006,
+ "loss": 5.998208618164062,
+ "step": 380
+ },
+ {
+ "epoch": 5.419488748088267,
+ "grad_norm": 0.2266741544008255,
+ "learning_rate": 0.0006,
+ "loss": 5.9553264617919925,
+ "step": 390
+ },
+ {
+ "epoch": 5.5593183307843566,
+ "grad_norm": 0.2521528899669647,
+ "learning_rate": 0.0006,
+ "loss": 5.89477310180664,
+ "step": 400
+ },
+ {
+ "epoch": 5.699147913480446,
+ "grad_norm": 0.43839821219444275,
+ "learning_rate": 0.0006,
+ "loss": 5.836813354492188,
+ "step": 410
+ },
+ {
+ "epoch": 5.838977496176534,
+ "grad_norm": 0.366489976644516,
+ "learning_rate": 0.0006,
+ "loss": 5.804706954956055,
+ "step": 420
+ },
+ {
+ "epoch": 5.978807078872624,
+ "grad_norm": 0.6092522740364075,
+ "learning_rate": 0.0006,
+ "loss": 5.7997489929199215,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 5.379720211029053,
+ "eval_runtime": 19.879,
+ "eval_samples_per_second": 122.793,
+ "eval_steps_per_second": 15.393,
+ "step": 432
+ },
+ {
+ "epoch": 6.1118636661568715,
+ "grad_norm": 0.26461780071258545,
+ "learning_rate": 0.0006,
+ "loss": 5.752685928344727,
+ "step": 440
+ },
+ {
+ "epoch": 6.251693248852961,
+ "grad_norm": 0.21015925705432892,
+ "learning_rate": 0.0006,
+ "loss": 5.668216705322266,
+ "step": 450
+ },
+ {
+ "epoch": 6.391522831549049,
+ "grad_norm": 0.43590185046195984,
+ "learning_rate": 0.0006,
+ "loss": 5.672888565063476,
+ "step": 460
+ },
+ {
+ "epoch": 6.531352414245139,
+ "grad_norm": 0.3031458854675293,
+ "learning_rate": 0.0006,
+ "loss": 5.634977340698242,
+ "step": 470
+ },
+ {
+ "epoch": 6.671181996941228,
+ "grad_norm": 0.31005996465682983,
+ "learning_rate": 0.0006,
+ "loss": 5.595721054077148,
+ "step": 480
+ },
+ {
+ "epoch": 6.811011579637317,
+ "grad_norm": 0.2860366404056549,
+ "learning_rate": 0.0006,
+ "loss": 5.547025680541992,
+ "step": 490
+ },
+ {
+ "epoch": 6.950841162333406,
+ "grad_norm": 0.27822229266166687,
+ "learning_rate": 0.0006,
+ "loss": 5.5134632110595705,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 5.08353853225708,
+ "eval_runtime": 19.8727,
+ "eval_samples_per_second": 122.832,
+ "eval_steps_per_second": 15.398,
+ "step": 504
+ },
+ {
+ "epoch": 7.083897749617654,
+ "grad_norm": 0.5675644278526306,
+ "learning_rate": 0.0006,
+ "loss": 5.4566387176513675,
+ "step": 510
+ },
+ {
+ "epoch": 7.223727332313743,
+ "grad_norm": 0.37517499923706055,
+ "learning_rate": 0.0006,
+ "loss": 5.47674674987793,
+ "step": 520
+ },
+ {
+ "epoch": 7.3635569150098314,
+ "grad_norm": 0.3244355320930481,
+ "learning_rate": 0.0006,
+ "loss": 5.433233642578125,
+ "step": 530
+ },
+ {
+ "epoch": 7.503386497705921,
+ "grad_norm": 0.2898264229297638,
+ "learning_rate": 0.0006,
+ "loss": 5.383018112182617,
+ "step": 540
+ },
+ {
+ "epoch": 7.64321608040201,
+ "grad_norm": 0.3088887333869934,
+ "learning_rate": 0.0006,
+ "loss": 5.351487350463867,
+ "step": 550
+ },
+ {
+ "epoch": 7.7830456630980995,
+ "grad_norm": 0.42422589659690857,
+ "learning_rate": 0.0006,
+ "loss": 5.283416748046875,
+ "step": 560
+ },
+ {
+ "epoch": 7.922875245794188,
+ "grad_norm": 0.33900678157806396,
+ "learning_rate": 0.0006,
+ "loss": 5.29944076538086,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.821493148803711,
+ "eval_runtime": 19.8817,
+ "eval_samples_per_second": 122.776,
+ "eval_steps_per_second": 15.391,
+ "step": 576
+ },
+ {
+ "epoch": 8.055931833078436,
+ "grad_norm": 0.6213559508323669,
+ "learning_rate": 0.0006,
+ "loss": 5.259173202514648,
+ "step": 580
+ },
+ {
+ "epoch": 8.195761415774525,
+ "grad_norm": 0.23778444528579712,
+ "learning_rate": 0.0006,
+ "loss": 5.240984344482422,
+ "step": 590
+ },
+ {
+ "epoch": 8.335590998470614,
+ "grad_norm": 0.3613894283771515,
+ "learning_rate": 0.0006,
+ "loss": 5.190341186523438,
+ "step": 600
+ },
+ {
+ "epoch": 8.475420581166704,
+ "grad_norm": 0.43025949597358704,
+ "learning_rate": 0.0006,
+ "loss": 5.168280792236328,
+ "step": 610
+ },
+ {
+ "epoch": 8.615250163862791,
+ "grad_norm": 0.3465190827846527,
+ "learning_rate": 0.0006,
+ "loss": 5.144497680664062,
+ "step": 620
+ },
+ {
+ "epoch": 8.75507974655888,
+ "grad_norm": 0.33380547165870667,
+ "learning_rate": 0.0006,
+ "loss": 5.1236320495605465,
+ "step": 630
+ },
+ {
+ "epoch": 8.89490932925497,
+ "grad_norm": 0.45730915665626526,
+ "learning_rate": 0.0006,
+ "loss": 5.113668823242188,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.614823818206787,
+ "eval_runtime": 19.8697,
+ "eval_samples_per_second": 122.851,
+ "eval_steps_per_second": 15.4,
+ "step": 648
+ },
+ {
+ "epoch": 9.027965916539218,
+ "grad_norm": 0.28316938877105713,
+ "learning_rate": 0.0006,
+ "loss": 5.052216339111328,
+ "step": 650
+ },
+ {
+ "epoch": 9.167795499235307,
+ "grad_norm": 0.3495168685913086,
+ "learning_rate": 0.0006,
+ "loss": 5.021400070190429,
+ "step": 660
+ },
+ {
+ "epoch": 9.307625081931397,
+ "grad_norm": 0.23191331326961517,
+ "learning_rate": 0.0006,
+ "loss": 5.003304672241211,
+ "step": 670
+ },
+ {
+ "epoch": 9.447454664627486,
+ "grad_norm": 0.30169883370399475,
+ "learning_rate": 0.0006,
+ "loss": 4.976224136352539,
+ "step": 680
+ },
+ {
+ "epoch": 9.587284247323574,
+ "grad_norm": 0.41019338369369507,
+ "learning_rate": 0.0006,
+ "loss": 4.991082382202149,
+ "step": 690
+ },
+ {
+ "epoch": 9.727113830019663,
+ "grad_norm": 0.3216799199581146,
+ "learning_rate": 0.0006,
+ "loss": 4.954632568359375,
+ "step": 700
+ },
+ {
+ "epoch": 9.866943412715752,
+ "grad_norm": 0.3435804545879364,
+ "learning_rate": 0.0006,
+ "loss": 4.9685718536376955,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.2994346618652344,
+ "learning_rate": 0.0006,
+ "loss": 4.951608657836914,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.471973896026611,
+ "eval_runtime": 19.8683,
+ "eval_samples_per_second": 122.859,
+ "eval_steps_per_second": 15.401,
+ "step": 720
+ },
+ {
+ "epoch": 10.13982958269609,
+ "grad_norm": 0.2370520979166031,
+ "learning_rate": 0.0006,
+ "loss": 4.911308288574219,
+ "step": 730
+ },
+ {
+ "epoch": 10.279659165392179,
+ "grad_norm": 0.22742299735546112,
+ "learning_rate": 0.0006,
+ "loss": 4.859556579589844,
+ "step": 740
+ },
+ {
+ "epoch": 10.419488748088268,
+ "grad_norm": 0.4794567823410034,
+ "learning_rate": 0.0006,
+ "loss": 4.864644622802734,
+ "step": 750
+ },
+ {
+ "epoch": 10.559318330784357,
+ "grad_norm": 0.36586564779281616,
+ "learning_rate": 0.0006,
+ "loss": 4.881325531005859,
+ "step": 760
+ },
+ {
+ "epoch": 10.699147913480445,
+ "grad_norm": 0.26905447244644165,
+ "learning_rate": 0.0006,
+ "loss": 4.8707633972167965,
+ "step": 770
+ },
+ {
+ "epoch": 10.838977496176534,
+ "grad_norm": 0.930326521396637,
+ "learning_rate": 0.0006,
+ "loss": 4.920488739013672,
+ "step": 780
+ },
+ {
+ "epoch": 10.978807078872624,
+ "grad_norm": 0.7027711272239685,
+ "learning_rate": 0.0006,
+ "loss": 5.044094467163086,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.559546947479248,
+ "eval_runtime": 19.8925,
+ "eval_samples_per_second": 122.709,
+ "eval_steps_per_second": 15.383,
+ "step": 792
+ },
+ {
+ "epoch": 11.111863666156871,
+ "grad_norm": 0.29737016558647156,
+ "learning_rate": 0.0006,
+ "loss": 4.965412521362305,
+ "step": 800
+ },
+ {
+ "epoch": 11.25169324885296,
+ "grad_norm": 0.16384148597717285,
+ "learning_rate": 0.0006,
+ "loss": 4.886568069458008,
+ "step": 810
+ },
+ {
+ "epoch": 11.39152283154905,
+ "grad_norm": 0.15612579882144928,
+ "learning_rate": 0.0006,
+ "loss": 4.853801727294922,
+ "step": 820
+ },
+ {
+ "epoch": 11.531352414245138,
+ "grad_norm": 0.16325823962688446,
+ "learning_rate": 0.0006,
+ "loss": 4.7977344512939455,
+ "step": 830
+ },
+ {
+ "epoch": 11.671181996941227,
+ "grad_norm": 0.1514514684677124,
+ "learning_rate": 0.0006,
+ "loss": 4.77706298828125,
+ "step": 840
+ },
+ {
+ "epoch": 11.811011579637317,
+ "grad_norm": 0.17136479914188385,
+ "learning_rate": 0.0006,
+ "loss": 4.776782608032226,
+ "step": 850
+ },
+ {
+ "epoch": 11.950841162333406,
+ "grad_norm": 0.1875540167093277,
+ "learning_rate": 0.0006,
+ "loss": 4.751203918457032,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.305402755737305,
+ "eval_runtime": 19.912,
+ "eval_samples_per_second": 122.59,
+ "eval_steps_per_second": 15.368,
+ "step": 864
+ },
+ {
+ "epoch": 12.083897749617654,
+ "grad_norm": 0.2090076506137848,
+ "learning_rate": 0.0006,
+ "loss": 4.742352294921875,
+ "step": 870
+ },
+ {
+ "epoch": 12.223727332313743,
+ "grad_norm": 0.19824068248271942,
+ "learning_rate": 0.0006,
+ "loss": 4.712979507446289,
+ "step": 880
+ },
+ {
+ "epoch": 12.363556915009832,
+ "grad_norm": 0.18527951836585999,
+ "learning_rate": 0.0006,
+ "loss": 4.720732498168945,
+ "step": 890
+ },
+ {
+ "epoch": 12.503386497705922,
+ "grad_norm": 0.21035733819007874,
+ "learning_rate": 0.0006,
+ "loss": 4.710549545288086,
+ "step": 900
+ },
+ {
+ "epoch": 12.64321608040201,
+ "grad_norm": 0.23118911683559418,
+ "learning_rate": 0.0006,
+ "loss": 4.696889495849609,
+ "step": 910
+ },
+ {
+ "epoch": 12.783045663098099,
+ "grad_norm": 0.27681127190589905,
+ "learning_rate": 0.0006,
+ "loss": 4.709384536743164,
+ "step": 920
+ },
+ {
+ "epoch": 12.922875245794188,
+ "grad_norm": 0.21971476078033447,
+ "learning_rate": 0.0006,
+ "loss": 4.691224670410156,
+ "step": 930
+ },
+ {
+ "epoch": 13.0,
+ "eval_loss": 4.2402753829956055,
+ "eval_runtime": 19.8836,
+ "eval_samples_per_second": 122.765,
+ "eval_steps_per_second": 15.39,
+ "step": 936
+ },
+ {
+ "epoch": 13.055931833078436,
+ "grad_norm": 0.24090802669525146,
+ "learning_rate": 0.0006,
+ "loss": 4.67314453125,
+ "step": 940
+ },
+ {
+ "epoch": 13.195761415774525,
+ "grad_norm": 0.25730693340301514,
+ "learning_rate": 0.0006,
+ "loss": 4.667090225219726,
+ "step": 950
+ },
+ {
+ "epoch": 13.335590998470614,
+ "grad_norm": 0.2326512485742569,
+ "learning_rate": 0.0006,
+ "loss": 4.660183715820312,
+ "step": 960
+ },
+ {
+ "epoch": 13.475420581166704,
+ "grad_norm": 0.27257394790649414,
+ "learning_rate": 0.0006,
+ "loss": 4.663516235351563,
+ "step": 970
+ },
+ {
+ "epoch": 13.615250163862791,
+ "grad_norm": 0.3018338084220886,
+ "learning_rate": 0.0006,
+ "loss": 4.664274978637695,
+ "step": 980
+ },
+ {
+ "epoch": 13.75507974655888,
+ "grad_norm": 0.2052220106124878,
+ "learning_rate": 0.0006,
+ "loss": 4.6318004608154295,
+ "step": 990
+ },
+ {
+ "epoch": 13.89490932925497,
+ "grad_norm": 0.19321006536483765,
+ "learning_rate": 0.0006,
+ "loss": 4.6148426055908205,
+ "step": 1000
+ },
+ {
+ "epoch": 14.0,
+ "eval_loss": 4.183226108551025,
+ "eval_runtime": 19.8764,
+ "eval_samples_per_second": 122.809,
+ "eval_steps_per_second": 15.395,
+ "step": 1008
+ },
+ {
+ "epoch": 14.027965916539218,
+ "grad_norm": 0.24733726680278778,
+ "learning_rate": 0.0006,
+ "loss": 4.6054435729980465,
+ "step": 1010
+ },
+ {
+ "epoch": 14.167795499235307,
+ "grad_norm": 0.2792834937572479,
+ "learning_rate": 0.0006,
+ "loss": 4.597950744628906,
+ "step": 1020
+ },
+ {
+ "epoch": 14.307625081931397,
+ "grad_norm": 0.28186720609664917,
+ "learning_rate": 0.0006,
+ "loss": 4.59100570678711,
+ "step": 1030
+ },
+ {
+ "epoch": 14.447454664627486,
+ "grad_norm": 0.30719828605651855,
+ "learning_rate": 0.0006,
+ "loss": 4.6152996063232425,
+ "step": 1040
+ },
+ {
+ "epoch": 14.587284247323574,
+ "grad_norm": 0.23397023975849152,
+ "learning_rate": 0.0006,
+ "loss": 4.568927764892578,
+ "step": 1050
+ },
+ {
+ "epoch": 14.727113830019663,
+ "grad_norm": 0.2331697642803192,
+ "learning_rate": 0.0006,
+ "loss": 4.594854736328125,
+ "step": 1060
+ },
+ {
+ "epoch": 14.866943412715752,
+ "grad_norm": 0.21673288941383362,
+ "learning_rate": 0.0006,
+ "loss": 4.593407440185547,
+ "step": 1070
+ },
+ {
+ "epoch": 15.0,
+ "grad_norm": 0.2536003589630127,
+ "learning_rate": 0.0006,
+ "loss": 4.579640579223633,
+ "step": 1080
+ },
+ {
+ "epoch": 15.0,
+ "eval_loss": 4.140101909637451,
+ "eval_runtime": 19.8879,
+ "eval_samples_per_second": 122.738,
+ "eval_steps_per_second": 15.386,
+ "step": 1080
+ },
+ {
+ "epoch": 15.13982958269609,
+ "grad_norm": 0.2548535466194153,
+ "learning_rate": 0.0006,
+ "loss": 4.561840438842774,
+ "step": 1090
+ },
+ {
+ "epoch": 15.279659165392179,
+ "grad_norm": 0.26136502623558044,
+ "learning_rate": 0.0006,
+ "loss": 4.54915771484375,
+ "step": 1100
+ },
+ {
+ "epoch": 15.419488748088268,
+ "grad_norm": 0.19713667035102844,
+ "learning_rate": 0.0006,
+ "loss": 4.515015411376953,
+ "step": 1110
+ },
+ {
+ "epoch": 15.559318330784357,
+ "grad_norm": 0.2860009968280792,
+ "learning_rate": 0.0006,
+ "loss": 4.512657165527344,
+ "step": 1120
+ },
+ {
+ "epoch": 15.699147913480445,
+ "grad_norm": 0.21634931862354279,
+ "learning_rate": 0.0006,
+ "loss": 4.510761260986328,
+ "step": 1130
+ },
+ {
+ "epoch": 15.838977496176534,
+ "grad_norm": 0.3173508644104004,
+ "learning_rate": 0.0006,
+ "loss": 4.52447395324707,
+ "step": 1140
+ },
+ {
+ "epoch": 15.978807078872624,
+ "grad_norm": 0.3568476140499115,
+ "learning_rate": 0.0006,
+ "loss": 4.5135456085205075,
+ "step": 1150
+ },
+ {
+ "epoch": 16.0,
+ "eval_loss": 4.103114604949951,
+ "eval_runtime": 19.8901,
+ "eval_samples_per_second": 122.724,
+ "eval_steps_per_second": 15.385,
+ "step": 1152
+ },
+ {
+ "epoch": 16.11186366615687,
+ "grad_norm": 0.28892287611961365,
+ "learning_rate": 0.0006,
+ "loss": 4.524422454833984,
+ "step": 1160
+ },
+ {
+ "epoch": 16.25169324885296,
+ "grad_norm": 0.23863045871257782,
+ "learning_rate": 0.0006,
+ "loss": 4.504647445678711,
+ "step": 1170
+ },
+ {
+ "epoch": 16.39152283154905,
+ "grad_norm": 0.2491164356470108,
+ "learning_rate": 0.0006,
+ "loss": 4.480694961547852,
+ "step": 1180
+ },
+ {
+ "epoch": 16.531352414245138,
+ "grad_norm": 0.2605704665184021,
+ "learning_rate": 0.0006,
+ "loss": 4.482597732543946,
+ "step": 1190
+ },
+ {
+ "epoch": 16.67118199694123,
+ "grad_norm": 0.2696908712387085,
+ "learning_rate": 0.0006,
+ "loss": 4.478039169311524,
+ "step": 1200
+ },
+ {
+ "epoch": 16.811011579637317,
+ "grad_norm": 0.2298387587070465,
+ "learning_rate": 0.0006,
+ "loss": 4.472258758544922,
+ "step": 1210
+ },
+ {
+ "epoch": 16.950841162333408,
+ "grad_norm": 0.25141581892967224,
+ "learning_rate": 0.0006,
+ "loss": 4.504211807250977,
+ "step": 1220
+ },
+ {
+ "epoch": 17.0,
+ "eval_loss": 4.065437316894531,
+ "eval_runtime": 19.8764,
+ "eval_samples_per_second": 122.809,
+ "eval_steps_per_second": 15.395,
+ "step": 1224
+ },
+ {
+ "epoch": 17.083897749617652,
+ "grad_norm": 0.2639216184616089,
+ "learning_rate": 0.0006,
+ "loss": 4.473295211791992,
+ "step": 1230
+ },
+ {
+ "epoch": 17.223727332313743,
+ "grad_norm": 0.26288121938705444,
+ "learning_rate": 0.0006,
+ "loss": 4.43352165222168,
+ "step": 1240
+ },
+ {
+ "epoch": 17.36355691500983,
+ "grad_norm": 0.27333375811576843,
+ "learning_rate": 0.0006,
+ "loss": 4.442911911010742,
+ "step": 1250
+ },
+ {
+ "epoch": 17.50338649770592,
+ "grad_norm": 0.25089168548583984,
+ "learning_rate": 0.0006,
+ "loss": 4.469575500488281,
+ "step": 1260
+ },
+ {
+ "epoch": 17.64321608040201,
+ "grad_norm": 0.2389490008354187,
+ "learning_rate": 0.0006,
+ "loss": 4.475759124755859,
+ "step": 1270
+ },
+ {
+ "epoch": 17.7830456630981,
+ "grad_norm": 0.2551150321960449,
+ "learning_rate": 0.0006,
+ "loss": 4.45277214050293,
+ "step": 1280
+ },
+ {
+ "epoch": 17.922875245794188,
+ "grad_norm": 0.2779748737812042,
+ "learning_rate": 0.0006,
+ "loss": 4.43327751159668,
+ "step": 1290
+ },
+ {
+ "epoch": 18.0,
+ "eval_loss": 4.038618564605713,
+ "eval_runtime": 19.8883,
+ "eval_samples_per_second": 122.736,
+ "eval_steps_per_second": 15.386,
+ "step": 1296
+ },
+ {
+ "epoch": 18.055931833078436,
+ "grad_norm": 0.2657301425933838,
+ "learning_rate": 0.0006,
+ "loss": 4.4483085632324215,
+ "step": 1300
+ },
+ {
+ "epoch": 18.195761415774523,
+ "grad_norm": 0.2548579275608063,
+ "learning_rate": 0.0006,
+ "loss": 4.398557662963867,
+ "step": 1310
+ },
+ {
+ "epoch": 18.335590998470614,
+ "grad_norm": 0.22614729404449463,
+ "learning_rate": 0.0006,
+ "loss": 4.422219085693359,
+ "step": 1320
+ },
+ {
+ "epoch": 18.475420581166702,
+ "grad_norm": 0.20935550332069397,
+ "learning_rate": 0.0006,
+ "loss": 4.401227951049805,
+ "step": 1330
+ },
+ {
+ "epoch": 18.615250163862793,
+ "grad_norm": 0.22054950892925262,
+ "learning_rate": 0.0006,
+ "loss": 4.416397476196289,
+ "step": 1340
+ },
+ {
+ "epoch": 18.75507974655888,
+ "grad_norm": 0.22855950891971588,
+ "learning_rate": 0.0006,
+ "loss": 4.436265563964843,
+ "step": 1350
+ },
+ {
+ "epoch": 18.894909329254972,
+ "grad_norm": 0.24926133453845978,
+ "learning_rate": 0.0006,
+ "loss": 4.4005176544189455,
+ "step": 1360
+ },
+ {
+ "epoch": 19.0,
+ "eval_loss": 4.012574672698975,
+ "eval_runtime": 19.897,
+ "eval_samples_per_second": 122.682,
+ "eval_steps_per_second": 15.379,
+ "step": 1368
+ },
+ {
+ "epoch": 19.027965916539216,
+ "grad_norm": 0.28323015570640564,
+ "learning_rate": 0.0006,
+ "loss": 4.406710052490235,
+ "step": 1370
+ },
+ {
+ "epoch": 19.167795499235307,
+ "grad_norm": 0.2382185012102127,
+ "learning_rate": 0.0006,
+ "loss": 4.406649398803711,
+ "step": 1380
+ },
+ {
+ "epoch": 19.307625081931395,
+ "grad_norm": 0.23818640410900116,
+ "learning_rate": 0.0006,
+ "loss": 4.373891448974609,
+ "step": 1390
+ },
+ {
+ "epoch": 19.447454664627486,
+ "grad_norm": 0.2785358130931854,
+ "learning_rate": 0.0006,
+ "loss": 4.3875068664550785,
+ "step": 1400
+ },
+ {
+ "epoch": 19.587284247323574,
+ "grad_norm": 0.24512676894664764,
+ "learning_rate": 0.0006,
+ "loss": 4.392554092407226,
+ "step": 1410
+ },
+ {
+ "epoch": 19.727113830019665,
+ "grad_norm": 0.2649998962879181,
+ "learning_rate": 0.0006,
+ "loss": 4.373421859741211,
+ "step": 1420
+ },
+ {
+ "epoch": 19.866943412715752,
+ "grad_norm": 0.2095542848110199,
+ "learning_rate": 0.0006,
+ "loss": 4.366753768920899,
+ "step": 1430
+ },
+ {
+ "epoch": 20.0,
+ "grad_norm": 0.2674562931060791,
+ "learning_rate": 0.0006,
+ "loss": 4.387431716918945,
+ "step": 1440
+ },
+ {
+ "epoch": 20.0,
+ "eval_loss": 3.989830255508423,
+ "eval_runtime": 19.9139,
+ "eval_samples_per_second": 122.578,
+ "eval_steps_per_second": 15.366,
+ "step": 1440
+ },
+ {
+ "epoch": 20.139829582696088,
+ "grad_norm": 0.32665759325027466,
+ "learning_rate": 0.0006,
+ "loss": 4.360237503051758,
+ "step": 1450
+ },
+ {
+ "epoch": 20.27965916539218,
+ "grad_norm": 1.410233736038208,
+ "learning_rate": 0.0006,
+ "loss": 4.528106689453125,
+ "step": 1460
+ },
+ {
+ "epoch": 20.419488748088266,
+ "grad_norm": 1.0754297971725464,
+ "learning_rate": 0.0006,
+ "loss": 4.761169815063477,
+ "step": 1470
+ },
+ {
+ "epoch": 20.559318330784357,
+ "grad_norm": 1.071813702583313,
+ "learning_rate": 0.0006,
+ "loss": 4.8363079071044925,
+ "step": 1480
+ },
+ {
+ "epoch": 20.699147913480445,
+ "grad_norm": 0.3152371644973755,
+ "learning_rate": 0.0006,
+ "loss": 4.819879150390625,
+ "step": 1490
+ },
+ {
+ "epoch": 20.838977496176536,
+ "grad_norm": 0.18142631649971008,
+ "learning_rate": 0.0006,
+ "loss": 4.649054718017578,
+ "step": 1500
+ },
+ {
+ "epoch": 20.978807078872624,
+ "grad_norm": 0.15613073110580444,
+ "learning_rate": 0.0006,
+ "loss": 4.554951858520508,
+ "step": 1510
+ },
+ {
+ "epoch": 21.0,
+ "eval_loss": 4.08000373840332,
+ "eval_runtime": 19.8769,
+ "eval_samples_per_second": 122.806,
+ "eval_steps_per_second": 15.395,
+ "step": 1512
+ },
+ {
+ "epoch": 21.11186366615687,
+ "grad_norm": 0.1434101015329361,
+ "learning_rate": 0.0006,
+ "loss": 4.464384841918945,
+ "step": 1520
+ },
+ {
+ "epoch": 21.25169324885296,
+ "grad_norm": 0.14000198245048523,
+ "learning_rate": 0.0006,
+ "loss": 4.396100234985352,
+ "step": 1530
+ },
+ {
+ "epoch": 21.39152283154905,
+ "grad_norm": 0.1461327075958252,
+ "learning_rate": 0.0006,
+ "loss": 4.3905376434326175,
+ "step": 1540
+ },
+ {
+ "epoch": 21.531352414245138,
+ "grad_norm": 0.1568935662508011,
+ "learning_rate": 0.0006,
+ "loss": 4.402006912231445,
+ "step": 1550
+ },
+ {
+ "epoch": 21.67118199694123,
+ "grad_norm": 0.16718244552612305,
+ "learning_rate": 0.0006,
+ "loss": 4.383187866210937,
+ "step": 1560
+ },
+ {
+ "epoch": 21.811011579637317,
+ "grad_norm": 0.14657054841518402,
+ "learning_rate": 0.0006,
+ "loss": 4.394550704956055,
+ "step": 1570
+ },
+ {
+ "epoch": 21.950841162333408,
+ "grad_norm": 0.16743497550487518,
+ "learning_rate": 0.0006,
+ "loss": 4.395104598999024,
+ "step": 1580
+ },
+ {
+ "epoch": 22.0,
+ "eval_loss": 3.9788756370544434,
+ "eval_runtime": 19.9135,
+ "eval_samples_per_second": 122.58,
+ "eval_steps_per_second": 15.366,
+ "step": 1584
+ },
+ {
+ "epoch": 22.083897749617652,
+ "grad_norm": 0.15204136073589325,
+ "learning_rate": 0.0006,
+ "loss": 4.366381072998047,
+ "step": 1590
+ },
+ {
+ "epoch": 22.223727332313743,
+ "grad_norm": 0.162191703915596,
+ "learning_rate": 0.0006,
+ "loss": 4.318870925903321,
+ "step": 1600
+ },
+ {
+ "epoch": 22.36355691500983,
+ "grad_norm": 0.15659810602664948,
+ "learning_rate": 0.0006,
+ "loss": 4.331067657470703,
+ "step": 1610
+ },
+ {
+ "epoch": 22.50338649770592,
+ "grad_norm": 0.16651582717895508,
+ "learning_rate": 0.0006,
+ "loss": 4.345297622680664,
+ "step": 1620
+ },
+ {
+ "epoch": 22.64321608040201,
+ "grad_norm": 0.1711549460887909,
+ "learning_rate": 0.0006,
+ "loss": 4.310105895996093,
+ "step": 1630
+ },
+ {
+ "epoch": 22.7830456630981,
+ "grad_norm": 0.16516436636447906,
+ "learning_rate": 0.0006,
+ "loss": 4.3384960174560545,
+ "step": 1640
+ },
+ {
+ "epoch": 22.922875245794188,
+ "grad_norm": 0.16404849290847778,
+ "learning_rate": 0.0006,
+ "loss": 4.3214569091796875,
+ "step": 1650
+ },
+ {
+ "epoch": 23.0,
+ "eval_loss": 3.952561616897583,
+ "eval_runtime": 19.8689,
+ "eval_samples_per_second": 122.855,
+ "eval_steps_per_second": 15.401,
+ "step": 1656
+ },
+ {
+ "epoch": 23.055931833078436,
+ "grad_norm": 0.19062557816505432,
+ "learning_rate": 0.0006,
+ "loss": 4.334538269042969,
+ "step": 1660
+ },
+ {
+ "epoch": 23.195761415774523,
+ "grad_norm": 0.17206895351409912,
+ "learning_rate": 0.0006,
+ "loss": 4.282015609741211,
+ "step": 1670
+ },
+ {
+ "epoch": 23.335590998470614,
+ "grad_norm": 0.20180630683898926,
+ "learning_rate": 0.0006,
+ "loss": 4.2861286163330075,
+ "step": 1680
+ },
+ {
+ "epoch": 23.475420581166702,
+ "grad_norm": 0.21979846060276031,
+ "learning_rate": 0.0006,
+ "loss": 4.320321273803711,
+ "step": 1690
+ },
+ {
+ "epoch": 23.615250163862793,
+ "grad_norm": 0.1712791621685028,
+ "learning_rate": 0.0006,
+ "loss": 4.299475479125976,
+ "step": 1700
+ },
+ {
+ "epoch": 23.75507974655888,
+ "grad_norm": 0.20566269755363464,
+ "learning_rate": 0.0006,
+ "loss": 4.338090515136718,
+ "step": 1710
+ },
+ {
+ "epoch": 23.894909329254972,
+ "grad_norm": 0.20964309573173523,
+ "learning_rate": 0.0006,
+ "loss": 4.302908325195313,
+ "step": 1720
+ },
+ {
+ "epoch": 24.0,
+ "eval_loss": 3.934921979904175,
+ "eval_runtime": 19.9067,
+ "eval_samples_per_second": 122.622,
+ "eval_steps_per_second": 15.372,
+ "step": 1728
+ },
+ {
+ "epoch": 24.027965916539216,
+ "grad_norm": 0.22795866429805756,
+ "learning_rate": 0.0006,
+ "loss": 4.306523132324219,
+ "step": 1730
+ },
+ {
+ "epoch": 24.167795499235307,
+ "grad_norm": 0.21343478560447693,
+ "learning_rate": 0.0006,
+ "loss": 4.256038665771484,
+ "step": 1740
+ },
+ {
+ "epoch": 24.307625081931395,
+ "grad_norm": 0.19679485261440277,
+ "learning_rate": 0.0006,
+ "loss": 4.2753551483154295,
+ "step": 1750
+ },
+ {
+ "epoch": 24.447454664627486,
+ "grad_norm": 0.2028961479663849,
+ "learning_rate": 0.0006,
+ "loss": 4.266365814208984,
+ "step": 1760
+ },
+ {
+ "epoch": 24.587284247323574,
+ "grad_norm": 0.1861930787563324,
+ "learning_rate": 0.0006,
+ "loss": 4.295475006103516,
+ "step": 1770
+ },
+ {
+ "epoch": 24.727113830019665,
+ "grad_norm": 0.1939159780740738,
+ "learning_rate": 0.0006,
+ "loss": 4.2797199249267575,
+ "step": 1780
+ },
+ {
+ "epoch": 24.866943412715752,
+ "grad_norm": 0.18230240046977997,
+ "learning_rate": 0.0006,
+ "loss": 4.273144149780274,
+ "step": 1790
+ },
+ {
+ "epoch": 25.0,
+ "grad_norm": 0.20727118849754333,
+ "learning_rate": 0.0006,
+ "loss": 4.269156646728516,
+ "step": 1800
+ },
+ {
+ "epoch": 25.0,
+ "eval_loss": 3.9196970462799072,
+ "eval_runtime": 19.8802,
+ "eval_samples_per_second": 122.785,
+ "eval_steps_per_second": 15.392,
+ "step": 1800
+ },
+ {
+ "epoch": 25.139829582696088,
+ "grad_norm": 0.18642550706863403,
+ "learning_rate": 0.0006,
+ "loss": 4.227664184570313,
+ "step": 1810
+ },
+ {
+ "epoch": 25.27965916539218,
+ "grad_norm": 0.2107246071100235,
+ "learning_rate": 0.0006,
+ "loss": 4.255146408081055,
+ "step": 1820
+ },
+ {
+ "epoch": 25.419488748088266,
+ "grad_norm": 0.20116163790225983,
+ "learning_rate": 0.0006,
+ "loss": 4.225793838500977,
+ "step": 1830
+ },
+ {
+ "epoch": 25.559318330784357,
+ "grad_norm": 0.21109884977340698,
+ "learning_rate": 0.0006,
+ "loss": 4.251416397094727,
+ "step": 1840
+ },
+ {
+ "epoch": 25.699147913480445,
+ "grad_norm": 0.20095095038414001,
+ "learning_rate": 0.0006,
+ "loss": 4.242668914794922,
+ "step": 1850
+ },
+ {
+ "epoch": 25.838977496176536,
+ "grad_norm": 0.20050594210624695,
+ "learning_rate": 0.0006,
+ "loss": 4.258688735961914,
+ "step": 1860
+ },
+ {
+ "epoch": 25.978807078872624,
+ "grad_norm": 0.202571839094162,
+ "learning_rate": 0.0006,
+ "loss": 4.251848602294922,
+ "step": 1870
+ },
+ {
+ "epoch": 26.0,
+ "eval_loss": 3.906507730484009,
+ "eval_runtime": 19.8714,
+ "eval_samples_per_second": 122.84,
+ "eval_steps_per_second": 15.399,
+ "step": 1872
+ },
+ {
+ "epoch": 26.11186366615687,
+ "grad_norm": 0.17989306151866913,
+ "learning_rate": 0.0006,
+ "loss": 4.22758903503418,
+ "step": 1880
+ },
+ {
+ "epoch": 26.25169324885296,
+ "grad_norm": 0.2802487313747406,
+ "learning_rate": 0.0006,
+ "loss": 4.192440795898437,
+ "step": 1890
+ },
+ {
+ "epoch": 26.39152283154905,
+ "grad_norm": 0.2439717948436737,
+ "learning_rate": 0.0006,
+ "loss": 4.2129661560058596,
+ "step": 1900
+ },
+ {
+ "epoch": 26.531352414245138,
+ "grad_norm": 0.21319910883903503,
+ "learning_rate": 0.0006,
+ "loss": 4.226039886474609,
+ "step": 1910
+ },
+ {
+ "epoch": 26.67118199694123,
+ "grad_norm": 0.21858461201190948,
+ "learning_rate": 0.0006,
+ "loss": 4.240176010131836,
+ "step": 1920
+ },
+ {
+ "epoch": 26.811011579637317,
+ "grad_norm": 0.19608710706233978,
+ "learning_rate": 0.0006,
+ "loss": 4.260812377929687,
+ "step": 1930
+ },
+ {
+ "epoch": 26.950841162333408,
+ "grad_norm": 0.1845732033252716,
+ "learning_rate": 0.0006,
+ "loss": 4.249131393432617,
+ "step": 1940
+ },
+ {
+ "epoch": 27.0,
+ "eval_loss": 3.8981809616088867,
+ "eval_runtime": 19.8774,
+ "eval_samples_per_second": 122.803,
+ "eval_steps_per_second": 15.394,
+ "step": 1944
+ },
+ {
+ "epoch": 27.083897749617652,
+ "grad_norm": 0.2204047590494156,
+ "learning_rate": 0.0006,
+ "loss": 4.219901657104492,
+ "step": 1950
+ },
+ {
+ "epoch": 27.223727332313743,
+ "grad_norm": 0.2252652645111084,
+ "learning_rate": 0.0006,
+ "loss": 4.1874244689941404,
+ "step": 1960
+ },
+ {
+ "epoch": 27.36355691500983,
+ "grad_norm": 0.2060825675725937,
+ "learning_rate": 0.0006,
+ "loss": 4.1874748229980465,
+ "step": 1970
+ },
+ {
+ "epoch": 27.50338649770592,
+ "grad_norm": 0.21779976785182953,
+ "learning_rate": 0.0006,
+ "loss": 4.211706161499023,
+ "step": 1980
+ },
+ {
+ "epoch": 27.64321608040201,
+ "grad_norm": 0.21283170580863953,
+ "learning_rate": 0.0006,
+ "loss": 4.219506072998047,
+ "step": 1990
+ },
+ {
+ "epoch": 27.7830456630981,
+ "grad_norm": 0.20306913554668427,
+ "learning_rate": 0.0006,
+ "loss": 4.204001617431641,
+ "step": 2000
+ },
+ {
+ "epoch": 27.922875245794188,
+ "grad_norm": 0.20212219655513763,
+ "learning_rate": 0.0006,
+ "loss": 4.217508316040039,
+ "step": 2010
+ },
+ {
+ "epoch": 28.0,
+ "eval_loss": 3.8878326416015625,
+ "eval_runtime": 19.8792,
+ "eval_samples_per_second": 122.792,
+ "eval_steps_per_second": 15.393,
+ "step": 2016
+ },
+ {
+ "epoch": 28.055931833078436,
+ "grad_norm": 0.21363605558872223,
+ "learning_rate": 0.0006,
+ "loss": 4.187773132324219,
+ "step": 2020
+ },
+ {
+ "epoch": 28.195761415774523,
+ "grad_norm": 0.2474210262298584,
+ "learning_rate": 0.0006,
+ "loss": 4.174351501464844,
+ "step": 2030
+ },
+ {
+ "epoch": 28.335590998470614,
+ "grad_norm": 0.22053049504756927,
+ "learning_rate": 0.0006,
+ "loss": 4.172845458984375,
+ "step": 2040
+ },
+ {
+ "epoch": 28.475420581166702,
+ "grad_norm": 0.23236072063446045,
+ "learning_rate": 0.0006,
+ "loss": 4.154056167602539,
+ "step": 2050
+ },
+ {
+ "epoch": 28.615250163862793,
+ "grad_norm": 0.20698638260364532,
+ "learning_rate": 0.0006,
+ "loss": 4.1694587707519535,
+ "step": 2060
+ },
+ {
+ "epoch": 28.75507974655888,
+ "grad_norm": 0.1927863508462906,
+ "learning_rate": 0.0006,
+ "loss": 4.1667224884033205,
+ "step": 2070
+ },
+ {
+ "epoch": 28.894909329254972,
+ "grad_norm": 0.20831426978111267,
+ "learning_rate": 0.0006,
+ "loss": 4.1969146728515625,
+ "step": 2080
+ },
+ {
+ "epoch": 29.0,
+ "eval_loss": 3.8821563720703125,
+ "eval_runtime": 19.8921,
+ "eval_samples_per_second": 122.712,
+ "eval_steps_per_second": 15.383,
+ "step": 2088
+ },
+ {
+ "epoch": 29.027965916539216,
+ "grad_norm": 0.30793341994285583,
+ "learning_rate": 0.0006,
+ "loss": 4.193627166748047,
+ "step": 2090
+ },
+ {
+ "epoch": 29.167795499235307,
+ "grad_norm": 0.24983562529087067,
+ "learning_rate": 0.0006,
+ "loss": 4.172861480712891,
+ "step": 2100
+ },
+ {
+ "epoch": 29.307625081931395,
+ "grad_norm": 0.23637285828590393,
+ "learning_rate": 0.0006,
+ "loss": 4.187952041625977,
+ "step": 2110
+ },
+ {
+ "epoch": 29.447454664627486,
+ "grad_norm": 0.20375877618789673,
+ "learning_rate": 0.0006,
+ "loss": 4.1752979278564455,
+ "step": 2120
+ },
+ {
+ "epoch": 29.587284247323574,
+ "grad_norm": 0.20303569734096527,
+ "learning_rate": 0.0006,
+ "loss": 4.151694488525391,
+ "step": 2130
+ },
+ {
+ "epoch": 29.727113830019665,
+ "grad_norm": 0.23724853992462158,
+ "learning_rate": 0.0006,
+ "loss": 4.165242004394531,
+ "step": 2140
+ },
+ {
+ "epoch": 29.866943412715752,
+ "grad_norm": 0.25325024127960205,
+ "learning_rate": 0.0006,
+ "loss": 4.146572875976562,
+ "step": 2150
+ },
+ {
+ "epoch": 30.0,
+ "grad_norm": 0.23261910676956177,
+ "learning_rate": 0.0006,
+ "loss": 4.198542785644531,
+ "step": 2160
+ },
+ {
+ "epoch": 30.0,
+ "eval_loss": 3.8707656860351562,
+ "eval_runtime": 19.8739,
+ "eval_samples_per_second": 122.825,
+ "eval_steps_per_second": 15.397,
+ "step": 2160
+ },
+ {
+ "epoch": 30.139829582696088,
+ "grad_norm": 0.22844703495502472,
+ "learning_rate": 0.0006,
+ "loss": 4.1447509765625,
+ "step": 2170
+ },
+ {
+ "epoch": 30.27965916539218,
+ "grad_norm": 0.23663879930973053,
+ "learning_rate": 0.0006,
+ "loss": 4.15783920288086,
+ "step": 2180
+ },
+ {
+ "epoch": 30.419488748088266,
+ "grad_norm": 0.20324403047561646,
+ "learning_rate": 0.0006,
+ "loss": 4.143719100952149,
+ "step": 2190
+ },
+ {
+ "epoch": 30.559318330784357,
+ "grad_norm": 0.21032340824604034,
+ "learning_rate": 0.0006,
+ "loss": 4.16072998046875,
+ "step": 2200
+ },
+ {
+ "epoch": 30.699147913480445,
+ "grad_norm": 0.18911229074001312,
+ "learning_rate": 0.0006,
+ "loss": 4.13366470336914,
+ "step": 2210
+ },
+ {
+ "epoch": 30.838977496176536,
+ "grad_norm": 0.21667662262916565,
+ "learning_rate": 0.0006,
+ "loss": 4.173492431640625,
+ "step": 2220
+ },
+ {
+ "epoch": 30.978807078872624,
+ "grad_norm": 0.22244423627853394,
+ "learning_rate": 0.0006,
+ "loss": 4.145239639282226,
+ "step": 2230
+ },
+ {
+ "epoch": 31.0,
+ "eval_loss": 3.8671557903289795,
+ "eval_runtime": 19.8641,
+ "eval_samples_per_second": 122.885,
+ "eval_steps_per_second": 15.405,
+ "step": 2232
+ },
+ {
+ "epoch": 31.11186366615687,
+ "grad_norm": 0.27511951327323914,
+ "learning_rate": 0.0006,
+ "loss": 4.129233932495117,
+ "step": 2240
+ },
+ {
+ "epoch": 31.25169324885296,
+ "grad_norm": 0.23246067762374878,
+ "learning_rate": 0.0006,
+ "loss": 4.1140899658203125,
+ "step": 2250
+ },
+ {
+ "epoch": 31.39152283154905,
+ "grad_norm": 0.21611282229423523,
+ "learning_rate": 0.0006,
+ "loss": 4.099480056762696,
+ "step": 2260
+ },
+ {
+ "epoch": 31.531352414245138,
+ "grad_norm": 0.22507013380527496,
+ "learning_rate": 0.0006,
+ "loss": 4.150814819335937,
+ "step": 2270
+ },
+ {
+ "epoch": 31.67118199694123,
+ "grad_norm": 0.24272406101226807,
+ "learning_rate": 0.0006,
+ "loss": 4.156952667236328,
+ "step": 2280
+ },
+ {
+ "epoch": 31.811011579637317,
+ "grad_norm": 0.2113332748413086,
+ "learning_rate": 0.0006,
+ "loss": 4.136573410034179,
+ "step": 2290
+ },
+ {
+ "epoch": 31.950841162333408,
+ "grad_norm": 0.21847917139530182,
+ "learning_rate": 0.0006,
+ "loss": 4.166263198852539,
+ "step": 2300
+ },
+ {
+ "epoch": 32.0,
+ "eval_loss": 3.8652803897857666,
+ "eval_runtime": 19.8683,
+ "eval_samples_per_second": 122.859,
+ "eval_steps_per_second": 15.401,
+ "step": 2304
+ },
+ {
+ "epoch": 32.083897749617655,
+ "grad_norm": 0.2485235184431076,
+ "learning_rate": 0.0006,
+ "loss": 4.126921081542969,
+ "step": 2310
+ },
+ {
+ "epoch": 32.22372733231374,
+ "grad_norm": 0.22406399250030518,
+ "learning_rate": 0.0006,
+ "loss": 4.101042556762695,
+ "step": 2320
+ },
+ {
+ "epoch": 32.36355691500983,
+ "grad_norm": 0.20872586965560913,
+ "learning_rate": 0.0006,
+ "loss": 4.122442626953125,
+ "step": 2330
+ },
+ {
+ "epoch": 32.50338649770592,
+ "grad_norm": 0.19452208280563354,
+ "learning_rate": 0.0006,
+ "loss": 4.118242263793945,
+ "step": 2340
+ },
+ {
+ "epoch": 32.64321608040201,
+ "grad_norm": 0.2058642953634262,
+ "learning_rate": 0.0006,
+ "loss": 4.140093231201172,
+ "step": 2350
+ },
+ {
+ "epoch": 32.7830456630981,
+ "grad_norm": 0.2443356215953827,
+ "learning_rate": 0.0006,
+ "loss": 4.129121398925781,
+ "step": 2360
+ },
+ {
+ "epoch": 32.92287524579419,
+ "grad_norm": 0.22833846509456635,
+ "learning_rate": 0.0006,
+ "loss": 4.125407028198242,
+ "step": 2370
+ },
+ {
+ "epoch": 33.0,
+ "eval_loss": 3.861725330352783,
+ "eval_runtime": 19.8831,
+ "eval_samples_per_second": 122.768,
+ "eval_steps_per_second": 15.39,
+ "step": 2376
+ },
+ {
+ "epoch": 33.05593183307843,
+ "grad_norm": 0.2334420382976532,
+ "learning_rate": 0.0006,
+ "loss": 4.115372848510742,
+ "step": 2380
+ },
+ {
+ "epoch": 33.19576141577453,
+ "grad_norm": 0.23251715302467346,
+ "learning_rate": 0.0006,
+ "loss": 4.101571273803711,
+ "step": 2390
+ },
+ {
+ "epoch": 33.335590998470614,
+ "grad_norm": 0.22560445964336395,
+ "learning_rate": 0.0006,
+ "loss": 4.075852966308593,
+ "step": 2400
+ },
+ {
+ "epoch": 33.4754205811667,
+ "grad_norm": 0.2696434259414673,
+ "learning_rate": 0.0006,
+ "loss": 4.103065872192383,
+ "step": 2410
+ },
+ {
+ "epoch": 33.61525016386279,
+ "grad_norm": 0.24127407371997833,
+ "learning_rate": 0.0006,
+ "loss": 4.101301574707032,
+ "step": 2420
+ },
+ {
+ "epoch": 33.755079746558884,
+ "grad_norm": 0.233340784907341,
+ "learning_rate": 0.0006,
+ "loss": 4.122187805175781,
+ "step": 2430
+ },
+ {
+ "epoch": 33.89490932925497,
+ "grad_norm": 0.23630867898464203,
+ "learning_rate": 0.0006,
+ "loss": 4.101116943359375,
+ "step": 2440
+ },
+ {
+ "epoch": 34.0,
+ "eval_loss": 3.85337233543396,
+ "eval_runtime": 19.8773,
+ "eval_samples_per_second": 122.803,
+ "eval_steps_per_second": 15.394,
+ "step": 2448
+ },
+ {
+ "epoch": 34.027965916539216,
+ "grad_norm": 0.21502625942230225,
+ "learning_rate": 0.0006,
+ "loss": 4.083441543579101,
+ "step": 2450
+ },
+ {
+ "epoch": 34.167795499235304,
+ "grad_norm": 0.26320427656173706,
+ "learning_rate": 0.0006,
+ "loss": 4.094919204711914,
+ "step": 2460
+ },
+ {
+ "epoch": 34.3076250819314,
+ "grad_norm": 0.21776047348976135,
+ "learning_rate": 0.0006,
+ "loss": 4.064165878295898,
+ "step": 2470
+ },
+ {
+ "epoch": 34.447454664627486,
+ "grad_norm": 0.27611953020095825,
+ "learning_rate": 0.0006,
+ "loss": 4.108417129516601,
+ "step": 2480
+ },
+ {
+ "epoch": 34.58728424732357,
+ "grad_norm": 0.2121397703886032,
+ "learning_rate": 0.0006,
+ "loss": 4.09581413269043,
+ "step": 2490
+ },
+ {
+ "epoch": 34.72711383001966,
+ "grad_norm": 0.218851238489151,
+ "learning_rate": 0.0006,
+ "loss": 4.098035049438477,
+ "step": 2500
+ },
+ {
+ "epoch": 34.866943412715756,
+ "grad_norm": 0.19818440079689026,
+ "learning_rate": 0.0006,
+ "loss": 4.110760498046875,
+ "step": 2510
+ },
+ {
+ "epoch": 35.0,
+ "grad_norm": 0.26101839542388916,
+ "learning_rate": 0.0006,
+ "loss": 4.104676818847656,
+ "step": 2520
+ },
+ {
+ "epoch": 35.0,
+ "eval_loss": 3.8500874042510986,
+ "eval_runtime": 19.8865,
+ "eval_samples_per_second": 122.747,
+ "eval_steps_per_second": 15.387,
+ "step": 2520
+ },
+ {
+ "epoch": 35.13982958269609,
+ "grad_norm": 0.24573564529418945,
+ "learning_rate": 0.0006,
+ "loss": 4.056778335571289,
+ "step": 2530
+ },
+ {
+ "epoch": 35.279659165392175,
+ "grad_norm": 0.25534752011299133,
+ "learning_rate": 0.0006,
+ "loss": 4.080355834960938,
+ "step": 2540
+ },
+ {
+ "epoch": 35.41948874808827,
+ "grad_norm": 0.22756867110729218,
+ "learning_rate": 0.0006,
+ "loss": 4.0860954284667965,
+ "step": 2550
+ },
+ {
+ "epoch": 35.55931833078436,
+ "grad_norm": 0.2059873342514038,
+ "learning_rate": 0.0006,
+ "loss": 4.08325080871582,
+ "step": 2560
+ },
+ {
+ "epoch": 35.699147913480445,
+ "grad_norm": 0.2249329537153244,
+ "learning_rate": 0.0006,
+ "loss": 4.077912521362305,
+ "step": 2570
+ },
+ {
+ "epoch": 35.83897749617653,
+ "grad_norm": 0.18873104453086853,
+ "learning_rate": 0.0006,
+ "loss": 4.079425430297851,
+ "step": 2580
+ },
+ {
+ "epoch": 35.97880707887263,
+ "grad_norm": 0.21719087660312653,
+ "learning_rate": 0.0006,
+ "loss": 4.101816558837891,
+ "step": 2590
+ },
+ {
+ "epoch": 36.0,
+ "eval_loss": 3.8457438945770264,
+ "eval_runtime": 19.8667,
+ "eval_samples_per_second": 122.869,
+ "eval_steps_per_second": 15.403,
+ "step": 2592
+ },
+ {
+ "epoch": 36.11186366615687,
+ "grad_norm": 0.24146319925785065,
+ "learning_rate": 0.0006,
+ "loss": 4.027724838256836,
+ "step": 2600
+ },
+ {
+ "epoch": 36.25169324885296,
+ "grad_norm": 0.25700923800468445,
+ "learning_rate": 0.0006,
+ "loss": 4.063429260253907,
+ "step": 2610
+ },
+ {
+ "epoch": 36.39152283154905,
+ "grad_norm": 0.24079975485801697,
+ "learning_rate": 0.0006,
+ "loss": 4.044811630249024,
+ "step": 2620
+ },
+ {
+ "epoch": 36.53135241424514,
+ "grad_norm": 0.22490380704402924,
+ "learning_rate": 0.0006,
+ "loss": 4.047432708740234,
+ "step": 2630
+ },
+ {
+ "epoch": 36.67118199694123,
+ "grad_norm": 0.21978461742401123,
+ "learning_rate": 0.0006,
+ "loss": 4.062580108642578,
+ "step": 2640
+ },
+ {
+ "epoch": 36.81101157963732,
+ "grad_norm": 0.22579775750637054,
+ "learning_rate": 0.0006,
+ "loss": 4.077471923828125,
+ "step": 2650
+ },
+ {
+ "epoch": 36.950841162333404,
+ "grad_norm": 0.20858746767044067,
+ "learning_rate": 0.0006,
+ "loss": 4.096049499511719,
+ "step": 2660
+ },
+ {
+ "epoch": 37.0,
+ "eval_loss": 3.8466100692749023,
+ "eval_runtime": 19.8626,
+ "eval_samples_per_second": 122.895,
+ "eval_steps_per_second": 15.406,
+ "step": 2664
+ },
+ {
+ "epoch": 37.083897749617655,
+ "grad_norm": 0.23128123581409454,
+ "learning_rate": 0.0006,
+ "loss": 4.020519638061524,
+ "step": 2670
+ },
+ {
+ "epoch": 37.22372733231374,
+ "grad_norm": 0.2819666564464569,
+ "learning_rate": 0.0006,
+ "loss": 4.016598129272461,
+ "step": 2680
+ },
+ {
+ "epoch": 37.36355691500983,
+ "grad_norm": 0.20954866707324982,
+ "learning_rate": 0.0006,
+ "loss": 4.0719951629638675,
+ "step": 2690
+ },
+ {
+ "epoch": 37.50338649770592,
+ "grad_norm": 0.252839595079422,
+ "learning_rate": 0.0006,
+ "loss": 4.052394104003906,
+ "step": 2700
+ },
+ {
+ "epoch": 37.64321608040201,
+ "grad_norm": 0.23314858973026276,
+ "learning_rate": 0.0006,
+ "loss": 4.0449882507324215,
+ "step": 2710
+ },
+ {
+ "epoch": 37.7830456630981,
+ "grad_norm": 0.22690322995185852,
+ "learning_rate": 0.0006,
+ "loss": 4.054507064819336,
+ "step": 2720
+ },
+ {
+ "epoch": 37.92287524579419,
+ "grad_norm": 0.23433133959770203,
+ "learning_rate": 0.0006,
+ "loss": 4.088650894165039,
+ "step": 2730
+ },
+ {
+ "epoch": 38.0,
+ "eval_loss": 3.8431267738342285,
+ "eval_runtime": 19.8719,
+ "eval_samples_per_second": 122.837,
+ "eval_steps_per_second": 15.399,
+ "step": 2736
+ },
+ {
+ "epoch": 38.05593183307843,
+ "grad_norm": 0.2319370061159134,
+ "learning_rate": 0.0006,
+ "loss": 4.0494140625,
+ "step": 2740
+ },
+ {
+ "epoch": 38.19576141577453,
+ "grad_norm": 0.3028583824634552,
+ "learning_rate": 0.0006,
+ "loss": 4.028265380859375,
+ "step": 2750
+ },
+ {
+ "epoch": 38.335590998470614,
+ "grad_norm": 0.26089686155319214,
+ "learning_rate": 0.0006,
+ "loss": 4.039775466918945,
+ "step": 2760
+ },
+ {
+ "epoch": 38.4754205811667,
+ "grad_norm": 0.20860013365745544,
+ "learning_rate": 0.0006,
+ "loss": 4.035010528564453,
+ "step": 2770
+ },
+ {
+ "epoch": 38.61525016386279,
+ "grad_norm": 0.20843325555324554,
+ "learning_rate": 0.0006,
+ "loss": 4.03155517578125,
+ "step": 2780
+ },
+ {
+ "epoch": 38.755079746558884,
+ "grad_norm": 0.2232809066772461,
+ "learning_rate": 0.0006,
+ "loss": 4.072791671752929,
+ "step": 2790
+ },
+ {
+ "epoch": 38.89490932925497,
+ "grad_norm": 0.1945791244506836,
+ "learning_rate": 0.0006,
+ "loss": 4.040574264526367,
+ "step": 2800
+ },
+ {
+ "epoch": 39.0,
+ "eval_loss": 3.835956335067749,
+ "eval_runtime": 19.8739,
+ "eval_samples_per_second": 122.825,
+ "eval_steps_per_second": 15.397,
+ "step": 2808
+ },
+ {
+ "epoch": 39.027965916539216,
+ "grad_norm": 0.2554682195186615,
+ "learning_rate": 0.0006,
+ "loss": 4.071317291259765,
+ "step": 2810
+ },
+ {
+ "epoch": 39.167795499235304,
+ "grad_norm": 0.2313416302204132,
+ "learning_rate": 0.0006,
+ "loss": 4.024376296997071,
+ "step": 2820
+ },
+ {
+ "epoch": 39.3076250819314,
+ "grad_norm": 0.2648560404777527,
+ "learning_rate": 0.0006,
+ "loss": 4.010868835449219,
+ "step": 2830
+ },
+ {
+ "epoch": 39.447454664627486,
+ "grad_norm": 0.22031910717487335,
+ "learning_rate": 0.0006,
+ "loss": 4.0235858917236325,
+ "step": 2840
+ },
+ {
+ "epoch": 39.58728424732357,
+ "grad_norm": 0.2400754690170288,
+ "learning_rate": 0.0006,
+ "loss": 4.018476486206055,
+ "step": 2850
+ },
+ {
+ "epoch": 39.72711383001966,
+ "grad_norm": 0.2010141760110855,
+ "learning_rate": 0.0006,
+ "loss": 4.049578475952148,
+ "step": 2860
+ },
+ {
+ "epoch": 39.866943412715756,
+ "grad_norm": 0.19912582635879517,
+ "learning_rate": 0.0006,
+ "loss": 4.029824447631836,
+ "step": 2870
+ },
+ {
+ "epoch": 40.0,
+ "grad_norm": 0.24000345170497894,
+ "learning_rate": 0.0006,
+ "loss": 4.025931549072266,
+ "step": 2880
+ },
+ {
+ "epoch": 40.0,
+ "eval_loss": 3.836862802505493,
+ "eval_runtime": 19.8743,
+ "eval_samples_per_second": 122.822,
+ "eval_steps_per_second": 15.397,
+ "step": 2880
+ },
+ {
+ "epoch": 40.13982958269609,
+ "grad_norm": 0.24903830885887146,
+ "learning_rate": 0.0006,
+ "loss": 3.988302993774414,
+ "step": 2890
+ },
+ {
+ "epoch": 40.279659165392175,
+ "grad_norm": 0.28228405117988586,
+ "learning_rate": 0.0006,
+ "loss": 4.027346801757813,
+ "step": 2900
+ },
+ {
+ "epoch": 40.41948874808827,
+ "grad_norm": 0.28054535388946533,
+ "learning_rate": 0.0006,
+ "loss": 4.017681503295899,
+ "step": 2910
+ },
+ {
+ "epoch": 40.55931833078436,
+ "grad_norm": 0.2016926407814026,
+ "learning_rate": 0.0006,
+ "loss": 4.014138031005859,
+ "step": 2920
+ },
+ {
+ "epoch": 40.699147913480445,
+ "grad_norm": 0.2401750236749649,
+ "learning_rate": 0.0006,
+ "loss": 4.023651885986328,
+ "step": 2930
+ },
+ {
+ "epoch": 40.83897749617653,
+ "grad_norm": 0.19922944903373718,
+ "learning_rate": 0.0006,
+ "loss": 4.024148941040039,
+ "step": 2940
+ },
+ {
+ "epoch": 40.97880707887263,
+ "grad_norm": 0.2199106067419052,
+ "learning_rate": 0.0006,
+ "loss": 4.0512535095214846,
+ "step": 2950
+ },
+ {
+ "epoch": 41.0,
+ "eval_loss": 3.836369276046753,
+ "eval_runtime": 19.8646,
+ "eval_samples_per_second": 122.882,
+ "eval_steps_per_second": 15.404,
+ "step": 2952
+ },
+ {
+ "epoch": 41.11186366615687,
+ "grad_norm": 0.2492665946483612,
+ "learning_rate": 0.0006,
+ "loss": 3.9661277770996093,
+ "step": 2960
+ },
+ {
+ "epoch": 41.25169324885296,
+ "grad_norm": 0.21760401129722595,
+ "learning_rate": 0.0006,
+ "loss": 3.9954681396484375,
+ "step": 2970
+ },
+ {
+ "epoch": 41.39152283154905,
+ "grad_norm": 0.21843111515045166,
+ "learning_rate": 0.0006,
+ "loss": 4.004875946044922,
+ "step": 2980
+ },
+ {
+ "epoch": 41.53135241424514,
+ "grad_norm": 0.2930936813354492,
+ "learning_rate": 0.0006,
+ "loss": 4.039703750610352,
+ "step": 2990
+ },
+ {
+ "epoch": 41.67118199694123,
+ "grad_norm": 0.2324160635471344,
+ "learning_rate": 0.0006,
+ "loss": 4.005767440795898,
+ "step": 3000
+ },
+ {
+ "epoch": 41.81101157963732,
+ "grad_norm": 0.2255973368883133,
+ "learning_rate": 0.0006,
+ "loss": 4.026718521118164,
+ "step": 3010
+ },
+ {
+ "epoch": 41.950841162333404,
+ "grad_norm": 0.2313026636838913,
+ "learning_rate": 0.0006,
+ "loss": 4.009809875488282,
+ "step": 3020
+ },
+ {
+ "epoch": 42.0,
+ "eval_loss": 3.834172487258911,
+ "eval_runtime": 19.8883,
+ "eval_samples_per_second": 122.735,
+ "eval_steps_per_second": 15.386,
+ "step": 3024
+ },
+ {
+ "epoch": 42.083897749617655,
+ "grad_norm": 0.2603314220905304,
+ "learning_rate": 0.0006,
+ "loss": 3.997126007080078,
+ "step": 3030
+ },
+ {
+ "epoch": 42.22372733231374,
+ "grad_norm": 0.21272456645965576,
+ "learning_rate": 0.0006,
+ "loss": 3.9705890655517577,
+ "step": 3040
+ },
+ {
+ "epoch": 42.36355691500983,
+ "grad_norm": 0.23962801694869995,
+ "learning_rate": 0.0006,
+ "loss": 3.981968307495117,
+ "step": 3050
+ },
+ {
+ "epoch": 42.50338649770592,
+ "grad_norm": 0.23225201666355133,
+ "learning_rate": 0.0006,
+ "loss": 4.003329849243164,
+ "step": 3060
+ },
+ {
+ "epoch": 42.64321608040201,
+ "grad_norm": 0.23806191980838776,
+ "learning_rate": 0.0006,
+ "loss": 4.015252304077149,
+ "step": 3070
+ },
+ {
+ "epoch": 42.7830456630981,
+ "grad_norm": 0.19710606336593628,
+ "learning_rate": 0.0006,
+ "loss": 4.033316802978516,
+ "step": 3080
+ },
+ {
+ "epoch": 42.92287524579419,
+ "grad_norm": 0.20284956693649292,
+ "learning_rate": 0.0006,
+ "loss": 4.021148681640625,
+ "step": 3090
+ },
+ {
+ "epoch": 43.0,
+ "eval_loss": 3.8335306644439697,
+ "eval_runtime": 19.8661,
+ "eval_samples_per_second": 122.873,
+ "eval_steps_per_second": 15.403,
+ "step": 3096
+ },
+ {
+ "epoch": 43.05593183307843,
+ "grad_norm": 0.21578088402748108,
+ "learning_rate": 0.0006,
+ "loss": 4.000060653686523,
+ "step": 3100
+ },
+ {
+ "epoch": 43.19576141577453,
+ "grad_norm": 0.2411411553621292,
+ "learning_rate": 0.0006,
+ "loss": 3.9653553009033202,
+ "step": 3110
+ },
+ {
+ "epoch": 43.335590998470614,
+ "grad_norm": 0.2169276475906372,
+ "learning_rate": 0.0006,
+ "loss": 3.976711654663086,
+ "step": 3120
+ },
+ {
+ "epoch": 43.4754205811667,
+ "grad_norm": 0.21465395390987396,
+ "learning_rate": 0.0006,
+ "loss": 3.9969348907470703,
+ "step": 3130
+ },
+ {
+ "epoch": 43.61525016386279,
+ "grad_norm": 0.22791877388954163,
+ "learning_rate": 0.0006,
+ "loss": 3.986632537841797,
+ "step": 3140
+ },
+ {
+ "epoch": 43.755079746558884,
+ "grad_norm": 0.22883419692516327,
+ "learning_rate": 0.0006,
+ "loss": 4.008829116821289,
+ "step": 3150
+ },
+ {
+ "epoch": 43.89490932925497,
+ "grad_norm": 0.21380403637886047,
+ "learning_rate": 0.0006,
+ "loss": 3.9839096069335938,
+ "step": 3160
+ },
+ {
+ "epoch": 44.0,
+ "eval_loss": 3.8331058025360107,
+ "eval_runtime": 19.8801,
+ "eval_samples_per_second": 122.786,
+ "eval_steps_per_second": 15.392,
+ "step": 3168
+ },
+ {
+ "epoch": 44.027965916539216,
+ "grad_norm": 0.21809452772140503,
+ "learning_rate": 0.0006,
+ "loss": 4.008271789550781,
+ "step": 3170
+ },
+ {
+ "epoch": 44.167795499235304,
+ "grad_norm": 0.26727327704429626,
+ "learning_rate": 0.0006,
+ "loss": 3.972920608520508,
+ "step": 3180
+ },
+ {
+ "epoch": 44.3076250819314,
+ "grad_norm": 0.23454609513282776,
+ "learning_rate": 0.0006,
+ "loss": 3.955358123779297,
+ "step": 3190
+ },
+ {
+ "epoch": 44.447454664627486,
+ "grad_norm": 0.2097703069448471,
+ "learning_rate": 0.0006,
+ "loss": 3.9747833251953124,
+ "step": 3200
+ },
+ {
+ "epoch": 44.58728424732357,
+ "grad_norm": 0.23480531573295593,
+ "learning_rate": 0.0006,
+ "loss": 3.9892623901367186,
+ "step": 3210
+ },
+ {
+ "epoch": 44.72711383001966,
+ "grad_norm": 0.23049408197402954,
+ "learning_rate": 0.0006,
+ "loss": 3.9728965759277344,
+ "step": 3220
+ },
+ {
+ "epoch": 44.866943412715756,
+ "grad_norm": 0.2217615842819214,
+ "learning_rate": 0.0006,
+ "loss": 3.999728775024414,
+ "step": 3230
+ },
+ {
+ "epoch": 45.0,
+ "grad_norm": 0.2554522454738617,
+ "learning_rate": 0.0006,
+ "loss": 4.002607727050782,
+ "step": 3240
+ },
+ {
+ "epoch": 45.0,
+ "eval_loss": 3.8344907760620117,
+ "eval_runtime": 19.8751,
+ "eval_samples_per_second": 122.817,
+ "eval_steps_per_second": 15.396,
+ "step": 3240
+ },
+ {
+ "epoch": 45.13982958269609,
+ "grad_norm": 0.3009033203125,
+ "learning_rate": 0.0006,
+ "loss": 3.9571372985839846,
+ "step": 3250
+ },
+ {
+ "epoch": 45.279659165392175,
+ "grad_norm": 0.2306888848543167,
+ "learning_rate": 0.0006,
+ "loss": 3.9317432403564454,
+ "step": 3260
+ },
+ {
+ "epoch": 45.41948874808827,
+ "grad_norm": 0.2440161257982254,
+ "learning_rate": 0.0006,
+ "loss": 3.963428497314453,
+ "step": 3270
+ },
+ {
+ "epoch": 45.55931833078436,
+ "grad_norm": 0.2652886211872101,
+ "learning_rate": 0.0006,
+ "loss": 3.9786258697509767,
+ "step": 3280
+ },
+ {
+ "epoch": 45.699147913480445,
+ "grad_norm": 0.2337835133075714,
+ "learning_rate": 0.0006,
+ "loss": 3.9886405944824217,
+ "step": 3290
+ },
+ {
+ "epoch": 45.83897749617653,
+ "grad_norm": 0.22756986320018768,
+ "learning_rate": 0.0006,
+ "loss": 3.9842605590820312,
+ "step": 3300
+ },
+ {
+ "epoch": 45.97880707887263,
+ "grad_norm": 0.19978488981723785,
+ "learning_rate": 0.0006,
+ "loss": 3.989816665649414,
+ "step": 3310
+ },
+ {
+ "epoch": 46.0,
+ "eval_loss": 3.839942216873169,
+ "eval_runtime": 19.8509,
+ "eval_samples_per_second": 122.966,
+ "eval_steps_per_second": 15.415,
+ "step": 3312
+ },
+ {
+ "epoch": 46.11186366615687,
+ "grad_norm": 0.2835380434989929,
+ "learning_rate": 0.0006,
+ "loss": 3.9531864166259765,
+ "step": 3320
+ },
+ {
+ "epoch": 46.25169324885296,
+ "grad_norm": 0.22056172788143158,
+ "learning_rate": 0.0006,
+ "loss": 3.939357376098633,
+ "step": 3330
+ },
+ {
+ "epoch": 46.39152283154905,
+ "grad_norm": 0.20216430723667145,
+ "learning_rate": 0.0006,
+ "loss": 3.9580955505371094,
+ "step": 3340
+ },
+ {
+ "epoch": 46.53135241424514,
+ "grad_norm": 0.21733932197093964,
+ "learning_rate": 0.0006,
+ "loss": 3.9605991363525392,
+ "step": 3350
+ },
+ {
+ "epoch": 46.67118199694123,
+ "grad_norm": 0.23969708383083344,
+ "learning_rate": 0.0006,
+ "loss": 3.9617855072021486,
+ "step": 3360
+ },
+ {
+ "epoch": 46.81101157963732,
+ "grad_norm": 0.22183765470981598,
+ "learning_rate": 0.0006,
+ "loss": 3.989068603515625,
+ "step": 3370
+ },
+ {
+ "epoch": 46.950841162333404,
+ "grad_norm": 0.2626979947090149,
+ "learning_rate": 0.0006,
+ "loss": 4.008098983764649,
+ "step": 3380
+ },
+ {
+ "epoch": 47.0,
+ "eval_loss": 3.837268829345703,
+ "eval_runtime": 19.8738,
+ "eval_samples_per_second": 122.825,
+ "eval_steps_per_second": 15.397,
+ "step": 3384
+ },
+ {
+ "epoch": 47.083897749617655,
+ "grad_norm": 0.25732189416885376,
+ "learning_rate": 0.0006,
+ "loss": 3.9304744720458986,
+ "step": 3390
+ },
+ {
+ "epoch": 47.22372733231374,
+ "grad_norm": 0.24863329529762268,
+ "learning_rate": 0.0006,
+ "loss": 3.9565696716308594,
+ "step": 3400
+ },
+ {
+ "epoch": 47.36355691500983,
+ "grad_norm": 0.25764819979667664,
+ "learning_rate": 0.0006,
+ "loss": 3.9362415313720702,
+ "step": 3410
+ },
+ {
+ "epoch": 47.50338649770592,
+ "grad_norm": 0.27024373412132263,
+ "learning_rate": 0.0006,
+ "loss": 3.940362548828125,
+ "step": 3420
+ },
+ {
+ "epoch": 47.64321608040201,
+ "grad_norm": 0.24121522903442383,
+ "learning_rate": 0.0006,
+ "loss": 3.956980514526367,
+ "step": 3430
+ },
+ {
+ "epoch": 47.7830456630981,
+ "grad_norm": 0.22264418005943298,
+ "learning_rate": 0.0006,
+ "loss": 3.9536224365234376,
+ "step": 3440
+ },
+ {
+ "epoch": 47.92287524579419,
+ "grad_norm": 0.22733484208583832,
+ "learning_rate": 0.0006,
+ "loss": 3.980303955078125,
+ "step": 3450
+ },
+ {
+ "epoch": 48.0,
+ "eval_loss": 3.8341763019561768,
+ "eval_runtime": 19.8675,
+ "eval_samples_per_second": 122.864,
+ "eval_steps_per_second": 15.402,
+ "step": 3456
+ },
+ {
+ "epoch": 48.05593183307843,
+ "grad_norm": 0.26292598247528076,
+ "learning_rate": 0.0006,
+ "loss": 3.9665447235107423,
+ "step": 3460
+ },
+ {
+ "epoch": 48.19576141577453,
+ "grad_norm": 0.22519847750663757,
+ "learning_rate": 0.0006,
+ "loss": 3.9315723419189452,
+ "step": 3470
+ },
+ {
+ "epoch": 48.335590998470614,
+ "grad_norm": 0.22407089173793793,
+ "learning_rate": 0.0006,
+ "loss": 3.916579818725586,
+ "step": 3480
+ },
+ {
+ "epoch": 48.4754205811667,
+ "grad_norm": 0.20874175429344177,
+ "learning_rate": 0.0006,
+ "loss": 3.9572383880615236,
+ "step": 3490
+ },
+ {
+ "epoch": 48.61525016386279,
+ "grad_norm": 0.2706707715988159,
+ "learning_rate": 0.0006,
+ "loss": 3.9582786560058594,
+ "step": 3500
+ },
+ {
+ "epoch": 48.755079746558884,
+ "grad_norm": 0.2232997566461563,
+ "learning_rate": 0.0006,
+ "loss": 3.9701709747314453,
+ "step": 3510
+ },
+ {
+ "epoch": 48.89490932925497,
+ "grad_norm": 0.22266264259815216,
+ "learning_rate": 0.0006,
+ "loss": 3.939614105224609,
+ "step": 3520
+ },
+ {
+ "epoch": 49.0,
+ "eval_loss": 3.8400838375091553,
+ "eval_runtime": 19.878,
+ "eval_samples_per_second": 122.799,
+ "eval_steps_per_second": 15.394,
+ "step": 3528
+ },
+ {
+ "epoch": 49.027965916539216,
+ "grad_norm": 0.2621210217475891,
+ "learning_rate": 0.0006,
+ "loss": 3.961019515991211,
+ "step": 3530
+ },
+ {
+ "epoch": 49.167795499235304,
+ "grad_norm": 0.25230973958969116,
+ "learning_rate": 0.0006,
+ "loss": 3.9141735076904296,
+ "step": 3540
+ },
+ {
+ "epoch": 49.3076250819314,
+ "grad_norm": 0.23983345925807953,
+ "learning_rate": 0.0006,
+ "loss": 3.9238140106201174,
+ "step": 3550
+ },
+ {
+ "epoch": 49.447454664627486,
+ "grad_norm": 0.234044149518013,
+ "learning_rate": 0.0006,
+ "loss": 3.9433780670166017,
+ "step": 3560
+ },
+ {
+ "epoch": 49.58728424732357,
+ "grad_norm": 0.22092850506305695,
+ "learning_rate": 0.0006,
+ "loss": 3.942871856689453,
+ "step": 3570
+ },
+ {
+ "epoch": 49.72711383001966,
+ "grad_norm": 0.2227325737476349,
+ "learning_rate": 0.0006,
+ "loss": 3.926601791381836,
+ "step": 3580
+ },
+ {
+ "epoch": 49.866943412715756,
+ "grad_norm": 0.20453153550624847,
+ "learning_rate": 0.0006,
+ "loss": 3.9523616790771485,
+ "step": 3590
+ },
+ {
+ "epoch": 50.0,
+ "grad_norm": 0.25963273644447327,
+ "learning_rate": 0.0006,
+ "loss": 3.9487586975097657,
+ "step": 3600
+ },
+ {
+ "epoch": 50.0,
+ "eval_loss": 3.8405935764312744,
+ "eval_runtime": 19.8765,
+ "eval_samples_per_second": 122.809,
+ "eval_steps_per_second": 15.395,
+ "step": 3600
+ },
+ {
+ "epoch": 50.13982958269609,
+ "grad_norm": 0.22767190635204315,
+ "learning_rate": 0.0006,
+ "loss": 3.8952316284179687,
+ "step": 3610
+ },
+ {
+ "epoch": 50.279659165392175,
+ "grad_norm": 0.26602327823638916,
+ "learning_rate": 0.0006,
+ "loss": 3.8838024139404297,
+ "step": 3620
+ },
+ {
+ "epoch": 50.41948874808827,
+ "grad_norm": 0.21582292020320892,
+ "learning_rate": 0.0006,
+ "loss": 3.928426742553711,
+ "step": 3630
+ },
+ {
+ "epoch": 50.55931833078436,
+ "grad_norm": 0.23087719082832336,
+ "learning_rate": 0.0006,
+ "loss": 3.948084259033203,
+ "step": 3640
+ },
+ {
+ "epoch": 50.699147913480445,
+ "grad_norm": 0.23141410946846008,
+ "learning_rate": 0.0006,
+ "loss": 3.932170104980469,
+ "step": 3650
+ },
+ {
+ "epoch": 50.83897749617653,
+ "grad_norm": 0.23195211589336395,
+ "learning_rate": 0.0006,
+ "loss": 3.925021743774414,
+ "step": 3660
+ },
+ {
+ "epoch": 50.97880707887263,
+ "grad_norm": 0.21663589775562286,
+ "learning_rate": 0.0006,
+ "loss": 3.947396469116211,
+ "step": 3670
+ },
+ {
+ "epoch": 51.0,
+ "eval_loss": 3.835434913635254,
+ "eval_runtime": 19.8868,
+ "eval_samples_per_second": 122.745,
+ "eval_steps_per_second": 15.387,
+ "step": 3672
+ },
+ {
+ "epoch": 51.11186366615687,
+ "grad_norm": 0.23005850613117218,
+ "learning_rate": 0.0006,
+ "loss": 3.928109359741211,
+ "step": 3680
+ },
+ {
+ "epoch": 51.25169324885296,
+ "grad_norm": 0.2656649947166443,
+ "learning_rate": 0.0006,
+ "loss": 3.911693572998047,
+ "step": 3690
+ },
+ {
+ "epoch": 51.39152283154905,
+ "grad_norm": 0.27117303013801575,
+ "learning_rate": 0.0006,
+ "loss": 3.902240753173828,
+ "step": 3700
+ },
+ {
+ "epoch": 51.53135241424514,
+ "grad_norm": 0.22568346560001373,
+ "learning_rate": 0.0006,
+ "loss": 3.938275146484375,
+ "step": 3710
+ },
+ {
+ "epoch": 51.67118199694123,
+ "grad_norm": 0.21113169193267822,
+ "learning_rate": 0.0006,
+ "loss": 3.9464157104492186,
+ "step": 3720
+ },
+ {
+ "epoch": 51.81101157963732,
+ "grad_norm": 0.2244177907705307,
+ "learning_rate": 0.0006,
+ "loss": 3.917864990234375,
+ "step": 3730
+ },
+ {
+ "epoch": 51.950841162333404,
+ "grad_norm": 0.21250809729099274,
+ "learning_rate": 0.0006,
+ "loss": 3.959328460693359,
+ "step": 3740
+ },
+ {
+ "epoch": 52.0,
+ "eval_loss": 3.8328680992126465,
+ "eval_runtime": 19.8889,
+ "eval_samples_per_second": 122.732,
+ "eval_steps_per_second": 15.385,
+ "step": 3744
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 7200,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 100,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.59506627396685e+18,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/train.log b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..62ac6fa1b7ecdbb331f2f61496748ef2a8ddb1d8
--- /dev/null
+++ b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s43/train.log
@@ -0,0 +1,12756 @@
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 100
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 8
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 64
+[transformers] Total optimization steps = 7,200
+[transformers] Number of trainable parameters = 145,836,544
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: setting up run cj32nrrj
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260917_031121-cj32nrrj
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t1a_S2_C2_r2l50_i5exp_s43
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/cj32nrrj
+
0%| | 0/7200 [00:00system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/checkpoint-3744/config.json b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/checkpoint-3744/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3474619f10931027097cc05d837296fa4f33a6c
--- /dev/null
+++ b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/checkpoint-3744/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 512,
+ "initializer_range": 0.02,
+ "intermediate_size": 1536,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 4,
+ "num_hidden_layers": 20,
+ "num_key_value_heads": 4,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151676
+}
diff --git a/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/checkpoint-3744/generation_config.json b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/checkpoint-3744/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/checkpoint-3744/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/checkpoint-3744/tokenizer_config.json b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/checkpoint-3744/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d43c97a447f4b25a8c36be5873f7b622935fa90
--- /dev/null
+++ b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/checkpoint-3744/tokenizer_config.json
@@ -0,0 +1,24 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>",
+ "<|next_1_pred|>",
+ "<|next_2_pred|>",
+ "<|next_3_pred|>",
+ "<|next_4_pred|>",
+ "<|next_5_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/checkpoint-3744/trainer_state.json b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/checkpoint-3744/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..0686e753c10353c1b13cd9630c59593efeb12157
--- /dev/null
+++ b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/checkpoint-3744/trainer_state.json
@@ -0,0 +1,3075 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 52.0,
+ "eval_steps": 72,
+ "global_step": 3744,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013982958269608914,
+ "grad_norm": 2.716968059539795,
+ "learning_rate": 0.0,
+ "loss": 12.023050308227539,
+ "step": 1
+ },
+ {
+ "epoch": 0.13982958269608914,
+ "grad_norm": 1.7333585023880005,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.668445163302952,
+ "step": 10
+ },
+ {
+ "epoch": 0.2796591653921783,
+ "grad_norm": 1.674895167350769,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 10.976360321044922,
+ "step": 20
+ },
+ {
+ "epoch": 0.4194887480882674,
+ "grad_norm": 1.6312729120254517,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 10.191786956787109,
+ "step": 30
+ },
+ {
+ "epoch": 0.5593183307843566,
+ "grad_norm": 1.4509426355361938,
+ "learning_rate": 0.000234,
+ "loss": 9.150357818603515,
+ "step": 40
+ },
+ {
+ "epoch": 0.6991479134804457,
+ "grad_norm": 0.8043447732925415,
+ "learning_rate": 0.000294,
+ "loss": 8.19864730834961,
+ "step": 50
+ },
+ {
+ "epoch": 0.8389774961765348,
+ "grad_norm": 0.285184383392334,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 7.694214630126953,
+ "step": 60
+ },
+ {
+ "epoch": 0.978807078872624,
+ "grad_norm": 0.19545111060142517,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 7.582648468017578,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.4924163818359375,
+ "eval_runtime": 43.5798,
+ "eval_samples_per_second": 56.012,
+ "eval_steps_per_second": 7.022,
+ "step": 72
+ },
+ {
+ "epoch": 1.1118636661568713,
+ "grad_norm": 0.4004134237766266,
+ "learning_rate": 0.000474,
+ "loss": 7.519743347167969,
+ "step": 80
+ },
+ {
+ "epoch": 1.2516932488529604,
+ "grad_norm": 0.23644468188285828,
+ "learning_rate": 0.000534,
+ "loss": 7.442707824707031,
+ "step": 90
+ },
+ {
+ "epoch": 1.3915228315490495,
+ "grad_norm": 0.31028902530670166,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 7.406687164306641,
+ "step": 100
+ },
+ {
+ "epoch": 1.5313524142451387,
+ "grad_norm": 0.20037660002708435,
+ "learning_rate": 0.0006,
+ "loss": 7.391224670410156,
+ "step": 110
+ },
+ {
+ "epoch": 1.671181996941228,
+ "grad_norm": 0.20043577253818512,
+ "learning_rate": 0.0006,
+ "loss": 7.3573463439941404,
+ "step": 120
+ },
+ {
+ "epoch": 1.811011579637317,
+ "grad_norm": 0.19670860469341278,
+ "learning_rate": 0.0006,
+ "loss": 7.322447204589844,
+ "step": 130
+ },
+ {
+ "epoch": 1.9508411623334063,
+ "grad_norm": 0.38997384905815125,
+ "learning_rate": 0.0006,
+ "loss": 7.298429870605469,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 7.155577659606934,
+ "eval_runtime": 19.8721,
+ "eval_samples_per_second": 122.836,
+ "eval_steps_per_second": 15.398,
+ "step": 144
+ },
+ {
+ "epoch": 2.0838977496176536,
+ "grad_norm": 0.2497471123933792,
+ "learning_rate": 0.0006,
+ "loss": 7.248709106445313,
+ "step": 150
+ },
+ {
+ "epoch": 2.2237273323137425,
+ "grad_norm": 0.29498448967933655,
+ "learning_rate": 0.0006,
+ "loss": 7.218118286132812,
+ "step": 160
+ },
+ {
+ "epoch": 2.363556915009832,
+ "grad_norm": 0.2413434386253357,
+ "learning_rate": 0.0006,
+ "loss": 7.174160003662109,
+ "step": 170
+ },
+ {
+ "epoch": 2.503386497705921,
+ "grad_norm": 0.7065138220787048,
+ "learning_rate": 0.0006,
+ "loss": 7.1481071472167965,
+ "step": 180
+ },
+ {
+ "epoch": 2.64321608040201,
+ "grad_norm": 0.3012448847293854,
+ "learning_rate": 0.0006,
+ "loss": 7.091287994384766,
+ "step": 190
+ },
+ {
+ "epoch": 2.783045663098099,
+ "grad_norm": 0.6359490752220154,
+ "learning_rate": 0.0006,
+ "loss": 7.021817779541015,
+ "step": 200
+ },
+ {
+ "epoch": 2.9228752457941884,
+ "grad_norm": 0.6794478893280029,
+ "learning_rate": 0.0006,
+ "loss": 6.968614196777343,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 6.6224188804626465,
+ "eval_runtime": 19.8756,
+ "eval_samples_per_second": 122.814,
+ "eval_steps_per_second": 15.396,
+ "step": 216
+ },
+ {
+ "epoch": 3.0559318330784357,
+ "grad_norm": 0.25412076711654663,
+ "learning_rate": 0.0006,
+ "loss": 6.8762260437011715,
+ "step": 220
+ },
+ {
+ "epoch": 3.1957614157745247,
+ "grad_norm": 0.2262560874223709,
+ "learning_rate": 0.0006,
+ "loss": 6.767330169677734,
+ "step": 230
+ },
+ {
+ "epoch": 3.335590998470614,
+ "grad_norm": 0.5441655516624451,
+ "learning_rate": 0.0006,
+ "loss": 6.695590972900391,
+ "step": 240
+ },
+ {
+ "epoch": 3.475420581166703,
+ "grad_norm": 0.3228594660758972,
+ "learning_rate": 0.0006,
+ "loss": 6.621391296386719,
+ "step": 250
+ },
+ {
+ "epoch": 3.6152501638627923,
+ "grad_norm": 0.2617524564266205,
+ "learning_rate": 0.0006,
+ "loss": 6.566825103759766,
+ "step": 260
+ },
+ {
+ "epoch": 3.755079746558881,
+ "grad_norm": 0.5654861927032471,
+ "learning_rate": 0.0006,
+ "loss": 6.522350311279297,
+ "step": 270
+ },
+ {
+ "epoch": 3.8949093292549706,
+ "grad_norm": 0.3757103681564331,
+ "learning_rate": 0.0006,
+ "loss": 6.475904083251953,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 6.099272727966309,
+ "eval_runtime": 19.8891,
+ "eval_samples_per_second": 122.73,
+ "eval_steps_per_second": 15.385,
+ "step": 288
+ },
+ {
+ "epoch": 4.027965916539218,
+ "grad_norm": 0.5570326447486877,
+ "learning_rate": 0.0006,
+ "loss": 6.409722137451172,
+ "step": 290
+ },
+ {
+ "epoch": 4.167795499235307,
+ "grad_norm": 0.28371670842170715,
+ "learning_rate": 0.0006,
+ "loss": 6.379258728027343,
+ "step": 300
+ },
+ {
+ "epoch": 4.307625081931396,
+ "grad_norm": 0.35956287384033203,
+ "learning_rate": 0.0006,
+ "loss": 6.299806213378906,
+ "step": 310
+ },
+ {
+ "epoch": 4.447454664627485,
+ "grad_norm": 0.7799175977706909,
+ "learning_rate": 0.0006,
+ "loss": 6.255207443237305,
+ "step": 320
+ },
+ {
+ "epoch": 4.587284247323574,
+ "grad_norm": 0.5951794385910034,
+ "learning_rate": 0.0006,
+ "loss": 6.2078605651855465,
+ "step": 330
+ },
+ {
+ "epoch": 4.727113830019664,
+ "grad_norm": 0.5990219712257385,
+ "learning_rate": 0.0006,
+ "loss": 6.172432327270508,
+ "step": 340
+ },
+ {
+ "epoch": 4.866943412715752,
+ "grad_norm": 0.5756080746650696,
+ "learning_rate": 0.0006,
+ "loss": 6.122249984741211,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.5394350290298462,
+ "learning_rate": 0.0006,
+ "loss": 6.109564208984375,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 5.717533111572266,
+ "eval_runtime": 19.8836,
+ "eval_samples_per_second": 122.764,
+ "eval_steps_per_second": 15.39,
+ "step": 360
+ },
+ {
+ "epoch": 5.139829582696089,
+ "grad_norm": 0.23153530061244965,
+ "learning_rate": 0.0006,
+ "loss": 6.047309494018554,
+ "step": 370
+ },
+ {
+ "epoch": 5.279659165392179,
+ "grad_norm": 0.5463221073150635,
+ "learning_rate": 0.0006,
+ "loss": 5.977159881591797,
+ "step": 380
+ },
+ {
+ "epoch": 5.419488748088267,
+ "grad_norm": 0.6735644340515137,
+ "learning_rate": 0.0006,
+ "loss": 5.995932388305664,
+ "step": 390
+ },
+ {
+ "epoch": 5.5593183307843566,
+ "grad_norm": 0.2792721390724182,
+ "learning_rate": 0.0006,
+ "loss": 5.981506729125977,
+ "step": 400
+ },
+ {
+ "epoch": 5.699147913480446,
+ "grad_norm": 0.17974849045276642,
+ "learning_rate": 0.0006,
+ "loss": 5.879684829711914,
+ "step": 410
+ },
+ {
+ "epoch": 5.838977496176534,
+ "grad_norm": 0.3589136600494385,
+ "learning_rate": 0.0006,
+ "loss": 5.832319641113282,
+ "step": 420
+ },
+ {
+ "epoch": 5.978807078872624,
+ "grad_norm": 0.6607341766357422,
+ "learning_rate": 0.0006,
+ "loss": 5.847049713134766,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 5.435446739196777,
+ "eval_runtime": 19.8953,
+ "eval_samples_per_second": 122.692,
+ "eval_steps_per_second": 15.381,
+ "step": 432
+ },
+ {
+ "epoch": 6.1118636661568715,
+ "grad_norm": 0.25870782136917114,
+ "learning_rate": 0.0006,
+ "loss": 5.788164520263672,
+ "step": 440
+ },
+ {
+ "epoch": 6.251693248852961,
+ "grad_norm": 0.22492191195487976,
+ "learning_rate": 0.0006,
+ "loss": 5.721003723144531,
+ "step": 450
+ },
+ {
+ "epoch": 6.391522831549049,
+ "grad_norm": 0.5248213410377502,
+ "learning_rate": 0.0006,
+ "loss": 5.681539916992188,
+ "step": 460
+ },
+ {
+ "epoch": 6.531352414245139,
+ "grad_norm": 0.318158894777298,
+ "learning_rate": 0.0006,
+ "loss": 5.695879364013672,
+ "step": 470
+ },
+ {
+ "epoch": 6.671181996941228,
+ "grad_norm": 0.24011744558811188,
+ "learning_rate": 0.0006,
+ "loss": 5.630021667480468,
+ "step": 480
+ },
+ {
+ "epoch": 6.811011579637317,
+ "grad_norm": 0.39051371812820435,
+ "learning_rate": 0.0006,
+ "loss": 5.590905380249024,
+ "step": 490
+ },
+ {
+ "epoch": 6.950841162333406,
+ "grad_norm": 0.5720425248146057,
+ "learning_rate": 0.0006,
+ "loss": 5.578094100952148,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 5.130133628845215,
+ "eval_runtime": 19.8856,
+ "eval_samples_per_second": 122.752,
+ "eval_steps_per_second": 15.388,
+ "step": 504
+ },
+ {
+ "epoch": 7.083897749617654,
+ "grad_norm": 0.4470057785511017,
+ "learning_rate": 0.0006,
+ "loss": 5.535285186767578,
+ "step": 510
+ },
+ {
+ "epoch": 7.223727332313743,
+ "grad_norm": 0.29476362466812134,
+ "learning_rate": 0.0006,
+ "loss": 5.491307067871094,
+ "step": 520
+ },
+ {
+ "epoch": 7.3635569150098314,
+ "grad_norm": 0.41337665915489197,
+ "learning_rate": 0.0006,
+ "loss": 5.453485107421875,
+ "step": 530
+ },
+ {
+ "epoch": 7.503386497705921,
+ "grad_norm": 0.5688338279724121,
+ "learning_rate": 0.0006,
+ "loss": 5.437624740600586,
+ "step": 540
+ },
+ {
+ "epoch": 7.64321608040201,
+ "grad_norm": 0.29169437289237976,
+ "learning_rate": 0.0006,
+ "loss": 5.427764511108398,
+ "step": 550
+ },
+ {
+ "epoch": 7.7830456630980995,
+ "grad_norm": 0.22401539981365204,
+ "learning_rate": 0.0006,
+ "loss": 5.353483963012695,
+ "step": 560
+ },
+ {
+ "epoch": 7.922875245794188,
+ "grad_norm": 0.4272184371948242,
+ "learning_rate": 0.0006,
+ "loss": 5.3211315155029295,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.847578048706055,
+ "eval_runtime": 19.8896,
+ "eval_samples_per_second": 122.727,
+ "eval_steps_per_second": 15.385,
+ "step": 576
+ },
+ {
+ "epoch": 8.055931833078436,
+ "grad_norm": 0.35211363434791565,
+ "learning_rate": 0.0006,
+ "loss": 5.277224349975586,
+ "step": 580
+ },
+ {
+ "epoch": 8.195761415774525,
+ "grad_norm": 0.46232008934020996,
+ "learning_rate": 0.0006,
+ "loss": 5.236256408691406,
+ "step": 590
+ },
+ {
+ "epoch": 8.335590998470614,
+ "grad_norm": 0.31235164403915405,
+ "learning_rate": 0.0006,
+ "loss": 5.2266803741455075,
+ "step": 600
+ },
+ {
+ "epoch": 8.475420581166704,
+ "grad_norm": 0.2106219381093979,
+ "learning_rate": 0.0006,
+ "loss": 5.162085342407226,
+ "step": 610
+ },
+ {
+ "epoch": 8.615250163862791,
+ "grad_norm": 0.27147120237350464,
+ "learning_rate": 0.0006,
+ "loss": 5.1779296875,
+ "step": 620
+ },
+ {
+ "epoch": 8.75507974655888,
+ "grad_norm": 0.33056122064590454,
+ "learning_rate": 0.0006,
+ "loss": 5.140441513061523,
+ "step": 630
+ },
+ {
+ "epoch": 8.89490932925497,
+ "grad_norm": 0.385438472032547,
+ "learning_rate": 0.0006,
+ "loss": 5.113404846191406,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.659756183624268,
+ "eval_runtime": 19.8832,
+ "eval_samples_per_second": 122.767,
+ "eval_steps_per_second": 15.39,
+ "step": 648
+ },
+ {
+ "epoch": 9.027965916539218,
+ "grad_norm": 0.38440024852752686,
+ "learning_rate": 0.0006,
+ "loss": 5.108001708984375,
+ "step": 650
+ },
+ {
+ "epoch": 9.167795499235307,
+ "grad_norm": 0.3112102448940277,
+ "learning_rate": 0.0006,
+ "loss": 5.07113037109375,
+ "step": 660
+ },
+ {
+ "epoch": 9.307625081931397,
+ "grad_norm": 0.3126074969768524,
+ "learning_rate": 0.0006,
+ "loss": 5.028032684326172,
+ "step": 670
+ },
+ {
+ "epoch": 9.447454664627486,
+ "grad_norm": 0.23846904933452606,
+ "learning_rate": 0.0006,
+ "loss": 5.016418838500977,
+ "step": 680
+ },
+ {
+ "epoch": 9.587284247323574,
+ "grad_norm": 0.32244792580604553,
+ "learning_rate": 0.0006,
+ "loss": 5.0063835144042965,
+ "step": 690
+ },
+ {
+ "epoch": 9.727113830019663,
+ "grad_norm": 0.30531519651412964,
+ "learning_rate": 0.0006,
+ "loss": 4.988818359375,
+ "step": 700
+ },
+ {
+ "epoch": 9.866943412715752,
+ "grad_norm": 0.318118691444397,
+ "learning_rate": 0.0006,
+ "loss": 5.00194091796875,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.3512510657310486,
+ "learning_rate": 0.0006,
+ "loss": 4.952599716186524,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.51050329208374,
+ "eval_runtime": 19.8841,
+ "eval_samples_per_second": 122.762,
+ "eval_steps_per_second": 15.389,
+ "step": 720
+ },
+ {
+ "epoch": 10.13982958269609,
+ "grad_norm": 0.4009549617767334,
+ "learning_rate": 0.0006,
+ "loss": 4.926473236083984,
+ "step": 730
+ },
+ {
+ "epoch": 10.279659165392179,
+ "grad_norm": 1.5823439359664917,
+ "learning_rate": 0.0006,
+ "loss": 5.045626068115235,
+ "step": 740
+ },
+ {
+ "epoch": 10.419488748088268,
+ "grad_norm": 0.9733429551124573,
+ "learning_rate": 0.0006,
+ "loss": 5.245881652832031,
+ "step": 750
+ },
+ {
+ "epoch": 10.559318330784357,
+ "grad_norm": 0.6840938925743103,
+ "learning_rate": 0.0006,
+ "loss": 5.181113433837891,
+ "step": 760
+ },
+ {
+ "epoch": 10.699147913480445,
+ "grad_norm": 0.21602104604244232,
+ "learning_rate": 0.0006,
+ "loss": 5.0521598815917965,
+ "step": 770
+ },
+ {
+ "epoch": 10.838977496176534,
+ "grad_norm": 0.16830213367938995,
+ "learning_rate": 0.0006,
+ "loss": 4.968352508544922,
+ "step": 780
+ },
+ {
+ "epoch": 10.978807078872624,
+ "grad_norm": 0.19245372712612152,
+ "learning_rate": 0.0006,
+ "loss": 4.957720184326172,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.460480690002441,
+ "eval_runtime": 19.8817,
+ "eval_samples_per_second": 122.776,
+ "eval_steps_per_second": 15.391,
+ "step": 792
+ },
+ {
+ "epoch": 11.111863666156871,
+ "grad_norm": 0.16914726793766022,
+ "learning_rate": 0.0006,
+ "loss": 4.882259750366211,
+ "step": 800
+ },
+ {
+ "epoch": 11.25169324885296,
+ "grad_norm": 0.1492013782262802,
+ "learning_rate": 0.0006,
+ "loss": 4.8725944519042965,
+ "step": 810
+ },
+ {
+ "epoch": 11.39152283154905,
+ "grad_norm": 0.1677086055278778,
+ "learning_rate": 0.0006,
+ "loss": 4.8613433837890625,
+ "step": 820
+ },
+ {
+ "epoch": 11.531352414245138,
+ "grad_norm": 0.17891162633895874,
+ "learning_rate": 0.0006,
+ "loss": 4.826210784912109,
+ "step": 830
+ },
+ {
+ "epoch": 11.671181996941227,
+ "grad_norm": 0.1669146567583084,
+ "learning_rate": 0.0006,
+ "loss": 4.8305011749267575,
+ "step": 840
+ },
+ {
+ "epoch": 11.811011579637317,
+ "grad_norm": 0.2779156267642975,
+ "learning_rate": 0.0006,
+ "loss": 4.8207557678222654,
+ "step": 850
+ },
+ {
+ "epoch": 11.950841162333406,
+ "grad_norm": 0.17464399337768555,
+ "learning_rate": 0.0006,
+ "loss": 4.797770309448242,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.346872329711914,
+ "eval_runtime": 19.8492,
+ "eval_samples_per_second": 122.977,
+ "eval_steps_per_second": 15.416,
+ "step": 864
+ },
+ {
+ "epoch": 12.083897749617654,
+ "grad_norm": 0.23368847370147705,
+ "learning_rate": 0.0006,
+ "loss": 4.751280975341797,
+ "step": 870
+ },
+ {
+ "epoch": 12.223727332313743,
+ "grad_norm": 0.31480148434638977,
+ "learning_rate": 0.0006,
+ "loss": 4.772942733764649,
+ "step": 880
+ },
+ {
+ "epoch": 12.363556915009832,
+ "grad_norm": 0.1941957026720047,
+ "learning_rate": 0.0006,
+ "loss": 4.7654777526855465,
+ "step": 890
+ },
+ {
+ "epoch": 12.503386497705922,
+ "grad_norm": 0.2883318066596985,
+ "learning_rate": 0.0006,
+ "loss": 4.746197509765625,
+ "step": 900
+ },
+ {
+ "epoch": 12.64321608040201,
+ "grad_norm": 0.21352677047252655,
+ "learning_rate": 0.0006,
+ "loss": 4.748310089111328,
+ "step": 910
+ },
+ {
+ "epoch": 12.783045663098099,
+ "grad_norm": 0.2179621458053589,
+ "learning_rate": 0.0006,
+ "loss": 4.746096801757813,
+ "step": 920
+ },
+ {
+ "epoch": 12.922875245794188,
+ "grad_norm": 0.2569679915904999,
+ "learning_rate": 0.0006,
+ "loss": 4.7146240234375,
+ "step": 930
+ },
+ {
+ "epoch": 13.0,
+ "eval_loss": 4.275017261505127,
+ "eval_runtime": 19.8584,
+ "eval_samples_per_second": 122.92,
+ "eval_steps_per_second": 15.409,
+ "step": 936
+ },
+ {
+ "epoch": 13.055931833078436,
+ "grad_norm": 0.32321077585220337,
+ "learning_rate": 0.0006,
+ "loss": 4.711003112792969,
+ "step": 940
+ },
+ {
+ "epoch": 13.195761415774525,
+ "grad_norm": 0.24348779022693634,
+ "learning_rate": 0.0006,
+ "loss": 4.7029365539550785,
+ "step": 950
+ },
+ {
+ "epoch": 13.335590998470614,
+ "grad_norm": 0.26076191663742065,
+ "learning_rate": 0.0006,
+ "loss": 4.690544128417969,
+ "step": 960
+ },
+ {
+ "epoch": 13.475420581166704,
+ "grad_norm": 0.3161758482456207,
+ "learning_rate": 0.0006,
+ "loss": 4.690534591674805,
+ "step": 970
+ },
+ {
+ "epoch": 13.615250163862791,
+ "grad_norm": 0.2838556170463562,
+ "learning_rate": 0.0006,
+ "loss": 4.665595245361328,
+ "step": 980
+ },
+ {
+ "epoch": 13.75507974655888,
+ "grad_norm": 0.27124014496803284,
+ "learning_rate": 0.0006,
+ "loss": 4.692533874511719,
+ "step": 990
+ },
+ {
+ "epoch": 13.89490932925497,
+ "grad_norm": 0.2232910692691803,
+ "learning_rate": 0.0006,
+ "loss": 4.658015441894531,
+ "step": 1000
+ },
+ {
+ "epoch": 14.0,
+ "eval_loss": 4.217504024505615,
+ "eval_runtime": 19.8873,
+ "eval_samples_per_second": 122.741,
+ "eval_steps_per_second": 15.387,
+ "step": 1008
+ },
+ {
+ "epoch": 14.027965916539218,
+ "grad_norm": 0.2071516513824463,
+ "learning_rate": 0.0006,
+ "loss": 4.665455627441406,
+ "step": 1010
+ },
+ {
+ "epoch": 14.167795499235307,
+ "grad_norm": 0.29424795508384705,
+ "learning_rate": 0.0006,
+ "loss": 4.641806411743164,
+ "step": 1020
+ },
+ {
+ "epoch": 14.307625081931397,
+ "grad_norm": 0.25501739978790283,
+ "learning_rate": 0.0006,
+ "loss": 4.651218032836914,
+ "step": 1030
+ },
+ {
+ "epoch": 14.447454664627486,
+ "grad_norm": 0.21987558901309967,
+ "learning_rate": 0.0006,
+ "loss": 4.639051818847657,
+ "step": 1040
+ },
+ {
+ "epoch": 14.587284247323574,
+ "grad_norm": 0.25687575340270996,
+ "learning_rate": 0.0006,
+ "loss": 4.626768493652344,
+ "step": 1050
+ },
+ {
+ "epoch": 14.727113830019663,
+ "grad_norm": 0.21611753106117249,
+ "learning_rate": 0.0006,
+ "loss": 4.610270690917969,
+ "step": 1060
+ },
+ {
+ "epoch": 14.866943412715752,
+ "grad_norm": 0.23722431063652039,
+ "learning_rate": 0.0006,
+ "loss": 4.621290588378907,
+ "step": 1070
+ },
+ {
+ "epoch": 15.0,
+ "grad_norm": 0.24093493819236755,
+ "learning_rate": 0.0006,
+ "loss": 4.593955993652344,
+ "step": 1080
+ },
+ {
+ "epoch": 15.0,
+ "eval_loss": 4.171814918518066,
+ "eval_runtime": 19.8854,
+ "eval_samples_per_second": 122.754,
+ "eval_steps_per_second": 15.388,
+ "step": 1080
+ },
+ {
+ "epoch": 15.13982958269609,
+ "grad_norm": 0.28158125281333923,
+ "learning_rate": 0.0006,
+ "loss": 4.604044723510742,
+ "step": 1090
+ },
+ {
+ "epoch": 15.279659165392179,
+ "grad_norm": 0.32815518975257874,
+ "learning_rate": 0.0006,
+ "loss": 4.582751083374023,
+ "step": 1100
+ },
+ {
+ "epoch": 15.419488748088268,
+ "grad_norm": 0.2801415026187897,
+ "learning_rate": 0.0006,
+ "loss": 4.593767547607422,
+ "step": 1110
+ },
+ {
+ "epoch": 15.559318330784357,
+ "grad_norm": 0.22562557458877563,
+ "learning_rate": 0.0006,
+ "loss": 4.593327331542969,
+ "step": 1120
+ },
+ {
+ "epoch": 15.699147913480445,
+ "grad_norm": 0.22806066274642944,
+ "learning_rate": 0.0006,
+ "loss": 4.535354232788086,
+ "step": 1130
+ },
+ {
+ "epoch": 15.838977496176534,
+ "grad_norm": 0.2634636461734772,
+ "learning_rate": 0.0006,
+ "loss": 4.543705749511719,
+ "step": 1140
+ },
+ {
+ "epoch": 15.978807078872624,
+ "grad_norm": 0.21634097397327423,
+ "learning_rate": 0.0006,
+ "loss": 4.5338890075683596,
+ "step": 1150
+ },
+ {
+ "epoch": 16.0,
+ "eval_loss": 4.131479263305664,
+ "eval_runtime": 19.8573,
+ "eval_samples_per_second": 122.927,
+ "eval_steps_per_second": 15.41,
+ "step": 1152
+ },
+ {
+ "epoch": 16.11186366615687,
+ "grad_norm": 0.2419835329055786,
+ "learning_rate": 0.0006,
+ "loss": 4.5284889221191404,
+ "step": 1160
+ },
+ {
+ "epoch": 16.25169324885296,
+ "grad_norm": 0.3165692687034607,
+ "learning_rate": 0.0006,
+ "loss": 4.513931655883789,
+ "step": 1170
+ },
+ {
+ "epoch": 16.39152283154905,
+ "grad_norm": 0.2896401286125183,
+ "learning_rate": 0.0006,
+ "loss": 4.528255844116211,
+ "step": 1180
+ },
+ {
+ "epoch": 16.531352414245138,
+ "grad_norm": 0.194514662027359,
+ "learning_rate": 0.0006,
+ "loss": 4.537871170043945,
+ "step": 1190
+ },
+ {
+ "epoch": 16.67118199694123,
+ "grad_norm": 0.25165173411369324,
+ "learning_rate": 0.0006,
+ "loss": 4.523268508911133,
+ "step": 1200
+ },
+ {
+ "epoch": 16.811011579637317,
+ "grad_norm": 0.2500514090061188,
+ "learning_rate": 0.0006,
+ "loss": 4.501129150390625,
+ "step": 1210
+ },
+ {
+ "epoch": 16.950841162333408,
+ "grad_norm": 0.2884882688522339,
+ "learning_rate": 0.0006,
+ "loss": 4.51097297668457,
+ "step": 1220
+ },
+ {
+ "epoch": 17.0,
+ "eval_loss": 4.097098350524902,
+ "eval_runtime": 19.8837,
+ "eval_samples_per_second": 122.764,
+ "eval_steps_per_second": 15.39,
+ "step": 1224
+ },
+ {
+ "epoch": 17.083897749617652,
+ "grad_norm": 0.36393827199935913,
+ "learning_rate": 0.0006,
+ "loss": 4.483684539794922,
+ "step": 1230
+ },
+ {
+ "epoch": 17.223727332313743,
+ "grad_norm": 0.31065353751182556,
+ "learning_rate": 0.0006,
+ "loss": 4.451641845703125,
+ "step": 1240
+ },
+ {
+ "epoch": 17.36355691500983,
+ "grad_norm": 0.25007814168930054,
+ "learning_rate": 0.0006,
+ "loss": 4.49823112487793,
+ "step": 1250
+ },
+ {
+ "epoch": 17.50338649770592,
+ "grad_norm": 0.22179868817329407,
+ "learning_rate": 0.0006,
+ "loss": 4.492696762084961,
+ "step": 1260
+ },
+ {
+ "epoch": 17.64321608040201,
+ "grad_norm": 0.21064846217632294,
+ "learning_rate": 0.0006,
+ "loss": 4.490806198120117,
+ "step": 1270
+ },
+ {
+ "epoch": 17.7830456630981,
+ "grad_norm": 0.25718459486961365,
+ "learning_rate": 0.0006,
+ "loss": 4.494770812988281,
+ "step": 1280
+ },
+ {
+ "epoch": 17.922875245794188,
+ "grad_norm": 0.2847071886062622,
+ "learning_rate": 0.0006,
+ "loss": 4.49290771484375,
+ "step": 1290
+ },
+ {
+ "epoch": 18.0,
+ "eval_loss": 4.067162990570068,
+ "eval_runtime": 19.8778,
+ "eval_samples_per_second": 122.8,
+ "eval_steps_per_second": 15.394,
+ "step": 1296
+ },
+ {
+ "epoch": 18.055931833078436,
+ "grad_norm": 0.2647097110748291,
+ "learning_rate": 0.0006,
+ "loss": 4.449942398071289,
+ "step": 1300
+ },
+ {
+ "epoch": 18.195761415774523,
+ "grad_norm": 0.253885954618454,
+ "learning_rate": 0.0006,
+ "loss": 4.418960952758789,
+ "step": 1310
+ },
+ {
+ "epoch": 18.335590998470614,
+ "grad_norm": 0.23532676696777344,
+ "learning_rate": 0.0006,
+ "loss": 4.452471160888672,
+ "step": 1320
+ },
+ {
+ "epoch": 18.475420581166702,
+ "grad_norm": 0.20562738180160522,
+ "learning_rate": 0.0006,
+ "loss": 4.417825698852539,
+ "step": 1330
+ },
+ {
+ "epoch": 18.615250163862793,
+ "grad_norm": 0.20547060668468475,
+ "learning_rate": 0.0006,
+ "loss": 4.445845031738282,
+ "step": 1340
+ },
+ {
+ "epoch": 18.75507974655888,
+ "grad_norm": 0.24597883224487305,
+ "learning_rate": 0.0006,
+ "loss": 4.4469043731689455,
+ "step": 1350
+ },
+ {
+ "epoch": 18.894909329254972,
+ "grad_norm": 0.2579265236854553,
+ "learning_rate": 0.0006,
+ "loss": 4.482316970825195,
+ "step": 1360
+ },
+ {
+ "epoch": 19.0,
+ "eval_loss": 4.040419101715088,
+ "eval_runtime": 19.8742,
+ "eval_samples_per_second": 122.823,
+ "eval_steps_per_second": 15.397,
+ "step": 1368
+ },
+ {
+ "epoch": 19.027965916539216,
+ "grad_norm": 0.26405206322669983,
+ "learning_rate": 0.0006,
+ "loss": 4.427297973632813,
+ "step": 1370
+ },
+ {
+ "epoch": 19.167795499235307,
+ "grad_norm": 0.3001002371311188,
+ "learning_rate": 0.0006,
+ "loss": 4.376183319091797,
+ "step": 1380
+ },
+ {
+ "epoch": 19.307625081931395,
+ "grad_norm": 0.225426584482193,
+ "learning_rate": 0.0006,
+ "loss": 4.411513137817383,
+ "step": 1390
+ },
+ {
+ "epoch": 19.447454664627486,
+ "grad_norm": 0.24004659056663513,
+ "learning_rate": 0.0006,
+ "loss": 4.408379745483399,
+ "step": 1400
+ },
+ {
+ "epoch": 19.587284247323574,
+ "grad_norm": 0.26392531394958496,
+ "learning_rate": 0.0006,
+ "loss": 4.405710220336914,
+ "step": 1410
+ },
+ {
+ "epoch": 19.727113830019665,
+ "grad_norm": 3.5607895851135254,
+ "learning_rate": 0.0006,
+ "loss": 4.8364215850830075,
+ "step": 1420
+ },
+ {
+ "epoch": 19.866943412715752,
+ "grad_norm": 1.176288366317749,
+ "learning_rate": 0.0006,
+ "loss": 5.104848098754883,
+ "step": 1430
+ },
+ {
+ "epoch": 20.0,
+ "grad_norm": 0.9165453314781189,
+ "learning_rate": 0.0006,
+ "loss": 5.082521057128906,
+ "step": 1440
+ },
+ {
+ "epoch": 20.0,
+ "eval_loss": 4.527106285095215,
+ "eval_runtime": 19.8888,
+ "eval_samples_per_second": 122.732,
+ "eval_steps_per_second": 15.386,
+ "step": 1440
+ },
+ {
+ "epoch": 20.139829582696088,
+ "grad_norm": 0.3012751340866089,
+ "learning_rate": 0.0006,
+ "loss": 4.852622222900391,
+ "step": 1450
+ },
+ {
+ "epoch": 20.27965916539218,
+ "grad_norm": 0.1736743450164795,
+ "learning_rate": 0.0006,
+ "loss": 4.676758193969727,
+ "step": 1460
+ },
+ {
+ "epoch": 20.419488748088266,
+ "grad_norm": 0.17676010727882385,
+ "learning_rate": 0.0006,
+ "loss": 4.5816596984863285,
+ "step": 1470
+ },
+ {
+ "epoch": 20.559318330784357,
+ "grad_norm": 0.18164357542991638,
+ "learning_rate": 0.0006,
+ "loss": 4.555968856811523,
+ "step": 1480
+ },
+ {
+ "epoch": 20.699147913480445,
+ "grad_norm": 0.14983311295509338,
+ "learning_rate": 0.0006,
+ "loss": 4.475181579589844,
+ "step": 1490
+ },
+ {
+ "epoch": 20.838977496176536,
+ "grad_norm": 0.14589060842990875,
+ "learning_rate": 0.0006,
+ "loss": 4.490390777587891,
+ "step": 1500
+ },
+ {
+ "epoch": 20.978807078872624,
+ "grad_norm": 0.15187016129493713,
+ "learning_rate": 0.0006,
+ "loss": 4.446943664550782,
+ "step": 1510
+ },
+ {
+ "epoch": 21.0,
+ "eval_loss": 4.041965484619141,
+ "eval_runtime": 19.8885,
+ "eval_samples_per_second": 122.734,
+ "eval_steps_per_second": 15.386,
+ "step": 1512
+ },
+ {
+ "epoch": 21.11186366615687,
+ "grad_norm": 0.1582321971654892,
+ "learning_rate": 0.0006,
+ "loss": 4.4322154998779295,
+ "step": 1520
+ },
+ {
+ "epoch": 21.25169324885296,
+ "grad_norm": 0.15622903406620026,
+ "learning_rate": 0.0006,
+ "loss": 4.423052978515625,
+ "step": 1530
+ },
+ {
+ "epoch": 21.39152283154905,
+ "grad_norm": 0.16327011585235596,
+ "learning_rate": 0.0006,
+ "loss": 4.390462112426758,
+ "step": 1540
+ },
+ {
+ "epoch": 21.531352414245138,
+ "grad_norm": 0.1648152470588684,
+ "learning_rate": 0.0006,
+ "loss": 4.420136260986328,
+ "step": 1550
+ },
+ {
+ "epoch": 21.67118199694123,
+ "grad_norm": 0.17459477484226227,
+ "learning_rate": 0.0006,
+ "loss": 4.404214477539062,
+ "step": 1560
+ },
+ {
+ "epoch": 21.811011579637317,
+ "grad_norm": 0.18171466886997223,
+ "learning_rate": 0.0006,
+ "loss": 4.3947807312011715,
+ "step": 1570
+ },
+ {
+ "epoch": 21.950841162333408,
+ "grad_norm": 0.18350829184055328,
+ "learning_rate": 0.0006,
+ "loss": 4.385610961914063,
+ "step": 1580
+ },
+ {
+ "epoch": 22.0,
+ "eval_loss": 3.9978244304656982,
+ "eval_runtime": 19.8923,
+ "eval_samples_per_second": 122.711,
+ "eval_steps_per_second": 15.383,
+ "step": 1584
+ },
+ {
+ "epoch": 22.083897749617652,
+ "grad_norm": 0.18790191411972046,
+ "learning_rate": 0.0006,
+ "loss": 4.371077346801758,
+ "step": 1590
+ },
+ {
+ "epoch": 22.223727332313743,
+ "grad_norm": 0.192194864153862,
+ "learning_rate": 0.0006,
+ "loss": 4.375743103027344,
+ "step": 1600
+ },
+ {
+ "epoch": 22.36355691500983,
+ "grad_norm": 0.2051071673631668,
+ "learning_rate": 0.0006,
+ "loss": 4.340732574462891,
+ "step": 1610
+ },
+ {
+ "epoch": 22.50338649770592,
+ "grad_norm": 0.3336673974990845,
+ "learning_rate": 0.0006,
+ "loss": 4.346902847290039,
+ "step": 1620
+ },
+ {
+ "epoch": 22.64321608040201,
+ "grad_norm": 0.5095342397689819,
+ "learning_rate": 0.0006,
+ "loss": 4.604707336425781,
+ "step": 1630
+ },
+ {
+ "epoch": 22.7830456630981,
+ "grad_norm": 0.19417355954647064,
+ "learning_rate": 0.0006,
+ "loss": 4.451902389526367,
+ "step": 1640
+ },
+ {
+ "epoch": 22.922875245794188,
+ "grad_norm": 0.1654558628797531,
+ "learning_rate": 0.0006,
+ "loss": 4.431742858886719,
+ "step": 1650
+ },
+ {
+ "epoch": 23.0,
+ "eval_loss": 3.991173028945923,
+ "eval_runtime": 19.8631,
+ "eval_samples_per_second": 122.891,
+ "eval_steps_per_second": 15.405,
+ "step": 1656
+ },
+ {
+ "epoch": 23.055931833078436,
+ "grad_norm": 0.1633480191230774,
+ "learning_rate": 0.0006,
+ "loss": 4.371438980102539,
+ "step": 1660
+ },
+ {
+ "epoch": 23.195761415774523,
+ "grad_norm": 0.18645387887954712,
+ "learning_rate": 0.0006,
+ "loss": 4.342131805419922,
+ "step": 1670
+ },
+ {
+ "epoch": 23.335590998470614,
+ "grad_norm": 0.16122855246067047,
+ "learning_rate": 0.0006,
+ "loss": 4.370444488525391,
+ "step": 1680
+ },
+ {
+ "epoch": 23.475420581166702,
+ "grad_norm": 0.15771615505218506,
+ "learning_rate": 0.0006,
+ "loss": 4.319095230102539,
+ "step": 1690
+ },
+ {
+ "epoch": 23.615250163862793,
+ "grad_norm": 0.15940791368484497,
+ "learning_rate": 0.0006,
+ "loss": 4.327032852172851,
+ "step": 1700
+ },
+ {
+ "epoch": 23.75507974655888,
+ "grad_norm": 0.17007584869861603,
+ "learning_rate": 0.0006,
+ "loss": 4.324097442626953,
+ "step": 1710
+ },
+ {
+ "epoch": 23.894909329254972,
+ "grad_norm": 0.16725797951221466,
+ "learning_rate": 0.0006,
+ "loss": 4.328718566894532,
+ "step": 1720
+ },
+ {
+ "epoch": 24.0,
+ "eval_loss": 3.958721160888672,
+ "eval_runtime": 19.8826,
+ "eval_samples_per_second": 122.771,
+ "eval_steps_per_second": 15.39,
+ "step": 1728
+ },
+ {
+ "epoch": 24.027965916539216,
+ "grad_norm": 0.16644851863384247,
+ "learning_rate": 0.0006,
+ "loss": 4.296227264404297,
+ "step": 1730
+ },
+ {
+ "epoch": 24.167795499235307,
+ "grad_norm": 0.18710844218730927,
+ "learning_rate": 0.0006,
+ "loss": 4.289547348022461,
+ "step": 1740
+ },
+ {
+ "epoch": 24.307625081931395,
+ "grad_norm": 0.14764000475406647,
+ "learning_rate": 0.0006,
+ "loss": 4.3060558319091795,
+ "step": 1750
+ },
+ {
+ "epoch": 24.447454664627486,
+ "grad_norm": 0.16646794974803925,
+ "learning_rate": 0.0006,
+ "loss": 4.303120040893555,
+ "step": 1760
+ },
+ {
+ "epoch": 24.587284247323574,
+ "grad_norm": 0.1819610595703125,
+ "learning_rate": 0.0006,
+ "loss": 4.280985641479492,
+ "step": 1770
+ },
+ {
+ "epoch": 24.727113830019665,
+ "grad_norm": 0.1946498453617096,
+ "learning_rate": 0.0006,
+ "loss": 4.320237350463867,
+ "step": 1780
+ },
+ {
+ "epoch": 24.866943412715752,
+ "grad_norm": 0.17951808869838715,
+ "learning_rate": 0.0006,
+ "loss": 4.330652236938477,
+ "step": 1790
+ },
+ {
+ "epoch": 25.0,
+ "grad_norm": 0.22762995958328247,
+ "learning_rate": 0.0006,
+ "loss": 4.302712249755859,
+ "step": 1800
+ },
+ {
+ "epoch": 25.0,
+ "eval_loss": 3.941317319869995,
+ "eval_runtime": 19.865,
+ "eval_samples_per_second": 122.879,
+ "eval_steps_per_second": 15.404,
+ "step": 1800
+ },
+ {
+ "epoch": 25.139829582696088,
+ "grad_norm": 0.18263204395771027,
+ "learning_rate": 0.0006,
+ "loss": 4.266713333129883,
+ "step": 1810
+ },
+ {
+ "epoch": 25.27965916539218,
+ "grad_norm": 0.1817312240600586,
+ "learning_rate": 0.0006,
+ "loss": 4.245487213134766,
+ "step": 1820
+ },
+ {
+ "epoch": 25.419488748088266,
+ "grad_norm": 0.21272337436676025,
+ "learning_rate": 0.0006,
+ "loss": 4.265260696411133,
+ "step": 1830
+ },
+ {
+ "epoch": 25.559318330784357,
+ "grad_norm": 0.20256032049655914,
+ "learning_rate": 0.0006,
+ "loss": 4.249737167358399,
+ "step": 1840
+ },
+ {
+ "epoch": 25.699147913480445,
+ "grad_norm": 0.19611196219921112,
+ "learning_rate": 0.0006,
+ "loss": 4.284307479858398,
+ "step": 1850
+ },
+ {
+ "epoch": 25.838977496176536,
+ "grad_norm": 0.21398484706878662,
+ "learning_rate": 0.0006,
+ "loss": 4.286024475097657,
+ "step": 1860
+ },
+ {
+ "epoch": 25.978807078872624,
+ "grad_norm": 0.19678419828414917,
+ "learning_rate": 0.0006,
+ "loss": 4.285805511474609,
+ "step": 1870
+ },
+ {
+ "epoch": 26.0,
+ "eval_loss": 3.9300894737243652,
+ "eval_runtime": 19.8853,
+ "eval_samples_per_second": 122.754,
+ "eval_steps_per_second": 15.388,
+ "step": 1872
+ },
+ {
+ "epoch": 26.11186366615687,
+ "grad_norm": 0.21788373589515686,
+ "learning_rate": 0.0006,
+ "loss": 4.262743377685547,
+ "step": 1880
+ },
+ {
+ "epoch": 26.25169324885296,
+ "grad_norm": 0.20581643283367157,
+ "learning_rate": 0.0006,
+ "loss": 4.24034538269043,
+ "step": 1890
+ },
+ {
+ "epoch": 26.39152283154905,
+ "grad_norm": 0.19301028549671173,
+ "learning_rate": 0.0006,
+ "loss": 4.254768753051758,
+ "step": 1900
+ },
+ {
+ "epoch": 26.531352414245138,
+ "grad_norm": 0.18821723759174347,
+ "learning_rate": 0.0006,
+ "loss": 4.2541767120361325,
+ "step": 1910
+ },
+ {
+ "epoch": 26.67118199694123,
+ "grad_norm": 0.20018640160560608,
+ "learning_rate": 0.0006,
+ "loss": 4.259914398193359,
+ "step": 1920
+ },
+ {
+ "epoch": 26.811011579637317,
+ "grad_norm": 0.19723200798034668,
+ "learning_rate": 0.0006,
+ "loss": 4.272549057006836,
+ "step": 1930
+ },
+ {
+ "epoch": 26.950841162333408,
+ "grad_norm": 0.1991368979215622,
+ "learning_rate": 0.0006,
+ "loss": 4.259438705444336,
+ "step": 1940
+ },
+ {
+ "epoch": 27.0,
+ "eval_loss": 3.917532205581665,
+ "eval_runtime": 19.8852,
+ "eval_samples_per_second": 122.755,
+ "eval_steps_per_second": 15.388,
+ "step": 1944
+ },
+ {
+ "epoch": 27.083897749617652,
+ "grad_norm": 0.2797158658504486,
+ "learning_rate": 0.0006,
+ "loss": 4.2379203796386715,
+ "step": 1950
+ },
+ {
+ "epoch": 27.223727332313743,
+ "grad_norm": 0.1977781057357788,
+ "learning_rate": 0.0006,
+ "loss": 4.218112564086914,
+ "step": 1960
+ },
+ {
+ "epoch": 27.36355691500983,
+ "grad_norm": 0.21210049092769623,
+ "learning_rate": 0.0006,
+ "loss": 4.2191017150878904,
+ "step": 1970
+ },
+ {
+ "epoch": 27.50338649770592,
+ "grad_norm": 0.2273968905210495,
+ "learning_rate": 0.0006,
+ "loss": 4.236033248901367,
+ "step": 1980
+ },
+ {
+ "epoch": 27.64321608040201,
+ "grad_norm": 0.23919540643692017,
+ "learning_rate": 0.0006,
+ "loss": 4.223239517211914,
+ "step": 1990
+ },
+ {
+ "epoch": 27.7830456630981,
+ "grad_norm": 0.1874573826789856,
+ "learning_rate": 0.0006,
+ "loss": 4.244692611694336,
+ "step": 2000
+ },
+ {
+ "epoch": 27.922875245794188,
+ "grad_norm": 0.1952115297317505,
+ "learning_rate": 0.0006,
+ "loss": 4.242477798461914,
+ "step": 2010
+ },
+ {
+ "epoch": 28.0,
+ "eval_loss": 3.906435966491699,
+ "eval_runtime": 19.871,
+ "eval_samples_per_second": 122.842,
+ "eval_steps_per_second": 15.399,
+ "step": 2016
+ },
+ {
+ "epoch": 28.055931833078436,
+ "grad_norm": 0.2088126838207245,
+ "learning_rate": 0.0006,
+ "loss": 4.247038269042969,
+ "step": 2020
+ },
+ {
+ "epoch": 28.195761415774523,
+ "grad_norm": 0.2384193390607834,
+ "learning_rate": 0.0006,
+ "loss": 4.1952362060546875,
+ "step": 2030
+ },
+ {
+ "epoch": 28.335590998470614,
+ "grad_norm": 0.2215941995382309,
+ "learning_rate": 0.0006,
+ "loss": 4.213097763061524,
+ "step": 2040
+ },
+ {
+ "epoch": 28.475420581166702,
+ "grad_norm": 0.22252431511878967,
+ "learning_rate": 0.0006,
+ "loss": 4.208810806274414,
+ "step": 2050
+ },
+ {
+ "epoch": 28.615250163862793,
+ "grad_norm": 0.19061279296875,
+ "learning_rate": 0.0006,
+ "loss": 4.22558364868164,
+ "step": 2060
+ },
+ {
+ "epoch": 28.75507974655888,
+ "grad_norm": 0.22449851036071777,
+ "learning_rate": 0.0006,
+ "loss": 4.213565444946289,
+ "step": 2070
+ },
+ {
+ "epoch": 28.894909329254972,
+ "grad_norm": 0.2126193344593048,
+ "learning_rate": 0.0006,
+ "loss": 4.215901565551758,
+ "step": 2080
+ },
+ {
+ "epoch": 29.0,
+ "eval_loss": 3.9004921913146973,
+ "eval_runtime": 19.8632,
+ "eval_samples_per_second": 122.891,
+ "eval_steps_per_second": 15.405,
+ "step": 2088
+ },
+ {
+ "epoch": 29.027965916539216,
+ "grad_norm": 0.2392246276140213,
+ "learning_rate": 0.0006,
+ "loss": 4.214061737060547,
+ "step": 2090
+ },
+ {
+ "epoch": 29.167795499235307,
+ "grad_norm": 0.20037686824798584,
+ "learning_rate": 0.0006,
+ "loss": 4.196040344238281,
+ "step": 2100
+ },
+ {
+ "epoch": 29.307625081931395,
+ "grad_norm": 0.20283105969429016,
+ "learning_rate": 0.0006,
+ "loss": 4.194776916503907,
+ "step": 2110
+ },
+ {
+ "epoch": 29.447454664627486,
+ "grad_norm": 0.240804523229599,
+ "learning_rate": 0.0006,
+ "loss": 4.185153198242188,
+ "step": 2120
+ },
+ {
+ "epoch": 29.587284247323574,
+ "grad_norm": 0.24765098094940186,
+ "learning_rate": 0.0006,
+ "loss": 4.200851058959961,
+ "step": 2130
+ },
+ {
+ "epoch": 29.727113830019665,
+ "grad_norm": 0.19377271831035614,
+ "learning_rate": 0.0006,
+ "loss": 4.181426239013672,
+ "step": 2140
+ },
+ {
+ "epoch": 29.866943412715752,
+ "grad_norm": 0.19239689409732819,
+ "learning_rate": 0.0006,
+ "loss": 4.202949905395508,
+ "step": 2150
+ },
+ {
+ "epoch": 30.0,
+ "grad_norm": 0.22955702245235443,
+ "learning_rate": 0.0006,
+ "loss": 4.193132400512695,
+ "step": 2160
+ },
+ {
+ "epoch": 30.0,
+ "eval_loss": 3.889738082885742,
+ "eval_runtime": 19.8924,
+ "eval_samples_per_second": 122.71,
+ "eval_steps_per_second": 15.383,
+ "step": 2160
+ },
+ {
+ "epoch": 30.139829582696088,
+ "grad_norm": 0.2727241516113281,
+ "learning_rate": 0.0006,
+ "loss": 4.1729576110839846,
+ "step": 2170
+ },
+ {
+ "epoch": 30.27965916539218,
+ "grad_norm": 0.22165383398532867,
+ "learning_rate": 0.0006,
+ "loss": 4.178167724609375,
+ "step": 2180
+ },
+ {
+ "epoch": 30.419488748088266,
+ "grad_norm": 0.2023787498474121,
+ "learning_rate": 0.0006,
+ "loss": 4.186256790161133,
+ "step": 2190
+ },
+ {
+ "epoch": 30.559318330784357,
+ "grad_norm": 0.2281879335641861,
+ "learning_rate": 0.0006,
+ "loss": 4.201462173461914,
+ "step": 2200
+ },
+ {
+ "epoch": 30.699147913480445,
+ "grad_norm": 0.2198174148797989,
+ "learning_rate": 0.0006,
+ "loss": 4.169087600708008,
+ "step": 2210
+ },
+ {
+ "epoch": 30.838977496176536,
+ "grad_norm": 0.19493478536605835,
+ "learning_rate": 0.0006,
+ "loss": 4.1776588439941404,
+ "step": 2220
+ },
+ {
+ "epoch": 30.978807078872624,
+ "grad_norm": 0.20501400530338287,
+ "learning_rate": 0.0006,
+ "loss": 4.172566986083984,
+ "step": 2230
+ },
+ {
+ "epoch": 31.0,
+ "eval_loss": 3.8863086700439453,
+ "eval_runtime": 19.8898,
+ "eval_samples_per_second": 122.726,
+ "eval_steps_per_second": 15.385,
+ "step": 2232
+ },
+ {
+ "epoch": 31.11186366615687,
+ "grad_norm": 0.24052457511425018,
+ "learning_rate": 0.0006,
+ "loss": 4.175981521606445,
+ "step": 2240
+ },
+ {
+ "epoch": 31.25169324885296,
+ "grad_norm": 0.2178056836128235,
+ "learning_rate": 0.0006,
+ "loss": 4.136004638671875,
+ "step": 2250
+ },
+ {
+ "epoch": 31.39152283154905,
+ "grad_norm": 0.24538201093673706,
+ "learning_rate": 0.0006,
+ "loss": 4.169219207763672,
+ "step": 2260
+ },
+ {
+ "epoch": 31.531352414245138,
+ "grad_norm": 0.2414664775133133,
+ "learning_rate": 0.0006,
+ "loss": 4.148029708862305,
+ "step": 2270
+ },
+ {
+ "epoch": 31.67118199694123,
+ "grad_norm": 0.23237508535385132,
+ "learning_rate": 0.0006,
+ "loss": 4.168670272827148,
+ "step": 2280
+ },
+ {
+ "epoch": 31.811011579637317,
+ "grad_norm": 0.2166748344898224,
+ "learning_rate": 0.0006,
+ "loss": 4.16821060180664,
+ "step": 2290
+ },
+ {
+ "epoch": 31.950841162333408,
+ "grad_norm": 0.1907017081975937,
+ "learning_rate": 0.0006,
+ "loss": 4.162161254882813,
+ "step": 2300
+ },
+ {
+ "epoch": 32.0,
+ "eval_loss": 3.8760528564453125,
+ "eval_runtime": 19.876,
+ "eval_samples_per_second": 122.811,
+ "eval_steps_per_second": 15.395,
+ "step": 2304
+ },
+ {
+ "epoch": 32.083897749617655,
+ "grad_norm": 0.23043058812618256,
+ "learning_rate": 0.0006,
+ "loss": 4.1766307830810545,
+ "step": 2310
+ },
+ {
+ "epoch": 32.22372733231374,
+ "grad_norm": 0.22683262825012207,
+ "learning_rate": 0.0006,
+ "loss": 4.116132354736328,
+ "step": 2320
+ },
+ {
+ "epoch": 32.36355691500983,
+ "grad_norm": 0.2286939024925232,
+ "learning_rate": 0.0006,
+ "loss": 4.143845748901367,
+ "step": 2330
+ },
+ {
+ "epoch": 32.50338649770592,
+ "grad_norm": 0.23664161562919617,
+ "learning_rate": 0.0006,
+ "loss": 4.148048019409179,
+ "step": 2340
+ },
+ {
+ "epoch": 32.64321608040201,
+ "grad_norm": 0.21717309951782227,
+ "learning_rate": 0.0006,
+ "loss": 4.128271484375,
+ "step": 2350
+ },
+ {
+ "epoch": 32.7830456630981,
+ "grad_norm": 0.22340939939022064,
+ "learning_rate": 0.0006,
+ "loss": 4.159354019165039,
+ "step": 2360
+ },
+ {
+ "epoch": 32.92287524579419,
+ "grad_norm": 0.233748197555542,
+ "learning_rate": 0.0006,
+ "loss": 4.1642333984375,
+ "step": 2370
+ },
+ {
+ "epoch": 33.0,
+ "eval_loss": 3.8757057189941406,
+ "eval_runtime": 19.8566,
+ "eval_samples_per_second": 122.931,
+ "eval_steps_per_second": 15.41,
+ "step": 2376
+ },
+ {
+ "epoch": 33.05593183307843,
+ "grad_norm": 0.28126245737075806,
+ "learning_rate": 0.0006,
+ "loss": 4.1295326232910154,
+ "step": 2380
+ },
+ {
+ "epoch": 33.19576141577453,
+ "grad_norm": 0.2256774604320526,
+ "learning_rate": 0.0006,
+ "loss": 4.123966979980469,
+ "step": 2390
+ },
+ {
+ "epoch": 33.335590998470614,
+ "grad_norm": 0.21874992549419403,
+ "learning_rate": 0.0006,
+ "loss": 4.1306617736816404,
+ "step": 2400
+ },
+ {
+ "epoch": 33.4754205811667,
+ "grad_norm": 0.24607613682746887,
+ "learning_rate": 0.0006,
+ "loss": 4.122125625610352,
+ "step": 2410
+ },
+ {
+ "epoch": 33.61525016386279,
+ "grad_norm": 0.21854019165039062,
+ "learning_rate": 0.0006,
+ "loss": 4.1378215789794925,
+ "step": 2420
+ },
+ {
+ "epoch": 33.755079746558884,
+ "grad_norm": 0.20719152688980103,
+ "learning_rate": 0.0006,
+ "loss": 4.124840545654297,
+ "step": 2430
+ },
+ {
+ "epoch": 33.89490932925497,
+ "grad_norm": 0.24441562592983246,
+ "learning_rate": 0.0006,
+ "loss": 4.158369064331055,
+ "step": 2440
+ },
+ {
+ "epoch": 34.0,
+ "eval_loss": 3.8689537048339844,
+ "eval_runtime": 19.8813,
+ "eval_samples_per_second": 122.779,
+ "eval_steps_per_second": 15.391,
+ "step": 2448
+ },
+ {
+ "epoch": 34.027965916539216,
+ "grad_norm": 0.23353195190429688,
+ "learning_rate": 0.0006,
+ "loss": 4.129092025756836,
+ "step": 2450
+ },
+ {
+ "epoch": 34.167795499235304,
+ "grad_norm": 0.23389264941215515,
+ "learning_rate": 0.0006,
+ "loss": 4.105047225952148,
+ "step": 2460
+ },
+ {
+ "epoch": 34.3076250819314,
+ "grad_norm": 0.22500835359096527,
+ "learning_rate": 0.0006,
+ "loss": 4.106203842163086,
+ "step": 2470
+ },
+ {
+ "epoch": 34.447454664627486,
+ "grad_norm": 0.20221099257469177,
+ "learning_rate": 0.0006,
+ "loss": 4.116315841674805,
+ "step": 2480
+ },
+ {
+ "epoch": 34.58728424732357,
+ "grad_norm": 0.23113557696342468,
+ "learning_rate": 0.0006,
+ "loss": 4.103522491455078,
+ "step": 2490
+ },
+ {
+ "epoch": 34.72711383001966,
+ "grad_norm": 0.2406412810087204,
+ "learning_rate": 0.0006,
+ "loss": 4.107744979858398,
+ "step": 2500
+ },
+ {
+ "epoch": 34.866943412715756,
+ "grad_norm": 0.22139394283294678,
+ "learning_rate": 0.0006,
+ "loss": 4.123571395874023,
+ "step": 2510
+ },
+ {
+ "epoch": 35.0,
+ "grad_norm": 0.24141380190849304,
+ "learning_rate": 0.0006,
+ "loss": 4.135809326171875,
+ "step": 2520
+ },
+ {
+ "epoch": 35.0,
+ "eval_loss": 3.8626644611358643,
+ "eval_runtime": 19.8871,
+ "eval_samples_per_second": 122.743,
+ "eval_steps_per_second": 15.387,
+ "step": 2520
+ },
+ {
+ "epoch": 35.13982958269609,
+ "grad_norm": 0.21153542399406433,
+ "learning_rate": 0.0006,
+ "loss": 4.086721801757813,
+ "step": 2530
+ },
+ {
+ "epoch": 35.279659165392175,
+ "grad_norm": 0.21188288927078247,
+ "learning_rate": 0.0006,
+ "loss": 4.0864921569824215,
+ "step": 2540
+ },
+ {
+ "epoch": 35.41948874808827,
+ "grad_norm": 0.2161928415298462,
+ "learning_rate": 0.0006,
+ "loss": 4.117094421386719,
+ "step": 2550
+ },
+ {
+ "epoch": 35.55931833078436,
+ "grad_norm": 0.24187582731246948,
+ "learning_rate": 0.0006,
+ "loss": 4.105147171020508,
+ "step": 2560
+ },
+ {
+ "epoch": 35.699147913480445,
+ "grad_norm": 0.24268373847007751,
+ "learning_rate": 0.0006,
+ "loss": 4.123013305664062,
+ "step": 2570
+ },
+ {
+ "epoch": 35.83897749617653,
+ "grad_norm": 0.21142488718032837,
+ "learning_rate": 0.0006,
+ "loss": 4.1278221130371096,
+ "step": 2580
+ },
+ {
+ "epoch": 35.97880707887263,
+ "grad_norm": 0.20133273303508759,
+ "learning_rate": 0.0006,
+ "loss": 4.110447692871094,
+ "step": 2590
+ },
+ {
+ "epoch": 36.0,
+ "eval_loss": 3.858666181564331,
+ "eval_runtime": 19.8936,
+ "eval_samples_per_second": 122.703,
+ "eval_steps_per_second": 15.382,
+ "step": 2592
+ },
+ {
+ "epoch": 36.11186366615687,
+ "grad_norm": 0.23672761023044586,
+ "learning_rate": 0.0006,
+ "loss": 4.057193756103516,
+ "step": 2600
+ },
+ {
+ "epoch": 36.25169324885296,
+ "grad_norm": 0.22852548956871033,
+ "learning_rate": 0.0006,
+ "loss": 4.071565246582031,
+ "step": 2610
+ },
+ {
+ "epoch": 36.39152283154905,
+ "grad_norm": 0.23083142936229706,
+ "learning_rate": 0.0006,
+ "loss": 4.088025665283203,
+ "step": 2620
+ },
+ {
+ "epoch": 36.53135241424514,
+ "grad_norm": 0.21557074785232544,
+ "learning_rate": 0.0006,
+ "loss": 4.0878650665283205,
+ "step": 2630
+ },
+ {
+ "epoch": 36.67118199694123,
+ "grad_norm": 0.20414093136787415,
+ "learning_rate": 0.0006,
+ "loss": 4.088549423217773,
+ "step": 2640
+ },
+ {
+ "epoch": 36.81101157963732,
+ "grad_norm": 0.2646992802619934,
+ "learning_rate": 0.0006,
+ "loss": 4.100538635253907,
+ "step": 2650
+ },
+ {
+ "epoch": 36.950841162333404,
+ "grad_norm": 0.23596541583538055,
+ "learning_rate": 0.0006,
+ "loss": 4.131303024291992,
+ "step": 2660
+ },
+ {
+ "epoch": 37.0,
+ "eval_loss": 3.8548638820648193,
+ "eval_runtime": 19.8961,
+ "eval_samples_per_second": 122.688,
+ "eval_steps_per_second": 15.38,
+ "step": 2664
+ },
+ {
+ "epoch": 37.083897749617655,
+ "grad_norm": 0.24256756901741028,
+ "learning_rate": 0.0006,
+ "loss": 4.044707489013672,
+ "step": 2670
+ },
+ {
+ "epoch": 37.22372733231374,
+ "grad_norm": 0.21312902867794037,
+ "learning_rate": 0.0006,
+ "loss": 4.05272331237793,
+ "step": 2680
+ },
+ {
+ "epoch": 37.36355691500983,
+ "grad_norm": 0.23208224773406982,
+ "learning_rate": 0.0006,
+ "loss": 4.053055191040039,
+ "step": 2690
+ },
+ {
+ "epoch": 37.50338649770592,
+ "grad_norm": 0.21824559569358826,
+ "learning_rate": 0.0006,
+ "loss": 4.093597412109375,
+ "step": 2700
+ },
+ {
+ "epoch": 37.64321608040201,
+ "grad_norm": 0.23761679232120514,
+ "learning_rate": 0.0006,
+ "loss": 4.08574104309082,
+ "step": 2710
+ },
+ {
+ "epoch": 37.7830456630981,
+ "grad_norm": 0.23472781479358673,
+ "learning_rate": 0.0006,
+ "loss": 4.049263763427734,
+ "step": 2720
+ },
+ {
+ "epoch": 37.92287524579419,
+ "grad_norm": 0.201343834400177,
+ "learning_rate": 0.0006,
+ "loss": 4.113669586181641,
+ "step": 2730
+ },
+ {
+ "epoch": 38.0,
+ "eval_loss": 3.850585460662842,
+ "eval_runtime": 19.9061,
+ "eval_samples_per_second": 122.626,
+ "eval_steps_per_second": 15.372,
+ "step": 2736
+ },
+ {
+ "epoch": 38.05593183307843,
+ "grad_norm": 0.24569375813007355,
+ "learning_rate": 0.0006,
+ "loss": 4.076086807250976,
+ "step": 2740
+ },
+ {
+ "epoch": 38.19576141577453,
+ "grad_norm": 0.2517682909965515,
+ "learning_rate": 0.0006,
+ "loss": 4.042757797241211,
+ "step": 2750
+ },
+ {
+ "epoch": 38.335590998470614,
+ "grad_norm": 0.2101573646068573,
+ "learning_rate": 0.0006,
+ "loss": 4.062640762329101,
+ "step": 2760
+ },
+ {
+ "epoch": 38.4754205811667,
+ "grad_norm": 0.20859113335609436,
+ "learning_rate": 0.0006,
+ "loss": 4.065978240966797,
+ "step": 2770
+ },
+ {
+ "epoch": 38.61525016386279,
+ "grad_norm": 0.21619582176208496,
+ "learning_rate": 0.0006,
+ "loss": 4.072301864624023,
+ "step": 2780
+ },
+ {
+ "epoch": 38.755079746558884,
+ "grad_norm": 0.23485828936100006,
+ "learning_rate": 0.0006,
+ "loss": 4.0832267761230465,
+ "step": 2790
+ },
+ {
+ "epoch": 38.89490932925497,
+ "grad_norm": 0.2651204466819763,
+ "learning_rate": 0.0006,
+ "loss": 4.071917343139648,
+ "step": 2800
+ },
+ {
+ "epoch": 39.0,
+ "eval_loss": 3.9559922218322754,
+ "eval_runtime": 19.8975,
+ "eval_samples_per_second": 122.679,
+ "eval_steps_per_second": 15.379,
+ "step": 2808
+ },
+ {
+ "epoch": 39.027965916539216,
+ "grad_norm": 0.9851544499397278,
+ "learning_rate": 0.0006,
+ "loss": 4.151388168334961,
+ "step": 2810
+ },
+ {
+ "epoch": 39.167795499235304,
+ "grad_norm": 1.0761138200759888,
+ "learning_rate": 0.0006,
+ "loss": 4.414295196533203,
+ "step": 2820
+ },
+ {
+ "epoch": 39.3076250819314,
+ "grad_norm": 1.107331395149231,
+ "learning_rate": 0.0006,
+ "loss": 4.610064315795898,
+ "step": 2830
+ },
+ {
+ "epoch": 39.447454664627486,
+ "grad_norm": 0.25186386704444885,
+ "learning_rate": 0.0006,
+ "loss": 4.4520305633544925,
+ "step": 2840
+ },
+ {
+ "epoch": 39.58728424732357,
+ "grad_norm": 0.18421608209609985,
+ "learning_rate": 0.0006,
+ "loss": 4.3207855224609375,
+ "step": 2850
+ },
+ {
+ "epoch": 39.72711383001966,
+ "grad_norm": 0.15925951302051544,
+ "learning_rate": 0.0006,
+ "loss": 4.231083679199219,
+ "step": 2860
+ },
+ {
+ "epoch": 39.866943412715756,
+ "grad_norm": 0.15482063591480255,
+ "learning_rate": 0.0006,
+ "loss": 4.183073425292969,
+ "step": 2870
+ },
+ {
+ "epoch": 40.0,
+ "grad_norm": 0.19148693978786469,
+ "learning_rate": 0.0006,
+ "loss": 4.176522064208984,
+ "step": 2880
+ },
+ {
+ "epoch": 40.0,
+ "eval_loss": 3.881002902984619,
+ "eval_runtime": 19.8855,
+ "eval_samples_per_second": 122.753,
+ "eval_steps_per_second": 15.388,
+ "step": 2880
+ },
+ {
+ "epoch": 40.13982958269609,
+ "grad_norm": 0.15897083282470703,
+ "learning_rate": 0.0006,
+ "loss": 4.0932159423828125,
+ "step": 2890
+ },
+ {
+ "epoch": 40.279659165392175,
+ "grad_norm": 0.1569022685289383,
+ "learning_rate": 0.0006,
+ "loss": 4.10535888671875,
+ "step": 2900
+ },
+ {
+ "epoch": 40.41948874808827,
+ "grad_norm": 0.19591611623764038,
+ "learning_rate": 0.0006,
+ "loss": 4.087108612060547,
+ "step": 2910
+ },
+ {
+ "epoch": 40.55931833078436,
+ "grad_norm": 0.17719434201717377,
+ "learning_rate": 0.0006,
+ "loss": 4.089181900024414,
+ "step": 2920
+ },
+ {
+ "epoch": 40.699147913480445,
+ "grad_norm": 0.1780233383178711,
+ "learning_rate": 0.0006,
+ "loss": 4.104327392578125,
+ "step": 2930
+ },
+ {
+ "epoch": 40.83897749617653,
+ "grad_norm": 0.17799495160579681,
+ "learning_rate": 0.0006,
+ "loss": 4.109510803222657,
+ "step": 2940
+ },
+ {
+ "epoch": 40.97880707887263,
+ "grad_norm": 0.18227098882198334,
+ "learning_rate": 0.0006,
+ "loss": 4.099219131469726,
+ "step": 2950
+ },
+ {
+ "epoch": 41.0,
+ "eval_loss": 3.8475587368011475,
+ "eval_runtime": 19.8729,
+ "eval_samples_per_second": 122.83,
+ "eval_steps_per_second": 15.398,
+ "step": 2952
+ },
+ {
+ "epoch": 41.11186366615687,
+ "grad_norm": 0.1759035289287567,
+ "learning_rate": 0.0006,
+ "loss": 4.047430038452148,
+ "step": 2960
+ },
+ {
+ "epoch": 41.25169324885296,
+ "grad_norm": 0.1729285717010498,
+ "learning_rate": 0.0006,
+ "loss": 4.046004867553711,
+ "step": 2970
+ },
+ {
+ "epoch": 41.39152283154905,
+ "grad_norm": 0.18820366263389587,
+ "learning_rate": 0.0006,
+ "loss": 4.045801162719727,
+ "step": 2980
+ },
+ {
+ "epoch": 41.53135241424514,
+ "grad_norm": 0.1819244921207428,
+ "learning_rate": 0.0006,
+ "loss": 4.034719467163086,
+ "step": 2990
+ },
+ {
+ "epoch": 41.67118199694123,
+ "grad_norm": 0.17209376394748688,
+ "learning_rate": 0.0006,
+ "loss": 4.065438842773437,
+ "step": 3000
+ },
+ {
+ "epoch": 41.81101157963732,
+ "grad_norm": 0.1764725297689438,
+ "learning_rate": 0.0006,
+ "loss": 4.060000610351563,
+ "step": 3010
+ },
+ {
+ "epoch": 41.950841162333404,
+ "grad_norm": 0.18680505454540253,
+ "learning_rate": 0.0006,
+ "loss": 4.0759326934814455,
+ "step": 3020
+ },
+ {
+ "epoch": 42.0,
+ "eval_loss": 3.846500873565674,
+ "eval_runtime": 19.9106,
+ "eval_samples_per_second": 122.598,
+ "eval_steps_per_second": 15.369,
+ "step": 3024
+ },
+ {
+ "epoch": 42.083897749617655,
+ "grad_norm": 0.19561053812503815,
+ "learning_rate": 0.0006,
+ "loss": 4.0523326873779295,
+ "step": 3030
+ },
+ {
+ "epoch": 42.22372733231374,
+ "grad_norm": 0.21906520426273346,
+ "learning_rate": 0.0006,
+ "loss": 4.03740234375,
+ "step": 3040
+ },
+ {
+ "epoch": 42.36355691500983,
+ "grad_norm": 0.20610764622688293,
+ "learning_rate": 0.0006,
+ "loss": 4.035350799560547,
+ "step": 3050
+ },
+ {
+ "epoch": 42.50338649770592,
+ "grad_norm": 0.1980750560760498,
+ "learning_rate": 0.0006,
+ "loss": 4.028232192993164,
+ "step": 3060
+ },
+ {
+ "epoch": 42.64321608040201,
+ "grad_norm": 0.17973868548870087,
+ "learning_rate": 0.0006,
+ "loss": 4.066466522216797,
+ "step": 3070
+ },
+ {
+ "epoch": 42.7830456630981,
+ "grad_norm": 0.18086765706539154,
+ "learning_rate": 0.0006,
+ "loss": 4.0102886199951175,
+ "step": 3080
+ },
+ {
+ "epoch": 42.92287524579419,
+ "grad_norm": 0.19552463293075562,
+ "learning_rate": 0.0006,
+ "loss": 4.057781219482422,
+ "step": 3090
+ },
+ {
+ "epoch": 43.0,
+ "eval_loss": 3.8388819694519043,
+ "eval_runtime": 19.8813,
+ "eval_samples_per_second": 122.779,
+ "eval_steps_per_second": 15.391,
+ "step": 3096
+ },
+ {
+ "epoch": 43.05593183307843,
+ "grad_norm": 0.18557795882225037,
+ "learning_rate": 0.0006,
+ "loss": 4.027313613891602,
+ "step": 3100
+ },
+ {
+ "epoch": 43.19576141577453,
+ "grad_norm": 0.19318890571594238,
+ "learning_rate": 0.0006,
+ "loss": 4.027843475341797,
+ "step": 3110
+ },
+ {
+ "epoch": 43.335590998470614,
+ "grad_norm": 0.18623752892017365,
+ "learning_rate": 0.0006,
+ "loss": 4.004922485351562,
+ "step": 3120
+ },
+ {
+ "epoch": 43.4754205811667,
+ "grad_norm": 0.1875496208667755,
+ "learning_rate": 0.0006,
+ "loss": 4.030118560791015,
+ "step": 3130
+ },
+ {
+ "epoch": 43.61525016386279,
+ "grad_norm": 0.2201668918132782,
+ "learning_rate": 0.0006,
+ "loss": 4.045892715454102,
+ "step": 3140
+ },
+ {
+ "epoch": 43.755079746558884,
+ "grad_norm": 0.20447023212909698,
+ "learning_rate": 0.0006,
+ "loss": 4.016044616699219,
+ "step": 3150
+ },
+ {
+ "epoch": 43.89490932925497,
+ "grad_norm": 0.19765615463256836,
+ "learning_rate": 0.0006,
+ "loss": 4.03606071472168,
+ "step": 3160
+ },
+ {
+ "epoch": 44.0,
+ "eval_loss": 3.837571382522583,
+ "eval_runtime": 19.884,
+ "eval_samples_per_second": 122.762,
+ "eval_steps_per_second": 15.389,
+ "step": 3168
+ },
+ {
+ "epoch": 44.027965916539216,
+ "grad_norm": 0.20829571783542633,
+ "learning_rate": 0.0006,
+ "loss": 4.021365737915039,
+ "step": 3170
+ },
+ {
+ "epoch": 44.167795499235304,
+ "grad_norm": 0.22128954529762268,
+ "learning_rate": 0.0006,
+ "loss": 3.9627857208251953,
+ "step": 3180
+ },
+ {
+ "epoch": 44.3076250819314,
+ "grad_norm": 0.19976745545864105,
+ "learning_rate": 0.0006,
+ "loss": 3.9967437744140626,
+ "step": 3190
+ },
+ {
+ "epoch": 44.447454664627486,
+ "grad_norm": 0.2479678839445114,
+ "learning_rate": 0.0006,
+ "loss": 4.017975616455078,
+ "step": 3200
+ },
+ {
+ "epoch": 44.58728424732357,
+ "grad_norm": 0.196462020277977,
+ "learning_rate": 0.0006,
+ "loss": 4.013183212280273,
+ "step": 3210
+ },
+ {
+ "epoch": 44.72711383001966,
+ "grad_norm": 0.2005736529827118,
+ "learning_rate": 0.0006,
+ "loss": 4.022643661499023,
+ "step": 3220
+ },
+ {
+ "epoch": 44.866943412715756,
+ "grad_norm": 0.2102411985397339,
+ "learning_rate": 0.0006,
+ "loss": 4.029917144775391,
+ "step": 3230
+ },
+ {
+ "epoch": 45.0,
+ "grad_norm": 0.22666622698307037,
+ "learning_rate": 0.0006,
+ "loss": 4.015410232543945,
+ "step": 3240
+ },
+ {
+ "epoch": 45.0,
+ "eval_loss": 3.836167097091675,
+ "eval_runtime": 19.871,
+ "eval_samples_per_second": 122.842,
+ "eval_steps_per_second": 15.399,
+ "step": 3240
+ },
+ {
+ "epoch": 45.13982958269609,
+ "grad_norm": 0.1865929216146469,
+ "learning_rate": 0.0006,
+ "loss": 3.962997817993164,
+ "step": 3250
+ },
+ {
+ "epoch": 45.279659165392175,
+ "grad_norm": 0.19806267321109772,
+ "learning_rate": 0.0006,
+ "loss": 3.985525131225586,
+ "step": 3260
+ },
+ {
+ "epoch": 45.41948874808827,
+ "grad_norm": 0.20194685459136963,
+ "learning_rate": 0.0006,
+ "loss": 3.9974178314208983,
+ "step": 3270
+ },
+ {
+ "epoch": 45.55931833078436,
+ "grad_norm": 0.216617614030838,
+ "learning_rate": 0.0006,
+ "loss": 3.971715545654297,
+ "step": 3280
+ },
+ {
+ "epoch": 45.699147913480445,
+ "grad_norm": 0.21372166275978088,
+ "learning_rate": 0.0006,
+ "loss": 4.020486450195312,
+ "step": 3290
+ },
+ {
+ "epoch": 45.83897749617653,
+ "grad_norm": 0.19971641898155212,
+ "learning_rate": 0.0006,
+ "loss": 4.013861846923828,
+ "step": 3300
+ },
+ {
+ "epoch": 45.97880707887263,
+ "grad_norm": 0.20691554248332977,
+ "learning_rate": 0.0006,
+ "loss": 4.003434753417968,
+ "step": 3310
+ },
+ {
+ "epoch": 46.0,
+ "eval_loss": 3.8369977474212646,
+ "eval_runtime": 19.9028,
+ "eval_samples_per_second": 122.646,
+ "eval_steps_per_second": 15.375,
+ "step": 3312
+ },
+ {
+ "epoch": 46.11186366615687,
+ "grad_norm": 0.21140055358409882,
+ "learning_rate": 0.0006,
+ "loss": 3.9648902893066404,
+ "step": 3320
+ },
+ {
+ "epoch": 46.25169324885296,
+ "grad_norm": 0.22331228852272034,
+ "learning_rate": 0.0006,
+ "loss": 3.9585491180419923,
+ "step": 3330
+ },
+ {
+ "epoch": 46.39152283154905,
+ "grad_norm": 0.2113780379295349,
+ "learning_rate": 0.0006,
+ "loss": 3.981510543823242,
+ "step": 3340
+ },
+ {
+ "epoch": 46.53135241424514,
+ "grad_norm": 0.1941022127866745,
+ "learning_rate": 0.0006,
+ "loss": 3.9624217987060546,
+ "step": 3350
+ },
+ {
+ "epoch": 46.67118199694123,
+ "grad_norm": 0.2140984833240509,
+ "learning_rate": 0.0006,
+ "loss": 3.9639034271240234,
+ "step": 3360
+ },
+ {
+ "epoch": 46.81101157963732,
+ "grad_norm": 0.22650973498821259,
+ "learning_rate": 0.0006,
+ "loss": 3.987116241455078,
+ "step": 3370
+ },
+ {
+ "epoch": 46.950841162333404,
+ "grad_norm": 0.22001947462558746,
+ "learning_rate": 0.0006,
+ "loss": 4.016731262207031,
+ "step": 3380
+ },
+ {
+ "epoch": 47.0,
+ "eval_loss": 3.837308168411255,
+ "eval_runtime": 19.9074,
+ "eval_samples_per_second": 122.618,
+ "eval_steps_per_second": 15.371,
+ "step": 3384
+ },
+ {
+ "epoch": 47.083897749617655,
+ "grad_norm": 0.24726083874702454,
+ "learning_rate": 0.0006,
+ "loss": 3.985137176513672,
+ "step": 3390
+ },
+ {
+ "epoch": 47.22372733231374,
+ "grad_norm": 0.2092239111661911,
+ "learning_rate": 0.0006,
+ "loss": 3.988035202026367,
+ "step": 3400
+ },
+ {
+ "epoch": 47.36355691500983,
+ "grad_norm": 0.22165395319461823,
+ "learning_rate": 0.0006,
+ "loss": 3.9932319641113283,
+ "step": 3410
+ },
+ {
+ "epoch": 47.50338649770592,
+ "grad_norm": 0.23019757866859436,
+ "learning_rate": 0.0006,
+ "loss": 3.9902557373046874,
+ "step": 3420
+ },
+ {
+ "epoch": 47.64321608040201,
+ "grad_norm": 0.20552881062030792,
+ "learning_rate": 0.0006,
+ "loss": 3.975299072265625,
+ "step": 3430
+ },
+ {
+ "epoch": 47.7830456630981,
+ "grad_norm": 0.2133014053106308,
+ "learning_rate": 0.0006,
+ "loss": 3.966014862060547,
+ "step": 3440
+ },
+ {
+ "epoch": 47.92287524579419,
+ "grad_norm": 0.22695177793502808,
+ "learning_rate": 0.0006,
+ "loss": 4.0052978515625,
+ "step": 3450
+ },
+ {
+ "epoch": 48.0,
+ "eval_loss": 3.834726095199585,
+ "eval_runtime": 19.9111,
+ "eval_samples_per_second": 122.595,
+ "eval_steps_per_second": 15.368,
+ "step": 3456
+ },
+ {
+ "epoch": 48.05593183307843,
+ "grad_norm": 0.2078237235546112,
+ "learning_rate": 0.0006,
+ "loss": 3.9635799407958983,
+ "step": 3460
+ },
+ {
+ "epoch": 48.19576141577453,
+ "grad_norm": 0.254269540309906,
+ "learning_rate": 0.0006,
+ "loss": 3.9350101470947267,
+ "step": 3470
+ },
+ {
+ "epoch": 48.335590998470614,
+ "grad_norm": 0.2060205638408661,
+ "learning_rate": 0.0006,
+ "loss": 3.999435806274414,
+ "step": 3480
+ },
+ {
+ "epoch": 48.4754205811667,
+ "grad_norm": 0.20986555516719818,
+ "learning_rate": 0.0006,
+ "loss": 3.97415771484375,
+ "step": 3490
+ },
+ {
+ "epoch": 48.61525016386279,
+ "grad_norm": 0.19596225023269653,
+ "learning_rate": 0.0006,
+ "loss": 3.968153381347656,
+ "step": 3500
+ },
+ {
+ "epoch": 48.755079746558884,
+ "grad_norm": 0.21542511880397797,
+ "learning_rate": 0.0006,
+ "loss": 3.979802703857422,
+ "step": 3510
+ },
+ {
+ "epoch": 48.89490932925497,
+ "grad_norm": 0.2204463630914688,
+ "learning_rate": 0.0006,
+ "loss": 3.986585998535156,
+ "step": 3520
+ },
+ {
+ "epoch": 49.0,
+ "eval_loss": 3.8454229831695557,
+ "eval_runtime": 19.9199,
+ "eval_samples_per_second": 122.541,
+ "eval_steps_per_second": 15.362,
+ "step": 3528
+ },
+ {
+ "epoch": 49.027965916539216,
+ "grad_norm": 0.24546828866004944,
+ "learning_rate": 0.0006,
+ "loss": 4.015388488769531,
+ "step": 3530
+ },
+ {
+ "epoch": 49.167795499235304,
+ "grad_norm": 0.24072720110416412,
+ "learning_rate": 0.0006,
+ "loss": 3.956148147583008,
+ "step": 3540
+ },
+ {
+ "epoch": 49.3076250819314,
+ "grad_norm": 0.25352367758750916,
+ "learning_rate": 0.0006,
+ "loss": 3.9292694091796876,
+ "step": 3550
+ },
+ {
+ "epoch": 49.447454664627486,
+ "grad_norm": 0.23329177498817444,
+ "learning_rate": 0.0006,
+ "loss": 3.9898456573486327,
+ "step": 3560
+ },
+ {
+ "epoch": 49.58728424732357,
+ "grad_norm": 0.20554102957248688,
+ "learning_rate": 0.0006,
+ "loss": 3.9599224090576173,
+ "step": 3570
+ },
+ {
+ "epoch": 49.72711383001966,
+ "grad_norm": 0.20182636380195618,
+ "learning_rate": 0.0006,
+ "loss": 3.9677288055419924,
+ "step": 3580
+ },
+ {
+ "epoch": 49.866943412715756,
+ "grad_norm": 0.21841393411159515,
+ "learning_rate": 0.0006,
+ "loss": 3.989236068725586,
+ "step": 3590
+ },
+ {
+ "epoch": 50.0,
+ "grad_norm": 0.23754680156707764,
+ "learning_rate": 0.0006,
+ "loss": 3.9796295166015625,
+ "step": 3600
+ },
+ {
+ "epoch": 50.0,
+ "eval_loss": 3.8374722003936768,
+ "eval_runtime": 19.8914,
+ "eval_samples_per_second": 122.717,
+ "eval_steps_per_second": 15.384,
+ "step": 3600
+ },
+ {
+ "epoch": 50.13982958269609,
+ "grad_norm": 0.21172602474689484,
+ "learning_rate": 0.0006,
+ "loss": 3.9041725158691407,
+ "step": 3610
+ },
+ {
+ "epoch": 50.279659165392175,
+ "grad_norm": 0.22906376421451569,
+ "learning_rate": 0.0006,
+ "loss": 3.967184829711914,
+ "step": 3620
+ },
+ {
+ "epoch": 50.41948874808827,
+ "grad_norm": 0.20152941346168518,
+ "learning_rate": 0.0006,
+ "loss": 3.9225933074951174,
+ "step": 3630
+ },
+ {
+ "epoch": 50.55931833078436,
+ "grad_norm": 0.24042142927646637,
+ "learning_rate": 0.0006,
+ "loss": 3.9427978515625,
+ "step": 3640
+ },
+ {
+ "epoch": 50.699147913480445,
+ "grad_norm": 0.20214207470417023,
+ "learning_rate": 0.0006,
+ "loss": 3.9555389404296877,
+ "step": 3650
+ },
+ {
+ "epoch": 50.83897749617653,
+ "grad_norm": 0.20780637860298157,
+ "learning_rate": 0.0006,
+ "loss": 3.965659332275391,
+ "step": 3660
+ },
+ {
+ "epoch": 50.97880707887263,
+ "grad_norm": 0.20646893978118896,
+ "learning_rate": 0.0006,
+ "loss": 3.98671875,
+ "step": 3670
+ },
+ {
+ "epoch": 51.0,
+ "eval_loss": 3.8382489681243896,
+ "eval_runtime": 19.9074,
+ "eval_samples_per_second": 122.617,
+ "eval_steps_per_second": 15.371,
+ "step": 3672
+ },
+ {
+ "epoch": 51.11186366615687,
+ "grad_norm": 0.25553613901138306,
+ "learning_rate": 0.0006,
+ "loss": 3.948938751220703,
+ "step": 3680
+ },
+ {
+ "epoch": 51.25169324885296,
+ "grad_norm": 0.2221221923828125,
+ "learning_rate": 0.0006,
+ "loss": 3.9367401123046877,
+ "step": 3690
+ },
+ {
+ "epoch": 51.39152283154905,
+ "grad_norm": 0.21815785765647888,
+ "learning_rate": 0.0006,
+ "loss": 3.931085205078125,
+ "step": 3700
+ },
+ {
+ "epoch": 51.53135241424514,
+ "grad_norm": 0.21005959808826447,
+ "learning_rate": 0.0006,
+ "loss": 3.9245285034179687,
+ "step": 3710
+ },
+ {
+ "epoch": 51.67118199694123,
+ "grad_norm": 0.21702027320861816,
+ "learning_rate": 0.0006,
+ "loss": 3.9639286041259765,
+ "step": 3720
+ },
+ {
+ "epoch": 51.81101157963732,
+ "grad_norm": 0.19672633707523346,
+ "learning_rate": 0.0006,
+ "loss": 3.9614242553710937,
+ "step": 3730
+ },
+ {
+ "epoch": 51.950841162333404,
+ "grad_norm": 0.21263723075389862,
+ "learning_rate": 0.0006,
+ "loss": 3.9813636779785155,
+ "step": 3740
+ },
+ {
+ "epoch": 52.0,
+ "eval_loss": 3.8346211910247803,
+ "eval_runtime": 19.9083,
+ "eval_samples_per_second": 122.612,
+ "eval_steps_per_second": 15.37,
+ "step": 3744
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 7200,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 100,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.59506627396685e+18,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/run.lock b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/run.lock
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/train.log b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..65e4134386ecc340f390760882112ec479f92d43
--- /dev/null
+++ b/runs/seeds/t1a_S2_C2_r2l50_i5exp_s44/train.log
@@ -0,0 +1,12789 @@
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 100
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 8
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 64
+[transformers] Total optimization steps = 7,200
+[transformers] Number of trainable parameters = 145,836,544
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: setting up run wayep2d8
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260917_105923-wayep2d8
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t1a_S2_C2_r2l50_i5exp_s44
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/wayep2d8
+
0%| | 0/7200 [00:00system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/checkpoint-6984/config.json b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/checkpoint-6984/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3474619f10931027097cc05d837296fa4f33a6c
--- /dev/null
+++ b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/checkpoint-6984/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 512,
+ "initializer_range": 0.02,
+ "intermediate_size": 1536,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 4,
+ "num_hidden_layers": 20,
+ "num_key_value_heads": 4,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151676
+}
diff --git a/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/checkpoint-6984/generation_config.json b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/checkpoint-6984/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/checkpoint-6984/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/checkpoint-6984/tokenizer_config.json b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/checkpoint-6984/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d43c97a447f4b25a8c36be5873f7b622935fa90
--- /dev/null
+++ b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/checkpoint-6984/tokenizer_config.json
@@ -0,0 +1,24 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>",
+ "<|next_1_pred|>",
+ "<|next_2_pred|>",
+ "<|next_3_pred|>",
+ "<|next_4_pred|>",
+ "<|next_5_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/checkpoint-6984/trainer_state.json b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/checkpoint-6984/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c075d82e4d775de4f44d91bf0881ddad9429ccf
--- /dev/null
+++ b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/checkpoint-6984/trainer_state.json
@@ -0,0 +1,5703 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 97.0,
+ "eval_steps": 72,
+ "global_step": 6984,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013982958269608914,
+ "grad_norm": 2.309645891189575,
+ "learning_rate": 0.0,
+ "loss": 12.038338661193848,
+ "step": 1
+ },
+ {
+ "epoch": 0.13982958269608914,
+ "grad_norm": 1.6804709434509277,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.730747646755642,
+ "step": 10
+ },
+ {
+ "epoch": 0.2796591653921783,
+ "grad_norm": 1.5878028869628906,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 11.073248291015625,
+ "step": 20
+ },
+ {
+ "epoch": 0.4194887480882674,
+ "grad_norm": 1.5632054805755615,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 10.350322723388672,
+ "step": 30
+ },
+ {
+ "epoch": 0.5593183307843566,
+ "grad_norm": 1.4121705293655396,
+ "learning_rate": 0.000234,
+ "loss": 9.367364501953125,
+ "step": 40
+ },
+ {
+ "epoch": 0.6991479134804457,
+ "grad_norm": 0.7714095711708069,
+ "learning_rate": 0.000294,
+ "loss": 8.448271179199219,
+ "step": 50
+ },
+ {
+ "epoch": 0.8389774961765348,
+ "grad_norm": 0.624386191368103,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 7.994615173339843,
+ "step": 60
+ },
+ {
+ "epoch": 0.978807078872624,
+ "grad_norm": 0.22891412675380707,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 7.909974670410156,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.5048112869262695,
+ "eval_runtime": 44.1159,
+ "eval_samples_per_second": 55.332,
+ "eval_steps_per_second": 6.936,
+ "step": 72
+ },
+ {
+ "epoch": 1.1118636661568713,
+ "grad_norm": 0.2561050355434418,
+ "learning_rate": 0.000474,
+ "loss": 7.828945159912109,
+ "step": 80
+ },
+ {
+ "epoch": 1.2516932488529604,
+ "grad_norm": 0.2886507511138916,
+ "learning_rate": 0.000534,
+ "loss": 7.7764640808105465,
+ "step": 90
+ },
+ {
+ "epoch": 1.3915228315490495,
+ "grad_norm": 0.17726194858551025,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 7.74832763671875,
+ "step": 100
+ },
+ {
+ "epoch": 1.5313524142451387,
+ "grad_norm": 0.2412502020597458,
+ "learning_rate": 0.0006,
+ "loss": 7.733228302001953,
+ "step": 110
+ },
+ {
+ "epoch": 1.671181996941228,
+ "grad_norm": 0.2463846057653427,
+ "learning_rate": 0.0006,
+ "loss": 7.703779602050782,
+ "step": 120
+ },
+ {
+ "epoch": 1.811011579637317,
+ "grad_norm": 0.2046874761581421,
+ "learning_rate": 0.0006,
+ "loss": 7.652253723144531,
+ "step": 130
+ },
+ {
+ "epoch": 1.9508411623334063,
+ "grad_norm": 0.6344203352928162,
+ "learning_rate": 0.0006,
+ "loss": 7.667548370361328,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 7.180665493011475,
+ "eval_runtime": 19.8801,
+ "eval_samples_per_second": 122.786,
+ "eval_steps_per_second": 15.392,
+ "step": 144
+ },
+ {
+ "epoch": 2.0838977496176536,
+ "grad_norm": 0.22687020897865295,
+ "learning_rate": 0.0006,
+ "loss": 7.627081298828125,
+ "step": 150
+ },
+ {
+ "epoch": 2.2237273323137425,
+ "grad_norm": 0.15179501473903656,
+ "learning_rate": 0.0006,
+ "loss": 7.577226257324218,
+ "step": 160
+ },
+ {
+ "epoch": 2.363556915009832,
+ "grad_norm": 0.31009194254875183,
+ "learning_rate": 0.0006,
+ "loss": 7.537248229980468,
+ "step": 170
+ },
+ {
+ "epoch": 2.503386497705921,
+ "grad_norm": 0.5839681625366211,
+ "learning_rate": 0.0006,
+ "loss": 7.5106758117675785,
+ "step": 180
+ },
+ {
+ "epoch": 2.64321608040201,
+ "grad_norm": 0.2834503948688507,
+ "learning_rate": 0.0006,
+ "loss": 7.471561431884766,
+ "step": 190
+ },
+ {
+ "epoch": 2.783045663098099,
+ "grad_norm": 0.6746534109115601,
+ "learning_rate": 0.0006,
+ "loss": 7.403218841552734,
+ "step": 200
+ },
+ {
+ "epoch": 2.9228752457941884,
+ "grad_norm": 0.27713218331336975,
+ "learning_rate": 0.0006,
+ "loss": 7.31884536743164,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 6.5756754875183105,
+ "eval_runtime": 19.8855,
+ "eval_samples_per_second": 122.753,
+ "eval_steps_per_second": 15.388,
+ "step": 216
+ },
+ {
+ "epoch": 3.0559318330784357,
+ "grad_norm": 0.8567643761634827,
+ "learning_rate": 0.0006,
+ "loss": 7.218125915527343,
+ "step": 220
+ },
+ {
+ "epoch": 3.1957614157745247,
+ "grad_norm": 0.5024699568748474,
+ "learning_rate": 0.0006,
+ "loss": 7.144245147705078,
+ "step": 230
+ },
+ {
+ "epoch": 3.335590998470614,
+ "grad_norm": 0.5251261591911316,
+ "learning_rate": 0.0006,
+ "loss": 7.059344482421875,
+ "step": 240
+ },
+ {
+ "epoch": 3.475420581166703,
+ "grad_norm": 0.5262770056724548,
+ "learning_rate": 0.0006,
+ "loss": 7.000555419921875,
+ "step": 250
+ },
+ {
+ "epoch": 3.6152501638627923,
+ "grad_norm": 0.30603206157684326,
+ "learning_rate": 0.0006,
+ "loss": 6.9616035461425785,
+ "step": 260
+ },
+ {
+ "epoch": 3.755079746558881,
+ "grad_norm": 0.28493693470954895,
+ "learning_rate": 0.0006,
+ "loss": 6.887174987792969,
+ "step": 270
+ },
+ {
+ "epoch": 3.8949093292549706,
+ "grad_norm": 0.6188507080078125,
+ "learning_rate": 0.0006,
+ "loss": 6.820857238769531,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 6.02525520324707,
+ "eval_runtime": 19.901,
+ "eval_samples_per_second": 122.657,
+ "eval_steps_per_second": 15.376,
+ "step": 288
+ },
+ {
+ "epoch": 4.027965916539218,
+ "grad_norm": 0.4271056056022644,
+ "learning_rate": 0.0006,
+ "loss": 6.812914276123047,
+ "step": 290
+ },
+ {
+ "epoch": 4.167795499235307,
+ "grad_norm": 0.19648367166519165,
+ "learning_rate": 0.0006,
+ "loss": 6.736518096923828,
+ "step": 300
+ },
+ {
+ "epoch": 4.307625081931396,
+ "grad_norm": 0.7077914476394653,
+ "learning_rate": 0.0006,
+ "loss": 6.69824447631836,
+ "step": 310
+ },
+ {
+ "epoch": 4.447454664627485,
+ "grad_norm": 0.39409831166267395,
+ "learning_rate": 0.0006,
+ "loss": 6.657051086425781,
+ "step": 320
+ },
+ {
+ "epoch": 4.587284247323574,
+ "grad_norm": 0.29634472727775574,
+ "learning_rate": 0.0006,
+ "loss": 6.596706390380859,
+ "step": 330
+ },
+ {
+ "epoch": 4.727113830019664,
+ "grad_norm": 0.4454764425754547,
+ "learning_rate": 0.0006,
+ "loss": 6.541728973388672,
+ "step": 340
+ },
+ {
+ "epoch": 4.866943412715752,
+ "grad_norm": 0.7296797037124634,
+ "learning_rate": 0.0006,
+ "loss": 6.544454956054688,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.24267183244228363,
+ "learning_rate": 0.0006,
+ "loss": 6.513423156738281,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 5.6134724617004395,
+ "eval_runtime": 19.8796,
+ "eval_samples_per_second": 122.789,
+ "eval_steps_per_second": 15.393,
+ "step": 360
+ },
+ {
+ "epoch": 5.139829582696089,
+ "grad_norm": 0.3262389004230499,
+ "learning_rate": 0.0006,
+ "loss": 6.446672821044922,
+ "step": 370
+ },
+ {
+ "epoch": 5.279659165392179,
+ "grad_norm": 0.4822391867637634,
+ "learning_rate": 0.0006,
+ "loss": 6.3959705352783205,
+ "step": 380
+ },
+ {
+ "epoch": 5.419488748088267,
+ "grad_norm": 0.4351154863834381,
+ "learning_rate": 0.0006,
+ "loss": 6.368149948120117,
+ "step": 390
+ },
+ {
+ "epoch": 5.5593183307843566,
+ "grad_norm": 0.22013914585113525,
+ "learning_rate": 0.0006,
+ "loss": 6.32379379272461,
+ "step": 400
+ },
+ {
+ "epoch": 5.699147913480446,
+ "grad_norm": 0.2950364649295807,
+ "learning_rate": 0.0006,
+ "loss": 6.288776779174805,
+ "step": 410
+ },
+ {
+ "epoch": 5.838977496176534,
+ "grad_norm": 0.4387047588825226,
+ "learning_rate": 0.0006,
+ "loss": 6.251935958862305,
+ "step": 420
+ },
+ {
+ "epoch": 5.978807078872624,
+ "grad_norm": 0.5122392773628235,
+ "learning_rate": 0.0006,
+ "loss": 6.241645812988281,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 5.321452617645264,
+ "eval_runtime": 19.888,
+ "eval_samples_per_second": 122.737,
+ "eval_steps_per_second": 15.386,
+ "step": 432
+ },
+ {
+ "epoch": 6.1118636661568715,
+ "grad_norm": 0.35598278045654297,
+ "learning_rate": 0.0006,
+ "loss": 6.211026382446289,
+ "step": 440
+ },
+ {
+ "epoch": 6.251693248852961,
+ "grad_norm": 0.1975601613521576,
+ "learning_rate": 0.0006,
+ "loss": 6.165004730224609,
+ "step": 450
+ },
+ {
+ "epoch": 6.391522831549049,
+ "grad_norm": 0.4413703382015228,
+ "learning_rate": 0.0006,
+ "loss": 6.114179992675782,
+ "step": 460
+ },
+ {
+ "epoch": 6.531352414245139,
+ "grad_norm": 0.6126284003257751,
+ "learning_rate": 0.0006,
+ "loss": 6.109112930297852,
+ "step": 470
+ },
+ {
+ "epoch": 6.671181996941228,
+ "grad_norm": 0.2846424877643585,
+ "learning_rate": 0.0006,
+ "loss": 6.125335311889648,
+ "step": 480
+ },
+ {
+ "epoch": 6.811011579637317,
+ "grad_norm": 0.19271819293498993,
+ "learning_rate": 0.0006,
+ "loss": 6.065495300292969,
+ "step": 490
+ },
+ {
+ "epoch": 6.950841162333406,
+ "grad_norm": 0.2632502019405365,
+ "learning_rate": 0.0006,
+ "loss": 5.98363151550293,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 5.0162434577941895,
+ "eval_runtime": 19.8854,
+ "eval_samples_per_second": 122.754,
+ "eval_steps_per_second": 15.388,
+ "step": 504
+ },
+ {
+ "epoch": 7.083897749617654,
+ "grad_norm": 0.4452163279056549,
+ "learning_rate": 0.0006,
+ "loss": 5.9663440704345705,
+ "step": 510
+ },
+ {
+ "epoch": 7.223727332313743,
+ "grad_norm": 0.32781338691711426,
+ "learning_rate": 0.0006,
+ "loss": 5.955696105957031,
+ "step": 520
+ },
+ {
+ "epoch": 7.3635569150098314,
+ "grad_norm": 0.2896769940853119,
+ "learning_rate": 0.0006,
+ "loss": 5.904755783081055,
+ "step": 530
+ },
+ {
+ "epoch": 7.503386497705921,
+ "grad_norm": 0.24288485944271088,
+ "learning_rate": 0.0006,
+ "loss": 5.8729698181152346,
+ "step": 540
+ },
+ {
+ "epoch": 7.64321608040201,
+ "grad_norm": 0.2532391846179962,
+ "learning_rate": 0.0006,
+ "loss": 5.854362487792969,
+ "step": 550
+ },
+ {
+ "epoch": 7.7830456630980995,
+ "grad_norm": 0.2693583369255066,
+ "learning_rate": 0.0006,
+ "loss": 5.79533805847168,
+ "step": 560
+ },
+ {
+ "epoch": 7.922875245794188,
+ "grad_norm": 0.38787806034088135,
+ "learning_rate": 0.0006,
+ "loss": 5.799647521972656,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.7656731605529785,
+ "eval_runtime": 19.8753,
+ "eval_samples_per_second": 122.816,
+ "eval_steps_per_second": 15.396,
+ "step": 576
+ },
+ {
+ "epoch": 8.055931833078436,
+ "grad_norm": 0.8482298851013184,
+ "learning_rate": 0.0006,
+ "loss": 5.771756362915039,
+ "step": 580
+ },
+ {
+ "epoch": 8.195761415774525,
+ "grad_norm": 1.7572287321090698,
+ "learning_rate": 0.0006,
+ "loss": 6.278157043457031,
+ "step": 590
+ },
+ {
+ "epoch": 8.335590998470614,
+ "grad_norm": 0.6067526936531067,
+ "learning_rate": 0.0006,
+ "loss": 6.15958251953125,
+ "step": 600
+ },
+ {
+ "epoch": 8.475420581166704,
+ "grad_norm": 0.2980787456035614,
+ "learning_rate": 0.0006,
+ "loss": 5.9723663330078125,
+ "step": 610
+ },
+ {
+ "epoch": 8.615250163862791,
+ "grad_norm": 0.18016934394836426,
+ "learning_rate": 0.0006,
+ "loss": 5.847333526611328,
+ "step": 620
+ },
+ {
+ "epoch": 8.75507974655888,
+ "grad_norm": 0.1433102786540985,
+ "learning_rate": 0.0006,
+ "loss": 5.790487670898438,
+ "step": 630
+ },
+ {
+ "epoch": 8.89490932925497,
+ "grad_norm": 0.1377483308315277,
+ "learning_rate": 0.0006,
+ "loss": 5.747037506103515,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.676959037780762,
+ "eval_runtime": 19.887,
+ "eval_samples_per_second": 122.744,
+ "eval_steps_per_second": 15.387,
+ "step": 648
+ },
+ {
+ "epoch": 9.027965916539218,
+ "grad_norm": 0.17598064243793488,
+ "learning_rate": 0.0006,
+ "loss": 5.69544563293457,
+ "step": 650
+ },
+ {
+ "epoch": 9.167795499235307,
+ "grad_norm": 0.23173747956752777,
+ "learning_rate": 0.0006,
+ "loss": 5.695872116088867,
+ "step": 660
+ },
+ {
+ "epoch": 9.307625081931397,
+ "grad_norm": 0.17895843088626862,
+ "learning_rate": 0.0006,
+ "loss": 5.652565002441406,
+ "step": 670
+ },
+ {
+ "epoch": 9.447454664627486,
+ "grad_norm": 0.24098169803619385,
+ "learning_rate": 0.0006,
+ "loss": 5.629655456542968,
+ "step": 680
+ },
+ {
+ "epoch": 9.587284247323574,
+ "grad_norm": 0.24608640372753143,
+ "learning_rate": 0.0006,
+ "loss": 5.649494934082031,
+ "step": 690
+ },
+ {
+ "epoch": 9.727113830019663,
+ "grad_norm": 0.20562946796417236,
+ "learning_rate": 0.0006,
+ "loss": 5.624085998535156,
+ "step": 700
+ },
+ {
+ "epoch": 9.866943412715752,
+ "grad_norm": 0.21710269153118134,
+ "learning_rate": 0.0006,
+ "loss": 5.596005630493164,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.3345872759819031,
+ "learning_rate": 0.0006,
+ "loss": 5.596768951416015,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.544719219207764,
+ "eval_runtime": 19.8888,
+ "eval_samples_per_second": 122.732,
+ "eval_steps_per_second": 15.386,
+ "step": 720
+ },
+ {
+ "epoch": 10.13982958269609,
+ "grad_norm": 0.2221258282661438,
+ "learning_rate": 0.0006,
+ "loss": 5.595984268188476,
+ "step": 730
+ },
+ {
+ "epoch": 10.279659165392179,
+ "grad_norm": 0.16621924936771393,
+ "learning_rate": 0.0006,
+ "loss": 5.541681289672852,
+ "step": 740
+ },
+ {
+ "epoch": 10.419488748088268,
+ "grad_norm": 0.357794851064682,
+ "learning_rate": 0.0006,
+ "loss": 5.574164962768554,
+ "step": 750
+ },
+ {
+ "epoch": 10.559318330784357,
+ "grad_norm": 0.195899099111557,
+ "learning_rate": 0.0006,
+ "loss": 5.5584228515625,
+ "step": 760
+ },
+ {
+ "epoch": 10.699147913480445,
+ "grad_norm": 0.18474528193473816,
+ "learning_rate": 0.0006,
+ "loss": 5.520959091186524,
+ "step": 770
+ },
+ {
+ "epoch": 10.838977496176534,
+ "grad_norm": 0.2380070835351944,
+ "learning_rate": 0.0006,
+ "loss": 5.519348526000977,
+ "step": 780
+ },
+ {
+ "epoch": 10.978807078872624,
+ "grad_norm": 0.18198099732398987,
+ "learning_rate": 0.0006,
+ "loss": 5.517754364013672,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.450497627258301,
+ "eval_runtime": 19.8758,
+ "eval_samples_per_second": 122.813,
+ "eval_steps_per_second": 15.396,
+ "step": 792
+ },
+ {
+ "epoch": 11.111863666156871,
+ "grad_norm": 0.3660085201263428,
+ "learning_rate": 0.0006,
+ "loss": 5.495576095581055,
+ "step": 800
+ },
+ {
+ "epoch": 11.25169324885296,
+ "grad_norm": 0.22333239018917084,
+ "learning_rate": 0.0006,
+ "loss": 5.4862628936767575,
+ "step": 810
+ },
+ {
+ "epoch": 11.39152283154905,
+ "grad_norm": 0.21112127602100372,
+ "learning_rate": 0.0006,
+ "loss": 5.466662979125976,
+ "step": 820
+ },
+ {
+ "epoch": 11.531352414245138,
+ "grad_norm": 1.529754400253296,
+ "learning_rate": 0.0006,
+ "loss": 5.490034484863282,
+ "step": 830
+ },
+ {
+ "epoch": 11.671181996941227,
+ "grad_norm": 0.7365639805793762,
+ "learning_rate": 0.0006,
+ "loss": 5.709087753295899,
+ "step": 840
+ },
+ {
+ "epoch": 11.811011579637317,
+ "grad_norm": 0.21305428445339203,
+ "learning_rate": 0.0006,
+ "loss": 5.6541259765625,
+ "step": 850
+ },
+ {
+ "epoch": 11.950841162333406,
+ "grad_norm": 0.1859208047389984,
+ "learning_rate": 0.0006,
+ "loss": 5.542103576660156,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.440022945404053,
+ "eval_runtime": 19.8745,
+ "eval_samples_per_second": 122.82,
+ "eval_steps_per_second": 15.397,
+ "step": 864
+ },
+ {
+ "epoch": 12.083897749617654,
+ "grad_norm": 0.13940244913101196,
+ "learning_rate": 0.0006,
+ "loss": 5.476589584350586,
+ "step": 870
+ },
+ {
+ "epoch": 12.223727332313743,
+ "grad_norm": 0.17301498353481293,
+ "learning_rate": 0.0006,
+ "loss": 5.463667678833008,
+ "step": 880
+ },
+ {
+ "epoch": 12.363556915009832,
+ "grad_norm": 0.1589861959218979,
+ "learning_rate": 0.0006,
+ "loss": 5.455442428588867,
+ "step": 890
+ },
+ {
+ "epoch": 12.503386497705922,
+ "grad_norm": 0.1611655354499817,
+ "learning_rate": 0.0006,
+ "loss": 5.4270366668701175,
+ "step": 900
+ },
+ {
+ "epoch": 12.64321608040201,
+ "grad_norm": 0.1630093902349472,
+ "learning_rate": 0.0006,
+ "loss": 5.414126205444336,
+ "step": 910
+ },
+ {
+ "epoch": 12.783045663098099,
+ "grad_norm": 0.22469639778137207,
+ "learning_rate": 0.0006,
+ "loss": 5.4185630798339846,
+ "step": 920
+ },
+ {
+ "epoch": 12.922875245794188,
+ "grad_norm": 0.17767876386642456,
+ "learning_rate": 0.0006,
+ "loss": 5.392729949951172,
+ "step": 930
+ },
+ {
+ "epoch": 13.0,
+ "eval_loss": 4.33057165145874,
+ "eval_runtime": 19.8754,
+ "eval_samples_per_second": 122.815,
+ "eval_steps_per_second": 15.396,
+ "step": 936
+ },
+ {
+ "epoch": 13.055931833078436,
+ "grad_norm": 0.1855991929769516,
+ "learning_rate": 0.0006,
+ "loss": 5.374072647094726,
+ "step": 940
+ },
+ {
+ "epoch": 13.195761415774525,
+ "grad_norm": 0.20089074969291687,
+ "learning_rate": 0.0006,
+ "loss": 5.37099609375,
+ "step": 950
+ },
+ {
+ "epoch": 13.335590998470614,
+ "grad_norm": 0.17839208245277405,
+ "learning_rate": 0.0006,
+ "loss": 5.3802490234375,
+ "step": 960
+ },
+ {
+ "epoch": 13.475420581166704,
+ "grad_norm": 0.1547849178314209,
+ "learning_rate": 0.0006,
+ "loss": 5.381869125366211,
+ "step": 970
+ },
+ {
+ "epoch": 13.615250163862791,
+ "grad_norm": 0.2046428918838501,
+ "learning_rate": 0.0006,
+ "loss": 5.352056884765625,
+ "step": 980
+ },
+ {
+ "epoch": 13.75507974655888,
+ "grad_norm": 0.22845542430877686,
+ "learning_rate": 0.0006,
+ "loss": 5.3497802734375,
+ "step": 990
+ },
+ {
+ "epoch": 13.89490932925497,
+ "grad_norm": 0.21850530803203583,
+ "learning_rate": 0.0006,
+ "loss": 5.333636474609375,
+ "step": 1000
+ },
+ {
+ "epoch": 14.0,
+ "eval_loss": 4.272489547729492,
+ "eval_runtime": 19.8686,
+ "eval_samples_per_second": 122.857,
+ "eval_steps_per_second": 15.401,
+ "step": 1008
+ },
+ {
+ "epoch": 14.027965916539218,
+ "grad_norm": 0.2228608876466751,
+ "learning_rate": 0.0006,
+ "loss": 5.3223724365234375,
+ "step": 1010
+ },
+ {
+ "epoch": 14.167795499235307,
+ "grad_norm": 0.19374042749404907,
+ "learning_rate": 0.0006,
+ "loss": 5.3154346466064455,
+ "step": 1020
+ },
+ {
+ "epoch": 14.307625081931397,
+ "grad_norm": 0.21237777173519135,
+ "learning_rate": 0.0006,
+ "loss": 5.299874496459961,
+ "step": 1030
+ },
+ {
+ "epoch": 14.447454664627486,
+ "grad_norm": 0.20794178545475006,
+ "learning_rate": 0.0006,
+ "loss": 5.304273223876953,
+ "step": 1040
+ },
+ {
+ "epoch": 14.587284247323574,
+ "grad_norm": 0.2232571840286255,
+ "learning_rate": 0.0006,
+ "loss": 5.309600448608398,
+ "step": 1050
+ },
+ {
+ "epoch": 14.727113830019663,
+ "grad_norm": 0.1920774281024933,
+ "learning_rate": 0.0006,
+ "loss": 5.325111770629883,
+ "step": 1060
+ },
+ {
+ "epoch": 14.866943412715752,
+ "grad_norm": 0.19044768810272217,
+ "learning_rate": 0.0006,
+ "loss": 5.28817138671875,
+ "step": 1070
+ },
+ {
+ "epoch": 15.0,
+ "grad_norm": 0.21018308401107788,
+ "learning_rate": 0.0006,
+ "loss": 5.286215972900391,
+ "step": 1080
+ },
+ {
+ "epoch": 15.0,
+ "eval_loss": 4.227161884307861,
+ "eval_runtime": 19.8673,
+ "eval_samples_per_second": 122.865,
+ "eval_steps_per_second": 15.402,
+ "step": 1080
+ },
+ {
+ "epoch": 15.13982958269609,
+ "grad_norm": 0.2824840247631073,
+ "learning_rate": 0.0006,
+ "loss": 5.27600326538086,
+ "step": 1090
+ },
+ {
+ "epoch": 15.279659165392179,
+ "grad_norm": 0.23559047281742096,
+ "learning_rate": 0.0006,
+ "loss": 5.26240348815918,
+ "step": 1100
+ },
+ {
+ "epoch": 15.419488748088268,
+ "grad_norm": 0.23002703487873077,
+ "learning_rate": 0.0006,
+ "loss": 5.271746826171875,
+ "step": 1110
+ },
+ {
+ "epoch": 15.559318330784357,
+ "grad_norm": 0.21145953238010406,
+ "learning_rate": 0.0006,
+ "loss": 5.246661758422851,
+ "step": 1120
+ },
+ {
+ "epoch": 15.699147913480445,
+ "grad_norm": 0.20860028266906738,
+ "learning_rate": 0.0006,
+ "loss": 5.243120574951172,
+ "step": 1130
+ },
+ {
+ "epoch": 15.838977496176534,
+ "grad_norm": 0.20988169312477112,
+ "learning_rate": 0.0006,
+ "loss": 5.28558578491211,
+ "step": 1140
+ },
+ {
+ "epoch": 15.978807078872624,
+ "grad_norm": 0.2011093646287918,
+ "learning_rate": 0.0006,
+ "loss": 5.240071868896484,
+ "step": 1150
+ },
+ {
+ "epoch": 16.0,
+ "eval_loss": 4.187058448791504,
+ "eval_runtime": 19.8796,
+ "eval_samples_per_second": 122.789,
+ "eval_steps_per_second": 15.393,
+ "step": 1152
+ },
+ {
+ "epoch": 16.11186366615687,
+ "grad_norm": 0.30222463607788086,
+ "learning_rate": 0.0006,
+ "loss": 5.245944213867188,
+ "step": 1160
+ },
+ {
+ "epoch": 16.25169324885296,
+ "grad_norm": 0.23616470396518707,
+ "learning_rate": 0.0006,
+ "loss": 5.226858520507813,
+ "step": 1170
+ },
+ {
+ "epoch": 16.39152283154905,
+ "grad_norm": 0.23074433207511902,
+ "learning_rate": 0.0006,
+ "loss": 5.21971549987793,
+ "step": 1180
+ },
+ {
+ "epoch": 16.531352414245138,
+ "grad_norm": 0.18964725732803345,
+ "learning_rate": 0.0006,
+ "loss": 5.226218414306641,
+ "step": 1190
+ },
+ {
+ "epoch": 16.67118199694123,
+ "grad_norm": 0.18440239131450653,
+ "learning_rate": 0.0006,
+ "loss": 5.187334060668945,
+ "step": 1200
+ },
+ {
+ "epoch": 16.811011579637317,
+ "grad_norm": 0.23255857825279236,
+ "learning_rate": 0.0006,
+ "loss": 5.214393997192383,
+ "step": 1210
+ },
+ {
+ "epoch": 16.950841162333408,
+ "grad_norm": 0.34394240379333496,
+ "learning_rate": 0.0006,
+ "loss": 5.210863876342773,
+ "step": 1220
+ },
+ {
+ "epoch": 17.0,
+ "eval_loss": 4.161089897155762,
+ "eval_runtime": 19.8821,
+ "eval_samples_per_second": 122.774,
+ "eval_steps_per_second": 15.391,
+ "step": 1224
+ },
+ {
+ "epoch": 17.083897749617652,
+ "grad_norm": 0.23063066601753235,
+ "learning_rate": 0.0006,
+ "loss": 5.227059173583984,
+ "step": 1230
+ },
+ {
+ "epoch": 17.223727332313743,
+ "grad_norm": 0.2626548409461975,
+ "learning_rate": 0.0006,
+ "loss": 5.188357543945313,
+ "step": 1240
+ },
+ {
+ "epoch": 17.36355691500983,
+ "grad_norm": 0.1966681331396103,
+ "learning_rate": 0.0006,
+ "loss": 5.174765777587891,
+ "step": 1250
+ },
+ {
+ "epoch": 17.50338649770592,
+ "grad_norm": 0.23317351937294006,
+ "learning_rate": 0.0006,
+ "loss": 5.202006912231445,
+ "step": 1260
+ },
+ {
+ "epoch": 17.64321608040201,
+ "grad_norm": 0.19967889785766602,
+ "learning_rate": 0.0006,
+ "loss": 5.187762069702148,
+ "step": 1270
+ },
+ {
+ "epoch": 17.7830456630981,
+ "grad_norm": 0.19705133140087128,
+ "learning_rate": 0.0006,
+ "loss": 5.186738204956055,
+ "step": 1280
+ },
+ {
+ "epoch": 17.922875245794188,
+ "grad_norm": 0.2006278932094574,
+ "learning_rate": 0.0006,
+ "loss": 5.169035720825195,
+ "step": 1290
+ },
+ {
+ "epoch": 18.0,
+ "eval_loss": 4.123867988586426,
+ "eval_runtime": 19.9074,
+ "eval_samples_per_second": 122.618,
+ "eval_steps_per_second": 15.371,
+ "step": 1296
+ },
+ {
+ "epoch": 18.055931833078436,
+ "grad_norm": 0.25349748134613037,
+ "learning_rate": 0.0006,
+ "loss": 5.182059478759766,
+ "step": 1300
+ },
+ {
+ "epoch": 18.195761415774523,
+ "grad_norm": 0.21297487616539001,
+ "learning_rate": 0.0006,
+ "loss": 5.137909317016602,
+ "step": 1310
+ },
+ {
+ "epoch": 18.335590998470614,
+ "grad_norm": 0.21930384635925293,
+ "learning_rate": 0.0006,
+ "loss": 5.140713882446289,
+ "step": 1320
+ },
+ {
+ "epoch": 18.475420581166702,
+ "grad_norm": 0.24430759251117706,
+ "learning_rate": 0.0006,
+ "loss": 5.172813796997071,
+ "step": 1330
+ },
+ {
+ "epoch": 18.615250163862793,
+ "grad_norm": 0.25275173783302307,
+ "learning_rate": 0.0006,
+ "loss": 5.157328414916992,
+ "step": 1340
+ },
+ {
+ "epoch": 18.75507974655888,
+ "grad_norm": 0.20037779211997986,
+ "learning_rate": 0.0006,
+ "loss": 5.166316223144531,
+ "step": 1350
+ },
+ {
+ "epoch": 18.894909329254972,
+ "grad_norm": 0.18744786083698273,
+ "learning_rate": 0.0006,
+ "loss": 5.139706802368164,
+ "step": 1360
+ },
+ {
+ "epoch": 19.0,
+ "eval_loss": 4.091379165649414,
+ "eval_runtime": 19.8946,
+ "eval_samples_per_second": 122.696,
+ "eval_steps_per_second": 15.381,
+ "step": 1368
+ },
+ {
+ "epoch": 19.027965916539216,
+ "grad_norm": 0.20728544890880585,
+ "learning_rate": 0.0006,
+ "loss": 5.173317718505859,
+ "step": 1370
+ },
+ {
+ "epoch": 19.167795499235307,
+ "grad_norm": 0.21604777872562408,
+ "learning_rate": 0.0006,
+ "loss": 5.1302234649658205,
+ "step": 1380
+ },
+ {
+ "epoch": 19.307625081931395,
+ "grad_norm": 0.2569308280944824,
+ "learning_rate": 0.0006,
+ "loss": 5.104495620727539,
+ "step": 1390
+ },
+ {
+ "epoch": 19.447454664627486,
+ "grad_norm": 0.21548086404800415,
+ "learning_rate": 0.0006,
+ "loss": 5.141900634765625,
+ "step": 1400
+ },
+ {
+ "epoch": 19.587284247323574,
+ "grad_norm": 0.21986989676952362,
+ "learning_rate": 0.0006,
+ "loss": 5.132741928100586,
+ "step": 1410
+ },
+ {
+ "epoch": 19.727113830019665,
+ "grad_norm": 0.19563151895999908,
+ "learning_rate": 0.0006,
+ "loss": 5.119315719604492,
+ "step": 1420
+ },
+ {
+ "epoch": 19.866943412715752,
+ "grad_norm": 0.1977599561214447,
+ "learning_rate": 0.0006,
+ "loss": 5.104138946533203,
+ "step": 1430
+ },
+ {
+ "epoch": 20.0,
+ "grad_norm": 0.22033612430095673,
+ "learning_rate": 0.0006,
+ "loss": 5.110370635986328,
+ "step": 1440
+ },
+ {
+ "epoch": 20.0,
+ "eval_loss": 4.064973831176758,
+ "eval_runtime": 19.9186,
+ "eval_samples_per_second": 122.549,
+ "eval_steps_per_second": 15.363,
+ "step": 1440
+ },
+ {
+ "epoch": 20.139829582696088,
+ "grad_norm": 0.23588906228542328,
+ "learning_rate": 0.0006,
+ "loss": 5.099114608764649,
+ "step": 1450
+ },
+ {
+ "epoch": 20.27965916539218,
+ "grad_norm": 0.23648838698863983,
+ "learning_rate": 0.0006,
+ "loss": 5.101178741455078,
+ "step": 1460
+ },
+ {
+ "epoch": 20.419488748088266,
+ "grad_norm": 0.26270610094070435,
+ "learning_rate": 0.0006,
+ "loss": 5.11432113647461,
+ "step": 1470
+ },
+ {
+ "epoch": 20.559318330784357,
+ "grad_norm": 0.2400597631931305,
+ "learning_rate": 0.0006,
+ "loss": 5.0783130645751955,
+ "step": 1480
+ },
+ {
+ "epoch": 20.699147913480445,
+ "grad_norm": 0.18867403268814087,
+ "learning_rate": 0.0006,
+ "loss": 5.1002544403076175,
+ "step": 1490
+ },
+ {
+ "epoch": 20.838977496176536,
+ "grad_norm": 0.20418550074100494,
+ "learning_rate": 0.0006,
+ "loss": 5.0886985778808596,
+ "step": 1500
+ },
+ {
+ "epoch": 20.978807078872624,
+ "grad_norm": 0.22204099595546722,
+ "learning_rate": 0.0006,
+ "loss": 5.089921569824218,
+ "step": 1510
+ },
+ {
+ "epoch": 21.0,
+ "eval_loss": 4.049679756164551,
+ "eval_runtime": 19.8858,
+ "eval_samples_per_second": 122.751,
+ "eval_steps_per_second": 15.388,
+ "step": 1512
+ },
+ {
+ "epoch": 21.11186366615687,
+ "grad_norm": 0.28355762362480164,
+ "learning_rate": 0.0006,
+ "loss": 5.067924499511719,
+ "step": 1520
+ },
+ {
+ "epoch": 21.25169324885296,
+ "grad_norm": 0.19949384033679962,
+ "learning_rate": 0.0006,
+ "loss": 5.074605178833008,
+ "step": 1530
+ },
+ {
+ "epoch": 21.39152283154905,
+ "grad_norm": 0.2125411033630371,
+ "learning_rate": 0.0006,
+ "loss": 5.072315216064453,
+ "step": 1540
+ },
+ {
+ "epoch": 21.531352414245138,
+ "grad_norm": 0.20479200780391693,
+ "learning_rate": 0.0006,
+ "loss": 5.056680679321289,
+ "step": 1550
+ },
+ {
+ "epoch": 21.67118199694123,
+ "grad_norm": 0.18927094340324402,
+ "learning_rate": 0.0006,
+ "loss": 5.061420059204101,
+ "step": 1560
+ },
+ {
+ "epoch": 21.811011579637317,
+ "grad_norm": 0.2311350405216217,
+ "learning_rate": 0.0006,
+ "loss": 5.069339370727539,
+ "step": 1570
+ },
+ {
+ "epoch": 21.950841162333408,
+ "grad_norm": 0.23102980852127075,
+ "learning_rate": 0.0006,
+ "loss": 5.0675910949707035,
+ "step": 1580
+ },
+ {
+ "epoch": 22.0,
+ "eval_loss": 4.031795024871826,
+ "eval_runtime": 19.9039,
+ "eval_samples_per_second": 122.639,
+ "eval_steps_per_second": 15.374,
+ "step": 1584
+ },
+ {
+ "epoch": 22.083897749617652,
+ "grad_norm": 0.22683364152908325,
+ "learning_rate": 0.0006,
+ "loss": 5.072455978393554,
+ "step": 1590
+ },
+ {
+ "epoch": 22.223727332313743,
+ "grad_norm": 0.22752700746059418,
+ "learning_rate": 0.0006,
+ "loss": 5.039709091186523,
+ "step": 1600
+ },
+ {
+ "epoch": 22.36355691500983,
+ "grad_norm": 0.17384633421897888,
+ "learning_rate": 0.0006,
+ "loss": 5.054748916625977,
+ "step": 1610
+ },
+ {
+ "epoch": 22.50338649770592,
+ "grad_norm": 0.23104862868785858,
+ "learning_rate": 0.0006,
+ "loss": 5.053108978271484,
+ "step": 1620
+ },
+ {
+ "epoch": 22.64321608040201,
+ "grad_norm": 0.23286665976047516,
+ "learning_rate": 0.0006,
+ "loss": 5.0459449768066404,
+ "step": 1630
+ },
+ {
+ "epoch": 22.7830456630981,
+ "grad_norm": 0.20633983612060547,
+ "learning_rate": 0.0006,
+ "loss": 5.039584732055664,
+ "step": 1640
+ },
+ {
+ "epoch": 22.922875245794188,
+ "grad_norm": 0.20277750492095947,
+ "learning_rate": 0.0006,
+ "loss": 5.049232482910156,
+ "step": 1650
+ },
+ {
+ "epoch": 23.0,
+ "eval_loss": 4.009687423706055,
+ "eval_runtime": 19.9015,
+ "eval_samples_per_second": 122.654,
+ "eval_steps_per_second": 15.376,
+ "step": 1656
+ },
+ {
+ "epoch": 23.055931833078436,
+ "grad_norm": 0.2552662193775177,
+ "learning_rate": 0.0006,
+ "loss": 5.031326675415039,
+ "step": 1660
+ },
+ {
+ "epoch": 23.195761415774523,
+ "grad_norm": 0.21026426553726196,
+ "learning_rate": 0.0006,
+ "loss": 5.015238189697266,
+ "step": 1670
+ },
+ {
+ "epoch": 23.335590998470614,
+ "grad_norm": 0.2333090454339981,
+ "learning_rate": 0.0006,
+ "loss": 5.024247360229492,
+ "step": 1680
+ },
+ {
+ "epoch": 23.475420581166702,
+ "grad_norm": 0.22201821208000183,
+ "learning_rate": 0.0006,
+ "loss": 5.037444305419922,
+ "step": 1690
+ },
+ {
+ "epoch": 23.615250163862793,
+ "grad_norm": 0.20002739131450653,
+ "learning_rate": 0.0006,
+ "loss": 5.030801773071289,
+ "step": 1700
+ },
+ {
+ "epoch": 23.75507974655888,
+ "grad_norm": 0.23732616007328033,
+ "learning_rate": 0.0006,
+ "loss": 5.030628204345703,
+ "step": 1710
+ },
+ {
+ "epoch": 23.894909329254972,
+ "grad_norm": 0.20032425224781036,
+ "learning_rate": 0.0006,
+ "loss": 5.034336090087891,
+ "step": 1720
+ },
+ {
+ "epoch": 24.0,
+ "eval_loss": 3.9944827556610107,
+ "eval_runtime": 19.9213,
+ "eval_samples_per_second": 122.532,
+ "eval_steps_per_second": 15.36,
+ "step": 1728
+ },
+ {
+ "epoch": 24.027965916539216,
+ "grad_norm": 0.20102764666080475,
+ "learning_rate": 0.0006,
+ "loss": 5.02490234375,
+ "step": 1730
+ },
+ {
+ "epoch": 24.167795499235307,
+ "grad_norm": 0.19266290962696075,
+ "learning_rate": 0.0006,
+ "loss": 4.983257293701172,
+ "step": 1740
+ },
+ {
+ "epoch": 24.307625081931395,
+ "grad_norm": 0.20911532640457153,
+ "learning_rate": 0.0006,
+ "loss": 5.0158226013183596,
+ "step": 1750
+ },
+ {
+ "epoch": 24.447454664627486,
+ "grad_norm": 0.2266414314508438,
+ "learning_rate": 0.0006,
+ "loss": 4.997885894775391,
+ "step": 1760
+ },
+ {
+ "epoch": 24.587284247323574,
+ "grad_norm": 0.21538399159908295,
+ "learning_rate": 0.0006,
+ "loss": 5.017646026611328,
+ "step": 1770
+ },
+ {
+ "epoch": 24.727113830019665,
+ "grad_norm": 0.2014819085597992,
+ "learning_rate": 0.0006,
+ "loss": 5.003549575805664,
+ "step": 1780
+ },
+ {
+ "epoch": 24.866943412715752,
+ "grad_norm": 0.18814566731452942,
+ "learning_rate": 0.0006,
+ "loss": 4.994387817382813,
+ "step": 1790
+ },
+ {
+ "epoch": 25.0,
+ "grad_norm": 0.22586414217948914,
+ "learning_rate": 0.0006,
+ "loss": 4.994391632080078,
+ "step": 1800
+ },
+ {
+ "epoch": 25.0,
+ "eval_loss": 3.976128101348877,
+ "eval_runtime": 19.9252,
+ "eval_samples_per_second": 122.508,
+ "eval_steps_per_second": 15.357,
+ "step": 1800
+ },
+ {
+ "epoch": 25.139829582696088,
+ "grad_norm": 0.20727957785129547,
+ "learning_rate": 0.0006,
+ "loss": 4.962083435058593,
+ "step": 1810
+ },
+ {
+ "epoch": 25.27965916539218,
+ "grad_norm": 0.20561031997203827,
+ "learning_rate": 0.0006,
+ "loss": 4.985612106323242,
+ "step": 1820
+ },
+ {
+ "epoch": 25.419488748088266,
+ "grad_norm": 0.23401397466659546,
+ "learning_rate": 0.0006,
+ "loss": 4.974888610839844,
+ "step": 1830
+ },
+ {
+ "epoch": 25.559318330784357,
+ "grad_norm": 0.20052191615104675,
+ "learning_rate": 0.0006,
+ "loss": 4.976966857910156,
+ "step": 1840
+ },
+ {
+ "epoch": 25.699147913480445,
+ "grad_norm": 0.21759679913520813,
+ "learning_rate": 0.0006,
+ "loss": 5.003248977661133,
+ "step": 1850
+ },
+ {
+ "epoch": 25.838977496176536,
+ "grad_norm": 0.19265390932559967,
+ "learning_rate": 0.0006,
+ "loss": 5.0107379913330075,
+ "step": 1860
+ },
+ {
+ "epoch": 25.978807078872624,
+ "grad_norm": 0.24477092921733856,
+ "learning_rate": 0.0006,
+ "loss": 4.9985393524169925,
+ "step": 1870
+ },
+ {
+ "epoch": 26.0,
+ "eval_loss": 3.969492197036743,
+ "eval_runtime": 19.9546,
+ "eval_samples_per_second": 122.328,
+ "eval_steps_per_second": 15.335,
+ "step": 1872
+ },
+ {
+ "epoch": 26.11186366615687,
+ "grad_norm": 0.22920085489749908,
+ "learning_rate": 0.0006,
+ "loss": 4.955302810668945,
+ "step": 1880
+ },
+ {
+ "epoch": 26.25169324885296,
+ "grad_norm": 0.2883584797382355,
+ "learning_rate": 0.0006,
+ "loss": 4.954433441162109,
+ "step": 1890
+ },
+ {
+ "epoch": 26.39152283154905,
+ "grad_norm": 0.20454390347003937,
+ "learning_rate": 0.0006,
+ "loss": 4.968096542358398,
+ "step": 1900
+ },
+ {
+ "epoch": 26.531352414245138,
+ "grad_norm": 0.19263872504234314,
+ "learning_rate": 0.0006,
+ "loss": 4.963127899169922,
+ "step": 1910
+ },
+ {
+ "epoch": 26.67118199694123,
+ "grad_norm": 0.19295209646224976,
+ "learning_rate": 0.0006,
+ "loss": 4.971673202514649,
+ "step": 1920
+ },
+ {
+ "epoch": 26.811011579637317,
+ "grad_norm": 0.19493436813354492,
+ "learning_rate": 0.0006,
+ "loss": 4.968273544311524,
+ "step": 1930
+ },
+ {
+ "epoch": 26.950841162333408,
+ "grad_norm": 0.20436061918735504,
+ "learning_rate": 0.0006,
+ "loss": 4.9730266571044925,
+ "step": 1940
+ },
+ {
+ "epoch": 27.0,
+ "eval_loss": 3.950493335723877,
+ "eval_runtime": 19.9056,
+ "eval_samples_per_second": 122.629,
+ "eval_steps_per_second": 15.373,
+ "step": 1944
+ },
+ {
+ "epoch": 27.083897749617652,
+ "grad_norm": 0.24062883853912354,
+ "learning_rate": 0.0006,
+ "loss": 4.9808917999267575,
+ "step": 1950
+ },
+ {
+ "epoch": 27.223727332313743,
+ "grad_norm": 0.2629009187221527,
+ "learning_rate": 0.0006,
+ "loss": 4.935837936401367,
+ "step": 1960
+ },
+ {
+ "epoch": 27.36355691500983,
+ "grad_norm": 0.19377221167087555,
+ "learning_rate": 0.0006,
+ "loss": 4.939309310913086,
+ "step": 1970
+ },
+ {
+ "epoch": 27.50338649770592,
+ "grad_norm": 0.22026003897190094,
+ "learning_rate": 0.0006,
+ "loss": 4.930693817138672,
+ "step": 1980
+ },
+ {
+ "epoch": 27.64321608040201,
+ "grad_norm": 0.21435536444187164,
+ "learning_rate": 0.0006,
+ "loss": 4.967763519287109,
+ "step": 1990
+ },
+ {
+ "epoch": 27.7830456630981,
+ "grad_norm": 0.18789879977703094,
+ "learning_rate": 0.0006,
+ "loss": 4.94080924987793,
+ "step": 2000
+ },
+ {
+ "epoch": 27.922875245794188,
+ "grad_norm": 0.21347181499004364,
+ "learning_rate": 0.0006,
+ "loss": 4.957330703735352,
+ "step": 2010
+ },
+ {
+ "epoch": 28.0,
+ "eval_loss": 3.9429216384887695,
+ "eval_runtime": 19.9321,
+ "eval_samples_per_second": 122.466,
+ "eval_steps_per_second": 15.352,
+ "step": 2016
+ },
+ {
+ "epoch": 28.055931833078436,
+ "grad_norm": 0.234244704246521,
+ "learning_rate": 0.0006,
+ "loss": 4.945362091064453,
+ "step": 2020
+ },
+ {
+ "epoch": 28.195761415774523,
+ "grad_norm": 0.1905280202627182,
+ "learning_rate": 0.0006,
+ "loss": 4.931621551513672,
+ "step": 2030
+ },
+ {
+ "epoch": 28.335590998470614,
+ "grad_norm": 0.25853726267814636,
+ "learning_rate": 0.0006,
+ "loss": 4.928206253051758,
+ "step": 2040
+ },
+ {
+ "epoch": 28.475420581166702,
+ "grad_norm": 0.23747533559799194,
+ "learning_rate": 0.0006,
+ "loss": 4.939360809326172,
+ "step": 2050
+ },
+ {
+ "epoch": 28.615250163862793,
+ "grad_norm": 0.19749325513839722,
+ "learning_rate": 0.0006,
+ "loss": 4.936768341064453,
+ "step": 2060
+ },
+ {
+ "epoch": 28.75507974655888,
+ "grad_norm": 0.2041572630405426,
+ "learning_rate": 0.0006,
+ "loss": 4.938620376586914,
+ "step": 2070
+ },
+ {
+ "epoch": 28.894909329254972,
+ "grad_norm": 0.19621336460113525,
+ "learning_rate": 0.0006,
+ "loss": 4.9312492370605465,
+ "step": 2080
+ },
+ {
+ "epoch": 29.0,
+ "eval_loss": 3.9307069778442383,
+ "eval_runtime": 19.9084,
+ "eval_samples_per_second": 122.612,
+ "eval_steps_per_second": 15.37,
+ "step": 2088
+ },
+ {
+ "epoch": 29.027965916539216,
+ "grad_norm": 0.23488345742225647,
+ "learning_rate": 0.0006,
+ "loss": 4.957594299316407,
+ "step": 2090
+ },
+ {
+ "epoch": 29.167795499235307,
+ "grad_norm": 0.23697027564048767,
+ "learning_rate": 0.0006,
+ "loss": 4.930992889404297,
+ "step": 2100
+ },
+ {
+ "epoch": 29.307625081931395,
+ "grad_norm": 0.20611435174942017,
+ "learning_rate": 0.0006,
+ "loss": 4.925578308105469,
+ "step": 2110
+ },
+ {
+ "epoch": 29.447454664627486,
+ "grad_norm": 0.22285430133342743,
+ "learning_rate": 0.0006,
+ "loss": 4.9125713348388675,
+ "step": 2120
+ },
+ {
+ "epoch": 29.587284247323574,
+ "grad_norm": 0.19881220161914825,
+ "learning_rate": 0.0006,
+ "loss": 4.9074348449707035,
+ "step": 2130
+ },
+ {
+ "epoch": 29.727113830019665,
+ "grad_norm": 0.20417934656143188,
+ "learning_rate": 0.0006,
+ "loss": 4.893528747558594,
+ "step": 2140
+ },
+ {
+ "epoch": 29.866943412715752,
+ "grad_norm": 0.2119961827993393,
+ "learning_rate": 0.0006,
+ "loss": 4.93562126159668,
+ "step": 2150
+ },
+ {
+ "epoch": 30.0,
+ "grad_norm": 0.22734826803207397,
+ "learning_rate": 0.0006,
+ "loss": 4.894660568237304,
+ "step": 2160
+ },
+ {
+ "epoch": 30.0,
+ "eval_loss": 3.920050621032715,
+ "eval_runtime": 19.9166,
+ "eval_samples_per_second": 122.561,
+ "eval_steps_per_second": 15.364,
+ "step": 2160
+ },
+ {
+ "epoch": 30.139829582696088,
+ "grad_norm": 0.2078607976436615,
+ "learning_rate": 0.0006,
+ "loss": 4.9121147155761715,
+ "step": 2170
+ },
+ {
+ "epoch": 30.27965916539218,
+ "grad_norm": 0.2205888032913208,
+ "learning_rate": 0.0006,
+ "loss": 4.899023056030273,
+ "step": 2180
+ },
+ {
+ "epoch": 30.419488748088266,
+ "grad_norm": 0.20526516437530518,
+ "learning_rate": 0.0006,
+ "loss": 4.8953392028808596,
+ "step": 2190
+ },
+ {
+ "epoch": 30.559318330784357,
+ "grad_norm": 0.20096731185913086,
+ "learning_rate": 0.0006,
+ "loss": 4.889182662963867,
+ "step": 2200
+ },
+ {
+ "epoch": 30.699147913480445,
+ "grad_norm": 0.194410040974617,
+ "learning_rate": 0.0006,
+ "loss": 4.904763031005859,
+ "step": 2210
+ },
+ {
+ "epoch": 30.838977496176536,
+ "grad_norm": 0.18821871280670166,
+ "learning_rate": 0.0006,
+ "loss": 4.886811447143555,
+ "step": 2220
+ },
+ {
+ "epoch": 30.978807078872624,
+ "grad_norm": 0.18639743328094482,
+ "learning_rate": 0.0006,
+ "loss": 4.908057022094726,
+ "step": 2230
+ },
+ {
+ "epoch": 31.0,
+ "eval_loss": 3.912003755569458,
+ "eval_runtime": 19.9097,
+ "eval_samples_per_second": 122.603,
+ "eval_steps_per_second": 15.369,
+ "step": 2232
+ },
+ {
+ "epoch": 31.11186366615687,
+ "grad_norm": 0.2815580368041992,
+ "learning_rate": 0.0006,
+ "loss": 4.8922382354736325,
+ "step": 2240
+ },
+ {
+ "epoch": 31.25169324885296,
+ "grad_norm": 6.756948947906494,
+ "learning_rate": 0.0006,
+ "loss": 5.476439666748047,
+ "step": 2250
+ },
+ {
+ "epoch": 31.39152283154905,
+ "grad_norm": 1.6655176877975464,
+ "learning_rate": 0.0006,
+ "loss": 5.90833854675293,
+ "step": 2260
+ },
+ {
+ "epoch": 31.531352414245138,
+ "grad_norm": 0.7790367603302002,
+ "learning_rate": 0.0006,
+ "loss": 5.738950729370117,
+ "step": 2270
+ },
+ {
+ "epoch": 31.67118199694123,
+ "grad_norm": 0.313378244638443,
+ "learning_rate": 0.0006,
+ "loss": 5.497149658203125,
+ "step": 2280
+ },
+ {
+ "epoch": 31.811011579637317,
+ "grad_norm": 0.2152474820613861,
+ "learning_rate": 0.0006,
+ "loss": 5.257891464233398,
+ "step": 2290
+ },
+ {
+ "epoch": 31.950841162333408,
+ "grad_norm": 0.16692155599594116,
+ "learning_rate": 0.0006,
+ "loss": 5.151300430297852,
+ "step": 2300
+ },
+ {
+ "epoch": 32.0,
+ "eval_loss": 4.0487060546875,
+ "eval_runtime": 19.9114,
+ "eval_samples_per_second": 122.593,
+ "eval_steps_per_second": 15.368,
+ "step": 2304
+ },
+ {
+ "epoch": 32.083897749617655,
+ "grad_norm": 0.15213488042354584,
+ "learning_rate": 0.0006,
+ "loss": 5.073321914672851,
+ "step": 2310
+ },
+ {
+ "epoch": 32.22372733231374,
+ "grad_norm": 0.1355232447385788,
+ "learning_rate": 0.0006,
+ "loss": 5.023114013671875,
+ "step": 2320
+ },
+ {
+ "epoch": 32.36355691500983,
+ "grad_norm": 0.1537737101316452,
+ "learning_rate": 0.0006,
+ "loss": 5.019962310791016,
+ "step": 2330
+ },
+ {
+ "epoch": 32.50338649770592,
+ "grad_norm": 0.14369140565395355,
+ "learning_rate": 0.0006,
+ "loss": 5.001427841186524,
+ "step": 2340
+ },
+ {
+ "epoch": 32.64321608040201,
+ "grad_norm": 0.135459765791893,
+ "learning_rate": 0.0006,
+ "loss": 4.967114639282227,
+ "step": 2350
+ },
+ {
+ "epoch": 32.7830456630981,
+ "grad_norm": 0.1543261855840683,
+ "learning_rate": 0.0006,
+ "loss": 4.939690017700196,
+ "step": 2360
+ },
+ {
+ "epoch": 32.92287524579419,
+ "grad_norm": 0.15223033726215363,
+ "learning_rate": 0.0006,
+ "loss": 4.9412891387939455,
+ "step": 2370
+ },
+ {
+ "epoch": 33.0,
+ "eval_loss": 3.9344186782836914,
+ "eval_runtime": 19.8977,
+ "eval_samples_per_second": 122.678,
+ "eval_steps_per_second": 15.379,
+ "step": 2376
+ },
+ {
+ "epoch": 33.05593183307843,
+ "grad_norm": 0.16246828436851501,
+ "learning_rate": 0.0006,
+ "loss": 4.9708984375,
+ "step": 2380
+ },
+ {
+ "epoch": 33.19576141577453,
+ "grad_norm": 0.15825116634368896,
+ "learning_rate": 0.0006,
+ "loss": 4.921619033813476,
+ "step": 2390
+ },
+ {
+ "epoch": 33.335590998470614,
+ "grad_norm": 0.14632634818553925,
+ "learning_rate": 0.0006,
+ "loss": 4.932820510864258,
+ "step": 2400
+ },
+ {
+ "epoch": 33.4754205811667,
+ "grad_norm": 0.16575364768505096,
+ "learning_rate": 0.0006,
+ "loss": 4.912859344482422,
+ "step": 2410
+ },
+ {
+ "epoch": 33.61525016386279,
+ "grad_norm": 0.15159428119659424,
+ "learning_rate": 0.0006,
+ "loss": 4.90052375793457,
+ "step": 2420
+ },
+ {
+ "epoch": 33.755079746558884,
+ "grad_norm": 0.15570823848247528,
+ "learning_rate": 0.0006,
+ "loss": 4.902946472167969,
+ "step": 2430
+ },
+ {
+ "epoch": 33.89490932925497,
+ "grad_norm": 0.14956501126289368,
+ "learning_rate": 0.0006,
+ "loss": 4.926367568969726,
+ "step": 2440
+ },
+ {
+ "epoch": 34.0,
+ "eval_loss": 3.91025710105896,
+ "eval_runtime": 19.9139,
+ "eval_samples_per_second": 122.578,
+ "eval_steps_per_second": 15.366,
+ "step": 2448
+ },
+ {
+ "epoch": 34.027965916539216,
+ "grad_norm": 0.15200866758823395,
+ "learning_rate": 0.0006,
+ "loss": 4.878167724609375,
+ "step": 2450
+ },
+ {
+ "epoch": 34.167795499235304,
+ "grad_norm": 0.16813978552818298,
+ "learning_rate": 0.0006,
+ "loss": 4.886758041381836,
+ "step": 2460
+ },
+ {
+ "epoch": 34.3076250819314,
+ "grad_norm": 0.14943887293338776,
+ "learning_rate": 0.0006,
+ "loss": 4.894789505004883,
+ "step": 2470
+ },
+ {
+ "epoch": 34.447454664627486,
+ "grad_norm": 0.17818871140480042,
+ "learning_rate": 0.0006,
+ "loss": 4.883363342285156,
+ "step": 2480
+ },
+ {
+ "epoch": 34.58728424732357,
+ "grad_norm": 0.16085270047187805,
+ "learning_rate": 0.0006,
+ "loss": 4.879457473754883,
+ "step": 2490
+ },
+ {
+ "epoch": 34.72711383001966,
+ "grad_norm": 0.16816453635692596,
+ "learning_rate": 0.0006,
+ "loss": 4.886758041381836,
+ "step": 2500
+ },
+ {
+ "epoch": 34.866943412715756,
+ "grad_norm": 0.175967738032341,
+ "learning_rate": 0.0006,
+ "loss": 4.890745544433594,
+ "step": 2510
+ },
+ {
+ "epoch": 35.0,
+ "grad_norm": 0.1912197321653366,
+ "learning_rate": 0.0006,
+ "loss": 4.899010848999024,
+ "step": 2520
+ },
+ {
+ "epoch": 35.0,
+ "eval_loss": 3.8980422019958496,
+ "eval_runtime": 19.8985,
+ "eval_samples_per_second": 122.672,
+ "eval_steps_per_second": 15.378,
+ "step": 2520
+ },
+ {
+ "epoch": 35.13982958269609,
+ "grad_norm": 0.17858198285102844,
+ "learning_rate": 0.0006,
+ "loss": 4.877031326293945,
+ "step": 2530
+ },
+ {
+ "epoch": 35.279659165392175,
+ "grad_norm": 0.17253074049949646,
+ "learning_rate": 0.0006,
+ "loss": 4.855865097045898,
+ "step": 2540
+ },
+ {
+ "epoch": 35.41948874808827,
+ "grad_norm": 0.17144449055194855,
+ "learning_rate": 0.0006,
+ "loss": 4.849673461914063,
+ "step": 2550
+ },
+ {
+ "epoch": 35.55931833078436,
+ "grad_norm": 0.16289135813713074,
+ "learning_rate": 0.0006,
+ "loss": 4.882015991210937,
+ "step": 2560
+ },
+ {
+ "epoch": 35.699147913480445,
+ "grad_norm": 0.16993777453899384,
+ "learning_rate": 0.0006,
+ "loss": 4.897224044799804,
+ "step": 2570
+ },
+ {
+ "epoch": 35.83897749617653,
+ "grad_norm": 0.16243058443069458,
+ "learning_rate": 0.0006,
+ "loss": 4.8659820556640625,
+ "step": 2580
+ },
+ {
+ "epoch": 35.97880707887263,
+ "grad_norm": 0.17329761385917664,
+ "learning_rate": 0.0006,
+ "loss": 4.88317985534668,
+ "step": 2590
+ },
+ {
+ "epoch": 36.0,
+ "eval_loss": 3.889164686203003,
+ "eval_runtime": 19.8825,
+ "eval_samples_per_second": 122.771,
+ "eval_steps_per_second": 15.39,
+ "step": 2592
+ },
+ {
+ "epoch": 36.11186366615687,
+ "grad_norm": 0.16629834473133087,
+ "learning_rate": 0.0006,
+ "loss": 4.831401443481445,
+ "step": 2600
+ },
+ {
+ "epoch": 36.25169324885296,
+ "grad_norm": 0.1753329634666443,
+ "learning_rate": 0.0006,
+ "loss": 4.850980377197265,
+ "step": 2610
+ },
+ {
+ "epoch": 36.39152283154905,
+ "grad_norm": 0.16380569338798523,
+ "learning_rate": 0.0006,
+ "loss": 4.847088241577149,
+ "step": 2620
+ },
+ {
+ "epoch": 36.53135241424514,
+ "grad_norm": 0.17095185816287994,
+ "learning_rate": 0.0006,
+ "loss": 4.867258453369141,
+ "step": 2630
+ },
+ {
+ "epoch": 36.67118199694123,
+ "grad_norm": 0.18793116509914398,
+ "learning_rate": 0.0006,
+ "loss": 4.849470520019532,
+ "step": 2640
+ },
+ {
+ "epoch": 36.81101157963732,
+ "grad_norm": 0.17469213902950287,
+ "learning_rate": 0.0006,
+ "loss": 4.843376922607422,
+ "step": 2650
+ },
+ {
+ "epoch": 36.950841162333404,
+ "grad_norm": 0.1608351469039917,
+ "learning_rate": 0.0006,
+ "loss": 4.861382675170899,
+ "step": 2660
+ },
+ {
+ "epoch": 37.0,
+ "eval_loss": 3.881481409072876,
+ "eval_runtime": 19.8909,
+ "eval_samples_per_second": 122.72,
+ "eval_steps_per_second": 15.384,
+ "step": 2664
+ },
+ {
+ "epoch": 37.083897749617655,
+ "grad_norm": 0.17486630380153656,
+ "learning_rate": 0.0006,
+ "loss": 4.822615051269532,
+ "step": 2670
+ },
+ {
+ "epoch": 37.22372733231374,
+ "grad_norm": 0.18961143493652344,
+ "learning_rate": 0.0006,
+ "loss": 4.8173072814941404,
+ "step": 2680
+ },
+ {
+ "epoch": 37.36355691500983,
+ "grad_norm": 0.18518193066120148,
+ "learning_rate": 0.0006,
+ "loss": 4.8454032897949215,
+ "step": 2690
+ },
+ {
+ "epoch": 37.50338649770592,
+ "grad_norm": 0.17764227092266083,
+ "learning_rate": 0.0006,
+ "loss": 4.844810867309571,
+ "step": 2700
+ },
+ {
+ "epoch": 37.64321608040201,
+ "grad_norm": 0.1789189875125885,
+ "learning_rate": 0.0006,
+ "loss": 4.825796508789063,
+ "step": 2710
+ },
+ {
+ "epoch": 37.7830456630981,
+ "grad_norm": 0.16197077929973602,
+ "learning_rate": 0.0006,
+ "loss": 4.836392211914062,
+ "step": 2720
+ },
+ {
+ "epoch": 37.92287524579419,
+ "grad_norm": 0.15372976660728455,
+ "learning_rate": 0.0006,
+ "loss": 4.854349136352539,
+ "step": 2730
+ },
+ {
+ "epoch": 38.0,
+ "eval_loss": 3.872185230255127,
+ "eval_runtime": 19.8834,
+ "eval_samples_per_second": 122.766,
+ "eval_steps_per_second": 15.39,
+ "step": 2736
+ },
+ {
+ "epoch": 38.05593183307843,
+ "grad_norm": 0.19343572854995728,
+ "learning_rate": 0.0006,
+ "loss": 4.856714248657227,
+ "step": 2740
+ },
+ {
+ "epoch": 38.19576141577453,
+ "grad_norm": 0.1915244162082672,
+ "learning_rate": 0.0006,
+ "loss": 4.821109390258789,
+ "step": 2750
+ },
+ {
+ "epoch": 38.335590998470614,
+ "grad_norm": 0.18525931239128113,
+ "learning_rate": 0.0006,
+ "loss": 4.833884048461914,
+ "step": 2760
+ },
+ {
+ "epoch": 38.4754205811667,
+ "grad_norm": 0.17032141983509064,
+ "learning_rate": 0.0006,
+ "loss": 4.841216278076172,
+ "step": 2770
+ },
+ {
+ "epoch": 38.61525016386279,
+ "grad_norm": 0.217947855591774,
+ "learning_rate": 0.0006,
+ "loss": 4.855699920654297,
+ "step": 2780
+ },
+ {
+ "epoch": 38.755079746558884,
+ "grad_norm": 0.17534874379634857,
+ "learning_rate": 0.0006,
+ "loss": 4.827981948852539,
+ "step": 2790
+ },
+ {
+ "epoch": 38.89490932925497,
+ "grad_norm": 0.1840103566646576,
+ "learning_rate": 0.0006,
+ "loss": 4.844279861450195,
+ "step": 2800
+ },
+ {
+ "epoch": 39.0,
+ "eval_loss": 3.8703455924987793,
+ "eval_runtime": 19.8663,
+ "eval_samples_per_second": 122.872,
+ "eval_steps_per_second": 15.403,
+ "step": 2808
+ },
+ {
+ "epoch": 39.027965916539216,
+ "grad_norm": 0.16699740290641785,
+ "learning_rate": 0.0006,
+ "loss": 4.835521697998047,
+ "step": 2810
+ },
+ {
+ "epoch": 39.167795499235304,
+ "grad_norm": 0.18986597657203674,
+ "learning_rate": 0.0006,
+ "loss": 4.812285232543945,
+ "step": 2820
+ },
+ {
+ "epoch": 39.3076250819314,
+ "grad_norm": 0.17179296910762787,
+ "learning_rate": 0.0006,
+ "loss": 4.803129577636719,
+ "step": 2830
+ },
+ {
+ "epoch": 39.447454664627486,
+ "grad_norm": 0.16693517565727234,
+ "learning_rate": 0.0006,
+ "loss": 4.80714225769043,
+ "step": 2840
+ },
+ {
+ "epoch": 39.58728424732357,
+ "grad_norm": 0.1756565123796463,
+ "learning_rate": 0.0006,
+ "loss": 4.8322502136230465,
+ "step": 2850
+ },
+ {
+ "epoch": 39.72711383001966,
+ "grad_norm": 0.17199955880641937,
+ "learning_rate": 0.0006,
+ "loss": 4.851013946533203,
+ "step": 2860
+ },
+ {
+ "epoch": 39.866943412715756,
+ "grad_norm": 0.1791684627532959,
+ "learning_rate": 0.0006,
+ "loss": 4.8342845916748045,
+ "step": 2870
+ },
+ {
+ "epoch": 40.0,
+ "grad_norm": 0.20261150598526,
+ "learning_rate": 0.0006,
+ "loss": 4.821056365966797,
+ "step": 2880
+ },
+ {
+ "epoch": 40.0,
+ "eval_loss": 3.864354133605957,
+ "eval_runtime": 19.8825,
+ "eval_samples_per_second": 122.771,
+ "eval_steps_per_second": 15.39,
+ "step": 2880
+ },
+ {
+ "epoch": 40.13982958269609,
+ "grad_norm": 0.17401723563671112,
+ "learning_rate": 0.0006,
+ "loss": 4.776754379272461,
+ "step": 2890
+ },
+ {
+ "epoch": 40.279659165392175,
+ "grad_norm": 0.17896327376365662,
+ "learning_rate": 0.0006,
+ "loss": 4.796353530883789,
+ "step": 2900
+ },
+ {
+ "epoch": 40.41948874808827,
+ "grad_norm": 0.1984410136938095,
+ "learning_rate": 0.0006,
+ "loss": 4.806838607788086,
+ "step": 2910
+ },
+ {
+ "epoch": 40.55931833078436,
+ "grad_norm": 0.1861223578453064,
+ "learning_rate": 0.0006,
+ "loss": 4.83325424194336,
+ "step": 2920
+ },
+ {
+ "epoch": 40.699147913480445,
+ "grad_norm": 0.1716955453157425,
+ "learning_rate": 0.0006,
+ "loss": 4.826747131347656,
+ "step": 2930
+ },
+ {
+ "epoch": 40.83897749617653,
+ "grad_norm": 0.19898776710033417,
+ "learning_rate": 0.0006,
+ "loss": 4.819411087036133,
+ "step": 2940
+ },
+ {
+ "epoch": 40.97880707887263,
+ "grad_norm": 0.19300246238708496,
+ "learning_rate": 0.0006,
+ "loss": 4.828544616699219,
+ "step": 2950
+ },
+ {
+ "epoch": 41.0,
+ "eval_loss": 3.8604962825775146,
+ "eval_runtime": 19.8825,
+ "eval_samples_per_second": 122.771,
+ "eval_steps_per_second": 15.39,
+ "step": 2952
+ },
+ {
+ "epoch": 41.11186366615687,
+ "grad_norm": 0.20119570195674896,
+ "learning_rate": 0.0006,
+ "loss": 4.790631866455078,
+ "step": 2960
+ },
+ {
+ "epoch": 41.25169324885296,
+ "grad_norm": 0.17397235333919525,
+ "learning_rate": 0.0006,
+ "loss": 4.781435775756836,
+ "step": 2970
+ },
+ {
+ "epoch": 41.39152283154905,
+ "grad_norm": 0.17674098908901215,
+ "learning_rate": 0.0006,
+ "loss": 4.787834548950196,
+ "step": 2980
+ },
+ {
+ "epoch": 41.53135241424514,
+ "grad_norm": 0.18732580542564392,
+ "learning_rate": 0.0006,
+ "loss": 4.8054157257080075,
+ "step": 2990
+ },
+ {
+ "epoch": 41.67118199694123,
+ "grad_norm": 0.19507819414138794,
+ "learning_rate": 0.0006,
+ "loss": 4.792606353759766,
+ "step": 3000
+ },
+ {
+ "epoch": 41.81101157963732,
+ "grad_norm": 0.20151498913764954,
+ "learning_rate": 0.0006,
+ "loss": 4.805193328857422,
+ "step": 3010
+ },
+ {
+ "epoch": 41.950841162333404,
+ "grad_norm": 0.17535369098186493,
+ "learning_rate": 0.0006,
+ "loss": 4.815096282958985,
+ "step": 3020
+ },
+ {
+ "epoch": 42.0,
+ "eval_loss": 3.8537988662719727,
+ "eval_runtime": 19.8799,
+ "eval_samples_per_second": 122.788,
+ "eval_steps_per_second": 15.392,
+ "step": 3024
+ },
+ {
+ "epoch": 42.083897749617655,
+ "grad_norm": 0.21616001427173615,
+ "learning_rate": 0.0006,
+ "loss": 4.780527114868164,
+ "step": 3030
+ },
+ {
+ "epoch": 42.22372733231374,
+ "grad_norm": 0.19461651146411896,
+ "learning_rate": 0.0006,
+ "loss": 4.763539886474609,
+ "step": 3040
+ },
+ {
+ "epoch": 42.36355691500983,
+ "grad_norm": 0.18285132944583893,
+ "learning_rate": 0.0006,
+ "loss": 4.764186859130859,
+ "step": 3050
+ },
+ {
+ "epoch": 42.50338649770592,
+ "grad_norm": 0.1856558620929718,
+ "learning_rate": 0.0006,
+ "loss": 4.819770431518554,
+ "step": 3060
+ },
+ {
+ "epoch": 42.64321608040201,
+ "grad_norm": 0.18869610130786896,
+ "learning_rate": 0.0006,
+ "loss": 4.7930137634277346,
+ "step": 3070
+ },
+ {
+ "epoch": 42.7830456630981,
+ "grad_norm": 0.18057413399219513,
+ "learning_rate": 0.0006,
+ "loss": 4.804056167602539,
+ "step": 3080
+ },
+ {
+ "epoch": 42.92287524579419,
+ "grad_norm": 0.1747533529996872,
+ "learning_rate": 0.0006,
+ "loss": 4.796476745605469,
+ "step": 3090
+ },
+ {
+ "epoch": 43.0,
+ "eval_loss": 3.8545141220092773,
+ "eval_runtime": 19.8761,
+ "eval_samples_per_second": 122.811,
+ "eval_steps_per_second": 15.395,
+ "step": 3096
+ },
+ {
+ "epoch": 43.05593183307843,
+ "grad_norm": 0.19284771382808685,
+ "learning_rate": 0.0006,
+ "loss": 4.799809265136719,
+ "step": 3100
+ },
+ {
+ "epoch": 43.19576141577453,
+ "grad_norm": 0.19247348606586456,
+ "learning_rate": 0.0006,
+ "loss": 4.767494583129883,
+ "step": 3110
+ },
+ {
+ "epoch": 43.335590998470614,
+ "grad_norm": 0.19674018025398254,
+ "learning_rate": 0.0006,
+ "loss": 4.781412887573242,
+ "step": 3120
+ },
+ {
+ "epoch": 43.4754205811667,
+ "grad_norm": 0.18593478202819824,
+ "learning_rate": 0.0006,
+ "loss": 4.75941162109375,
+ "step": 3130
+ },
+ {
+ "epoch": 43.61525016386279,
+ "grad_norm": 0.18039630353450775,
+ "learning_rate": 0.0006,
+ "loss": 4.807698440551758,
+ "step": 3140
+ },
+ {
+ "epoch": 43.755079746558884,
+ "grad_norm": 0.19801653921604156,
+ "learning_rate": 0.0006,
+ "loss": 4.77733154296875,
+ "step": 3150
+ },
+ {
+ "epoch": 43.89490932925497,
+ "grad_norm": 0.17759037017822266,
+ "learning_rate": 0.0006,
+ "loss": 4.772528839111328,
+ "step": 3160
+ },
+ {
+ "epoch": 44.0,
+ "eval_loss": 3.8480331897735596,
+ "eval_runtime": 19.8827,
+ "eval_samples_per_second": 122.77,
+ "eval_steps_per_second": 15.39,
+ "step": 3168
+ },
+ {
+ "epoch": 44.027965916539216,
+ "grad_norm": 0.18391110002994537,
+ "learning_rate": 0.0006,
+ "loss": 4.771400833129883,
+ "step": 3170
+ },
+ {
+ "epoch": 44.167795499235304,
+ "grad_norm": 0.1799115091562271,
+ "learning_rate": 0.0006,
+ "loss": 4.760651779174805,
+ "step": 3180
+ },
+ {
+ "epoch": 44.3076250819314,
+ "grad_norm": 0.19926029443740845,
+ "learning_rate": 0.0006,
+ "loss": 4.783069610595703,
+ "step": 3190
+ },
+ {
+ "epoch": 44.447454664627486,
+ "grad_norm": 0.22545892000198364,
+ "learning_rate": 0.0006,
+ "loss": 4.761959075927734,
+ "step": 3200
+ },
+ {
+ "epoch": 44.58728424732357,
+ "grad_norm": 0.202196404337883,
+ "learning_rate": 0.0006,
+ "loss": 4.770500946044922,
+ "step": 3210
+ },
+ {
+ "epoch": 44.72711383001966,
+ "grad_norm": 0.22643990814685822,
+ "learning_rate": 0.0006,
+ "loss": 4.784806442260742,
+ "step": 3220
+ },
+ {
+ "epoch": 44.866943412715756,
+ "grad_norm": 0.18608754873275757,
+ "learning_rate": 0.0006,
+ "loss": 4.816587829589844,
+ "step": 3230
+ },
+ {
+ "epoch": 45.0,
+ "grad_norm": 0.21552273631095886,
+ "learning_rate": 0.0006,
+ "loss": 4.767296981811524,
+ "step": 3240
+ },
+ {
+ "epoch": 45.0,
+ "eval_loss": 3.843341827392578,
+ "eval_runtime": 19.8736,
+ "eval_samples_per_second": 122.827,
+ "eval_steps_per_second": 15.397,
+ "step": 3240
+ },
+ {
+ "epoch": 45.13982958269609,
+ "grad_norm": 0.19632185995578766,
+ "learning_rate": 0.0006,
+ "loss": 4.756115341186524,
+ "step": 3250
+ },
+ {
+ "epoch": 45.279659165392175,
+ "grad_norm": 0.16628552973270416,
+ "learning_rate": 0.0006,
+ "loss": 4.743072509765625,
+ "step": 3260
+ },
+ {
+ "epoch": 45.41948874808827,
+ "grad_norm": 0.17863409221172333,
+ "learning_rate": 0.0006,
+ "loss": 4.751149749755859,
+ "step": 3270
+ },
+ {
+ "epoch": 45.55931833078436,
+ "grad_norm": 0.18641866743564606,
+ "learning_rate": 0.0006,
+ "loss": 4.778366470336914,
+ "step": 3280
+ },
+ {
+ "epoch": 45.699147913480445,
+ "grad_norm": 0.18890663981437683,
+ "learning_rate": 0.0006,
+ "loss": 4.7641548156738285,
+ "step": 3290
+ },
+ {
+ "epoch": 45.83897749617653,
+ "grad_norm": 0.1934252381324768,
+ "learning_rate": 0.0006,
+ "loss": 4.760050201416016,
+ "step": 3300
+ },
+ {
+ "epoch": 45.97880707887263,
+ "grad_norm": 0.18169492483139038,
+ "learning_rate": 0.0006,
+ "loss": 4.789390563964844,
+ "step": 3310
+ },
+ {
+ "epoch": 46.0,
+ "eval_loss": 3.841214418411255,
+ "eval_runtime": 19.8922,
+ "eval_samples_per_second": 122.711,
+ "eval_steps_per_second": 15.383,
+ "step": 3312
+ },
+ {
+ "epoch": 46.11186366615687,
+ "grad_norm": 0.23312604427337646,
+ "learning_rate": 0.0006,
+ "loss": 4.758332443237305,
+ "step": 3320
+ },
+ {
+ "epoch": 46.25169324885296,
+ "grad_norm": 0.21910175681114197,
+ "learning_rate": 0.0006,
+ "loss": 4.749750137329102,
+ "step": 3330
+ },
+ {
+ "epoch": 46.39152283154905,
+ "grad_norm": 0.18404679000377655,
+ "learning_rate": 0.0006,
+ "loss": 4.746379089355469,
+ "step": 3340
+ },
+ {
+ "epoch": 46.53135241424514,
+ "grad_norm": 0.18589167296886444,
+ "learning_rate": 0.0006,
+ "loss": 4.746405410766601,
+ "step": 3350
+ },
+ {
+ "epoch": 46.67118199694123,
+ "grad_norm": 0.1808452606201172,
+ "learning_rate": 0.0006,
+ "loss": 4.742364120483399,
+ "step": 3360
+ },
+ {
+ "epoch": 46.81101157963732,
+ "grad_norm": 0.18686114251613617,
+ "learning_rate": 0.0006,
+ "loss": 4.772963333129883,
+ "step": 3370
+ },
+ {
+ "epoch": 46.950841162333404,
+ "grad_norm": 0.19424539804458618,
+ "learning_rate": 0.0006,
+ "loss": 4.7697795867919925,
+ "step": 3380
+ },
+ {
+ "epoch": 47.0,
+ "eval_loss": 3.8399569988250732,
+ "eval_runtime": 19.8756,
+ "eval_samples_per_second": 122.814,
+ "eval_steps_per_second": 15.396,
+ "step": 3384
+ },
+ {
+ "epoch": 47.083897749617655,
+ "grad_norm": 0.19590376317501068,
+ "learning_rate": 0.0006,
+ "loss": 4.746635437011719,
+ "step": 3390
+ },
+ {
+ "epoch": 47.22372733231374,
+ "grad_norm": 0.18070504069328308,
+ "learning_rate": 0.0006,
+ "loss": 4.7387939453125,
+ "step": 3400
+ },
+ {
+ "epoch": 47.36355691500983,
+ "grad_norm": 0.2168056219816208,
+ "learning_rate": 0.0006,
+ "loss": 4.746350860595703,
+ "step": 3410
+ },
+ {
+ "epoch": 47.50338649770592,
+ "grad_norm": 0.2072831243276596,
+ "learning_rate": 0.0006,
+ "loss": 4.757722854614258,
+ "step": 3420
+ },
+ {
+ "epoch": 47.64321608040201,
+ "grad_norm": 0.18567147850990295,
+ "learning_rate": 0.0006,
+ "loss": 4.736513900756836,
+ "step": 3430
+ },
+ {
+ "epoch": 47.7830456630981,
+ "grad_norm": 0.18916817009449005,
+ "learning_rate": 0.0006,
+ "loss": 4.742779541015625,
+ "step": 3440
+ },
+ {
+ "epoch": 47.92287524579419,
+ "grad_norm": 0.19555436074733734,
+ "learning_rate": 0.0006,
+ "loss": 4.762164688110351,
+ "step": 3450
+ },
+ {
+ "epoch": 48.0,
+ "eval_loss": 3.839027166366577,
+ "eval_runtime": 19.8866,
+ "eval_samples_per_second": 122.746,
+ "eval_steps_per_second": 15.387,
+ "step": 3456
+ },
+ {
+ "epoch": 48.05593183307843,
+ "grad_norm": 0.18981625139713287,
+ "learning_rate": 0.0006,
+ "loss": 4.7446025848388675,
+ "step": 3460
+ },
+ {
+ "epoch": 48.19576141577453,
+ "grad_norm": 0.1893685907125473,
+ "learning_rate": 0.0006,
+ "loss": 4.741929244995117,
+ "step": 3470
+ },
+ {
+ "epoch": 48.335590998470614,
+ "grad_norm": 0.18987691402435303,
+ "learning_rate": 0.0006,
+ "loss": 4.715566253662109,
+ "step": 3480
+ },
+ {
+ "epoch": 48.4754205811667,
+ "grad_norm": 0.1902618259191513,
+ "learning_rate": 0.0006,
+ "loss": 4.7344520568847654,
+ "step": 3490
+ },
+ {
+ "epoch": 48.61525016386279,
+ "grad_norm": 0.1950402855873108,
+ "learning_rate": 0.0006,
+ "loss": 4.744307327270508,
+ "step": 3500
+ },
+ {
+ "epoch": 48.755079746558884,
+ "grad_norm": 0.18797294795513153,
+ "learning_rate": 0.0006,
+ "loss": 4.740202331542969,
+ "step": 3510
+ },
+ {
+ "epoch": 48.89490932925497,
+ "grad_norm": 0.1715318262577057,
+ "learning_rate": 0.0006,
+ "loss": 4.741127014160156,
+ "step": 3520
+ },
+ {
+ "epoch": 49.0,
+ "eval_loss": 3.83217716217041,
+ "eval_runtime": 19.8839,
+ "eval_samples_per_second": 122.762,
+ "eval_steps_per_second": 15.389,
+ "step": 3528
+ },
+ {
+ "epoch": 49.027965916539216,
+ "grad_norm": 0.19864198565483093,
+ "learning_rate": 0.0006,
+ "loss": 4.728072738647461,
+ "step": 3530
+ },
+ {
+ "epoch": 49.167795499235304,
+ "grad_norm": 0.18943680822849274,
+ "learning_rate": 0.0006,
+ "loss": 4.703773498535156,
+ "step": 3540
+ },
+ {
+ "epoch": 49.3076250819314,
+ "grad_norm": 0.20500534772872925,
+ "learning_rate": 0.0006,
+ "loss": 4.742556381225586,
+ "step": 3550
+ },
+ {
+ "epoch": 49.447454664627486,
+ "grad_norm": 0.18511460721492767,
+ "learning_rate": 0.0006,
+ "loss": 4.715683364868164,
+ "step": 3560
+ },
+ {
+ "epoch": 49.58728424732357,
+ "grad_norm": 0.21317464113235474,
+ "learning_rate": 0.0006,
+ "loss": 4.7390178680419925,
+ "step": 3570
+ },
+ {
+ "epoch": 49.72711383001966,
+ "grad_norm": 0.16918888688087463,
+ "learning_rate": 0.0006,
+ "loss": 4.724227523803711,
+ "step": 3580
+ },
+ {
+ "epoch": 49.866943412715756,
+ "grad_norm": 0.19701409339904785,
+ "learning_rate": 0.0006,
+ "loss": 4.736687088012696,
+ "step": 3590
+ },
+ {
+ "epoch": 50.0,
+ "grad_norm": 0.23444531857967377,
+ "learning_rate": 0.0006,
+ "loss": 4.765250396728516,
+ "step": 3600
+ },
+ {
+ "epoch": 50.0,
+ "eval_loss": 3.833585500717163,
+ "eval_runtime": 19.8638,
+ "eval_samples_per_second": 122.887,
+ "eval_steps_per_second": 15.405,
+ "step": 3600
+ },
+ {
+ "epoch": 50.13982958269609,
+ "grad_norm": 0.19077414274215698,
+ "learning_rate": 0.0006,
+ "loss": 4.716189575195313,
+ "step": 3610
+ },
+ {
+ "epoch": 50.279659165392175,
+ "grad_norm": 0.21307913959026337,
+ "learning_rate": 0.0006,
+ "loss": 4.694992446899414,
+ "step": 3620
+ },
+ {
+ "epoch": 50.41948874808827,
+ "grad_norm": 0.20392359793186188,
+ "learning_rate": 0.0006,
+ "loss": 4.732614517211914,
+ "step": 3630
+ },
+ {
+ "epoch": 50.55931833078436,
+ "grad_norm": 0.1844174563884735,
+ "learning_rate": 0.0006,
+ "loss": 4.728442764282226,
+ "step": 3640
+ },
+ {
+ "epoch": 50.699147913480445,
+ "grad_norm": 0.18802128732204437,
+ "learning_rate": 0.0006,
+ "loss": 4.736631774902344,
+ "step": 3650
+ },
+ {
+ "epoch": 50.83897749617653,
+ "grad_norm": 0.19038118422031403,
+ "learning_rate": 0.0006,
+ "loss": 4.754737091064453,
+ "step": 3660
+ },
+ {
+ "epoch": 50.97880707887263,
+ "grad_norm": 0.1851847618818283,
+ "learning_rate": 0.0006,
+ "loss": 4.7560173034667965,
+ "step": 3670
+ },
+ {
+ "epoch": 51.0,
+ "eval_loss": 3.8279497623443604,
+ "eval_runtime": 19.8773,
+ "eval_samples_per_second": 122.803,
+ "eval_steps_per_second": 15.394,
+ "step": 3672
+ },
+ {
+ "epoch": 51.11186366615687,
+ "grad_norm": 0.19092661142349243,
+ "learning_rate": 0.0006,
+ "loss": 4.715619277954102,
+ "step": 3680
+ },
+ {
+ "epoch": 51.25169324885296,
+ "grad_norm": 0.1901930719614029,
+ "learning_rate": 0.0006,
+ "loss": 4.7030292510986325,
+ "step": 3690
+ },
+ {
+ "epoch": 51.39152283154905,
+ "grad_norm": 0.1821942776441574,
+ "learning_rate": 0.0006,
+ "loss": 4.715106582641601,
+ "step": 3700
+ },
+ {
+ "epoch": 51.53135241424514,
+ "grad_norm": 0.16833125054836273,
+ "learning_rate": 0.0006,
+ "loss": 4.702355194091797,
+ "step": 3710
+ },
+ {
+ "epoch": 51.67118199694123,
+ "grad_norm": 0.19603586196899414,
+ "learning_rate": 0.0006,
+ "loss": 4.765693664550781,
+ "step": 3720
+ },
+ {
+ "epoch": 51.81101157963732,
+ "grad_norm": 0.16649724543094635,
+ "learning_rate": 0.0006,
+ "loss": 4.715229034423828,
+ "step": 3730
+ },
+ {
+ "epoch": 51.950841162333404,
+ "grad_norm": 0.18829794228076935,
+ "learning_rate": 0.0006,
+ "loss": 4.749346923828125,
+ "step": 3740
+ },
+ {
+ "epoch": 52.0,
+ "eval_loss": 3.8237857818603516,
+ "eval_runtime": 19.8776,
+ "eval_samples_per_second": 122.801,
+ "eval_steps_per_second": 15.394,
+ "step": 3744
+ },
+ {
+ "epoch": 52.083897749617655,
+ "grad_norm": 0.20663219690322876,
+ "learning_rate": 0.0006,
+ "loss": 4.724364852905273,
+ "step": 3750
+ },
+ {
+ "epoch": 52.22372733231374,
+ "grad_norm": 0.23545010387897491,
+ "learning_rate": 0.0006,
+ "loss": 4.720872116088867,
+ "step": 3760
+ },
+ {
+ "epoch": 52.36355691500983,
+ "grad_norm": 0.18443231284618378,
+ "learning_rate": 0.0006,
+ "loss": 4.731410980224609,
+ "step": 3770
+ },
+ {
+ "epoch": 52.50338649770592,
+ "grad_norm": 0.18234513700008392,
+ "learning_rate": 0.0006,
+ "loss": 4.715656661987305,
+ "step": 3780
+ },
+ {
+ "epoch": 52.64321608040201,
+ "grad_norm": 0.18054722249507904,
+ "learning_rate": 0.0006,
+ "loss": 4.716067123413086,
+ "step": 3790
+ },
+ {
+ "epoch": 52.7830456630981,
+ "grad_norm": 0.20188716053962708,
+ "learning_rate": 0.0006,
+ "loss": 4.725110626220703,
+ "step": 3800
+ },
+ {
+ "epoch": 52.92287524579419,
+ "grad_norm": 0.18996301293373108,
+ "learning_rate": 0.0006,
+ "loss": 4.710061264038086,
+ "step": 3810
+ },
+ {
+ "epoch": 53.0,
+ "eval_loss": 3.8232734203338623,
+ "eval_runtime": 19.8738,
+ "eval_samples_per_second": 122.825,
+ "eval_steps_per_second": 15.397,
+ "step": 3816
+ },
+ {
+ "epoch": 53.05593183307843,
+ "grad_norm": 0.18608900904655457,
+ "learning_rate": 0.0006,
+ "loss": 4.744919204711914,
+ "step": 3820
+ },
+ {
+ "epoch": 53.19576141577453,
+ "grad_norm": 0.188418447971344,
+ "learning_rate": 0.0006,
+ "loss": 4.694300842285156,
+ "step": 3830
+ },
+ {
+ "epoch": 53.335590998470614,
+ "grad_norm": 0.20073595643043518,
+ "learning_rate": 0.0006,
+ "loss": 4.70234260559082,
+ "step": 3840
+ },
+ {
+ "epoch": 53.4754205811667,
+ "grad_norm": 0.1863900125026703,
+ "learning_rate": 0.0006,
+ "loss": 4.715248107910156,
+ "step": 3850
+ },
+ {
+ "epoch": 53.61525016386279,
+ "grad_norm": 0.19397905468940735,
+ "learning_rate": 0.0006,
+ "loss": 4.719331359863281,
+ "step": 3860
+ },
+ {
+ "epoch": 53.755079746558884,
+ "grad_norm": 0.18346035480499268,
+ "learning_rate": 0.0006,
+ "loss": 4.7100261688232425,
+ "step": 3870
+ },
+ {
+ "epoch": 53.89490932925497,
+ "grad_norm": 0.184013232588768,
+ "learning_rate": 0.0006,
+ "loss": 4.736498260498047,
+ "step": 3880
+ },
+ {
+ "epoch": 54.0,
+ "eval_loss": 3.8217175006866455,
+ "eval_runtime": 19.8681,
+ "eval_samples_per_second": 122.86,
+ "eval_steps_per_second": 15.402,
+ "step": 3888
+ },
+ {
+ "epoch": 54.027965916539216,
+ "grad_norm": 0.22240453958511353,
+ "learning_rate": 0.0006,
+ "loss": 4.7125396728515625,
+ "step": 3890
+ },
+ {
+ "epoch": 54.167795499235304,
+ "grad_norm": 0.1980697363615036,
+ "learning_rate": 0.0006,
+ "loss": 4.697897338867188,
+ "step": 3900
+ },
+ {
+ "epoch": 54.3076250819314,
+ "grad_norm": 0.18887707591056824,
+ "learning_rate": 0.0006,
+ "loss": 4.689971923828125,
+ "step": 3910
+ },
+ {
+ "epoch": 54.447454664627486,
+ "grad_norm": 0.19104623794555664,
+ "learning_rate": 0.0006,
+ "loss": 4.698477554321289,
+ "step": 3920
+ },
+ {
+ "epoch": 54.58728424732357,
+ "grad_norm": 0.20511414110660553,
+ "learning_rate": 0.0006,
+ "loss": 4.706096649169922,
+ "step": 3930
+ },
+ {
+ "epoch": 54.72711383001966,
+ "grad_norm": 0.19156834483146667,
+ "learning_rate": 0.0006,
+ "loss": 4.712075424194336,
+ "step": 3940
+ },
+ {
+ "epoch": 54.866943412715756,
+ "grad_norm": 0.2024255096912384,
+ "learning_rate": 0.0006,
+ "loss": 4.717763137817383,
+ "step": 3950
+ },
+ {
+ "epoch": 55.0,
+ "grad_norm": 0.2041298747062683,
+ "learning_rate": 0.0006,
+ "loss": 4.707680511474609,
+ "step": 3960
+ },
+ {
+ "epoch": 55.0,
+ "eval_loss": 3.8258237838745117,
+ "eval_runtime": 19.8723,
+ "eval_samples_per_second": 122.834,
+ "eval_steps_per_second": 15.398,
+ "step": 3960
+ },
+ {
+ "epoch": 55.13982958269609,
+ "grad_norm": 0.18334828317165375,
+ "learning_rate": 0.0006,
+ "loss": 4.646780395507813,
+ "step": 3970
+ },
+ {
+ "epoch": 55.279659165392175,
+ "grad_norm": 0.20664778351783752,
+ "learning_rate": 0.0006,
+ "loss": 4.69473876953125,
+ "step": 3980
+ },
+ {
+ "epoch": 55.41948874808827,
+ "grad_norm": 0.2237965166568756,
+ "learning_rate": 0.0006,
+ "loss": 4.673326873779297,
+ "step": 3990
+ },
+ {
+ "epoch": 55.55931833078436,
+ "grad_norm": 0.19321177899837494,
+ "learning_rate": 0.0006,
+ "loss": 4.707321548461914,
+ "step": 4000
+ },
+ {
+ "epoch": 55.699147913480445,
+ "grad_norm": 0.20430657267570496,
+ "learning_rate": 0.0006,
+ "loss": 4.734820175170898,
+ "step": 4010
+ },
+ {
+ "epoch": 55.83897749617653,
+ "grad_norm": 0.1818908452987671,
+ "learning_rate": 0.0006,
+ "loss": 4.666254043579102,
+ "step": 4020
+ },
+ {
+ "epoch": 55.97880707887263,
+ "grad_norm": 0.21304383873939514,
+ "learning_rate": 0.0006,
+ "loss": 4.722911834716797,
+ "step": 4030
+ },
+ {
+ "epoch": 56.0,
+ "eval_loss": 3.8244595527648926,
+ "eval_runtime": 19.8899,
+ "eval_samples_per_second": 122.726,
+ "eval_steps_per_second": 15.385,
+ "step": 4032
+ },
+ {
+ "epoch": 56.11186366615687,
+ "grad_norm": 0.22179262340068817,
+ "learning_rate": 0.0006,
+ "loss": 4.656169509887695,
+ "step": 4040
+ },
+ {
+ "epoch": 56.25169324885296,
+ "grad_norm": 0.20124323666095734,
+ "learning_rate": 0.0006,
+ "loss": 4.683835220336914,
+ "step": 4050
+ },
+ {
+ "epoch": 56.39152283154905,
+ "grad_norm": 0.17302779853343964,
+ "learning_rate": 0.0006,
+ "loss": 4.712786102294922,
+ "step": 4060
+ },
+ {
+ "epoch": 56.53135241424514,
+ "grad_norm": 0.19677670300006866,
+ "learning_rate": 0.0006,
+ "loss": 4.688117599487304,
+ "step": 4070
+ },
+ {
+ "epoch": 56.67118199694123,
+ "grad_norm": 0.16738387942314148,
+ "learning_rate": 0.0006,
+ "loss": 4.665131759643555,
+ "step": 4080
+ },
+ {
+ "epoch": 56.81101157963732,
+ "grad_norm": 0.1883835643529892,
+ "learning_rate": 0.0006,
+ "loss": 4.688896942138672,
+ "step": 4090
+ },
+ {
+ "epoch": 56.950841162333404,
+ "grad_norm": 0.19651229679584503,
+ "learning_rate": 0.0006,
+ "loss": 4.700971984863282,
+ "step": 4100
+ },
+ {
+ "epoch": 57.0,
+ "eval_loss": 3.8181145191192627,
+ "eval_runtime": 19.8811,
+ "eval_samples_per_second": 122.78,
+ "eval_steps_per_second": 15.392,
+ "step": 4104
+ },
+ {
+ "epoch": 57.083897749617655,
+ "grad_norm": 0.22105193138122559,
+ "learning_rate": 0.0006,
+ "loss": 4.6802925109863285,
+ "step": 4110
+ },
+ {
+ "epoch": 57.22372733231374,
+ "grad_norm": 0.21743889153003693,
+ "learning_rate": 0.0006,
+ "loss": 4.648151397705078,
+ "step": 4120
+ },
+ {
+ "epoch": 57.36355691500983,
+ "grad_norm": 0.17246513068675995,
+ "learning_rate": 0.0006,
+ "loss": 4.663088607788086,
+ "step": 4130
+ },
+ {
+ "epoch": 57.50338649770592,
+ "grad_norm": 0.19539178907871246,
+ "learning_rate": 0.0006,
+ "loss": 4.687311172485352,
+ "step": 4140
+ },
+ {
+ "epoch": 57.64321608040201,
+ "grad_norm": 0.2126225382089615,
+ "learning_rate": 0.0006,
+ "loss": 4.684661483764648,
+ "step": 4150
+ },
+ {
+ "epoch": 57.7830456630981,
+ "grad_norm": 0.19517719745635986,
+ "learning_rate": 0.0006,
+ "loss": 4.689257431030273,
+ "step": 4160
+ },
+ {
+ "epoch": 57.92287524579419,
+ "grad_norm": 0.18389222025871277,
+ "learning_rate": 0.0006,
+ "loss": 4.6811878204345705,
+ "step": 4170
+ },
+ {
+ "epoch": 58.0,
+ "eval_loss": 3.8137402534484863,
+ "eval_runtime": 19.8856,
+ "eval_samples_per_second": 122.752,
+ "eval_steps_per_second": 15.388,
+ "step": 4176
+ },
+ {
+ "epoch": 58.05593183307843,
+ "grad_norm": 0.2017996907234192,
+ "learning_rate": 0.0006,
+ "loss": 4.672654342651367,
+ "step": 4180
+ },
+ {
+ "epoch": 58.19576141577453,
+ "grad_norm": 0.20609158277511597,
+ "learning_rate": 0.0006,
+ "loss": 4.67772331237793,
+ "step": 4190
+ },
+ {
+ "epoch": 58.335590998470614,
+ "grad_norm": 0.18405425548553467,
+ "learning_rate": 0.0006,
+ "loss": 4.650214767456054,
+ "step": 4200
+ },
+ {
+ "epoch": 58.4754205811667,
+ "grad_norm": 0.19392850995063782,
+ "learning_rate": 0.0006,
+ "loss": 4.6834980010986325,
+ "step": 4210
+ },
+ {
+ "epoch": 58.61525016386279,
+ "grad_norm": 0.18864279985427856,
+ "learning_rate": 0.0006,
+ "loss": 4.693776321411133,
+ "step": 4220
+ },
+ {
+ "epoch": 58.755079746558884,
+ "grad_norm": 0.19657056033611298,
+ "learning_rate": 0.0006,
+ "loss": 4.6955207824707035,
+ "step": 4230
+ },
+ {
+ "epoch": 58.89490932925497,
+ "grad_norm": 0.179567351937294,
+ "learning_rate": 0.0006,
+ "loss": 4.664589309692383,
+ "step": 4240
+ },
+ {
+ "epoch": 59.0,
+ "eval_loss": 3.8216488361358643,
+ "eval_runtime": 19.8856,
+ "eval_samples_per_second": 122.752,
+ "eval_steps_per_second": 15.388,
+ "step": 4248
+ },
+ {
+ "epoch": 59.027965916539216,
+ "grad_norm": 0.21359020471572876,
+ "learning_rate": 0.0006,
+ "loss": 4.707389068603516,
+ "step": 4250
+ },
+ {
+ "epoch": 59.167795499235304,
+ "grad_norm": 0.21011729538440704,
+ "learning_rate": 0.0006,
+ "loss": 4.649345016479492,
+ "step": 4260
+ },
+ {
+ "epoch": 59.3076250819314,
+ "grad_norm": 0.20205551385879517,
+ "learning_rate": 0.0006,
+ "loss": 4.656140899658203,
+ "step": 4270
+ },
+ {
+ "epoch": 59.447454664627486,
+ "grad_norm": 0.19951649010181427,
+ "learning_rate": 0.0006,
+ "loss": 4.661082077026367,
+ "step": 4280
+ },
+ {
+ "epoch": 59.58728424732357,
+ "grad_norm": 0.2001890242099762,
+ "learning_rate": 0.0006,
+ "loss": 4.677437591552734,
+ "step": 4290
+ },
+ {
+ "epoch": 59.72711383001966,
+ "grad_norm": 0.20710663497447968,
+ "learning_rate": 0.0006,
+ "loss": 4.702688980102539,
+ "step": 4300
+ },
+ {
+ "epoch": 59.866943412715756,
+ "grad_norm": 0.18877770006656647,
+ "learning_rate": 0.0006,
+ "loss": 4.703515625,
+ "step": 4310
+ },
+ {
+ "epoch": 60.0,
+ "grad_norm": 0.23295427858829498,
+ "learning_rate": 0.0006,
+ "loss": 4.694321060180664,
+ "step": 4320
+ },
+ {
+ "epoch": 60.0,
+ "eval_loss": 3.8161001205444336,
+ "eval_runtime": 19.8799,
+ "eval_samples_per_second": 122.788,
+ "eval_steps_per_second": 15.392,
+ "step": 4320
+ },
+ {
+ "epoch": 60.13982958269609,
+ "grad_norm": 0.19182275235652924,
+ "learning_rate": 0.0006,
+ "loss": 4.6506603240966795,
+ "step": 4330
+ },
+ {
+ "epoch": 60.279659165392175,
+ "grad_norm": 0.1830444633960724,
+ "learning_rate": 0.0006,
+ "loss": 4.65746955871582,
+ "step": 4340
+ },
+ {
+ "epoch": 60.41948874808827,
+ "grad_norm": 0.1964675635099411,
+ "learning_rate": 0.0006,
+ "loss": 4.667615509033203,
+ "step": 4350
+ },
+ {
+ "epoch": 60.55931833078436,
+ "grad_norm": 0.1889505535364151,
+ "learning_rate": 0.0006,
+ "loss": 4.664949035644531,
+ "step": 4360
+ },
+ {
+ "epoch": 60.699147913480445,
+ "grad_norm": 0.18944698572158813,
+ "learning_rate": 0.0006,
+ "loss": 4.681035995483398,
+ "step": 4370
+ },
+ {
+ "epoch": 60.83897749617653,
+ "grad_norm": 0.18462738394737244,
+ "learning_rate": 0.0006,
+ "loss": 4.681563949584961,
+ "step": 4380
+ },
+ {
+ "epoch": 60.97880707887263,
+ "grad_norm": 0.20614218711853027,
+ "learning_rate": 0.0006,
+ "loss": 4.69165153503418,
+ "step": 4390
+ },
+ {
+ "epoch": 61.0,
+ "eval_loss": 3.815230369567871,
+ "eval_runtime": 19.8724,
+ "eval_samples_per_second": 122.834,
+ "eval_steps_per_second": 15.398,
+ "step": 4392
+ },
+ {
+ "epoch": 61.11186366615687,
+ "grad_norm": 0.21028798818588257,
+ "learning_rate": 0.0006,
+ "loss": 4.659085083007812,
+ "step": 4400
+ },
+ {
+ "epoch": 61.25169324885296,
+ "grad_norm": 0.18306267261505127,
+ "learning_rate": 0.0006,
+ "loss": 4.658502578735352,
+ "step": 4410
+ },
+ {
+ "epoch": 61.39152283154905,
+ "grad_norm": 0.191131591796875,
+ "learning_rate": 0.0006,
+ "loss": 4.671865463256836,
+ "step": 4420
+ },
+ {
+ "epoch": 61.53135241424514,
+ "grad_norm": 0.18265610933303833,
+ "learning_rate": 0.0006,
+ "loss": 4.671598815917969,
+ "step": 4430
+ },
+ {
+ "epoch": 61.67118199694123,
+ "grad_norm": 0.1852072924375534,
+ "learning_rate": 0.0006,
+ "loss": 4.6644245147705075,
+ "step": 4440
+ },
+ {
+ "epoch": 61.81101157963732,
+ "grad_norm": 0.1862410306930542,
+ "learning_rate": 0.0006,
+ "loss": 4.670680999755859,
+ "step": 4450
+ },
+ {
+ "epoch": 61.950841162333404,
+ "grad_norm": 0.19163239002227783,
+ "learning_rate": 0.0006,
+ "loss": 4.695830917358398,
+ "step": 4460
+ },
+ {
+ "epoch": 62.0,
+ "eval_loss": 3.8148200511932373,
+ "eval_runtime": 19.8836,
+ "eval_samples_per_second": 122.765,
+ "eval_steps_per_second": 15.39,
+ "step": 4464
+ },
+ {
+ "epoch": 62.083897749617655,
+ "grad_norm": 0.21463270485401154,
+ "learning_rate": 0.0006,
+ "loss": 4.657954406738281,
+ "step": 4470
+ },
+ {
+ "epoch": 62.22372733231374,
+ "grad_norm": 0.1950284242630005,
+ "learning_rate": 0.0006,
+ "loss": 4.6307228088378904,
+ "step": 4480
+ },
+ {
+ "epoch": 62.36355691500983,
+ "grad_norm": 0.19737891852855682,
+ "learning_rate": 0.0006,
+ "loss": 4.670150375366211,
+ "step": 4490
+ },
+ {
+ "epoch": 62.50338649770592,
+ "grad_norm": 0.1774602234363556,
+ "learning_rate": 0.0006,
+ "loss": 4.663933944702149,
+ "step": 4500
+ },
+ {
+ "epoch": 62.64321608040201,
+ "grad_norm": 0.18456695973873138,
+ "learning_rate": 0.0006,
+ "loss": 4.654977416992187,
+ "step": 4510
+ },
+ {
+ "epoch": 62.7830456630981,
+ "grad_norm": 0.1863299310207367,
+ "learning_rate": 0.0006,
+ "loss": 4.671571731567383,
+ "step": 4520
+ },
+ {
+ "epoch": 62.92287524579419,
+ "grad_norm": 0.21943876147270203,
+ "learning_rate": 0.0006,
+ "loss": 4.669335174560547,
+ "step": 4530
+ },
+ {
+ "epoch": 63.0,
+ "eval_loss": 3.812978506088257,
+ "eval_runtime": 19.8864,
+ "eval_samples_per_second": 122.747,
+ "eval_steps_per_second": 15.387,
+ "step": 4536
+ },
+ {
+ "epoch": 63.05593183307843,
+ "grad_norm": 0.20031523704528809,
+ "learning_rate": 0.0006,
+ "loss": 4.659534072875976,
+ "step": 4540
+ },
+ {
+ "epoch": 63.19576141577453,
+ "grad_norm": 0.1890549212694168,
+ "learning_rate": 0.0006,
+ "loss": 4.6399391174316404,
+ "step": 4550
+ },
+ {
+ "epoch": 63.335590998470614,
+ "grad_norm": 0.18944664299488068,
+ "learning_rate": 0.0006,
+ "loss": 4.6267646789550785,
+ "step": 4560
+ },
+ {
+ "epoch": 63.4754205811667,
+ "grad_norm": 0.2193065732717514,
+ "learning_rate": 0.0006,
+ "loss": 4.658189392089843,
+ "step": 4570
+ },
+ {
+ "epoch": 63.61525016386279,
+ "grad_norm": 0.19303832948207855,
+ "learning_rate": 0.0006,
+ "loss": 4.6728271484375,
+ "step": 4580
+ },
+ {
+ "epoch": 63.755079746558884,
+ "grad_norm": 0.19138774275779724,
+ "learning_rate": 0.0006,
+ "loss": 4.662395095825195,
+ "step": 4590
+ },
+ {
+ "epoch": 63.89490932925497,
+ "grad_norm": 0.19790250062942505,
+ "learning_rate": 0.0006,
+ "loss": 4.641337966918945,
+ "step": 4600
+ },
+ {
+ "epoch": 64.0,
+ "eval_loss": 3.8099043369293213,
+ "eval_runtime": 19.8644,
+ "eval_samples_per_second": 122.883,
+ "eval_steps_per_second": 15.404,
+ "step": 4608
+ },
+ {
+ "epoch": 64.02796591653922,
+ "grad_norm": 0.22475092113018036,
+ "learning_rate": 0.0006,
+ "loss": 4.633626937866211,
+ "step": 4610
+ },
+ {
+ "epoch": 64.16779549923531,
+ "grad_norm": 0.1930069625377655,
+ "learning_rate": 0.0006,
+ "loss": 4.663767623901367,
+ "step": 4620
+ },
+ {
+ "epoch": 64.30762508193139,
+ "grad_norm": 0.17936798930168152,
+ "learning_rate": 0.0006,
+ "loss": 4.649365615844727,
+ "step": 4630
+ },
+ {
+ "epoch": 64.44745466462749,
+ "grad_norm": 0.2004196047782898,
+ "learning_rate": 0.0006,
+ "loss": 4.629872512817383,
+ "step": 4640
+ },
+ {
+ "epoch": 64.58728424732358,
+ "grad_norm": 0.2259526401758194,
+ "learning_rate": 0.0006,
+ "loss": 4.670361709594727,
+ "step": 4650
+ },
+ {
+ "epoch": 64.72711383001966,
+ "grad_norm": 0.19929975271224976,
+ "learning_rate": 0.0006,
+ "loss": 4.661606216430664,
+ "step": 4660
+ },
+ {
+ "epoch": 64.86694341271576,
+ "grad_norm": 0.18241754174232483,
+ "learning_rate": 0.0006,
+ "loss": 4.644285583496094,
+ "step": 4670
+ },
+ {
+ "epoch": 65.0,
+ "grad_norm": 0.22225119173526764,
+ "learning_rate": 0.0006,
+ "loss": 4.663331985473633,
+ "step": 4680
+ },
+ {
+ "epoch": 65.0,
+ "eval_loss": 3.8096163272857666,
+ "eval_runtime": 19.8783,
+ "eval_samples_per_second": 122.798,
+ "eval_steps_per_second": 15.394,
+ "step": 4680
+ },
+ {
+ "epoch": 65.1398295826961,
+ "grad_norm": 0.20077547430992126,
+ "learning_rate": 0.0006,
+ "loss": 4.657568740844726,
+ "step": 4690
+ },
+ {
+ "epoch": 65.27965916539218,
+ "grad_norm": 0.19460268318653107,
+ "learning_rate": 0.0006,
+ "loss": 4.6413829803466795,
+ "step": 4700
+ },
+ {
+ "epoch": 65.41948874808827,
+ "grad_norm": 0.2003551870584488,
+ "learning_rate": 0.0006,
+ "loss": 4.621867752075195,
+ "step": 4710
+ },
+ {
+ "epoch": 65.55931833078435,
+ "grad_norm": 0.1902441382408142,
+ "learning_rate": 0.0006,
+ "loss": 4.669316101074219,
+ "step": 4720
+ },
+ {
+ "epoch": 65.69914791348045,
+ "grad_norm": 0.18279673159122467,
+ "learning_rate": 0.0006,
+ "loss": 4.671043014526367,
+ "step": 4730
+ },
+ {
+ "epoch": 65.83897749617654,
+ "grad_norm": 0.18528275191783905,
+ "learning_rate": 0.0006,
+ "loss": 4.65008773803711,
+ "step": 4740
+ },
+ {
+ "epoch": 65.97880707887262,
+ "grad_norm": 0.17923274636268616,
+ "learning_rate": 0.0006,
+ "loss": 4.651177215576172,
+ "step": 4750
+ },
+ {
+ "epoch": 66.0,
+ "eval_loss": 3.8098883628845215,
+ "eval_runtime": 19.876,
+ "eval_samples_per_second": 122.811,
+ "eval_steps_per_second": 15.395,
+ "step": 4752
+ },
+ {
+ "epoch": 66.11186366615686,
+ "grad_norm": 0.2216683030128479,
+ "learning_rate": 0.0006,
+ "loss": 4.640898513793945,
+ "step": 4760
+ },
+ {
+ "epoch": 66.25169324885296,
+ "grad_norm": 0.21217995882034302,
+ "learning_rate": 0.0006,
+ "loss": 4.631869506835938,
+ "step": 4770
+ },
+ {
+ "epoch": 66.39152283154905,
+ "grad_norm": 0.20881089568138123,
+ "learning_rate": 0.0006,
+ "loss": 4.621099472045898,
+ "step": 4780
+ },
+ {
+ "epoch": 66.53135241424513,
+ "grad_norm": 0.19826291501522064,
+ "learning_rate": 0.0006,
+ "loss": 4.640815734863281,
+ "step": 4790
+ },
+ {
+ "epoch": 66.67118199694123,
+ "grad_norm": 0.19001665711402893,
+ "learning_rate": 0.0006,
+ "loss": 4.655009078979492,
+ "step": 4800
+ },
+ {
+ "epoch": 66.81101157963732,
+ "grad_norm": 0.19571472704410553,
+ "learning_rate": 0.0006,
+ "loss": 4.662299728393554,
+ "step": 4810
+ },
+ {
+ "epoch": 66.9508411623334,
+ "grad_norm": 0.1844869703054428,
+ "learning_rate": 0.0006,
+ "loss": 4.678953170776367,
+ "step": 4820
+ },
+ {
+ "epoch": 67.0,
+ "eval_loss": 3.8090898990631104,
+ "eval_runtime": 19.8868,
+ "eval_samples_per_second": 122.745,
+ "eval_steps_per_second": 15.387,
+ "step": 4824
+ },
+ {
+ "epoch": 67.08389774961765,
+ "grad_norm": 0.2028668075799942,
+ "learning_rate": 0.0006,
+ "loss": 4.654681777954101,
+ "step": 4830
+ },
+ {
+ "epoch": 67.22372733231374,
+ "grad_norm": 0.24819622933864594,
+ "learning_rate": 0.0006,
+ "loss": 4.612421035766602,
+ "step": 4840
+ },
+ {
+ "epoch": 67.36355691500984,
+ "grad_norm": 0.18000318109989166,
+ "learning_rate": 0.0006,
+ "loss": 4.627009582519531,
+ "step": 4850
+ },
+ {
+ "epoch": 67.50338649770592,
+ "grad_norm": 0.19026058912277222,
+ "learning_rate": 0.0006,
+ "loss": 4.612002563476563,
+ "step": 4860
+ },
+ {
+ "epoch": 67.64321608040201,
+ "grad_norm": 0.20325525104999542,
+ "learning_rate": 0.0006,
+ "loss": 4.6504364013671875,
+ "step": 4870
+ },
+ {
+ "epoch": 67.7830456630981,
+ "grad_norm": 0.18711955845355988,
+ "learning_rate": 0.0006,
+ "loss": 4.665063095092774,
+ "step": 4880
+ },
+ {
+ "epoch": 67.92287524579419,
+ "grad_norm": 0.20104920864105225,
+ "learning_rate": 0.0006,
+ "loss": 4.657225036621094,
+ "step": 4890
+ },
+ {
+ "epoch": 68.0,
+ "eval_loss": 3.8145639896392822,
+ "eval_runtime": 19.8861,
+ "eval_samples_per_second": 122.749,
+ "eval_steps_per_second": 15.388,
+ "step": 4896
+ },
+ {
+ "epoch": 68.05593183307843,
+ "grad_norm": 0.19064010679721832,
+ "learning_rate": 0.0006,
+ "loss": 4.635558319091797,
+ "step": 4900
+ },
+ {
+ "epoch": 68.19576141577453,
+ "grad_norm": 0.19875690340995789,
+ "learning_rate": 0.0006,
+ "loss": 4.620716094970703,
+ "step": 4910
+ },
+ {
+ "epoch": 68.33559099847061,
+ "grad_norm": 0.1913611888885498,
+ "learning_rate": 0.0006,
+ "loss": 4.617000961303711,
+ "step": 4920
+ },
+ {
+ "epoch": 68.4754205811667,
+ "grad_norm": 0.20077291131019592,
+ "learning_rate": 0.0006,
+ "loss": 4.6278644561767575,
+ "step": 4930
+ },
+ {
+ "epoch": 68.6152501638628,
+ "grad_norm": 0.21102823317050934,
+ "learning_rate": 0.0006,
+ "loss": 4.65388298034668,
+ "step": 4940
+ },
+ {
+ "epoch": 68.75507974655888,
+ "grad_norm": 0.21921777725219727,
+ "learning_rate": 0.0006,
+ "loss": 4.648143005371094,
+ "step": 4950
+ },
+ {
+ "epoch": 68.89490932925497,
+ "grad_norm": 0.18093104660511017,
+ "learning_rate": 0.0006,
+ "loss": 4.673964309692383,
+ "step": 4960
+ },
+ {
+ "epoch": 69.0,
+ "eval_loss": 3.817399263381958,
+ "eval_runtime": 19.8788,
+ "eval_samples_per_second": 122.794,
+ "eval_steps_per_second": 15.393,
+ "step": 4968
+ },
+ {
+ "epoch": 69.02796591653922,
+ "grad_norm": 0.20031331479549408,
+ "learning_rate": 0.0006,
+ "loss": 4.620370483398437,
+ "step": 4970
+ },
+ {
+ "epoch": 69.16779549923531,
+ "grad_norm": 0.2188914716243744,
+ "learning_rate": 0.0006,
+ "loss": 4.598160171508789,
+ "step": 4980
+ },
+ {
+ "epoch": 69.30762508193139,
+ "grad_norm": 0.1853947937488556,
+ "learning_rate": 0.0006,
+ "loss": 4.608010101318359,
+ "step": 4990
+ },
+ {
+ "epoch": 69.44745466462749,
+ "grad_norm": 0.19699349999427795,
+ "learning_rate": 0.0006,
+ "loss": 4.617240142822266,
+ "step": 5000
+ },
+ {
+ "epoch": 69.58728424732358,
+ "grad_norm": 0.19049297273159027,
+ "learning_rate": 0.0006,
+ "loss": 4.6363872528076175,
+ "step": 5010
+ },
+ {
+ "epoch": 69.72711383001966,
+ "grad_norm": 0.2052137702703476,
+ "learning_rate": 0.0006,
+ "loss": 4.626726150512695,
+ "step": 5020
+ },
+ {
+ "epoch": 69.86694341271576,
+ "grad_norm": 0.20372851192951202,
+ "learning_rate": 0.0006,
+ "loss": 4.615378570556641,
+ "step": 5030
+ },
+ {
+ "epoch": 70.0,
+ "grad_norm": 0.2202913463115692,
+ "learning_rate": 0.0006,
+ "loss": 4.652328872680664,
+ "step": 5040
+ },
+ {
+ "epoch": 70.0,
+ "eval_loss": 3.813464879989624,
+ "eval_runtime": 19.8846,
+ "eval_samples_per_second": 122.758,
+ "eval_steps_per_second": 15.389,
+ "step": 5040
+ },
+ {
+ "epoch": 70.1398295826961,
+ "grad_norm": 0.21354041993618011,
+ "learning_rate": 0.0006,
+ "loss": 4.6309162139892575,
+ "step": 5050
+ },
+ {
+ "epoch": 70.27965916539218,
+ "grad_norm": 0.1999189555644989,
+ "learning_rate": 0.0006,
+ "loss": 4.599420547485352,
+ "step": 5060
+ },
+ {
+ "epoch": 70.41948874808827,
+ "grad_norm": 0.18350717425346375,
+ "learning_rate": 0.0006,
+ "loss": 4.634274673461914,
+ "step": 5070
+ },
+ {
+ "epoch": 70.55931833078435,
+ "grad_norm": 0.18830187618732452,
+ "learning_rate": 0.0006,
+ "loss": 4.629424667358398,
+ "step": 5080
+ },
+ {
+ "epoch": 70.69914791348045,
+ "grad_norm": 0.19745728373527527,
+ "learning_rate": 0.0006,
+ "loss": 4.643351745605469,
+ "step": 5090
+ },
+ {
+ "epoch": 70.83897749617654,
+ "grad_norm": 0.19092801213264465,
+ "learning_rate": 0.0006,
+ "loss": 4.625621795654297,
+ "step": 5100
+ },
+ {
+ "epoch": 70.97880707887262,
+ "grad_norm": 0.1901710331439972,
+ "learning_rate": 0.0006,
+ "loss": 4.652716445922851,
+ "step": 5110
+ },
+ {
+ "epoch": 71.0,
+ "eval_loss": 3.8083794116973877,
+ "eval_runtime": 19.877,
+ "eval_samples_per_second": 122.806,
+ "eval_steps_per_second": 15.395,
+ "step": 5112
+ },
+ {
+ "epoch": 71.11186366615686,
+ "grad_norm": 0.22699449956417084,
+ "learning_rate": 0.0006,
+ "loss": 4.626126098632812,
+ "step": 5120
+ },
+ {
+ "epoch": 71.25169324885296,
+ "grad_norm": 0.21119394898414612,
+ "learning_rate": 0.0006,
+ "loss": 4.636773681640625,
+ "step": 5130
+ },
+ {
+ "epoch": 71.39152283154905,
+ "grad_norm": 0.19554030895233154,
+ "learning_rate": 0.0006,
+ "loss": 4.6165626525878904,
+ "step": 5140
+ },
+ {
+ "epoch": 71.53135241424513,
+ "grad_norm": 0.19632764160633087,
+ "learning_rate": 0.0006,
+ "loss": 4.612465667724609,
+ "step": 5150
+ },
+ {
+ "epoch": 71.67118199694123,
+ "grad_norm": 0.21347951889038086,
+ "learning_rate": 0.0006,
+ "loss": 4.597922515869141,
+ "step": 5160
+ },
+ {
+ "epoch": 71.81101157963732,
+ "grad_norm": 0.193961039185524,
+ "learning_rate": 0.0006,
+ "loss": 4.624491119384766,
+ "step": 5170
+ },
+ {
+ "epoch": 71.9508411623334,
+ "grad_norm": 0.18764105439186096,
+ "learning_rate": 0.0006,
+ "loss": 4.640554428100586,
+ "step": 5180
+ },
+ {
+ "epoch": 72.0,
+ "eval_loss": 3.804595470428467,
+ "eval_runtime": 19.8787,
+ "eval_samples_per_second": 122.795,
+ "eval_steps_per_second": 15.393,
+ "step": 5184
+ },
+ {
+ "epoch": 72.08389774961765,
+ "grad_norm": 0.20070268213748932,
+ "learning_rate": 0.0006,
+ "loss": 4.622536849975586,
+ "step": 5190
+ },
+ {
+ "epoch": 72.22372733231374,
+ "grad_norm": 0.1959109902381897,
+ "learning_rate": 0.0006,
+ "loss": 4.623437118530274,
+ "step": 5200
+ },
+ {
+ "epoch": 72.36355691500984,
+ "grad_norm": 0.22956733405590057,
+ "learning_rate": 0.0006,
+ "loss": 4.615106201171875,
+ "step": 5210
+ },
+ {
+ "epoch": 72.50338649770592,
+ "grad_norm": 0.20801733434200287,
+ "learning_rate": 0.0006,
+ "loss": 4.599320983886718,
+ "step": 5220
+ },
+ {
+ "epoch": 72.64321608040201,
+ "grad_norm": 0.1874663382768631,
+ "learning_rate": 0.0006,
+ "loss": 4.620613098144531,
+ "step": 5230
+ },
+ {
+ "epoch": 72.7830456630981,
+ "grad_norm": 0.21099282801151276,
+ "learning_rate": 0.0006,
+ "loss": 4.61894416809082,
+ "step": 5240
+ },
+ {
+ "epoch": 72.92287524579419,
+ "grad_norm": 0.203792005777359,
+ "learning_rate": 0.0006,
+ "loss": 4.637520980834961,
+ "step": 5250
+ },
+ {
+ "epoch": 73.0,
+ "eval_loss": 3.8045051097869873,
+ "eval_runtime": 19.8767,
+ "eval_samples_per_second": 122.807,
+ "eval_steps_per_second": 15.395,
+ "step": 5256
+ },
+ {
+ "epoch": 73.05593183307843,
+ "grad_norm": 0.20452506840229034,
+ "learning_rate": 0.0006,
+ "loss": 4.6207130432128904,
+ "step": 5260
+ },
+ {
+ "epoch": 73.19576141577453,
+ "grad_norm": 0.21968676149845123,
+ "learning_rate": 0.0006,
+ "loss": 4.596218109130859,
+ "step": 5270
+ },
+ {
+ "epoch": 73.33559099847061,
+ "grad_norm": 0.21587903797626495,
+ "learning_rate": 0.0006,
+ "loss": 4.5994819641113285,
+ "step": 5280
+ },
+ {
+ "epoch": 73.4754205811667,
+ "grad_norm": 0.20859280228614807,
+ "learning_rate": 0.0006,
+ "loss": 4.568918991088867,
+ "step": 5290
+ },
+ {
+ "epoch": 73.6152501638628,
+ "grad_norm": 0.23974867165088654,
+ "learning_rate": 0.0006,
+ "loss": 4.59791145324707,
+ "step": 5300
+ },
+ {
+ "epoch": 73.75507974655888,
+ "grad_norm": 0.19860383868217468,
+ "learning_rate": 0.0006,
+ "loss": 4.633654022216797,
+ "step": 5310
+ },
+ {
+ "epoch": 73.89490932925497,
+ "grad_norm": 0.1856987327337265,
+ "learning_rate": 0.0006,
+ "loss": 4.657920455932617,
+ "step": 5320
+ },
+ {
+ "epoch": 74.0,
+ "eval_loss": 3.807504653930664,
+ "eval_runtime": 19.8847,
+ "eval_samples_per_second": 122.757,
+ "eval_steps_per_second": 15.389,
+ "step": 5328
+ },
+ {
+ "epoch": 74.02796591653922,
+ "grad_norm": 0.24725547432899475,
+ "learning_rate": 0.0006,
+ "loss": 4.603820037841797,
+ "step": 5330
+ },
+ {
+ "epoch": 74.16779549923531,
+ "grad_norm": 0.2029736191034317,
+ "learning_rate": 0.0006,
+ "loss": 4.598336410522461,
+ "step": 5340
+ },
+ {
+ "epoch": 74.30762508193139,
+ "grad_norm": 0.2073630392551422,
+ "learning_rate": 0.0006,
+ "loss": 4.6051677703857425,
+ "step": 5350
+ },
+ {
+ "epoch": 74.44745466462749,
+ "grad_norm": 0.19294817745685577,
+ "learning_rate": 0.0006,
+ "loss": 4.598844909667969,
+ "step": 5360
+ },
+ {
+ "epoch": 74.58728424732358,
+ "grad_norm": 0.19145922362804413,
+ "learning_rate": 0.0006,
+ "loss": 4.599289321899414,
+ "step": 5370
+ },
+ {
+ "epoch": 74.72711383001966,
+ "grad_norm": 0.17760322988033295,
+ "learning_rate": 0.0006,
+ "loss": 4.613811874389649,
+ "step": 5380
+ },
+ {
+ "epoch": 74.86694341271576,
+ "grad_norm": 0.18685883283615112,
+ "learning_rate": 0.0006,
+ "loss": 4.608720016479492,
+ "step": 5390
+ },
+ {
+ "epoch": 75.0,
+ "grad_norm": 0.20717309415340424,
+ "learning_rate": 0.0006,
+ "loss": 4.633865737915039,
+ "step": 5400
+ },
+ {
+ "epoch": 75.0,
+ "eval_loss": 3.806734561920166,
+ "eval_runtime": 19.8652,
+ "eval_samples_per_second": 122.878,
+ "eval_steps_per_second": 15.404,
+ "step": 5400
+ },
+ {
+ "epoch": 75.1398295826961,
+ "grad_norm": 0.20117467641830444,
+ "learning_rate": 0.0006,
+ "loss": 4.587604141235351,
+ "step": 5410
+ },
+ {
+ "epoch": 75.27965916539218,
+ "grad_norm": 0.19121626019477844,
+ "learning_rate": 0.0006,
+ "loss": 4.587888717651367,
+ "step": 5420
+ },
+ {
+ "epoch": 75.41948874808827,
+ "grad_norm": 0.20793040096759796,
+ "learning_rate": 0.0006,
+ "loss": 4.605117416381836,
+ "step": 5430
+ },
+ {
+ "epoch": 75.55931833078435,
+ "grad_norm": 0.1943058967590332,
+ "learning_rate": 0.0006,
+ "loss": 4.594827270507812,
+ "step": 5440
+ },
+ {
+ "epoch": 75.69914791348045,
+ "grad_norm": 0.18714813888072968,
+ "learning_rate": 0.0006,
+ "loss": 4.600225448608398,
+ "step": 5450
+ },
+ {
+ "epoch": 75.83897749617654,
+ "grad_norm": 0.22596855461597443,
+ "learning_rate": 0.0006,
+ "loss": 4.624265670776367,
+ "step": 5460
+ },
+ {
+ "epoch": 75.97880707887262,
+ "grad_norm": 0.2115841805934906,
+ "learning_rate": 0.0006,
+ "loss": 4.6443733215332035,
+ "step": 5470
+ },
+ {
+ "epoch": 76.0,
+ "eval_loss": 3.8091273307800293,
+ "eval_runtime": 19.8832,
+ "eval_samples_per_second": 122.767,
+ "eval_steps_per_second": 15.39,
+ "step": 5472
+ },
+ {
+ "epoch": 76.11186366615686,
+ "grad_norm": 0.21845823526382446,
+ "learning_rate": 0.0006,
+ "loss": 4.56871223449707,
+ "step": 5480
+ },
+ {
+ "epoch": 76.25169324885296,
+ "grad_norm": 0.20992623269557953,
+ "learning_rate": 0.0006,
+ "loss": 4.601784133911133,
+ "step": 5490
+ },
+ {
+ "epoch": 76.39152283154905,
+ "grad_norm": 0.1943107545375824,
+ "learning_rate": 0.0006,
+ "loss": 4.602323150634765,
+ "step": 5500
+ },
+ {
+ "epoch": 76.53135241424513,
+ "grad_norm": 0.18950358033180237,
+ "learning_rate": 0.0006,
+ "loss": 4.622146987915039,
+ "step": 5510
+ },
+ {
+ "epoch": 76.67118199694123,
+ "grad_norm": 0.20178720355033875,
+ "learning_rate": 0.0006,
+ "loss": 4.5987194061279295,
+ "step": 5520
+ },
+ {
+ "epoch": 76.81101157963732,
+ "grad_norm": 0.18404407799243927,
+ "learning_rate": 0.0006,
+ "loss": 4.619661331176758,
+ "step": 5530
+ },
+ {
+ "epoch": 76.9508411623334,
+ "grad_norm": 0.2018350511789322,
+ "learning_rate": 0.0006,
+ "loss": 4.581767272949219,
+ "step": 5540
+ },
+ {
+ "epoch": 77.0,
+ "eval_loss": 3.8075482845306396,
+ "eval_runtime": 19.8737,
+ "eval_samples_per_second": 122.826,
+ "eval_steps_per_second": 15.397,
+ "step": 5544
+ },
+ {
+ "epoch": 77.08389774961765,
+ "grad_norm": 0.2422284334897995,
+ "learning_rate": 0.0006,
+ "loss": 4.5938873291015625,
+ "step": 5550
+ },
+ {
+ "epoch": 77.22372733231374,
+ "grad_norm": 0.2221692055463791,
+ "learning_rate": 0.0006,
+ "loss": 4.581957244873047,
+ "step": 5560
+ },
+ {
+ "epoch": 77.36355691500984,
+ "grad_norm": 0.21280072629451752,
+ "learning_rate": 0.0006,
+ "loss": 4.603216171264648,
+ "step": 5570
+ },
+ {
+ "epoch": 77.50338649770592,
+ "grad_norm": 0.23058749735355377,
+ "learning_rate": 0.0006,
+ "loss": 4.597089004516602,
+ "step": 5580
+ },
+ {
+ "epoch": 77.64321608040201,
+ "grad_norm": 0.20739658176898956,
+ "learning_rate": 0.0006,
+ "loss": 4.606765365600586,
+ "step": 5590
+ },
+ {
+ "epoch": 77.7830456630981,
+ "grad_norm": 0.18038851022720337,
+ "learning_rate": 0.0006,
+ "loss": 4.613376235961914,
+ "step": 5600
+ },
+ {
+ "epoch": 77.92287524579419,
+ "grad_norm": 0.1987231820821762,
+ "learning_rate": 0.0006,
+ "loss": 4.606674575805664,
+ "step": 5610
+ },
+ {
+ "epoch": 78.0,
+ "eval_loss": 3.802713632583618,
+ "eval_runtime": 19.8735,
+ "eval_samples_per_second": 122.827,
+ "eval_steps_per_second": 15.397,
+ "step": 5616
+ },
+ {
+ "epoch": 78.05593183307843,
+ "grad_norm": 0.20318147540092468,
+ "learning_rate": 0.0006,
+ "loss": 4.590503311157226,
+ "step": 5620
+ },
+ {
+ "epoch": 78.19576141577453,
+ "grad_norm": 0.20581156015396118,
+ "learning_rate": 0.0006,
+ "loss": 4.5913654327392575,
+ "step": 5630
+ },
+ {
+ "epoch": 78.33559099847061,
+ "grad_norm": 0.20032081007957458,
+ "learning_rate": 0.0006,
+ "loss": 4.584352111816406,
+ "step": 5640
+ },
+ {
+ "epoch": 78.4754205811667,
+ "grad_norm": 0.20647548139095306,
+ "learning_rate": 0.0006,
+ "loss": 4.609855270385742,
+ "step": 5650
+ },
+ {
+ "epoch": 78.6152501638628,
+ "grad_norm": 0.2067854106426239,
+ "learning_rate": 0.0006,
+ "loss": 4.6134075164794925,
+ "step": 5660
+ },
+ {
+ "epoch": 78.75507974655888,
+ "grad_norm": 0.19249257445335388,
+ "learning_rate": 0.0006,
+ "loss": 4.595817947387696,
+ "step": 5670
+ },
+ {
+ "epoch": 78.89490932925497,
+ "grad_norm": 0.1957986205816269,
+ "learning_rate": 0.0006,
+ "loss": 4.625430297851563,
+ "step": 5680
+ },
+ {
+ "epoch": 79.0,
+ "eval_loss": 3.8076088428497314,
+ "eval_runtime": 19.8798,
+ "eval_samples_per_second": 122.788,
+ "eval_steps_per_second": 15.392,
+ "step": 5688
+ },
+ {
+ "epoch": 79.02796591653922,
+ "grad_norm": 0.2209218144416809,
+ "learning_rate": 0.0006,
+ "loss": 4.618154907226563,
+ "step": 5690
+ },
+ {
+ "epoch": 79.16779549923531,
+ "grad_norm": 0.20958910882472992,
+ "learning_rate": 0.0006,
+ "loss": 4.601102828979492,
+ "step": 5700
+ },
+ {
+ "epoch": 79.30762508193139,
+ "grad_norm": 0.18710540235042572,
+ "learning_rate": 0.0006,
+ "loss": 4.589111709594727,
+ "step": 5710
+ },
+ {
+ "epoch": 79.44745466462749,
+ "grad_norm": 0.18651942908763885,
+ "learning_rate": 0.0006,
+ "loss": 4.59680404663086,
+ "step": 5720
+ },
+ {
+ "epoch": 79.58728424732358,
+ "grad_norm": 0.2064020037651062,
+ "learning_rate": 0.0006,
+ "loss": 4.608306121826172,
+ "step": 5730
+ },
+ {
+ "epoch": 79.72711383001966,
+ "grad_norm": 0.17298419773578644,
+ "learning_rate": 0.0006,
+ "loss": 4.57537956237793,
+ "step": 5740
+ },
+ {
+ "epoch": 79.86694341271576,
+ "grad_norm": 0.19675157964229584,
+ "learning_rate": 0.0006,
+ "loss": 4.61697998046875,
+ "step": 5750
+ },
+ {
+ "epoch": 80.0,
+ "grad_norm": 0.21655811369419098,
+ "learning_rate": 0.0006,
+ "loss": 4.6062774658203125,
+ "step": 5760
+ },
+ {
+ "epoch": 80.0,
+ "eval_loss": 3.8018345832824707,
+ "eval_runtime": 19.8716,
+ "eval_samples_per_second": 122.839,
+ "eval_steps_per_second": 15.399,
+ "step": 5760
+ },
+ {
+ "epoch": 80.1398295826961,
+ "grad_norm": 0.21224157512187958,
+ "learning_rate": 0.0006,
+ "loss": 4.573962783813476,
+ "step": 5770
+ },
+ {
+ "epoch": 80.27965916539218,
+ "grad_norm": 0.1942117065191269,
+ "learning_rate": 0.0006,
+ "loss": 4.578574371337891,
+ "step": 5780
+ },
+ {
+ "epoch": 80.41948874808827,
+ "grad_norm": 0.22042931616306305,
+ "learning_rate": 0.0006,
+ "loss": 4.552381896972657,
+ "step": 5790
+ },
+ {
+ "epoch": 80.55931833078435,
+ "grad_norm": 0.22603461146354675,
+ "learning_rate": 0.0006,
+ "loss": 4.618230819702148,
+ "step": 5800
+ },
+ {
+ "epoch": 80.69914791348045,
+ "grad_norm": 0.19256076216697693,
+ "learning_rate": 0.0006,
+ "loss": 4.605718994140625,
+ "step": 5810
+ },
+ {
+ "epoch": 80.83897749617654,
+ "grad_norm": 0.20895510911941528,
+ "learning_rate": 0.0006,
+ "loss": 4.5987907409667965,
+ "step": 5820
+ },
+ {
+ "epoch": 80.97880707887262,
+ "grad_norm": 0.187264546751976,
+ "learning_rate": 0.0006,
+ "loss": 4.609566116333008,
+ "step": 5830
+ },
+ {
+ "epoch": 81.0,
+ "eval_loss": 3.8088114261627197,
+ "eval_runtime": 19.8592,
+ "eval_samples_per_second": 122.915,
+ "eval_steps_per_second": 15.408,
+ "step": 5832
+ },
+ {
+ "epoch": 81.11186366615686,
+ "grad_norm": 0.20549099147319794,
+ "learning_rate": 0.0006,
+ "loss": 4.572750854492187,
+ "step": 5840
+ },
+ {
+ "epoch": 81.25169324885296,
+ "grad_norm": 0.21270322799682617,
+ "learning_rate": 0.0006,
+ "loss": 4.616018676757813,
+ "step": 5850
+ },
+ {
+ "epoch": 81.39152283154905,
+ "grad_norm": 0.2284741997718811,
+ "learning_rate": 0.0006,
+ "loss": 4.584511184692383,
+ "step": 5860
+ },
+ {
+ "epoch": 81.53135241424513,
+ "grad_norm": 0.21854864060878754,
+ "learning_rate": 0.0006,
+ "loss": 4.594888687133789,
+ "step": 5870
+ },
+ {
+ "epoch": 81.67118199694123,
+ "grad_norm": 0.18461118638515472,
+ "learning_rate": 0.0006,
+ "loss": 4.587589263916016,
+ "step": 5880
+ },
+ {
+ "epoch": 81.81101157963732,
+ "grad_norm": 0.195165753364563,
+ "learning_rate": 0.0006,
+ "loss": 4.604648590087891,
+ "step": 5890
+ },
+ {
+ "epoch": 81.9508411623334,
+ "grad_norm": 0.1992371827363968,
+ "learning_rate": 0.0006,
+ "loss": 4.607958221435547,
+ "step": 5900
+ },
+ {
+ "epoch": 82.0,
+ "eval_loss": 3.8060965538024902,
+ "eval_runtime": 19.8812,
+ "eval_samples_per_second": 122.779,
+ "eval_steps_per_second": 15.391,
+ "step": 5904
+ },
+ {
+ "epoch": 82.08389774961765,
+ "grad_norm": 0.20829026401042938,
+ "learning_rate": 0.0006,
+ "loss": 4.577935791015625,
+ "step": 5910
+ },
+ {
+ "epoch": 82.22372733231374,
+ "grad_norm": 0.19801479578018188,
+ "learning_rate": 0.0006,
+ "loss": 4.5797676086425785,
+ "step": 5920
+ },
+ {
+ "epoch": 82.36355691500984,
+ "grad_norm": 0.23395755887031555,
+ "learning_rate": 0.0006,
+ "loss": 4.582308197021485,
+ "step": 5930
+ },
+ {
+ "epoch": 82.50338649770592,
+ "grad_norm": 0.1842985600233078,
+ "learning_rate": 0.0006,
+ "loss": 4.594236373901367,
+ "step": 5940
+ },
+ {
+ "epoch": 82.64321608040201,
+ "grad_norm": 0.2022172063589096,
+ "learning_rate": 0.0006,
+ "loss": 4.601956176757812,
+ "step": 5950
+ },
+ {
+ "epoch": 82.7830456630981,
+ "grad_norm": 0.19542860984802246,
+ "learning_rate": 0.0006,
+ "loss": 4.596880340576172,
+ "step": 5960
+ },
+ {
+ "epoch": 82.92287524579419,
+ "grad_norm": 0.20934593677520752,
+ "learning_rate": 0.0006,
+ "loss": 4.6065715789794925,
+ "step": 5970
+ },
+ {
+ "epoch": 83.0,
+ "eval_loss": 3.805053949356079,
+ "eval_runtime": 19.8795,
+ "eval_samples_per_second": 122.79,
+ "eval_steps_per_second": 15.393,
+ "step": 5976
+ },
+ {
+ "epoch": 83.05593183307843,
+ "grad_norm": 0.22285516560077667,
+ "learning_rate": 0.0006,
+ "loss": 4.58252067565918,
+ "step": 5980
+ },
+ {
+ "epoch": 83.19576141577453,
+ "grad_norm": 0.18773552775382996,
+ "learning_rate": 0.0006,
+ "loss": 4.551812362670899,
+ "step": 5990
+ },
+ {
+ "epoch": 83.33559099847061,
+ "grad_norm": 0.18697050213813782,
+ "learning_rate": 0.0006,
+ "loss": 4.5652423858642575,
+ "step": 6000
+ },
+ {
+ "epoch": 83.4754205811667,
+ "grad_norm": 0.20959897339344025,
+ "learning_rate": 0.0006,
+ "loss": 4.5766242980957035,
+ "step": 6010
+ },
+ {
+ "epoch": 83.6152501638628,
+ "grad_norm": 0.20520903170108795,
+ "learning_rate": 0.0006,
+ "loss": 4.57137451171875,
+ "step": 6020
+ },
+ {
+ "epoch": 83.75507974655888,
+ "grad_norm": 0.18049179017543793,
+ "learning_rate": 0.0006,
+ "loss": 4.590155410766601,
+ "step": 6030
+ },
+ {
+ "epoch": 83.89490932925497,
+ "grad_norm": 0.20023898780345917,
+ "learning_rate": 0.0006,
+ "loss": 4.5983329772949215,
+ "step": 6040
+ },
+ {
+ "epoch": 84.0,
+ "eval_loss": 3.8064918518066406,
+ "eval_runtime": 19.8833,
+ "eval_samples_per_second": 122.766,
+ "eval_steps_per_second": 15.39,
+ "step": 6048
+ },
+ {
+ "epoch": 84.02796591653922,
+ "grad_norm": 0.20501694083213806,
+ "learning_rate": 0.0006,
+ "loss": 4.56171989440918,
+ "step": 6050
+ },
+ {
+ "epoch": 84.16779549923531,
+ "grad_norm": 0.2150523066520691,
+ "learning_rate": 0.0006,
+ "loss": 4.579818344116211,
+ "step": 6060
+ },
+ {
+ "epoch": 84.30762508193139,
+ "grad_norm": 0.19770589470863342,
+ "learning_rate": 0.0006,
+ "loss": 4.567891693115234,
+ "step": 6070
+ },
+ {
+ "epoch": 84.44745466462749,
+ "grad_norm": 0.21325121819972992,
+ "learning_rate": 0.0006,
+ "loss": 4.579122543334961,
+ "step": 6080
+ },
+ {
+ "epoch": 84.58728424732358,
+ "grad_norm": 0.22174964845180511,
+ "learning_rate": 0.0006,
+ "loss": 4.5678447723388675,
+ "step": 6090
+ },
+ {
+ "epoch": 84.72711383001966,
+ "grad_norm": 0.20849718153476715,
+ "learning_rate": 0.0006,
+ "loss": 4.605095672607422,
+ "step": 6100
+ },
+ {
+ "epoch": 84.86694341271576,
+ "grad_norm": 0.19705182313919067,
+ "learning_rate": 0.0006,
+ "loss": 4.613206481933593,
+ "step": 6110
+ },
+ {
+ "epoch": 85.0,
+ "grad_norm": 0.2299671769142151,
+ "learning_rate": 0.0006,
+ "loss": 4.583039474487305,
+ "step": 6120
+ },
+ {
+ "epoch": 85.0,
+ "eval_loss": 3.8035073280334473,
+ "eval_runtime": 19.8771,
+ "eval_samples_per_second": 122.805,
+ "eval_steps_per_second": 15.395,
+ "step": 6120
+ },
+ {
+ "epoch": 85.1398295826961,
+ "grad_norm": 0.22410725057125092,
+ "learning_rate": 0.0006,
+ "loss": 4.540530395507813,
+ "step": 6130
+ },
+ {
+ "epoch": 85.27965916539218,
+ "grad_norm": 0.20432931184768677,
+ "learning_rate": 0.0006,
+ "loss": 4.556621170043945,
+ "step": 6140
+ },
+ {
+ "epoch": 85.41948874808827,
+ "grad_norm": 0.18392537534236908,
+ "learning_rate": 0.0006,
+ "loss": 4.586653137207032,
+ "step": 6150
+ },
+ {
+ "epoch": 85.55931833078435,
+ "grad_norm": 0.19733254611492157,
+ "learning_rate": 0.0006,
+ "loss": 4.5644184112548825,
+ "step": 6160
+ },
+ {
+ "epoch": 85.69914791348045,
+ "grad_norm": 0.19133655726909637,
+ "learning_rate": 0.0006,
+ "loss": 4.585682678222656,
+ "step": 6170
+ },
+ {
+ "epoch": 85.83897749617654,
+ "grad_norm": 0.2014480084180832,
+ "learning_rate": 0.0006,
+ "loss": 4.576129150390625,
+ "step": 6180
+ },
+ {
+ "epoch": 85.97880707887262,
+ "grad_norm": 0.2147936373949051,
+ "learning_rate": 0.0006,
+ "loss": 4.586693572998047,
+ "step": 6190
+ },
+ {
+ "epoch": 86.0,
+ "eval_loss": 3.804001808166504,
+ "eval_runtime": 19.9387,
+ "eval_samples_per_second": 122.425,
+ "eval_steps_per_second": 15.347,
+ "step": 6192
+ },
+ {
+ "epoch": 86.11186366615686,
+ "grad_norm": 0.27694109082221985,
+ "learning_rate": 0.0006,
+ "loss": 4.569868469238282,
+ "step": 6200
+ },
+ {
+ "epoch": 86.25169324885296,
+ "grad_norm": 0.2367578148841858,
+ "learning_rate": 0.0006,
+ "loss": 4.551297378540039,
+ "step": 6210
+ },
+ {
+ "epoch": 86.39152283154905,
+ "grad_norm": 0.2255973517894745,
+ "learning_rate": 0.0006,
+ "loss": 4.557403945922852,
+ "step": 6220
+ },
+ {
+ "epoch": 86.53135241424513,
+ "grad_norm": 0.20496904850006104,
+ "learning_rate": 0.0006,
+ "loss": 4.537801361083984,
+ "step": 6230
+ },
+ {
+ "epoch": 86.67118199694123,
+ "grad_norm": 0.18589381873607635,
+ "learning_rate": 0.0006,
+ "loss": 4.603959274291992,
+ "step": 6240
+ },
+ {
+ "epoch": 86.81101157963732,
+ "grad_norm": 0.1892382651567459,
+ "learning_rate": 0.0006,
+ "loss": 4.570262145996094,
+ "step": 6250
+ },
+ {
+ "epoch": 86.9508411623334,
+ "grad_norm": 0.19356517493724823,
+ "learning_rate": 0.0006,
+ "loss": 4.580677795410156,
+ "step": 6260
+ },
+ {
+ "epoch": 87.0,
+ "eval_loss": 3.811758279800415,
+ "eval_runtime": 19.9042,
+ "eval_samples_per_second": 122.637,
+ "eval_steps_per_second": 15.374,
+ "step": 6264
+ },
+ {
+ "epoch": 87.08389774961765,
+ "grad_norm": 0.2236512154340744,
+ "learning_rate": 0.0006,
+ "loss": 4.557270431518555,
+ "step": 6270
+ },
+ {
+ "epoch": 87.22372733231374,
+ "grad_norm": 0.21406367421150208,
+ "learning_rate": 0.0006,
+ "loss": 4.557614517211914,
+ "step": 6280
+ },
+ {
+ "epoch": 87.36355691500984,
+ "grad_norm": 0.2245820164680481,
+ "learning_rate": 0.0006,
+ "loss": 4.549554443359375,
+ "step": 6290
+ },
+ {
+ "epoch": 87.50338649770592,
+ "grad_norm": 0.1989087164402008,
+ "learning_rate": 0.0006,
+ "loss": 4.563199234008789,
+ "step": 6300
+ },
+ {
+ "epoch": 87.64321608040201,
+ "grad_norm": 0.21840374171733856,
+ "learning_rate": 0.0006,
+ "loss": 4.57634391784668,
+ "step": 6310
+ },
+ {
+ "epoch": 87.7830456630981,
+ "grad_norm": 0.22034838795661926,
+ "learning_rate": 0.0006,
+ "loss": 4.595157623291016,
+ "step": 6320
+ },
+ {
+ "epoch": 87.92287524579419,
+ "grad_norm": 0.19467967748641968,
+ "learning_rate": 0.0006,
+ "loss": 4.5950569152832035,
+ "step": 6330
+ },
+ {
+ "epoch": 88.0,
+ "eval_loss": 3.810549259185791,
+ "eval_runtime": 19.906,
+ "eval_samples_per_second": 122.626,
+ "eval_steps_per_second": 15.372,
+ "step": 6336
+ },
+ {
+ "epoch": 88.05593183307843,
+ "grad_norm": 0.19336706399917603,
+ "learning_rate": 0.0006,
+ "loss": 4.5729827880859375,
+ "step": 6340
+ },
+ {
+ "epoch": 88.19576141577453,
+ "grad_norm": 0.22518819570541382,
+ "learning_rate": 0.0006,
+ "loss": 4.551432037353516,
+ "step": 6350
+ },
+ {
+ "epoch": 88.33559099847061,
+ "grad_norm": 0.21392221748828888,
+ "learning_rate": 0.0006,
+ "loss": 4.557023620605468,
+ "step": 6360
+ },
+ {
+ "epoch": 88.4754205811667,
+ "grad_norm": 0.23801741003990173,
+ "learning_rate": 0.0006,
+ "loss": 4.5937965393066404,
+ "step": 6370
+ },
+ {
+ "epoch": 88.6152501638628,
+ "grad_norm": 0.20078833401203156,
+ "learning_rate": 0.0006,
+ "loss": 4.553228378295898,
+ "step": 6380
+ },
+ {
+ "epoch": 88.75507974655888,
+ "grad_norm": 0.18830180168151855,
+ "learning_rate": 0.0006,
+ "loss": 4.590095138549804,
+ "step": 6390
+ },
+ {
+ "epoch": 88.89490932925497,
+ "grad_norm": 0.19501690566539764,
+ "learning_rate": 0.0006,
+ "loss": 4.563954544067383,
+ "step": 6400
+ },
+ {
+ "epoch": 89.0,
+ "eval_loss": 3.8103811740875244,
+ "eval_runtime": 19.9381,
+ "eval_samples_per_second": 122.429,
+ "eval_steps_per_second": 15.348,
+ "step": 6408
+ },
+ {
+ "epoch": 89.02796591653922,
+ "grad_norm": 0.21821734309196472,
+ "learning_rate": 0.0006,
+ "loss": 4.585327529907227,
+ "step": 6410
+ },
+ {
+ "epoch": 89.16779549923531,
+ "grad_norm": 0.2193078249692917,
+ "learning_rate": 0.0006,
+ "loss": 4.549476623535156,
+ "step": 6420
+ },
+ {
+ "epoch": 89.30762508193139,
+ "grad_norm": 0.20491138100624084,
+ "learning_rate": 0.0006,
+ "loss": 4.544483184814453,
+ "step": 6430
+ },
+ {
+ "epoch": 89.44745466462749,
+ "grad_norm": 0.20006248354911804,
+ "learning_rate": 0.0006,
+ "loss": 4.561048889160157,
+ "step": 6440
+ },
+ {
+ "epoch": 89.58728424732358,
+ "grad_norm": 0.21788370609283447,
+ "learning_rate": 0.0006,
+ "loss": 4.578559875488281,
+ "step": 6450
+ },
+ {
+ "epoch": 89.72711383001966,
+ "grad_norm": 0.21937519311904907,
+ "learning_rate": 0.0006,
+ "loss": 4.603849411010742,
+ "step": 6460
+ },
+ {
+ "epoch": 89.86694341271576,
+ "grad_norm": 0.19886308908462524,
+ "learning_rate": 0.0006,
+ "loss": 4.600399780273437,
+ "step": 6470
+ },
+ {
+ "epoch": 90.0,
+ "grad_norm": 0.22430722415447235,
+ "learning_rate": 0.0006,
+ "loss": 4.581440734863281,
+ "step": 6480
+ },
+ {
+ "epoch": 90.0,
+ "eval_loss": 3.804628372192383,
+ "eval_runtime": 19.9046,
+ "eval_samples_per_second": 122.635,
+ "eval_steps_per_second": 15.373,
+ "step": 6480
+ },
+ {
+ "epoch": 90.1398295826961,
+ "grad_norm": 0.22914966940879822,
+ "learning_rate": 0.0006,
+ "loss": 4.543235015869141,
+ "step": 6490
+ },
+ {
+ "epoch": 90.27965916539218,
+ "grad_norm": 0.20187883079051971,
+ "learning_rate": 0.0006,
+ "loss": 4.579413604736328,
+ "step": 6500
+ },
+ {
+ "epoch": 90.41948874808827,
+ "grad_norm": 0.20426665246486664,
+ "learning_rate": 0.0006,
+ "loss": 4.560892105102539,
+ "step": 6510
+ },
+ {
+ "epoch": 90.55931833078435,
+ "grad_norm": 0.21732565760612488,
+ "learning_rate": 0.0006,
+ "loss": 4.562885665893555,
+ "step": 6520
+ },
+ {
+ "epoch": 90.69914791348045,
+ "grad_norm": 0.21286481618881226,
+ "learning_rate": 0.0006,
+ "loss": 4.568899917602539,
+ "step": 6530
+ },
+ {
+ "epoch": 90.83897749617654,
+ "grad_norm": 0.18496796488761902,
+ "learning_rate": 0.0006,
+ "loss": 4.562284088134765,
+ "step": 6540
+ },
+ {
+ "epoch": 90.97880707887262,
+ "grad_norm": 0.18660616874694824,
+ "learning_rate": 0.0006,
+ "loss": 4.550055694580078,
+ "step": 6550
+ },
+ {
+ "epoch": 91.0,
+ "eval_loss": 3.8095858097076416,
+ "eval_runtime": 19.9042,
+ "eval_samples_per_second": 122.638,
+ "eval_steps_per_second": 15.374,
+ "step": 6552
+ },
+ {
+ "epoch": 91.11186366615686,
+ "grad_norm": 0.22705161571502686,
+ "learning_rate": 0.0006,
+ "loss": 4.533221817016601,
+ "step": 6560
+ },
+ {
+ "epoch": 91.25169324885296,
+ "grad_norm": 0.19260120391845703,
+ "learning_rate": 0.0006,
+ "loss": 4.549694061279297,
+ "step": 6570
+ },
+ {
+ "epoch": 91.39152283154905,
+ "grad_norm": 0.18537335097789764,
+ "learning_rate": 0.0006,
+ "loss": 4.539786529541016,
+ "step": 6580
+ },
+ {
+ "epoch": 91.53135241424513,
+ "grad_norm": 0.20670363306999207,
+ "learning_rate": 0.0006,
+ "loss": 4.559244537353516,
+ "step": 6590
+ },
+ {
+ "epoch": 91.67118199694123,
+ "grad_norm": 0.20007802546024323,
+ "learning_rate": 0.0006,
+ "loss": 4.5579994201660154,
+ "step": 6600
+ },
+ {
+ "epoch": 91.81101157963732,
+ "grad_norm": 0.2167370468378067,
+ "learning_rate": 0.0006,
+ "loss": 4.576282501220703,
+ "step": 6610
+ },
+ {
+ "epoch": 91.9508411623334,
+ "grad_norm": 0.24413971602916718,
+ "learning_rate": 0.0006,
+ "loss": 4.582468795776367,
+ "step": 6620
+ },
+ {
+ "epoch": 92.0,
+ "eval_loss": 3.808347702026367,
+ "eval_runtime": 19.9194,
+ "eval_samples_per_second": 122.544,
+ "eval_steps_per_second": 15.362,
+ "step": 6624
+ },
+ {
+ "epoch": 92.08389774961765,
+ "grad_norm": 0.20741702616214752,
+ "learning_rate": 0.0006,
+ "loss": 4.533404541015625,
+ "step": 6630
+ },
+ {
+ "epoch": 92.22372733231374,
+ "grad_norm": 0.1921168565750122,
+ "learning_rate": 0.0006,
+ "loss": 4.53277473449707,
+ "step": 6640
+ },
+ {
+ "epoch": 92.36355691500984,
+ "grad_norm": 0.20250281691551208,
+ "learning_rate": 0.0006,
+ "loss": 4.542144393920898,
+ "step": 6650
+ },
+ {
+ "epoch": 92.50338649770592,
+ "grad_norm": 0.22585487365722656,
+ "learning_rate": 0.0006,
+ "loss": 4.542268753051758,
+ "step": 6660
+ },
+ {
+ "epoch": 92.64321608040201,
+ "grad_norm": 0.21949340403079987,
+ "learning_rate": 0.0006,
+ "loss": 4.576710510253906,
+ "step": 6670
+ },
+ {
+ "epoch": 92.7830456630981,
+ "grad_norm": 0.19863633811473846,
+ "learning_rate": 0.0006,
+ "loss": 4.560410690307617,
+ "step": 6680
+ },
+ {
+ "epoch": 92.92287524579419,
+ "grad_norm": 0.19653229415416718,
+ "learning_rate": 0.0006,
+ "loss": 4.57304573059082,
+ "step": 6690
+ },
+ {
+ "epoch": 93.0,
+ "eval_loss": 3.8125948905944824,
+ "eval_runtime": 19.9001,
+ "eval_samples_per_second": 122.663,
+ "eval_steps_per_second": 15.377,
+ "step": 6696
+ },
+ {
+ "epoch": 93.05593183307843,
+ "grad_norm": 0.20713748037815094,
+ "learning_rate": 0.0006,
+ "loss": 4.572751998901367,
+ "step": 6700
+ },
+ {
+ "epoch": 93.19576141577453,
+ "grad_norm": 0.21599775552749634,
+ "learning_rate": 0.0006,
+ "loss": 4.554477691650391,
+ "step": 6710
+ },
+ {
+ "epoch": 93.33559099847061,
+ "grad_norm": 0.2048768848180771,
+ "learning_rate": 0.0006,
+ "loss": 4.565523529052735,
+ "step": 6720
+ },
+ {
+ "epoch": 93.4754205811667,
+ "grad_norm": 0.20093759894371033,
+ "learning_rate": 0.0006,
+ "loss": 4.554126739501953,
+ "step": 6730
+ },
+ {
+ "epoch": 93.6152501638628,
+ "grad_norm": 0.21041150391101837,
+ "learning_rate": 0.0006,
+ "loss": 4.543162155151367,
+ "step": 6740
+ },
+ {
+ "epoch": 93.75507974655888,
+ "grad_norm": 0.20444220304489136,
+ "learning_rate": 0.0006,
+ "loss": 4.540937423706055,
+ "step": 6750
+ },
+ {
+ "epoch": 93.89490932925497,
+ "grad_norm": 0.22046950459480286,
+ "learning_rate": 0.0006,
+ "loss": 4.563838958740234,
+ "step": 6760
+ },
+ {
+ "epoch": 94.0,
+ "eval_loss": 3.807183027267456,
+ "eval_runtime": 19.9077,
+ "eval_samples_per_second": 122.616,
+ "eval_steps_per_second": 15.371,
+ "step": 6768
+ },
+ {
+ "epoch": 94.02796591653922,
+ "grad_norm": 0.21147844195365906,
+ "learning_rate": 0.0006,
+ "loss": 4.580686950683594,
+ "step": 6770
+ },
+ {
+ "epoch": 94.16779549923531,
+ "grad_norm": 0.19452129304409027,
+ "learning_rate": 0.0006,
+ "loss": 4.533623886108399,
+ "step": 6780
+ },
+ {
+ "epoch": 94.30762508193139,
+ "grad_norm": 0.2030157595872879,
+ "learning_rate": 0.0006,
+ "loss": 4.5371856689453125,
+ "step": 6790
+ },
+ {
+ "epoch": 94.44745466462749,
+ "grad_norm": 0.2147054523229599,
+ "learning_rate": 0.0006,
+ "loss": 4.512752914428711,
+ "step": 6800
+ },
+ {
+ "epoch": 94.58728424732358,
+ "grad_norm": 0.1930948942899704,
+ "learning_rate": 0.0006,
+ "loss": 4.560096359252929,
+ "step": 6810
+ },
+ {
+ "epoch": 94.72711383001966,
+ "grad_norm": 0.1974930465221405,
+ "learning_rate": 0.0006,
+ "loss": 4.550091934204102,
+ "step": 6820
+ },
+ {
+ "epoch": 94.86694341271576,
+ "grad_norm": 0.19735991954803467,
+ "learning_rate": 0.0006,
+ "loss": 4.552406311035156,
+ "step": 6830
+ },
+ {
+ "epoch": 95.0,
+ "grad_norm": 0.236751526594162,
+ "learning_rate": 0.0006,
+ "loss": 4.565310668945313,
+ "step": 6840
+ },
+ {
+ "epoch": 95.0,
+ "eval_loss": 3.8101532459259033,
+ "eval_runtime": 19.9088,
+ "eval_samples_per_second": 122.609,
+ "eval_steps_per_second": 15.37,
+ "step": 6840
+ },
+ {
+ "epoch": 95.1398295826961,
+ "grad_norm": 0.21721166372299194,
+ "learning_rate": 0.0006,
+ "loss": 4.528188705444336,
+ "step": 6850
+ },
+ {
+ "epoch": 95.27965916539218,
+ "grad_norm": 0.20605036616325378,
+ "learning_rate": 0.0006,
+ "loss": 4.550911331176758,
+ "step": 6860
+ },
+ {
+ "epoch": 95.41948874808827,
+ "grad_norm": 0.21444012224674225,
+ "learning_rate": 0.0006,
+ "loss": 4.5492809295654295,
+ "step": 6870
+ },
+ {
+ "epoch": 95.55931833078435,
+ "grad_norm": 0.2271149754524231,
+ "learning_rate": 0.0006,
+ "loss": 4.547186660766601,
+ "step": 6880
+ },
+ {
+ "epoch": 95.69914791348045,
+ "grad_norm": 0.19431202113628387,
+ "learning_rate": 0.0006,
+ "loss": 4.547607421875,
+ "step": 6890
+ },
+ {
+ "epoch": 95.83897749617654,
+ "grad_norm": 0.2061498910188675,
+ "learning_rate": 0.0006,
+ "loss": 4.574040222167969,
+ "step": 6900
+ },
+ {
+ "epoch": 95.97880707887262,
+ "grad_norm": 0.19307735562324524,
+ "learning_rate": 0.0006,
+ "loss": 4.557284545898438,
+ "step": 6910
+ },
+ {
+ "epoch": 96.0,
+ "eval_loss": 3.8081448078155518,
+ "eval_runtime": 19.9109,
+ "eval_samples_per_second": 122.596,
+ "eval_steps_per_second": 15.368,
+ "step": 6912
+ },
+ {
+ "epoch": 96.11186366615686,
+ "grad_norm": 0.20300351083278656,
+ "learning_rate": 0.0006,
+ "loss": 4.521833801269532,
+ "step": 6920
+ },
+ {
+ "epoch": 96.25169324885296,
+ "grad_norm": 0.22221265733242035,
+ "learning_rate": 0.0006,
+ "loss": 4.52725830078125,
+ "step": 6930
+ },
+ {
+ "epoch": 96.39152283154905,
+ "grad_norm": 0.2252635657787323,
+ "learning_rate": 0.0006,
+ "loss": 4.540511703491211,
+ "step": 6940
+ },
+ {
+ "epoch": 96.53135241424513,
+ "grad_norm": 0.20095887780189514,
+ "learning_rate": 0.0006,
+ "loss": 4.539092254638672,
+ "step": 6950
+ },
+ {
+ "epoch": 96.67118199694123,
+ "grad_norm": 0.18414869904518127,
+ "learning_rate": 0.0006,
+ "loss": 4.5429729461669925,
+ "step": 6960
+ },
+ {
+ "epoch": 96.81101157963732,
+ "grad_norm": 0.2045433670282364,
+ "learning_rate": 0.0006,
+ "loss": 4.5543476104736325,
+ "step": 6970
+ },
+ {
+ "epoch": 96.9508411623334,
+ "grad_norm": 0.24516333639621735,
+ "learning_rate": 0.0006,
+ "loss": 4.55430908203125,
+ "step": 6980
+ },
+ {
+ "epoch": 97.0,
+ "eval_loss": 3.8011868000030518,
+ "eval_runtime": 19.9161,
+ "eval_samples_per_second": 122.564,
+ "eval_steps_per_second": 15.364,
+ "step": 6984
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 7200,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 100,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 2.975412087976624e+18,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/run.lock b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/run.lock
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/train.log b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..3937541e0ddd5e6138a7b24e7b86701431b82ee8
--- /dev/null
+++ b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s43/train.log
@@ -0,0 +1,17736 @@
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 100
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 8
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 64
+[transformers] Total optimization steps = 7,200
+[transformers] Number of trainable parameters = 145,836,544
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: setting up run 7qdqed0q
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260917_064532-7qdqed0q
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t1a_S2_C3_rand5_r2l50_i5exp_s43
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/7qdqed0q
+
0%| | 0/7200 [00:00system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/checkpoint-5184/config.json b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/checkpoint-5184/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3474619f10931027097cc05d837296fa4f33a6c
--- /dev/null
+++ b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/checkpoint-5184/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 512,
+ "initializer_range": 0.02,
+ "intermediate_size": 1536,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 4,
+ "num_hidden_layers": 20,
+ "num_key_value_heads": 4,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151676
+}
diff --git a/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/checkpoint-5184/generation_config.json b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/checkpoint-5184/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/checkpoint-5184/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/checkpoint-5184/tokenizer_config.json b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/checkpoint-5184/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d43c97a447f4b25a8c36be5873f7b622935fa90
--- /dev/null
+++ b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/checkpoint-5184/tokenizer_config.json
@@ -0,0 +1,24 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>",
+ "<|next_1_pred|>",
+ "<|next_2_pred|>",
+ "<|next_3_pred|>",
+ "<|next_4_pred|>",
+ "<|next_5_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/checkpoint-5184/trainer_state.json b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/checkpoint-5184/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bca76064c96711cccfa5c630f88eeeb491bd435c
--- /dev/null
+++ b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/checkpoint-5184/trainer_state.json
@@ -0,0 +1,4243 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 72.0,
+ "eval_steps": 72,
+ "global_step": 5184,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013982958269608914,
+ "grad_norm": 2.3550753593444824,
+ "learning_rate": 0.0,
+ "loss": 12.043922424316406,
+ "step": 1
+ },
+ {
+ "epoch": 0.13982958269608914,
+ "grad_norm": 1.6696112155914307,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.727321200900608,
+ "step": 10
+ },
+ {
+ "epoch": 0.2796591653921783,
+ "grad_norm": 1.5979626178741455,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 11.075974273681641,
+ "step": 20
+ },
+ {
+ "epoch": 0.4194887480882674,
+ "grad_norm": 1.5628509521484375,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 10.351950073242188,
+ "step": 30
+ },
+ {
+ "epoch": 0.5593183307843566,
+ "grad_norm": 1.4085773229599,
+ "learning_rate": 0.000234,
+ "loss": 9.378990936279298,
+ "step": 40
+ },
+ {
+ "epoch": 0.6991479134804457,
+ "grad_norm": 0.7446804642677307,
+ "learning_rate": 0.000294,
+ "loss": 8.47821502685547,
+ "step": 50
+ },
+ {
+ "epoch": 0.8389774961765348,
+ "grad_norm": 0.2203596830368042,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 8.003546905517577,
+ "step": 60
+ },
+ {
+ "epoch": 0.978807078872624,
+ "grad_norm": 0.5071847438812256,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 7.9124092102050785,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.52958869934082,
+ "eval_runtime": 43.8843,
+ "eval_samples_per_second": 55.624,
+ "eval_steps_per_second": 6.973,
+ "step": 72
+ },
+ {
+ "epoch": 1.1118636661568713,
+ "grad_norm": 0.22265546023845673,
+ "learning_rate": 0.000474,
+ "loss": 7.845391845703125,
+ "step": 80
+ },
+ {
+ "epoch": 1.2516932488529604,
+ "grad_norm": 0.2484208643436432,
+ "learning_rate": 0.000534,
+ "loss": 7.784606170654297,
+ "step": 90
+ },
+ {
+ "epoch": 1.3915228315490495,
+ "grad_norm": 0.48633983731269836,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 7.758364868164063,
+ "step": 100
+ },
+ {
+ "epoch": 1.5313524142451387,
+ "grad_norm": 0.21603888273239136,
+ "learning_rate": 0.0006,
+ "loss": 7.746398162841797,
+ "step": 110
+ },
+ {
+ "epoch": 1.671181996941228,
+ "grad_norm": 0.2396312803030014,
+ "learning_rate": 0.0006,
+ "loss": 7.712017822265625,
+ "step": 120
+ },
+ {
+ "epoch": 1.811011579637317,
+ "grad_norm": 0.4377139210700989,
+ "learning_rate": 0.0006,
+ "loss": 7.680126190185547,
+ "step": 130
+ },
+ {
+ "epoch": 1.9508411623334063,
+ "grad_norm": 0.6607714891433716,
+ "learning_rate": 0.0006,
+ "loss": 7.6617889404296875,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 7.146584987640381,
+ "eval_runtime": 19.9264,
+ "eval_samples_per_second": 122.501,
+ "eval_steps_per_second": 15.357,
+ "step": 144
+ },
+ {
+ "epoch": 2.0838977496176536,
+ "grad_norm": 0.22717265784740448,
+ "learning_rate": 0.0006,
+ "loss": 7.6038368225097654,
+ "step": 150
+ },
+ {
+ "epoch": 2.2237273323137425,
+ "grad_norm": 0.34783488512039185,
+ "learning_rate": 0.0006,
+ "loss": 7.572261047363281,
+ "step": 160
+ },
+ {
+ "epoch": 2.363556915009832,
+ "grad_norm": 0.4001064598560333,
+ "learning_rate": 0.0006,
+ "loss": 7.553607177734375,
+ "step": 170
+ },
+ {
+ "epoch": 2.503386497705921,
+ "grad_norm": 0.9263410568237305,
+ "learning_rate": 0.0006,
+ "loss": 7.490562438964844,
+ "step": 180
+ },
+ {
+ "epoch": 2.64321608040201,
+ "grad_norm": 0.5998312830924988,
+ "learning_rate": 0.0006,
+ "loss": 7.445381164550781,
+ "step": 190
+ },
+ {
+ "epoch": 2.783045663098099,
+ "grad_norm": 0.27508029341697693,
+ "learning_rate": 0.0006,
+ "loss": 7.334153747558593,
+ "step": 200
+ },
+ {
+ "epoch": 2.9228752457941884,
+ "grad_norm": 0.3955696225166321,
+ "learning_rate": 0.0006,
+ "loss": 7.245298767089844,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 6.503957271575928,
+ "eval_runtime": 19.8721,
+ "eval_samples_per_second": 122.835,
+ "eval_steps_per_second": 15.398,
+ "step": 216
+ },
+ {
+ "epoch": 3.0559318330784357,
+ "grad_norm": 0.42500039935112,
+ "learning_rate": 0.0006,
+ "loss": 7.17149658203125,
+ "step": 220
+ },
+ {
+ "epoch": 3.1957614157745247,
+ "grad_norm": 0.37210938334465027,
+ "learning_rate": 0.0006,
+ "loss": 7.088895416259765,
+ "step": 230
+ },
+ {
+ "epoch": 3.335590998470614,
+ "grad_norm": 0.3639368712902069,
+ "learning_rate": 0.0006,
+ "loss": 7.046165466308594,
+ "step": 240
+ },
+ {
+ "epoch": 3.475420581166703,
+ "grad_norm": 0.7530733346939087,
+ "learning_rate": 0.0006,
+ "loss": 6.96002197265625,
+ "step": 250
+ },
+ {
+ "epoch": 3.6152501638627923,
+ "grad_norm": 0.2216365486383438,
+ "learning_rate": 0.0006,
+ "loss": 6.903421783447266,
+ "step": 260
+ },
+ {
+ "epoch": 3.755079746558881,
+ "grad_norm": 0.32837608456611633,
+ "learning_rate": 0.0006,
+ "loss": 6.85345458984375,
+ "step": 270
+ },
+ {
+ "epoch": 3.8949093292549706,
+ "grad_norm": 0.3560459613800049,
+ "learning_rate": 0.0006,
+ "loss": 6.806476593017578,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 6.022144317626953,
+ "eval_runtime": 19.8809,
+ "eval_samples_per_second": 122.781,
+ "eval_steps_per_second": 15.392,
+ "step": 288
+ },
+ {
+ "epoch": 4.027965916539218,
+ "grad_norm": 0.7235914468765259,
+ "learning_rate": 0.0006,
+ "loss": 6.768614196777344,
+ "step": 290
+ },
+ {
+ "epoch": 4.167795499235307,
+ "grad_norm": 0.28256329894065857,
+ "learning_rate": 0.0006,
+ "loss": 6.743829345703125,
+ "step": 300
+ },
+ {
+ "epoch": 4.307625081931396,
+ "grad_norm": 0.24897515773773193,
+ "learning_rate": 0.0006,
+ "loss": 6.668219757080078,
+ "step": 310
+ },
+ {
+ "epoch": 4.447454664627485,
+ "grad_norm": 0.25434738397598267,
+ "learning_rate": 0.0006,
+ "loss": 6.61641845703125,
+ "step": 320
+ },
+ {
+ "epoch": 4.587284247323574,
+ "grad_norm": 0.3578528165817261,
+ "learning_rate": 0.0006,
+ "loss": 6.5698402404785154,
+ "step": 330
+ },
+ {
+ "epoch": 4.727113830019664,
+ "grad_norm": 0.31078484654426575,
+ "learning_rate": 0.0006,
+ "loss": 6.5328727722167965,
+ "step": 340
+ },
+ {
+ "epoch": 4.866943412715752,
+ "grad_norm": 0.3765343427658081,
+ "learning_rate": 0.0006,
+ "loss": 6.483283996582031,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.3727656900882721,
+ "learning_rate": 0.0006,
+ "loss": 6.4663551330566404,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 5.5900397300720215,
+ "eval_runtime": 19.8801,
+ "eval_samples_per_second": 122.786,
+ "eval_steps_per_second": 15.392,
+ "step": 360
+ },
+ {
+ "epoch": 5.139829582696089,
+ "grad_norm": 0.44575974345207214,
+ "learning_rate": 0.0006,
+ "loss": 6.436042785644531,
+ "step": 370
+ },
+ {
+ "epoch": 5.279659165392179,
+ "grad_norm": 0.24130988121032715,
+ "learning_rate": 0.0006,
+ "loss": 6.3821063995361325,
+ "step": 380
+ },
+ {
+ "epoch": 5.419488748088267,
+ "grad_norm": 0.2145642340183258,
+ "learning_rate": 0.0006,
+ "loss": 6.320829772949219,
+ "step": 390
+ },
+ {
+ "epoch": 5.5593183307843566,
+ "grad_norm": 0.2756289541721344,
+ "learning_rate": 0.0006,
+ "loss": 6.293283843994141,
+ "step": 400
+ },
+ {
+ "epoch": 5.699147913480446,
+ "grad_norm": 0.88419109582901,
+ "learning_rate": 0.0006,
+ "loss": 6.2766155242919925,
+ "step": 410
+ },
+ {
+ "epoch": 5.838977496176534,
+ "grad_norm": 0.6335473656654358,
+ "learning_rate": 0.0006,
+ "loss": 6.376976013183594,
+ "step": 420
+ },
+ {
+ "epoch": 5.978807078872624,
+ "grad_norm": 0.22760677337646484,
+ "learning_rate": 0.0006,
+ "loss": 6.313069534301758,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 5.333829879760742,
+ "eval_runtime": 19.8755,
+ "eval_samples_per_second": 122.815,
+ "eval_steps_per_second": 15.396,
+ "step": 432
+ },
+ {
+ "epoch": 6.1118636661568715,
+ "grad_norm": 0.18680767714977264,
+ "learning_rate": 0.0006,
+ "loss": 6.2059070587158205,
+ "step": 440
+ },
+ {
+ "epoch": 6.251693248852961,
+ "grad_norm": 0.14362162351608276,
+ "learning_rate": 0.0006,
+ "loss": 6.16846809387207,
+ "step": 450
+ },
+ {
+ "epoch": 6.391522831549049,
+ "grad_norm": 0.21500176191329956,
+ "learning_rate": 0.0006,
+ "loss": 6.13460693359375,
+ "step": 460
+ },
+ {
+ "epoch": 6.531352414245139,
+ "grad_norm": 0.23855455219745636,
+ "learning_rate": 0.0006,
+ "loss": 6.095898818969727,
+ "step": 470
+ },
+ {
+ "epoch": 6.671181996941228,
+ "grad_norm": 0.6023666262626648,
+ "learning_rate": 0.0006,
+ "loss": 6.075651550292969,
+ "step": 480
+ },
+ {
+ "epoch": 6.811011579637317,
+ "grad_norm": 0.3553936183452606,
+ "learning_rate": 0.0006,
+ "loss": 6.068154525756836,
+ "step": 490
+ },
+ {
+ "epoch": 6.950841162333406,
+ "grad_norm": 0.2022867351770401,
+ "learning_rate": 0.0006,
+ "loss": 6.049007415771484,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 5.02913236618042,
+ "eval_runtime": 19.877,
+ "eval_samples_per_second": 122.805,
+ "eval_steps_per_second": 15.395,
+ "step": 504
+ },
+ {
+ "epoch": 7.083897749617654,
+ "grad_norm": 0.31047022342681885,
+ "learning_rate": 0.0006,
+ "loss": 6.002838897705078,
+ "step": 510
+ },
+ {
+ "epoch": 7.223727332313743,
+ "grad_norm": 0.7759034037590027,
+ "learning_rate": 0.0006,
+ "loss": 5.993117904663086,
+ "step": 520
+ },
+ {
+ "epoch": 7.3635569150098314,
+ "grad_norm": 0.4120352864265442,
+ "learning_rate": 0.0006,
+ "loss": 6.030142974853516,
+ "step": 530
+ },
+ {
+ "epoch": 7.503386497705921,
+ "grad_norm": 0.20562559366226196,
+ "learning_rate": 0.0006,
+ "loss": 5.970446395874023,
+ "step": 540
+ },
+ {
+ "epoch": 7.64321608040201,
+ "grad_norm": 0.2305469512939453,
+ "learning_rate": 0.0006,
+ "loss": 5.930681991577148,
+ "step": 550
+ },
+ {
+ "epoch": 7.7830456630980995,
+ "grad_norm": 0.20169003307819366,
+ "learning_rate": 0.0006,
+ "loss": 5.888197326660157,
+ "step": 560
+ },
+ {
+ "epoch": 7.922875245794188,
+ "grad_norm": 0.4079500436782837,
+ "learning_rate": 0.0006,
+ "loss": 5.829338455200196,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.814265251159668,
+ "eval_runtime": 19.8764,
+ "eval_samples_per_second": 122.809,
+ "eval_steps_per_second": 15.395,
+ "step": 576
+ },
+ {
+ "epoch": 8.055931833078436,
+ "grad_norm": 0.24522879719734192,
+ "learning_rate": 0.0006,
+ "loss": 5.798547744750977,
+ "step": 580
+ },
+ {
+ "epoch": 8.195761415774525,
+ "grad_norm": 0.3815036416053772,
+ "learning_rate": 0.0006,
+ "loss": 5.778239440917969,
+ "step": 590
+ },
+ {
+ "epoch": 8.335590998470614,
+ "grad_norm": 0.4450983703136444,
+ "learning_rate": 0.0006,
+ "loss": 5.785980987548828,
+ "step": 600
+ },
+ {
+ "epoch": 8.475420581166704,
+ "grad_norm": 0.22726193070411682,
+ "learning_rate": 0.0006,
+ "loss": 5.754364776611328,
+ "step": 610
+ },
+ {
+ "epoch": 8.615250163862791,
+ "grad_norm": 0.25717779994010925,
+ "learning_rate": 0.0006,
+ "loss": 5.735243225097657,
+ "step": 620
+ },
+ {
+ "epoch": 8.75507974655888,
+ "grad_norm": 0.3881776034832001,
+ "learning_rate": 0.0006,
+ "loss": 5.730010604858398,
+ "step": 630
+ },
+ {
+ "epoch": 8.89490932925497,
+ "grad_norm": 0.31444108486175537,
+ "learning_rate": 0.0006,
+ "loss": 5.722562408447265,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.645870685577393,
+ "eval_runtime": 19.8597,
+ "eval_samples_per_second": 122.912,
+ "eval_steps_per_second": 15.408,
+ "step": 648
+ },
+ {
+ "epoch": 9.027965916539218,
+ "grad_norm": 0.29299837350845337,
+ "learning_rate": 0.0006,
+ "loss": 5.696731948852539,
+ "step": 650
+ },
+ {
+ "epoch": 9.167795499235307,
+ "grad_norm": 0.24998565018177032,
+ "learning_rate": 0.0006,
+ "loss": 5.660461807250977,
+ "step": 660
+ },
+ {
+ "epoch": 9.307625081931397,
+ "grad_norm": 0.2535006105899811,
+ "learning_rate": 0.0006,
+ "loss": 5.621578598022461,
+ "step": 670
+ },
+ {
+ "epoch": 9.447454664627486,
+ "grad_norm": 0.3315477669239044,
+ "learning_rate": 0.0006,
+ "loss": 5.6454109191894535,
+ "step": 680
+ },
+ {
+ "epoch": 9.587284247323574,
+ "grad_norm": 0.5049772262573242,
+ "learning_rate": 0.0006,
+ "loss": 5.600779342651367,
+ "step": 690
+ },
+ {
+ "epoch": 9.727113830019663,
+ "grad_norm": 0.3210202753543854,
+ "learning_rate": 0.0006,
+ "loss": 5.617647171020508,
+ "step": 700
+ },
+ {
+ "epoch": 9.866943412715752,
+ "grad_norm": 0.2265830636024475,
+ "learning_rate": 0.0006,
+ "loss": 5.591337203979492,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.22018803656101227,
+ "learning_rate": 0.0006,
+ "loss": 5.584283828735352,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.520809650421143,
+ "eval_runtime": 19.8543,
+ "eval_samples_per_second": 122.945,
+ "eval_steps_per_second": 15.412,
+ "step": 720
+ },
+ {
+ "epoch": 10.13982958269609,
+ "grad_norm": 0.3432421088218689,
+ "learning_rate": 0.0006,
+ "loss": 5.540806579589844,
+ "step": 730
+ },
+ {
+ "epoch": 10.279659165392179,
+ "grad_norm": 0.3163314163684845,
+ "learning_rate": 0.0006,
+ "loss": 5.536584091186524,
+ "step": 740
+ },
+ {
+ "epoch": 10.419488748088268,
+ "grad_norm": 0.3583109974861145,
+ "learning_rate": 0.0006,
+ "loss": 5.539588546752929,
+ "step": 750
+ },
+ {
+ "epoch": 10.559318330784357,
+ "grad_norm": 0.2924076318740845,
+ "learning_rate": 0.0006,
+ "loss": 5.537279510498047,
+ "step": 760
+ },
+ {
+ "epoch": 10.699147913480445,
+ "grad_norm": 0.25934353470802307,
+ "learning_rate": 0.0006,
+ "loss": 5.5250202178955075,
+ "step": 770
+ },
+ {
+ "epoch": 10.838977496176534,
+ "grad_norm": 0.21091929078102112,
+ "learning_rate": 0.0006,
+ "loss": 5.50440559387207,
+ "step": 780
+ },
+ {
+ "epoch": 10.978807078872624,
+ "grad_norm": 0.21586337685585022,
+ "learning_rate": 0.0006,
+ "loss": 5.48821792602539,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.424731254577637,
+ "eval_runtime": 19.858,
+ "eval_samples_per_second": 122.923,
+ "eval_steps_per_second": 15.409,
+ "step": 792
+ },
+ {
+ "epoch": 11.111863666156871,
+ "grad_norm": 0.2433748096227646,
+ "learning_rate": 0.0006,
+ "loss": 5.457383728027343,
+ "step": 800
+ },
+ {
+ "epoch": 11.25169324885296,
+ "grad_norm": 0.34305232763290405,
+ "learning_rate": 0.0006,
+ "loss": 5.457944869995117,
+ "step": 810
+ },
+ {
+ "epoch": 11.39152283154905,
+ "grad_norm": 0.3128548860549927,
+ "learning_rate": 0.0006,
+ "loss": 5.459765625,
+ "step": 820
+ },
+ {
+ "epoch": 11.531352414245138,
+ "grad_norm": 0.3107026517391205,
+ "learning_rate": 0.0006,
+ "loss": 5.450466918945312,
+ "step": 830
+ },
+ {
+ "epoch": 11.671181996941227,
+ "grad_norm": 0.21432644128799438,
+ "learning_rate": 0.0006,
+ "loss": 5.466511917114258,
+ "step": 840
+ },
+ {
+ "epoch": 11.811011579637317,
+ "grad_norm": 0.22154822945594788,
+ "learning_rate": 0.0006,
+ "loss": 5.4184917449951175,
+ "step": 850
+ },
+ {
+ "epoch": 11.950841162333406,
+ "grad_norm": 0.2612152397632599,
+ "learning_rate": 0.0006,
+ "loss": 5.419964599609375,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.35953950881958,
+ "eval_runtime": 19.8608,
+ "eval_samples_per_second": 122.906,
+ "eval_steps_per_second": 15.407,
+ "step": 864
+ },
+ {
+ "epoch": 12.083897749617654,
+ "grad_norm": 0.26321491599082947,
+ "learning_rate": 0.0006,
+ "loss": 5.396856307983398,
+ "step": 870
+ },
+ {
+ "epoch": 12.223727332313743,
+ "grad_norm": 0.4384300112724304,
+ "learning_rate": 0.0006,
+ "loss": 5.394463348388672,
+ "step": 880
+ },
+ {
+ "epoch": 12.363556915009832,
+ "grad_norm": 0.2962022125720978,
+ "learning_rate": 0.0006,
+ "loss": 5.3839069366455075,
+ "step": 890
+ },
+ {
+ "epoch": 12.503386497705922,
+ "grad_norm": 0.21519114077091217,
+ "learning_rate": 0.0006,
+ "loss": 5.380209350585938,
+ "step": 900
+ },
+ {
+ "epoch": 12.64321608040201,
+ "grad_norm": 0.2335522323846817,
+ "learning_rate": 0.0006,
+ "loss": 5.384426498413086,
+ "step": 910
+ },
+ {
+ "epoch": 12.783045663098099,
+ "grad_norm": 0.23483894765377045,
+ "learning_rate": 0.0006,
+ "loss": 5.360023117065429,
+ "step": 920
+ },
+ {
+ "epoch": 12.922875245794188,
+ "grad_norm": 0.26582035422325134,
+ "learning_rate": 0.0006,
+ "loss": 5.34907112121582,
+ "step": 930
+ },
+ {
+ "epoch": 13.0,
+ "eval_loss": 4.292357921600342,
+ "eval_runtime": 19.8807,
+ "eval_samples_per_second": 122.782,
+ "eval_steps_per_second": 15.392,
+ "step": 936
+ },
+ {
+ "epoch": 13.055931833078436,
+ "grad_norm": 0.2583659589290619,
+ "learning_rate": 0.0006,
+ "loss": 5.332267761230469,
+ "step": 940
+ },
+ {
+ "epoch": 13.195761415774525,
+ "grad_norm": 0.29705196619033813,
+ "learning_rate": 0.0006,
+ "loss": 5.325775909423828,
+ "step": 950
+ },
+ {
+ "epoch": 13.335590998470614,
+ "grad_norm": 0.2906164526939392,
+ "learning_rate": 0.0006,
+ "loss": 5.329711151123047,
+ "step": 960
+ },
+ {
+ "epoch": 13.475420581166704,
+ "grad_norm": 0.23155604302883148,
+ "learning_rate": 0.0006,
+ "loss": 5.350077056884766,
+ "step": 970
+ },
+ {
+ "epoch": 13.615250163862791,
+ "grad_norm": 0.2496420294046402,
+ "learning_rate": 0.0006,
+ "loss": 5.308800506591797,
+ "step": 980
+ },
+ {
+ "epoch": 13.75507974655888,
+ "grad_norm": 0.25255584716796875,
+ "learning_rate": 0.0006,
+ "loss": 5.332855224609375,
+ "step": 990
+ },
+ {
+ "epoch": 13.89490932925497,
+ "grad_norm": 0.21292808651924133,
+ "learning_rate": 0.0006,
+ "loss": 5.2939697265625,
+ "step": 1000
+ },
+ {
+ "epoch": 14.0,
+ "eval_loss": 4.240225791931152,
+ "eval_runtime": 19.8838,
+ "eval_samples_per_second": 122.763,
+ "eval_steps_per_second": 15.389,
+ "step": 1008
+ },
+ {
+ "epoch": 14.027965916539218,
+ "grad_norm": 0.23205769062042236,
+ "learning_rate": 0.0006,
+ "loss": 5.296220016479492,
+ "step": 1010
+ },
+ {
+ "epoch": 14.167795499235307,
+ "grad_norm": 0.2593178451061249,
+ "learning_rate": 0.0006,
+ "loss": 5.285311126708985,
+ "step": 1020
+ },
+ {
+ "epoch": 14.307625081931397,
+ "grad_norm": 0.3025290369987488,
+ "learning_rate": 0.0006,
+ "loss": 5.292156600952149,
+ "step": 1030
+ },
+ {
+ "epoch": 14.447454664627486,
+ "grad_norm": 0.2441108077764511,
+ "learning_rate": 0.0006,
+ "loss": 5.2836051940917965,
+ "step": 1040
+ },
+ {
+ "epoch": 14.587284247323574,
+ "grad_norm": 0.2786442041397095,
+ "learning_rate": 0.0006,
+ "loss": 5.251115417480468,
+ "step": 1050
+ },
+ {
+ "epoch": 14.727113830019663,
+ "grad_norm": 0.273461252450943,
+ "learning_rate": 0.0006,
+ "loss": 5.2666667938232425,
+ "step": 1060
+ },
+ {
+ "epoch": 14.866943412715752,
+ "grad_norm": 0.22930866479873657,
+ "learning_rate": 0.0006,
+ "loss": 5.270443344116211,
+ "step": 1070
+ },
+ {
+ "epoch": 15.0,
+ "grad_norm": 0.2404669225215912,
+ "learning_rate": 0.0006,
+ "loss": 5.253658294677734,
+ "step": 1080
+ },
+ {
+ "epoch": 15.0,
+ "eval_loss": 4.196949005126953,
+ "eval_runtime": 19.8762,
+ "eval_samples_per_second": 122.81,
+ "eval_steps_per_second": 15.395,
+ "step": 1080
+ },
+ {
+ "epoch": 15.13982958269609,
+ "grad_norm": 0.2094593048095703,
+ "learning_rate": 0.0006,
+ "loss": 5.258792495727539,
+ "step": 1090
+ },
+ {
+ "epoch": 15.279659165392179,
+ "grad_norm": 0.24499695003032684,
+ "learning_rate": 0.0006,
+ "loss": 5.242315673828125,
+ "step": 1100
+ },
+ {
+ "epoch": 15.419488748088268,
+ "grad_norm": 0.24898479878902435,
+ "learning_rate": 0.0006,
+ "loss": 5.222119522094727,
+ "step": 1110
+ },
+ {
+ "epoch": 15.559318330784357,
+ "grad_norm": 0.26408272981643677,
+ "learning_rate": 0.0006,
+ "loss": 5.231377792358399,
+ "step": 1120
+ },
+ {
+ "epoch": 15.699147913480445,
+ "grad_norm": 0.30738386511802673,
+ "learning_rate": 0.0006,
+ "loss": 5.229532623291016,
+ "step": 1130
+ },
+ {
+ "epoch": 15.838977496176534,
+ "grad_norm": 0.2342987358570099,
+ "learning_rate": 0.0006,
+ "loss": 5.212747955322266,
+ "step": 1140
+ },
+ {
+ "epoch": 15.978807078872624,
+ "grad_norm": 0.23493576049804688,
+ "learning_rate": 0.0006,
+ "loss": 5.212433624267578,
+ "step": 1150
+ },
+ {
+ "epoch": 16.0,
+ "eval_loss": 4.162764072418213,
+ "eval_runtime": 19.8819,
+ "eval_samples_per_second": 122.775,
+ "eval_steps_per_second": 15.391,
+ "step": 1152
+ },
+ {
+ "epoch": 16.11186366615687,
+ "grad_norm": 0.22780407965183258,
+ "learning_rate": 0.0006,
+ "loss": 5.212160873413086,
+ "step": 1160
+ },
+ {
+ "epoch": 16.25169324885296,
+ "grad_norm": 0.2650904953479767,
+ "learning_rate": 0.0006,
+ "loss": 5.174283218383789,
+ "step": 1170
+ },
+ {
+ "epoch": 16.39152283154905,
+ "grad_norm": 0.21618317067623138,
+ "learning_rate": 0.0006,
+ "loss": 5.18211669921875,
+ "step": 1180
+ },
+ {
+ "epoch": 16.531352414245138,
+ "grad_norm": 0.2616349458694458,
+ "learning_rate": 0.0006,
+ "loss": 5.192581558227539,
+ "step": 1190
+ },
+ {
+ "epoch": 16.67118199694123,
+ "grad_norm": 0.21440915763378143,
+ "learning_rate": 0.0006,
+ "loss": 5.194570922851563,
+ "step": 1200
+ },
+ {
+ "epoch": 16.811011579637317,
+ "grad_norm": 0.24742798507213593,
+ "learning_rate": 0.0006,
+ "loss": 5.1946556091308596,
+ "step": 1210
+ },
+ {
+ "epoch": 16.950841162333408,
+ "grad_norm": 0.23486006259918213,
+ "learning_rate": 0.0006,
+ "loss": 5.1789085388183596,
+ "step": 1220
+ },
+ {
+ "epoch": 17.0,
+ "eval_loss": 4.1234331130981445,
+ "eval_runtime": 19.8845,
+ "eval_samples_per_second": 122.759,
+ "eval_steps_per_second": 15.389,
+ "step": 1224
+ },
+ {
+ "epoch": 17.083897749617652,
+ "grad_norm": 0.2517339885234833,
+ "learning_rate": 0.0006,
+ "loss": 5.148545837402343,
+ "step": 1230
+ },
+ {
+ "epoch": 17.223727332313743,
+ "grad_norm": 0.26246583461761475,
+ "learning_rate": 0.0006,
+ "loss": 5.1458251953125,
+ "step": 1240
+ },
+ {
+ "epoch": 17.36355691500983,
+ "grad_norm": 0.2603471279144287,
+ "learning_rate": 0.0006,
+ "loss": 5.151715087890625,
+ "step": 1250
+ },
+ {
+ "epoch": 17.50338649770592,
+ "grad_norm": 0.22735387086868286,
+ "learning_rate": 0.0006,
+ "loss": 5.161113739013672,
+ "step": 1260
+ },
+ {
+ "epoch": 17.64321608040201,
+ "grad_norm": 0.2330193668603897,
+ "learning_rate": 0.0006,
+ "loss": 5.152214050292969,
+ "step": 1270
+ },
+ {
+ "epoch": 17.7830456630981,
+ "grad_norm": 0.24208895862102509,
+ "learning_rate": 0.0006,
+ "loss": 5.157229232788086,
+ "step": 1280
+ },
+ {
+ "epoch": 17.922875245794188,
+ "grad_norm": 0.2033202201128006,
+ "learning_rate": 0.0006,
+ "loss": 5.169243621826172,
+ "step": 1290
+ },
+ {
+ "epoch": 18.0,
+ "eval_loss": 4.096863269805908,
+ "eval_runtime": 19.8835,
+ "eval_samples_per_second": 122.765,
+ "eval_steps_per_second": 15.39,
+ "step": 1296
+ },
+ {
+ "epoch": 18.055931833078436,
+ "grad_norm": 0.267029345035553,
+ "learning_rate": 0.0006,
+ "loss": 5.139271545410156,
+ "step": 1300
+ },
+ {
+ "epoch": 18.195761415774523,
+ "grad_norm": 0.25402936339378357,
+ "learning_rate": 0.0006,
+ "loss": 5.125152587890625,
+ "step": 1310
+ },
+ {
+ "epoch": 18.335590998470614,
+ "grad_norm": 0.26124241948127747,
+ "learning_rate": 0.0006,
+ "loss": 5.141316604614258,
+ "step": 1320
+ },
+ {
+ "epoch": 18.475420581166702,
+ "grad_norm": 0.23416341841220856,
+ "learning_rate": 0.0006,
+ "loss": 5.139893341064453,
+ "step": 1330
+ },
+ {
+ "epoch": 18.615250163862793,
+ "grad_norm": 0.22165881097316742,
+ "learning_rate": 0.0006,
+ "loss": 5.1016193389892575,
+ "step": 1340
+ },
+ {
+ "epoch": 18.75507974655888,
+ "grad_norm": 0.27904248237609863,
+ "learning_rate": 0.0006,
+ "loss": 5.128104400634766,
+ "step": 1350
+ },
+ {
+ "epoch": 18.894909329254972,
+ "grad_norm": 0.22776001691818237,
+ "learning_rate": 0.0006,
+ "loss": 5.1150562286376955,
+ "step": 1360
+ },
+ {
+ "epoch": 19.0,
+ "eval_loss": 4.072588920593262,
+ "eval_runtime": 19.8722,
+ "eval_samples_per_second": 122.835,
+ "eval_steps_per_second": 15.398,
+ "step": 1368
+ },
+ {
+ "epoch": 19.027965916539216,
+ "grad_norm": 0.25991615653038025,
+ "learning_rate": 0.0006,
+ "loss": 5.125886535644531,
+ "step": 1370
+ },
+ {
+ "epoch": 19.167795499235307,
+ "grad_norm": 0.23771052062511444,
+ "learning_rate": 0.0006,
+ "loss": 5.100527572631836,
+ "step": 1380
+ },
+ {
+ "epoch": 19.307625081931395,
+ "grad_norm": 0.2185024470090866,
+ "learning_rate": 0.0006,
+ "loss": 5.112496566772461,
+ "step": 1390
+ },
+ {
+ "epoch": 19.447454664627486,
+ "grad_norm": 0.272063672542572,
+ "learning_rate": 0.0006,
+ "loss": 5.090521621704101,
+ "step": 1400
+ },
+ {
+ "epoch": 19.587284247323574,
+ "grad_norm": 0.22067144513130188,
+ "learning_rate": 0.0006,
+ "loss": 5.077079010009766,
+ "step": 1410
+ },
+ {
+ "epoch": 19.727113830019665,
+ "grad_norm": 0.2031974196434021,
+ "learning_rate": 0.0006,
+ "loss": 5.078173446655273,
+ "step": 1420
+ },
+ {
+ "epoch": 19.866943412715752,
+ "grad_norm": 0.2309705764055252,
+ "learning_rate": 0.0006,
+ "loss": 5.096664047241211,
+ "step": 1430
+ },
+ {
+ "epoch": 20.0,
+ "grad_norm": 0.2505143880844116,
+ "learning_rate": 0.0006,
+ "loss": 5.117228698730469,
+ "step": 1440
+ },
+ {
+ "epoch": 20.0,
+ "eval_loss": 4.049932956695557,
+ "eval_runtime": 19.8881,
+ "eval_samples_per_second": 122.737,
+ "eval_steps_per_second": 15.386,
+ "step": 1440
+ },
+ {
+ "epoch": 20.139829582696088,
+ "grad_norm": 0.22238174080848694,
+ "learning_rate": 0.0006,
+ "loss": 5.065390014648438,
+ "step": 1450
+ },
+ {
+ "epoch": 20.27965916539218,
+ "grad_norm": 0.2451140284538269,
+ "learning_rate": 0.0006,
+ "loss": 5.052023315429688,
+ "step": 1460
+ },
+ {
+ "epoch": 20.419488748088266,
+ "grad_norm": 0.27595046162605286,
+ "learning_rate": 0.0006,
+ "loss": 5.073014450073242,
+ "step": 1470
+ },
+ {
+ "epoch": 20.559318330784357,
+ "grad_norm": 0.20936991274356842,
+ "learning_rate": 0.0006,
+ "loss": 5.079638290405273,
+ "step": 1480
+ },
+ {
+ "epoch": 20.699147913480445,
+ "grad_norm": 0.229985311627388,
+ "learning_rate": 0.0006,
+ "loss": 5.0666851043701175,
+ "step": 1490
+ },
+ {
+ "epoch": 20.838977496176536,
+ "grad_norm": 0.22804521024227142,
+ "learning_rate": 0.0006,
+ "loss": 5.0595237731933596,
+ "step": 1500
+ },
+ {
+ "epoch": 20.978807078872624,
+ "grad_norm": 0.23293128609657288,
+ "learning_rate": 0.0006,
+ "loss": 5.072089004516601,
+ "step": 1510
+ },
+ {
+ "epoch": 21.0,
+ "eval_loss": 4.0278801918029785,
+ "eval_runtime": 19.8948,
+ "eval_samples_per_second": 122.695,
+ "eval_steps_per_second": 15.381,
+ "step": 1512
+ },
+ {
+ "epoch": 21.11186366615687,
+ "grad_norm": 0.36356237530708313,
+ "learning_rate": 0.0006,
+ "loss": 5.059445571899414,
+ "step": 1520
+ },
+ {
+ "epoch": 21.25169324885296,
+ "grad_norm": 0.2529871165752411,
+ "learning_rate": 0.0006,
+ "loss": 5.055412292480469,
+ "step": 1530
+ },
+ {
+ "epoch": 21.39152283154905,
+ "grad_norm": 0.20508664846420288,
+ "learning_rate": 0.0006,
+ "loss": 5.044166564941406,
+ "step": 1540
+ },
+ {
+ "epoch": 21.531352414245138,
+ "grad_norm": 0.224314883351326,
+ "learning_rate": 0.0006,
+ "loss": 5.057932662963867,
+ "step": 1550
+ },
+ {
+ "epoch": 21.67118199694123,
+ "grad_norm": 0.21916615962982178,
+ "learning_rate": 0.0006,
+ "loss": 5.052794647216797,
+ "step": 1560
+ },
+ {
+ "epoch": 21.811011579637317,
+ "grad_norm": 0.2536468207836151,
+ "learning_rate": 0.0006,
+ "loss": 5.0565437316894535,
+ "step": 1570
+ },
+ {
+ "epoch": 21.950841162333408,
+ "grad_norm": 0.21259447932243347,
+ "learning_rate": 0.0006,
+ "loss": 5.046916198730469,
+ "step": 1580
+ },
+ {
+ "epoch": 22.0,
+ "eval_loss": 4.008029937744141,
+ "eval_runtime": 19.8741,
+ "eval_samples_per_second": 122.823,
+ "eval_steps_per_second": 15.397,
+ "step": 1584
+ },
+ {
+ "epoch": 22.083897749617652,
+ "grad_norm": 0.21582898497581482,
+ "learning_rate": 0.0006,
+ "loss": 5.008998870849609,
+ "step": 1590
+ },
+ {
+ "epoch": 22.223727332313743,
+ "grad_norm": 0.23864123225212097,
+ "learning_rate": 0.0006,
+ "loss": 5.021179962158203,
+ "step": 1600
+ },
+ {
+ "epoch": 22.36355691500983,
+ "grad_norm": 0.22727112472057343,
+ "learning_rate": 0.0006,
+ "loss": 5.022564697265625,
+ "step": 1610
+ },
+ {
+ "epoch": 22.50338649770592,
+ "grad_norm": 0.21434837579727173,
+ "learning_rate": 0.0006,
+ "loss": 4.99401741027832,
+ "step": 1620
+ },
+ {
+ "epoch": 22.64321608040201,
+ "grad_norm": 0.21472220122814178,
+ "learning_rate": 0.0006,
+ "loss": 5.028687286376953,
+ "step": 1630
+ },
+ {
+ "epoch": 22.7830456630981,
+ "grad_norm": 0.22586843371391296,
+ "learning_rate": 0.0006,
+ "loss": 4.987936401367188,
+ "step": 1640
+ },
+ {
+ "epoch": 22.922875245794188,
+ "grad_norm": 0.22765567898750305,
+ "learning_rate": 0.0006,
+ "loss": 5.03822021484375,
+ "step": 1650
+ },
+ {
+ "epoch": 23.0,
+ "eval_loss": 3.9949896335601807,
+ "eval_runtime": 19.8579,
+ "eval_samples_per_second": 122.923,
+ "eval_steps_per_second": 15.409,
+ "step": 1656
+ },
+ {
+ "epoch": 23.055931833078436,
+ "grad_norm": 0.2958192229270935,
+ "learning_rate": 0.0006,
+ "loss": 5.015535736083985,
+ "step": 1660
+ },
+ {
+ "epoch": 23.195761415774523,
+ "grad_norm": 0.24347978830337524,
+ "learning_rate": 0.0006,
+ "loss": 4.981285476684571,
+ "step": 1670
+ },
+ {
+ "epoch": 23.335590998470614,
+ "grad_norm": 0.2121012955904007,
+ "learning_rate": 0.0006,
+ "loss": 4.988442611694336,
+ "step": 1680
+ },
+ {
+ "epoch": 23.475420581166702,
+ "grad_norm": 0.20856034755706787,
+ "learning_rate": 0.0006,
+ "loss": 4.994302368164062,
+ "step": 1690
+ },
+ {
+ "epoch": 23.615250163862793,
+ "grad_norm": 0.23616427183151245,
+ "learning_rate": 0.0006,
+ "loss": 5.010406112670898,
+ "step": 1700
+ },
+ {
+ "epoch": 23.75507974655888,
+ "grad_norm": 0.24862425029277802,
+ "learning_rate": 0.0006,
+ "loss": 5.015080261230469,
+ "step": 1710
+ },
+ {
+ "epoch": 23.894909329254972,
+ "grad_norm": 0.2052837759256363,
+ "learning_rate": 0.0006,
+ "loss": 4.9907276153564455,
+ "step": 1720
+ },
+ {
+ "epoch": 24.0,
+ "eval_loss": 3.9752299785614014,
+ "eval_runtime": 19.8657,
+ "eval_samples_per_second": 122.875,
+ "eval_steps_per_second": 15.403,
+ "step": 1728
+ },
+ {
+ "epoch": 24.027965916539216,
+ "grad_norm": 0.21541047096252441,
+ "learning_rate": 0.0006,
+ "loss": 4.972194671630859,
+ "step": 1730
+ },
+ {
+ "epoch": 24.167795499235307,
+ "grad_norm": 0.20455493032932281,
+ "learning_rate": 0.0006,
+ "loss": 4.972018814086914,
+ "step": 1740
+ },
+ {
+ "epoch": 24.307625081931395,
+ "grad_norm": 0.25495803356170654,
+ "learning_rate": 0.0006,
+ "loss": 4.991477203369141,
+ "step": 1750
+ },
+ {
+ "epoch": 24.447454664627486,
+ "grad_norm": 0.19167377054691315,
+ "learning_rate": 0.0006,
+ "loss": 4.985448837280273,
+ "step": 1760
+ },
+ {
+ "epoch": 24.587284247323574,
+ "grad_norm": 0.21120870113372803,
+ "learning_rate": 0.0006,
+ "loss": 4.990003967285157,
+ "step": 1770
+ },
+ {
+ "epoch": 24.727113830019665,
+ "grad_norm": 0.27185723185539246,
+ "learning_rate": 0.0006,
+ "loss": 4.988444900512695,
+ "step": 1780
+ },
+ {
+ "epoch": 24.866943412715752,
+ "grad_norm": 0.22553864121437073,
+ "learning_rate": 0.0006,
+ "loss": 4.983925628662109,
+ "step": 1790
+ },
+ {
+ "epoch": 25.0,
+ "grad_norm": 0.2647480368614197,
+ "learning_rate": 0.0006,
+ "loss": 4.990047454833984,
+ "step": 1800
+ },
+ {
+ "epoch": 25.0,
+ "eval_loss": 3.96997332572937,
+ "eval_runtime": 19.8643,
+ "eval_samples_per_second": 122.884,
+ "eval_steps_per_second": 15.405,
+ "step": 1800
+ },
+ {
+ "epoch": 25.139829582696088,
+ "grad_norm": 0.19566938281059265,
+ "learning_rate": 0.0006,
+ "loss": 4.9477394104003904,
+ "step": 1810
+ },
+ {
+ "epoch": 25.27965916539218,
+ "grad_norm": 0.21404828131198883,
+ "learning_rate": 0.0006,
+ "loss": 4.937141799926758,
+ "step": 1820
+ },
+ {
+ "epoch": 25.419488748088266,
+ "grad_norm": 0.23815017938613892,
+ "learning_rate": 0.0006,
+ "loss": 4.950643920898438,
+ "step": 1830
+ },
+ {
+ "epoch": 25.559318330784357,
+ "grad_norm": 0.20241430401802063,
+ "learning_rate": 0.0006,
+ "loss": 4.965119171142578,
+ "step": 1840
+ },
+ {
+ "epoch": 25.699147913480445,
+ "grad_norm": 0.3020039498806,
+ "learning_rate": 0.0006,
+ "loss": 4.960115432739258,
+ "step": 1850
+ },
+ {
+ "epoch": 25.838977496176536,
+ "grad_norm": 0.2118341326713562,
+ "learning_rate": 0.0006,
+ "loss": 4.98785514831543,
+ "step": 1860
+ },
+ {
+ "epoch": 25.978807078872624,
+ "grad_norm": 0.2224600911140442,
+ "learning_rate": 0.0006,
+ "loss": 4.976942443847657,
+ "step": 1870
+ },
+ {
+ "epoch": 26.0,
+ "eval_loss": 3.9550652503967285,
+ "eval_runtime": 19.8655,
+ "eval_samples_per_second": 122.876,
+ "eval_steps_per_second": 15.404,
+ "step": 1872
+ },
+ {
+ "epoch": 26.11186366615687,
+ "grad_norm": 0.21752822399139404,
+ "learning_rate": 0.0006,
+ "loss": 4.934904861450195,
+ "step": 1880
+ },
+ {
+ "epoch": 26.25169324885296,
+ "grad_norm": 0.21701398491859436,
+ "learning_rate": 0.0006,
+ "loss": 4.9411567687988285,
+ "step": 1890
+ },
+ {
+ "epoch": 26.39152283154905,
+ "grad_norm": 0.23068198561668396,
+ "learning_rate": 0.0006,
+ "loss": 4.925229263305664,
+ "step": 1900
+ },
+ {
+ "epoch": 26.531352414245138,
+ "grad_norm": 0.23771940171718597,
+ "learning_rate": 0.0006,
+ "loss": 4.951910018920898,
+ "step": 1910
+ },
+ {
+ "epoch": 26.67118199694123,
+ "grad_norm": 0.19659607112407684,
+ "learning_rate": 0.0006,
+ "loss": 4.946131896972656,
+ "step": 1920
+ },
+ {
+ "epoch": 26.811011579637317,
+ "grad_norm": 0.25778424739837646,
+ "learning_rate": 0.0006,
+ "loss": 4.954918670654297,
+ "step": 1930
+ },
+ {
+ "epoch": 26.950841162333408,
+ "grad_norm": 0.19610807299613953,
+ "learning_rate": 0.0006,
+ "loss": 4.931307601928711,
+ "step": 1940
+ },
+ {
+ "epoch": 27.0,
+ "eval_loss": 3.9393091201782227,
+ "eval_runtime": 19.8806,
+ "eval_samples_per_second": 122.783,
+ "eval_steps_per_second": 15.392,
+ "step": 1944
+ },
+ {
+ "epoch": 27.083897749617652,
+ "grad_norm": 0.20383642613887787,
+ "learning_rate": 0.0006,
+ "loss": 4.925513839721679,
+ "step": 1950
+ },
+ {
+ "epoch": 27.223727332313743,
+ "grad_norm": 0.2158823162317276,
+ "learning_rate": 0.0006,
+ "loss": 4.9211467742919925,
+ "step": 1960
+ },
+ {
+ "epoch": 27.36355691500983,
+ "grad_norm": 0.22252072393894196,
+ "learning_rate": 0.0006,
+ "loss": 4.903144073486328,
+ "step": 1970
+ },
+ {
+ "epoch": 27.50338649770592,
+ "grad_norm": 0.23872476816177368,
+ "learning_rate": 0.0006,
+ "loss": 4.908475112915039,
+ "step": 1980
+ },
+ {
+ "epoch": 27.64321608040201,
+ "grad_norm": 0.21195088326931,
+ "learning_rate": 0.0006,
+ "loss": 4.933717727661133,
+ "step": 1990
+ },
+ {
+ "epoch": 27.7830456630981,
+ "grad_norm": 0.2233554720878601,
+ "learning_rate": 0.0006,
+ "loss": 4.937528228759765,
+ "step": 2000
+ },
+ {
+ "epoch": 27.922875245794188,
+ "grad_norm": 0.1994478702545166,
+ "learning_rate": 0.0006,
+ "loss": 4.919642639160156,
+ "step": 2010
+ },
+ {
+ "epoch": 28.0,
+ "eval_loss": 3.9329097270965576,
+ "eval_runtime": 19.8805,
+ "eval_samples_per_second": 122.784,
+ "eval_steps_per_second": 15.392,
+ "step": 2016
+ },
+ {
+ "epoch": 28.055931833078436,
+ "grad_norm": 0.24859251081943512,
+ "learning_rate": 0.0006,
+ "loss": 4.9434154510498045,
+ "step": 2020
+ },
+ {
+ "epoch": 28.195761415774523,
+ "grad_norm": 0.22961944341659546,
+ "learning_rate": 0.0006,
+ "loss": 4.9187572479248045,
+ "step": 2030
+ },
+ {
+ "epoch": 28.335590998470614,
+ "grad_norm": 0.2131347954273224,
+ "learning_rate": 0.0006,
+ "loss": 4.8861431121826175,
+ "step": 2040
+ },
+ {
+ "epoch": 28.475420581166702,
+ "grad_norm": 0.24457871913909912,
+ "learning_rate": 0.0006,
+ "loss": 4.8864295959472654,
+ "step": 2050
+ },
+ {
+ "epoch": 28.615250163862793,
+ "grad_norm": 0.20861995220184326,
+ "learning_rate": 0.0006,
+ "loss": 4.909621047973633,
+ "step": 2060
+ },
+ {
+ "epoch": 28.75507974655888,
+ "grad_norm": 0.24604618549346924,
+ "learning_rate": 0.0006,
+ "loss": 4.9301097869873045,
+ "step": 2070
+ },
+ {
+ "epoch": 28.894909329254972,
+ "grad_norm": 0.19704554975032806,
+ "learning_rate": 0.0006,
+ "loss": 4.90391845703125,
+ "step": 2080
+ },
+ {
+ "epoch": 29.0,
+ "eval_loss": 3.921823740005493,
+ "eval_runtime": 19.8853,
+ "eval_samples_per_second": 122.754,
+ "eval_steps_per_second": 15.388,
+ "step": 2088
+ },
+ {
+ "epoch": 29.027965916539216,
+ "grad_norm": 0.2266867458820343,
+ "learning_rate": 0.0006,
+ "loss": 4.902531433105469,
+ "step": 2090
+ },
+ {
+ "epoch": 29.167795499235307,
+ "grad_norm": 0.2299785614013672,
+ "learning_rate": 0.0006,
+ "loss": 4.9037025451660154,
+ "step": 2100
+ },
+ {
+ "epoch": 29.307625081931395,
+ "grad_norm": 0.21560785174369812,
+ "learning_rate": 0.0006,
+ "loss": 4.897410202026367,
+ "step": 2110
+ },
+ {
+ "epoch": 29.447454664627486,
+ "grad_norm": 0.21066813170909882,
+ "learning_rate": 0.0006,
+ "loss": 4.886149597167969,
+ "step": 2120
+ },
+ {
+ "epoch": 29.587284247323574,
+ "grad_norm": 0.24212266504764557,
+ "learning_rate": 0.0006,
+ "loss": 4.9057472229003904,
+ "step": 2130
+ },
+ {
+ "epoch": 29.727113830019665,
+ "grad_norm": 0.21783293783664703,
+ "learning_rate": 0.0006,
+ "loss": 4.903200912475586,
+ "step": 2140
+ },
+ {
+ "epoch": 29.866943412715752,
+ "grad_norm": 0.18009400367736816,
+ "learning_rate": 0.0006,
+ "loss": 4.885364532470703,
+ "step": 2150
+ },
+ {
+ "epoch": 30.0,
+ "grad_norm": 0.2251339852809906,
+ "learning_rate": 0.0006,
+ "loss": 4.892536544799805,
+ "step": 2160
+ },
+ {
+ "epoch": 30.0,
+ "eval_loss": 3.9129135608673096,
+ "eval_runtime": 19.9057,
+ "eval_samples_per_second": 122.628,
+ "eval_steps_per_second": 15.372,
+ "step": 2160
+ },
+ {
+ "epoch": 30.139829582696088,
+ "grad_norm": 0.2512070834636688,
+ "learning_rate": 0.0006,
+ "loss": 4.872610092163086,
+ "step": 2170
+ },
+ {
+ "epoch": 30.27965916539218,
+ "grad_norm": 0.23662729561328888,
+ "learning_rate": 0.0006,
+ "loss": 4.869965362548828,
+ "step": 2180
+ },
+ {
+ "epoch": 30.419488748088266,
+ "grad_norm": 0.19750620424747467,
+ "learning_rate": 0.0006,
+ "loss": 4.888255310058594,
+ "step": 2190
+ },
+ {
+ "epoch": 30.559318330784357,
+ "grad_norm": 0.21780364215373993,
+ "learning_rate": 0.0006,
+ "loss": 4.881540298461914,
+ "step": 2200
+ },
+ {
+ "epoch": 30.699147913480445,
+ "grad_norm": 0.21209587156772614,
+ "learning_rate": 0.0006,
+ "loss": 4.890599060058594,
+ "step": 2210
+ },
+ {
+ "epoch": 30.838977496176536,
+ "grad_norm": 0.22255845367908478,
+ "learning_rate": 0.0006,
+ "loss": 4.876234817504883,
+ "step": 2220
+ },
+ {
+ "epoch": 30.978807078872624,
+ "grad_norm": 0.22943098843097687,
+ "learning_rate": 0.0006,
+ "loss": 4.89000129699707,
+ "step": 2230
+ },
+ {
+ "epoch": 31.0,
+ "eval_loss": 3.9040637016296387,
+ "eval_runtime": 19.8863,
+ "eval_samples_per_second": 122.748,
+ "eval_steps_per_second": 15.387,
+ "step": 2232
+ },
+ {
+ "epoch": 31.11186366615687,
+ "grad_norm": 0.20729172229766846,
+ "learning_rate": 0.0006,
+ "loss": 4.876800537109375,
+ "step": 2240
+ },
+ {
+ "epoch": 31.25169324885296,
+ "grad_norm": 0.21609972417354584,
+ "learning_rate": 0.0006,
+ "loss": 4.880703735351562,
+ "step": 2250
+ },
+ {
+ "epoch": 31.39152283154905,
+ "grad_norm": 0.2123723030090332,
+ "learning_rate": 0.0006,
+ "loss": 4.8724933624267575,
+ "step": 2260
+ },
+ {
+ "epoch": 31.531352414245138,
+ "grad_norm": 0.1936194747686386,
+ "learning_rate": 0.0006,
+ "loss": 4.861216354370117,
+ "step": 2270
+ },
+ {
+ "epoch": 31.67118199694123,
+ "grad_norm": 0.23735114932060242,
+ "learning_rate": 0.0006,
+ "loss": 4.8814647674560545,
+ "step": 2280
+ },
+ {
+ "epoch": 31.811011579637317,
+ "grad_norm": 0.2286250740289688,
+ "learning_rate": 0.0006,
+ "loss": 4.857819366455078,
+ "step": 2290
+ },
+ {
+ "epoch": 31.950841162333408,
+ "grad_norm": 0.20370683073997498,
+ "learning_rate": 0.0006,
+ "loss": 4.8777915954589846,
+ "step": 2300
+ },
+ {
+ "epoch": 32.0,
+ "eval_loss": 3.8942551612854004,
+ "eval_runtime": 19.8595,
+ "eval_samples_per_second": 122.914,
+ "eval_steps_per_second": 15.408,
+ "step": 2304
+ },
+ {
+ "epoch": 32.083897749617655,
+ "grad_norm": 0.24912215769290924,
+ "learning_rate": 0.0006,
+ "loss": 4.854056549072266,
+ "step": 2310
+ },
+ {
+ "epoch": 32.22372733231374,
+ "grad_norm": 0.2454812377691269,
+ "learning_rate": 0.0006,
+ "loss": 4.853765487670898,
+ "step": 2320
+ },
+ {
+ "epoch": 32.36355691500983,
+ "grad_norm": 0.21579904854297638,
+ "learning_rate": 0.0006,
+ "loss": 4.847312545776367,
+ "step": 2330
+ },
+ {
+ "epoch": 32.50338649770592,
+ "grad_norm": 0.23074951767921448,
+ "learning_rate": 0.0006,
+ "loss": 4.861853408813476,
+ "step": 2340
+ },
+ {
+ "epoch": 32.64321608040201,
+ "grad_norm": 0.2100847214460373,
+ "learning_rate": 0.0006,
+ "loss": 4.843603515625,
+ "step": 2350
+ },
+ {
+ "epoch": 32.7830456630981,
+ "grad_norm": 0.20412872731685638,
+ "learning_rate": 0.0006,
+ "loss": 4.88000373840332,
+ "step": 2360
+ },
+ {
+ "epoch": 32.92287524579419,
+ "grad_norm": 0.21376822888851166,
+ "learning_rate": 0.0006,
+ "loss": 4.859452438354492,
+ "step": 2370
+ },
+ {
+ "epoch": 33.0,
+ "eval_loss": 3.891066789627075,
+ "eval_runtime": 19.8283,
+ "eval_samples_per_second": 123.107,
+ "eval_steps_per_second": 15.433,
+ "step": 2376
+ },
+ {
+ "epoch": 33.05593183307843,
+ "grad_norm": 0.20631594955921173,
+ "learning_rate": 0.0006,
+ "loss": 4.832675170898438,
+ "step": 2380
+ },
+ {
+ "epoch": 33.19576141577453,
+ "grad_norm": 0.22685754299163818,
+ "learning_rate": 0.0006,
+ "loss": 4.837306213378906,
+ "step": 2390
+ },
+ {
+ "epoch": 33.335590998470614,
+ "grad_norm": 0.22275273501873016,
+ "learning_rate": 0.0006,
+ "loss": 4.828560638427734,
+ "step": 2400
+ },
+ {
+ "epoch": 33.4754205811667,
+ "grad_norm": 0.19824115931987762,
+ "learning_rate": 0.0006,
+ "loss": 4.843119049072266,
+ "step": 2410
+ },
+ {
+ "epoch": 33.61525016386279,
+ "grad_norm": 0.1886703372001648,
+ "learning_rate": 0.0006,
+ "loss": 4.844092941284179,
+ "step": 2420
+ },
+ {
+ "epoch": 33.755079746558884,
+ "grad_norm": 0.24356143176555634,
+ "learning_rate": 0.0006,
+ "loss": 4.867326736450195,
+ "step": 2430
+ },
+ {
+ "epoch": 33.89490932925497,
+ "grad_norm": 0.25885090231895447,
+ "learning_rate": 0.0006,
+ "loss": 4.852556610107422,
+ "step": 2440
+ },
+ {
+ "epoch": 34.0,
+ "eval_loss": 3.889261484146118,
+ "eval_runtime": 19.8837,
+ "eval_samples_per_second": 122.764,
+ "eval_steps_per_second": 15.389,
+ "step": 2448
+ },
+ {
+ "epoch": 34.027965916539216,
+ "grad_norm": 0.23677904903888702,
+ "learning_rate": 0.0006,
+ "loss": 4.847293472290039,
+ "step": 2450
+ },
+ {
+ "epoch": 34.167795499235304,
+ "grad_norm": 0.24346138536930084,
+ "learning_rate": 0.0006,
+ "loss": 4.821566772460938,
+ "step": 2460
+ },
+ {
+ "epoch": 34.3076250819314,
+ "grad_norm": 0.21538189053535461,
+ "learning_rate": 0.0006,
+ "loss": 4.822731399536133,
+ "step": 2470
+ },
+ {
+ "epoch": 34.447454664627486,
+ "grad_norm": 0.18960842490196228,
+ "learning_rate": 0.0006,
+ "loss": 4.815583801269531,
+ "step": 2480
+ },
+ {
+ "epoch": 34.58728424732357,
+ "grad_norm": 0.21497246623039246,
+ "learning_rate": 0.0006,
+ "loss": 4.854027557373047,
+ "step": 2490
+ },
+ {
+ "epoch": 34.72711383001966,
+ "grad_norm": 0.2013980597257614,
+ "learning_rate": 0.0006,
+ "loss": 4.823099899291992,
+ "step": 2500
+ },
+ {
+ "epoch": 34.866943412715756,
+ "grad_norm": 0.217397540807724,
+ "learning_rate": 0.0006,
+ "loss": 4.85218505859375,
+ "step": 2510
+ },
+ {
+ "epoch": 35.0,
+ "grad_norm": 0.22843845188617706,
+ "learning_rate": 0.0006,
+ "loss": 4.843357086181641,
+ "step": 2520
+ },
+ {
+ "epoch": 35.0,
+ "eval_loss": 3.877617120742798,
+ "eval_runtime": 19.8488,
+ "eval_samples_per_second": 122.98,
+ "eval_steps_per_second": 15.417,
+ "step": 2520
+ },
+ {
+ "epoch": 35.13982958269609,
+ "grad_norm": 0.2526199519634247,
+ "learning_rate": 0.0006,
+ "loss": 4.810260391235351,
+ "step": 2530
+ },
+ {
+ "epoch": 35.279659165392175,
+ "grad_norm": 0.1938101053237915,
+ "learning_rate": 0.0006,
+ "loss": 4.815373229980469,
+ "step": 2540
+ },
+ {
+ "epoch": 35.41948874808827,
+ "grad_norm": 0.21734291315078735,
+ "learning_rate": 0.0006,
+ "loss": 4.812586975097656,
+ "step": 2550
+ },
+ {
+ "epoch": 35.55931833078436,
+ "grad_norm": 0.26203104853630066,
+ "learning_rate": 0.0006,
+ "loss": 4.848589706420898,
+ "step": 2560
+ },
+ {
+ "epoch": 35.699147913480445,
+ "grad_norm": 0.2047021985054016,
+ "learning_rate": 0.0006,
+ "loss": 4.8345378875732425,
+ "step": 2570
+ },
+ {
+ "epoch": 35.83897749617653,
+ "grad_norm": 0.2075943648815155,
+ "learning_rate": 0.0006,
+ "loss": 4.8235118865966795,
+ "step": 2580
+ },
+ {
+ "epoch": 35.97880707887263,
+ "grad_norm": 0.20531374216079712,
+ "learning_rate": 0.0006,
+ "loss": 4.82477912902832,
+ "step": 2590
+ },
+ {
+ "epoch": 36.0,
+ "eval_loss": 3.873983860015869,
+ "eval_runtime": 19.9088,
+ "eval_samples_per_second": 122.609,
+ "eval_steps_per_second": 15.37,
+ "step": 2592
+ },
+ {
+ "epoch": 36.11186366615687,
+ "grad_norm": 0.20720267295837402,
+ "learning_rate": 0.0006,
+ "loss": 4.8084972381591795,
+ "step": 2600
+ },
+ {
+ "epoch": 36.25169324885296,
+ "grad_norm": 0.22165866196155548,
+ "learning_rate": 0.0006,
+ "loss": 4.7961265563964846,
+ "step": 2610
+ },
+ {
+ "epoch": 36.39152283154905,
+ "grad_norm": 0.2634962797164917,
+ "learning_rate": 0.0006,
+ "loss": 4.8182018280029295,
+ "step": 2620
+ },
+ {
+ "epoch": 36.53135241424514,
+ "grad_norm": 0.23495934903621674,
+ "learning_rate": 0.0006,
+ "loss": 4.820383834838867,
+ "step": 2630
+ },
+ {
+ "epoch": 36.67118199694123,
+ "grad_norm": 0.194443479180336,
+ "learning_rate": 0.0006,
+ "loss": 4.80553092956543,
+ "step": 2640
+ },
+ {
+ "epoch": 36.81101157963732,
+ "grad_norm": 0.2008592188358307,
+ "learning_rate": 0.0006,
+ "loss": 4.816376113891602,
+ "step": 2650
+ },
+ {
+ "epoch": 36.950841162333404,
+ "grad_norm": 0.21756094694137573,
+ "learning_rate": 0.0006,
+ "loss": 4.8351997375488285,
+ "step": 2660
+ },
+ {
+ "epoch": 37.0,
+ "eval_loss": 3.8646721839904785,
+ "eval_runtime": 19.8743,
+ "eval_samples_per_second": 122.822,
+ "eval_steps_per_second": 15.397,
+ "step": 2664
+ },
+ {
+ "epoch": 37.083897749617655,
+ "grad_norm": 0.2149820625782013,
+ "learning_rate": 0.0006,
+ "loss": 4.798689270019532,
+ "step": 2670
+ },
+ {
+ "epoch": 37.22372733231374,
+ "grad_norm": 0.20645718276500702,
+ "learning_rate": 0.0006,
+ "loss": 4.796624374389649,
+ "step": 2680
+ },
+ {
+ "epoch": 37.36355691500983,
+ "grad_norm": 0.2111601084470749,
+ "learning_rate": 0.0006,
+ "loss": 4.804593658447265,
+ "step": 2690
+ },
+ {
+ "epoch": 37.50338649770592,
+ "grad_norm": 0.18802596628665924,
+ "learning_rate": 0.0006,
+ "loss": 4.803740692138672,
+ "step": 2700
+ },
+ {
+ "epoch": 37.64321608040201,
+ "grad_norm": 0.22289212048053741,
+ "learning_rate": 0.0006,
+ "loss": 4.803132247924805,
+ "step": 2710
+ },
+ {
+ "epoch": 37.7830456630981,
+ "grad_norm": 0.21882526576519012,
+ "learning_rate": 0.0006,
+ "loss": 4.8198692321777346,
+ "step": 2720
+ },
+ {
+ "epoch": 37.92287524579419,
+ "grad_norm": 0.20797757804393768,
+ "learning_rate": 0.0006,
+ "loss": 4.805184555053711,
+ "step": 2730
+ },
+ {
+ "epoch": 38.0,
+ "eval_loss": 3.8626515865325928,
+ "eval_runtime": 19.8711,
+ "eval_samples_per_second": 122.841,
+ "eval_steps_per_second": 15.399,
+ "step": 2736
+ },
+ {
+ "epoch": 38.05593183307843,
+ "grad_norm": 0.20918527245521545,
+ "learning_rate": 0.0006,
+ "loss": 4.786815643310547,
+ "step": 2740
+ },
+ {
+ "epoch": 38.19576141577453,
+ "grad_norm": 0.19230300188064575,
+ "learning_rate": 0.0006,
+ "loss": 4.800808334350586,
+ "step": 2750
+ },
+ {
+ "epoch": 38.335590998470614,
+ "grad_norm": 0.18650050461292267,
+ "learning_rate": 0.0006,
+ "loss": 4.804238891601562,
+ "step": 2760
+ },
+ {
+ "epoch": 38.4754205811667,
+ "grad_norm": 0.23730531334877014,
+ "learning_rate": 0.0006,
+ "loss": 4.799304580688476,
+ "step": 2770
+ },
+ {
+ "epoch": 38.61525016386279,
+ "grad_norm": 0.20464745163917542,
+ "learning_rate": 0.0006,
+ "loss": 4.812607192993164,
+ "step": 2780
+ },
+ {
+ "epoch": 38.755079746558884,
+ "grad_norm": 0.20051081478595734,
+ "learning_rate": 0.0006,
+ "loss": 4.812133407592773,
+ "step": 2790
+ },
+ {
+ "epoch": 38.89490932925497,
+ "grad_norm": 0.2172926664352417,
+ "learning_rate": 0.0006,
+ "loss": 4.825130844116211,
+ "step": 2800
+ },
+ {
+ "epoch": 39.0,
+ "eval_loss": 3.860419273376465,
+ "eval_runtime": 19.8835,
+ "eval_samples_per_second": 122.765,
+ "eval_steps_per_second": 15.39,
+ "step": 2808
+ },
+ {
+ "epoch": 39.027965916539216,
+ "grad_norm": 0.21916452050209045,
+ "learning_rate": 0.0006,
+ "loss": 4.787854766845703,
+ "step": 2810
+ },
+ {
+ "epoch": 39.167795499235304,
+ "grad_norm": 0.29523247480392456,
+ "learning_rate": 0.0006,
+ "loss": 4.781070327758789,
+ "step": 2820
+ },
+ {
+ "epoch": 39.3076250819314,
+ "grad_norm": 0.22156883776187897,
+ "learning_rate": 0.0006,
+ "loss": 4.800021743774414,
+ "step": 2830
+ },
+ {
+ "epoch": 39.447454664627486,
+ "grad_norm": 0.22916612029075623,
+ "learning_rate": 0.0006,
+ "loss": 4.804826354980468,
+ "step": 2840
+ },
+ {
+ "epoch": 39.58728424732357,
+ "grad_norm": 0.20375584065914154,
+ "learning_rate": 0.0006,
+ "loss": 4.783047103881836,
+ "step": 2850
+ },
+ {
+ "epoch": 39.72711383001966,
+ "grad_norm": 0.184370756149292,
+ "learning_rate": 0.0006,
+ "loss": 4.785774612426758,
+ "step": 2860
+ },
+ {
+ "epoch": 39.866943412715756,
+ "grad_norm": 0.1971350461244583,
+ "learning_rate": 0.0006,
+ "loss": 4.783952713012695,
+ "step": 2870
+ },
+ {
+ "epoch": 40.0,
+ "grad_norm": 0.2512478232383728,
+ "learning_rate": 0.0006,
+ "loss": 4.816450881958008,
+ "step": 2880
+ },
+ {
+ "epoch": 40.0,
+ "eval_loss": 3.854426145553589,
+ "eval_runtime": 19.8773,
+ "eval_samples_per_second": 122.803,
+ "eval_steps_per_second": 15.394,
+ "step": 2880
+ },
+ {
+ "epoch": 40.13982958269609,
+ "grad_norm": 0.25572943687438965,
+ "learning_rate": 0.0006,
+ "loss": 4.75367317199707,
+ "step": 2890
+ },
+ {
+ "epoch": 40.279659165392175,
+ "grad_norm": 0.2112315148115158,
+ "learning_rate": 0.0006,
+ "loss": 4.745833206176758,
+ "step": 2900
+ },
+ {
+ "epoch": 40.41948874808827,
+ "grad_norm": 0.19919830560684204,
+ "learning_rate": 0.0006,
+ "loss": 4.774380493164062,
+ "step": 2910
+ },
+ {
+ "epoch": 40.55931833078436,
+ "grad_norm": 0.2059367448091507,
+ "learning_rate": 0.0006,
+ "loss": 4.782925796508789,
+ "step": 2920
+ },
+ {
+ "epoch": 40.699147913480445,
+ "grad_norm": 0.1972065418958664,
+ "learning_rate": 0.0006,
+ "loss": 4.787817764282226,
+ "step": 2930
+ },
+ {
+ "epoch": 40.83897749617653,
+ "grad_norm": 0.19386503100395203,
+ "learning_rate": 0.0006,
+ "loss": 4.793806838989258,
+ "step": 2940
+ },
+ {
+ "epoch": 40.97880707887263,
+ "grad_norm": 0.22329017519950867,
+ "learning_rate": 0.0006,
+ "loss": 4.777092361450196,
+ "step": 2950
+ },
+ {
+ "epoch": 41.0,
+ "eval_loss": 3.8565285205841064,
+ "eval_runtime": 19.8761,
+ "eval_samples_per_second": 122.811,
+ "eval_steps_per_second": 15.395,
+ "step": 2952
+ },
+ {
+ "epoch": 41.11186366615687,
+ "grad_norm": 0.23527269065380096,
+ "learning_rate": 0.0006,
+ "loss": 4.759013748168945,
+ "step": 2960
+ },
+ {
+ "epoch": 41.25169324885296,
+ "grad_norm": 0.21171024441719055,
+ "learning_rate": 0.0006,
+ "loss": 4.746056747436524,
+ "step": 2970
+ },
+ {
+ "epoch": 41.39152283154905,
+ "grad_norm": 22.525999069213867,
+ "learning_rate": 0.0006,
+ "loss": 4.937899398803711,
+ "step": 2980
+ },
+ {
+ "epoch": 41.53135241424514,
+ "grad_norm": 1.6331260204315186,
+ "learning_rate": 0.0006,
+ "loss": 5.848717880249024,
+ "step": 2990
+ },
+ {
+ "epoch": 41.67118199694123,
+ "grad_norm": 2.0371134281158447,
+ "learning_rate": 0.0006,
+ "loss": 5.8080589294433596,
+ "step": 3000
+ },
+ {
+ "epoch": 41.81101157963732,
+ "grad_norm": 0.7730643153190613,
+ "learning_rate": 0.0006,
+ "loss": 5.532208633422852,
+ "step": 3010
+ },
+ {
+ "epoch": 41.950841162333404,
+ "grad_norm": 0.2524166703224182,
+ "learning_rate": 0.0006,
+ "loss": 5.289742279052734,
+ "step": 3020
+ },
+ {
+ "epoch": 42.0,
+ "eval_loss": 4.1060261726379395,
+ "eval_runtime": 19.8732,
+ "eval_samples_per_second": 122.829,
+ "eval_steps_per_second": 15.398,
+ "step": 3024
+ },
+ {
+ "epoch": 42.083897749617655,
+ "grad_norm": 0.19597682356834412,
+ "learning_rate": 0.0006,
+ "loss": 5.104341888427735,
+ "step": 3030
+ },
+ {
+ "epoch": 42.22372733231374,
+ "grad_norm": 0.14660239219665527,
+ "learning_rate": 0.0006,
+ "loss": 5.022753524780273,
+ "step": 3040
+ },
+ {
+ "epoch": 42.36355691500983,
+ "grad_norm": 0.15924286842346191,
+ "learning_rate": 0.0006,
+ "loss": 4.951805496215821,
+ "step": 3050
+ },
+ {
+ "epoch": 42.50338649770592,
+ "grad_norm": 0.15215617418289185,
+ "learning_rate": 0.0006,
+ "loss": 4.930615997314453,
+ "step": 3060
+ },
+ {
+ "epoch": 42.64321608040201,
+ "grad_norm": 0.169259175658226,
+ "learning_rate": 0.0006,
+ "loss": 4.885701751708984,
+ "step": 3070
+ },
+ {
+ "epoch": 42.7830456630981,
+ "grad_norm": 0.15493115782737732,
+ "learning_rate": 0.0006,
+ "loss": 4.865037155151367,
+ "step": 3080
+ },
+ {
+ "epoch": 42.92287524579419,
+ "grad_norm": 0.14189353585243225,
+ "learning_rate": 0.0006,
+ "loss": 4.855790710449218,
+ "step": 3090
+ },
+ {
+ "epoch": 43.0,
+ "eval_loss": 3.8872692584991455,
+ "eval_runtime": 19.8617,
+ "eval_samples_per_second": 122.9,
+ "eval_steps_per_second": 15.407,
+ "step": 3096
+ },
+ {
+ "epoch": 43.05593183307843,
+ "grad_norm": 0.35166364908218384,
+ "learning_rate": 0.0006,
+ "loss": 4.871305847167969,
+ "step": 3100
+ },
+ {
+ "epoch": 43.19576141577453,
+ "grad_norm": 0.17600642144680023,
+ "learning_rate": 0.0006,
+ "loss": 4.8058326721191404,
+ "step": 3110
+ },
+ {
+ "epoch": 43.335590998470614,
+ "grad_norm": 0.17076772451400757,
+ "learning_rate": 0.0006,
+ "loss": 4.820144271850586,
+ "step": 3120
+ },
+ {
+ "epoch": 43.4754205811667,
+ "grad_norm": 0.17997494339942932,
+ "learning_rate": 0.0006,
+ "loss": 4.822472000122071,
+ "step": 3130
+ },
+ {
+ "epoch": 43.61525016386279,
+ "grad_norm": 0.3493572771549225,
+ "learning_rate": 0.0006,
+ "loss": 4.825169372558594,
+ "step": 3140
+ },
+ {
+ "epoch": 43.755079746558884,
+ "grad_norm": 0.3211308717727661,
+ "learning_rate": 0.0006,
+ "loss": 4.894686126708985,
+ "step": 3150
+ },
+ {
+ "epoch": 43.89490932925497,
+ "grad_norm": 0.23874182999134064,
+ "learning_rate": 0.0006,
+ "loss": 4.934762573242187,
+ "step": 3160
+ },
+ {
+ "epoch": 44.0,
+ "eval_loss": 3.8827362060546875,
+ "eval_runtime": 19.8832,
+ "eval_samples_per_second": 122.767,
+ "eval_steps_per_second": 15.39,
+ "step": 3168
+ },
+ {
+ "epoch": 44.027965916539216,
+ "grad_norm": 0.16899354755878448,
+ "learning_rate": 0.0006,
+ "loss": 4.863615417480469,
+ "step": 3170
+ },
+ {
+ "epoch": 44.167795499235304,
+ "grad_norm": 0.15521778166294098,
+ "learning_rate": 0.0006,
+ "loss": 4.841251373291016,
+ "step": 3180
+ },
+ {
+ "epoch": 44.3076250819314,
+ "grad_norm": 0.15920916199684143,
+ "learning_rate": 0.0006,
+ "loss": 4.784141921997071,
+ "step": 3190
+ },
+ {
+ "epoch": 44.447454664627486,
+ "grad_norm": 0.15417060256004333,
+ "learning_rate": 0.0006,
+ "loss": 4.7895256042480465,
+ "step": 3200
+ },
+ {
+ "epoch": 44.58728424732357,
+ "grad_norm": 0.1521017849445343,
+ "learning_rate": 0.0006,
+ "loss": 4.791777420043945,
+ "step": 3210
+ },
+ {
+ "epoch": 44.72711383001966,
+ "grad_norm": 0.14990563690662384,
+ "learning_rate": 0.0006,
+ "loss": 4.781497955322266,
+ "step": 3220
+ },
+ {
+ "epoch": 44.866943412715756,
+ "grad_norm": 0.15872526168823242,
+ "learning_rate": 0.0006,
+ "loss": 4.7754676818847654,
+ "step": 3230
+ },
+ {
+ "epoch": 45.0,
+ "grad_norm": 0.18911069631576538,
+ "learning_rate": 0.0006,
+ "loss": 4.798118591308594,
+ "step": 3240
+ },
+ {
+ "epoch": 45.0,
+ "eval_loss": 3.846724033355713,
+ "eval_runtime": 19.8811,
+ "eval_samples_per_second": 122.78,
+ "eval_steps_per_second": 15.392,
+ "step": 3240
+ },
+ {
+ "epoch": 45.13982958269609,
+ "grad_norm": 0.16680815815925598,
+ "learning_rate": 0.0006,
+ "loss": 4.75772705078125,
+ "step": 3250
+ },
+ {
+ "epoch": 45.279659165392175,
+ "grad_norm": 0.1755628138780594,
+ "learning_rate": 0.0006,
+ "loss": 4.769796752929688,
+ "step": 3260
+ },
+ {
+ "epoch": 45.41948874808827,
+ "grad_norm": 0.16424906253814697,
+ "learning_rate": 0.0006,
+ "loss": 4.774546051025391,
+ "step": 3270
+ },
+ {
+ "epoch": 45.55931833078436,
+ "grad_norm": 0.16644906997680664,
+ "learning_rate": 0.0006,
+ "loss": 4.779018783569336,
+ "step": 3280
+ },
+ {
+ "epoch": 45.699147913480445,
+ "grad_norm": 0.17116601765155792,
+ "learning_rate": 0.0006,
+ "loss": 4.764167785644531,
+ "step": 3290
+ },
+ {
+ "epoch": 45.83897749617653,
+ "grad_norm": 0.17005392909049988,
+ "learning_rate": 0.0006,
+ "loss": 4.779270172119141,
+ "step": 3300
+ },
+ {
+ "epoch": 45.97880707887263,
+ "grad_norm": 0.16521161794662476,
+ "learning_rate": 0.0006,
+ "loss": 4.779656219482422,
+ "step": 3310
+ },
+ {
+ "epoch": 46.0,
+ "eval_loss": 3.8409879207611084,
+ "eval_runtime": 19.8917,
+ "eval_samples_per_second": 122.715,
+ "eval_steps_per_second": 15.383,
+ "step": 3312
+ },
+ {
+ "epoch": 46.11186366615687,
+ "grad_norm": 0.1703106015920639,
+ "learning_rate": 0.0006,
+ "loss": 4.740508270263672,
+ "step": 3320
+ },
+ {
+ "epoch": 46.25169324885296,
+ "grad_norm": 0.1715848296880722,
+ "learning_rate": 0.0006,
+ "loss": 4.724883270263672,
+ "step": 3330
+ },
+ {
+ "epoch": 46.39152283154905,
+ "grad_norm": 0.17324848473072052,
+ "learning_rate": 0.0006,
+ "loss": 4.728617095947266,
+ "step": 3340
+ },
+ {
+ "epoch": 46.53135241424514,
+ "grad_norm": 0.1682024449110031,
+ "learning_rate": 0.0006,
+ "loss": 4.758925247192383,
+ "step": 3350
+ },
+ {
+ "epoch": 46.67118199694123,
+ "grad_norm": 0.16215704381465912,
+ "learning_rate": 0.0006,
+ "loss": 4.766384124755859,
+ "step": 3360
+ },
+ {
+ "epoch": 46.81101157963732,
+ "grad_norm": 0.16584189236164093,
+ "learning_rate": 0.0006,
+ "loss": 4.764976882934571,
+ "step": 3370
+ },
+ {
+ "epoch": 46.950841162333404,
+ "grad_norm": 0.17134441435337067,
+ "learning_rate": 0.0006,
+ "loss": 4.779911041259766,
+ "step": 3380
+ },
+ {
+ "epoch": 47.0,
+ "eval_loss": 3.8372416496276855,
+ "eval_runtime": 19.8742,
+ "eval_samples_per_second": 122.822,
+ "eval_steps_per_second": 15.397,
+ "step": 3384
+ },
+ {
+ "epoch": 47.083897749617655,
+ "grad_norm": 0.1665998101234436,
+ "learning_rate": 0.0006,
+ "loss": 4.742333602905274,
+ "step": 3390
+ },
+ {
+ "epoch": 47.22372733231374,
+ "grad_norm": 0.18139144778251648,
+ "learning_rate": 0.0006,
+ "loss": 4.720782852172851,
+ "step": 3400
+ },
+ {
+ "epoch": 47.36355691500983,
+ "grad_norm": 0.185211643576622,
+ "learning_rate": 0.0006,
+ "loss": 4.7349296569824215,
+ "step": 3410
+ },
+ {
+ "epoch": 47.50338649770592,
+ "grad_norm": 0.1670907735824585,
+ "learning_rate": 0.0006,
+ "loss": 4.763270568847656,
+ "step": 3420
+ },
+ {
+ "epoch": 47.64321608040201,
+ "grad_norm": 0.18716077506542206,
+ "learning_rate": 0.0006,
+ "loss": 4.727042388916016,
+ "step": 3430
+ },
+ {
+ "epoch": 47.7830456630981,
+ "grad_norm": 0.17152276635169983,
+ "learning_rate": 0.0006,
+ "loss": 4.745809936523438,
+ "step": 3440
+ },
+ {
+ "epoch": 47.92287524579419,
+ "grad_norm": 0.17694507539272308,
+ "learning_rate": 0.0006,
+ "loss": 4.781035232543945,
+ "step": 3450
+ },
+ {
+ "epoch": 48.0,
+ "eval_loss": 3.8324384689331055,
+ "eval_runtime": 19.8845,
+ "eval_samples_per_second": 122.759,
+ "eval_steps_per_second": 15.389,
+ "step": 3456
+ },
+ {
+ "epoch": 48.05593183307843,
+ "grad_norm": 0.18532711267471313,
+ "learning_rate": 0.0006,
+ "loss": 4.743651580810547,
+ "step": 3460
+ },
+ {
+ "epoch": 48.19576141577453,
+ "grad_norm": 0.18051403760910034,
+ "learning_rate": 0.0006,
+ "loss": 4.728671646118164,
+ "step": 3470
+ },
+ {
+ "epoch": 48.335590998470614,
+ "grad_norm": 0.17348633706569672,
+ "learning_rate": 0.0006,
+ "loss": 4.750390243530274,
+ "step": 3480
+ },
+ {
+ "epoch": 48.4754205811667,
+ "grad_norm": 0.19282595813274384,
+ "learning_rate": 0.0006,
+ "loss": 4.735567855834961,
+ "step": 3490
+ },
+ {
+ "epoch": 48.61525016386279,
+ "grad_norm": 0.17296482622623444,
+ "learning_rate": 0.0006,
+ "loss": 4.742820358276367,
+ "step": 3500
+ },
+ {
+ "epoch": 48.755079746558884,
+ "grad_norm": 0.18879961967468262,
+ "learning_rate": 0.0006,
+ "loss": 4.737877655029297,
+ "step": 3510
+ },
+ {
+ "epoch": 48.89490932925497,
+ "grad_norm": 0.17374156415462494,
+ "learning_rate": 0.0006,
+ "loss": 4.732737350463867,
+ "step": 3520
+ },
+ {
+ "epoch": 49.0,
+ "eval_loss": 3.832265615463257,
+ "eval_runtime": 19.8834,
+ "eval_samples_per_second": 122.766,
+ "eval_steps_per_second": 15.39,
+ "step": 3528
+ },
+ {
+ "epoch": 49.027965916539216,
+ "grad_norm": 0.19252924621105194,
+ "learning_rate": 0.0006,
+ "loss": 4.729864120483398,
+ "step": 3530
+ },
+ {
+ "epoch": 49.167795499235304,
+ "grad_norm": 0.1831343024969101,
+ "learning_rate": 0.0006,
+ "loss": 4.711724472045899,
+ "step": 3540
+ },
+ {
+ "epoch": 49.3076250819314,
+ "grad_norm": 0.18384422361850739,
+ "learning_rate": 0.0006,
+ "loss": 4.710269546508789,
+ "step": 3550
+ },
+ {
+ "epoch": 49.447454664627486,
+ "grad_norm": 0.18014152348041534,
+ "learning_rate": 0.0006,
+ "loss": 4.731889724731445,
+ "step": 3560
+ },
+ {
+ "epoch": 49.58728424732357,
+ "grad_norm": 0.1778501272201538,
+ "learning_rate": 0.0006,
+ "loss": 4.745777893066406,
+ "step": 3570
+ },
+ {
+ "epoch": 49.72711383001966,
+ "grad_norm": 0.1705511063337326,
+ "learning_rate": 0.0006,
+ "loss": 4.73709716796875,
+ "step": 3580
+ },
+ {
+ "epoch": 49.866943412715756,
+ "grad_norm": 0.17540687322616577,
+ "learning_rate": 0.0006,
+ "loss": 4.722479248046875,
+ "step": 3590
+ },
+ {
+ "epoch": 50.0,
+ "grad_norm": 0.22017529606819153,
+ "learning_rate": 0.0006,
+ "loss": 4.723981475830078,
+ "step": 3600
+ },
+ {
+ "epoch": 50.0,
+ "eval_loss": 3.8285906314849854,
+ "eval_runtime": 19.8826,
+ "eval_samples_per_second": 122.77,
+ "eval_steps_per_second": 15.39,
+ "step": 3600
+ },
+ {
+ "epoch": 50.13982958269609,
+ "grad_norm": 0.18896132707595825,
+ "learning_rate": 0.0006,
+ "loss": 4.682983016967773,
+ "step": 3610
+ },
+ {
+ "epoch": 50.279659165392175,
+ "grad_norm": 0.17524129152297974,
+ "learning_rate": 0.0006,
+ "loss": 4.709360885620117,
+ "step": 3620
+ },
+ {
+ "epoch": 50.41948874808827,
+ "grad_norm": 0.1805744469165802,
+ "learning_rate": 0.0006,
+ "loss": 4.719758605957031,
+ "step": 3630
+ },
+ {
+ "epoch": 50.55931833078436,
+ "grad_norm": 0.190475195646286,
+ "learning_rate": 0.0006,
+ "loss": 4.729971694946289,
+ "step": 3640
+ },
+ {
+ "epoch": 50.699147913480445,
+ "grad_norm": 0.1775844395160675,
+ "learning_rate": 0.0006,
+ "loss": 4.743276214599609,
+ "step": 3650
+ },
+ {
+ "epoch": 50.83897749617653,
+ "grad_norm": 0.18260429799556732,
+ "learning_rate": 0.0006,
+ "loss": 4.7101280212402346,
+ "step": 3660
+ },
+ {
+ "epoch": 50.97880707887263,
+ "grad_norm": 0.1999107152223587,
+ "learning_rate": 0.0006,
+ "loss": 4.75849494934082,
+ "step": 3670
+ },
+ {
+ "epoch": 51.0,
+ "eval_loss": 3.8242807388305664,
+ "eval_runtime": 19.8792,
+ "eval_samples_per_second": 122.792,
+ "eval_steps_per_second": 15.393,
+ "step": 3672
+ },
+ {
+ "epoch": 51.11186366615687,
+ "grad_norm": 0.20437408983707428,
+ "learning_rate": 0.0006,
+ "loss": 4.70637321472168,
+ "step": 3680
+ },
+ {
+ "epoch": 51.25169324885296,
+ "grad_norm": 0.21942563354969025,
+ "learning_rate": 0.0006,
+ "loss": 4.694840621948242,
+ "step": 3690
+ },
+ {
+ "epoch": 51.39152283154905,
+ "grad_norm": 0.18444079160690308,
+ "learning_rate": 0.0006,
+ "loss": 4.72169075012207,
+ "step": 3700
+ },
+ {
+ "epoch": 51.53135241424514,
+ "grad_norm": 0.17947980761528015,
+ "learning_rate": 0.0006,
+ "loss": 4.674046325683594,
+ "step": 3710
+ },
+ {
+ "epoch": 51.67118199694123,
+ "grad_norm": 0.1736469715833664,
+ "learning_rate": 0.0006,
+ "loss": 4.725739669799805,
+ "step": 3720
+ },
+ {
+ "epoch": 51.81101157963732,
+ "grad_norm": 0.1828382909297943,
+ "learning_rate": 0.0006,
+ "loss": 4.721571731567383,
+ "step": 3730
+ },
+ {
+ "epoch": 51.950841162333404,
+ "grad_norm": 0.19675672054290771,
+ "learning_rate": 0.0006,
+ "loss": 4.716368103027344,
+ "step": 3740
+ },
+ {
+ "epoch": 52.0,
+ "eval_loss": 3.828145742416382,
+ "eval_runtime": 19.8556,
+ "eval_samples_per_second": 122.938,
+ "eval_steps_per_second": 15.411,
+ "step": 3744
+ },
+ {
+ "epoch": 52.083897749617655,
+ "grad_norm": 0.18220734596252441,
+ "learning_rate": 0.0006,
+ "loss": 4.699446105957032,
+ "step": 3750
+ },
+ {
+ "epoch": 52.22372733231374,
+ "grad_norm": 0.1834433674812317,
+ "learning_rate": 0.0006,
+ "loss": 4.696271896362305,
+ "step": 3760
+ },
+ {
+ "epoch": 52.36355691500983,
+ "grad_norm": 0.18527671694755554,
+ "learning_rate": 0.0006,
+ "loss": 4.698380661010742,
+ "step": 3770
+ },
+ {
+ "epoch": 52.50338649770592,
+ "grad_norm": 0.18875326216220856,
+ "learning_rate": 0.0006,
+ "loss": 4.712321472167969,
+ "step": 3780
+ },
+ {
+ "epoch": 52.64321608040201,
+ "grad_norm": 0.1843811720609665,
+ "learning_rate": 0.0006,
+ "loss": 4.678805541992188,
+ "step": 3790
+ },
+ {
+ "epoch": 52.7830456630981,
+ "grad_norm": 0.18752862513065338,
+ "learning_rate": 0.0006,
+ "loss": 4.712589263916016,
+ "step": 3800
+ },
+ {
+ "epoch": 52.92287524579419,
+ "grad_norm": 0.1879226565361023,
+ "learning_rate": 0.0006,
+ "loss": 4.7094581604003904,
+ "step": 3810
+ },
+ {
+ "epoch": 53.0,
+ "eval_loss": 3.8249356746673584,
+ "eval_runtime": 19.8788,
+ "eval_samples_per_second": 122.794,
+ "eval_steps_per_second": 15.393,
+ "step": 3816
+ },
+ {
+ "epoch": 53.05593183307843,
+ "grad_norm": 0.20928291976451874,
+ "learning_rate": 0.0006,
+ "loss": 4.696858596801758,
+ "step": 3820
+ },
+ {
+ "epoch": 53.19576141577453,
+ "grad_norm": 0.17810186743736267,
+ "learning_rate": 0.0006,
+ "loss": 4.687096786499024,
+ "step": 3830
+ },
+ {
+ "epoch": 53.335590998470614,
+ "grad_norm": 0.19799920916557312,
+ "learning_rate": 0.0006,
+ "loss": 4.681804275512695,
+ "step": 3840
+ },
+ {
+ "epoch": 53.4754205811667,
+ "grad_norm": 0.18917086720466614,
+ "learning_rate": 0.0006,
+ "loss": 4.698408508300782,
+ "step": 3850
+ },
+ {
+ "epoch": 53.61525016386279,
+ "grad_norm": 0.20475496351718903,
+ "learning_rate": 0.0006,
+ "loss": 4.714855194091797,
+ "step": 3860
+ },
+ {
+ "epoch": 53.755079746558884,
+ "grad_norm": 0.18814033269882202,
+ "learning_rate": 0.0006,
+ "loss": 4.671658706665039,
+ "step": 3870
+ },
+ {
+ "epoch": 53.89490932925497,
+ "grad_norm": 0.1799933761358261,
+ "learning_rate": 0.0006,
+ "loss": 4.714011001586914,
+ "step": 3880
+ },
+ {
+ "epoch": 54.0,
+ "eval_loss": 3.824115514755249,
+ "eval_runtime": 19.8616,
+ "eval_samples_per_second": 122.901,
+ "eval_steps_per_second": 15.407,
+ "step": 3888
+ },
+ {
+ "epoch": 54.027965916539216,
+ "grad_norm": 0.19464558362960815,
+ "learning_rate": 0.0006,
+ "loss": 4.714732360839844,
+ "step": 3890
+ },
+ {
+ "epoch": 54.167795499235304,
+ "grad_norm": 0.20483791828155518,
+ "learning_rate": 0.0006,
+ "loss": 4.669854736328125,
+ "step": 3900
+ },
+ {
+ "epoch": 54.3076250819314,
+ "grad_norm": 0.18571896851062775,
+ "learning_rate": 0.0006,
+ "loss": 4.659539031982422,
+ "step": 3910
+ },
+ {
+ "epoch": 54.447454664627486,
+ "grad_norm": 0.1880280077457428,
+ "learning_rate": 0.0006,
+ "loss": 4.6629180908203125,
+ "step": 3920
+ },
+ {
+ "epoch": 54.58728424732357,
+ "grad_norm": 0.18686671555042267,
+ "learning_rate": 0.0006,
+ "loss": 4.681613540649414,
+ "step": 3930
+ },
+ {
+ "epoch": 54.72711383001966,
+ "grad_norm": 0.1900676190853119,
+ "learning_rate": 0.0006,
+ "loss": 4.718310546875,
+ "step": 3940
+ },
+ {
+ "epoch": 54.866943412715756,
+ "grad_norm": 0.19094745814800262,
+ "learning_rate": 0.0006,
+ "loss": 4.701691818237305,
+ "step": 3950
+ },
+ {
+ "epoch": 55.0,
+ "grad_norm": 0.22744594514369965,
+ "learning_rate": 0.0006,
+ "loss": 4.709153366088867,
+ "step": 3960
+ },
+ {
+ "epoch": 55.0,
+ "eval_loss": 3.824770927429199,
+ "eval_runtime": 19.868,
+ "eval_samples_per_second": 122.861,
+ "eval_steps_per_second": 15.402,
+ "step": 3960
+ },
+ {
+ "epoch": 55.13982958269609,
+ "grad_norm": 0.2019844502210617,
+ "learning_rate": 0.0006,
+ "loss": 4.668522644042969,
+ "step": 3970
+ },
+ {
+ "epoch": 55.279659165392175,
+ "grad_norm": 0.19143150746822357,
+ "learning_rate": 0.0006,
+ "loss": 4.670612716674805,
+ "step": 3980
+ },
+ {
+ "epoch": 55.41948874808827,
+ "grad_norm": 0.18436941504478455,
+ "learning_rate": 0.0006,
+ "loss": 4.698069000244141,
+ "step": 3990
+ },
+ {
+ "epoch": 55.55931833078436,
+ "grad_norm": 0.19224430620670319,
+ "learning_rate": 0.0006,
+ "loss": 4.6969482421875,
+ "step": 4000
+ },
+ {
+ "epoch": 55.699147913480445,
+ "grad_norm": 0.16750234365463257,
+ "learning_rate": 0.0006,
+ "loss": 4.6871898651123045,
+ "step": 4010
+ },
+ {
+ "epoch": 55.83897749617653,
+ "grad_norm": 0.20357821881771088,
+ "learning_rate": 0.0006,
+ "loss": 4.6840965270996096,
+ "step": 4020
+ },
+ {
+ "epoch": 55.97880707887263,
+ "grad_norm": 0.21988941729068756,
+ "learning_rate": 0.0006,
+ "loss": 4.702257919311523,
+ "step": 4030
+ },
+ {
+ "epoch": 56.0,
+ "eval_loss": 3.820272445678711,
+ "eval_runtime": 19.8727,
+ "eval_samples_per_second": 122.832,
+ "eval_steps_per_second": 15.398,
+ "step": 4032
+ },
+ {
+ "epoch": 56.11186366615687,
+ "grad_norm": 0.191788449883461,
+ "learning_rate": 0.0006,
+ "loss": 4.672414779663086,
+ "step": 4040
+ },
+ {
+ "epoch": 56.25169324885296,
+ "grad_norm": 0.21312043070793152,
+ "learning_rate": 0.0006,
+ "loss": 4.677421188354492,
+ "step": 4050
+ },
+ {
+ "epoch": 56.39152283154905,
+ "grad_norm": 0.1835770606994629,
+ "learning_rate": 0.0006,
+ "loss": 4.659929275512695,
+ "step": 4060
+ },
+ {
+ "epoch": 56.53135241424514,
+ "grad_norm": 0.22035381197929382,
+ "learning_rate": 0.0006,
+ "loss": 4.665010452270508,
+ "step": 4070
+ },
+ {
+ "epoch": 56.67118199694123,
+ "grad_norm": 0.18291093409061432,
+ "learning_rate": 0.0006,
+ "loss": 4.708282470703125,
+ "step": 4080
+ },
+ {
+ "epoch": 56.81101157963732,
+ "grad_norm": 0.19147560000419617,
+ "learning_rate": 0.0006,
+ "loss": 4.695187377929687,
+ "step": 4090
+ },
+ {
+ "epoch": 56.950841162333404,
+ "grad_norm": 0.21278563141822815,
+ "learning_rate": 0.0006,
+ "loss": 4.704980850219727,
+ "step": 4100
+ },
+ {
+ "epoch": 57.0,
+ "eval_loss": 3.817530393600464,
+ "eval_runtime": 19.8624,
+ "eval_samples_per_second": 122.896,
+ "eval_steps_per_second": 15.406,
+ "step": 4104
+ },
+ {
+ "epoch": 57.083897749617655,
+ "grad_norm": 0.21941319108009338,
+ "learning_rate": 0.0006,
+ "loss": 4.665832901000977,
+ "step": 4110
+ },
+ {
+ "epoch": 57.22372733231374,
+ "grad_norm": 0.19666622579097748,
+ "learning_rate": 0.0006,
+ "loss": 4.656256866455078,
+ "step": 4120
+ },
+ {
+ "epoch": 57.36355691500983,
+ "grad_norm": 0.19325830042362213,
+ "learning_rate": 0.0006,
+ "loss": 4.671561050415039,
+ "step": 4130
+ },
+ {
+ "epoch": 57.50338649770592,
+ "grad_norm": 0.18281924724578857,
+ "learning_rate": 0.0006,
+ "loss": 4.661495208740234,
+ "step": 4140
+ },
+ {
+ "epoch": 57.64321608040201,
+ "grad_norm": 0.1920410692691803,
+ "learning_rate": 0.0006,
+ "loss": 4.688328552246094,
+ "step": 4150
+ },
+ {
+ "epoch": 57.7830456630981,
+ "grad_norm": 0.19096143543720245,
+ "learning_rate": 0.0006,
+ "loss": 4.709062194824218,
+ "step": 4160
+ },
+ {
+ "epoch": 57.92287524579419,
+ "grad_norm": 0.17966091632843018,
+ "learning_rate": 0.0006,
+ "loss": 4.706349563598633,
+ "step": 4170
+ },
+ {
+ "epoch": 58.0,
+ "eval_loss": 3.8186466693878174,
+ "eval_runtime": 19.8867,
+ "eval_samples_per_second": 122.745,
+ "eval_steps_per_second": 15.387,
+ "step": 4176
+ },
+ {
+ "epoch": 58.05593183307843,
+ "grad_norm": 0.22185473144054413,
+ "learning_rate": 0.0006,
+ "loss": 4.691797256469727,
+ "step": 4180
+ },
+ {
+ "epoch": 58.19576141577453,
+ "grad_norm": 0.19726398587226868,
+ "learning_rate": 0.0006,
+ "loss": 4.651484680175781,
+ "step": 4190
+ },
+ {
+ "epoch": 58.335590998470614,
+ "grad_norm": 0.18144628405570984,
+ "learning_rate": 0.0006,
+ "loss": 4.661507797241211,
+ "step": 4200
+ },
+ {
+ "epoch": 58.4754205811667,
+ "grad_norm": 0.18910720944404602,
+ "learning_rate": 0.0006,
+ "loss": 4.660706329345703,
+ "step": 4210
+ },
+ {
+ "epoch": 58.61525016386279,
+ "grad_norm": 0.22065356373786926,
+ "learning_rate": 0.0006,
+ "loss": 4.707099914550781,
+ "step": 4220
+ },
+ {
+ "epoch": 58.755079746558884,
+ "grad_norm": 0.1941348910331726,
+ "learning_rate": 0.0006,
+ "loss": 4.673603057861328,
+ "step": 4230
+ },
+ {
+ "epoch": 58.89490932925497,
+ "grad_norm": 0.18585318326950073,
+ "learning_rate": 0.0006,
+ "loss": 4.69781379699707,
+ "step": 4240
+ },
+ {
+ "epoch": 59.0,
+ "eval_loss": 3.8172175884246826,
+ "eval_runtime": 19.8665,
+ "eval_samples_per_second": 122.87,
+ "eval_steps_per_second": 15.403,
+ "step": 4248
+ },
+ {
+ "epoch": 59.027965916539216,
+ "grad_norm": 0.2061402052640915,
+ "learning_rate": 0.0006,
+ "loss": 4.673849105834961,
+ "step": 4250
+ },
+ {
+ "epoch": 59.167795499235304,
+ "grad_norm": 0.20132972300052643,
+ "learning_rate": 0.0006,
+ "loss": 4.655863189697266,
+ "step": 4260
+ },
+ {
+ "epoch": 59.3076250819314,
+ "grad_norm": 0.1974571794271469,
+ "learning_rate": 0.0006,
+ "loss": 4.667364120483398,
+ "step": 4270
+ },
+ {
+ "epoch": 59.447454664627486,
+ "grad_norm": 0.18729819357395172,
+ "learning_rate": 0.0006,
+ "loss": 4.66797103881836,
+ "step": 4280
+ },
+ {
+ "epoch": 59.58728424732357,
+ "grad_norm": 0.1784348487854004,
+ "learning_rate": 0.0006,
+ "loss": 4.664365005493164,
+ "step": 4290
+ },
+ {
+ "epoch": 59.72711383001966,
+ "grad_norm": 0.1926375925540924,
+ "learning_rate": 0.0006,
+ "loss": 4.658140563964844,
+ "step": 4300
+ },
+ {
+ "epoch": 59.866943412715756,
+ "grad_norm": 0.19018849730491638,
+ "learning_rate": 0.0006,
+ "loss": 4.652832794189453,
+ "step": 4310
+ },
+ {
+ "epoch": 60.0,
+ "grad_norm": 0.22277569770812988,
+ "learning_rate": 0.0006,
+ "loss": 4.675009536743164,
+ "step": 4320
+ },
+ {
+ "epoch": 60.0,
+ "eval_loss": 3.8137154579162598,
+ "eval_runtime": 19.868,
+ "eval_samples_per_second": 122.861,
+ "eval_steps_per_second": 15.402,
+ "step": 4320
+ },
+ {
+ "epoch": 60.13982958269609,
+ "grad_norm": 0.2021809220314026,
+ "learning_rate": 0.0006,
+ "loss": 4.651922607421875,
+ "step": 4330
+ },
+ {
+ "epoch": 60.279659165392175,
+ "grad_norm": 0.18974873423576355,
+ "learning_rate": 0.0006,
+ "loss": 4.64575424194336,
+ "step": 4340
+ },
+ {
+ "epoch": 60.41948874808827,
+ "grad_norm": 0.1774570196866989,
+ "learning_rate": 0.0006,
+ "loss": 4.656733703613281,
+ "step": 4350
+ },
+ {
+ "epoch": 60.55931833078436,
+ "grad_norm": 0.2151338905096054,
+ "learning_rate": 0.0006,
+ "loss": 4.669946670532227,
+ "step": 4360
+ },
+ {
+ "epoch": 60.699147913480445,
+ "grad_norm": 0.1834229975938797,
+ "learning_rate": 0.0006,
+ "loss": 4.657568359375,
+ "step": 4370
+ },
+ {
+ "epoch": 60.83897749617653,
+ "grad_norm": 0.1966671198606491,
+ "learning_rate": 0.0006,
+ "loss": 4.682220840454102,
+ "step": 4380
+ },
+ {
+ "epoch": 60.97880707887263,
+ "grad_norm": 0.19139347970485687,
+ "learning_rate": 0.0006,
+ "loss": 4.681847381591797,
+ "step": 4390
+ },
+ {
+ "epoch": 61.0,
+ "eval_loss": 3.814025402069092,
+ "eval_runtime": 19.8703,
+ "eval_samples_per_second": 122.847,
+ "eval_steps_per_second": 15.4,
+ "step": 4392
+ },
+ {
+ "epoch": 61.11186366615687,
+ "grad_norm": 0.20676615834236145,
+ "learning_rate": 0.0006,
+ "loss": 4.639699554443359,
+ "step": 4400
+ },
+ {
+ "epoch": 61.25169324885296,
+ "grad_norm": 0.21310955286026,
+ "learning_rate": 0.0006,
+ "loss": 4.649012756347656,
+ "step": 4410
+ },
+ {
+ "epoch": 61.39152283154905,
+ "grad_norm": 0.18878060579299927,
+ "learning_rate": 0.0006,
+ "loss": 4.650315856933593,
+ "step": 4420
+ },
+ {
+ "epoch": 61.53135241424514,
+ "grad_norm": 0.2027527391910553,
+ "learning_rate": 0.0006,
+ "loss": 4.6452491760253904,
+ "step": 4430
+ },
+ {
+ "epoch": 61.67118199694123,
+ "grad_norm": 0.20255820453166962,
+ "learning_rate": 0.0006,
+ "loss": 4.63293571472168,
+ "step": 4440
+ },
+ {
+ "epoch": 61.81101157963732,
+ "grad_norm": 0.20470598340034485,
+ "learning_rate": 0.0006,
+ "loss": 4.667168045043946,
+ "step": 4450
+ },
+ {
+ "epoch": 61.950841162333404,
+ "grad_norm": 0.20371755957603455,
+ "learning_rate": 0.0006,
+ "loss": 4.667899322509766,
+ "step": 4460
+ },
+ {
+ "epoch": 62.0,
+ "eval_loss": 3.814819574356079,
+ "eval_runtime": 19.8574,
+ "eval_samples_per_second": 122.926,
+ "eval_steps_per_second": 15.41,
+ "step": 4464
+ },
+ {
+ "epoch": 62.083897749617655,
+ "grad_norm": 0.22562506794929504,
+ "learning_rate": 0.0006,
+ "loss": 4.645642852783203,
+ "step": 4470
+ },
+ {
+ "epoch": 62.22372733231374,
+ "grad_norm": 0.19694721698760986,
+ "learning_rate": 0.0006,
+ "loss": 4.631222152709961,
+ "step": 4480
+ },
+ {
+ "epoch": 62.36355691500983,
+ "grad_norm": 0.22761695086956024,
+ "learning_rate": 0.0006,
+ "loss": 4.641876602172852,
+ "step": 4490
+ },
+ {
+ "epoch": 62.50338649770592,
+ "grad_norm": 0.1962880939245224,
+ "learning_rate": 0.0006,
+ "loss": 4.648233795166016,
+ "step": 4500
+ },
+ {
+ "epoch": 62.64321608040201,
+ "grad_norm": 0.19471420347690582,
+ "learning_rate": 0.0006,
+ "loss": 4.6440387725830075,
+ "step": 4510
+ },
+ {
+ "epoch": 62.7830456630981,
+ "grad_norm": 0.20076067745685577,
+ "learning_rate": 0.0006,
+ "loss": 4.64909553527832,
+ "step": 4520
+ },
+ {
+ "epoch": 62.92287524579419,
+ "grad_norm": 0.21344459056854248,
+ "learning_rate": 0.0006,
+ "loss": 4.649943923950195,
+ "step": 4530
+ },
+ {
+ "epoch": 63.0,
+ "eval_loss": 3.805960178375244,
+ "eval_runtime": 19.8744,
+ "eval_samples_per_second": 122.821,
+ "eval_steps_per_second": 15.397,
+ "step": 4536
+ },
+ {
+ "epoch": 63.05593183307843,
+ "grad_norm": 0.2157433182001114,
+ "learning_rate": 0.0006,
+ "loss": 4.665023422241211,
+ "step": 4540
+ },
+ {
+ "epoch": 63.19576141577453,
+ "grad_norm": 0.19380687177181244,
+ "learning_rate": 0.0006,
+ "loss": 4.650500869750976,
+ "step": 4550
+ },
+ {
+ "epoch": 63.335590998470614,
+ "grad_norm": 0.22612617909908295,
+ "learning_rate": 0.0006,
+ "loss": 4.641617202758789,
+ "step": 4560
+ },
+ {
+ "epoch": 63.4754205811667,
+ "grad_norm": 0.19319947063922882,
+ "learning_rate": 0.0006,
+ "loss": 4.641976928710937,
+ "step": 4570
+ },
+ {
+ "epoch": 63.61525016386279,
+ "grad_norm": 0.2167237251996994,
+ "learning_rate": 0.0006,
+ "loss": 4.6466117858886715,
+ "step": 4580
+ },
+ {
+ "epoch": 63.755079746558884,
+ "grad_norm": 0.20178721845149994,
+ "learning_rate": 0.0006,
+ "loss": 4.657366561889648,
+ "step": 4590
+ },
+ {
+ "epoch": 63.89490932925497,
+ "grad_norm": 0.18932299315929413,
+ "learning_rate": 0.0006,
+ "loss": 4.621096420288086,
+ "step": 4600
+ },
+ {
+ "epoch": 64.0,
+ "eval_loss": 3.81069016456604,
+ "eval_runtime": 19.8746,
+ "eval_samples_per_second": 122.82,
+ "eval_steps_per_second": 15.397,
+ "step": 4608
+ },
+ {
+ "epoch": 64.02796591653922,
+ "grad_norm": 0.23000451922416687,
+ "learning_rate": 0.0006,
+ "loss": 4.658928298950196,
+ "step": 4610
+ },
+ {
+ "epoch": 64.16779549923531,
+ "grad_norm": 0.202028289437294,
+ "learning_rate": 0.0006,
+ "loss": 4.644448089599609,
+ "step": 4620
+ },
+ {
+ "epoch": 64.30762508193139,
+ "grad_norm": 0.18988314270973206,
+ "learning_rate": 0.0006,
+ "loss": 4.666038131713867,
+ "step": 4630
+ },
+ {
+ "epoch": 64.44745466462749,
+ "grad_norm": 0.22012928128242493,
+ "learning_rate": 0.0006,
+ "loss": 4.63026123046875,
+ "step": 4640
+ },
+ {
+ "epoch": 64.58728424732358,
+ "grad_norm": 0.2103138417005539,
+ "learning_rate": 0.0006,
+ "loss": 4.656526947021485,
+ "step": 4650
+ },
+ {
+ "epoch": 64.72711383001966,
+ "grad_norm": 0.18606559932231903,
+ "learning_rate": 0.0006,
+ "loss": 4.646159744262695,
+ "step": 4660
+ },
+ {
+ "epoch": 64.86694341271576,
+ "grad_norm": 0.18499816954135895,
+ "learning_rate": 0.0006,
+ "loss": 4.637837982177734,
+ "step": 4670
+ },
+ {
+ "epoch": 65.0,
+ "grad_norm": 0.2283986210823059,
+ "learning_rate": 0.0006,
+ "loss": 4.624805068969726,
+ "step": 4680
+ },
+ {
+ "epoch": 65.0,
+ "eval_loss": 3.8084254264831543,
+ "eval_runtime": 19.8787,
+ "eval_samples_per_second": 122.795,
+ "eval_steps_per_second": 15.393,
+ "step": 4680
+ },
+ {
+ "epoch": 65.1398295826961,
+ "grad_norm": 0.20705945789813995,
+ "learning_rate": 0.0006,
+ "loss": 4.589903259277344,
+ "step": 4690
+ },
+ {
+ "epoch": 65.27965916539218,
+ "grad_norm": 0.19774095714092255,
+ "learning_rate": 0.0006,
+ "loss": 4.6366527557373045,
+ "step": 4700
+ },
+ {
+ "epoch": 65.41948874808827,
+ "grad_norm": 0.2037009745836258,
+ "learning_rate": 0.0006,
+ "loss": 4.64111442565918,
+ "step": 4710
+ },
+ {
+ "epoch": 65.55931833078435,
+ "grad_norm": 0.1995818018913269,
+ "learning_rate": 0.0006,
+ "loss": 4.640586853027344,
+ "step": 4720
+ },
+ {
+ "epoch": 65.69914791348045,
+ "grad_norm": 0.19600741565227509,
+ "learning_rate": 0.0006,
+ "loss": 4.629389953613281,
+ "step": 4730
+ },
+ {
+ "epoch": 65.83897749617654,
+ "grad_norm": 0.19653038680553436,
+ "learning_rate": 0.0006,
+ "loss": 4.641580200195312,
+ "step": 4740
+ },
+ {
+ "epoch": 65.97880707887262,
+ "grad_norm": 0.1832682341337204,
+ "learning_rate": 0.0006,
+ "loss": 4.654794311523437,
+ "step": 4750
+ },
+ {
+ "epoch": 66.0,
+ "eval_loss": 3.810702085494995,
+ "eval_runtime": 19.8737,
+ "eval_samples_per_second": 122.826,
+ "eval_steps_per_second": 15.397,
+ "step": 4752
+ },
+ {
+ "epoch": 66.11186366615686,
+ "grad_norm": 0.22284665703773499,
+ "learning_rate": 0.0006,
+ "loss": 4.632387924194336,
+ "step": 4760
+ },
+ {
+ "epoch": 66.25169324885296,
+ "grad_norm": 0.21953873336315155,
+ "learning_rate": 0.0006,
+ "loss": 4.613433837890625,
+ "step": 4770
+ },
+ {
+ "epoch": 66.39152283154905,
+ "grad_norm": 0.19770269095897675,
+ "learning_rate": 0.0006,
+ "loss": 4.62382926940918,
+ "step": 4780
+ },
+ {
+ "epoch": 66.53135241424513,
+ "grad_norm": 0.19758054614067078,
+ "learning_rate": 0.0006,
+ "loss": 4.615694427490235,
+ "step": 4790
+ },
+ {
+ "epoch": 66.67118199694123,
+ "grad_norm": 0.1848340779542923,
+ "learning_rate": 0.0006,
+ "loss": 4.646462631225586,
+ "step": 4800
+ },
+ {
+ "epoch": 66.81101157963732,
+ "grad_norm": 0.20013093948364258,
+ "learning_rate": 0.0006,
+ "loss": 4.6562744140625,
+ "step": 4810
+ },
+ {
+ "epoch": 66.9508411623334,
+ "grad_norm": 0.22081327438354492,
+ "learning_rate": 0.0006,
+ "loss": 4.654232406616211,
+ "step": 4820
+ },
+ {
+ "epoch": 67.0,
+ "eval_loss": 3.81170392036438,
+ "eval_runtime": 19.8665,
+ "eval_samples_per_second": 122.87,
+ "eval_steps_per_second": 15.403,
+ "step": 4824
+ },
+ {
+ "epoch": 67.08389774961765,
+ "grad_norm": 0.24410399794578552,
+ "learning_rate": 0.0006,
+ "loss": 4.6375988006591795,
+ "step": 4830
+ },
+ {
+ "epoch": 67.22372733231374,
+ "grad_norm": 0.20865154266357422,
+ "learning_rate": 0.0006,
+ "loss": 4.616934967041016,
+ "step": 4840
+ },
+ {
+ "epoch": 67.36355691500984,
+ "grad_norm": 0.22167667746543884,
+ "learning_rate": 0.0006,
+ "loss": 4.61735954284668,
+ "step": 4850
+ },
+ {
+ "epoch": 67.50338649770592,
+ "grad_norm": 0.18847821652889252,
+ "learning_rate": 0.0006,
+ "loss": 4.631498336791992,
+ "step": 4860
+ },
+ {
+ "epoch": 67.64321608040201,
+ "grad_norm": 0.25281551480293274,
+ "learning_rate": 0.0006,
+ "loss": 4.645813751220703,
+ "step": 4870
+ },
+ {
+ "epoch": 67.7830456630981,
+ "grad_norm": 0.192857027053833,
+ "learning_rate": 0.0006,
+ "loss": 4.635842895507812,
+ "step": 4880
+ },
+ {
+ "epoch": 67.92287524579419,
+ "grad_norm": 0.19587306678295135,
+ "learning_rate": 0.0006,
+ "loss": 4.634440994262695,
+ "step": 4890
+ },
+ {
+ "epoch": 68.0,
+ "eval_loss": 3.8126003742218018,
+ "eval_runtime": 19.868,
+ "eval_samples_per_second": 122.861,
+ "eval_steps_per_second": 15.402,
+ "step": 4896
+ },
+ {
+ "epoch": 68.05593183307843,
+ "grad_norm": 0.1873190850019455,
+ "learning_rate": 0.0006,
+ "loss": 4.619278335571289,
+ "step": 4900
+ },
+ {
+ "epoch": 68.19576141577453,
+ "grad_norm": 0.2564001977443695,
+ "learning_rate": 0.0006,
+ "loss": 4.619945526123047,
+ "step": 4910
+ },
+ {
+ "epoch": 68.33559099847061,
+ "grad_norm": 0.23214736580848694,
+ "learning_rate": 0.0006,
+ "loss": 4.614174270629883,
+ "step": 4920
+ },
+ {
+ "epoch": 68.4754205811667,
+ "grad_norm": 0.23541651666164398,
+ "learning_rate": 0.0006,
+ "loss": 4.643715286254883,
+ "step": 4930
+ },
+ {
+ "epoch": 68.6152501638628,
+ "grad_norm": 0.19236932694911957,
+ "learning_rate": 0.0006,
+ "loss": 4.646054458618164,
+ "step": 4940
+ },
+ {
+ "epoch": 68.75507974655888,
+ "grad_norm": 0.18536677956581116,
+ "learning_rate": 0.0006,
+ "loss": 4.626923751831055,
+ "step": 4950
+ },
+ {
+ "epoch": 68.89490932925497,
+ "grad_norm": 0.22821229696273804,
+ "learning_rate": 0.0006,
+ "loss": 4.64448356628418,
+ "step": 4960
+ },
+ {
+ "epoch": 69.0,
+ "eval_loss": 3.814791679382324,
+ "eval_runtime": 19.8683,
+ "eval_samples_per_second": 122.859,
+ "eval_steps_per_second": 15.401,
+ "step": 4968
+ },
+ {
+ "epoch": 69.02796591653922,
+ "grad_norm": 0.20987339317798615,
+ "learning_rate": 0.0006,
+ "loss": 4.643074417114258,
+ "step": 4970
+ },
+ {
+ "epoch": 69.16779549923531,
+ "grad_norm": 0.19386881589889526,
+ "learning_rate": 0.0006,
+ "loss": 4.615528869628906,
+ "step": 4980
+ },
+ {
+ "epoch": 69.30762508193139,
+ "grad_norm": 0.2236853390932083,
+ "learning_rate": 0.0006,
+ "loss": 4.620489120483398,
+ "step": 4990
+ },
+ {
+ "epoch": 69.44745466462749,
+ "grad_norm": 0.19033971428871155,
+ "learning_rate": 0.0006,
+ "loss": 4.631692504882812,
+ "step": 5000
+ },
+ {
+ "epoch": 69.58728424732358,
+ "grad_norm": 0.2054239958524704,
+ "learning_rate": 0.0006,
+ "loss": 4.637352371215821,
+ "step": 5010
+ },
+ {
+ "epoch": 69.72711383001966,
+ "grad_norm": 0.23273080587387085,
+ "learning_rate": 0.0006,
+ "loss": 4.636100769042969,
+ "step": 5020
+ },
+ {
+ "epoch": 69.86694341271576,
+ "grad_norm": 0.20068477094173431,
+ "learning_rate": 0.0006,
+ "loss": 4.616515731811523,
+ "step": 5030
+ },
+ {
+ "epoch": 70.0,
+ "grad_norm": 0.23735693097114563,
+ "learning_rate": 0.0006,
+ "loss": 4.660518646240234,
+ "step": 5040
+ },
+ {
+ "epoch": 70.0,
+ "eval_loss": 3.813115358352661,
+ "eval_runtime": 19.8767,
+ "eval_samples_per_second": 122.807,
+ "eval_steps_per_second": 15.395,
+ "step": 5040
+ },
+ {
+ "epoch": 70.1398295826961,
+ "grad_norm": 0.20828500390052795,
+ "learning_rate": 0.0006,
+ "loss": 4.624510955810547,
+ "step": 5050
+ },
+ {
+ "epoch": 70.27965916539218,
+ "grad_norm": 0.21619421243667603,
+ "learning_rate": 0.0006,
+ "loss": 4.616556549072266,
+ "step": 5060
+ },
+ {
+ "epoch": 70.41948874808827,
+ "grad_norm": 0.2034153938293457,
+ "learning_rate": 0.0006,
+ "loss": 4.606290054321289,
+ "step": 5070
+ },
+ {
+ "epoch": 70.55931833078435,
+ "grad_norm": 0.19158920645713806,
+ "learning_rate": 0.0006,
+ "loss": 4.613040542602539,
+ "step": 5080
+ },
+ {
+ "epoch": 70.69914791348045,
+ "grad_norm": 0.2007114142179489,
+ "learning_rate": 0.0006,
+ "loss": 4.618235778808594,
+ "step": 5090
+ },
+ {
+ "epoch": 70.83897749617654,
+ "grad_norm": 0.22325071692466736,
+ "learning_rate": 0.0006,
+ "loss": 4.622517013549805,
+ "step": 5100
+ },
+ {
+ "epoch": 70.97880707887262,
+ "grad_norm": 0.19615410268306732,
+ "learning_rate": 0.0006,
+ "loss": 4.647401428222656,
+ "step": 5110
+ },
+ {
+ "epoch": 71.0,
+ "eval_loss": 3.808810234069824,
+ "eval_runtime": 19.8644,
+ "eval_samples_per_second": 122.883,
+ "eval_steps_per_second": 15.404,
+ "step": 5112
+ },
+ {
+ "epoch": 71.11186366615686,
+ "grad_norm": 0.20997561514377594,
+ "learning_rate": 0.0006,
+ "loss": 4.628616714477539,
+ "step": 5120
+ },
+ {
+ "epoch": 71.25169324885296,
+ "grad_norm": 0.19207453727722168,
+ "learning_rate": 0.0006,
+ "loss": 4.596881103515625,
+ "step": 5130
+ },
+ {
+ "epoch": 71.39152283154905,
+ "grad_norm": 0.1934022605419159,
+ "learning_rate": 0.0006,
+ "loss": 4.5937644958496096,
+ "step": 5140
+ },
+ {
+ "epoch": 71.53135241424513,
+ "grad_norm": 0.20189517736434937,
+ "learning_rate": 0.0006,
+ "loss": 4.608246612548828,
+ "step": 5150
+ },
+ {
+ "epoch": 71.67118199694123,
+ "grad_norm": 0.21113520860671997,
+ "learning_rate": 0.0006,
+ "loss": 4.6216388702392575,
+ "step": 5160
+ },
+ {
+ "epoch": 71.81101157963732,
+ "grad_norm": 0.20459023118019104,
+ "learning_rate": 0.0006,
+ "loss": 4.6358898162841795,
+ "step": 5170
+ },
+ {
+ "epoch": 71.9508411623334,
+ "grad_norm": 0.19118602573871613,
+ "learning_rate": 0.0006,
+ "loss": 4.630368423461914,
+ "step": 5180
+ },
+ {
+ "epoch": 72.0,
+ "eval_loss": 3.804443359375,
+ "eval_runtime": 19.8611,
+ "eval_samples_per_second": 122.904,
+ "eval_steps_per_second": 15.407,
+ "step": 5184
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 7200,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 100,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 2.2085533024156385e+18,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/run.lock b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/run.lock
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/train.log b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..c2003f50d42299cb038ac44856e5bc5f4f437fc4
--- /dev/null
+++ b/runs/seeds/t1a_S2_C3_rand5_r2l50_i5exp_s44/train.log
@@ -0,0 +1,16294 @@
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 100
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 8
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 64
+[transformers] Total optimization steps = 7,200
+[transformers] Number of trainable parameters = 145,836,544
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: setting up run 3zo08n8j
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260917_184808-3zo08n8j
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t1a_S2_C3_rand5_r2l50_i5exp_s44
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/3zo08n8j
+
0%| | 0/7200 [00:00system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/t0/t0_S3_lr0.0006/checkpoint-864/config.json b/runs/t0/t0_S3_lr0.0006/checkpoint-864/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..edf37b33d02b391ecdf2ca887c98eec60d9c7686
--- /dev/null
+++ b/runs/t0/t0_S3_lr0.0006/checkpoint-864/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 768,
+ "initializer_range": 0.02,
+ "intermediate_size": 2048,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 6,
+ "num_hidden_layers": 20,
+ "num_key_value_heads": 6,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151671
+}
diff --git a/runs/t0/t0_S3_lr0.0006/checkpoint-864/tokenizer_config.json b/runs/t0/t0_S3_lr0.0006/checkpoint-864/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..839a3743177a3ed77e1ffbae67520e8500ca0532
--- /dev/null
+++ b/runs/t0/t0_S3_lr0.0006/checkpoint-864/tokenizer_config.json
@@ -0,0 +1,19 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/t0/t0_S3_lr0.0006/checkpoint-864/trainer_state.json b/runs/t0/t0_S3_lr0.0006/checkpoint-864/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a5f53e3191901c7c5e0d6753fd0caf3ebae4c75a
--- /dev/null
+++ b/runs/t0/t0_S3_lr0.0006/checkpoint-864/trainer_state.json
@@ -0,0 +1,739 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 12.0,
+ "eval_steps": 72,
+ "global_step": 864,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013961605584642234,
+ "grad_norm": 2.611384153366089,
+ "learning_rate": 0.0,
+ "loss": 12.058305740356445,
+ "step": 1
+ },
+ {
+ "epoch": 0.13961605584642234,
+ "grad_norm": 2.055915355682373,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.514022827148438,
+ "step": 10
+ },
+ {
+ "epoch": 0.2792321116928447,
+ "grad_norm": 1.8875689506530762,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 10.464095306396484,
+ "step": 20
+ },
+ {
+ "epoch": 0.418848167539267,
+ "grad_norm": 1.523587942123413,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 9.40309829711914,
+ "step": 30
+ },
+ {
+ "epoch": 0.5584642233856894,
+ "grad_norm": 0.6947948336601257,
+ "learning_rate": 0.000234,
+ "loss": 8.224795532226562,
+ "step": 40
+ },
+ {
+ "epoch": 0.6980802792321117,
+ "grad_norm": 0.3216397166252136,
+ "learning_rate": 0.000294,
+ "loss": 7.628060913085937,
+ "step": 50
+ },
+ {
+ "epoch": 0.837696335078534,
+ "grad_norm": 0.552848219871521,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 7.487551879882813,
+ "step": 60
+ },
+ {
+ "epoch": 0.9773123909249564,
+ "grad_norm": 1.7517889738082886,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 7.324363708496094,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.211400032043457,
+ "eval_runtime": 28.716,
+ "eval_samples_per_second": 85.005,
+ "eval_steps_per_second": 1.358,
+ "step": 72
+ },
+ {
+ "epoch": 1.111692844677138,
+ "grad_norm": 0.5828327536582947,
+ "learning_rate": 0.000474,
+ "loss": 7.129631042480469,
+ "step": 80
+ },
+ {
+ "epoch": 1.2513089005235603,
+ "grad_norm": 0.8480843305587769,
+ "learning_rate": 0.000534,
+ "loss": 6.9373291015625,
+ "step": 90
+ },
+ {
+ "epoch": 1.3909249563699826,
+ "grad_norm": 0.696884036064148,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 6.788343048095703,
+ "step": 100
+ },
+ {
+ "epoch": 1.530541012216405,
+ "grad_norm": 0.7813565731048584,
+ "learning_rate": 0.0006,
+ "loss": 6.701974487304687,
+ "step": 110
+ },
+ {
+ "epoch": 1.6701570680628273,
+ "grad_norm": 1.1431225538253784,
+ "learning_rate": 0.0006,
+ "loss": 6.54716796875,
+ "step": 120
+ },
+ {
+ "epoch": 1.8097731239092496,
+ "grad_norm": 0.64098060131073,
+ "learning_rate": 0.0006,
+ "loss": 6.435434722900391,
+ "step": 130
+ },
+ {
+ "epoch": 1.949389179755672,
+ "grad_norm": 0.3750019967556,
+ "learning_rate": 0.0006,
+ "loss": 6.350676727294922,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 6.276365280151367,
+ "eval_runtime": 4.105,
+ "eval_samples_per_second": 594.638,
+ "eval_steps_per_second": 9.501,
+ "step": 144
+ },
+ {
+ "epoch": 2.0837696335078535,
+ "grad_norm": 0.5605656504631042,
+ "learning_rate": 0.0006,
+ "loss": 6.234184265136719,
+ "step": 150
+ },
+ {
+ "epoch": 2.223385689354276,
+ "grad_norm": 0.41968899965286255,
+ "learning_rate": 0.0006,
+ "loss": 6.15125732421875,
+ "step": 160
+ },
+ {
+ "epoch": 2.363001745200698,
+ "grad_norm": 0.592795193195343,
+ "learning_rate": 0.0006,
+ "loss": 6.060428619384766,
+ "step": 170
+ },
+ {
+ "epoch": 2.5026178010471205,
+ "grad_norm": 0.88614422082901,
+ "learning_rate": 0.0006,
+ "loss": 6.083383941650391,
+ "step": 180
+ },
+ {
+ "epoch": 2.642233856893543,
+ "grad_norm": 0.3843565285205841,
+ "learning_rate": 0.0006,
+ "loss": 5.966249465942383,
+ "step": 190
+ },
+ {
+ "epoch": 2.781849912739965,
+ "grad_norm": 1.1335476636886597,
+ "learning_rate": 0.0006,
+ "loss": 5.871111297607422,
+ "step": 200
+ },
+ {
+ "epoch": 2.9214659685863875,
+ "grad_norm": 0.5506201982498169,
+ "learning_rate": 0.0006,
+ "loss": 5.832624816894532,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 5.773789405822754,
+ "eval_runtime": 4.0949,
+ "eval_samples_per_second": 596.11,
+ "eval_steps_per_second": 9.524,
+ "step": 216
+ },
+ {
+ "epoch": 3.055846422338569,
+ "grad_norm": 0.39692822098731995,
+ "learning_rate": 0.0006,
+ "loss": 5.73065414428711,
+ "step": 220
+ },
+ {
+ "epoch": 3.195462478184991,
+ "grad_norm": 0.9978296756744385,
+ "learning_rate": 0.0006,
+ "loss": 5.671360778808594,
+ "step": 230
+ },
+ {
+ "epoch": 3.3350785340314135,
+ "grad_norm": 0.3980925381183624,
+ "learning_rate": 0.0006,
+ "loss": 5.656559753417969,
+ "step": 240
+ },
+ {
+ "epoch": 3.474694589877836,
+ "grad_norm": 0.22790172696113586,
+ "learning_rate": 0.0006,
+ "loss": 5.569467163085937,
+ "step": 250
+ },
+ {
+ "epoch": 3.614310645724258,
+ "grad_norm": 0.5183550119400024,
+ "learning_rate": 0.0006,
+ "loss": 5.4890899658203125,
+ "step": 260
+ },
+ {
+ "epoch": 3.7539267015706805,
+ "grad_norm": 0.6049178838729858,
+ "learning_rate": 0.0006,
+ "loss": 5.470254135131836,
+ "step": 270
+ },
+ {
+ "epoch": 3.893542757417103,
+ "grad_norm": 0.23422406613826752,
+ "learning_rate": 0.0006,
+ "loss": 5.389688110351562,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 5.350030899047852,
+ "eval_runtime": 4.1141,
+ "eval_samples_per_second": 593.321,
+ "eval_steps_per_second": 9.48,
+ "step": 288
+ },
+ {
+ "epoch": 4.027923211169284,
+ "grad_norm": 0.7217665910720825,
+ "learning_rate": 0.0006,
+ "loss": 5.291028594970703,
+ "step": 290
+ },
+ {
+ "epoch": 4.167539267015707,
+ "grad_norm": 0.6471975445747375,
+ "learning_rate": 0.0006,
+ "loss": 5.301509094238281,
+ "step": 300
+ },
+ {
+ "epoch": 4.307155322862129,
+ "grad_norm": 0.27962714433670044,
+ "learning_rate": 0.0006,
+ "loss": 5.2473197937011715,
+ "step": 310
+ },
+ {
+ "epoch": 4.446771378708552,
+ "grad_norm": 0.19167020916938782,
+ "learning_rate": 0.0006,
+ "loss": 5.154059600830078,
+ "step": 320
+ },
+ {
+ "epoch": 4.5863874345549736,
+ "grad_norm": 0.5546616911888123,
+ "learning_rate": 0.0006,
+ "loss": 5.114247894287109,
+ "step": 330
+ },
+ {
+ "epoch": 4.726003490401396,
+ "grad_norm": 0.7748067378997803,
+ "learning_rate": 0.0006,
+ "loss": 5.056869125366211,
+ "step": 340
+ },
+ {
+ "epoch": 4.865619546247818,
+ "grad_norm": 0.4927058517932892,
+ "learning_rate": 0.0006,
+ "loss": 5.059400177001953,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.19630888104438782,
+ "learning_rate": 0.0006,
+ "loss": 4.985496520996094,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 5.01917839050293,
+ "eval_runtime": 4.1016,
+ "eval_samples_per_second": 595.139,
+ "eval_steps_per_second": 9.509,
+ "step": 360
+ },
+ {
+ "epoch": 5.139616055846423,
+ "grad_norm": 0.24849317967891693,
+ "learning_rate": 0.0006,
+ "loss": 4.859331893920898,
+ "step": 370
+ },
+ {
+ "epoch": 5.279232111692845,
+ "grad_norm": 0.5930282473564148,
+ "learning_rate": 0.0006,
+ "loss": 4.8379966735839846,
+ "step": 380
+ },
+ {
+ "epoch": 5.418848167539267,
+ "grad_norm": 0.7550261616706848,
+ "learning_rate": 0.0006,
+ "loss": 4.835939025878906,
+ "step": 390
+ },
+ {
+ "epoch": 5.558464223385689,
+ "grad_norm": 0.3209051787853241,
+ "learning_rate": 0.0006,
+ "loss": 4.792094421386719,
+ "step": 400
+ },
+ {
+ "epoch": 5.698080279232112,
+ "grad_norm": 0.20111848413944244,
+ "learning_rate": 0.0006,
+ "loss": 4.7063037872314455,
+ "step": 410
+ },
+ {
+ "epoch": 5.837696335078534,
+ "grad_norm": 0.5071753859519958,
+ "learning_rate": 0.0006,
+ "loss": 4.655247497558594,
+ "step": 420
+ },
+ {
+ "epoch": 5.977312390924956,
+ "grad_norm": 0.5023978352546692,
+ "learning_rate": 0.0006,
+ "loss": 4.654132843017578,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 4.724213123321533,
+ "eval_runtime": 4.1131,
+ "eval_samples_per_second": 593.47,
+ "eval_steps_per_second": 9.482,
+ "step": 432
+ },
+ {
+ "epoch": 6.111692844677138,
+ "grad_norm": 0.24052776396274567,
+ "learning_rate": 0.0006,
+ "loss": 4.563269805908203,
+ "step": 440
+ },
+ {
+ "epoch": 6.2513089005235605,
+ "grad_norm": 0.44461575150489807,
+ "learning_rate": 0.0006,
+ "loss": 4.4963134765625,
+ "step": 450
+ },
+ {
+ "epoch": 6.390924956369982,
+ "grad_norm": 0.4214881658554077,
+ "learning_rate": 0.0006,
+ "loss": 4.46234245300293,
+ "step": 460
+ },
+ {
+ "epoch": 6.530541012216405,
+ "grad_norm": 0.3929758071899414,
+ "learning_rate": 0.0006,
+ "loss": 4.436433410644531,
+ "step": 470
+ },
+ {
+ "epoch": 6.670157068062827,
+ "grad_norm": 0.45893657207489014,
+ "learning_rate": 0.0006,
+ "loss": 4.399936676025391,
+ "step": 480
+ },
+ {
+ "epoch": 6.80977312390925,
+ "grad_norm": 0.5055980682373047,
+ "learning_rate": 0.0006,
+ "loss": 4.420112609863281,
+ "step": 490
+ },
+ {
+ "epoch": 6.949389179755672,
+ "grad_norm": 0.23332498967647552,
+ "learning_rate": 0.0006,
+ "loss": 4.359828186035156,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 4.46279239654541,
+ "eval_runtime": 4.1006,
+ "eval_samples_per_second": 595.277,
+ "eval_steps_per_second": 9.511,
+ "step": 504
+ },
+ {
+ "epoch": 7.0837696335078535,
+ "grad_norm": 0.4353914260864258,
+ "learning_rate": 0.0006,
+ "loss": 4.284747314453125,
+ "step": 510
+ },
+ {
+ "epoch": 7.223385689354275,
+ "grad_norm": 0.36101245880126953,
+ "learning_rate": 0.0006,
+ "loss": 4.24701156616211,
+ "step": 520
+ },
+ {
+ "epoch": 7.363001745200698,
+ "grad_norm": 0.5949675440788269,
+ "learning_rate": 0.0006,
+ "loss": 4.242698669433594,
+ "step": 530
+ },
+ {
+ "epoch": 7.50261780104712,
+ "grad_norm": 0.365689218044281,
+ "learning_rate": 0.0006,
+ "loss": 4.253655242919922,
+ "step": 540
+ },
+ {
+ "epoch": 7.642233856893543,
+ "grad_norm": 0.2188694030046463,
+ "learning_rate": 0.0006,
+ "loss": 4.2057655334472654,
+ "step": 550
+ },
+ {
+ "epoch": 7.781849912739965,
+ "grad_norm": 0.2015274465084076,
+ "learning_rate": 0.0006,
+ "loss": 4.171692657470703,
+ "step": 560
+ },
+ {
+ "epoch": 7.9214659685863875,
+ "grad_norm": 0.32952603697776794,
+ "learning_rate": 0.0006,
+ "loss": 4.146934890747071,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.314759731292725,
+ "eval_runtime": 4.098,
+ "eval_samples_per_second": 595.66,
+ "eval_steps_per_second": 9.517,
+ "step": 576
+ },
+ {
+ "epoch": 8.055846422338568,
+ "grad_norm": 0.473550021648407,
+ "learning_rate": 0.0006,
+ "loss": 4.101259994506836,
+ "step": 580
+ },
+ {
+ "epoch": 8.195462478184991,
+ "grad_norm": 0.7781897187232971,
+ "learning_rate": 0.0006,
+ "loss": 4.117421340942383,
+ "step": 590
+ },
+ {
+ "epoch": 8.335078534031414,
+ "grad_norm": 0.9461731314659119,
+ "learning_rate": 0.0006,
+ "loss": 4.236537933349609,
+ "step": 600
+ },
+ {
+ "epoch": 8.474694589877837,
+ "grad_norm": 0.5159804224967957,
+ "learning_rate": 0.0006,
+ "loss": 4.2641654968261715,
+ "step": 610
+ },
+ {
+ "epoch": 8.614310645724258,
+ "grad_norm": 0.18528199195861816,
+ "learning_rate": 0.0006,
+ "loss": 4.1504676818847654,
+ "step": 620
+ },
+ {
+ "epoch": 8.75392670157068,
+ "grad_norm": 0.146124929189682,
+ "learning_rate": 0.0006,
+ "loss": 4.078816223144531,
+ "step": 630
+ },
+ {
+ "epoch": 8.893542757417103,
+ "grad_norm": 0.15202854573726654,
+ "learning_rate": 0.0006,
+ "loss": 4.052716064453125,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.224334716796875,
+ "eval_runtime": 4.0986,
+ "eval_samples_per_second": 595.562,
+ "eval_steps_per_second": 9.515,
+ "step": 648
+ },
+ {
+ "epoch": 9.027923211169284,
+ "grad_norm": 0.17902769148349762,
+ "learning_rate": 0.0006,
+ "loss": 4.001434326171875,
+ "step": 650
+ },
+ {
+ "epoch": 9.167539267015707,
+ "grad_norm": 0.18123631179332733,
+ "learning_rate": 0.0006,
+ "loss": 3.9232330322265625,
+ "step": 660
+ },
+ {
+ "epoch": 9.30715532286213,
+ "grad_norm": 0.40141674876213074,
+ "learning_rate": 0.0006,
+ "loss": 3.919754409790039,
+ "step": 670
+ },
+ {
+ "epoch": 9.44677137870855,
+ "grad_norm": 0.18586568534374237,
+ "learning_rate": 0.0006,
+ "loss": 3.920806121826172,
+ "step": 680
+ },
+ {
+ "epoch": 9.586387434554974,
+ "grad_norm": 0.17802633345127106,
+ "learning_rate": 0.0006,
+ "loss": 3.8947032928466796,
+ "step": 690
+ },
+ {
+ "epoch": 9.726003490401396,
+ "grad_norm": 0.22717879712581635,
+ "learning_rate": 0.0006,
+ "loss": 3.8957778930664064,
+ "step": 700
+ },
+ {
+ "epoch": 9.86561954624782,
+ "grad_norm": 0.18440183997154236,
+ "learning_rate": 0.0006,
+ "loss": 3.8769763946533202,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.22135742008686066,
+ "learning_rate": 0.0006,
+ "loss": 3.8590770721435548,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.130075454711914,
+ "eval_runtime": 4.0974,
+ "eval_samples_per_second": 595.745,
+ "eval_steps_per_second": 9.518,
+ "step": 720
+ },
+ {
+ "epoch": 10.139616055846423,
+ "grad_norm": 0.41075822710990906,
+ "learning_rate": 0.0006,
+ "loss": 3.777687835693359,
+ "step": 730
+ },
+ {
+ "epoch": 10.279232111692846,
+ "grad_norm": 0.4864703118801117,
+ "learning_rate": 0.0006,
+ "loss": 3.8111328125,
+ "step": 740
+ },
+ {
+ "epoch": 10.418848167539267,
+ "grad_norm": 0.23122157156467438,
+ "learning_rate": 0.0006,
+ "loss": 3.795425796508789,
+ "step": 750
+ },
+ {
+ "epoch": 10.55846422338569,
+ "grad_norm": 0.21544651687145233,
+ "learning_rate": 0.0006,
+ "loss": 3.7907455444335936,
+ "step": 760
+ },
+ {
+ "epoch": 10.698080279232112,
+ "grad_norm": 0.18590371310710907,
+ "learning_rate": 0.0006,
+ "loss": 3.771350860595703,
+ "step": 770
+ },
+ {
+ "epoch": 10.837696335078533,
+ "grad_norm": 0.1815602034330368,
+ "learning_rate": 0.0006,
+ "loss": 3.766162872314453,
+ "step": 780
+ },
+ {
+ "epoch": 10.977312390924956,
+ "grad_norm": 0.2208927869796753,
+ "learning_rate": 0.0006,
+ "loss": 3.752302551269531,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.0743303298950195,
+ "eval_runtime": 4.1063,
+ "eval_samples_per_second": 594.446,
+ "eval_steps_per_second": 9.497,
+ "step": 792
+ },
+ {
+ "epoch": 11.111692844677139,
+ "grad_norm": 0.41857829689979553,
+ "learning_rate": 0.0006,
+ "loss": 3.67520751953125,
+ "step": 800
+ },
+ {
+ "epoch": 11.25130890052356,
+ "grad_norm": 0.5337375402450562,
+ "learning_rate": 0.0006,
+ "loss": 3.6692478179931642,
+ "step": 810
+ },
+ {
+ "epoch": 11.390924956369982,
+ "grad_norm": 0.3253119885921478,
+ "learning_rate": 0.0006,
+ "loss": 3.6862701416015624,
+ "step": 820
+ },
+ {
+ "epoch": 11.530541012216405,
+ "grad_norm": 0.22796425223350525,
+ "learning_rate": 0.0006,
+ "loss": 3.6763381958007812,
+ "step": 830
+ },
+ {
+ "epoch": 11.670157068062828,
+ "grad_norm": 0.19382311403751373,
+ "learning_rate": 0.0006,
+ "loss": 3.6609451293945314,
+ "step": 840
+ },
+ {
+ "epoch": 11.809773123909249,
+ "grad_norm": 0.19692926108837128,
+ "learning_rate": 0.0006,
+ "loss": 3.655015563964844,
+ "step": 850
+ },
+ {
+ "epoch": 11.949389179755672,
+ "grad_norm": 0.19138450920581818,
+ "learning_rate": 0.0006,
+ "loss": 3.6566497802734377,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.031644344329834,
+ "eval_runtime": 4.1133,
+ "eval_samples_per_second": 593.439,
+ "eval_steps_per_second": 9.481,
+ "step": 864
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 864,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 12,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": true
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 7.656447700673495e+17,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/t0/t0_S4_lr0.0003/checkpoint-864/chat_template.jinja b/runs/t0/t0_S4_lr0.0003/checkpoint-864/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..699ff8df401fe4788525e9c1f9b86a99eadd6230
--- /dev/null
+++ b/runs/t0/t0_S4_lr0.0003/checkpoint-864/chat_template.jinja
@@ -0,0 +1,85 @@
+{%- if tools %}
+ {{- '<|im_start|>system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/t0/t0_S4_lr0.0003/checkpoint-864/config.json b/runs/t0/t0_S4_lr0.0003/checkpoint-864/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..ef8760b121e086715460fb47a7acf8f81e61b283
--- /dev/null
+++ b/runs/t0/t0_S4_lr0.0003/checkpoint-864/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 2816,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 8,
+ "num_hidden_layers": 24,
+ "num_key_value_heads": 8,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151671
+}
diff --git a/runs/t0/t0_S4_lr0.0003/checkpoint-864/generation_config.json b/runs/t0/t0_S4_lr0.0003/checkpoint-864/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/t0/t0_S4_lr0.0003/checkpoint-864/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/t0/t0_S4_lr0.0003/checkpoint-864/tokenizer_config.json b/runs/t0/t0_S4_lr0.0003/checkpoint-864/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..839a3743177a3ed77e1ffbae67520e8500ca0532
--- /dev/null
+++ b/runs/t0/t0_S4_lr0.0003/checkpoint-864/tokenizer_config.json
@@ -0,0 +1,19 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/t0/t0_S4_lr0.0003/checkpoint-864/trainer_state.json b/runs/t0/t0_S4_lr0.0003/checkpoint-864/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..da5ebb343ec60232b037346027fda9e1065148f9
--- /dev/null
+++ b/runs/t0/t0_S4_lr0.0003/checkpoint-864/trainer_state.json
@@ -0,0 +1,739 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 12.0,
+ "eval_steps": 72,
+ "global_step": 864,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013961605584642234,
+ "grad_norm": 3.0108652114868164,
+ "learning_rate": 0.0,
+ "loss": 12.143596649169922,
+ "step": 1
+ },
+ {
+ "epoch": 0.13961605584642234,
+ "grad_norm": 2.5072195529937744,
+ "learning_rate": 2.6999999999999996e-05,
+ "loss": 11.637612236870659,
+ "step": 10
+ },
+ {
+ "epoch": 0.2792321116928447,
+ "grad_norm": 1.999590277671814,
+ "learning_rate": 5.6999999999999996e-05,
+ "loss": 10.444085693359375,
+ "step": 20
+ },
+ {
+ "epoch": 0.418848167539267,
+ "grad_norm": 1.7020442485809326,
+ "learning_rate": 8.699999999999999e-05,
+ "loss": 9.67947769165039,
+ "step": 30
+ },
+ {
+ "epoch": 0.5584642233856894,
+ "grad_norm": 1.142599105834961,
+ "learning_rate": 0.000117,
+ "loss": 8.673612976074219,
+ "step": 40
+ },
+ {
+ "epoch": 0.6980802792321117,
+ "grad_norm": 0.5679003000259399,
+ "learning_rate": 0.000147,
+ "loss": 7.807984924316406,
+ "step": 50
+ },
+ {
+ "epoch": 0.837696335078534,
+ "grad_norm": 0.9599733352661133,
+ "learning_rate": 0.00017699999999999997,
+ "loss": 7.524662780761719,
+ "step": 60
+ },
+ {
+ "epoch": 0.9773123909249564,
+ "grad_norm": 0.839605450630188,
+ "learning_rate": 0.00020699999999999996,
+ "loss": 7.358512878417969,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.235728740692139,
+ "eval_runtime": 29.225,
+ "eval_samples_per_second": 83.524,
+ "eval_steps_per_second": 1.334,
+ "step": 72
+ },
+ {
+ "epoch": 1.111692844677138,
+ "grad_norm": 0.9182751178741455,
+ "learning_rate": 0.000237,
+ "loss": 7.150469970703125,
+ "step": 80
+ },
+ {
+ "epoch": 1.2513089005235603,
+ "grad_norm": 1.30881667137146,
+ "learning_rate": 0.000267,
+ "loss": 6.937275695800781,
+ "step": 90
+ },
+ {
+ "epoch": 1.3909249563699826,
+ "grad_norm": 1.0096999406814575,
+ "learning_rate": 0.00029699999999999996,
+ "loss": 6.755093383789062,
+ "step": 100
+ },
+ {
+ "epoch": 1.530541012216405,
+ "grad_norm": 0.6322146058082581,
+ "learning_rate": 0.0003,
+ "loss": 6.605305480957031,
+ "step": 110
+ },
+ {
+ "epoch": 1.6701570680628273,
+ "grad_norm": 0.9903938174247742,
+ "learning_rate": 0.0003,
+ "loss": 6.469866943359375,
+ "step": 120
+ },
+ {
+ "epoch": 1.8097731239092496,
+ "grad_norm": 0.3880707025527954,
+ "learning_rate": 0.0003,
+ "loss": 6.343294525146485,
+ "step": 130
+ },
+ {
+ "epoch": 1.949389179755672,
+ "grad_norm": 1.379593014717102,
+ "learning_rate": 0.0003,
+ "loss": 6.234751892089844,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 6.195944786071777,
+ "eval_runtime": 5.45,
+ "eval_samples_per_second": 447.888,
+ "eval_steps_per_second": 7.156,
+ "step": 144
+ },
+ {
+ "epoch": 2.0837696335078535,
+ "grad_norm": 0.9042283892631531,
+ "learning_rate": 0.0003,
+ "loss": 6.154948425292969,
+ "step": 150
+ },
+ {
+ "epoch": 2.223385689354276,
+ "grad_norm": 0.46852168440818787,
+ "learning_rate": 0.0003,
+ "loss": 6.034410095214843,
+ "step": 160
+ },
+ {
+ "epoch": 2.363001745200698,
+ "grad_norm": 0.8142240047454834,
+ "learning_rate": 0.0003,
+ "loss": 5.959646224975586,
+ "step": 170
+ },
+ {
+ "epoch": 2.5026178010471205,
+ "grad_norm": 0.6119681000709534,
+ "learning_rate": 0.0003,
+ "loss": 5.900678253173828,
+ "step": 180
+ },
+ {
+ "epoch": 2.642233856893543,
+ "grad_norm": 0.6856432557106018,
+ "learning_rate": 0.0003,
+ "loss": 5.835366821289062,
+ "step": 190
+ },
+ {
+ "epoch": 2.781849912739965,
+ "grad_norm": 0.37913742661476135,
+ "learning_rate": 0.0003,
+ "loss": 5.72945556640625,
+ "step": 200
+ },
+ {
+ "epoch": 2.9214659685863875,
+ "grad_norm": 1.0967448949813843,
+ "learning_rate": 0.0003,
+ "loss": 5.672224426269532,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 5.731411457061768,
+ "eval_runtime": 5.4593,
+ "eval_samples_per_second": 447.13,
+ "eval_steps_per_second": 7.144,
+ "step": 216
+ },
+ {
+ "epoch": 3.055846422338569,
+ "grad_norm": 0.5412494540214539,
+ "learning_rate": 0.0003,
+ "loss": 5.653792190551758,
+ "step": 220
+ },
+ {
+ "epoch": 3.195462478184991,
+ "grad_norm": 0.2865621745586395,
+ "learning_rate": 0.0003,
+ "loss": 5.54034194946289,
+ "step": 230
+ },
+ {
+ "epoch": 3.3350785340314135,
+ "grad_norm": 0.940181314945221,
+ "learning_rate": 0.0003,
+ "loss": 5.451759338378906,
+ "step": 240
+ },
+ {
+ "epoch": 3.474694589877836,
+ "grad_norm": 0.41092240810394287,
+ "learning_rate": 0.0003,
+ "loss": 5.458601760864258,
+ "step": 250
+ },
+ {
+ "epoch": 3.614310645724258,
+ "grad_norm": 0.31816035509109497,
+ "learning_rate": 0.0003,
+ "loss": 5.354522705078125,
+ "step": 260
+ },
+ {
+ "epoch": 3.7539267015706805,
+ "grad_norm": 0.9422369599342346,
+ "learning_rate": 0.0003,
+ "loss": 5.291055297851562,
+ "step": 270
+ },
+ {
+ "epoch": 3.893542757417103,
+ "grad_norm": 0.7183839678764343,
+ "learning_rate": 0.0003,
+ "loss": 5.273518371582031,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 5.251414775848389,
+ "eval_runtime": 5.4426,
+ "eval_samples_per_second": 448.496,
+ "eval_steps_per_second": 7.166,
+ "step": 288
+ },
+ {
+ "epoch": 4.027923211169284,
+ "grad_norm": 0.33922079205513,
+ "learning_rate": 0.0003,
+ "loss": 5.1952766418457035,
+ "step": 290
+ },
+ {
+ "epoch": 4.167539267015707,
+ "grad_norm": 0.3060228228569031,
+ "learning_rate": 0.0003,
+ "loss": 5.100880432128906,
+ "step": 300
+ },
+ {
+ "epoch": 4.307155322862129,
+ "grad_norm": 0.810433566570282,
+ "learning_rate": 0.0003,
+ "loss": 5.039925384521484,
+ "step": 310
+ },
+ {
+ "epoch": 4.446771378708552,
+ "grad_norm": 0.8437997102737427,
+ "learning_rate": 0.0003,
+ "loss": 5.056494140625,
+ "step": 320
+ },
+ {
+ "epoch": 4.5863874345549736,
+ "grad_norm": 0.5162206888198853,
+ "learning_rate": 0.0003,
+ "loss": 5.043321990966797,
+ "step": 330
+ },
+ {
+ "epoch": 4.726003490401396,
+ "grad_norm": 0.339292049407959,
+ "learning_rate": 0.0003,
+ "loss": 4.952778625488281,
+ "step": 340
+ },
+ {
+ "epoch": 4.865619546247818,
+ "grad_norm": 0.47109776735305786,
+ "learning_rate": 0.0003,
+ "loss": 4.878589630126953,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.6511371731758118,
+ "learning_rate": 0.0003,
+ "loss": 4.85400161743164,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 4.929421424865723,
+ "eval_runtime": 5.4373,
+ "eval_samples_per_second": 448.933,
+ "eval_steps_per_second": 7.173,
+ "step": 360
+ },
+ {
+ "epoch": 5.139616055846423,
+ "grad_norm": 0.7077712416648865,
+ "learning_rate": 0.0003,
+ "loss": 4.800180053710937,
+ "step": 370
+ },
+ {
+ "epoch": 5.279232111692845,
+ "grad_norm": 0.34735652804374695,
+ "learning_rate": 0.0003,
+ "loss": 4.75678596496582,
+ "step": 380
+ },
+ {
+ "epoch": 5.418848167539267,
+ "grad_norm": 0.3806954026222229,
+ "learning_rate": 0.0003,
+ "loss": 4.681983184814453,
+ "step": 390
+ },
+ {
+ "epoch": 5.558464223385689,
+ "grad_norm": 0.5873163938522339,
+ "learning_rate": 0.0003,
+ "loss": 4.633249664306641,
+ "step": 400
+ },
+ {
+ "epoch": 5.698080279232112,
+ "grad_norm": 0.6403170228004456,
+ "learning_rate": 0.0003,
+ "loss": 4.635204315185547,
+ "step": 410
+ },
+ {
+ "epoch": 5.837696335078534,
+ "grad_norm": 0.37991443276405334,
+ "learning_rate": 0.0003,
+ "loss": 4.603053283691406,
+ "step": 420
+ },
+ {
+ "epoch": 5.977312390924956,
+ "grad_norm": 0.35368460416793823,
+ "learning_rate": 0.0003,
+ "loss": 4.547142028808594,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 4.637178421020508,
+ "eval_runtime": 5.4586,
+ "eval_samples_per_second": 447.181,
+ "eval_steps_per_second": 7.145,
+ "step": 432
+ },
+ {
+ "epoch": 6.111692844677138,
+ "grad_norm": 0.5179786682128906,
+ "learning_rate": 0.0003,
+ "loss": 4.461312103271484,
+ "step": 440
+ },
+ {
+ "epoch": 6.2513089005235605,
+ "grad_norm": 0.6137361526489258,
+ "learning_rate": 0.0003,
+ "loss": 4.4385627746582035,
+ "step": 450
+ },
+ {
+ "epoch": 6.390924956369982,
+ "grad_norm": 0.6672863960266113,
+ "learning_rate": 0.0003,
+ "loss": 4.426197052001953,
+ "step": 460
+ },
+ {
+ "epoch": 6.530541012216405,
+ "grad_norm": 0.3582875728607178,
+ "learning_rate": 0.0003,
+ "loss": 4.408975601196289,
+ "step": 470
+ },
+ {
+ "epoch": 6.670157068062827,
+ "grad_norm": 0.28696542978286743,
+ "learning_rate": 0.0003,
+ "loss": 4.354529190063476,
+ "step": 480
+ },
+ {
+ "epoch": 6.80977312390925,
+ "grad_norm": 0.39151230454444885,
+ "learning_rate": 0.0003,
+ "loss": 4.325887298583984,
+ "step": 490
+ },
+ {
+ "epoch": 6.949389179755672,
+ "grad_norm": 0.3595994710922241,
+ "learning_rate": 0.0003,
+ "loss": 4.289802551269531,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 4.437053203582764,
+ "eval_runtime": 5.459,
+ "eval_samples_per_second": 447.148,
+ "eval_steps_per_second": 7.144,
+ "step": 504
+ },
+ {
+ "epoch": 7.0837696335078535,
+ "grad_norm": 0.6919127106666565,
+ "learning_rate": 0.0003,
+ "loss": 4.235585403442383,
+ "step": 510
+ },
+ {
+ "epoch": 7.223385689354275,
+ "grad_norm": 0.904406726360321,
+ "learning_rate": 0.0003,
+ "loss": 4.264218139648437,
+ "step": 520
+ },
+ {
+ "epoch": 7.363001745200698,
+ "grad_norm": 0.9251596927642822,
+ "learning_rate": 0.0003,
+ "loss": 4.338154602050781,
+ "step": 530
+ },
+ {
+ "epoch": 7.50261780104712,
+ "grad_norm": 0.47216707468032837,
+ "learning_rate": 0.0003,
+ "loss": 4.318355560302734,
+ "step": 540
+ },
+ {
+ "epoch": 7.642233856893543,
+ "grad_norm": 0.22250303626060486,
+ "learning_rate": 0.0003,
+ "loss": 4.223075485229492,
+ "step": 550
+ },
+ {
+ "epoch": 7.781849912739965,
+ "grad_norm": 0.20152734220027924,
+ "learning_rate": 0.0003,
+ "loss": 4.170987701416015,
+ "step": 560
+ },
+ {
+ "epoch": 7.9214659685863875,
+ "grad_norm": 0.20165708661079407,
+ "learning_rate": 0.0003,
+ "loss": 4.1383930206298825,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.319216728210449,
+ "eval_runtime": 5.4427,
+ "eval_samples_per_second": 448.493,
+ "eval_steps_per_second": 7.166,
+ "step": 576
+ },
+ {
+ "epoch": 8.055846422338568,
+ "grad_norm": 0.4072367250919342,
+ "learning_rate": 0.0003,
+ "loss": 4.08062973022461,
+ "step": 580
+ },
+ {
+ "epoch": 8.195462478184991,
+ "grad_norm": 0.3458138406276703,
+ "learning_rate": 0.0003,
+ "loss": 4.027294158935547,
+ "step": 590
+ },
+ {
+ "epoch": 8.335078534031414,
+ "grad_norm": 0.35914185643196106,
+ "learning_rate": 0.0003,
+ "loss": 4.023153686523438,
+ "step": 600
+ },
+ {
+ "epoch": 8.474694589877837,
+ "grad_norm": 0.35142284631729126,
+ "learning_rate": 0.0003,
+ "loss": 4.002416610717773,
+ "step": 610
+ },
+ {
+ "epoch": 8.614310645724258,
+ "grad_norm": 0.2793391942977905,
+ "learning_rate": 0.0003,
+ "loss": 3.986408233642578,
+ "step": 620
+ },
+ {
+ "epoch": 8.75392670157068,
+ "grad_norm": 0.3329005241394043,
+ "learning_rate": 0.0003,
+ "loss": 3.9794677734375,
+ "step": 630
+ },
+ {
+ "epoch": 8.893542757417103,
+ "grad_norm": 0.32278040051460266,
+ "learning_rate": 0.0003,
+ "loss": 3.9819564819335938,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.219235897064209,
+ "eval_runtime": 5.4619,
+ "eval_samples_per_second": 446.916,
+ "eval_steps_per_second": 7.14,
+ "step": 648
+ },
+ {
+ "epoch": 9.027923211169284,
+ "grad_norm": 0.49991485476493835,
+ "learning_rate": 0.0003,
+ "loss": 3.944264221191406,
+ "step": 650
+ },
+ {
+ "epoch": 9.167539267015707,
+ "grad_norm": 1.1535576581954956,
+ "learning_rate": 0.0003,
+ "loss": 3.9001365661621095,
+ "step": 660
+ },
+ {
+ "epoch": 9.30715532286213,
+ "grad_norm": 1.0573060512542725,
+ "learning_rate": 0.0003,
+ "loss": 3.9769973754882812,
+ "step": 670
+ },
+ {
+ "epoch": 9.44677137870855,
+ "grad_norm": 0.33998435735702515,
+ "learning_rate": 0.0003,
+ "loss": 3.9572212219238283,
+ "step": 680
+ },
+ {
+ "epoch": 9.586387434554974,
+ "grad_norm": 0.2560954689979553,
+ "learning_rate": 0.0003,
+ "loss": 3.8848876953125,
+ "step": 690
+ },
+ {
+ "epoch": 9.726003490401396,
+ "grad_norm": 0.22545966506004333,
+ "learning_rate": 0.0003,
+ "loss": 3.864862060546875,
+ "step": 700
+ },
+ {
+ "epoch": 9.86561954624782,
+ "grad_norm": 0.2370660901069641,
+ "learning_rate": 0.0003,
+ "loss": 3.835978317260742,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.3350813388824463,
+ "learning_rate": 0.0003,
+ "loss": 3.81748046875,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.153279781341553,
+ "eval_runtime": 5.4429,
+ "eval_samples_per_second": 448.47,
+ "eval_steps_per_second": 7.165,
+ "step": 720
+ },
+ {
+ "epoch": 10.139616055846423,
+ "grad_norm": 0.6783317923545837,
+ "learning_rate": 0.0003,
+ "loss": 3.7290679931640627,
+ "step": 730
+ },
+ {
+ "epoch": 10.279232111692846,
+ "grad_norm": 0.6124918460845947,
+ "learning_rate": 0.0003,
+ "loss": 3.748491668701172,
+ "step": 740
+ },
+ {
+ "epoch": 10.418848167539267,
+ "grad_norm": 0.3630979359149933,
+ "learning_rate": 0.0003,
+ "loss": 3.7257305145263673,
+ "step": 750
+ },
+ {
+ "epoch": 10.55846422338569,
+ "grad_norm": 0.2911542057991028,
+ "learning_rate": 0.0003,
+ "loss": 3.7298557281494142,
+ "step": 760
+ },
+ {
+ "epoch": 10.698080279232112,
+ "grad_norm": 0.2577025890350342,
+ "learning_rate": 0.0003,
+ "loss": 3.713488006591797,
+ "step": 770
+ },
+ {
+ "epoch": 10.837696335078533,
+ "grad_norm": 0.2614780366420746,
+ "learning_rate": 0.0003,
+ "loss": 3.713165283203125,
+ "step": 780
+ },
+ {
+ "epoch": 10.977312390924956,
+ "grad_norm": 0.27999359369277954,
+ "learning_rate": 0.0003,
+ "loss": 3.7012664794921877,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.10227632522583,
+ "eval_runtime": 5.4422,
+ "eval_samples_per_second": 448.532,
+ "eval_steps_per_second": 7.166,
+ "step": 792
+ },
+ {
+ "epoch": 11.111692844677139,
+ "grad_norm": 0.8187613487243652,
+ "learning_rate": 0.0003,
+ "loss": 3.6175563812255858,
+ "step": 800
+ },
+ {
+ "epoch": 11.25130890052356,
+ "grad_norm": 0.6838123202323914,
+ "learning_rate": 0.0003,
+ "loss": 3.6162586212158203,
+ "step": 810
+ },
+ {
+ "epoch": 11.390924956369982,
+ "grad_norm": 0.5785560607910156,
+ "learning_rate": 0.0003,
+ "loss": 3.628948974609375,
+ "step": 820
+ },
+ {
+ "epoch": 11.530541012216405,
+ "grad_norm": 0.42999038100242615,
+ "learning_rate": 0.0003,
+ "loss": 3.6205848693847655,
+ "step": 830
+ },
+ {
+ "epoch": 11.670157068062828,
+ "grad_norm": 0.3406503200531006,
+ "learning_rate": 0.0003,
+ "loss": 3.5991973876953125,
+ "step": 840
+ },
+ {
+ "epoch": 11.809773123909249,
+ "grad_norm": 0.2777751386165619,
+ "learning_rate": 0.0003,
+ "loss": 3.5885921478271485,
+ "step": 850
+ },
+ {
+ "epoch": 11.949389179755672,
+ "grad_norm": 0.284288227558136,
+ "learning_rate": 0.0003,
+ "loss": 3.5896976470947264,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.083232879638672,
+ "eval_runtime": 5.4586,
+ "eval_samples_per_second": 447.18,
+ "eval_steps_per_second": 7.145,
+ "step": 864
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 864,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 12,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": true
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.6673046800042557e+18,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/t0/t0_S4_lr0.0003/train.log b/runs/t0/t0_S4_lr0.0003/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..6c1e61cae19ded0d60c4339ffa7729a23daee13f
--- /dev/null
+++ b/runs/t0/t0_S4_lr0.0003/train.log
@@ -0,0 +1,1666 @@
+W0817 20:22:56.546000 2250018 torch/distributed/run.py:851]
+W0817 20:22:56.546000 2250018 torch/distributed/run.py:851] *****************************************
+W0817 20:22:56.546000 2250018 torch/distributed/run.py:851] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
+W0817 20:22:56.546000 2250018 torch/distributed/run.py:851] *****************************************
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+EOS token: <|im_end|>
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 463642624
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 463642624
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 463642624
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 463642624
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 463642624
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 463642624
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 463642624
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 463642624
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 12
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 8
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 8
+[transformers] Total optimization steps = 864
+[transformers] Number of trainable parameters = 463,642,624
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: setting up run 7op1rnvz
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260817_202321-7op1rnvz
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t0_S4_lr0.0003
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/7op1rnvz
+
0%| | 0/864 [00:00 will be ignored
+[rank7]:W0817 20:23:44.644000 2250124 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[rank3]:W0817 20:23:44.716000 2250120 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank4]:W0817 20:23:44.873000 2250121 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank2]:W0817 20:23:44.897000 2250119 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank5]:W0817 20:23:45.002000 2250122 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[rank1]:W0817 20:23:45.256000 2250118 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank0]:W0817 20:23:45.777000 2250117 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+
0%| | 1/864 [01:03<15:19:45, 63.95s/it]
{'loss': '12.14', 'grad_norm': '3.011', 'learning_rate': '0', 'epoch': '0.01396'}
+
0%| | 1/864 [01:03<15:19:45, 63.95s/it]
0%| | 2/864 [01:05<6:29:30, 27.11s/it]
0%| | 3/864 [01:06<3:39:58, 15.33s/it]
0%| | 4/864 [01:07<2:20:20, 9.79s/it]
1%| | 5/864 [01:09<1:36:20, 6.73s/it]
1%| | 6/864 [01:10<1:09:50, 4.88s/it]
1%| | 7/864 [01:11<53:01, 3.71s/it]
1%| | 8/864 [01:13<42:00, 2.94s/it]
1%| | 9/864 [01:14<34:38, 2.43s/it]
1%| | 10/864 [01:15<29:37, 2.08s/it]
{'loss': '11.64', 'grad_norm': '2.507', 'learning_rate': '2.7e-05', 'epoch': '0.1396'}
+
1%| | 10/864 [01:15<29:37, 2.08s/it]
1%|▏ | 11/864 [01:16<26:12, 1.84s/it]
1%|▏ | 12/864 [01:18<23:49, 1.68s/it]
2%|▏ | 13/864 [01:19<22:10, 1.56s/it]
2%|▏ | 14/864 [01:20<21:01, 1.48s/it]
2%|▏ | 15/864 [01:22<20:12, 1.43s/it]
2%|▏ | 16/864 [01:23<19:38, 1.39s/it]
2%|▏ | 17/864 [01:24<19:13, 1.36s/it]
2%|▏ | 18/864 [01:26<18:56, 1.34s/it]
2%|▏ | 19/864 [01:27<18:43, 1.33s/it]
2%|▏ | 20/864 [01:28<18:35, 1.32s/it]
{'loss': '10.44', 'grad_norm': '2', 'learning_rate': '5.7e-05', 'epoch': '0.2792'}
+
2%|▏ | 20/864 [01:28<18:35, 1.32s/it]
2%|▏ | 21/864 [01:29<18:28, 1.31s/it]
3%|▎ | 22/864 [01:31<18:22, 1.31s/it]
3%|▎ | 23/864 [01:32<18:18, 1.31s/it]
3%|▎ | 24/864 [01:33<18:15, 1.30s/it]
3%|▎ | 25/864 [01:35<18:12, 1.30s/it]
3%|▎ | 26/864 [01:36<18:11, 1.30s/it]
3%|▎ | 27/864 [01:37<18:09, 1.30s/it]
3%|▎ | 28/864 [01:39<18:07, 1.30s/it]
3%|▎ | 29/864 [01:40<18:06, 1.30s/it]
3%|▎ | 30/864 [01:41<18:04, 1.30s/it]
{'loss': '9.679', 'grad_norm': '1.702', 'learning_rate': '8.7e-05', 'epoch': '0.4188'}
+
3%|▎ | 30/864 [01:41<18:04, 1.30s/it]
4%|▎ | 31/864 [01:42<18:03, 1.30s/it]
4%|▎ | 32/864 [01:44<18:01, 1.30s/it]
4%|▍ | 33/864 [01:45<18:00, 1.30s/it]
4%|▍ | 34/864 [01:46<17:59, 1.30s/it]
4%|▍ | 35/864 [01:48<17:57, 1.30s/it]
4%|▍ | 36/864 [01:49<17:56, 1.30s/it]
4%|▍ | 37/864 [01:50<17:55, 1.30s/it]
4%|▍ | 38/864 [01:52<17:53, 1.30s/it]
5%|▍ | 39/864 [01:53<17:52, 1.30s/it]
5%|▍ | 40/864 [01:54<17:51, 1.30s/it]
{'loss': '8.674', 'grad_norm': '1.143', 'learning_rate': '0.000117', 'epoch': '0.5585'}
+
5%|▍ | 40/864 [01:54<17:51, 1.30s/it]
5%|▍ | 41/864 [01:55<17:50, 1.30s/it]
5%|▍ | 42/864 [01:57<17:48, 1.30s/it]
5%|▍ | 43/864 [01:58<17:47, 1.30s/it]
5%|▌ | 44/864 [01:59<17:45, 1.30s/it]
5%|▌ | 45/864 [02:01<17:45, 1.30s/it]
5%|▌ | 46/864 [02:02<17:43, 1.30s/it]
5%|▌ | 47/864 [02:03<17:42, 1.30s/it]
6%|▌ | 48/864 [02:05<17:42, 1.30s/it]
6%|▌ | 49/864 [02:06<17:41, 1.30s/it]
6%|▌ | 50/864 [02:07<17:40, 1.30s/it]
{'loss': '7.808', 'grad_norm': '0.5679', 'learning_rate': '0.000147', 'epoch': '0.6981'}
+
6%|▌ | 50/864 [02:07<17:40, 1.30s/it]
6%|▌ | 51/864 [02:09<17:39, 1.30s/it]
6%|▌ | 52/864 [02:10<17:38, 1.30s/it]
6%|▌ | 53/864 [02:11<17:37, 1.30s/it]
6%|▋ | 54/864 [02:12<17:35, 1.30s/it]
6%|▋ | 55/864 [02:14<17:33, 1.30s/it]
6%|▋ | 56/864 [02:15<17:33, 1.30s/it]
7%|▋ | 57/864 [02:16<17:32, 1.30s/it]
7%|▋ | 58/864 [02:18<17:31, 1.30s/it]
7%|▋ | 59/864 [02:19<17:29, 1.30s/it]
7%|▋ | 60/864 [02:20<17:27, 1.30s/it]
{'loss': '7.525', 'grad_norm': '0.96', 'learning_rate': '0.000177', 'epoch': '0.8377'}
+
7%|▋ | 60/864 [02:20<17:27, 1.30s/it]
7%|▋ | 61/864 [02:22<17:27, 1.30s/it]
7%|▋ | 62/864 [02:23<17:26, 1.30s/it]
7%|▋ | 63/864 [02:24<17:25, 1.30s/it]
7%|▋ | 64/864 [02:25<17:25, 1.31s/it]
8%|▊ | 65/864 [02:27<17:23, 1.31s/it]
8%|▊ | 66/864 [02:28<17:23, 1.31s/it]
8%|▊ | 67/864 [02:29<17:22, 1.31s/it]
8%|▊ | 68/864 [02:31<17:21, 1.31s/it]
8%|▊ | 69/864 [02:32<17:20, 1.31s/it]
8%|▊ | 70/864 [02:33<17:20, 1.31s/it]
{'loss': '7.359', 'grad_norm': '0.8396', 'learning_rate': '0.000207', 'epoch': '0.9773'}
+
8%|▊ | 70/864 [02:33<17:20, 1.31s/it]
8%|▊ | 71/864 [02:35<17:18, 1.31s/it]
8%|▊ | 72/864 [02:35<15:23, 1.17s/it][transformers]
+***** Running Evaluation *****
+[transformers] Num examples = 2441
+[transformers] Batch size = 8
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+
+
0%| | 0/39 [00:00system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/t0/t0_S5_lr0.0006/checkpoint-648/config.json b/runs/t0/t0_S5_lr0.0006/checkpoint-648/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..58ed141896af03a1496676b060879ee607fff4bb
--- /dev/null
+++ b/runs/t0/t0_S5_lr0.0006/checkpoint-648/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1536,
+ "initializer_range": 0.02,
+ "intermediate_size": 4096,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 12,
+ "num_hidden_layers": 24,
+ "num_key_value_heads": 12,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151671
+}
diff --git a/runs/t0/t0_S5_lr0.0006/checkpoint-648/generation_config.json b/runs/t0/t0_S5_lr0.0006/checkpoint-648/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/t0/t0_S5_lr0.0006/checkpoint-648/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/t0/t0_S5_lr0.0006/checkpoint-648/tokenizer_config.json b/runs/t0/t0_S5_lr0.0006/checkpoint-648/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..839a3743177a3ed77e1ffbae67520e8500ca0532
--- /dev/null
+++ b/runs/t0/t0_S5_lr0.0006/checkpoint-648/tokenizer_config.json
@@ -0,0 +1,19 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/t0/t0_S5_lr0.0006/checkpoint-648/trainer_state.json b/runs/t0/t0_S5_lr0.0006/checkpoint-648/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b57d0d0be9f42fca98347b7f161b4d88692606c0
--- /dev/null
+++ b/runs/t0/t0_S5_lr0.0006/checkpoint-648/trainer_state.json
@@ -0,0 +1,561 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 9.0,
+ "eval_steps": 72,
+ "global_step": 648,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013973799126637555,
+ "grad_norm": 3.63700532913208,
+ "learning_rate": 0.0,
+ "loss": 12.25691032409668,
+ "step": 1
+ },
+ {
+ "epoch": 0.13973799126637554,
+ "grad_norm": 2.0760152339935303,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.061671786838108,
+ "step": 10
+ },
+ {
+ "epoch": 0.2794759825327511,
+ "grad_norm": 1.2159764766693115,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 9.3091796875,
+ "step": 20
+ },
+ {
+ "epoch": 0.4192139737991266,
+ "grad_norm": 0.6038848757743835,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 7.946849060058594,
+ "step": 30
+ },
+ {
+ "epoch": 0.5589519650655022,
+ "grad_norm": 1.1402862071990967,
+ "learning_rate": 0.000234,
+ "loss": 7.514804077148438,
+ "step": 40
+ },
+ {
+ "epoch": 0.6986899563318777,
+ "grad_norm": 1.3050649166107178,
+ "learning_rate": 0.000294,
+ "loss": 7.308438110351562,
+ "step": 50
+ },
+ {
+ "epoch": 0.8384279475982532,
+ "grad_norm": 1.3311524391174316,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 7.054753112792969,
+ "step": 60
+ },
+ {
+ "epoch": 0.9781659388646288,
+ "grad_norm": 1.1332180500030518,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 6.819429016113281,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 6.700946807861328,
+ "eval_runtime": 32.0489,
+ "eval_samples_per_second": 76.165,
+ "eval_steps_per_second": 2.403,
+ "step": 72
+ },
+ {
+ "epoch": 1.1117903930131003,
+ "grad_norm": 0.8679574131965637,
+ "learning_rate": 0.000474,
+ "loss": 6.635536956787109,
+ "step": 80
+ },
+ {
+ "epoch": 1.251528384279476,
+ "grad_norm": 1.000284194946289,
+ "learning_rate": 0.000534,
+ "loss": 6.494485473632812,
+ "step": 90
+ },
+ {
+ "epoch": 1.3912663755458516,
+ "grad_norm": 0.980858325958252,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 6.378277969360352,
+ "step": 100
+ },
+ {
+ "epoch": 1.531004366812227,
+ "grad_norm": 1.2506310939788818,
+ "learning_rate": 0.0006,
+ "loss": 6.2798408508300785,
+ "step": 110
+ },
+ {
+ "epoch": 1.6707423580786025,
+ "grad_norm": 0.4951258897781372,
+ "learning_rate": 0.0006,
+ "loss": 6.146320343017578,
+ "step": 120
+ },
+ {
+ "epoch": 1.8104803493449781,
+ "grad_norm": 0.566660463809967,
+ "learning_rate": 0.0006,
+ "loss": 6.022150039672852,
+ "step": 130
+ },
+ {
+ "epoch": 1.9502183406113538,
+ "grad_norm": 0.7159810662269592,
+ "learning_rate": 0.0006,
+ "loss": 5.944305038452148,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 5.880650520324707,
+ "eval_runtime": 7.9249,
+ "eval_samples_per_second": 308.015,
+ "eval_steps_per_second": 9.716,
+ "step": 144
+ },
+ {
+ "epoch": 2.0838427947598253,
+ "grad_norm": 0.7987729907035828,
+ "learning_rate": 0.0006,
+ "loss": 5.845639038085937,
+ "step": 150
+ },
+ {
+ "epoch": 2.2235807860262007,
+ "grad_norm": 0.35364511609077454,
+ "learning_rate": 0.0006,
+ "loss": 5.713481903076172,
+ "step": 160
+ },
+ {
+ "epoch": 2.3633187772925766,
+ "grad_norm": 0.6100532412528992,
+ "learning_rate": 0.0006,
+ "loss": 5.675626373291015,
+ "step": 170
+ },
+ {
+ "epoch": 2.503056768558952,
+ "grad_norm": 0.2931850552558899,
+ "learning_rate": 0.0006,
+ "loss": 5.587120819091797,
+ "step": 180
+ },
+ {
+ "epoch": 2.6427947598253274,
+ "grad_norm": 0.3854564428329468,
+ "learning_rate": 0.0006,
+ "loss": 5.475901412963867,
+ "step": 190
+ },
+ {
+ "epoch": 2.7825327510917033,
+ "grad_norm": 0.6930826306343079,
+ "learning_rate": 0.0006,
+ "loss": 5.407214736938476,
+ "step": 200
+ },
+ {
+ "epoch": 2.9222707423580787,
+ "grad_norm": 0.36691468954086304,
+ "learning_rate": 0.0006,
+ "loss": 5.332212829589844,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 5.250132083892822,
+ "eval_runtime": 7.9236,
+ "eval_samples_per_second": 308.066,
+ "eval_steps_per_second": 9.718,
+ "step": 216
+ },
+ {
+ "epoch": 3.05589519650655,
+ "grad_norm": 0.4266318082809448,
+ "learning_rate": 0.0006,
+ "loss": 5.183085250854492,
+ "step": 220
+ },
+ {
+ "epoch": 3.1956331877729256,
+ "grad_norm": 0.524971067905426,
+ "learning_rate": 0.0006,
+ "loss": 5.106776428222656,
+ "step": 230
+ },
+ {
+ "epoch": 3.3353711790393015,
+ "grad_norm": 0.28536364436149597,
+ "learning_rate": 0.0006,
+ "loss": 5.022960662841797,
+ "step": 240
+ },
+ {
+ "epoch": 3.475109170305677,
+ "grad_norm": 0.40682587027549744,
+ "learning_rate": 0.0006,
+ "loss": 4.939643478393554,
+ "step": 250
+ },
+ {
+ "epoch": 3.6148471615720523,
+ "grad_norm": 0.5927167534828186,
+ "learning_rate": 0.0006,
+ "loss": 4.890360260009766,
+ "step": 260
+ },
+ {
+ "epoch": 3.7545851528384278,
+ "grad_norm": 0.674172580242157,
+ "learning_rate": 0.0006,
+ "loss": 4.916819000244141,
+ "step": 270
+ },
+ {
+ "epoch": 3.8943231441048036,
+ "grad_norm": 0.32566314935684204,
+ "learning_rate": 0.0006,
+ "loss": 4.933982467651367,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 4.803877830505371,
+ "eval_runtime": 7.9356,
+ "eval_samples_per_second": 307.6,
+ "eval_steps_per_second": 9.703,
+ "step": 288
+ },
+ {
+ "epoch": 4.0279475982532755,
+ "grad_norm": 0.25885775685310364,
+ "learning_rate": 0.0006,
+ "loss": 4.72625961303711,
+ "step": 290
+ },
+ {
+ "epoch": 4.1676855895196505,
+ "grad_norm": 0.27622267603874207,
+ "learning_rate": 0.0006,
+ "loss": 4.565699768066406,
+ "step": 300
+ },
+ {
+ "epoch": 4.307423580786026,
+ "grad_norm": 0.34537044167518616,
+ "learning_rate": 0.0006,
+ "loss": 4.506320953369141,
+ "step": 310
+ },
+ {
+ "epoch": 4.447161572052401,
+ "grad_norm": 0.36128517985343933,
+ "learning_rate": 0.0006,
+ "loss": 4.436258316040039,
+ "step": 320
+ },
+ {
+ "epoch": 4.586899563318777,
+ "grad_norm": 0.36944088339805603,
+ "learning_rate": 0.0006,
+ "loss": 4.394168853759766,
+ "step": 330
+ },
+ {
+ "epoch": 4.726637554585153,
+ "grad_norm": 0.3527185320854187,
+ "learning_rate": 0.0006,
+ "loss": 4.368709564208984,
+ "step": 340
+ },
+ {
+ "epoch": 4.866375545851528,
+ "grad_norm": 0.2085314393043518,
+ "learning_rate": 0.0006,
+ "loss": 4.305722045898437,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.3686930239200592,
+ "learning_rate": 0.0006,
+ "loss": 4.263705444335938,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 4.388052463531494,
+ "eval_runtime": 7.9251,
+ "eval_samples_per_second": 308.007,
+ "eval_steps_per_second": 9.716,
+ "step": 360
+ },
+ {
+ "epoch": 5.139737991266376,
+ "grad_norm": 0.6086430549621582,
+ "learning_rate": 0.0006,
+ "loss": 4.14691047668457,
+ "step": 370
+ },
+ {
+ "epoch": 5.279475982532751,
+ "grad_norm": 0.3700956404209137,
+ "learning_rate": 0.0006,
+ "loss": 4.1726539611816404,
+ "step": 380
+ },
+ {
+ "epoch": 5.419213973799127,
+ "grad_norm": 0.20227950811386108,
+ "learning_rate": 0.0006,
+ "loss": 4.090512466430664,
+ "step": 390
+ },
+ {
+ "epoch": 5.558951965065502,
+ "grad_norm": 0.19781829416751862,
+ "learning_rate": 0.0006,
+ "loss": 4.018944549560547,
+ "step": 400
+ },
+ {
+ "epoch": 5.698689956331878,
+ "grad_norm": 0.15543371438980103,
+ "learning_rate": 0.0006,
+ "loss": 3.991476821899414,
+ "step": 410
+ },
+ {
+ "epoch": 5.8384279475982535,
+ "grad_norm": 0.3352770209312439,
+ "learning_rate": 0.0006,
+ "loss": 3.97552490234375,
+ "step": 420
+ },
+ {
+ "epoch": 5.978165938864628,
+ "grad_norm": 0.413346529006958,
+ "learning_rate": 0.0006,
+ "loss": 3.983864974975586,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 4.199523448944092,
+ "eval_runtime": 7.924,
+ "eval_samples_per_second": 308.051,
+ "eval_steps_per_second": 9.717,
+ "step": 432
+ },
+ {
+ "epoch": 6.1117903930131,
+ "grad_norm": 0.7355369925498962,
+ "learning_rate": 0.0006,
+ "loss": 3.909012222290039,
+ "step": 440
+ },
+ {
+ "epoch": 6.251528384279476,
+ "grad_norm": 0.7885279059410095,
+ "learning_rate": 0.0006,
+ "loss": 4.066497802734375,
+ "step": 450
+ },
+ {
+ "epoch": 6.391266375545851,
+ "grad_norm": 1.1330866813659668,
+ "learning_rate": 0.0006,
+ "loss": 4.440969848632813,
+ "step": 460
+ },
+ {
+ "epoch": 6.531004366812227,
+ "grad_norm": 0.8877047896385193,
+ "learning_rate": 0.0006,
+ "loss": 4.542274093627929,
+ "step": 470
+ },
+ {
+ "epoch": 6.670742358078603,
+ "grad_norm": 0.2488427311182022,
+ "learning_rate": 0.0006,
+ "loss": 4.350867462158203,
+ "step": 480
+ },
+ {
+ "epoch": 6.810480349344978,
+ "grad_norm": 0.16023924946784973,
+ "learning_rate": 0.0006,
+ "loss": 4.121267700195313,
+ "step": 490
+ },
+ {
+ "epoch": 6.950218340611354,
+ "grad_norm": 0.19913139939308167,
+ "learning_rate": 0.0006,
+ "loss": 3.9925003051757812,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 4.189695835113525,
+ "eval_runtime": 7.9274,
+ "eval_samples_per_second": 307.918,
+ "eval_steps_per_second": 9.713,
+ "step": 504
+ },
+ {
+ "epoch": 7.083842794759826,
+ "grad_norm": 0.3292243778705597,
+ "learning_rate": 0.0006,
+ "loss": 3.8803970336914064,
+ "step": 510
+ },
+ {
+ "epoch": 7.223580786026201,
+ "grad_norm": 0.20132949948310852,
+ "learning_rate": 0.0006,
+ "loss": 3.8114532470703124,
+ "step": 520
+ },
+ {
+ "epoch": 7.3633187772925766,
+ "grad_norm": 0.14111486077308655,
+ "learning_rate": 0.0006,
+ "loss": 3.7562381744384767,
+ "step": 530
+ },
+ {
+ "epoch": 7.503056768558952,
+ "grad_norm": 0.13241393864154816,
+ "learning_rate": 0.0006,
+ "loss": 3.7331714630126953,
+ "step": 540
+ },
+ {
+ "epoch": 7.642794759825327,
+ "grad_norm": 0.1305728554725647,
+ "learning_rate": 0.0006,
+ "loss": 3.715618133544922,
+ "step": 550
+ },
+ {
+ "epoch": 7.782532751091703,
+ "grad_norm": 0.1556539684534073,
+ "learning_rate": 0.0006,
+ "loss": 3.6987678527832033,
+ "step": 560
+ },
+ {
+ "epoch": 7.922270742358078,
+ "grad_norm": 0.15663324296474457,
+ "learning_rate": 0.0006,
+ "loss": 3.685735321044922,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.001772880554199,
+ "eval_runtime": 7.9284,
+ "eval_samples_per_second": 307.881,
+ "eval_steps_per_second": 9.712,
+ "step": 576
+ },
+ {
+ "epoch": 8.055895196506551,
+ "grad_norm": 0.31078657507896423,
+ "learning_rate": 0.0006,
+ "loss": 3.590984344482422,
+ "step": 580
+ },
+ {
+ "epoch": 8.195633187772925,
+ "grad_norm": 0.5049765110015869,
+ "learning_rate": 0.0006,
+ "loss": 3.536634826660156,
+ "step": 590
+ },
+ {
+ "epoch": 8.335371179039301,
+ "grad_norm": 0.2599044144153595,
+ "learning_rate": 0.0006,
+ "loss": 3.548571014404297,
+ "step": 600
+ },
+ {
+ "epoch": 8.475109170305677,
+ "grad_norm": 0.26677677035331726,
+ "learning_rate": 0.0006,
+ "loss": 3.5195777893066404,
+ "step": 610
+ },
+ {
+ "epoch": 8.614847161572053,
+ "grad_norm": 0.16387900710105896,
+ "learning_rate": 0.0006,
+ "loss": 3.501542663574219,
+ "step": 620
+ },
+ {
+ "epoch": 8.754585152838429,
+ "grad_norm": 0.1453789323568344,
+ "learning_rate": 0.0006,
+ "loss": 3.489454650878906,
+ "step": 630
+ },
+ {
+ "epoch": 8.894323144104803,
+ "grad_norm": 0.1279163956642151,
+ "learning_rate": 0.0006,
+ "loss": 3.4968822479248045,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 3.952561616897583,
+ "eval_runtime": 7.9265,
+ "eval_samples_per_second": 307.956,
+ "eval_steps_per_second": 9.714,
+ "step": 648
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 864,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 12,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 2.753608419788194e+18,
+ "train_batch_size": 4,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/t0/t0_S5_lr0.0006/train.log b/runs/t0/t0_S5_lr0.0006/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..aa8a022fd41325d1f5d82994a7ae05bb0a36e581
--- /dev/null
+++ b/runs/t0/t0_S5_lr0.0006/train.log
@@ -0,0 +1,2023 @@
+W0817 22:12:46.656000 2318409 torch/distributed/run.py:851]
+W0817 22:12:46.656000 2318409 torch/distributed/run.py:851] *****************************************
+W0817 22:12:46.656000 2318409 torch/distributed/run.py:851] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
+W0817 22:12:46.656000 2318409 torch/distributed/run.py:851] *****************************************
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 912519168
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+total params: 912519168
+[transformers] PyTorch: setting up devices
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 912519168
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 912519168
+total params:[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+912519168
+[transformers] PyTorch: setting up devices
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 912519168
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 912519168
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 912519168
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 12
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 4
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 16
+[transformers] Total optimization steps = 864
+[transformers] Number of trainable parameters = 912,519,168
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: setting up run a35dlqvz
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260817_221314-a35dlqvz
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t0_S5_lr0.0006
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/a35dlqvz
+
0%| | 0/864 [00:00 will be ignored
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[rank4]:W0817 22:13:37.714000 2318517 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[rank1]:W0817 22:13:37.824000 2318514 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank3]:W0817 22:13:37.940000 2318516 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank7]:W0817 22:13:38.132000 2318520 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank6]:W0817 22:13:38.152000 2318519 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank2]:W0817 22:13:38.294000 2318515 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[rank0]:W0817 22:13:41.149000 2318513 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+
0%| | 1/864 [00:47<11:21:02, 47.35s/it]
{'loss': '12.26', 'grad_norm': '3.637', 'learning_rate': '0', 'epoch': '0.01397'}
+
0%| | 1/864 [00:47<11:21:02, 47.35s/it]
0%| | 2/864 [00:49<5:00:55, 20.95s/it]
0%| | 3/864 [00:52<2:58:25, 12.43s/it]
0%| | 4/864 [00:54<2:00:54, 8.44s/it]
1%| | 5/864 [00:56<1:29:07, 6.23s/it]
1%| | 6/864 [00:59<1:09:57, 4.89s/it]
1%| | 7/864 [01:01<57:50, 4.05s/it]
1%| | 8/864 [01:03<49:50, 3.49s/it]
1%| | 9/864 [01:05<44:28, 3.12s/it]
1%| | 10/864 [01:08<40:48, 2.87s/it]
{'loss': '11.06', 'grad_norm': '2.076', 'learning_rate': '5.4e-05', 'epoch': '0.1397'}
+
1%| | 10/864 [01:08<40:48, 2.87s/it]
1%|▏ | 11/864 [01:10<38:17, 2.69s/it]
1%|▏ | 12/864 [01:12<36:32, 2.57s/it]
2%|▏ | 13/864 [01:15<35:19, 2.49s/it]
2%|▏ | 14/864 [01:17<34:27, 2.43s/it]
2%|▏ | 15/864 [01:19<33:53, 2.39s/it]
2%|▏ | 16/864 [01:22<33:27, 2.37s/it]
2%|▏ | 17/864 [01:24<33:11, 2.35s/it]
2%|▏ | 18/864 [01:26<32:56, 2.34s/it]
2%|▏ | 19/864 [01:28<32:44, 2.32s/it]
2%|▏ | 20/864 [01:31<32:35, 2.32s/it]
{'loss': '9.309', 'grad_norm': '1.216', 'learning_rate': '0.000114', 'epoch': '0.2795'}
+
2%|▏ | 20/864 [01:31<32:35, 2.32s/it]
2%|▏ | 21/864 [01:33<32:29, 2.31s/it]
3%|▎ | 22/864 [01:35<32:25, 2.31s/it]
3%|▎ | 23/864 [01:38<32:23, 2.31s/it]
3%|▎ | 24/864 [01:40<32:18, 2.31s/it]
3%|▎ | 25/864 [01:42<32:14, 2.31s/it]
3%|▎ | 26/864 [01:45<32:11, 2.31s/it]
3%|▎ | 27/864 [01:47<32:08, 2.30s/it]
3%|▎ | 28/864 [01:49<32:05, 2.30s/it]
3%|▎ | 29/864 [01:52<32:05, 2.31s/it]
3%|▎ | 30/864 [01:54<33:48, 2.43s/it]
{'loss': '7.947', 'grad_norm': '0.6039', 'learning_rate': '0.000174', 'epoch': '0.4192'}
+
3%|▎ | 30/864 [01:54<33:48, 2.43s/it]
4%|▎ | 31/864 [01:57<33:13, 2.39s/it]
4%|▎ | 32/864 [01:59<32:49, 2.37s/it]
4%|▍ | 33/864 [02:01<32:32, 2.35s/it]
4%|▍ | 34/864 [02:03<32:19, 2.34s/it]
4%|▍ | 35/864 [02:06<32:09, 2.33s/it]
4%|▍ | 36/864 [02:08<32:01, 2.32s/it]
4%|▍ | 37/864 [02:10<31:57, 2.32s/it]
4%|▍ | 38/864 [02:13<31:53, 2.32s/it]
5%|▍ | 39/864 [02:15<31:51, 2.32s/it]
5%|▍ | 40/864 [02:17<31:48, 2.32s/it]
{'loss': '7.515', 'grad_norm': '1.14', 'learning_rate': '0.000234', 'epoch': '0.559'}
+
5%|▍ | 40/864 [02:17<31:48, 2.32s/it]
5%|▍ | 41/864 [02:20<31:46, 2.32s/it]
5%|▍ | 42/864 [02:22<33:39, 2.46s/it]
5%|▍ | 43/864 [02:25<33:02, 2.42s/it]
5%|▌ | 44/864 [02:27<32:35, 2.38s/it]
5%|▌ | 45/864 [02:29<32:16, 2.36s/it]
5%|▌ | 46/864 [02:32<32:02, 2.35s/it]
5%|▌ | 47/864 [02:34<31:52, 2.34s/it]
6%|▌ | 48/864 [02:36<31:45, 2.33s/it]
6%|▌ | 49/864 [02:39<31:39, 2.33s/it]
6%|▌ | 50/864 [02:41<31:33, 2.33s/it]
{'loss': '7.308', 'grad_norm': '1.305', 'learning_rate': '0.000294', 'epoch': '0.6987'}
+
6%|▌ | 50/864 [02:41<31:33, 2.33s/it]
6%|▌ | 51/864 [02:43<31:30, 2.32s/it]
6%|▌ | 52/864 [02:46<31:27, 2.32s/it]
6%|▌ | 53/864 [02:48<31:23, 2.32s/it]
6%|▋ | 54/864 [02:50<31:21, 2.32s/it]
6%|▋ | 55/864 [02:53<31:18, 2.32s/it]
6%|▋ | 56/864 [02:55<31:16, 2.32s/it]
7%|▋ | 57/864 [02:57<31:15, 2.32s/it]
7%|▋ | 58/864 [03:00<31:11, 2.32s/it]
7%|▋ | 59/864 [03:02<31:09, 2.32s/it]
7%|▋ | 60/864 [03:04<31:07, 2.32s/it]
{'loss': '7.055', 'grad_norm': '1.331', 'learning_rate': '0.000354', 'epoch': '0.8384'}
+
7%|▋ | 60/864 [03:04<31:07, 2.32s/it]
7%|▋ | 61/864 [03:07<31:05, 2.32s/it]
7%|▋ | 62/864 [03:09<31:03, 2.32s/it]
7%|▋ | 63/864 [03:11<31:02, 2.32s/it]
7%|▋ | 64/864 [03:13<30:58, 2.32s/it]
8%|▊ | 65/864 [03:16<30:56, 2.32s/it]
8%|▊ | 66/864 [03:18<30:54, 2.32s/it]
8%|▊ | 67/864 [03:20<30:53, 2.33s/it]
8%|▊ | 68/864 [03:23<30:51, 2.33s/it]
8%|▊ | 69/864 [03:25<30:48, 2.33s/it]
8%|▊ | 70/864 [03:27<30:46, 2.33s/it]
{'loss': '6.819', 'grad_norm': '1.133', 'learning_rate': '0.000414', 'epoch': '0.9782'}
+
8%|▊ | 70/864 [03:27<30:46, 2.33s/it]
8%|▊ | 71/864 [03:30<30:44, 2.33s/it]
8%|▊ | 72/864 [03:31<26:45, 2.03s/it][transformers]
+***** Running Evaluation *****
+[transformers] Num examples = 2441
+[transformers] Batch size = 4
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+
+
0%| | 0/77 [00:00system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/t1a/t1a_S2_C0_baseline/checkpoint-1224/config.json b/runs/t1a/t1a_S2_C0_baseline/checkpoint-1224/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8ca584a0751672f280da7c54ff3916cdf24d6f4
--- /dev/null
+++ b/runs/t1a/t1a_S2_C0_baseline/checkpoint-1224/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 512,
+ "initializer_range": 0.02,
+ "intermediate_size": 1536,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 4,
+ "num_hidden_layers": 20,
+ "num_key_value_heads": 4,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151671
+}
diff --git a/runs/t1a/t1a_S2_C0_baseline/checkpoint-1224/generation_config.json b/runs/t1a/t1a_S2_C0_baseline/checkpoint-1224/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/t1a/t1a_S2_C0_baseline/checkpoint-1224/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/t1a/t1a_S2_C0_baseline/checkpoint-1224/tokenizer_config.json b/runs/t1a/t1a_S2_C0_baseline/checkpoint-1224/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..839a3743177a3ed77e1ffbae67520e8500ca0532
--- /dev/null
+++ b/runs/t1a/t1a_S2_C0_baseline/checkpoint-1224/tokenizer_config.json
@@ -0,0 +1,19 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/t1a/t1a_S2_C0_baseline/checkpoint-1224/trainer_state.json b/runs/t1a/t1a_S2_C0_baseline/checkpoint-1224/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4ad2f600e0fd9808d2ddc185469ef405580cd52
--- /dev/null
+++ b/runs/t1a/t1a_S2_C0_baseline/checkpoint-1224/trainer_state.json
@@ -0,0 +1,1031 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 17.0,
+ "eval_steps": 72,
+ "global_step": 1224,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013961605584642234,
+ "grad_norm": 1.8900947570800781,
+ "learning_rate": 0.0,
+ "loss": 12.03172492980957,
+ "step": 1
+ },
+ {
+ "epoch": 0.13961605584642234,
+ "grad_norm": 1.765875220298767,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.73646714952257,
+ "step": 10
+ },
+ {
+ "epoch": 0.2792321116928447,
+ "grad_norm": 1.694366216659546,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 11.039095306396485,
+ "step": 20
+ },
+ {
+ "epoch": 0.418848167539267,
+ "grad_norm": 1.6475836038589478,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 10.272348785400391,
+ "step": 30
+ },
+ {
+ "epoch": 0.5584642233856894,
+ "grad_norm": 1.5271304845809937,
+ "learning_rate": 0.000234,
+ "loss": 9.232383728027344,
+ "step": 40
+ },
+ {
+ "epoch": 0.6980802792321117,
+ "grad_norm": 0.945402979850769,
+ "learning_rate": 0.000294,
+ "loss": 8.251237487792968,
+ "step": 50
+ },
+ {
+ "epoch": 0.837696335078534,
+ "grad_norm": 0.3249334394931793,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 7.700569152832031,
+ "step": 60
+ },
+ {
+ "epoch": 0.9773123909249564,
+ "grad_norm": 1.6057826280593872,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 7.523677062988281,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.4746503829956055,
+ "eval_runtime": 27.4901,
+ "eval_samples_per_second": 88.795,
+ "eval_steps_per_second": 1.419,
+ "step": 72
+ },
+ {
+ "epoch": 1.111692844677138,
+ "grad_norm": 0.550653338432312,
+ "learning_rate": 0.000474,
+ "loss": 7.411495208740234,
+ "step": 80
+ },
+ {
+ "epoch": 1.2513089005235603,
+ "grad_norm": 0.526059091091156,
+ "learning_rate": 0.000534,
+ "loss": 7.210162353515625,
+ "step": 90
+ },
+ {
+ "epoch": 1.3909249563699826,
+ "grad_norm": 0.3909878730773926,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 7.044137573242187,
+ "step": 100
+ },
+ {
+ "epoch": 1.530541012216405,
+ "grad_norm": 0.3361167311668396,
+ "learning_rate": 0.0006,
+ "loss": 6.893026733398438,
+ "step": 110
+ },
+ {
+ "epoch": 1.6701570680628273,
+ "grad_norm": 0.6221396923065186,
+ "learning_rate": 0.0006,
+ "loss": 6.761862945556641,
+ "step": 120
+ },
+ {
+ "epoch": 1.8097731239092496,
+ "grad_norm": 0.697410523891449,
+ "learning_rate": 0.0006,
+ "loss": 6.660725402832031,
+ "step": 130
+ },
+ {
+ "epoch": 1.949389179755672,
+ "grad_norm": 0.7825177907943726,
+ "learning_rate": 0.0006,
+ "loss": 6.542182922363281,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 6.464291095733643,
+ "eval_runtime": 3.2557,
+ "eval_samples_per_second": 749.771,
+ "eval_steps_per_second": 11.979,
+ "step": 144
+ },
+ {
+ "epoch": 2.0837696335078535,
+ "grad_norm": 0.7686965465545654,
+ "learning_rate": 0.0006,
+ "loss": 6.454528045654297,
+ "step": 150
+ },
+ {
+ "epoch": 2.223385689354276,
+ "grad_norm": 0.9595056176185608,
+ "learning_rate": 0.0006,
+ "loss": 6.370765686035156,
+ "step": 160
+ },
+ {
+ "epoch": 2.363001745200698,
+ "grad_norm": 0.37826409935951233,
+ "learning_rate": 0.0006,
+ "loss": 6.292685317993164,
+ "step": 170
+ },
+ {
+ "epoch": 2.5026178010471205,
+ "grad_norm": 0.5043449997901917,
+ "learning_rate": 0.0006,
+ "loss": 6.2167106628417965,
+ "step": 180
+ },
+ {
+ "epoch": 2.642233856893543,
+ "grad_norm": 0.5590866804122925,
+ "learning_rate": 0.0006,
+ "loss": 6.202901840209961,
+ "step": 190
+ },
+ {
+ "epoch": 2.781849912739965,
+ "grad_norm": 0.5644322633743286,
+ "learning_rate": 0.0006,
+ "loss": 6.110133361816406,
+ "step": 200
+ },
+ {
+ "epoch": 2.9214659685863875,
+ "grad_norm": 0.41593649983406067,
+ "learning_rate": 0.0006,
+ "loss": 6.03997688293457,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 6.000840187072754,
+ "eval_runtime": 3.2655,
+ "eval_samples_per_second": 747.508,
+ "eval_steps_per_second": 11.943,
+ "step": 216
+ },
+ {
+ "epoch": 3.055846422338569,
+ "grad_norm": 0.8875220417976379,
+ "learning_rate": 0.0006,
+ "loss": 5.967375183105469,
+ "step": 220
+ },
+ {
+ "epoch": 3.195462478184991,
+ "grad_norm": 0.7434491515159607,
+ "learning_rate": 0.0006,
+ "loss": 5.9561717987060545,
+ "step": 230
+ },
+ {
+ "epoch": 3.3350785340314135,
+ "grad_norm": 0.3451848328113556,
+ "learning_rate": 0.0006,
+ "loss": 5.85661506652832,
+ "step": 240
+ },
+ {
+ "epoch": 3.474694589877836,
+ "grad_norm": 1.0598573684692383,
+ "learning_rate": 0.0006,
+ "loss": 5.834651184082031,
+ "step": 250
+ },
+ {
+ "epoch": 3.614310645724258,
+ "grad_norm": 0.4185563325881958,
+ "learning_rate": 0.0006,
+ "loss": 5.774275588989258,
+ "step": 260
+ },
+ {
+ "epoch": 3.7539267015706805,
+ "grad_norm": 1.0291014909744263,
+ "learning_rate": 0.0006,
+ "loss": 5.725038146972656,
+ "step": 270
+ },
+ {
+ "epoch": 3.893542757417103,
+ "grad_norm": 0.2985861003398895,
+ "learning_rate": 0.0006,
+ "loss": 5.680973815917969,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 5.631031036376953,
+ "eval_runtime": 3.2654,
+ "eval_samples_per_second": 747.538,
+ "eval_steps_per_second": 11.943,
+ "step": 288
+ },
+ {
+ "epoch": 4.027923211169284,
+ "grad_norm": 0.5913441181182861,
+ "learning_rate": 0.0006,
+ "loss": 5.58748779296875,
+ "step": 290
+ },
+ {
+ "epoch": 4.167539267015707,
+ "grad_norm": 0.8499543070793152,
+ "learning_rate": 0.0006,
+ "loss": 5.547278594970703,
+ "step": 300
+ },
+ {
+ "epoch": 4.307155322862129,
+ "grad_norm": 0.38768234848976135,
+ "learning_rate": 0.0006,
+ "loss": 5.487092971801758,
+ "step": 310
+ },
+ {
+ "epoch": 4.446771378708552,
+ "grad_norm": 0.9851556420326233,
+ "learning_rate": 0.0006,
+ "loss": 5.4364665985107425,
+ "step": 320
+ },
+ {
+ "epoch": 4.5863874345549736,
+ "grad_norm": 0.4321438670158386,
+ "learning_rate": 0.0006,
+ "loss": 5.441534423828125,
+ "step": 330
+ },
+ {
+ "epoch": 4.726003490401396,
+ "grad_norm": 0.521005392074585,
+ "learning_rate": 0.0006,
+ "loss": 5.357921981811524,
+ "step": 340
+ },
+ {
+ "epoch": 4.865619546247818,
+ "grad_norm": 0.45179012417793274,
+ "learning_rate": 0.0006,
+ "loss": 5.309077453613281,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.3567813038825989,
+ "learning_rate": 0.0006,
+ "loss": 5.269740295410156,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 5.295504093170166,
+ "eval_runtime": 3.2523,
+ "eval_samples_per_second": 750.546,
+ "eval_steps_per_second": 11.992,
+ "step": 360
+ },
+ {
+ "epoch": 5.139616055846423,
+ "grad_norm": 0.40960413217544556,
+ "learning_rate": 0.0006,
+ "loss": 5.198007965087891,
+ "step": 370
+ },
+ {
+ "epoch": 5.279232111692845,
+ "grad_norm": 0.7737030386924744,
+ "learning_rate": 0.0006,
+ "loss": 5.168359375,
+ "step": 380
+ },
+ {
+ "epoch": 5.418848167539267,
+ "grad_norm": 0.5689374208450317,
+ "learning_rate": 0.0006,
+ "loss": 5.1186878204345705,
+ "step": 390
+ },
+ {
+ "epoch": 5.558464223385689,
+ "grad_norm": 0.6907999515533447,
+ "learning_rate": 0.0006,
+ "loss": 5.05829086303711,
+ "step": 400
+ },
+ {
+ "epoch": 5.698080279232112,
+ "grad_norm": 0.34229475259780884,
+ "learning_rate": 0.0006,
+ "loss": 5.028703689575195,
+ "step": 410
+ },
+ {
+ "epoch": 5.837696335078534,
+ "grad_norm": 0.5962613224983215,
+ "learning_rate": 0.0006,
+ "loss": 4.971548080444336,
+ "step": 420
+ },
+ {
+ "epoch": 5.977312390924956,
+ "grad_norm": 0.6673392653465271,
+ "learning_rate": 0.0006,
+ "loss": 4.964711761474609,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 5.0211181640625,
+ "eval_runtime": 3.2639,
+ "eval_samples_per_second": 747.87,
+ "eval_steps_per_second": 11.949,
+ "step": 432
+ },
+ {
+ "epoch": 6.111692844677138,
+ "grad_norm": 0.6230090856552124,
+ "learning_rate": 0.0006,
+ "loss": 4.910141372680664,
+ "step": 440
+ },
+ {
+ "epoch": 6.2513089005235605,
+ "grad_norm": 0.5077657103538513,
+ "learning_rate": 0.0006,
+ "loss": 4.831769180297852,
+ "step": 450
+ },
+ {
+ "epoch": 6.390924956369982,
+ "grad_norm": 0.410927414894104,
+ "learning_rate": 0.0006,
+ "loss": 4.802286911010742,
+ "step": 460
+ },
+ {
+ "epoch": 6.530541012216405,
+ "grad_norm": 0.33632829785346985,
+ "learning_rate": 0.0006,
+ "loss": 4.736461639404297,
+ "step": 470
+ },
+ {
+ "epoch": 6.670157068062827,
+ "grad_norm": 0.710969865322113,
+ "learning_rate": 0.0006,
+ "loss": 4.708580780029297,
+ "step": 480
+ },
+ {
+ "epoch": 6.80977312390925,
+ "grad_norm": 0.4061291515827179,
+ "learning_rate": 0.0006,
+ "loss": 4.7073310852050785,
+ "step": 490
+ },
+ {
+ "epoch": 6.949389179755672,
+ "grad_norm": 0.34648239612579346,
+ "learning_rate": 0.0006,
+ "loss": 4.628459167480469,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 4.7059807777404785,
+ "eval_runtime": 3.2645,
+ "eval_samples_per_second": 747.749,
+ "eval_steps_per_second": 11.947,
+ "step": 504
+ },
+ {
+ "epoch": 7.0837696335078535,
+ "grad_norm": 0.5380810499191284,
+ "learning_rate": 0.0006,
+ "loss": 4.575440216064453,
+ "step": 510
+ },
+ {
+ "epoch": 7.223385689354275,
+ "grad_norm": 0.36076590418815613,
+ "learning_rate": 0.0006,
+ "loss": 4.5593925476074215,
+ "step": 520
+ },
+ {
+ "epoch": 7.363001745200698,
+ "grad_norm": 0.2604350447654724,
+ "learning_rate": 0.0006,
+ "loss": 4.49984130859375,
+ "step": 530
+ },
+ {
+ "epoch": 7.50261780104712,
+ "grad_norm": 0.5101083517074585,
+ "learning_rate": 0.0006,
+ "loss": 4.471659088134766,
+ "step": 540
+ },
+ {
+ "epoch": 7.642233856893543,
+ "grad_norm": 0.5122832655906677,
+ "learning_rate": 0.0006,
+ "loss": 4.441104125976563,
+ "step": 550
+ },
+ {
+ "epoch": 7.781849912739965,
+ "grad_norm": 0.4772845506668091,
+ "learning_rate": 0.0006,
+ "loss": 4.428692626953125,
+ "step": 560
+ },
+ {
+ "epoch": 7.9214659685863875,
+ "grad_norm": 0.33619433641433716,
+ "learning_rate": 0.0006,
+ "loss": 4.402286529541016,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.493332862854004,
+ "eval_runtime": 3.2628,
+ "eval_samples_per_second": 748.125,
+ "eval_steps_per_second": 11.953,
+ "step": 576
+ },
+ {
+ "epoch": 8.055846422338568,
+ "grad_norm": 0.4145880937576294,
+ "learning_rate": 0.0006,
+ "loss": 4.345246887207031,
+ "step": 580
+ },
+ {
+ "epoch": 8.195462478184991,
+ "grad_norm": 0.432656854391098,
+ "learning_rate": 0.0006,
+ "loss": 4.303118896484375,
+ "step": 590
+ },
+ {
+ "epoch": 8.335078534031414,
+ "grad_norm": 0.30922794342041016,
+ "learning_rate": 0.0006,
+ "loss": 4.28332748413086,
+ "step": 600
+ },
+ {
+ "epoch": 8.474694589877837,
+ "grad_norm": 0.29728612303733826,
+ "learning_rate": 0.0006,
+ "loss": 4.251079177856445,
+ "step": 610
+ },
+ {
+ "epoch": 8.614310645724258,
+ "grad_norm": 0.346055805683136,
+ "learning_rate": 0.0006,
+ "loss": 4.234855651855469,
+ "step": 620
+ },
+ {
+ "epoch": 8.75392670157068,
+ "grad_norm": 0.4120888113975525,
+ "learning_rate": 0.0006,
+ "loss": 4.2234031677246096,
+ "step": 630
+ },
+ {
+ "epoch": 8.893542757417103,
+ "grad_norm": 0.32442861795425415,
+ "learning_rate": 0.0006,
+ "loss": 4.218895721435547,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.336294174194336,
+ "eval_runtime": 4.2376,
+ "eval_samples_per_second": 576.029,
+ "eval_steps_per_second": 9.203,
+ "step": 648
+ },
+ {
+ "epoch": 9.027923211169284,
+ "grad_norm": 0.4526400864124298,
+ "learning_rate": 0.0006,
+ "loss": 4.176125335693359,
+ "step": 650
+ },
+ {
+ "epoch": 9.167539267015707,
+ "grad_norm": 0.6241201758384705,
+ "learning_rate": 0.0006,
+ "loss": 4.141399383544922,
+ "step": 660
+ },
+ {
+ "epoch": 9.30715532286213,
+ "grad_norm": 0.507562518119812,
+ "learning_rate": 0.0006,
+ "loss": 4.1806785583496096,
+ "step": 670
+ },
+ {
+ "epoch": 9.44677137870855,
+ "grad_norm": 0.23918727040290833,
+ "learning_rate": 0.0006,
+ "loss": 4.146372222900391,
+ "step": 680
+ },
+ {
+ "epoch": 9.586387434554974,
+ "grad_norm": 0.2581324577331543,
+ "learning_rate": 0.0006,
+ "loss": 4.096488189697266,
+ "step": 690
+ },
+ {
+ "epoch": 9.726003490401396,
+ "grad_norm": 0.25462764501571655,
+ "learning_rate": 0.0006,
+ "loss": 4.083126831054687,
+ "step": 700
+ },
+ {
+ "epoch": 9.86561954624782,
+ "grad_norm": 0.22756332159042358,
+ "learning_rate": 0.0006,
+ "loss": 4.053802871704102,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.35741207003593445,
+ "learning_rate": 0.0006,
+ "loss": 4.035561370849609,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.232204437255859,
+ "eval_runtime": 4.9974,
+ "eval_samples_per_second": 488.45,
+ "eval_steps_per_second": 7.804,
+ "step": 720
+ },
+ {
+ "epoch": 10.139616055846423,
+ "grad_norm": 0.42792415618896484,
+ "learning_rate": 0.0006,
+ "loss": 3.974091339111328,
+ "step": 730
+ },
+ {
+ "epoch": 10.279232111692846,
+ "grad_norm": 0.44319143891334534,
+ "learning_rate": 0.0006,
+ "loss": 3.981563186645508,
+ "step": 740
+ },
+ {
+ "epoch": 10.418848167539267,
+ "grad_norm": 0.5030475854873657,
+ "learning_rate": 0.0006,
+ "loss": 3.9796348571777345,
+ "step": 750
+ },
+ {
+ "epoch": 10.55846422338569,
+ "grad_norm": 0.27747198939323425,
+ "learning_rate": 0.0006,
+ "loss": 3.985432815551758,
+ "step": 760
+ },
+ {
+ "epoch": 10.698080279232112,
+ "grad_norm": 0.26435166597366333,
+ "learning_rate": 0.0006,
+ "loss": 3.9562255859375,
+ "step": 770
+ },
+ {
+ "epoch": 10.837696335078533,
+ "grad_norm": 0.22242967784404755,
+ "learning_rate": 0.0006,
+ "loss": 3.9437767028808595,
+ "step": 780
+ },
+ {
+ "epoch": 10.977312390924956,
+ "grad_norm": 0.23960651457309723,
+ "learning_rate": 0.0006,
+ "loss": 3.9271080017089846,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.146022796630859,
+ "eval_runtime": 3.4186,
+ "eval_samples_per_second": 714.039,
+ "eval_steps_per_second": 11.408,
+ "step": 792
+ },
+ {
+ "epoch": 11.111692844677139,
+ "grad_norm": 0.512363612651825,
+ "learning_rate": 0.0006,
+ "loss": 3.8826152801513674,
+ "step": 800
+ },
+ {
+ "epoch": 11.25130890052356,
+ "grad_norm": 0.5255026817321777,
+ "learning_rate": 0.0006,
+ "loss": 3.8584999084472655,
+ "step": 810
+ },
+ {
+ "epoch": 11.390924956369982,
+ "grad_norm": 0.47391441464424133,
+ "learning_rate": 0.0006,
+ "loss": 3.8792823791503905,
+ "step": 820
+ },
+ {
+ "epoch": 11.530541012216405,
+ "grad_norm": 0.28765520453453064,
+ "learning_rate": 0.0006,
+ "loss": 3.876079559326172,
+ "step": 830
+ },
+ {
+ "epoch": 11.670157068062828,
+ "grad_norm": 0.22929531335830688,
+ "learning_rate": 0.0006,
+ "loss": 3.8527259826660156,
+ "step": 840
+ },
+ {
+ "epoch": 11.809773123909249,
+ "grad_norm": 0.22583293914794922,
+ "learning_rate": 0.0006,
+ "loss": 3.8398895263671875,
+ "step": 850
+ },
+ {
+ "epoch": 11.949389179755672,
+ "grad_norm": 0.25066936016082764,
+ "learning_rate": 0.0006,
+ "loss": 3.838032531738281,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.0848612785339355,
+ "eval_runtime": 4.1891,
+ "eval_samples_per_second": 582.704,
+ "eval_steps_per_second": 9.31,
+ "step": 864
+ },
+ {
+ "epoch": 12.083769633507853,
+ "grad_norm": 0.42194268107414246,
+ "learning_rate": 0.0006,
+ "loss": 3.7769397735595702,
+ "step": 870
+ },
+ {
+ "epoch": 12.223385689354275,
+ "grad_norm": 0.5075748562812805,
+ "learning_rate": 0.0006,
+ "loss": 3.7586505889892576,
+ "step": 880
+ },
+ {
+ "epoch": 12.363001745200698,
+ "grad_norm": 0.34035468101501465,
+ "learning_rate": 0.0006,
+ "loss": 3.7768104553222654,
+ "step": 890
+ },
+ {
+ "epoch": 12.502617801047121,
+ "grad_norm": 0.28600114583969116,
+ "learning_rate": 0.0006,
+ "loss": 3.755086135864258,
+ "step": 900
+ },
+ {
+ "epoch": 12.642233856893542,
+ "grad_norm": 0.23299799859523773,
+ "learning_rate": 0.0006,
+ "loss": 3.7589508056640626,
+ "step": 910
+ },
+ {
+ "epoch": 12.781849912739965,
+ "grad_norm": 0.24068665504455566,
+ "learning_rate": 0.0006,
+ "loss": 3.7530860900878906,
+ "step": 920
+ },
+ {
+ "epoch": 12.921465968586388,
+ "grad_norm": 0.25879961252212524,
+ "learning_rate": 0.0006,
+ "loss": 3.743784713745117,
+ "step": 930
+ },
+ {
+ "epoch": 13.0,
+ "eval_loss": 4.047211170196533,
+ "eval_runtime": 5.5621,
+ "eval_samples_per_second": 438.862,
+ "eval_steps_per_second": 7.012,
+ "step": 936
+ },
+ {
+ "epoch": 13.055846422338568,
+ "grad_norm": 0.48456165194511414,
+ "learning_rate": 0.0006,
+ "loss": 3.696119689941406,
+ "step": 940
+ },
+ {
+ "epoch": 13.195462478184991,
+ "grad_norm": 0.4380106031894684,
+ "learning_rate": 0.0006,
+ "loss": 3.671025848388672,
+ "step": 950
+ },
+ {
+ "epoch": 13.335078534031414,
+ "grad_norm": 0.3542587459087372,
+ "learning_rate": 0.0006,
+ "loss": 3.6769004821777345,
+ "step": 960
+ },
+ {
+ "epoch": 13.474694589877837,
+ "grad_norm": 0.3082897663116455,
+ "learning_rate": 0.0006,
+ "loss": 3.664134979248047,
+ "step": 970
+ },
+ {
+ "epoch": 13.614310645724258,
+ "grad_norm": 0.24334463477134705,
+ "learning_rate": 0.0006,
+ "loss": 3.661321258544922,
+ "step": 980
+ },
+ {
+ "epoch": 13.75392670157068,
+ "grad_norm": 0.25771069526672363,
+ "learning_rate": 0.0006,
+ "loss": 3.6706939697265626,
+ "step": 990
+ },
+ {
+ "epoch": 13.893542757417103,
+ "grad_norm": 0.27326256036758423,
+ "learning_rate": 0.0006,
+ "loss": 3.6684364318847655,
+ "step": 1000
+ },
+ {
+ "epoch": 14.0,
+ "eval_loss": 4.020983695983887,
+ "eval_runtime": 4.6413,
+ "eval_samples_per_second": 525.93,
+ "eval_steps_per_second": 8.403,
+ "step": 1008
+ },
+ {
+ "epoch": 14.027923211169284,
+ "grad_norm": 0.35269543528556824,
+ "learning_rate": 0.0006,
+ "loss": 3.643656921386719,
+ "step": 1010
+ },
+ {
+ "epoch": 14.167539267015707,
+ "grad_norm": 0.559786856174469,
+ "learning_rate": 0.0006,
+ "loss": 3.5870620727539064,
+ "step": 1020
+ },
+ {
+ "epoch": 14.30715532286213,
+ "grad_norm": 0.4918050169944763,
+ "learning_rate": 0.0006,
+ "loss": 3.608507537841797,
+ "step": 1030
+ },
+ {
+ "epoch": 14.44677137870855,
+ "grad_norm": 0.3729175627231598,
+ "learning_rate": 0.0006,
+ "loss": 3.6270870208740233,
+ "step": 1040
+ },
+ {
+ "epoch": 14.586387434554974,
+ "grad_norm": 0.2599662244319916,
+ "learning_rate": 0.0006,
+ "loss": 3.603087615966797,
+ "step": 1050
+ },
+ {
+ "epoch": 14.726003490401396,
+ "grad_norm": 0.22056028246879578,
+ "learning_rate": 0.0006,
+ "loss": 3.5970054626464845,
+ "step": 1060
+ },
+ {
+ "epoch": 14.86561954624782,
+ "grad_norm": 0.24676060676574707,
+ "learning_rate": 0.0006,
+ "loss": 3.593349075317383,
+ "step": 1070
+ },
+ {
+ "epoch": 15.0,
+ "grad_norm": 0.23699252307415009,
+ "learning_rate": 0.0006,
+ "loss": 3.603044128417969,
+ "step": 1080
+ },
+ {
+ "epoch": 15.0,
+ "eval_loss": 3.9932801723480225,
+ "eval_runtime": 5.5907,
+ "eval_samples_per_second": 436.616,
+ "eval_steps_per_second": 6.976,
+ "step": 1080
+ },
+ {
+ "epoch": 15.139616055846423,
+ "grad_norm": 0.4945415258407593,
+ "learning_rate": 0.0006,
+ "loss": 3.4861793518066406,
+ "step": 1090
+ },
+ {
+ "epoch": 15.279232111692846,
+ "grad_norm": 0.6685895323753357,
+ "learning_rate": 0.0006,
+ "loss": 3.5365150451660154,
+ "step": 1100
+ },
+ {
+ "epoch": 15.418848167539267,
+ "grad_norm": 0.48016801476478577,
+ "learning_rate": 0.0006,
+ "loss": 3.5710548400878905,
+ "step": 1110
+ },
+ {
+ "epoch": 15.55846422338569,
+ "grad_norm": 0.27422818541526794,
+ "learning_rate": 0.0006,
+ "loss": 3.556343841552734,
+ "step": 1120
+ },
+ {
+ "epoch": 15.698080279232112,
+ "grad_norm": 0.24454139173030853,
+ "learning_rate": 0.0006,
+ "loss": 3.5544708251953123,
+ "step": 1130
+ },
+ {
+ "epoch": 15.837696335078533,
+ "grad_norm": 0.22212658822536469,
+ "learning_rate": 0.0006,
+ "loss": 3.540570831298828,
+ "step": 1140
+ },
+ {
+ "epoch": 15.977312390924956,
+ "grad_norm": 0.2754276692867279,
+ "learning_rate": 0.0006,
+ "loss": 3.5291683197021486,
+ "step": 1150
+ },
+ {
+ "epoch": 16.0,
+ "eval_loss": 3.9805431365966797,
+ "eval_runtime": 5.3776,
+ "eval_samples_per_second": 453.921,
+ "eval_steps_per_second": 7.252,
+ "step": 1152
+ },
+ {
+ "epoch": 16.111692844677137,
+ "grad_norm": 0.38299480080604553,
+ "learning_rate": 0.0006,
+ "loss": 3.4493717193603515,
+ "step": 1160
+ },
+ {
+ "epoch": 16.25130890052356,
+ "grad_norm": 0.3984591066837311,
+ "learning_rate": 0.0006,
+ "loss": 3.432342529296875,
+ "step": 1170
+ },
+ {
+ "epoch": 16.390924956369982,
+ "grad_norm": 0.308737188577652,
+ "learning_rate": 0.0006,
+ "loss": 3.441315460205078,
+ "step": 1180
+ },
+ {
+ "epoch": 16.530541012216403,
+ "grad_norm": 0.2914087772369385,
+ "learning_rate": 0.0006,
+ "loss": 3.4504776000976562,
+ "step": 1190
+ },
+ {
+ "epoch": 16.670157068062828,
+ "grad_norm": 0.28331220149993896,
+ "learning_rate": 0.0006,
+ "loss": 3.461971664428711,
+ "step": 1200
+ },
+ {
+ "epoch": 16.80977312390925,
+ "grad_norm": 0.27251577377319336,
+ "learning_rate": 0.0006,
+ "loss": 3.483176422119141,
+ "step": 1210
+ },
+ {
+ "epoch": 16.949389179755673,
+ "grad_norm": 0.24930404126644135,
+ "learning_rate": 0.0006,
+ "loss": 3.4626285552978517,
+ "step": 1220
+ },
+ {
+ "epoch": 17.0,
+ "eval_loss": 3.9779744148254395,
+ "eval_runtime": 5.7372,
+ "eval_samples_per_second": 425.467,
+ "eval_steps_per_second": 6.798,
+ "step": 1224
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 7200,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 100,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 5.2229002308983194e+17,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/t1a/t1a_S2_C0_baseline/train.log b/runs/t1a/t1a_S2_C0_baseline/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..84f4428bf3ddca9c2dabc69d0ab60abfc017901a
--- /dev/null
+++ b/runs/t1a/t1a_S2_C0_baseline/train.log
@@ -0,0 +1,2813 @@
+W0818 04:08:35.699000 2493017 torch/distributed/run.py:851]
+W0818 04:08:35.699000 2493017 torch/distributed/run.py:851] *****************************************
+W0818 04:08:35.699000 2493017 torch/distributed/run.py:851] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
+W0818 04:08:35.699000 2493017 torch/distributed/run.py:851] *****************************************
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+EOS token: <|im_end|>
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+EOS token: <|im_end|>
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+total params: 145833984
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 100
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 8
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 8
+[transformers] Total optimization steps = 7,200
+[transformers] Number of trainable parameters = 145,833,984
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: setting up run hh6dh348
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260818_040855-hh6dh348
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t1a_S2_C0_baseline
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/hh6dh348
+
0%| | 0/7200 [00:00 will be ignored
+[rank7]:W0818 04:09:19.155000 2493325 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank2]:W0818 04:09:19.179000 2493320 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[rank3]:W0818 04:09:19.202000 2493321 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank6]:W0818 04:09:19.345000 2493324 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank4]:W0818 04:09:19.721000 2493322 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank1]:W0818 04:09:19.862000 2493319 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[rank0]:W0818 04:09:21.057000 2493317 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+
0%| | 1/7200 [00:41<83:36:40, 41.81s/it]
{'loss': '12.03', 'grad_norm': '1.89', 'learning_rate': '0', 'epoch': '0.01396'}
+
0%| | 1/7200 [00:41<83:36:40, 41.81s/it]
0%| | 2/7200 [00:42<35:12:15, 17.61s/it]
0%| | 3/7200 [00:43<19:43:14, 9.86s/it]
0%| | 4/7200 [00:43<12:27:00, 6.23s/it]
0%| | 5/7200 [00:44<8:25:46, 4.22s/it]
0%| | 6/7200 [00:45<6:00:22, 3.01s/it]
0%| | 7/7200 [00:45<4:28:06, 2.24s/it]
0%| | 8/7200 [00:46<3:27:37, 1.73s/it]
0%| | 9/7200 [00:47<2:47:11, 1.39s/it]
0%| | 10/7200 [00:47<2:19:54, 1.17s/it]
{'loss': '11.74', 'grad_norm': '1.766', 'learning_rate': '5.4e-05', 'epoch': '0.1396'}
+
0%| | 10/7200 [00:47<2:19:54, 1.17s/it]
0%| | 11/7200 [00:48<2:01:04, 1.01s/it]
0%| | 12/7200 [00:49<1:48:00, 1.11it/s]
0%| | 13/7200 [00:49<1:38:54, 1.21it/s]
0%| | 14/7200 [00:50<1:32:40, 1.29it/s]
0%| | 15/7200 [00:50<1:28:15, 1.36it/s]
0%| | 16/7200 [00:51<1:25:09, 1.41it/s]
0%| | 17/7200 [00:52<1:23:07, 1.44it/s]
0%| | 18/7200 [00:52<1:21:40, 1.47it/s]
0%| | 19/7200 [00:53<1:20:32, 1.49it/s]
0%| | 20/7200 [00:54<1:19:51, 1.50it/s]
{'loss': '11.04', 'grad_norm': '1.694', 'learning_rate': '0.000114', 'epoch': '0.2792'}
+
0%| | 20/7200 [00:54<1:19:51, 1.50it/s]
0%| | 21/7200 [00:54<1:19:14, 1.51it/s]
0%| | 22/7200 [00:55<1:18:46, 1.52it/s]
0%| | 23/7200 [00:56<1:18:26, 1.53it/s]
0%| | 24/7200 [00:56<1:18:15, 1.53it/s]
0%| | 25/7200 [00:57<1:18:03, 1.53it/s]
0%| | 26/7200 [00:58<1:17:56, 1.53it/s]
0%| | 27/7200 [00:58<1:17:49, 1.54it/s]
0%| | 28/7200 [00:59<1:17:50, 1.54it/s]
0%| | 29/7200 [01:00<1:17:44, 1.54it/s]
0%| | 30/7200 [01:00<1:17:39, 1.54it/s]
{'loss': '10.27', 'grad_norm': '1.648', 'learning_rate': '0.000174', 'epoch': '0.4188'}
+
0%| | 30/7200 [01:00<1:17:39, 1.54it/s]
0%| | 31/7200 [01:01<1:17:36, 1.54it/s]
0%| | 32/7200 [01:02<1:17:40, 1.54it/s]
0%| | 33/7200 [01:02<1:17:37, 1.54it/s]
0%| | 34/7200 [01:03<1:17:33, 1.54it/s]
0%| | 35/7200 [01:03<1:17:33, 1.54it/s]
0%| | 36/7200 [01:04<1:17:32, 1.54it/s]
1%| | 37/7200 [01:05<1:17:35, 1.54it/s]
1%| | 38/7200 [01:05<1:17:34, 1.54it/s]
1%| | 39/7200 [01:06<1:17:34, 1.54it/s]
1%| | 40/7200 [01:07<1:17:31, 1.54it/s]
{'loss': '9.232', 'grad_norm': '1.527', 'learning_rate': '0.000234', 'epoch': '0.5585'}
+
1%| | 40/7200 [01:07<1:17:31, 1.54it/s]
1%| | 41/7200 [01:07<1:17:34, 1.54it/s]
1%| | 42/7200 [01:08<1:17:29, 1.54it/s]
1%| | 43/7200 [01:09<1:17:31, 1.54it/s]
1%| | 44/7200 [01:09<1:17:30, 1.54it/s]
1%| | 45/7200 [01:10<1:17:30, 1.54it/s]
1%| | 46/7200 [01:11<1:17:30, 1.54it/s]
1%| | 47/7200 [01:11<1:17:31, 1.54it/s]
1%| | 48/7200 [01:12<1:17:31, 1.54it/s]
1%| | 49/7200 [01:13<1:17:29, 1.54it/s]
1%| | 50/7200 [01:13<1:17:25, 1.54it/s]
{'loss': '8.251', 'grad_norm': '0.9454', 'learning_rate': '0.000294', 'epoch': '0.6981'}
+
1%| | 50/7200 [01:13<1:17:25, 1.54it/s]
1%| | 51/7200 [01:14<1:17:29, 1.54it/s]
1%| | 52/7200 [01:15<1:17:25, 1.54it/s]
1%| | 53/7200 [01:15<1:17:27, 1.54it/s]
1%| | 54/7200 [01:16<1:17:25, 1.54it/s]
1%| | 55/7200 [01:16<1:17:24, 1.54it/s]
1%| | 56/7200 [01:17<1:17:24, 1.54it/s]
1%| | 57/7200 [01:18<1:17:22, 1.54it/s]
1%| | 58/7200 [01:18<1:17:21, 1.54it/s]
1%| | 59/7200 [01:19<1:17:22, 1.54it/s]
1%| | 60/7200 [01:20<1:17:23, 1.54it/s]
{'loss': '7.701', 'grad_norm': '0.3249', 'learning_rate': '0.000354', 'epoch': '0.8377'}
+
1%| | 60/7200 [01:20<1:17:23, 1.54it/s]
1%| | 61/7200 [01:20<1:17:30, 1.54it/s]
1%| | 62/7200 [01:21<1:17:34, 1.53it/s]
1%| | 63/7200 [01:22<1:17:36, 1.53it/s]
1%| | 64/7200 [01:22<1:17:35, 1.53it/s]
1%| | 65/7200 [01:23<1:17:38, 1.53it/s]
1%| | 66/7200 [01:24<1:17:43, 1.53it/s]
1%| | 67/7200 [01:24<1:17:55, 1.53it/s]
1%| | 68/7200 [01:25<1:17:55, 1.53it/s]
1%| | 69/7200 [01:26<1:17:48, 1.53it/s]
1%| | 70/7200 [01:26<1:17:58, 1.52it/s]
{'loss': '7.524', 'grad_norm': '1.606', 'learning_rate': '0.000414', 'epoch': '0.9773'}
+
1%| | 70/7200 [01:26<1:17:58, 1.52it/s]
1%| | 71/7200 [01:27<1:17:52, 1.53it/s]
1%| | 72/7200 [01:27<1:09:04, 1.72it/s][transformers]
+***** Running Evaluation *****
+[transformers] Num examples = 2441
+[transformers] Batch size = 8
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+
+
0%| | 0/39 [00:00system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/t1a/t1a_S2_C1_rand15/checkpoint-2160/config.json b/runs/t1a/t1a_S2_C1_rand15/checkpoint-2160/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8ca584a0751672f280da7c54ff3916cdf24d6f4
--- /dev/null
+++ b/runs/t1a/t1a_S2_C1_rand15/checkpoint-2160/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 512,
+ "initializer_range": 0.02,
+ "intermediate_size": 1536,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 4,
+ "num_hidden_layers": 20,
+ "num_key_value_heads": 4,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151671
+}
diff --git a/runs/t1a/t1a_S2_C1_rand15/checkpoint-2160/generation_config.json b/runs/t1a/t1a_S2_C1_rand15/checkpoint-2160/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/t1a/t1a_S2_C1_rand15/checkpoint-2160/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/t1a/t1a_S2_C1_rand15/checkpoint-2160/tokenizer_config.json b/runs/t1a/t1a_S2_C1_rand15/checkpoint-2160/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..839a3743177a3ed77e1ffbae67520e8500ca0532
--- /dev/null
+++ b/runs/t1a/t1a_S2_C1_rand15/checkpoint-2160/tokenizer_config.json
@@ -0,0 +1,19 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/t1a/t1a_S2_C1_rand15/checkpoint-2160/trainer_state.json b/runs/t1a/t1a_S2_C1_rand15/checkpoint-2160/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f8cff2623d6d05454e8ac6b7bfd19a273319edd
--- /dev/null
+++ b/runs/t1a/t1a_S2_C1_rand15/checkpoint-2160/trainer_state.json
@@ -0,0 +1,1793 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 30.0,
+ "eval_steps": 72,
+ "global_step": 2160,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013961605584642234,
+ "grad_norm": 1.438960313796997,
+ "learning_rate": 0.0,
+ "loss": 12.021270751953125,
+ "step": 1
+ },
+ {
+ "epoch": 0.13961605584642234,
+ "grad_norm": 1.7513903379440308,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.761078728569878,
+ "step": 10
+ },
+ {
+ "epoch": 0.2792321116928447,
+ "grad_norm": 1.696722388267517,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 11.066200256347656,
+ "step": 20
+ },
+ {
+ "epoch": 0.418848167539267,
+ "grad_norm": 1.6440199613571167,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 10.309919738769532,
+ "step": 30
+ },
+ {
+ "epoch": 0.5584642233856894,
+ "grad_norm": 1.4849390983581543,
+ "learning_rate": 0.000234,
+ "loss": 9.273545074462891,
+ "step": 40
+ },
+ {
+ "epoch": 0.6980802792321117,
+ "grad_norm": 1.0249814987182617,
+ "learning_rate": 0.000294,
+ "loss": 8.27840805053711,
+ "step": 50
+ },
+ {
+ "epoch": 0.837696335078534,
+ "grad_norm": 0.28332021832466125,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 7.725849914550781,
+ "step": 60
+ },
+ {
+ "epoch": 0.9773123909249564,
+ "grad_norm": 1.4672844409942627,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 7.554132080078125,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.4967169761657715,
+ "eval_runtime": 27.3832,
+ "eval_samples_per_second": 89.142,
+ "eval_steps_per_second": 1.424,
+ "step": 72
+ },
+ {
+ "epoch": 1.111692844677138,
+ "grad_norm": 0.40830719470977783,
+ "learning_rate": 0.000474,
+ "loss": 7.4779502868652346,
+ "step": 80
+ },
+ {
+ "epoch": 1.2513089005235603,
+ "grad_norm": 0.3323667049407959,
+ "learning_rate": 0.000534,
+ "loss": 7.320184326171875,
+ "step": 90
+ },
+ {
+ "epoch": 1.3909249563699826,
+ "grad_norm": 0.48224157094955444,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 7.196783447265625,
+ "step": 100
+ },
+ {
+ "epoch": 1.530541012216405,
+ "grad_norm": 0.5221763253211975,
+ "learning_rate": 0.0006,
+ "loss": 7.082601928710938,
+ "step": 110
+ },
+ {
+ "epoch": 1.6701570680628273,
+ "grad_norm": 0.3978205621242523,
+ "learning_rate": 0.0006,
+ "loss": 6.955892944335938,
+ "step": 120
+ },
+ {
+ "epoch": 1.8097731239092496,
+ "grad_norm": 1.45212721824646,
+ "learning_rate": 0.0006,
+ "loss": 6.880769348144531,
+ "step": 130
+ },
+ {
+ "epoch": 1.949389179755672,
+ "grad_norm": 0.39447876811027527,
+ "learning_rate": 0.0006,
+ "loss": 6.787995910644531,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 7.036859512329102,
+ "eval_runtime": 3.2437,
+ "eval_samples_per_second": 752.541,
+ "eval_steps_per_second": 12.023,
+ "step": 144
+ },
+ {
+ "epoch": 2.0837696335078535,
+ "grad_norm": 0.9656530022621155,
+ "learning_rate": 0.0006,
+ "loss": 6.697367858886719,
+ "step": 150
+ },
+ {
+ "epoch": 2.223385689354276,
+ "grad_norm": 0.5059946179389954,
+ "learning_rate": 0.0006,
+ "loss": 6.621917724609375,
+ "step": 160
+ },
+ {
+ "epoch": 2.363001745200698,
+ "grad_norm": 0.701939046382904,
+ "learning_rate": 0.0006,
+ "loss": 6.5650169372558596,
+ "step": 170
+ },
+ {
+ "epoch": 2.5026178010471205,
+ "grad_norm": 0.42858022451400757,
+ "learning_rate": 0.0006,
+ "loss": 6.511680603027344,
+ "step": 180
+ },
+ {
+ "epoch": 2.642233856893543,
+ "grad_norm": 1.0025036334991455,
+ "learning_rate": 0.0006,
+ "loss": 6.450339508056641,
+ "step": 190
+ },
+ {
+ "epoch": 2.781849912739965,
+ "grad_norm": 0.8441563248634338,
+ "learning_rate": 0.0006,
+ "loss": 6.417574310302735,
+ "step": 200
+ },
+ {
+ "epoch": 2.9214659685863875,
+ "grad_norm": 0.6542202830314636,
+ "learning_rate": 0.0006,
+ "loss": 6.368536376953125,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 6.625904560089111,
+ "eval_runtime": 3.2598,
+ "eval_samples_per_second": 748.826,
+ "eval_steps_per_second": 11.964,
+ "step": 216
+ },
+ {
+ "epoch": 3.055846422338569,
+ "grad_norm": 0.6169397830963135,
+ "learning_rate": 0.0006,
+ "loss": 6.290319442749023,
+ "step": 220
+ },
+ {
+ "epoch": 3.195462478184991,
+ "grad_norm": 0.5342265367507935,
+ "learning_rate": 0.0006,
+ "loss": 6.258647918701172,
+ "step": 230
+ },
+ {
+ "epoch": 3.3350785340314135,
+ "grad_norm": 0.34757575392723083,
+ "learning_rate": 0.0006,
+ "loss": 6.183958053588867,
+ "step": 240
+ },
+ {
+ "epoch": 3.474694589877836,
+ "grad_norm": 0.7538574934005737,
+ "learning_rate": 0.0006,
+ "loss": 6.183706665039063,
+ "step": 250
+ },
+ {
+ "epoch": 3.614310645724258,
+ "grad_norm": 0.7209551334381104,
+ "learning_rate": 0.0006,
+ "loss": 6.154908752441406,
+ "step": 260
+ },
+ {
+ "epoch": 3.7539267015706805,
+ "grad_norm": 0.4867055118083954,
+ "learning_rate": 0.0006,
+ "loss": 6.074378967285156,
+ "step": 270
+ },
+ {
+ "epoch": 3.893542757417103,
+ "grad_norm": 0.7056074738502502,
+ "learning_rate": 0.0006,
+ "loss": 6.0574394226074215,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 6.1944122314453125,
+ "eval_runtime": 3.2429,
+ "eval_samples_per_second": 752.716,
+ "eval_steps_per_second": 12.026,
+ "step": 288
+ },
+ {
+ "epoch": 4.027923211169284,
+ "grad_norm": 0.36485642194747925,
+ "learning_rate": 0.0006,
+ "loss": 5.998421478271484,
+ "step": 290
+ },
+ {
+ "epoch": 4.167539267015707,
+ "grad_norm": 1.1686674356460571,
+ "learning_rate": 0.0006,
+ "loss": 5.932280731201172,
+ "step": 300
+ },
+ {
+ "epoch": 4.307155322862129,
+ "grad_norm": 0.7692177295684814,
+ "learning_rate": 0.0006,
+ "loss": 5.942491149902343,
+ "step": 310
+ },
+ {
+ "epoch": 4.446771378708552,
+ "grad_norm": 0.4043848216533661,
+ "learning_rate": 0.0006,
+ "loss": 5.892348098754883,
+ "step": 320
+ },
+ {
+ "epoch": 4.5863874345549736,
+ "grad_norm": 0.911917507648468,
+ "learning_rate": 0.0006,
+ "loss": 5.847317886352539,
+ "step": 330
+ },
+ {
+ "epoch": 4.726003490401396,
+ "grad_norm": 0.6406515836715698,
+ "learning_rate": 0.0006,
+ "loss": 5.82725715637207,
+ "step": 340
+ },
+ {
+ "epoch": 4.865619546247818,
+ "grad_norm": 0.2842332422733307,
+ "learning_rate": 0.0006,
+ "loss": 5.753609085083008,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.7789965867996216,
+ "learning_rate": 0.0006,
+ "loss": 5.765028381347657,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 5.913822650909424,
+ "eval_runtime": 3.2319,
+ "eval_samples_per_second": 755.284,
+ "eval_steps_per_second": 12.067,
+ "step": 360
+ },
+ {
+ "epoch": 5.139616055846423,
+ "grad_norm": 0.4087577760219574,
+ "learning_rate": 0.0006,
+ "loss": 5.741168212890625,
+ "step": 370
+ },
+ {
+ "epoch": 5.279232111692845,
+ "grad_norm": 0.295691579580307,
+ "learning_rate": 0.0006,
+ "loss": 5.668488693237305,
+ "step": 380
+ },
+ {
+ "epoch": 5.418848167539267,
+ "grad_norm": 0.3318697214126587,
+ "learning_rate": 0.0006,
+ "loss": 5.617916107177734,
+ "step": 390
+ },
+ {
+ "epoch": 5.558464223385689,
+ "grad_norm": 0.7654287815093994,
+ "learning_rate": 0.0006,
+ "loss": 5.5916282653808596,
+ "step": 400
+ },
+ {
+ "epoch": 5.698080279232112,
+ "grad_norm": 0.4317888915538788,
+ "learning_rate": 0.0006,
+ "loss": 5.598070526123047,
+ "step": 410
+ },
+ {
+ "epoch": 5.837696335078534,
+ "grad_norm": 0.30880123376846313,
+ "learning_rate": 0.0006,
+ "loss": 5.542387390136719,
+ "step": 420
+ },
+ {
+ "epoch": 5.977312390924956,
+ "grad_norm": 0.5962476134300232,
+ "learning_rate": 0.0006,
+ "loss": 5.502596664428711,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 5.518165111541748,
+ "eval_runtime": 3.244,
+ "eval_samples_per_second": 752.467,
+ "eval_steps_per_second": 12.022,
+ "step": 432
+ },
+ {
+ "epoch": 6.111692844677138,
+ "grad_norm": 0.4179597795009613,
+ "learning_rate": 0.0006,
+ "loss": 5.466204071044922,
+ "step": 440
+ },
+ {
+ "epoch": 6.2513089005235605,
+ "grad_norm": 0.5469250082969666,
+ "learning_rate": 0.0006,
+ "loss": 5.424791717529297,
+ "step": 450
+ },
+ {
+ "epoch": 6.390924956369982,
+ "grad_norm": 0.7651413083076477,
+ "learning_rate": 0.0006,
+ "loss": 5.387444305419922,
+ "step": 460
+ },
+ {
+ "epoch": 6.530541012216405,
+ "grad_norm": 0.4763544499874115,
+ "learning_rate": 0.0006,
+ "loss": 5.372788619995117,
+ "step": 470
+ },
+ {
+ "epoch": 6.670157068062827,
+ "grad_norm": 0.46549415588378906,
+ "learning_rate": 0.0006,
+ "loss": 5.327343368530274,
+ "step": 480
+ },
+ {
+ "epoch": 6.80977312390925,
+ "grad_norm": 0.6359885931015015,
+ "learning_rate": 0.0006,
+ "loss": 5.312184906005859,
+ "step": 490
+ },
+ {
+ "epoch": 6.949389179755672,
+ "grad_norm": 0.9627570509910583,
+ "learning_rate": 0.0006,
+ "loss": 5.310484313964844,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 5.279782772064209,
+ "eval_runtime": 3.2437,
+ "eval_samples_per_second": 752.535,
+ "eval_steps_per_second": 12.023,
+ "step": 504
+ },
+ {
+ "epoch": 7.0837696335078535,
+ "grad_norm": 0.4257539212703705,
+ "learning_rate": 0.0006,
+ "loss": 5.269083404541016,
+ "step": 510
+ },
+ {
+ "epoch": 7.223385689354275,
+ "grad_norm": 0.3374505341053009,
+ "learning_rate": 0.0006,
+ "loss": 5.212802124023438,
+ "step": 520
+ },
+ {
+ "epoch": 7.363001745200698,
+ "grad_norm": 0.5186793208122253,
+ "learning_rate": 0.0006,
+ "loss": 5.166359710693359,
+ "step": 530
+ },
+ {
+ "epoch": 7.50261780104712,
+ "grad_norm": 0.44578421115875244,
+ "learning_rate": 0.0006,
+ "loss": 5.147264862060547,
+ "step": 540
+ },
+ {
+ "epoch": 7.642233856893543,
+ "grad_norm": 0.5879621505737305,
+ "learning_rate": 0.0006,
+ "loss": 5.133060455322266,
+ "step": 550
+ },
+ {
+ "epoch": 7.781849912739965,
+ "grad_norm": 0.7359283566474915,
+ "learning_rate": 0.0006,
+ "loss": 5.144127655029297,
+ "step": 560
+ },
+ {
+ "epoch": 7.9214659685863875,
+ "grad_norm": 0.33267563581466675,
+ "learning_rate": 0.0006,
+ "loss": 5.115408325195313,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.982746124267578,
+ "eval_runtime": 3.2572,
+ "eval_samples_per_second": 749.418,
+ "eval_steps_per_second": 11.973,
+ "step": 576
+ },
+ {
+ "epoch": 8.055846422338568,
+ "grad_norm": 0.24862998723983765,
+ "learning_rate": 0.0006,
+ "loss": 5.04302749633789,
+ "step": 580
+ },
+ {
+ "epoch": 8.195462478184991,
+ "grad_norm": 0.2427068054676056,
+ "learning_rate": 0.0006,
+ "loss": 4.991822052001953,
+ "step": 590
+ },
+ {
+ "epoch": 8.335078534031414,
+ "grad_norm": 0.5329022407531738,
+ "learning_rate": 0.0006,
+ "loss": 4.980979919433594,
+ "step": 600
+ },
+ {
+ "epoch": 8.474694589877837,
+ "grad_norm": 0.531551718711853,
+ "learning_rate": 0.0006,
+ "loss": 4.976587295532227,
+ "step": 610
+ },
+ {
+ "epoch": 8.614310645724258,
+ "grad_norm": 0.36841699481010437,
+ "learning_rate": 0.0006,
+ "loss": 4.963835144042969,
+ "step": 620
+ },
+ {
+ "epoch": 8.75392670157068,
+ "grad_norm": 0.3721601665019989,
+ "learning_rate": 0.0006,
+ "loss": 4.921863555908203,
+ "step": 630
+ },
+ {
+ "epoch": 8.893542757417103,
+ "grad_norm": 0.2814253270626068,
+ "learning_rate": 0.0006,
+ "loss": 4.900112152099609,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.70157527923584,
+ "eval_runtime": 3.2577,
+ "eval_samples_per_second": 749.302,
+ "eval_steps_per_second": 11.972,
+ "step": 648
+ },
+ {
+ "epoch": 9.027923211169284,
+ "grad_norm": 0.2770821750164032,
+ "learning_rate": 0.0006,
+ "loss": 4.858328247070313,
+ "step": 650
+ },
+ {
+ "epoch": 9.167539267015707,
+ "grad_norm": 0.3335670530796051,
+ "learning_rate": 0.0006,
+ "loss": 4.806454467773437,
+ "step": 660
+ },
+ {
+ "epoch": 9.30715532286213,
+ "grad_norm": 0.5527796149253845,
+ "learning_rate": 0.0006,
+ "loss": 4.8165546417236325,
+ "step": 670
+ },
+ {
+ "epoch": 9.44677137870855,
+ "grad_norm": 0.3936336934566498,
+ "learning_rate": 0.0006,
+ "loss": 4.826232147216797,
+ "step": 680
+ },
+ {
+ "epoch": 9.586387434554974,
+ "grad_norm": 0.34287649393081665,
+ "learning_rate": 0.0006,
+ "loss": 4.7986595153808596,
+ "step": 690
+ },
+ {
+ "epoch": 9.726003490401396,
+ "grad_norm": 0.2968748211860657,
+ "learning_rate": 0.0006,
+ "loss": 4.766458892822266,
+ "step": 700
+ },
+ {
+ "epoch": 9.86561954624782,
+ "grad_norm": 0.24643062055110931,
+ "learning_rate": 0.0006,
+ "loss": 4.723728942871094,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.3842594623565674,
+ "learning_rate": 0.0006,
+ "loss": 4.694996643066406,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.531019687652588,
+ "eval_runtime": 3.2458,
+ "eval_samples_per_second": 752.058,
+ "eval_steps_per_second": 12.016,
+ "step": 720
+ },
+ {
+ "epoch": 10.139616055846423,
+ "grad_norm": 0.4676571190357208,
+ "learning_rate": 0.0006,
+ "loss": 4.67070198059082,
+ "step": 730
+ },
+ {
+ "epoch": 10.279232111692846,
+ "grad_norm": 0.4936009347438812,
+ "learning_rate": 0.0006,
+ "loss": 4.687448883056641,
+ "step": 740
+ },
+ {
+ "epoch": 10.418848167539267,
+ "grad_norm": 0.33258292078971863,
+ "learning_rate": 0.0006,
+ "loss": 4.670026016235352,
+ "step": 750
+ },
+ {
+ "epoch": 10.55846422338569,
+ "grad_norm": 0.2642180323600769,
+ "learning_rate": 0.0006,
+ "loss": 4.6580047607421875,
+ "step": 760
+ },
+ {
+ "epoch": 10.698080279232112,
+ "grad_norm": 0.18084998428821564,
+ "learning_rate": 0.0006,
+ "loss": 4.6212505340576175,
+ "step": 770
+ },
+ {
+ "epoch": 10.837696335078533,
+ "grad_norm": 0.24538297951221466,
+ "learning_rate": 0.0006,
+ "loss": 4.600024032592773,
+ "step": 780
+ },
+ {
+ "epoch": 10.977312390924956,
+ "grad_norm": 0.24823062121868134,
+ "learning_rate": 0.0006,
+ "loss": 4.585308837890625,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.378584384918213,
+ "eval_runtime": 3.2411,
+ "eval_samples_per_second": 753.13,
+ "eval_steps_per_second": 12.033,
+ "step": 792
+ },
+ {
+ "epoch": 11.111692844677139,
+ "grad_norm": 0.37619447708129883,
+ "learning_rate": 0.0006,
+ "loss": 4.562310028076172,
+ "step": 800
+ },
+ {
+ "epoch": 11.25130890052356,
+ "grad_norm": 0.356943815946579,
+ "learning_rate": 0.0006,
+ "loss": 4.530385589599609,
+ "step": 810
+ },
+ {
+ "epoch": 11.390924956369982,
+ "grad_norm": 0.34838002920150757,
+ "learning_rate": 0.0006,
+ "loss": 4.530497741699219,
+ "step": 820
+ },
+ {
+ "epoch": 11.530541012216405,
+ "grad_norm": 0.28737014532089233,
+ "learning_rate": 0.0006,
+ "loss": 4.528908538818359,
+ "step": 830
+ },
+ {
+ "epoch": 11.670157068062828,
+ "grad_norm": 0.26005175709724426,
+ "learning_rate": 0.0006,
+ "loss": 4.508984375,
+ "step": 840
+ },
+ {
+ "epoch": 11.809773123909249,
+ "grad_norm": 0.26725491881370544,
+ "learning_rate": 0.0006,
+ "loss": 4.491153335571289,
+ "step": 850
+ },
+ {
+ "epoch": 11.949389179755672,
+ "grad_norm": 0.22221490740776062,
+ "learning_rate": 0.0006,
+ "loss": 4.490458679199219,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.280045032501221,
+ "eval_runtime": 3.2461,
+ "eval_samples_per_second": 751.971,
+ "eval_steps_per_second": 12.014,
+ "step": 864
+ },
+ {
+ "epoch": 12.083769633507853,
+ "grad_norm": 0.3112393617630005,
+ "learning_rate": 0.0006,
+ "loss": 4.451058959960937,
+ "step": 870
+ },
+ {
+ "epoch": 12.223385689354275,
+ "grad_norm": 0.4979000687599182,
+ "learning_rate": 0.0006,
+ "loss": 4.428517913818359,
+ "step": 880
+ },
+ {
+ "epoch": 12.363001745200698,
+ "grad_norm": 0.343538373708725,
+ "learning_rate": 0.0006,
+ "loss": 4.459210205078125,
+ "step": 890
+ },
+ {
+ "epoch": 12.502617801047121,
+ "grad_norm": 0.3375740349292755,
+ "learning_rate": 0.0006,
+ "loss": 4.431212615966797,
+ "step": 900
+ },
+ {
+ "epoch": 12.642233856893542,
+ "grad_norm": 0.21244044601917267,
+ "learning_rate": 0.0006,
+ "loss": 4.428363037109375,
+ "step": 910
+ },
+ {
+ "epoch": 12.781849912739965,
+ "grad_norm": 0.22292549908161163,
+ "learning_rate": 0.0006,
+ "loss": 4.409710693359375,
+ "step": 920
+ },
+ {
+ "epoch": 12.921465968586388,
+ "grad_norm": 0.2266375720500946,
+ "learning_rate": 0.0006,
+ "loss": 4.391383361816406,
+ "step": 930
+ },
+ {
+ "epoch": 13.0,
+ "eval_loss": 4.193124771118164,
+ "eval_runtime": 3.243,
+ "eval_samples_per_second": 752.699,
+ "eval_steps_per_second": 12.026,
+ "step": 936
+ },
+ {
+ "epoch": 13.055846422338568,
+ "grad_norm": 0.2656148076057434,
+ "learning_rate": 0.0006,
+ "loss": 4.356477355957031,
+ "step": 940
+ },
+ {
+ "epoch": 13.195462478184991,
+ "grad_norm": 0.42341428995132446,
+ "learning_rate": 0.0006,
+ "loss": 4.350735092163086,
+ "step": 950
+ },
+ {
+ "epoch": 13.335078534031414,
+ "grad_norm": 0.4301425516605377,
+ "learning_rate": 0.0006,
+ "loss": 4.371968078613281,
+ "step": 960
+ },
+ {
+ "epoch": 13.474694589877837,
+ "grad_norm": 0.2854589521884918,
+ "learning_rate": 0.0006,
+ "loss": 4.349368286132813,
+ "step": 970
+ },
+ {
+ "epoch": 13.614310645724258,
+ "grad_norm": 0.2292899340391159,
+ "learning_rate": 0.0006,
+ "loss": 4.332583618164063,
+ "step": 980
+ },
+ {
+ "epoch": 13.75392670157068,
+ "grad_norm": 0.22543831169605255,
+ "learning_rate": 0.0006,
+ "loss": 4.328607177734375,
+ "step": 990
+ },
+ {
+ "epoch": 13.893542757417103,
+ "grad_norm": 0.26455968618392944,
+ "learning_rate": 0.0006,
+ "loss": 4.32293701171875,
+ "step": 1000
+ },
+ {
+ "epoch": 14.0,
+ "eval_loss": 4.136460781097412,
+ "eval_runtime": 3.2444,
+ "eval_samples_per_second": 752.38,
+ "eval_steps_per_second": 12.021,
+ "step": 1008
+ },
+ {
+ "epoch": 14.027923211169284,
+ "grad_norm": 0.3389642536640167,
+ "learning_rate": 0.0006,
+ "loss": 4.305249786376953,
+ "step": 1010
+ },
+ {
+ "epoch": 14.167539267015707,
+ "grad_norm": 0.3664684593677521,
+ "learning_rate": 0.0006,
+ "loss": 4.277291870117187,
+ "step": 1020
+ },
+ {
+ "epoch": 14.30715532286213,
+ "grad_norm": 0.287652850151062,
+ "learning_rate": 0.0006,
+ "loss": 4.27338752746582,
+ "step": 1030
+ },
+ {
+ "epoch": 14.44677137870855,
+ "grad_norm": 0.28992146253585815,
+ "learning_rate": 0.0006,
+ "loss": 4.279605484008789,
+ "step": 1040
+ },
+ {
+ "epoch": 14.586387434554974,
+ "grad_norm": 0.2611526548862457,
+ "learning_rate": 0.0006,
+ "loss": 4.263419342041016,
+ "step": 1050
+ },
+ {
+ "epoch": 14.726003490401396,
+ "grad_norm": 0.231856569647789,
+ "learning_rate": 0.0006,
+ "loss": 4.261050415039063,
+ "step": 1060
+ },
+ {
+ "epoch": 14.86561954624782,
+ "grad_norm": 0.23650093376636505,
+ "learning_rate": 0.0006,
+ "loss": 4.252535247802735,
+ "step": 1070
+ },
+ {
+ "epoch": 15.0,
+ "grad_norm": 0.32511240243911743,
+ "learning_rate": 0.0006,
+ "loss": 4.259856033325195,
+ "step": 1080
+ },
+ {
+ "epoch": 15.0,
+ "eval_loss": 4.084939479827881,
+ "eval_runtime": 3.2432,
+ "eval_samples_per_second": 752.662,
+ "eval_steps_per_second": 12.025,
+ "step": 1080
+ },
+ {
+ "epoch": 15.139616055846423,
+ "grad_norm": 0.36072540283203125,
+ "learning_rate": 0.0006,
+ "loss": 4.2040458679199215,
+ "step": 1090
+ },
+ {
+ "epoch": 15.279232111692846,
+ "grad_norm": 0.5160514116287231,
+ "learning_rate": 0.0006,
+ "loss": 4.217556762695312,
+ "step": 1100
+ },
+ {
+ "epoch": 15.418848167539267,
+ "grad_norm": 0.35392868518829346,
+ "learning_rate": 0.0006,
+ "loss": 4.250083923339844,
+ "step": 1110
+ },
+ {
+ "epoch": 15.55846422338569,
+ "grad_norm": 0.2551209628582001,
+ "learning_rate": 0.0006,
+ "loss": 4.230625915527344,
+ "step": 1120
+ },
+ {
+ "epoch": 15.698080279232112,
+ "grad_norm": 0.18756623566150665,
+ "learning_rate": 0.0006,
+ "loss": 4.225308227539062,
+ "step": 1130
+ },
+ {
+ "epoch": 15.837696335078533,
+ "grad_norm": 0.20765367150306702,
+ "learning_rate": 0.0006,
+ "loss": 4.208599090576172,
+ "step": 1140
+ },
+ {
+ "epoch": 15.977312390924956,
+ "grad_norm": 0.21032412350177765,
+ "learning_rate": 0.0006,
+ "loss": 4.1894683837890625,
+ "step": 1150
+ },
+ {
+ "epoch": 16.0,
+ "eval_loss": 4.038693428039551,
+ "eval_runtime": 3.2442,
+ "eval_samples_per_second": 752.429,
+ "eval_steps_per_second": 12.022,
+ "step": 1152
+ },
+ {
+ "epoch": 16.111692844677137,
+ "grad_norm": 0.2504013776779175,
+ "learning_rate": 0.0006,
+ "loss": 4.161359405517578,
+ "step": 1160
+ },
+ {
+ "epoch": 16.25130890052356,
+ "grad_norm": 0.26923617720603943,
+ "learning_rate": 0.0006,
+ "loss": 4.151763916015625,
+ "step": 1170
+ },
+ {
+ "epoch": 16.390924956369982,
+ "grad_norm": 0.24602985382080078,
+ "learning_rate": 0.0006,
+ "loss": 4.147951126098633,
+ "step": 1180
+ },
+ {
+ "epoch": 16.530541012216403,
+ "grad_norm": 0.2466205358505249,
+ "learning_rate": 0.0006,
+ "loss": 4.147978591918945,
+ "step": 1190
+ },
+ {
+ "epoch": 16.670157068062828,
+ "grad_norm": 0.2613685429096222,
+ "learning_rate": 0.0006,
+ "loss": 4.152555465698242,
+ "step": 1200
+ },
+ {
+ "epoch": 16.80977312390925,
+ "grad_norm": 0.28424179553985596,
+ "learning_rate": 0.0006,
+ "loss": 4.169595336914062,
+ "step": 1210
+ },
+ {
+ "epoch": 16.949389179755673,
+ "grad_norm": 0.2104388326406479,
+ "learning_rate": 0.0006,
+ "loss": 4.1387176513671875,
+ "step": 1220
+ },
+ {
+ "epoch": 17.0,
+ "eval_loss": 4.007999897003174,
+ "eval_runtime": 3.2558,
+ "eval_samples_per_second": 749.743,
+ "eval_steps_per_second": 11.979,
+ "step": 1224
+ },
+ {
+ "epoch": 17.083769633507853,
+ "grad_norm": 0.27143770456314087,
+ "learning_rate": 0.0006,
+ "loss": 4.1279747009277346,
+ "step": 1230
+ },
+ {
+ "epoch": 17.223385689354277,
+ "grad_norm": 0.23995907604694366,
+ "learning_rate": 0.0006,
+ "loss": 4.085084915161133,
+ "step": 1240
+ },
+ {
+ "epoch": 17.363001745200698,
+ "grad_norm": 0.27527937293052673,
+ "learning_rate": 0.0006,
+ "loss": 4.101887512207031,
+ "step": 1250
+ },
+ {
+ "epoch": 17.50261780104712,
+ "grad_norm": 0.34114211797714233,
+ "learning_rate": 0.0006,
+ "loss": 4.110912704467774,
+ "step": 1260
+ },
+ {
+ "epoch": 17.642233856893544,
+ "grad_norm": 0.31985175609588623,
+ "learning_rate": 0.0006,
+ "loss": 4.1121879577636715,
+ "step": 1270
+ },
+ {
+ "epoch": 17.781849912739965,
+ "grad_norm": 0.23774679005146027,
+ "learning_rate": 0.0006,
+ "loss": 4.099986267089844,
+ "step": 1280
+ },
+ {
+ "epoch": 17.921465968586386,
+ "grad_norm": 0.2219526767730713,
+ "learning_rate": 0.0006,
+ "loss": 4.114411163330078,
+ "step": 1290
+ },
+ {
+ "epoch": 18.0,
+ "eval_loss": 3.9769914150238037,
+ "eval_runtime": 3.2487,
+ "eval_samples_per_second": 751.374,
+ "eval_steps_per_second": 12.005,
+ "step": 1296
+ },
+ {
+ "epoch": 18.05584642233857,
+ "grad_norm": 0.2768155038356781,
+ "learning_rate": 0.0006,
+ "loss": 4.086803436279297,
+ "step": 1300
+ },
+ {
+ "epoch": 18.195462478184993,
+ "grad_norm": 0.2482820749282837,
+ "learning_rate": 0.0006,
+ "loss": 4.047364044189453,
+ "step": 1310
+ },
+ {
+ "epoch": 18.335078534031414,
+ "grad_norm": 0.22649119794368744,
+ "learning_rate": 0.0006,
+ "loss": 4.050416564941406,
+ "step": 1320
+ },
+ {
+ "epoch": 18.474694589877835,
+ "grad_norm": 0.27260687947273254,
+ "learning_rate": 0.0006,
+ "loss": 4.065253448486328,
+ "step": 1330
+ },
+ {
+ "epoch": 18.61431064572426,
+ "grad_norm": 0.2676929533481598,
+ "learning_rate": 0.0006,
+ "loss": 4.063740539550781,
+ "step": 1340
+ },
+ {
+ "epoch": 18.75392670157068,
+ "grad_norm": 0.27753403782844543,
+ "learning_rate": 0.0006,
+ "loss": 4.073302459716797,
+ "step": 1350
+ },
+ {
+ "epoch": 18.8935427574171,
+ "grad_norm": 0.24279844760894775,
+ "learning_rate": 0.0006,
+ "loss": 4.058047866821289,
+ "step": 1360
+ },
+ {
+ "epoch": 19.0,
+ "eval_loss": 3.949812889099121,
+ "eval_runtime": 3.2526,
+ "eval_samples_per_second": 750.481,
+ "eval_steps_per_second": 11.99,
+ "step": 1368
+ },
+ {
+ "epoch": 19.027923211169284,
+ "grad_norm": 0.2772994637489319,
+ "learning_rate": 0.0006,
+ "loss": 4.048649597167969,
+ "step": 1370
+ },
+ {
+ "epoch": 19.167539267015705,
+ "grad_norm": 0.2842816114425659,
+ "learning_rate": 0.0006,
+ "loss": 3.9972496032714844,
+ "step": 1380
+ },
+ {
+ "epoch": 19.30715532286213,
+ "grad_norm": 0.266347736120224,
+ "learning_rate": 0.0006,
+ "loss": 4.001751327514649,
+ "step": 1390
+ },
+ {
+ "epoch": 19.44677137870855,
+ "grad_norm": 0.2815294861793518,
+ "learning_rate": 0.0006,
+ "loss": 4.0252124786376955,
+ "step": 1400
+ },
+ {
+ "epoch": 19.586387434554975,
+ "grad_norm": 0.29131412506103516,
+ "learning_rate": 0.0006,
+ "loss": 4.028485870361328,
+ "step": 1410
+ },
+ {
+ "epoch": 19.726003490401396,
+ "grad_norm": 0.24175862967967987,
+ "learning_rate": 0.0006,
+ "loss": 4.022135925292969,
+ "step": 1420
+ },
+ {
+ "epoch": 19.865619546247817,
+ "grad_norm": 0.22423435747623444,
+ "learning_rate": 0.0006,
+ "loss": 4.0257820129394535,
+ "step": 1430
+ },
+ {
+ "epoch": 20.0,
+ "grad_norm": 0.24630773067474365,
+ "learning_rate": 0.0006,
+ "loss": 4.036606979370117,
+ "step": 1440
+ },
+ {
+ "epoch": 20.0,
+ "eval_loss": 3.9335389137268066,
+ "eval_runtime": 3.24,
+ "eval_samples_per_second": 753.402,
+ "eval_steps_per_second": 12.037,
+ "step": 1440
+ },
+ {
+ "epoch": 20.13961605584642,
+ "grad_norm": 0.2827102541923523,
+ "learning_rate": 0.0006,
+ "loss": 3.9775035858154295,
+ "step": 1450
+ },
+ {
+ "epoch": 20.279232111692846,
+ "grad_norm": 0.24853454530239105,
+ "learning_rate": 0.0006,
+ "loss": 3.9664005279541015,
+ "step": 1460
+ },
+ {
+ "epoch": 20.418848167539267,
+ "grad_norm": 0.323213666677475,
+ "learning_rate": 0.0006,
+ "loss": 3.989998626708984,
+ "step": 1470
+ },
+ {
+ "epoch": 20.55846422338569,
+ "grad_norm": 0.2930547893047333,
+ "learning_rate": 0.0006,
+ "loss": 3.992165374755859,
+ "step": 1480
+ },
+ {
+ "epoch": 20.698080279232112,
+ "grad_norm": 0.2545824646949768,
+ "learning_rate": 0.0006,
+ "loss": 3.982561492919922,
+ "step": 1490
+ },
+ {
+ "epoch": 20.837696335078533,
+ "grad_norm": 0.24988673627376556,
+ "learning_rate": 0.0006,
+ "loss": 3.9926902770996096,
+ "step": 1500
+ },
+ {
+ "epoch": 20.977312390924958,
+ "grad_norm": 0.22226130962371826,
+ "learning_rate": 0.0006,
+ "loss": 3.992121124267578,
+ "step": 1510
+ },
+ {
+ "epoch": 21.0,
+ "eval_loss": 3.9203379154205322,
+ "eval_runtime": 3.2442,
+ "eval_samples_per_second": 752.43,
+ "eval_steps_per_second": 12.022,
+ "step": 1512
+ },
+ {
+ "epoch": 21.111692844677137,
+ "grad_norm": 0.32606032490730286,
+ "learning_rate": 0.0006,
+ "loss": 3.947393035888672,
+ "step": 1520
+ },
+ {
+ "epoch": 21.25130890052356,
+ "grad_norm": 0.32446718215942383,
+ "learning_rate": 0.0006,
+ "loss": 3.9485897064208983,
+ "step": 1530
+ },
+ {
+ "epoch": 21.390924956369982,
+ "grad_norm": 0.2575104832649231,
+ "learning_rate": 0.0006,
+ "loss": 3.938507843017578,
+ "step": 1540
+ },
+ {
+ "epoch": 21.530541012216403,
+ "grad_norm": 0.2790096700191498,
+ "learning_rate": 0.0006,
+ "loss": 3.9489200592041014,
+ "step": 1550
+ },
+ {
+ "epoch": 21.670157068062828,
+ "grad_norm": 0.2349964827299118,
+ "learning_rate": 0.0006,
+ "loss": 3.9602699279785156,
+ "step": 1560
+ },
+ {
+ "epoch": 21.80977312390925,
+ "grad_norm": 0.23055724799633026,
+ "learning_rate": 0.0006,
+ "loss": 3.956432342529297,
+ "step": 1570
+ },
+ {
+ "epoch": 21.949389179755673,
+ "grad_norm": 0.2546336054801941,
+ "learning_rate": 0.0006,
+ "loss": 3.9740161895751953,
+ "step": 1580
+ },
+ {
+ "epoch": 22.0,
+ "eval_loss": 3.9034576416015625,
+ "eval_runtime": 3.256,
+ "eval_samples_per_second": 749.69,
+ "eval_steps_per_second": 11.978,
+ "step": 1584
+ },
+ {
+ "epoch": 22.083769633507853,
+ "grad_norm": 0.33441808819770813,
+ "learning_rate": 0.0006,
+ "loss": 3.919890594482422,
+ "step": 1590
+ },
+ {
+ "epoch": 22.223385689354277,
+ "grad_norm": 0.27338382601737976,
+ "learning_rate": 0.0006,
+ "loss": 3.8977729797363283,
+ "step": 1600
+ },
+ {
+ "epoch": 22.363001745200698,
+ "grad_norm": 0.23193812370300293,
+ "learning_rate": 0.0006,
+ "loss": 3.896637725830078,
+ "step": 1610
+ },
+ {
+ "epoch": 22.50261780104712,
+ "grad_norm": 0.25093206763267517,
+ "learning_rate": 0.0006,
+ "loss": 3.92679443359375,
+ "step": 1620
+ },
+ {
+ "epoch": 22.642233856893544,
+ "grad_norm": 0.2428864985704422,
+ "learning_rate": 0.0006,
+ "loss": 3.92083625793457,
+ "step": 1630
+ },
+ {
+ "epoch": 22.781849912739965,
+ "grad_norm": 0.2282162457704544,
+ "learning_rate": 0.0006,
+ "loss": 3.930418014526367,
+ "step": 1640
+ },
+ {
+ "epoch": 22.921465968586386,
+ "grad_norm": 0.2241750806570053,
+ "learning_rate": 0.0006,
+ "loss": 3.929116058349609,
+ "step": 1650
+ },
+ {
+ "epoch": 23.0,
+ "eval_loss": 3.9006025791168213,
+ "eval_runtime": 3.2598,
+ "eval_samples_per_second": 748.828,
+ "eval_steps_per_second": 11.964,
+ "step": 1656
+ },
+ {
+ "epoch": 23.05584642233857,
+ "grad_norm": 0.2826197147369385,
+ "learning_rate": 0.0006,
+ "loss": 3.915720748901367,
+ "step": 1660
+ },
+ {
+ "epoch": 23.195462478184993,
+ "grad_norm": 0.3196408152580261,
+ "learning_rate": 0.0006,
+ "loss": 3.8664260864257813,
+ "step": 1670
+ },
+ {
+ "epoch": 23.335078534031414,
+ "grad_norm": 0.2874223291873932,
+ "learning_rate": 0.0006,
+ "loss": 3.8865821838378904,
+ "step": 1680
+ },
+ {
+ "epoch": 23.474694589877835,
+ "grad_norm": 0.35815414786338806,
+ "learning_rate": 0.0006,
+ "loss": 3.900670623779297,
+ "step": 1690
+ },
+ {
+ "epoch": 23.61431064572426,
+ "grad_norm": 0.24199464917182922,
+ "learning_rate": 0.0006,
+ "loss": 3.8978904724121093,
+ "step": 1700
+ },
+ {
+ "epoch": 23.75392670157068,
+ "grad_norm": 0.22838062047958374,
+ "learning_rate": 0.0006,
+ "loss": 3.9044815063476563,
+ "step": 1710
+ },
+ {
+ "epoch": 23.8935427574171,
+ "grad_norm": 0.2266451120376587,
+ "learning_rate": 0.0006,
+ "loss": 3.8952095031738283,
+ "step": 1720
+ },
+ {
+ "epoch": 24.0,
+ "eval_loss": 3.8856241703033447,
+ "eval_runtime": 3.2426,
+ "eval_samples_per_second": 752.781,
+ "eval_steps_per_second": 12.027,
+ "step": 1728
+ },
+ {
+ "epoch": 24.027923211169284,
+ "grad_norm": 0.25755101442337036,
+ "learning_rate": 0.0006,
+ "loss": 3.886323165893555,
+ "step": 1730
+ },
+ {
+ "epoch": 24.167539267015705,
+ "grad_norm": 0.2644495666027069,
+ "learning_rate": 0.0006,
+ "loss": 3.838153076171875,
+ "step": 1740
+ },
+ {
+ "epoch": 24.30715532286213,
+ "grad_norm": 0.315290629863739,
+ "learning_rate": 0.0006,
+ "loss": 3.8442230224609375,
+ "step": 1750
+ },
+ {
+ "epoch": 24.44677137870855,
+ "grad_norm": 0.2414376586675644,
+ "learning_rate": 0.0006,
+ "loss": 3.8515029907226563,
+ "step": 1760
+ },
+ {
+ "epoch": 24.586387434554975,
+ "grad_norm": 0.222580686211586,
+ "learning_rate": 0.0006,
+ "loss": 3.8714954376220705,
+ "step": 1770
+ },
+ {
+ "epoch": 24.726003490401396,
+ "grad_norm": 0.2727249264717102,
+ "learning_rate": 0.0006,
+ "loss": 3.8663612365722657,
+ "step": 1780
+ },
+ {
+ "epoch": 24.865619546247817,
+ "grad_norm": 0.23636268079280853,
+ "learning_rate": 0.0006,
+ "loss": 3.881989288330078,
+ "step": 1790
+ },
+ {
+ "epoch": 25.0,
+ "grad_norm": 0.2542417645454407,
+ "learning_rate": 0.0006,
+ "loss": 3.8787811279296873,
+ "step": 1800
+ },
+ {
+ "epoch": 25.0,
+ "eval_loss": 3.879793643951416,
+ "eval_runtime": 3.2455,
+ "eval_samples_per_second": 752.129,
+ "eval_steps_per_second": 12.017,
+ "step": 1800
+ },
+ {
+ "epoch": 25.13961605584642,
+ "grad_norm": 0.30943143367767334,
+ "learning_rate": 0.0006,
+ "loss": 3.8055301666259767,
+ "step": 1810
+ },
+ {
+ "epoch": 25.279232111692846,
+ "grad_norm": 0.3107646405696869,
+ "learning_rate": 0.0006,
+ "loss": 3.830365753173828,
+ "step": 1820
+ },
+ {
+ "epoch": 25.418848167539267,
+ "grad_norm": 0.2445444017648697,
+ "learning_rate": 0.0006,
+ "loss": 3.8284027099609377,
+ "step": 1830
+ },
+ {
+ "epoch": 25.55846422338569,
+ "grad_norm": 0.27556443214416504,
+ "learning_rate": 0.0006,
+ "loss": 3.8477210998535156,
+ "step": 1840
+ },
+ {
+ "epoch": 25.698080279232112,
+ "grad_norm": 0.2213355302810669,
+ "learning_rate": 0.0006,
+ "loss": 3.8447418212890625,
+ "step": 1850
+ },
+ {
+ "epoch": 25.837696335078533,
+ "grad_norm": 0.23272836208343506,
+ "learning_rate": 0.0006,
+ "loss": 3.849355697631836,
+ "step": 1860
+ },
+ {
+ "epoch": 25.977312390924958,
+ "grad_norm": 0.22968950867652893,
+ "learning_rate": 0.0006,
+ "loss": 3.8472652435302734,
+ "step": 1870
+ },
+ {
+ "epoch": 26.0,
+ "eval_loss": 3.872666358947754,
+ "eval_runtime": 3.2447,
+ "eval_samples_per_second": 752.296,
+ "eval_steps_per_second": 12.019,
+ "step": 1872
+ },
+ {
+ "epoch": 26.111692844677137,
+ "grad_norm": 0.3149333894252777,
+ "learning_rate": 0.0006,
+ "loss": 3.7838531494140626,
+ "step": 1880
+ },
+ {
+ "epoch": 26.25130890052356,
+ "grad_norm": 0.39827290177345276,
+ "learning_rate": 0.0006,
+ "loss": 3.8100791931152345,
+ "step": 1890
+ },
+ {
+ "epoch": 26.390924956369982,
+ "grad_norm": 0.24689096212387085,
+ "learning_rate": 0.0006,
+ "loss": 3.8122238159179687,
+ "step": 1900
+ },
+ {
+ "epoch": 26.530541012216403,
+ "grad_norm": 0.24206413328647614,
+ "learning_rate": 0.0006,
+ "loss": 3.8097824096679687,
+ "step": 1910
+ },
+ {
+ "epoch": 26.670157068062828,
+ "grad_norm": 0.22627055644989014,
+ "learning_rate": 0.0006,
+ "loss": 3.8124542236328125,
+ "step": 1920
+ },
+ {
+ "epoch": 26.80977312390925,
+ "grad_norm": 0.23836272954940796,
+ "learning_rate": 0.0006,
+ "loss": 3.8373123168945313,
+ "step": 1930
+ },
+ {
+ "epoch": 26.949389179755673,
+ "grad_norm": 0.23511981964111328,
+ "learning_rate": 0.0006,
+ "loss": 3.834056091308594,
+ "step": 1940
+ },
+ {
+ "epoch": 27.0,
+ "eval_loss": 3.8710741996765137,
+ "eval_runtime": 3.2477,
+ "eval_samples_per_second": 751.617,
+ "eval_steps_per_second": 12.009,
+ "step": 1944
+ },
+ {
+ "epoch": 27.083769633507853,
+ "grad_norm": 0.2708267867565155,
+ "learning_rate": 0.0006,
+ "loss": 3.7948688507080077,
+ "step": 1950
+ },
+ {
+ "epoch": 27.223385689354277,
+ "grad_norm": 0.2775874733924866,
+ "learning_rate": 0.0006,
+ "loss": 3.7512874603271484,
+ "step": 1960
+ },
+ {
+ "epoch": 27.363001745200698,
+ "grad_norm": 0.277740478515625,
+ "learning_rate": 0.0006,
+ "loss": 3.7783355712890625,
+ "step": 1970
+ },
+ {
+ "epoch": 27.50261780104712,
+ "grad_norm": 0.26964810490608215,
+ "learning_rate": 0.0006,
+ "loss": 3.778544616699219,
+ "step": 1980
+ },
+ {
+ "epoch": 27.642233856893544,
+ "grad_norm": 0.25105977058410645,
+ "learning_rate": 0.0006,
+ "loss": 3.7948646545410156,
+ "step": 1990
+ },
+ {
+ "epoch": 27.781849912739965,
+ "grad_norm": 0.24751153588294983,
+ "learning_rate": 0.0006,
+ "loss": 3.809347915649414,
+ "step": 2000
+ },
+ {
+ "epoch": 27.921465968586386,
+ "grad_norm": 0.2541753649711609,
+ "learning_rate": 0.0006,
+ "loss": 3.82115478515625,
+ "step": 2010
+ },
+ {
+ "epoch": 28.0,
+ "eval_loss": 3.870692014694214,
+ "eval_runtime": 3.2579,
+ "eval_samples_per_second": 749.253,
+ "eval_steps_per_second": 11.971,
+ "step": 2016
+ },
+ {
+ "epoch": 28.05584642233857,
+ "grad_norm": 0.2538928985595703,
+ "learning_rate": 0.0006,
+ "loss": 3.795104217529297,
+ "step": 2020
+ },
+ {
+ "epoch": 28.195462478184993,
+ "grad_norm": 0.29538166522979736,
+ "learning_rate": 0.0006,
+ "loss": 3.7417327880859377,
+ "step": 2030
+ },
+ {
+ "epoch": 28.335078534031414,
+ "grad_norm": 0.2790539562702179,
+ "learning_rate": 0.0006,
+ "loss": 3.752838897705078,
+ "step": 2040
+ },
+ {
+ "epoch": 28.474694589877835,
+ "grad_norm": 0.31643059849739075,
+ "learning_rate": 0.0006,
+ "loss": 3.7693412780761717,
+ "step": 2050
+ },
+ {
+ "epoch": 28.61431064572426,
+ "grad_norm": 0.23396413028240204,
+ "learning_rate": 0.0006,
+ "loss": 3.7859622955322267,
+ "step": 2060
+ },
+ {
+ "epoch": 28.75392670157068,
+ "grad_norm": 0.24350608885288239,
+ "learning_rate": 0.0006,
+ "loss": 3.773362731933594,
+ "step": 2070
+ },
+ {
+ "epoch": 28.8935427574171,
+ "grad_norm": 0.2211272269487381,
+ "learning_rate": 0.0006,
+ "loss": 3.7897659301757813,
+ "step": 2080
+ },
+ {
+ "epoch": 29.0,
+ "eval_loss": 3.8660898208618164,
+ "eval_runtime": 3.2434,
+ "eval_samples_per_second": 752.604,
+ "eval_steps_per_second": 12.024,
+ "step": 2088
+ },
+ {
+ "epoch": 29.027923211169284,
+ "grad_norm": 0.27128738164901733,
+ "learning_rate": 0.0006,
+ "loss": 3.7713768005371096,
+ "step": 2090
+ },
+ {
+ "epoch": 29.167539267015705,
+ "grad_norm": 0.27414658665657043,
+ "learning_rate": 0.0006,
+ "loss": 3.7139110565185547,
+ "step": 2100
+ },
+ {
+ "epoch": 29.30715532286213,
+ "grad_norm": 0.27187302708625793,
+ "learning_rate": 0.0006,
+ "loss": 3.7351806640625,
+ "step": 2110
+ },
+ {
+ "epoch": 29.44677137870855,
+ "grad_norm": 0.258182168006897,
+ "learning_rate": 0.0006,
+ "loss": 3.7201263427734377,
+ "step": 2120
+ },
+ {
+ "epoch": 29.586387434554975,
+ "grad_norm": 0.26228925585746765,
+ "learning_rate": 0.0006,
+ "loss": 3.7570404052734374,
+ "step": 2130
+ },
+ {
+ "epoch": 29.726003490401396,
+ "grad_norm": 0.2470857799053192,
+ "learning_rate": 0.0006,
+ "loss": 3.7635124206542967,
+ "step": 2140
+ },
+ {
+ "epoch": 29.865619546247817,
+ "grad_norm": 0.23536743223667145,
+ "learning_rate": 0.0006,
+ "loss": 3.767218017578125,
+ "step": 2150
+ },
+ {
+ "epoch": 30.0,
+ "grad_norm": 0.25428009033203125,
+ "learning_rate": 0.0006,
+ "loss": 3.7901588439941407,
+ "step": 2160
+ },
+ {
+ "epoch": 30.0,
+ "eval_loss": 3.8617773056030273,
+ "eval_runtime": 3.2549,
+ "eval_samples_per_second": 749.935,
+ "eval_steps_per_second": 11.982,
+ "step": 2160
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 7200,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 100,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 9.216882763251057e+17,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/t1a/t1a_S2_C1_rand15/train.log b/runs/t1a/t1a_S2_C1_rand15/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..0e301fc306787e8b1721ed3c1ce0a6f6082885d8
--- /dev/null
+++ b/runs/t1a/t1a_S2_C1_rand15/train.log
@@ -0,0 +1,3035 @@
+W0818 05:02:17.472000 2549233 torch/distributed/run.py:851]
+W0818 05:02:17.472000 2549233 torch/distributed/run.py:851] *****************************************
+W0818 05:02:17.472000 2549233 torch/distributed/run.py:851] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
+W0818 05:02:17.472000 2549233 torch/distributed/run.py:851] *****************************************
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 145833984
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 145833984
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 145833984
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 145833984
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 100
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 8
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 8
+[transformers] Total optimization steps = 7,200
+[transformers] Number of trainable parameters = 145,833,984
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260818_050237-f0hal6fm
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t1a_S2_C1_rand15
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/f0hal6fm
+
0%| | 0/7200 [00:00 will be ignored
+[rank6]:W0818 05:03:00.536000 2549346 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank1]:W0818 05:03:00.537000 2549341 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank4]:W0818 05:03:00.549000 2549344 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank7]:W0818 05:03:00.567000 2549347 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank5]:W0818 05:03:00.791000 2549345 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank2]:W0818 05:03:00.872000 2549342 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[rank0]:W0818 05:03:01.762000 2549340 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+
0%| | 1/7200 [00:44<89:32:24, 44.78s/it]
{'loss': '12.02', 'grad_norm': '1.439', 'learning_rate': '0', 'epoch': '0.01396'}
+
0%| | 1/7200 [00:44<89:32:24, 44.78s/it]
0%| | 2/7200 [00:45<37:39:24, 18.83s/it]
0%| | 3/7200 [00:46<21:03:52, 10.54s/it]
0%| | 4/7200 [00:46<13:15:57, 6.64s/it]
0%| | 5/7200 [00:47<8:57:38, 4.48s/it]
0%| | 6/7200 [00:48<6:21:39, 3.18s/it]
0%| | 7/7200 [00:48<4:42:39, 2.36s/it]
0%| | 8/7200 [00:49<3:37:56, 1.82s/it]
0%| | 9/7200 [00:50<2:54:23, 1.46s/it]
0%| | 10/7200 [00:50<2:24:48, 1.21s/it]
{'loss': '11.76', 'grad_norm': '1.751', 'learning_rate': '5.4e-05', 'epoch': '0.1396'}
+
0%| | 10/7200 [00:50<2:24:48, 1.21s/it]
0%| | 11/7200 [00:52<2:27:03, 1.23s/it]
0%| | 12/7200 [00:52<2:12:34, 1.11s/it]
0%| | 13/7200 [00:53<1:56:12, 1.03it/s]
0%| | 14/7200 [00:54<1:44:52, 1.14it/s]
0%| | 15/7200 [00:54<1:37:18, 1.23it/s]
0%| | 16/7200 [00:55<1:31:39, 1.31it/s]
0%| | 17/7200 [00:56<1:27:43, 1.36it/s]
0%| | 18/7200 [00:56<1:25:13, 1.40it/s]
0%| | 19/7200 [00:57<1:23:13, 1.44it/s]
0%| | 20/7200 [00:58<1:21:50, 1.46it/s]
{'loss': '11.07', 'grad_norm': '1.697', 'learning_rate': '0.000114', 'epoch': '0.2792'}
+
0%| | 20/7200 [00:58<1:21:50, 1.46it/s]
0%| | 21/7200 [00:58<1:21:16, 1.47it/s]
0%| | 22/7200 [00:59<1:20:25, 1.49it/s]
0%| | 23/7200 [01:00<1:19:49, 1.50it/s]
0%| | 24/7200 [01:00<1:19:44, 1.50it/s]
0%| | 25/7200 [01:01<1:19:14, 1.51it/s]
0%| | 26/7200 [01:02<1:19:18, 1.51it/s]
0%| | 27/7200 [01:02<1:18:56, 1.51it/s]
0%| | 28/7200 [01:03<1:18:43, 1.52it/s]
0%| | 29/7200 [01:04<1:18:55, 1.51it/s]
0%| | 30/7200 [01:04<1:18:41, 1.52it/s]
{'loss': '10.31', 'grad_norm': '1.644', 'learning_rate': '0.000174', 'epoch': '0.4188'}
+
0%| | 30/7200 [01:04<1:18:41, 1.52it/s]
0%| | 31/7200 [01:05<1:18:35, 1.52it/s]
0%| | 32/7200 [01:06<1:18:51, 1.51it/s]
0%| | 33/7200 [01:06<1:18:38, 1.52it/s]
0%| | 34/7200 [01:07<1:18:31, 1.52it/s]
0%| | 35/7200 [01:07<1:18:45, 1.52it/s]
0%| | 36/7200 [01:08<1:18:34, 1.52it/s]
1%| | 37/7200 [01:09<1:18:31, 1.52it/s]
1%| | 38/7200 [01:09<1:18:32, 1.52it/s]
1%| | 39/7200 [01:10<1:18:23, 1.52it/s]
1%| | 40/7200 [01:11<1:18:39, 1.52it/s]
{'loss': '9.274', 'grad_norm': '1.485', 'learning_rate': '0.000234', 'epoch': '0.5585'}
+
1%| | 40/7200 [01:11<1:18:39, 1.52it/s]
1%| | 41/7200 [01:11<1:18:33, 1.52it/s]
1%| | 42/7200 [01:12<1:18:25, 1.52it/s]
1%| | 43/7200 [01:13<1:18:43, 1.52it/s]
1%| | 44/7200 [01:13<1:18:31, 1.52it/s]
1%| | 45/7200 [01:14<1:18:17, 1.52it/s]
1%| | 46/7200 [01:15<1:18:36, 1.52it/s]
1%| | 47/7200 [01:15<1:18:28, 1.52it/s]
1%| | 48/7200 [01:16<1:18:22, 1.52it/s]
1%| | 49/7200 [01:17<1:18:37, 1.52it/s]
1%| | 50/7200 [01:17<1:18:27, 1.52it/s]
{'loss': '8.278', 'grad_norm': '1.025', 'learning_rate': '0.000294', 'epoch': '0.6981'}
+
1%| | 50/7200 [01:17<1:18:27, 1.52it/s]
1%| | 51/7200 [01:18<1:18:20, 1.52it/s]
1%| | 52/7200 [01:19<1:18:38, 1.51it/s]
1%| | 53/7200 [01:19<1:18:27, 1.52it/s]
1%| | 54/7200 [01:20<1:18:15, 1.52it/s]
1%| | 55/7200 [01:21<1:18:36, 1.52it/s]
1%| | 56/7200 [01:21<1:18:26, 1.52it/s]
1%| | 57/7200 [01:22<1:18:44, 1.51it/s]
1%| | 58/7200 [01:23<1:18:34, 1.51it/s]
1%| | 59/7200 [01:23<1:18:27, 1.52it/s]
1%| | 60/7200 [01:24<1:18:43, 1.51it/s]
{'loss': '7.726', 'grad_norm': '0.2833', 'learning_rate': '0.000354', 'epoch': '0.8377'}
+
1%| | 60/7200 [01:24<1:18:43, 1.51it/s]
1%| | 61/7200 [01:25<1:18:30, 1.52it/s]
1%| | 62/7200 [01:25<1:18:23, 1.52it/s]
1%| | 63/7200 [01:26<1:18:33, 1.51it/s]
1%| | 64/7200 [01:27<1:18:27, 1.52it/s]
1%| | 65/7200 [01:27<1:18:20, 1.52it/s]
1%| | 66/7200 [01:28<1:18:37, 1.51it/s]
1%| | 67/7200 [01:29<1:18:25, 1.52it/s]
1%| | 68/7200 [01:29<1:18:36, 1.51it/s]
1%| | 69/7200 [01:30<1:18:28, 1.51it/s]
1%| | 70/7200 [01:31<1:18:24, 1.52it/s]
{'loss': '7.554', 'grad_norm': '1.467', 'learning_rate': '0.000414', 'epoch': '0.9773'}
+
1%| | 70/7200 [01:31<1:18:24, 1.52it/s]
1%| | 71/7200 [01:31<1:18:41, 1.51it/s]
1%| | 72/7200 [01:32<1:09:41, 1.70it/s][transformers]
+***** Running Evaluation *****
+[transformers] Num examples = 2441
+[transformers] Batch size = 8
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+
+
0%| | 0/39 [00:00system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/t1a/t1a_S2_C2_r2l50_i5exp/checkpoint-3240/config.json b/runs/t1a/t1a_S2_C2_r2l50_i5exp/checkpoint-3240/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3474619f10931027097cc05d837296fa4f33a6c
--- /dev/null
+++ b/runs/t1a/t1a_S2_C2_r2l50_i5exp/checkpoint-3240/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 512,
+ "initializer_range": 0.02,
+ "intermediate_size": 1536,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 4,
+ "num_hidden_layers": 20,
+ "num_key_value_heads": 4,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151676
+}
diff --git a/runs/t1a/t1a_S2_C2_r2l50_i5exp/checkpoint-3240/generation_config.json b/runs/t1a/t1a_S2_C2_r2l50_i5exp/checkpoint-3240/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/t1a/t1a_S2_C2_r2l50_i5exp/checkpoint-3240/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/t1a/t1a_S2_C2_r2l50_i5exp/checkpoint-3240/tokenizer_config.json b/runs/t1a/t1a_S2_C2_r2l50_i5exp/checkpoint-3240/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d43c97a447f4b25a8c36be5873f7b622935fa90
--- /dev/null
+++ b/runs/t1a/t1a_S2_C2_r2l50_i5exp/checkpoint-3240/tokenizer_config.json
@@ -0,0 +1,24 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>",
+ "<|next_1_pred|>",
+ "<|next_2_pred|>",
+ "<|next_3_pred|>",
+ "<|next_4_pred|>",
+ "<|next_5_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/t1a/t1a_S2_C2_r2l50_i5exp/checkpoint-3240/trainer_state.json b/runs/t1a/t1a_S2_C2_r2l50_i5exp/checkpoint-3240/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c643d7487a2d4dfd4181788e4a3788ddcf2f101
--- /dev/null
+++ b/runs/t1a/t1a_S2_C2_r2l50_i5exp/checkpoint-3240/trainer_state.json
@@ -0,0 +1,2669 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 45.0,
+ "eval_steps": 72,
+ "global_step": 3240,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013961605584642234,
+ "grad_norm": 2.748727560043335,
+ "learning_rate": 0.0,
+ "loss": 12.011189460754395,
+ "step": 1
+ },
+ {
+ "epoch": 0.13961605584642234,
+ "grad_norm": 1.7549406290054321,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.671863132052952,
+ "step": 10
+ },
+ {
+ "epoch": 0.2792321116928447,
+ "grad_norm": 1.6919854879379272,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 10.999891662597657,
+ "step": 20
+ },
+ {
+ "epoch": 0.418848167539267,
+ "grad_norm": 1.6407349109649658,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 10.233381652832032,
+ "step": 30
+ },
+ {
+ "epoch": 0.5584642233856894,
+ "grad_norm": 1.4805939197540283,
+ "learning_rate": 0.000234,
+ "loss": 9.196365356445312,
+ "step": 40
+ },
+ {
+ "epoch": 0.6980802792321117,
+ "grad_norm": 0.8693683743476868,
+ "learning_rate": 0.000294,
+ "loss": 8.213917541503907,
+ "step": 50
+ },
+ {
+ "epoch": 0.837696335078534,
+ "grad_norm": 0.26588985323905945,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 7.701846313476563,
+ "step": 60
+ },
+ {
+ "epoch": 0.9773123909249564,
+ "grad_norm": 0.24134407937526703,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 7.559868621826172,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.481448173522949,
+ "eval_runtime": 27.024,
+ "eval_samples_per_second": 90.327,
+ "eval_steps_per_second": 1.443,
+ "step": 72
+ },
+ {
+ "epoch": 1.111692844677138,
+ "grad_norm": 0.6404353380203247,
+ "learning_rate": 0.000474,
+ "loss": 7.488853454589844,
+ "step": 80
+ },
+ {
+ "epoch": 1.2513089005235603,
+ "grad_norm": 0.18342405557632446,
+ "learning_rate": 0.000534,
+ "loss": 7.422576141357422,
+ "step": 90
+ },
+ {
+ "epoch": 1.3909249563699826,
+ "grad_norm": 0.2256995141506195,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 7.387411499023438,
+ "step": 100
+ },
+ {
+ "epoch": 1.530541012216405,
+ "grad_norm": 0.19165629148483276,
+ "learning_rate": 0.0006,
+ "loss": 7.364878845214844,
+ "step": 110
+ },
+ {
+ "epoch": 1.6701570680628273,
+ "grad_norm": 0.32182881236076355,
+ "learning_rate": 0.0006,
+ "loss": 7.340007019042969,
+ "step": 120
+ },
+ {
+ "epoch": 1.8097731239092496,
+ "grad_norm": 0.18933962285518646,
+ "learning_rate": 0.0006,
+ "loss": 7.309558868408203,
+ "step": 130
+ },
+ {
+ "epoch": 1.949389179755672,
+ "grad_norm": 0.34846189618110657,
+ "learning_rate": 0.0006,
+ "loss": 7.322872924804687,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 7.1557416915893555,
+ "eval_runtime": 2.6233,
+ "eval_samples_per_second": 930.507,
+ "eval_steps_per_second": 14.867,
+ "step": 144
+ },
+ {
+ "epoch": 2.0837696335078535,
+ "grad_norm": 0.2591068744659424,
+ "learning_rate": 0.0006,
+ "loss": 7.26339111328125,
+ "step": 150
+ },
+ {
+ "epoch": 2.223385689354276,
+ "grad_norm": 0.3847464323043823,
+ "learning_rate": 0.0006,
+ "loss": 7.193439483642578,
+ "step": 160
+ },
+ {
+ "epoch": 2.363001745200698,
+ "grad_norm": 0.44022712111473083,
+ "learning_rate": 0.0006,
+ "loss": 7.166969299316406,
+ "step": 170
+ },
+ {
+ "epoch": 2.5026178010471205,
+ "grad_norm": 0.2414730340242386,
+ "learning_rate": 0.0006,
+ "loss": 7.098081970214844,
+ "step": 180
+ },
+ {
+ "epoch": 2.642233856893543,
+ "grad_norm": 0.3886379599571228,
+ "learning_rate": 0.0006,
+ "loss": 7.048555755615235,
+ "step": 190
+ },
+ {
+ "epoch": 2.781849912739965,
+ "grad_norm": 0.3026723861694336,
+ "learning_rate": 0.0006,
+ "loss": 6.947491455078125,
+ "step": 200
+ },
+ {
+ "epoch": 2.9214659685863875,
+ "grad_norm": 0.4259362518787384,
+ "learning_rate": 0.0006,
+ "loss": 6.872621154785156,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 6.512463092803955,
+ "eval_runtime": 2.6376,
+ "eval_samples_per_second": 925.454,
+ "eval_steps_per_second": 14.786,
+ "step": 216
+ },
+ {
+ "epoch": 3.055846422338569,
+ "grad_norm": 0.304609477519989,
+ "learning_rate": 0.0006,
+ "loss": 6.766963195800781,
+ "step": 220
+ },
+ {
+ "epoch": 3.195462478184991,
+ "grad_norm": 0.4567917287349701,
+ "learning_rate": 0.0006,
+ "loss": 6.694587707519531,
+ "step": 230
+ },
+ {
+ "epoch": 3.3350785340314135,
+ "grad_norm": 0.4419524073600769,
+ "learning_rate": 0.0006,
+ "loss": 6.669084167480468,
+ "step": 240
+ },
+ {
+ "epoch": 3.474694589877836,
+ "grad_norm": 0.23295481503009796,
+ "learning_rate": 0.0006,
+ "loss": 6.600353240966797,
+ "step": 250
+ },
+ {
+ "epoch": 3.614310645724258,
+ "grad_norm": 0.2101762592792511,
+ "learning_rate": 0.0006,
+ "loss": 6.495196533203125,
+ "step": 260
+ },
+ {
+ "epoch": 3.7539267015706805,
+ "grad_norm": 0.3698682188987732,
+ "learning_rate": 0.0006,
+ "loss": 6.441866302490235,
+ "step": 270
+ },
+ {
+ "epoch": 3.893542757417103,
+ "grad_norm": 0.25112441182136536,
+ "learning_rate": 0.0006,
+ "loss": 6.361153411865234,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 5.97499418258667,
+ "eval_runtime": 2.6138,
+ "eval_samples_per_second": 933.875,
+ "eval_steps_per_second": 14.921,
+ "step": 288
+ },
+ {
+ "epoch": 4.027923211169284,
+ "grad_norm": 0.5446335077285767,
+ "learning_rate": 0.0006,
+ "loss": 6.318148803710938,
+ "step": 290
+ },
+ {
+ "epoch": 4.167539267015707,
+ "grad_norm": 0.293059378862381,
+ "learning_rate": 0.0006,
+ "loss": 6.239653396606445,
+ "step": 300
+ },
+ {
+ "epoch": 4.307155322862129,
+ "grad_norm": 0.3282206952571869,
+ "learning_rate": 0.0006,
+ "loss": 6.208422088623047,
+ "step": 310
+ },
+ {
+ "epoch": 4.446771378708552,
+ "grad_norm": 0.5108447670936584,
+ "learning_rate": 0.0006,
+ "loss": 6.176555633544922,
+ "step": 320
+ },
+ {
+ "epoch": 4.5863874345549736,
+ "grad_norm": 0.4455273151397705,
+ "learning_rate": 0.0006,
+ "loss": 6.164451980590821,
+ "step": 330
+ },
+ {
+ "epoch": 4.726003490401396,
+ "grad_norm": 0.2580024302005768,
+ "learning_rate": 0.0006,
+ "loss": 6.107962417602539,
+ "step": 340
+ },
+ {
+ "epoch": 4.865619546247818,
+ "grad_norm": 0.30771341919898987,
+ "learning_rate": 0.0006,
+ "loss": 6.0555564880371096,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.22924596071243286,
+ "learning_rate": 0.0006,
+ "loss": 5.9695274353027346,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 5.57134485244751,
+ "eval_runtime": 2.6209,
+ "eval_samples_per_second": 931.356,
+ "eval_steps_per_second": 14.88,
+ "step": 360
+ },
+ {
+ "epoch": 5.139616055846423,
+ "grad_norm": 0.45722731947898865,
+ "learning_rate": 0.0006,
+ "loss": 5.922108459472656,
+ "step": 370
+ },
+ {
+ "epoch": 5.279232111692845,
+ "grad_norm": 0.31171005964279175,
+ "learning_rate": 0.0006,
+ "loss": 5.903049468994141,
+ "step": 380
+ },
+ {
+ "epoch": 5.418848167539267,
+ "grad_norm": 0.3982207775115967,
+ "learning_rate": 0.0006,
+ "loss": 5.857491302490234,
+ "step": 390
+ },
+ {
+ "epoch": 5.558464223385689,
+ "grad_norm": 0.3677067458629608,
+ "learning_rate": 0.0006,
+ "loss": 5.716806030273437,
+ "step": 400
+ },
+ {
+ "epoch": 5.698080279232112,
+ "grad_norm": 0.4849364459514618,
+ "learning_rate": 0.0006,
+ "loss": 5.727468872070313,
+ "step": 410
+ },
+ {
+ "epoch": 5.837696335078534,
+ "grad_norm": 0.39584583044052124,
+ "learning_rate": 0.0006,
+ "loss": 5.76154899597168,
+ "step": 420
+ },
+ {
+ "epoch": 5.977312390924956,
+ "grad_norm": 0.2571723759174347,
+ "learning_rate": 0.0006,
+ "loss": 5.697042083740234,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 5.238385200500488,
+ "eval_runtime": 2.6457,
+ "eval_samples_per_second": 922.628,
+ "eval_steps_per_second": 14.741,
+ "step": 432
+ },
+ {
+ "epoch": 6.111692844677138,
+ "grad_norm": 0.2874707877635956,
+ "learning_rate": 0.0006,
+ "loss": 5.664549255371094,
+ "step": 440
+ },
+ {
+ "epoch": 6.2513089005235605,
+ "grad_norm": 0.4763229787349701,
+ "learning_rate": 0.0006,
+ "loss": 5.60162353515625,
+ "step": 450
+ },
+ {
+ "epoch": 6.390924956369982,
+ "grad_norm": 0.3173147439956665,
+ "learning_rate": 0.0006,
+ "loss": 5.5615699768066404,
+ "step": 460
+ },
+ {
+ "epoch": 6.530541012216405,
+ "grad_norm": 0.604937732219696,
+ "learning_rate": 0.0006,
+ "loss": 5.561077880859375,
+ "step": 470
+ },
+ {
+ "epoch": 6.670157068062827,
+ "grad_norm": 0.33040156960487366,
+ "learning_rate": 0.0006,
+ "loss": 5.4935859680175785,
+ "step": 480
+ },
+ {
+ "epoch": 6.80977312390925,
+ "grad_norm": 0.2717789113521576,
+ "learning_rate": 0.0006,
+ "loss": 5.449597930908203,
+ "step": 490
+ },
+ {
+ "epoch": 6.949389179755672,
+ "grad_norm": 0.4181554317474365,
+ "learning_rate": 0.0006,
+ "loss": 5.474884033203125,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 4.924429893493652,
+ "eval_runtime": 2.6372,
+ "eval_samples_per_second": 925.604,
+ "eval_steps_per_second": 14.788,
+ "step": 504
+ },
+ {
+ "epoch": 7.0837696335078535,
+ "grad_norm": 0.39575958251953125,
+ "learning_rate": 0.0006,
+ "loss": 5.366435241699219,
+ "step": 510
+ },
+ {
+ "epoch": 7.223385689354275,
+ "grad_norm": 0.8365405201911926,
+ "learning_rate": 0.0006,
+ "loss": 5.405091857910156,
+ "step": 520
+ },
+ {
+ "epoch": 7.363001745200698,
+ "grad_norm": 0.6012616157531738,
+ "learning_rate": 0.0006,
+ "loss": 5.420777893066406,
+ "step": 530
+ },
+ {
+ "epoch": 7.50261780104712,
+ "grad_norm": 0.24743209779262543,
+ "learning_rate": 0.0006,
+ "loss": 5.346966934204102,
+ "step": 540
+ },
+ {
+ "epoch": 7.642233856893543,
+ "grad_norm": 0.21129906177520752,
+ "learning_rate": 0.0006,
+ "loss": 5.248374938964844,
+ "step": 550
+ },
+ {
+ "epoch": 7.781849912739965,
+ "grad_norm": 0.20006980001926422,
+ "learning_rate": 0.0006,
+ "loss": 5.239125442504883,
+ "step": 560
+ },
+ {
+ "epoch": 7.9214659685863875,
+ "grad_norm": 0.2738592028617859,
+ "learning_rate": 0.0006,
+ "loss": 5.203498077392578,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.709870338439941,
+ "eval_runtime": 2.6354,
+ "eval_samples_per_second": 926.245,
+ "eval_steps_per_second": 14.799,
+ "step": 576
+ },
+ {
+ "epoch": 8.055846422338568,
+ "grad_norm": 0.3069969117641449,
+ "learning_rate": 0.0006,
+ "loss": 5.195360565185547,
+ "step": 580
+ },
+ {
+ "epoch": 8.195462478184991,
+ "grad_norm": 0.3827716112136841,
+ "learning_rate": 0.0006,
+ "loss": 5.0992279052734375,
+ "step": 590
+ },
+ {
+ "epoch": 8.335078534031414,
+ "grad_norm": 0.6046319603919983,
+ "learning_rate": 0.0006,
+ "loss": 5.209037780761719,
+ "step": 600
+ },
+ {
+ "epoch": 8.474694589877837,
+ "grad_norm": 0.29844221472740173,
+ "learning_rate": 0.0006,
+ "loss": 5.197072219848633,
+ "step": 610
+ },
+ {
+ "epoch": 8.614310645724258,
+ "grad_norm": 0.22425347566604614,
+ "learning_rate": 0.0006,
+ "loss": 5.089417266845703,
+ "step": 620
+ },
+ {
+ "epoch": 8.75392670157068,
+ "grad_norm": 0.2454708069562912,
+ "learning_rate": 0.0006,
+ "loss": 5.082542419433594,
+ "step": 630
+ },
+ {
+ "epoch": 8.893542757417103,
+ "grad_norm": 0.2508460581302643,
+ "learning_rate": 0.0006,
+ "loss": 5.107080459594727,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.55564022064209,
+ "eval_runtime": 2.6371,
+ "eval_samples_per_second": 925.65,
+ "eval_steps_per_second": 14.789,
+ "step": 648
+ },
+ {
+ "epoch": 9.027923211169284,
+ "grad_norm": 0.2406289130449295,
+ "learning_rate": 0.0006,
+ "loss": 5.086015701293945,
+ "step": 650
+ },
+ {
+ "epoch": 9.167539267015707,
+ "grad_norm": 0.3326975107192993,
+ "learning_rate": 0.0006,
+ "loss": 5.0320793151855465,
+ "step": 660
+ },
+ {
+ "epoch": 9.30715532286213,
+ "grad_norm": 0.3569769263267517,
+ "learning_rate": 0.0006,
+ "loss": 4.983912658691406,
+ "step": 670
+ },
+ {
+ "epoch": 9.44677137870855,
+ "grad_norm": 0.23501677811145782,
+ "learning_rate": 0.0006,
+ "loss": 5.008148193359375,
+ "step": 680
+ },
+ {
+ "epoch": 9.586387434554974,
+ "grad_norm": 0.28273847699165344,
+ "learning_rate": 0.0006,
+ "loss": 4.913330078125,
+ "step": 690
+ },
+ {
+ "epoch": 9.726003490401396,
+ "grad_norm": 0.23736517131328583,
+ "learning_rate": 0.0006,
+ "loss": 5.003079986572265,
+ "step": 700
+ },
+ {
+ "epoch": 9.86561954624782,
+ "grad_norm": 0.2456035017967224,
+ "learning_rate": 0.0006,
+ "loss": 4.860411834716797,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.5961561799049377,
+ "learning_rate": 0.0006,
+ "loss": 4.870589065551758,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.47821569442749,
+ "eval_runtime": 2.6233,
+ "eval_samples_per_second": 930.513,
+ "eval_steps_per_second": 14.867,
+ "step": 720
+ },
+ {
+ "epoch": 10.139616055846423,
+ "grad_norm": 0.32937246561050415,
+ "learning_rate": 0.0006,
+ "loss": 4.955348968505859,
+ "step": 730
+ },
+ {
+ "epoch": 10.279232111692846,
+ "grad_norm": 0.26360464096069336,
+ "learning_rate": 0.0006,
+ "loss": 4.893888092041015,
+ "step": 740
+ },
+ {
+ "epoch": 10.418848167539267,
+ "grad_norm": 0.25288084149360657,
+ "learning_rate": 0.0006,
+ "loss": 4.831504821777344,
+ "step": 750
+ },
+ {
+ "epoch": 10.55846422338569,
+ "grad_norm": 0.21669960021972656,
+ "learning_rate": 0.0006,
+ "loss": 4.8575889587402346,
+ "step": 760
+ },
+ {
+ "epoch": 10.698080279232112,
+ "grad_norm": 0.1999340057373047,
+ "learning_rate": 0.0006,
+ "loss": 4.827686309814453,
+ "step": 770
+ },
+ {
+ "epoch": 10.837696335078533,
+ "grad_norm": 0.24877016246318817,
+ "learning_rate": 0.0006,
+ "loss": 4.827722549438477,
+ "step": 780
+ },
+ {
+ "epoch": 10.977312390924956,
+ "grad_norm": 0.24957390129566193,
+ "learning_rate": 0.0006,
+ "loss": 4.793301773071289,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.353127479553223,
+ "eval_runtime": 2.6215,
+ "eval_samples_per_second": 931.152,
+ "eval_steps_per_second": 14.877,
+ "step": 792
+ },
+ {
+ "epoch": 11.111692844677139,
+ "grad_norm": 0.4687240719795227,
+ "learning_rate": 0.0006,
+ "loss": 4.809085083007813,
+ "step": 800
+ },
+ {
+ "epoch": 11.25130890052356,
+ "grad_norm": 0.26527899503707886,
+ "learning_rate": 0.0006,
+ "loss": 4.849463653564453,
+ "step": 810
+ },
+ {
+ "epoch": 11.390924956369982,
+ "grad_norm": 0.25706419348716736,
+ "learning_rate": 0.0006,
+ "loss": 4.831657409667969,
+ "step": 820
+ },
+ {
+ "epoch": 11.530541012216405,
+ "grad_norm": 0.24298502504825592,
+ "learning_rate": 0.0006,
+ "loss": 4.803998565673828,
+ "step": 830
+ },
+ {
+ "epoch": 11.670157068062828,
+ "grad_norm": 0.2836157977581024,
+ "learning_rate": 0.0006,
+ "loss": 4.7751914978027346,
+ "step": 840
+ },
+ {
+ "epoch": 11.809773123909249,
+ "grad_norm": 0.28408530354499817,
+ "learning_rate": 0.0006,
+ "loss": 4.707385635375976,
+ "step": 850
+ },
+ {
+ "epoch": 11.949389179755672,
+ "grad_norm": 0.2384726107120514,
+ "learning_rate": 0.0006,
+ "loss": 4.747846221923828,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.282207012176514,
+ "eval_runtime": 2.6214,
+ "eval_samples_per_second": 931.186,
+ "eval_steps_per_second": 14.878,
+ "step": 864
+ },
+ {
+ "epoch": 12.083769633507853,
+ "grad_norm": 0.265219122171402,
+ "learning_rate": 0.0006,
+ "loss": 4.647481155395508,
+ "step": 870
+ },
+ {
+ "epoch": 12.223385689354275,
+ "grad_norm": 0.34703144431114197,
+ "learning_rate": 0.0006,
+ "loss": 4.6871086120605465,
+ "step": 880
+ },
+ {
+ "epoch": 12.363001745200698,
+ "grad_norm": 0.2610820531845093,
+ "learning_rate": 0.0006,
+ "loss": 4.715609741210938,
+ "step": 890
+ },
+ {
+ "epoch": 12.502617801047121,
+ "grad_norm": 0.331238329410553,
+ "learning_rate": 0.0006,
+ "loss": 4.748788070678711,
+ "step": 900
+ },
+ {
+ "epoch": 12.642233856893542,
+ "grad_norm": 0.2705678939819336,
+ "learning_rate": 0.0006,
+ "loss": 4.686032485961914,
+ "step": 910
+ },
+ {
+ "epoch": 12.781849912739965,
+ "grad_norm": 0.24467197060585022,
+ "learning_rate": 0.0006,
+ "loss": 4.6793701171875,
+ "step": 920
+ },
+ {
+ "epoch": 12.921465968586388,
+ "grad_norm": 0.2173224240541458,
+ "learning_rate": 0.0006,
+ "loss": 4.706697082519531,
+ "step": 930
+ },
+ {
+ "epoch": 13.0,
+ "eval_loss": 4.2196831703186035,
+ "eval_runtime": 2.6315,
+ "eval_samples_per_second": 927.609,
+ "eval_steps_per_second": 14.82,
+ "step": 936
+ },
+ {
+ "epoch": 13.055846422338568,
+ "grad_norm": 0.2356448620557785,
+ "learning_rate": 0.0006,
+ "loss": 4.69750862121582,
+ "step": 940
+ },
+ {
+ "epoch": 13.195462478184991,
+ "grad_norm": 0.4130285680294037,
+ "learning_rate": 0.0006,
+ "loss": 4.676899719238281,
+ "step": 950
+ },
+ {
+ "epoch": 13.335078534031414,
+ "grad_norm": 0.3391016721725464,
+ "learning_rate": 0.0006,
+ "loss": 4.635129165649414,
+ "step": 960
+ },
+ {
+ "epoch": 13.474694589877837,
+ "grad_norm": 0.2830184996128082,
+ "learning_rate": 0.0006,
+ "loss": 4.648286819458008,
+ "step": 970
+ },
+ {
+ "epoch": 13.614310645724258,
+ "grad_norm": 0.22988024353981018,
+ "learning_rate": 0.0006,
+ "loss": 4.642067718505859,
+ "step": 980
+ },
+ {
+ "epoch": 13.75392670157068,
+ "grad_norm": 0.20977698266506195,
+ "learning_rate": 0.0006,
+ "loss": 4.654012298583984,
+ "step": 990
+ },
+ {
+ "epoch": 13.893542757417103,
+ "grad_norm": 0.49698737263679504,
+ "learning_rate": 0.0006,
+ "loss": 4.726877593994141,
+ "step": 1000
+ },
+ {
+ "epoch": 14.0,
+ "eval_loss": 4.204152584075928,
+ "eval_runtime": 2.6184,
+ "eval_samples_per_second": 932.237,
+ "eval_steps_per_second": 14.894,
+ "step": 1008
+ },
+ {
+ "epoch": 14.027923211169284,
+ "grad_norm": 0.34385788440704346,
+ "learning_rate": 0.0006,
+ "loss": 4.696585845947266,
+ "step": 1010
+ },
+ {
+ "epoch": 14.167539267015707,
+ "grad_norm": 0.22285610437393188,
+ "learning_rate": 0.0006,
+ "loss": 4.6084037780761715,
+ "step": 1020
+ },
+ {
+ "epoch": 14.30715532286213,
+ "grad_norm": 0.2015100121498108,
+ "learning_rate": 0.0006,
+ "loss": 4.68123779296875,
+ "step": 1030
+ },
+ {
+ "epoch": 14.44677137870855,
+ "grad_norm": 0.1840844601392746,
+ "learning_rate": 0.0006,
+ "loss": 4.627776336669922,
+ "step": 1040
+ },
+ {
+ "epoch": 14.586387434554974,
+ "grad_norm": 0.21436038613319397,
+ "learning_rate": 0.0006,
+ "loss": 4.551625823974609,
+ "step": 1050
+ },
+ {
+ "epoch": 14.726003490401396,
+ "grad_norm": 0.23446917533874512,
+ "learning_rate": 0.0006,
+ "loss": 4.599505615234375,
+ "step": 1060
+ },
+ {
+ "epoch": 14.86561954624782,
+ "grad_norm": 0.27302172780036926,
+ "learning_rate": 0.0006,
+ "loss": 4.538143920898437,
+ "step": 1070
+ },
+ {
+ "epoch": 15.0,
+ "grad_norm": 0.2408692091703415,
+ "learning_rate": 0.0006,
+ "loss": 4.64341926574707,
+ "step": 1080
+ },
+ {
+ "epoch": 15.0,
+ "eval_loss": 4.129874229431152,
+ "eval_runtime": 2.6175,
+ "eval_samples_per_second": 932.577,
+ "eval_steps_per_second": 14.9,
+ "step": 1080
+ },
+ {
+ "epoch": 15.139616055846423,
+ "grad_norm": 0.27539339661598206,
+ "learning_rate": 0.0006,
+ "loss": 4.5645294189453125,
+ "step": 1090
+ },
+ {
+ "epoch": 15.279232111692846,
+ "grad_norm": 0.29777997732162476,
+ "learning_rate": 0.0006,
+ "loss": 4.572846984863281,
+ "step": 1100
+ },
+ {
+ "epoch": 15.418848167539267,
+ "grad_norm": 0.26377418637275696,
+ "learning_rate": 0.0006,
+ "loss": 4.535099029541016,
+ "step": 1110
+ },
+ {
+ "epoch": 15.55846422338569,
+ "grad_norm": 0.255034476518631,
+ "learning_rate": 0.0006,
+ "loss": 4.533952713012695,
+ "step": 1120
+ },
+ {
+ "epoch": 15.698080279232112,
+ "grad_norm": 0.5040932893753052,
+ "learning_rate": 0.0006,
+ "loss": 4.532304763793945,
+ "step": 1130
+ },
+ {
+ "epoch": 15.837696335078533,
+ "grad_norm": 0.6600229740142822,
+ "learning_rate": 0.0006,
+ "loss": 4.572650146484375,
+ "step": 1140
+ },
+ {
+ "epoch": 15.977312390924956,
+ "grad_norm": 1.0680906772613525,
+ "learning_rate": 0.0006,
+ "loss": 4.7913860321044925,
+ "step": 1150
+ },
+ {
+ "epoch": 16.0,
+ "eval_loss": 4.414292335510254,
+ "eval_runtime": 2.6333,
+ "eval_samples_per_second": 926.986,
+ "eval_steps_per_second": 14.811,
+ "step": 1152
+ },
+ {
+ "epoch": 16.111692844677137,
+ "grad_norm": 1.0122393369674683,
+ "learning_rate": 0.0006,
+ "loss": 4.8935295104980465,
+ "step": 1160
+ },
+ {
+ "epoch": 16.25130890052356,
+ "grad_norm": 0.30783557891845703,
+ "learning_rate": 0.0006,
+ "loss": 4.906919479370117,
+ "step": 1170
+ },
+ {
+ "epoch": 16.390924956369982,
+ "grad_norm": 0.17488312721252441,
+ "learning_rate": 0.0006,
+ "loss": 4.732842254638672,
+ "step": 1180
+ },
+ {
+ "epoch": 16.530541012216403,
+ "grad_norm": 0.1468682587146759,
+ "learning_rate": 0.0006,
+ "loss": 4.658345794677734,
+ "step": 1190
+ },
+ {
+ "epoch": 16.670157068062828,
+ "grad_norm": 0.15733979642391205,
+ "learning_rate": 0.0006,
+ "loss": 4.5733894348144535,
+ "step": 1200
+ },
+ {
+ "epoch": 16.80977312390925,
+ "grad_norm": 0.1471433937549591,
+ "learning_rate": 0.0006,
+ "loss": 4.5941619873046875,
+ "step": 1210
+ },
+ {
+ "epoch": 16.949389179755673,
+ "grad_norm": 0.15746451914310455,
+ "learning_rate": 0.0006,
+ "loss": 4.525423431396485,
+ "step": 1220
+ },
+ {
+ "epoch": 17.0,
+ "eval_loss": 4.110367298126221,
+ "eval_runtime": 2.633,
+ "eval_samples_per_second": 927.079,
+ "eval_steps_per_second": 14.812,
+ "step": 1224
+ },
+ {
+ "epoch": 17.083769633507853,
+ "grad_norm": 0.16006214916706085,
+ "learning_rate": 0.0006,
+ "loss": 4.538452911376953,
+ "step": 1230
+ },
+ {
+ "epoch": 17.223385689354277,
+ "grad_norm": 0.1758093535900116,
+ "learning_rate": 0.0006,
+ "loss": 4.525255584716797,
+ "step": 1240
+ },
+ {
+ "epoch": 17.363001745200698,
+ "grad_norm": 0.16477851569652557,
+ "learning_rate": 0.0006,
+ "loss": 4.541443252563477,
+ "step": 1250
+ },
+ {
+ "epoch": 17.50261780104712,
+ "grad_norm": 0.16630513966083527,
+ "learning_rate": 0.0006,
+ "loss": 4.523538970947266,
+ "step": 1260
+ },
+ {
+ "epoch": 17.642233856893544,
+ "grad_norm": 0.16956695914268494,
+ "learning_rate": 0.0006,
+ "loss": 4.4966178894042965,
+ "step": 1270
+ },
+ {
+ "epoch": 17.781849912739965,
+ "grad_norm": 0.2044956535100937,
+ "learning_rate": 0.0006,
+ "loss": 4.484248352050781,
+ "step": 1280
+ },
+ {
+ "epoch": 17.921465968586386,
+ "grad_norm": 0.19074811041355133,
+ "learning_rate": 0.0006,
+ "loss": 4.528261947631836,
+ "step": 1290
+ },
+ {
+ "epoch": 18.0,
+ "eval_loss": 4.061490058898926,
+ "eval_runtime": 2.6332,
+ "eval_samples_per_second": 926.992,
+ "eval_steps_per_second": 14.811,
+ "step": 1296
+ },
+ {
+ "epoch": 18.05584642233857,
+ "grad_norm": 0.20949822664260864,
+ "learning_rate": 0.0006,
+ "loss": 4.521551513671875,
+ "step": 1300
+ },
+ {
+ "epoch": 18.195462478184993,
+ "grad_norm": 0.20056022703647614,
+ "learning_rate": 0.0006,
+ "loss": 4.500559997558594,
+ "step": 1310
+ },
+ {
+ "epoch": 18.335078534031414,
+ "grad_norm": 0.2039003223180771,
+ "learning_rate": 0.0006,
+ "loss": 4.421464920043945,
+ "step": 1320
+ },
+ {
+ "epoch": 18.474694589877835,
+ "grad_norm": 0.16461309790611267,
+ "learning_rate": 0.0006,
+ "loss": 4.488716506958008,
+ "step": 1330
+ },
+ {
+ "epoch": 18.61431064572426,
+ "grad_norm": 0.16185447573661804,
+ "learning_rate": 0.0006,
+ "loss": 4.480081939697266,
+ "step": 1340
+ },
+ {
+ "epoch": 18.75392670157068,
+ "grad_norm": 0.1718590259552002,
+ "learning_rate": 0.0006,
+ "loss": 4.4592033386230465,
+ "step": 1350
+ },
+ {
+ "epoch": 18.8935427574171,
+ "grad_norm": 0.1993137151002884,
+ "learning_rate": 0.0006,
+ "loss": 4.3988502502441404,
+ "step": 1360
+ },
+ {
+ "epoch": 19.0,
+ "eval_loss": 4.02978515625,
+ "eval_runtime": 2.634,
+ "eval_samples_per_second": 926.745,
+ "eval_steps_per_second": 14.807,
+ "step": 1368
+ },
+ {
+ "epoch": 19.027923211169284,
+ "grad_norm": 0.20860058069229126,
+ "learning_rate": 0.0006,
+ "loss": 4.481473541259765,
+ "step": 1370
+ },
+ {
+ "epoch": 19.167539267015705,
+ "grad_norm": 0.23360496759414673,
+ "learning_rate": 0.0006,
+ "loss": 4.434365081787109,
+ "step": 1380
+ },
+ {
+ "epoch": 19.30715532286213,
+ "grad_norm": 0.19588778913021088,
+ "learning_rate": 0.0006,
+ "loss": 4.38256607055664,
+ "step": 1390
+ },
+ {
+ "epoch": 19.44677137870855,
+ "grad_norm": 0.21648041903972626,
+ "learning_rate": 0.0006,
+ "loss": 4.476564025878906,
+ "step": 1400
+ },
+ {
+ "epoch": 19.586387434554975,
+ "grad_norm": 0.21911892294883728,
+ "learning_rate": 0.0006,
+ "loss": 4.45465316772461,
+ "step": 1410
+ },
+ {
+ "epoch": 19.726003490401396,
+ "grad_norm": 0.18623879551887512,
+ "learning_rate": 0.0006,
+ "loss": 4.417459106445312,
+ "step": 1420
+ },
+ {
+ "epoch": 19.865619546247817,
+ "grad_norm": 0.2935650944709778,
+ "learning_rate": 0.0006,
+ "loss": 4.394468688964844,
+ "step": 1430
+ },
+ {
+ "epoch": 20.0,
+ "grad_norm": 0.21868087351322174,
+ "learning_rate": 0.0006,
+ "loss": 4.487446594238281,
+ "step": 1440
+ },
+ {
+ "epoch": 20.0,
+ "eval_loss": 4.0057854652404785,
+ "eval_runtime": 2.6236,
+ "eval_samples_per_second": 930.408,
+ "eval_steps_per_second": 14.865,
+ "step": 1440
+ },
+ {
+ "epoch": 20.13961605584642,
+ "grad_norm": 0.17526422441005707,
+ "learning_rate": 0.0006,
+ "loss": 4.39727783203125,
+ "step": 1450
+ },
+ {
+ "epoch": 20.279232111692846,
+ "grad_norm": 0.20493270456790924,
+ "learning_rate": 0.0006,
+ "loss": 4.411034393310547,
+ "step": 1460
+ },
+ {
+ "epoch": 20.418848167539267,
+ "grad_norm": 0.19484785199165344,
+ "learning_rate": 0.0006,
+ "loss": 4.385637664794922,
+ "step": 1470
+ },
+ {
+ "epoch": 20.55846422338569,
+ "grad_norm": 0.255136638879776,
+ "learning_rate": 0.0006,
+ "loss": 4.395726013183594,
+ "step": 1480
+ },
+ {
+ "epoch": 20.698080279232112,
+ "grad_norm": 0.2136113941669464,
+ "learning_rate": 0.0006,
+ "loss": 4.370745086669922,
+ "step": 1490
+ },
+ {
+ "epoch": 20.837696335078533,
+ "grad_norm": 0.2094653993844986,
+ "learning_rate": 0.0006,
+ "loss": 4.447227096557617,
+ "step": 1500
+ },
+ {
+ "epoch": 20.977312390924958,
+ "grad_norm": 0.1883687525987625,
+ "learning_rate": 0.0006,
+ "loss": 4.387507629394531,
+ "step": 1510
+ },
+ {
+ "epoch": 21.0,
+ "eval_loss": 3.9870479106903076,
+ "eval_runtime": 2.6365,
+ "eval_samples_per_second": 925.861,
+ "eval_steps_per_second": 14.793,
+ "step": 1512
+ },
+ {
+ "epoch": 21.111692844677137,
+ "grad_norm": 0.2296922355890274,
+ "learning_rate": 0.0006,
+ "loss": 4.382663726806641,
+ "step": 1520
+ },
+ {
+ "epoch": 21.25130890052356,
+ "grad_norm": 0.17691919207572937,
+ "learning_rate": 0.0006,
+ "loss": 4.4230400085449215,
+ "step": 1530
+ },
+ {
+ "epoch": 21.390924956369982,
+ "grad_norm": 0.24876897037029266,
+ "learning_rate": 0.0006,
+ "loss": 4.363992309570312,
+ "step": 1540
+ },
+ {
+ "epoch": 21.530541012216403,
+ "grad_norm": 0.225335493683815,
+ "learning_rate": 0.0006,
+ "loss": 4.384988403320312,
+ "step": 1550
+ },
+ {
+ "epoch": 21.670157068062828,
+ "grad_norm": 0.24899786710739136,
+ "learning_rate": 0.0006,
+ "loss": 4.385707092285156,
+ "step": 1560
+ },
+ {
+ "epoch": 21.80977312390925,
+ "grad_norm": 0.22294718027114868,
+ "learning_rate": 0.0006,
+ "loss": 4.376637268066406,
+ "step": 1570
+ },
+ {
+ "epoch": 21.949389179755673,
+ "grad_norm": 0.2113858163356781,
+ "learning_rate": 0.0006,
+ "loss": 4.3080589294433596,
+ "step": 1580
+ },
+ {
+ "epoch": 22.0,
+ "eval_loss": 3.970257043838501,
+ "eval_runtime": 2.6354,
+ "eval_samples_per_second": 926.235,
+ "eval_steps_per_second": 14.799,
+ "step": 1584
+ },
+ {
+ "epoch": 22.083769633507853,
+ "grad_norm": 0.24430984258651733,
+ "learning_rate": 0.0006,
+ "loss": 4.341077423095703,
+ "step": 1590
+ },
+ {
+ "epoch": 22.223385689354277,
+ "grad_norm": 0.22397567331790924,
+ "learning_rate": 0.0006,
+ "loss": 4.262723541259765,
+ "step": 1600
+ },
+ {
+ "epoch": 22.363001745200698,
+ "grad_norm": 0.2470531463623047,
+ "learning_rate": 0.0006,
+ "loss": 4.320391845703125,
+ "step": 1610
+ },
+ {
+ "epoch": 22.50261780104712,
+ "grad_norm": 0.2752299904823303,
+ "learning_rate": 0.0006,
+ "loss": 4.299174499511719,
+ "step": 1620
+ },
+ {
+ "epoch": 22.642233856893544,
+ "grad_norm": 0.23211416602134705,
+ "learning_rate": 0.0006,
+ "loss": 4.290595245361328,
+ "step": 1630
+ },
+ {
+ "epoch": 22.781849912739965,
+ "grad_norm": 0.24011847376823425,
+ "learning_rate": 0.0006,
+ "loss": 4.348600006103515,
+ "step": 1640
+ },
+ {
+ "epoch": 22.921465968586386,
+ "grad_norm": 0.20325970649719238,
+ "learning_rate": 0.0006,
+ "loss": 4.325626373291016,
+ "step": 1650
+ },
+ {
+ "epoch": 23.0,
+ "eval_loss": 3.952242374420166,
+ "eval_runtime": 2.6199,
+ "eval_samples_per_second": 931.717,
+ "eval_steps_per_second": 14.886,
+ "step": 1656
+ },
+ {
+ "epoch": 23.05584642233857,
+ "grad_norm": 0.25825536251068115,
+ "learning_rate": 0.0006,
+ "loss": 4.303960800170898,
+ "step": 1660
+ },
+ {
+ "epoch": 23.195462478184993,
+ "grad_norm": 0.22531944513320923,
+ "learning_rate": 0.0006,
+ "loss": 4.309435272216797,
+ "step": 1670
+ },
+ {
+ "epoch": 23.335078534031414,
+ "grad_norm": 0.21413488686084747,
+ "learning_rate": 0.0006,
+ "loss": 4.307373046875,
+ "step": 1680
+ },
+ {
+ "epoch": 23.474694589877835,
+ "grad_norm": 0.2230566293001175,
+ "learning_rate": 0.0006,
+ "loss": 4.3524635314941404,
+ "step": 1690
+ },
+ {
+ "epoch": 23.61431064572426,
+ "grad_norm": 0.22512280941009521,
+ "learning_rate": 0.0006,
+ "loss": 4.348959350585938,
+ "step": 1700
+ },
+ {
+ "epoch": 23.75392670157068,
+ "grad_norm": 0.21183687448501587,
+ "learning_rate": 0.0006,
+ "loss": 4.293547058105469,
+ "step": 1710
+ },
+ {
+ "epoch": 23.8935427574171,
+ "grad_norm": 0.223998561501503,
+ "learning_rate": 0.0006,
+ "loss": 4.274700164794922,
+ "step": 1720
+ },
+ {
+ "epoch": 24.0,
+ "eval_loss": 3.94085693359375,
+ "eval_runtime": 2.6269,
+ "eval_samples_per_second": 929.226,
+ "eval_steps_per_second": 14.846,
+ "step": 1728
+ },
+ {
+ "epoch": 24.027923211169284,
+ "grad_norm": 0.23592819273471832,
+ "learning_rate": 0.0006,
+ "loss": 4.2767585754394535,
+ "step": 1730
+ },
+ {
+ "epoch": 24.167539267015705,
+ "grad_norm": 0.22928845882415771,
+ "learning_rate": 0.0006,
+ "loss": 4.336571121215821,
+ "step": 1740
+ },
+ {
+ "epoch": 24.30715532286213,
+ "grad_norm": 0.22282704710960388,
+ "learning_rate": 0.0006,
+ "loss": 4.291425323486328,
+ "step": 1750
+ },
+ {
+ "epoch": 24.44677137870855,
+ "grad_norm": 0.25562313199043274,
+ "learning_rate": 0.0006,
+ "loss": 4.316172027587891,
+ "step": 1760
+ },
+ {
+ "epoch": 24.586387434554975,
+ "grad_norm": 0.2171746790409088,
+ "learning_rate": 0.0006,
+ "loss": 4.291923522949219,
+ "step": 1770
+ },
+ {
+ "epoch": 24.726003490401396,
+ "grad_norm": 0.2340134084224701,
+ "learning_rate": 0.0006,
+ "loss": 4.285986328125,
+ "step": 1780
+ },
+ {
+ "epoch": 24.865619546247817,
+ "grad_norm": 0.23133744299411774,
+ "learning_rate": 0.0006,
+ "loss": 4.215341949462891,
+ "step": 1790
+ },
+ {
+ "epoch": 25.0,
+ "grad_norm": 0.2402936965227127,
+ "learning_rate": 0.0006,
+ "loss": 4.204048156738281,
+ "step": 1800
+ },
+ {
+ "epoch": 25.0,
+ "eval_loss": 3.92448091506958,
+ "eval_runtime": 2.6203,
+ "eval_samples_per_second": 931.585,
+ "eval_steps_per_second": 14.884,
+ "step": 1800
+ },
+ {
+ "epoch": 25.13961605584642,
+ "grad_norm": 0.26366516947746277,
+ "learning_rate": 0.0006,
+ "loss": 4.274285888671875,
+ "step": 1810
+ },
+ {
+ "epoch": 25.279232111692846,
+ "grad_norm": 0.21639637649059296,
+ "learning_rate": 0.0006,
+ "loss": 4.231010437011719,
+ "step": 1820
+ },
+ {
+ "epoch": 25.418848167539267,
+ "grad_norm": 0.22888189554214478,
+ "learning_rate": 0.0006,
+ "loss": 4.246493530273438,
+ "step": 1830
+ },
+ {
+ "epoch": 25.55846422338569,
+ "grad_norm": 0.24619103968143463,
+ "learning_rate": 0.0006,
+ "loss": 4.259849929809571,
+ "step": 1840
+ },
+ {
+ "epoch": 25.698080279232112,
+ "grad_norm": 0.21995064616203308,
+ "learning_rate": 0.0006,
+ "loss": 4.272932052612305,
+ "step": 1850
+ },
+ {
+ "epoch": 25.837696335078533,
+ "grad_norm": 0.2358444780111313,
+ "learning_rate": 0.0006,
+ "loss": 4.271376037597657,
+ "step": 1860
+ },
+ {
+ "epoch": 25.977312390924958,
+ "grad_norm": 0.22276923060417175,
+ "learning_rate": 0.0006,
+ "loss": 4.271081161499024,
+ "step": 1870
+ },
+ {
+ "epoch": 26.0,
+ "eval_loss": 3.9179155826568604,
+ "eval_runtime": 2.6347,
+ "eval_samples_per_second": 926.474,
+ "eval_steps_per_second": 14.802,
+ "step": 1872
+ },
+ {
+ "epoch": 26.111692844677137,
+ "grad_norm": 0.20989879965782166,
+ "learning_rate": 0.0006,
+ "loss": 4.200099945068359,
+ "step": 1880
+ },
+ {
+ "epoch": 26.25130890052356,
+ "grad_norm": 0.27911919355392456,
+ "learning_rate": 0.0006,
+ "loss": 4.1891357421875,
+ "step": 1890
+ },
+ {
+ "epoch": 26.390924956369982,
+ "grad_norm": 0.2901744544506073,
+ "learning_rate": 0.0006,
+ "loss": 4.203256607055664,
+ "step": 1900
+ },
+ {
+ "epoch": 26.530541012216403,
+ "grad_norm": 0.21709665656089783,
+ "learning_rate": 0.0006,
+ "loss": 4.215483856201172,
+ "step": 1910
+ },
+ {
+ "epoch": 26.670157068062828,
+ "grad_norm": 0.24950534105300903,
+ "learning_rate": 0.0006,
+ "loss": 4.2203014373779295,
+ "step": 1920
+ },
+ {
+ "epoch": 26.80977312390925,
+ "grad_norm": 0.2143084555864334,
+ "learning_rate": 0.0006,
+ "loss": 4.234615325927734,
+ "step": 1930
+ },
+ {
+ "epoch": 26.949389179755673,
+ "grad_norm": 0.22018875181674957,
+ "learning_rate": 0.0006,
+ "loss": 4.245726776123047,
+ "step": 1940
+ },
+ {
+ "epoch": 27.0,
+ "eval_loss": 3.906799793243408,
+ "eval_runtime": 2.6208,
+ "eval_samples_per_second": 931.397,
+ "eval_steps_per_second": 14.881,
+ "step": 1944
+ },
+ {
+ "epoch": 27.083769633507853,
+ "grad_norm": 0.21637099981307983,
+ "learning_rate": 0.0006,
+ "loss": 4.24003791809082,
+ "step": 1950
+ },
+ {
+ "epoch": 27.223385689354277,
+ "grad_norm": 0.2187204509973526,
+ "learning_rate": 0.0006,
+ "loss": 4.159322357177734,
+ "step": 1960
+ },
+ {
+ "epoch": 27.363001745200698,
+ "grad_norm": 0.2560689449310303,
+ "learning_rate": 0.0006,
+ "loss": 4.147882080078125,
+ "step": 1970
+ },
+ {
+ "epoch": 27.50261780104712,
+ "grad_norm": 0.2346278876066208,
+ "learning_rate": 0.0006,
+ "loss": 4.213126754760742,
+ "step": 1980
+ },
+ {
+ "epoch": 27.642233856893544,
+ "grad_norm": 0.24696636199951172,
+ "learning_rate": 0.0006,
+ "loss": 4.218080139160156,
+ "step": 1990
+ },
+ {
+ "epoch": 27.781849912739965,
+ "grad_norm": 0.24423378705978394,
+ "learning_rate": 0.0006,
+ "loss": 4.183169174194336,
+ "step": 2000
+ },
+ {
+ "epoch": 27.921465968586386,
+ "grad_norm": 0.2107905000448227,
+ "learning_rate": 0.0006,
+ "loss": 4.217204666137695,
+ "step": 2010
+ },
+ {
+ "epoch": 28.0,
+ "eval_loss": 3.89740252494812,
+ "eval_runtime": 2.6187,
+ "eval_samples_per_second": 932.142,
+ "eval_steps_per_second": 14.893,
+ "step": 2016
+ },
+ {
+ "epoch": 28.05584642233857,
+ "grad_norm": 0.35934513807296753,
+ "learning_rate": 0.0006,
+ "loss": 4.1997215270996096,
+ "step": 2020
+ },
+ {
+ "epoch": 28.195462478184993,
+ "grad_norm": 0.2634032368659973,
+ "learning_rate": 0.0006,
+ "loss": 4.210472106933594,
+ "step": 2030
+ },
+ {
+ "epoch": 28.335078534031414,
+ "grad_norm": 0.24293319880962372,
+ "learning_rate": 0.0006,
+ "loss": 4.179236221313476,
+ "step": 2040
+ },
+ {
+ "epoch": 28.474694589877835,
+ "grad_norm": 0.2374039888381958,
+ "learning_rate": 0.0006,
+ "loss": 4.186960601806641,
+ "step": 2050
+ },
+ {
+ "epoch": 28.61431064572426,
+ "grad_norm": 0.21195976436138153,
+ "learning_rate": 0.0006,
+ "loss": 4.2232818603515625,
+ "step": 2060
+ },
+ {
+ "epoch": 28.75392670157068,
+ "grad_norm": 0.2089933454990387,
+ "learning_rate": 0.0006,
+ "loss": 4.177834701538086,
+ "step": 2070
+ },
+ {
+ "epoch": 28.8935427574171,
+ "grad_norm": 0.23219753801822662,
+ "learning_rate": 0.0006,
+ "loss": 4.201029205322266,
+ "step": 2080
+ },
+ {
+ "epoch": 29.0,
+ "eval_loss": 3.8887510299682617,
+ "eval_runtime": 2.6085,
+ "eval_samples_per_second": 935.781,
+ "eval_steps_per_second": 14.951,
+ "step": 2088
+ },
+ {
+ "epoch": 29.027923211169284,
+ "grad_norm": 0.2494087666273117,
+ "learning_rate": 0.0006,
+ "loss": 4.169363403320313,
+ "step": 2090
+ },
+ {
+ "epoch": 29.167539267015705,
+ "grad_norm": 0.23004640638828278,
+ "learning_rate": 0.0006,
+ "loss": 4.122265625,
+ "step": 2100
+ },
+ {
+ "epoch": 29.30715532286213,
+ "grad_norm": 0.20843105018138885,
+ "learning_rate": 0.0006,
+ "loss": 4.124951171875,
+ "step": 2110
+ },
+ {
+ "epoch": 29.44677137870855,
+ "grad_norm": 0.25792133808135986,
+ "learning_rate": 0.0006,
+ "loss": 4.145717620849609,
+ "step": 2120
+ },
+ {
+ "epoch": 29.586387434554975,
+ "grad_norm": 0.26935991644859314,
+ "learning_rate": 0.0006,
+ "loss": 4.208763885498047,
+ "step": 2130
+ },
+ {
+ "epoch": 29.726003490401396,
+ "grad_norm": 0.23253610730171204,
+ "learning_rate": 0.0006,
+ "loss": 4.1822052001953125,
+ "step": 2140
+ },
+ {
+ "epoch": 29.865619546247817,
+ "grad_norm": 0.23977990448474884,
+ "learning_rate": 0.0006,
+ "loss": 4.140848159790039,
+ "step": 2150
+ },
+ {
+ "epoch": 30.0,
+ "grad_norm": 0.26465073227882385,
+ "learning_rate": 0.0006,
+ "loss": 4.239994812011719,
+ "step": 2160
+ },
+ {
+ "epoch": 30.0,
+ "eval_loss": 3.8858282566070557,
+ "eval_runtime": 2.6201,
+ "eval_samples_per_second": 931.659,
+ "eval_steps_per_second": 14.885,
+ "step": 2160
+ },
+ {
+ "epoch": 30.13961605584642,
+ "grad_norm": 0.25698450207710266,
+ "learning_rate": 0.0006,
+ "loss": 4.1457672119140625,
+ "step": 2170
+ },
+ {
+ "epoch": 30.279232111692846,
+ "grad_norm": 0.23117145895957947,
+ "learning_rate": 0.0006,
+ "loss": 4.139920806884765,
+ "step": 2180
+ },
+ {
+ "epoch": 30.418848167539267,
+ "grad_norm": 0.23112572729587555,
+ "learning_rate": 0.0006,
+ "loss": 4.174540710449219,
+ "step": 2190
+ },
+ {
+ "epoch": 30.55846422338569,
+ "grad_norm": 0.21470710635185242,
+ "learning_rate": 0.0006,
+ "loss": 4.143948364257812,
+ "step": 2200
+ },
+ {
+ "epoch": 30.698080279232112,
+ "grad_norm": 0.23187771439552307,
+ "learning_rate": 0.0006,
+ "loss": 4.218885803222657,
+ "step": 2210
+ },
+ {
+ "epoch": 30.837696335078533,
+ "grad_norm": 0.22308728098869324,
+ "learning_rate": 0.0006,
+ "loss": 4.082337188720703,
+ "step": 2220
+ },
+ {
+ "epoch": 30.977312390924958,
+ "grad_norm": 0.2217535823583603,
+ "learning_rate": 0.0006,
+ "loss": 4.116067504882812,
+ "step": 2230
+ },
+ {
+ "epoch": 31.0,
+ "eval_loss": 3.8785805702209473,
+ "eval_runtime": 2.6215,
+ "eval_samples_per_second": 931.144,
+ "eval_steps_per_second": 14.877,
+ "step": 2232
+ },
+ {
+ "epoch": 31.111692844677137,
+ "grad_norm": 0.2576958239078522,
+ "learning_rate": 0.0006,
+ "loss": 4.133813095092774,
+ "step": 2240
+ },
+ {
+ "epoch": 31.25130890052356,
+ "grad_norm": 0.23338694870471954,
+ "learning_rate": 0.0006,
+ "loss": 4.206774139404297,
+ "step": 2250
+ },
+ {
+ "epoch": 31.390924956369982,
+ "grad_norm": 0.24903886020183563,
+ "learning_rate": 0.0006,
+ "loss": 4.16229476928711,
+ "step": 2260
+ },
+ {
+ "epoch": 31.530541012216403,
+ "grad_norm": 0.26954007148742676,
+ "learning_rate": 0.0006,
+ "loss": 4.10771484375,
+ "step": 2270
+ },
+ {
+ "epoch": 31.670157068062828,
+ "grad_norm": 0.2246377170085907,
+ "learning_rate": 0.0006,
+ "loss": 4.115251541137695,
+ "step": 2280
+ },
+ {
+ "epoch": 31.80977312390925,
+ "grad_norm": 0.2282128632068634,
+ "learning_rate": 0.0006,
+ "loss": 4.139779663085937,
+ "step": 2290
+ },
+ {
+ "epoch": 31.949389179755673,
+ "grad_norm": 0.2224973440170288,
+ "learning_rate": 0.0006,
+ "loss": 4.120811462402344,
+ "step": 2300
+ },
+ {
+ "epoch": 32.0,
+ "eval_loss": 3.8674914836883545,
+ "eval_runtime": 2.6351,
+ "eval_samples_per_second": 926.324,
+ "eval_steps_per_second": 14.8,
+ "step": 2304
+ },
+ {
+ "epoch": 32.083769633507856,
+ "grad_norm": 0.31452250480651855,
+ "learning_rate": 0.0006,
+ "loss": 4.1610572814941404,
+ "step": 2310
+ },
+ {
+ "epoch": 32.223385689354274,
+ "grad_norm": 0.2622784972190857,
+ "learning_rate": 0.0006,
+ "loss": 4.0874076843261715,
+ "step": 2320
+ },
+ {
+ "epoch": 32.3630017452007,
+ "grad_norm": 0.23233181238174438,
+ "learning_rate": 0.0006,
+ "loss": 4.1359100341796875,
+ "step": 2330
+ },
+ {
+ "epoch": 32.50261780104712,
+ "grad_norm": 0.22049587965011597,
+ "learning_rate": 0.0006,
+ "loss": 4.117932891845703,
+ "step": 2340
+ },
+ {
+ "epoch": 32.64223385689354,
+ "grad_norm": 0.28545913100242615,
+ "learning_rate": 0.0006,
+ "loss": 4.170756530761719,
+ "step": 2350
+ },
+ {
+ "epoch": 32.781849912739965,
+ "grad_norm": 0.22290664911270142,
+ "learning_rate": 0.0006,
+ "loss": 4.178508758544922,
+ "step": 2360
+ },
+ {
+ "epoch": 32.92146596858639,
+ "grad_norm": 0.21724696457386017,
+ "learning_rate": 0.0006,
+ "loss": 4.125429916381836,
+ "step": 2370
+ },
+ {
+ "epoch": 33.0,
+ "eval_loss": 3.868417263031006,
+ "eval_runtime": 2.6391,
+ "eval_samples_per_second": 924.941,
+ "eval_steps_per_second": 14.778,
+ "step": 2376
+ },
+ {
+ "epoch": 33.05584642233857,
+ "grad_norm": 0.256018728017807,
+ "learning_rate": 0.0006,
+ "loss": 4.098083114624023,
+ "step": 2380
+ },
+ {
+ "epoch": 33.19546247818499,
+ "grad_norm": 0.24846933782100677,
+ "learning_rate": 0.0006,
+ "loss": 4.063763427734375,
+ "step": 2390
+ },
+ {
+ "epoch": 33.33507853403141,
+ "grad_norm": 0.24831777811050415,
+ "learning_rate": 0.0006,
+ "loss": 4.077348709106445,
+ "step": 2400
+ },
+ {
+ "epoch": 33.474694589877835,
+ "grad_norm": 0.253953218460083,
+ "learning_rate": 0.0006,
+ "loss": 4.134611511230469,
+ "step": 2410
+ },
+ {
+ "epoch": 33.61431064572426,
+ "grad_norm": 0.2522493302822113,
+ "learning_rate": 0.0006,
+ "loss": 4.125479888916016,
+ "step": 2420
+ },
+ {
+ "epoch": 33.753926701570684,
+ "grad_norm": 0.21423368155956268,
+ "learning_rate": 0.0006,
+ "loss": 4.126710510253906,
+ "step": 2430
+ },
+ {
+ "epoch": 33.8935427574171,
+ "grad_norm": 0.23868481814861298,
+ "learning_rate": 0.0006,
+ "loss": 4.082051849365234,
+ "step": 2440
+ },
+ {
+ "epoch": 34.0,
+ "eval_loss": 3.86088490486145,
+ "eval_runtime": 2.6236,
+ "eval_samples_per_second": 930.388,
+ "eval_steps_per_second": 14.865,
+ "step": 2448
+ },
+ {
+ "epoch": 34.02792321116929,
+ "grad_norm": 0.24495521187782288,
+ "learning_rate": 0.0006,
+ "loss": 4.133685684204101,
+ "step": 2450
+ },
+ {
+ "epoch": 34.167539267015705,
+ "grad_norm": 0.24209558963775635,
+ "learning_rate": 0.0006,
+ "loss": 4.015423583984375,
+ "step": 2460
+ },
+ {
+ "epoch": 34.30715532286213,
+ "grad_norm": 0.21196018159389496,
+ "learning_rate": 0.0006,
+ "loss": 4.051766967773437,
+ "step": 2470
+ },
+ {
+ "epoch": 34.446771378708554,
+ "grad_norm": 0.25079014897346497,
+ "learning_rate": 0.0006,
+ "loss": 4.063561248779297,
+ "step": 2480
+ },
+ {
+ "epoch": 34.58638743455497,
+ "grad_norm": 0.2742410898208618,
+ "learning_rate": 0.0006,
+ "loss": 4.100325393676758,
+ "step": 2490
+ },
+ {
+ "epoch": 34.726003490401396,
+ "grad_norm": 0.27754494547843933,
+ "learning_rate": 0.0006,
+ "loss": 4.078593444824219,
+ "step": 2500
+ },
+ {
+ "epoch": 34.86561954624782,
+ "grad_norm": 0.2332543283700943,
+ "learning_rate": 0.0006,
+ "loss": 4.075416564941406,
+ "step": 2510
+ },
+ {
+ "epoch": 35.0,
+ "grad_norm": 0.24284276366233826,
+ "learning_rate": 0.0006,
+ "loss": 4.094195556640625,
+ "step": 2520
+ },
+ {
+ "epoch": 35.0,
+ "eval_loss": 3.8581764698028564,
+ "eval_runtime": 2.6224,
+ "eval_samples_per_second": 930.816,
+ "eval_steps_per_second": 14.872,
+ "step": 2520
+ },
+ {
+ "epoch": 35.139616055846425,
+ "grad_norm": 0.2262984961271286,
+ "learning_rate": 0.0006,
+ "loss": 4.038826751708984,
+ "step": 2530
+ },
+ {
+ "epoch": 35.27923211169284,
+ "grad_norm": 0.2867301404476166,
+ "learning_rate": 0.0006,
+ "loss": 4.09522705078125,
+ "step": 2540
+ },
+ {
+ "epoch": 35.41884816753927,
+ "grad_norm": 0.24296395480632782,
+ "learning_rate": 0.0006,
+ "loss": 4.082079315185547,
+ "step": 2550
+ },
+ {
+ "epoch": 35.55846422338569,
+ "grad_norm": 0.24271143972873688,
+ "learning_rate": 0.0006,
+ "loss": 4.0996440887451175,
+ "step": 2560
+ },
+ {
+ "epoch": 35.69808027923211,
+ "grad_norm": 0.2230568528175354,
+ "learning_rate": 0.0006,
+ "loss": 4.0597373962402346,
+ "step": 2570
+ },
+ {
+ "epoch": 35.83769633507853,
+ "grad_norm": 0.22024185955524445,
+ "learning_rate": 0.0006,
+ "loss": 4.10230598449707,
+ "step": 2580
+ },
+ {
+ "epoch": 35.97731239092496,
+ "grad_norm": 0.20378516614437103,
+ "learning_rate": 0.0006,
+ "loss": 4.1020057678222654,
+ "step": 2590
+ },
+ {
+ "epoch": 36.0,
+ "eval_loss": 3.859058141708374,
+ "eval_runtime": 2.6211,
+ "eval_samples_per_second": 931.277,
+ "eval_steps_per_second": 14.879,
+ "step": 2592
+ },
+ {
+ "epoch": 36.11169284467714,
+ "grad_norm": 0.2807842791080475,
+ "learning_rate": 0.0006,
+ "loss": 4.0657600402832035,
+ "step": 2600
+ },
+ {
+ "epoch": 36.25130890052356,
+ "grad_norm": 0.2572230100631714,
+ "learning_rate": 0.0006,
+ "loss": 4.036009979248047,
+ "step": 2610
+ },
+ {
+ "epoch": 36.390924956369986,
+ "grad_norm": 0.2482750564813614,
+ "learning_rate": 0.0006,
+ "loss": 4.077861785888672,
+ "step": 2620
+ },
+ {
+ "epoch": 36.5305410122164,
+ "grad_norm": 0.250893771648407,
+ "learning_rate": 0.0006,
+ "loss": 4.059228515625,
+ "step": 2630
+ },
+ {
+ "epoch": 36.67015706806283,
+ "grad_norm": 0.24665629863739014,
+ "learning_rate": 0.0006,
+ "loss": 4.072270202636719,
+ "step": 2640
+ },
+ {
+ "epoch": 36.80977312390925,
+ "grad_norm": 0.24624468386173248,
+ "learning_rate": 0.0006,
+ "loss": 4.082905197143555,
+ "step": 2650
+ },
+ {
+ "epoch": 36.94938917975567,
+ "grad_norm": 0.22487853467464447,
+ "learning_rate": 0.0006,
+ "loss": 4.067748641967773,
+ "step": 2660
+ },
+ {
+ "epoch": 37.0,
+ "eval_loss": 3.851016044616699,
+ "eval_runtime": 2.6179,
+ "eval_samples_per_second": 932.439,
+ "eval_steps_per_second": 14.898,
+ "step": 2664
+ },
+ {
+ "epoch": 37.083769633507856,
+ "grad_norm": 0.23729118704795837,
+ "learning_rate": 0.0006,
+ "loss": 4.073279953002929,
+ "step": 2670
+ },
+ {
+ "epoch": 37.223385689354274,
+ "grad_norm": 0.2738165259361267,
+ "learning_rate": 0.0006,
+ "loss": 4.019068908691406,
+ "step": 2680
+ },
+ {
+ "epoch": 37.3630017452007,
+ "grad_norm": 0.2373737096786499,
+ "learning_rate": 0.0006,
+ "loss": 3.9804473876953126,
+ "step": 2690
+ },
+ {
+ "epoch": 37.50261780104712,
+ "grad_norm": 0.2902696132659912,
+ "learning_rate": 0.0006,
+ "loss": 4.12921142578125,
+ "step": 2700
+ },
+ {
+ "epoch": 37.64223385689354,
+ "grad_norm": 0.25886985659599304,
+ "learning_rate": 0.0006,
+ "loss": 4.040469360351563,
+ "step": 2710
+ },
+ {
+ "epoch": 37.781849912739965,
+ "grad_norm": 0.22481660544872284,
+ "learning_rate": 0.0006,
+ "loss": 4.101229858398438,
+ "step": 2720
+ },
+ {
+ "epoch": 37.92146596858639,
+ "grad_norm": 0.22507773339748383,
+ "learning_rate": 0.0006,
+ "loss": 4.04801025390625,
+ "step": 2730
+ },
+ {
+ "epoch": 38.0,
+ "eval_loss": 3.8515193462371826,
+ "eval_runtime": 2.637,
+ "eval_samples_per_second": 925.675,
+ "eval_steps_per_second": 14.79,
+ "step": 2736
+ },
+ {
+ "epoch": 38.05584642233857,
+ "grad_norm": 0.23503729701042175,
+ "learning_rate": 0.0006,
+ "loss": 4.047577285766602,
+ "step": 2740
+ },
+ {
+ "epoch": 38.19546247818499,
+ "grad_norm": 0.23788076639175415,
+ "learning_rate": 0.0006,
+ "loss": 4.041569519042969,
+ "step": 2750
+ },
+ {
+ "epoch": 38.33507853403141,
+ "grad_norm": 0.2763453423976898,
+ "learning_rate": 0.0006,
+ "loss": 4.118089294433593,
+ "step": 2760
+ },
+ {
+ "epoch": 38.474694589877835,
+ "grad_norm": 0.25524625182151794,
+ "learning_rate": 0.0006,
+ "loss": 4.019875335693359,
+ "step": 2770
+ },
+ {
+ "epoch": 38.61431064572426,
+ "grad_norm": 0.23400157690048218,
+ "learning_rate": 0.0006,
+ "loss": 3.962510681152344,
+ "step": 2780
+ },
+ {
+ "epoch": 38.753926701570684,
+ "grad_norm": 0.26603859663009644,
+ "learning_rate": 0.0006,
+ "loss": 4.062420654296875,
+ "step": 2790
+ },
+ {
+ "epoch": 38.8935427574171,
+ "grad_norm": 0.23940995335578918,
+ "learning_rate": 0.0006,
+ "loss": 4.164274597167969,
+ "step": 2800
+ },
+ {
+ "epoch": 39.0,
+ "eval_loss": 3.848816394805908,
+ "eval_runtime": 2.6123,
+ "eval_samples_per_second": 934.433,
+ "eval_steps_per_second": 14.929,
+ "step": 2808
+ },
+ {
+ "epoch": 39.02792321116929,
+ "grad_norm": 0.245711550116539,
+ "learning_rate": 0.0006,
+ "loss": 4.050062561035157,
+ "step": 2810
+ },
+ {
+ "epoch": 39.167539267015705,
+ "grad_norm": 0.22280636429786682,
+ "learning_rate": 0.0006,
+ "loss": 4.011087799072266,
+ "step": 2820
+ },
+ {
+ "epoch": 39.30715532286213,
+ "grad_norm": 0.2487151026725769,
+ "learning_rate": 0.0006,
+ "loss": 4.033265686035156,
+ "step": 2830
+ },
+ {
+ "epoch": 39.446771378708554,
+ "grad_norm": 0.260999858379364,
+ "learning_rate": 0.0006,
+ "loss": 4.009197616577149,
+ "step": 2840
+ },
+ {
+ "epoch": 39.58638743455497,
+ "grad_norm": 0.43024319410324097,
+ "learning_rate": 0.0006,
+ "loss": 4.0970100402832035,
+ "step": 2850
+ },
+ {
+ "epoch": 39.726003490401396,
+ "grad_norm": 0.3226476311683655,
+ "learning_rate": 0.0006,
+ "loss": 4.089163208007813,
+ "step": 2860
+ },
+ {
+ "epoch": 39.86561954624782,
+ "grad_norm": 0.23710575699806213,
+ "learning_rate": 0.0006,
+ "loss": 4.05914535522461,
+ "step": 2870
+ },
+ {
+ "epoch": 40.0,
+ "grad_norm": 0.24438202381134033,
+ "learning_rate": 0.0006,
+ "loss": 4.071099853515625,
+ "step": 2880
+ },
+ {
+ "epoch": 40.0,
+ "eval_loss": 3.8521595001220703,
+ "eval_runtime": 2.6118,
+ "eval_samples_per_second": 934.608,
+ "eval_steps_per_second": 14.932,
+ "step": 2880
+ },
+ {
+ "epoch": 40.139616055846425,
+ "grad_norm": 0.2233884334564209,
+ "learning_rate": 0.0006,
+ "loss": 4.029550933837891,
+ "step": 2890
+ },
+ {
+ "epoch": 40.27923211169284,
+ "grad_norm": 0.23693658411502838,
+ "learning_rate": 0.0006,
+ "loss": 4.023176574707032,
+ "step": 2900
+ },
+ {
+ "epoch": 40.41884816753927,
+ "grad_norm": 0.23371657729148865,
+ "learning_rate": 0.0006,
+ "loss": 4.052134704589844,
+ "step": 2910
+ },
+ {
+ "epoch": 40.55846422338569,
+ "grad_norm": 0.239592507481575,
+ "learning_rate": 0.0006,
+ "loss": 4.033892059326172,
+ "step": 2920
+ },
+ {
+ "epoch": 40.69808027923211,
+ "grad_norm": 0.24122460186481476,
+ "learning_rate": 0.0006,
+ "loss": 4.117848205566406,
+ "step": 2930
+ },
+ {
+ "epoch": 40.83769633507853,
+ "grad_norm": 0.225126713514328,
+ "learning_rate": 0.0006,
+ "loss": 4.0176139831542965,
+ "step": 2940
+ },
+ {
+ "epoch": 40.97731239092496,
+ "grad_norm": 0.2192733883857727,
+ "learning_rate": 0.0006,
+ "loss": 4.072132110595703,
+ "step": 2950
+ },
+ {
+ "epoch": 41.0,
+ "eval_loss": 3.8469386100769043,
+ "eval_runtime": 2.6199,
+ "eval_samples_per_second": 931.716,
+ "eval_steps_per_second": 14.886,
+ "step": 2952
+ },
+ {
+ "epoch": 41.11169284467714,
+ "grad_norm": 0.3247385621070862,
+ "learning_rate": 0.0006,
+ "loss": 3.9490089416503906,
+ "step": 2960
+ },
+ {
+ "epoch": 41.25130890052356,
+ "grad_norm": 0.2508738338947296,
+ "learning_rate": 0.0006,
+ "loss": 4.0644981384277346,
+ "step": 2970
+ },
+ {
+ "epoch": 41.390924956369986,
+ "grad_norm": 0.24719730019569397,
+ "learning_rate": 0.0006,
+ "loss": 3.9777069091796875,
+ "step": 2980
+ },
+ {
+ "epoch": 41.5305410122164,
+ "grad_norm": 0.2791763246059418,
+ "learning_rate": 0.0006,
+ "loss": 4.140811920166016,
+ "step": 2990
+ },
+ {
+ "epoch": 41.67015706806283,
+ "grad_norm": 0.23849117755889893,
+ "learning_rate": 0.0006,
+ "loss": 3.9691139221191407,
+ "step": 3000
+ },
+ {
+ "epoch": 41.80977312390925,
+ "grad_norm": 0.21639244258403778,
+ "learning_rate": 0.0006,
+ "loss": 3.9948379516601564,
+ "step": 3010
+ },
+ {
+ "epoch": 41.94938917975567,
+ "grad_norm": 0.21242210268974304,
+ "learning_rate": 0.0006,
+ "loss": 4.056636047363281,
+ "step": 3020
+ },
+ {
+ "epoch": 42.0,
+ "eval_loss": 3.851733684539795,
+ "eval_runtime": 2.6183,
+ "eval_samples_per_second": 932.283,
+ "eval_steps_per_second": 14.895,
+ "step": 3024
+ },
+ {
+ "epoch": 42.083769633507856,
+ "grad_norm": 0.23022547364234924,
+ "learning_rate": 0.0006,
+ "loss": 3.9390525817871094,
+ "step": 3030
+ },
+ {
+ "epoch": 42.223385689354274,
+ "grad_norm": 0.2444913536310196,
+ "learning_rate": 0.0006,
+ "loss": 3.930714416503906,
+ "step": 3040
+ },
+ {
+ "epoch": 42.3630017452007,
+ "grad_norm": 0.2900421917438507,
+ "learning_rate": 0.0006,
+ "loss": 3.9969547271728514,
+ "step": 3050
+ },
+ {
+ "epoch": 42.50261780104712,
+ "grad_norm": 0.23384881019592285,
+ "learning_rate": 0.0006,
+ "loss": 4.0108074188232425,
+ "step": 3060
+ },
+ {
+ "epoch": 42.64223385689354,
+ "grad_norm": 0.25605347752571106,
+ "learning_rate": 0.0006,
+ "loss": 3.968499755859375,
+ "step": 3070
+ },
+ {
+ "epoch": 42.781849912739965,
+ "grad_norm": 0.23523737490177155,
+ "learning_rate": 0.0006,
+ "loss": 4.086688232421875,
+ "step": 3080
+ },
+ {
+ "epoch": 42.92146596858639,
+ "grad_norm": 0.23249155282974243,
+ "learning_rate": 0.0006,
+ "loss": 4.011849212646484,
+ "step": 3090
+ },
+ {
+ "epoch": 43.0,
+ "eval_loss": 3.844954252243042,
+ "eval_runtime": 2.6376,
+ "eval_samples_per_second": 925.458,
+ "eval_steps_per_second": 14.786,
+ "step": 3096
+ },
+ {
+ "epoch": 43.05584642233857,
+ "grad_norm": 0.24028173089027405,
+ "learning_rate": 0.0006,
+ "loss": 3.973236083984375,
+ "step": 3100
+ },
+ {
+ "epoch": 43.19546247818499,
+ "grad_norm": 0.28670504689216614,
+ "learning_rate": 0.0006,
+ "loss": 4.019177627563477,
+ "step": 3110
+ },
+ {
+ "epoch": 43.33507853403141,
+ "grad_norm": 0.2485414445400238,
+ "learning_rate": 0.0006,
+ "loss": 3.9387283325195312,
+ "step": 3120
+ },
+ {
+ "epoch": 43.474694589877835,
+ "grad_norm": 0.24301430583000183,
+ "learning_rate": 0.0006,
+ "loss": 4.01648178100586,
+ "step": 3130
+ },
+ {
+ "epoch": 43.61431064572426,
+ "grad_norm": 0.2293490469455719,
+ "learning_rate": 0.0006,
+ "loss": 4.034006500244141,
+ "step": 3140
+ },
+ {
+ "epoch": 43.753926701570684,
+ "grad_norm": 0.224209263920784,
+ "learning_rate": 0.0006,
+ "loss": 4.025347137451172,
+ "step": 3150
+ },
+ {
+ "epoch": 43.8935427574171,
+ "grad_norm": 0.22185851633548737,
+ "learning_rate": 0.0006,
+ "loss": 4.037354278564453,
+ "step": 3160
+ },
+ {
+ "epoch": 44.0,
+ "eval_loss": 3.8588380813598633,
+ "eval_runtime": 2.6377,
+ "eval_samples_per_second": 925.432,
+ "eval_steps_per_second": 14.786,
+ "step": 3168
+ },
+ {
+ "epoch": 44.02792321116929,
+ "grad_norm": 0.332205206155777,
+ "learning_rate": 0.0006,
+ "loss": 4.052106475830078,
+ "step": 3170
+ },
+ {
+ "epoch": 44.167539267015705,
+ "grad_norm": 0.2365674376487732,
+ "learning_rate": 0.0006,
+ "loss": 4.044369506835937,
+ "step": 3180
+ },
+ {
+ "epoch": 44.30715532286213,
+ "grad_norm": 0.23375248908996582,
+ "learning_rate": 0.0006,
+ "loss": 3.979005813598633,
+ "step": 3190
+ },
+ {
+ "epoch": 44.446771378708554,
+ "grad_norm": 0.2364208847284317,
+ "learning_rate": 0.0006,
+ "loss": 3.936968994140625,
+ "step": 3200
+ },
+ {
+ "epoch": 44.58638743455497,
+ "grad_norm": 0.2860606908798218,
+ "learning_rate": 0.0006,
+ "loss": 3.9862152099609376,
+ "step": 3210
+ },
+ {
+ "epoch": 44.726003490401396,
+ "grad_norm": 0.23332272469997406,
+ "learning_rate": 0.0006,
+ "loss": 4.02375602722168,
+ "step": 3220
+ },
+ {
+ "epoch": 44.86561954624782,
+ "grad_norm": 0.25934821367263794,
+ "learning_rate": 0.0006,
+ "loss": 3.986064910888672,
+ "step": 3230
+ },
+ {
+ "epoch": 45.0,
+ "grad_norm": 0.24828918278217316,
+ "learning_rate": 0.0006,
+ "loss": 4.0054443359375,
+ "step": 3240
+ },
+ {
+ "epoch": 45.0,
+ "eval_loss": 3.8444252014160156,
+ "eval_runtime": 2.6215,
+ "eval_samples_per_second": 931.139,
+ "eval_steps_per_second": 14.877,
+ "step": 3240
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 7200,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 100,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 1.3825324148903117e+18,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/t1a/t1a_S2_C2_r2l50_i5exp/train.log b/runs/t1a/t1a_S2_C2_r2l50_i5exp/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..711091ddf1fba611ebc329869a157535e0030244
--- /dev/null
+++ b/runs/t1a/t1a_S2_C2_r2l50_i5exp/train.log
@@ -0,0 +1,3603 @@
+W0818 05:47:41.793000 2570106 torch/distributed/run.py:851]
+W0818 05:47:41.793000 2570106 torch/distributed/run.py:851] *****************************************
+W0818 05:47:41.793000 2570106 torch/distributed/run.py:851] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
+W0818 05:47:41.793000 2570106 torch/distributed/run.py:851] *****************************************
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 100
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 8
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 8
+[transformers] Total optimization steps = 7,200
+[transformers] Number of trainable parameters = 145,836,544
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: setting up run krsk2578
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260818_054802-krsk2578
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t1a_S2_C2_r2l50_i5exp
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/krsk2578
+
0%| | 0/7200 [00:00 will be ignored
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[rank5]:W0818 05:48:25.587000 2570311 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank7]:W0818 05:48:25.592000 2570313 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[rank1]:W0818 05:48:25.796000 2570307 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[rank3]:W0818 05:48:26.126000 2570309 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank6]:W0818 05:48:26.279000 2570312 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank2]:W0818 05:48:26.462000 2570308 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[rank0]:W0818 05:48:27.354000 2570306 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+
0%| | 1/7200 [00:45<90:26:27, 45.23s/it]
{'loss': '12.01', 'grad_norm': '2.749', 'learning_rate': '0', 'epoch': '0.01396'}
+
0%| | 1/7200 [00:45<90:26:27, 45.23s/it]
0%| | 2/7200 [00:45<37:59:36, 19.00s/it]
0%| | 3/7200 [00:46<21:13:00, 10.61s/it]
0%| | 4/7200 [00:47<13:20:11, 6.67s/it]
0%| | 5/7200 [00:47<8:58:54, 4.49s/it]
0%| | 6/7200 [00:48<6:21:21, 3.18s/it]
0%| | 7/7200 [00:49<4:41:23, 2.35s/it]
0%| | 8/7200 [00:49<3:35:52, 1.80s/it]
0%| | 9/7200 [00:50<2:52:00, 1.44s/it]
0%| | 10/7200 [00:50<2:22:12, 1.19s/it]
{'loss': '11.67', 'grad_norm': '1.755', 'learning_rate': '5.4e-05', 'epoch': '0.1396'}
+
0%| | 10/7200 [00:50<2:22:12, 1.19s/it]
0%| | 11/7200 [00:51<2:01:50, 1.02s/it]
0%| | 12/7200 [00:52<1:47:45, 1.11it/s]
0%| | 13/7200 [00:52<1:38:02, 1.22it/s]
0%| | 14/7200 [00:53<1:31:14, 1.31it/s]
0%| | 15/7200 [00:54<1:26:27, 1.39it/s]
0%| | 16/7200 [00:54<1:23:07, 1.44it/s]
0%| | 17/7200 [00:55<1:20:46, 1.48it/s]
0%| | 18/7200 [00:55<1:19:09, 1.51it/s]
0%| | 19/7200 [00:56<1:17:59, 1.53it/s]
0%| | 20/7200 [00:57<1:17:12, 1.55it/s]
{'loss': '11', 'grad_norm': '1.692', 'learning_rate': '0.000114', 'epoch': '0.2792'}
+
0%| | 20/7200 [00:57<1:17:12, 1.55it/s]
0%| | 21/7200 [00:57<1:16:36, 1.56it/s]
0%| | 22/7200 [00:58<1:16:11, 1.57it/s]
0%| | 23/7200 [00:59<1:15:52, 1.58it/s]
0%| | 24/7200 [00:59<1:15:38, 1.58it/s]
0%| | 25/7200 [01:00<1:15:27, 1.58it/s]
0%| | 26/7200 [01:00<1:15:21, 1.59it/s]
0%| | 27/7200 [01:01<1:15:17, 1.59it/s]
0%| | 28/7200 [01:02<1:15:12, 1.59it/s]
0%| | 29/7200 [01:02<1:15:08, 1.59it/s]
0%| | 30/7200 [01:03<1:15:10, 1.59it/s]
{'loss': '10.23', 'grad_norm': '1.641', 'learning_rate': '0.000174', 'epoch': '0.4188'}
+
0%| | 30/7200 [01:03<1:15:10, 1.59it/s]
0%| | 31/7200 [01:04<1:15:08, 1.59it/s]
0%| | 32/7200 [01:04<1:15:05, 1.59it/s]
0%| | 33/7200 [01:05<1:15:05, 1.59it/s]
0%| | 34/7200 [01:06<1:15:05, 1.59it/s]
0%| | 35/7200 [01:06<1:15:02, 1.59it/s]
0%| | 36/7200 [01:07<1:15:03, 1.59it/s]
1%| | 37/7200 [01:07<1:15:02, 1.59it/s]
1%| | 38/7200 [01:08<1:15:02, 1.59it/s]
1%| | 39/7200 [01:09<1:15:01, 1.59it/s]
1%| | 40/7200 [01:09<1:14:59, 1.59it/s]
{'loss': '9.196', 'grad_norm': '1.481', 'learning_rate': '0.000234', 'epoch': '0.5585'}
+
1%| | 40/7200 [01:09<1:14:59, 1.59it/s]
1%| | 41/7200 [01:10<1:15:04, 1.59it/s]
1%| | 42/7200 [01:11<1:15:01, 1.59it/s]
1%| | 43/7200 [01:11<1:15:02, 1.59it/s]
1%| | 44/7200 [01:12<1:14:59, 1.59it/s]
1%| | 45/7200 [01:12<1:14:58, 1.59it/s]
1%| | 46/7200 [01:13<1:14:58, 1.59it/s]
1%| | 47/7200 [01:14<1:14:56, 1.59it/s]
1%| | 48/7200 [01:14<1:14:57, 1.59it/s]
1%| | 49/7200 [01:15<1:14:57, 1.59it/s]
1%| | 50/7200 [01:16<1:14:56, 1.59it/s]
{'loss': '8.214', 'grad_norm': '0.8694', 'learning_rate': '0.000294', 'epoch': '0.6981'}
+
1%| | 50/7200 [01:16<1:14:56, 1.59it/s]
1%| | 51/7200 [01:16<1:14:59, 1.59it/s]
1%| | 52/7200 [01:17<1:14:57, 1.59it/s]
1%| | 53/7200 [01:17<1:14:56, 1.59it/s]
1%| | 54/7200 [01:18<1:14:55, 1.59it/s]
1%| | 55/7200 [01:19<1:14:54, 1.59it/s]
1%| | 56/7200 [01:19<1:14:53, 1.59it/s]
1%| | 57/7200 [01:20<1:14:53, 1.59it/s]
1%| | 58/7200 [01:21<1:14:53, 1.59it/s]
1%| | 59/7200 [01:21<1:14:53, 1.59it/s]
1%| | 60/7200 [01:22<1:14:54, 1.59it/s]
{'loss': '7.702', 'grad_norm': '0.2659', 'learning_rate': '0.000354', 'epoch': '0.8377'}
+
1%| | 60/7200 [01:22<1:14:54, 1.59it/s]
1%| | 61/7200 [01:23<1:14:56, 1.59it/s]
1%| | 62/7200 [01:23<1:14:56, 1.59it/s]
1%| | 63/7200 [01:24<1:14:53, 1.59it/s]
1%| | 64/7200 [01:24<1:14:53, 1.59it/s]
1%| | 65/7200 [01:25<1:14:56, 1.59it/s]
1%| | 66/7200 [01:26<1:14:50, 1.59it/s]
1%| | 67/7200 [01:26<1:14:49, 1.59it/s]
1%| | 68/7200 [01:27<1:14:52, 1.59it/s]
1%| | 69/7200 [01:28<1:14:54, 1.59it/s]
1%| | 70/7200 [01:28<1:14:59, 1.58it/s]
{'loss': '7.56', 'grad_norm': '0.2413', 'learning_rate': '0.000414', 'epoch': '0.9773'}
+
1%| | 70/7200 [01:28<1:14:59, 1.58it/s]
1%| | 71/7200 [01:29<1:14:57, 1.59it/s]
1%| | 72/7200 [01:29<1:06:36, 1.78it/s][transformers]
+***** Running Evaluation *****
+[transformers] Num examples = 2441
+[transformers] Batch size = 8
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+
+
0%| | 0/39 [00:00system\n' }}
+ {%- if messages[0].role == 'system' %}
+ {{- messages[0].content + '\n\n' }}
+ {%- endif %}
+ {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within XML tags:\n" }}
+ {%- for tool in tools %}
+ {{- "\n" }}
+ {{- tool | tojson }}
+ {%- endfor %}
+ {{- "\n\n\nFor each function call, return a json object with function name and arguments within XML tags:\n\n{\"name\": , \"arguments\": }\n<|im_end|>\n" }}
+{%- else %}
+ {%- if messages[0].role == 'system' %}
+ {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+ {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+ {%- set index = (messages|length - 1) - loop.index0 %}
+ {%- if ns.multi_step_tool and message.role == "user" and not(message.content.startswith('') and message.content.endswith('')) %}
+ {%- set ns.multi_step_tool = false %}
+ {%- set ns.last_query_index = index %}
+ {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+ {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+ {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }}
+ {%- elif message.role == "assistant" %}
+ {%- set content = message.content %}
+ {%- set reasoning_content = '' %}
+ {%- if message.reasoning_content is defined and message.reasoning_content is not none %}
+ {%- set reasoning_content = message.reasoning_content %}
+ {%- else %}
+ {%- if '' in message.content %}
+ {%- set content = message.content.split('')[-1].lstrip('\n') %}
+ {%- set reasoning_content = message.content.split('')[0].rstrip('\n').split('')[-1].lstrip('\n') %}
+ {%- endif %}
+ {%- endif %}
+ {%- if loop.index0 > ns.last_query_index %}
+ {%- if loop.last or (not loop.last and reasoning_content) %}
+ {{- '<|im_start|>' + message.role + '\n\n' + reasoning_content.strip('\n') + '\n\n\n' + content.lstrip('\n') }}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- else %}
+ {{- '<|im_start|>' + message.role + '\n' + content }}
+ {%- endif %}
+ {%- if message.tool_calls %}
+ {%- for tool_call in message.tool_calls %}
+ {%- if (loop.first and content) or (not loop.first) %}
+ {{- '\n' }}
+ {%- endif %}
+ {%- if tool_call.function %}
+ {%- set tool_call = tool_call.function %}
+ {%- endif %}
+ {{- '\n{"name": "' }}
+ {{- tool_call.name }}
+ {{- '", "arguments": ' }}
+ {%- if tool_call.arguments is string %}
+ {{- tool_call.arguments }}
+ {%- else %}
+ {{- tool_call.arguments | tojson }}
+ {%- endif %}
+ {{- '}\n' }}
+ {%- endfor %}
+ {%- endif %}
+ {{- '<|im_end|>\n' }}
+ {%- elif message.role == "tool" %}
+ {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+ {{- '<|im_start|>user' }}
+ {%- endif %}
+ {{- '\n\n' }}
+ {{- message.content }}
+ {{- '\n' }}
+ {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+ {{- '<|im_end|>\n' }}
+ {%- endif %}
+ {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+ {{- '<|im_start|>assistant\n' }}
+ {%- if enable_thinking is defined and enable_thinking is false %}
+ {{- '\n\n\n\n' }}
+ {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/checkpoint-6552/config.json b/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/checkpoint-6552/config.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3474619f10931027097cc05d837296fa4f33a6c
--- /dev/null
+++ b/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/checkpoint-6552/config.json
@@ -0,0 +1,32 @@
+{
+ "architectures": [
+ "LlamaForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": null,
+ "dtype": "float32",
+ "eos_token_id": 151645,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 512,
+ "initializer_range": 0.02,
+ "intermediate_size": 1536,
+ "max_position_embeddings": 2048,
+ "mlp_bias": false,
+ "model_type": "llama",
+ "num_attention_heads": 4,
+ "num_hidden_layers": 20,
+ "num_key_value_heads": 4,
+ "pad_token_id": 151645,
+ "pretraining_tp": 1,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 10000.0,
+ "rope_type": "default"
+ },
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": false,
+ "vocab_size": 151676
+}
diff --git a/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/checkpoint-6552/generation_config.json b/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/checkpoint-6552/generation_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c4242f32d40df256857f87264f646130f684e33
--- /dev/null
+++ b/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/checkpoint-6552/generation_config.json
@@ -0,0 +1,11 @@
+{
+ "_from_model_config": true,
+ "eos_token_id": [
+ 151645
+ ],
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "transformers_version": "5.15.0",
+ "use_cache": true
+}
diff --git a/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/checkpoint-6552/tokenizer_config.json b/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/checkpoint-6552/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d43c97a447f4b25a8c36be5873f7b622935fa90
--- /dev/null
+++ b/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/checkpoint-6552/tokenizer_config.json
@@ -0,0 +1,24 @@
+{
+ "add_prefix_space": false,
+ "backend": "tokenizers",
+ "bos_token": null,
+ "clean_up_tokenization_spaces": false,
+ "eos_token": "<|im_end|>",
+ "errors": "replace",
+ "extra_special_tokens": [
+ "<|l2r_pred|>",
+ "<|r2l_pred|>",
+ "<|next_1_pred|>",
+ "<|next_2_pred|>",
+ "<|next_3_pred|>",
+ "<|next_4_pred|>",
+ "<|next_5_pred|>"
+ ],
+ "is_local": false,
+ "local_files_only": false,
+ "model_max_length": 131072,
+ "pad_token": "<|im_end|>",
+ "split_special_tokens": false,
+ "tokenizer_class": "Qwen2Tokenizer",
+ "unk_token": null
+}
diff --git a/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/checkpoint-6552/trainer_state.json b/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/checkpoint-6552/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..1a758de0218b6650cf7fae430ffd4f03daad821e
--- /dev/null
+++ b/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/checkpoint-6552/trainer_state.json
@@ -0,0 +1,5354 @@
+{
+ "best_global_step": null,
+ "best_metric": null,
+ "best_model_checkpoint": null,
+ "epoch": 91.0,
+ "eval_steps": 72,
+ "global_step": 6552,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.013961605584642234,
+ "grad_norm": 2.4882164001464844,
+ "learning_rate": 0.0,
+ "loss": 12.036874771118164,
+ "step": 1
+ },
+ {
+ "epoch": 0.13961605584642234,
+ "grad_norm": 1.6776295900344849,
+ "learning_rate": 5.399999999999999e-05,
+ "loss": 11.712064107259115,
+ "step": 10
+ },
+ {
+ "epoch": 0.2792321116928447,
+ "grad_norm": 1.600034475326538,
+ "learning_rate": 0.00011399999999999999,
+ "loss": 11.057333374023438,
+ "step": 20
+ },
+ {
+ "epoch": 0.418848167539267,
+ "grad_norm": 1.5412160158157349,
+ "learning_rate": 0.00017399999999999997,
+ "loss": 10.341563415527343,
+ "step": 30
+ },
+ {
+ "epoch": 0.5584642233856894,
+ "grad_norm": 1.3672738075256348,
+ "learning_rate": 0.000234,
+ "loss": 9.386703491210938,
+ "step": 40
+ },
+ {
+ "epoch": 0.6980802792321117,
+ "grad_norm": 0.8154869079589844,
+ "learning_rate": 0.000294,
+ "loss": 8.493916320800782,
+ "step": 50
+ },
+ {
+ "epoch": 0.837696335078534,
+ "grad_norm": 0.25931650400161743,
+ "learning_rate": 0.00035399999999999993,
+ "loss": 8.024773406982423,
+ "step": 60
+ },
+ {
+ "epoch": 0.9773123909249564,
+ "grad_norm": 0.2732985019683838,
+ "learning_rate": 0.0004139999999999999,
+ "loss": 7.90897216796875,
+ "step": 70
+ },
+ {
+ "epoch": 1.0,
+ "eval_loss": 7.521921157836914,
+ "eval_runtime": 26.7452,
+ "eval_samples_per_second": 91.269,
+ "eval_steps_per_second": 1.458,
+ "step": 72
+ },
+ {
+ "epoch": 1.111692844677138,
+ "grad_norm": 0.5093569159507751,
+ "learning_rate": 0.000474,
+ "loss": 7.841438293457031,
+ "step": 80
+ },
+ {
+ "epoch": 1.2513089005235603,
+ "grad_norm": 0.22970077395439148,
+ "learning_rate": 0.000534,
+ "loss": 7.795421600341797,
+ "step": 90
+ },
+ {
+ "epoch": 1.3909249563699826,
+ "grad_norm": 0.2142554670572281,
+ "learning_rate": 0.0005939999999999999,
+ "loss": 7.757545471191406,
+ "step": 100
+ },
+ {
+ "epoch": 1.530541012216405,
+ "grad_norm": 0.5944523811340332,
+ "learning_rate": 0.0006,
+ "loss": 7.742117309570313,
+ "step": 110
+ },
+ {
+ "epoch": 1.6701570680628273,
+ "grad_norm": 0.1773921251296997,
+ "learning_rate": 0.0006,
+ "loss": 7.706277465820312,
+ "step": 120
+ },
+ {
+ "epoch": 1.8097731239092496,
+ "grad_norm": 0.21996533870697021,
+ "learning_rate": 0.0006,
+ "loss": 7.663858795166016,
+ "step": 130
+ },
+ {
+ "epoch": 1.949389179755672,
+ "grad_norm": 0.5998210310935974,
+ "learning_rate": 0.0006,
+ "loss": 7.625840759277343,
+ "step": 140
+ },
+ {
+ "epoch": 2.0,
+ "eval_loss": 7.115081787109375,
+ "eval_runtime": 2.6329,
+ "eval_samples_per_second": 927.125,
+ "eval_steps_per_second": 14.813,
+ "step": 144
+ },
+ {
+ "epoch": 2.0837696335078535,
+ "grad_norm": 0.45933032035827637,
+ "learning_rate": 0.0006,
+ "loss": 7.588809204101563,
+ "step": 150
+ },
+ {
+ "epoch": 2.223385689354276,
+ "grad_norm": 0.3270089030265808,
+ "learning_rate": 0.0006,
+ "loss": 7.54593505859375,
+ "step": 160
+ },
+ {
+ "epoch": 2.363001745200698,
+ "grad_norm": 0.44116050004959106,
+ "learning_rate": 0.0006,
+ "loss": 7.550751495361328,
+ "step": 170
+ },
+ {
+ "epoch": 2.5026178010471205,
+ "grad_norm": 0.25038057565689087,
+ "learning_rate": 0.0006,
+ "loss": 7.484361267089843,
+ "step": 180
+ },
+ {
+ "epoch": 2.642233856893543,
+ "grad_norm": 0.5552337765693665,
+ "learning_rate": 0.0006,
+ "loss": 7.450635528564453,
+ "step": 190
+ },
+ {
+ "epoch": 2.781849912739965,
+ "grad_norm": 0.26204076409339905,
+ "learning_rate": 0.0006,
+ "loss": 7.347109985351563,
+ "step": 200
+ },
+ {
+ "epoch": 2.9214659685863875,
+ "grad_norm": 0.219796285033226,
+ "learning_rate": 0.0006,
+ "loss": 7.222745513916015,
+ "step": 210
+ },
+ {
+ "epoch": 3.0,
+ "eval_loss": 6.518674373626709,
+ "eval_runtime": 2.6308,
+ "eval_samples_per_second": 927.866,
+ "eval_steps_per_second": 14.825,
+ "step": 216
+ },
+ {
+ "epoch": 3.055846422338569,
+ "grad_norm": 0.28553327918052673,
+ "learning_rate": 0.0006,
+ "loss": 7.189297485351562,
+ "step": 220
+ },
+ {
+ "epoch": 3.195462478184991,
+ "grad_norm": 0.42079439759254456,
+ "learning_rate": 0.0006,
+ "loss": 7.103524780273437,
+ "step": 230
+ },
+ {
+ "epoch": 3.3350785340314135,
+ "grad_norm": 0.35625824332237244,
+ "learning_rate": 0.0006,
+ "loss": 7.043144226074219,
+ "step": 240
+ },
+ {
+ "epoch": 3.474694589877836,
+ "grad_norm": 0.6744439601898193,
+ "learning_rate": 0.0006,
+ "loss": 6.988848114013672,
+ "step": 250
+ },
+ {
+ "epoch": 3.614310645724258,
+ "grad_norm": 0.3082502782344818,
+ "learning_rate": 0.0006,
+ "loss": 6.954491424560547,
+ "step": 260
+ },
+ {
+ "epoch": 3.7539267015706805,
+ "grad_norm": 0.6940694451332092,
+ "learning_rate": 0.0006,
+ "loss": 6.865011596679688,
+ "step": 270
+ },
+ {
+ "epoch": 3.893542757417103,
+ "grad_norm": 0.25443607568740845,
+ "learning_rate": 0.0006,
+ "loss": 6.813713073730469,
+ "step": 280
+ },
+ {
+ "epoch": 4.0,
+ "eval_loss": 5.984549522399902,
+ "eval_runtime": 3.0189,
+ "eval_samples_per_second": 808.559,
+ "eval_steps_per_second": 12.918,
+ "step": 288
+ },
+ {
+ "epoch": 4.027923211169284,
+ "grad_norm": 0.48908087611198425,
+ "learning_rate": 0.0006,
+ "loss": 6.764627838134766,
+ "step": 290
+ },
+ {
+ "epoch": 4.167539267015707,
+ "grad_norm": 0.46522390842437744,
+ "learning_rate": 0.0006,
+ "loss": 6.725939178466797,
+ "step": 300
+ },
+ {
+ "epoch": 4.307155322862129,
+ "grad_norm": 0.22166909277439117,
+ "learning_rate": 0.0006,
+ "loss": 6.672442626953125,
+ "step": 310
+ },
+ {
+ "epoch": 4.446771378708552,
+ "grad_norm": 0.6852430701255798,
+ "learning_rate": 0.0006,
+ "loss": 6.647119140625,
+ "step": 320
+ },
+ {
+ "epoch": 4.5863874345549736,
+ "grad_norm": 0.3676527142524719,
+ "learning_rate": 0.0006,
+ "loss": 6.688330078125,
+ "step": 330
+ },
+ {
+ "epoch": 4.726003490401396,
+ "grad_norm": 0.2540346086025238,
+ "learning_rate": 0.0006,
+ "loss": 6.577582550048828,
+ "step": 340
+ },
+ {
+ "epoch": 4.865619546247818,
+ "grad_norm": 0.34375742077827454,
+ "learning_rate": 0.0006,
+ "loss": 6.527590942382813,
+ "step": 350
+ },
+ {
+ "epoch": 5.0,
+ "grad_norm": 0.41854578256607056,
+ "learning_rate": 0.0006,
+ "loss": 6.457218933105469,
+ "step": 360
+ },
+ {
+ "epoch": 5.0,
+ "eval_loss": 5.609179496765137,
+ "eval_runtime": 3.5733,
+ "eval_samples_per_second": 683.124,
+ "eval_steps_per_second": 10.914,
+ "step": 360
+ },
+ {
+ "epoch": 5.139616055846423,
+ "grad_norm": 0.47690168023109436,
+ "learning_rate": 0.0006,
+ "loss": 6.46545639038086,
+ "step": 370
+ },
+ {
+ "epoch": 5.279232111692845,
+ "grad_norm": 0.3670310378074646,
+ "learning_rate": 0.0006,
+ "loss": 6.414584350585938,
+ "step": 380
+ },
+ {
+ "epoch": 5.418848167539267,
+ "grad_norm": 0.22919140756130219,
+ "learning_rate": 0.0006,
+ "loss": 6.375809097290039,
+ "step": 390
+ },
+ {
+ "epoch": 5.558464223385689,
+ "grad_norm": 0.24118264019489288,
+ "learning_rate": 0.0006,
+ "loss": 6.33638916015625,
+ "step": 400
+ },
+ {
+ "epoch": 5.698080279232112,
+ "grad_norm": 0.4086466431617737,
+ "learning_rate": 0.0006,
+ "loss": 6.335563278198242,
+ "step": 410
+ },
+ {
+ "epoch": 5.837696335078534,
+ "grad_norm": 0.39880454540252686,
+ "learning_rate": 0.0006,
+ "loss": 6.2886604309082035,
+ "step": 420
+ },
+ {
+ "epoch": 5.977312390924956,
+ "grad_norm": 0.27013733983039856,
+ "learning_rate": 0.0006,
+ "loss": 6.262530517578125,
+ "step": 430
+ },
+ {
+ "epoch": 6.0,
+ "eval_loss": 5.305898189544678,
+ "eval_runtime": 4.4241,
+ "eval_samples_per_second": 551.746,
+ "eval_steps_per_second": 8.815,
+ "step": 432
+ },
+ {
+ "epoch": 6.111692844677138,
+ "grad_norm": 0.338518351316452,
+ "learning_rate": 0.0006,
+ "loss": 6.173490142822265,
+ "step": 440
+ },
+ {
+ "epoch": 6.2513089005235605,
+ "grad_norm": 0.4229814112186432,
+ "learning_rate": 0.0006,
+ "loss": 6.212611389160156,
+ "step": 450
+ },
+ {
+ "epoch": 6.390924956369982,
+ "grad_norm": 0.36979153752326965,
+ "learning_rate": 0.0006,
+ "loss": 6.162299346923828,
+ "step": 460
+ },
+ {
+ "epoch": 6.530541012216405,
+ "grad_norm": 0.2768906056880951,
+ "learning_rate": 0.0006,
+ "loss": 6.132846832275391,
+ "step": 470
+ },
+ {
+ "epoch": 6.670157068062827,
+ "grad_norm": 0.3919268846511841,
+ "learning_rate": 0.0006,
+ "loss": 6.097616577148438,
+ "step": 480
+ },
+ {
+ "epoch": 6.80977312390925,
+ "grad_norm": 0.44298413395881653,
+ "learning_rate": 0.0006,
+ "loss": 6.055324554443359,
+ "step": 490
+ },
+ {
+ "epoch": 6.949389179755672,
+ "grad_norm": 0.3626202642917633,
+ "learning_rate": 0.0006,
+ "loss": 6.069403076171875,
+ "step": 500
+ },
+ {
+ "epoch": 7.0,
+ "eval_loss": 5.03093147277832,
+ "eval_runtime": 2.8094,
+ "eval_samples_per_second": 868.878,
+ "eval_steps_per_second": 13.882,
+ "step": 504
+ },
+ {
+ "epoch": 7.0837696335078535,
+ "grad_norm": 0.26268091797828674,
+ "learning_rate": 0.0006,
+ "loss": 5.999639892578125,
+ "step": 510
+ },
+ {
+ "epoch": 7.223385689354275,
+ "grad_norm": 0.33666378259658813,
+ "learning_rate": 0.0006,
+ "loss": 6.000067138671875,
+ "step": 520
+ },
+ {
+ "epoch": 7.363001745200698,
+ "grad_norm": 0.44332125782966614,
+ "learning_rate": 0.0006,
+ "loss": 5.925631713867188,
+ "step": 530
+ },
+ {
+ "epoch": 7.50261780104712,
+ "grad_norm": 0.36089223623275757,
+ "learning_rate": 0.0006,
+ "loss": 5.9203338623046875,
+ "step": 540
+ },
+ {
+ "epoch": 7.642233856893543,
+ "grad_norm": 0.25677165389060974,
+ "learning_rate": 0.0006,
+ "loss": 5.8987586975097654,
+ "step": 550
+ },
+ {
+ "epoch": 7.781849912739965,
+ "grad_norm": 0.2519516050815582,
+ "learning_rate": 0.0006,
+ "loss": 5.847074127197265,
+ "step": 560
+ },
+ {
+ "epoch": 7.9214659685863875,
+ "grad_norm": 0.3302984833717346,
+ "learning_rate": 0.0006,
+ "loss": 5.798227691650391,
+ "step": 570
+ },
+ {
+ "epoch": 8.0,
+ "eval_loss": 4.789069652557373,
+ "eval_runtime": 3.6873,
+ "eval_samples_per_second": 662.01,
+ "eval_steps_per_second": 10.577,
+ "step": 576
+ },
+ {
+ "epoch": 8.055846422338568,
+ "grad_norm": 0.39834606647491455,
+ "learning_rate": 0.0006,
+ "loss": 5.775135040283203,
+ "step": 580
+ },
+ {
+ "epoch": 8.195462478184991,
+ "grad_norm": 0.29033949971199036,
+ "learning_rate": 0.0006,
+ "loss": 5.789869689941407,
+ "step": 590
+ },
+ {
+ "epoch": 8.335078534031414,
+ "grad_norm": 0.3337038457393646,
+ "learning_rate": 0.0006,
+ "loss": 5.746315002441406,
+ "step": 600
+ },
+ {
+ "epoch": 8.474694589877837,
+ "grad_norm": 2.489588499069214,
+ "learning_rate": 0.0006,
+ "loss": 5.807655715942383,
+ "step": 610
+ },
+ {
+ "epoch": 8.614310645724258,
+ "grad_norm": 1.0617883205413818,
+ "learning_rate": 0.0006,
+ "loss": 6.33654556274414,
+ "step": 620
+ },
+ {
+ "epoch": 8.75392670157068,
+ "grad_norm": 0.595336377620697,
+ "learning_rate": 0.0006,
+ "loss": 6.181840515136718,
+ "step": 630
+ },
+ {
+ "epoch": 8.893542757417103,
+ "grad_norm": 0.21573497354984283,
+ "learning_rate": 0.0006,
+ "loss": 5.997751235961914,
+ "step": 640
+ },
+ {
+ "epoch": 9.0,
+ "eval_loss": 4.802168846130371,
+ "eval_runtime": 2.6452,
+ "eval_samples_per_second": 922.799,
+ "eval_steps_per_second": 14.744,
+ "step": 648
+ },
+ {
+ "epoch": 9.027923211169284,
+ "grad_norm": 0.1687227040529251,
+ "learning_rate": 0.0006,
+ "loss": 5.860659790039063,
+ "step": 650
+ },
+ {
+ "epoch": 9.167539267015707,
+ "grad_norm": 0.13544581830501556,
+ "learning_rate": 0.0006,
+ "loss": 5.765312194824219,
+ "step": 660
+ },
+ {
+ "epoch": 9.30715532286213,
+ "grad_norm": 0.13602802157402039,
+ "learning_rate": 0.0006,
+ "loss": 5.754157257080078,
+ "step": 670
+ },
+ {
+ "epoch": 9.44677137870855,
+ "grad_norm": 0.14676471054553986,
+ "learning_rate": 0.0006,
+ "loss": 5.67625846862793,
+ "step": 680
+ },
+ {
+ "epoch": 9.586387434554974,
+ "grad_norm": 0.17356885969638824,
+ "learning_rate": 0.0006,
+ "loss": 5.70880012512207,
+ "step": 690
+ },
+ {
+ "epoch": 9.726003490401396,
+ "grad_norm": 0.1571578085422516,
+ "learning_rate": 0.0006,
+ "loss": 5.662619781494141,
+ "step": 700
+ },
+ {
+ "epoch": 9.86561954624782,
+ "grad_norm": 0.17748786509037018,
+ "learning_rate": 0.0006,
+ "loss": 5.6423591613769535,
+ "step": 710
+ },
+ {
+ "epoch": 10.0,
+ "grad_norm": 0.21707725524902344,
+ "learning_rate": 0.0006,
+ "loss": 5.621284103393554,
+ "step": 720
+ },
+ {
+ "epoch": 10.0,
+ "eval_loss": 4.573047161102295,
+ "eval_runtime": 2.6279,
+ "eval_samples_per_second": 928.89,
+ "eval_steps_per_second": 14.841,
+ "step": 720
+ },
+ {
+ "epoch": 10.139616055846423,
+ "grad_norm": 0.2633434534072876,
+ "learning_rate": 0.0006,
+ "loss": 5.589619827270508,
+ "step": 730
+ },
+ {
+ "epoch": 10.279232111692846,
+ "grad_norm": 0.2043304294347763,
+ "learning_rate": 0.0006,
+ "loss": 5.5800941467285154,
+ "step": 740
+ },
+ {
+ "epoch": 10.418848167539267,
+ "grad_norm": 0.2554124891757965,
+ "learning_rate": 0.0006,
+ "loss": 5.568388366699219,
+ "step": 750
+ },
+ {
+ "epoch": 10.55846422338569,
+ "grad_norm": 0.22094231843948364,
+ "learning_rate": 0.0006,
+ "loss": 5.5551597595214846,
+ "step": 760
+ },
+ {
+ "epoch": 10.698080279232112,
+ "grad_norm": 0.22957001626491547,
+ "learning_rate": 0.0006,
+ "loss": 5.591373443603516,
+ "step": 770
+ },
+ {
+ "epoch": 10.837696335078533,
+ "grad_norm": 0.205167755484581,
+ "learning_rate": 0.0006,
+ "loss": 5.579811096191406,
+ "step": 780
+ },
+ {
+ "epoch": 10.977312390924956,
+ "grad_norm": 0.20447632670402527,
+ "learning_rate": 0.0006,
+ "loss": 5.536233901977539,
+ "step": 790
+ },
+ {
+ "epoch": 11.0,
+ "eval_loss": 4.474292278289795,
+ "eval_runtime": 2.6448,
+ "eval_samples_per_second": 922.959,
+ "eval_steps_per_second": 14.746,
+ "step": 792
+ },
+ {
+ "epoch": 11.111692844677139,
+ "grad_norm": 0.2230602353811264,
+ "learning_rate": 0.0006,
+ "loss": 5.49072036743164,
+ "step": 800
+ },
+ {
+ "epoch": 11.25130890052356,
+ "grad_norm": 0.23244008421897888,
+ "learning_rate": 0.0006,
+ "loss": 5.46531867980957,
+ "step": 810
+ },
+ {
+ "epoch": 11.390924956369982,
+ "grad_norm": 0.3002164959907532,
+ "learning_rate": 0.0006,
+ "loss": 5.469110107421875,
+ "step": 820
+ },
+ {
+ "epoch": 11.530541012216405,
+ "grad_norm": 0.1926560401916504,
+ "learning_rate": 0.0006,
+ "loss": 5.48816146850586,
+ "step": 830
+ },
+ {
+ "epoch": 11.670157068062828,
+ "grad_norm": 0.19971534609794617,
+ "learning_rate": 0.0006,
+ "loss": 5.454804992675781,
+ "step": 840
+ },
+ {
+ "epoch": 11.809773123909249,
+ "grad_norm": 0.24356654286384583,
+ "learning_rate": 0.0006,
+ "loss": 5.5233512878417965,
+ "step": 850
+ },
+ {
+ "epoch": 11.949389179755672,
+ "grad_norm": 0.18953797221183777,
+ "learning_rate": 0.0006,
+ "loss": 5.440174102783203,
+ "step": 860
+ },
+ {
+ "epoch": 12.0,
+ "eval_loss": 4.403199195861816,
+ "eval_runtime": 2.6271,
+ "eval_samples_per_second": 929.179,
+ "eval_steps_per_second": 14.846,
+ "step": 864
+ },
+ {
+ "epoch": 12.083769633507853,
+ "grad_norm": 0.26803192496299744,
+ "learning_rate": 0.0006,
+ "loss": 5.464957046508789,
+ "step": 870
+ },
+ {
+ "epoch": 12.223385689354275,
+ "grad_norm": 0.2143821120262146,
+ "learning_rate": 0.0006,
+ "loss": 5.455624389648437,
+ "step": 880
+ },
+ {
+ "epoch": 12.363001745200698,
+ "grad_norm": 0.25315165519714355,
+ "learning_rate": 0.0006,
+ "loss": 5.452237319946289,
+ "step": 890
+ },
+ {
+ "epoch": 12.502617801047121,
+ "grad_norm": 0.239376962184906,
+ "learning_rate": 0.0006,
+ "loss": 5.402100372314453,
+ "step": 900
+ },
+ {
+ "epoch": 12.642233856893542,
+ "grad_norm": 0.20413321256637573,
+ "learning_rate": 0.0006,
+ "loss": 5.4060310363769535,
+ "step": 910
+ },
+ {
+ "epoch": 12.781849912739965,
+ "grad_norm": 0.20323936641216278,
+ "learning_rate": 0.0006,
+ "loss": 5.429556274414063,
+ "step": 920
+ },
+ {
+ "epoch": 12.921465968586388,
+ "grad_norm": 0.2838175892829895,
+ "learning_rate": 0.0006,
+ "loss": 5.413117218017578,
+ "step": 930
+ },
+ {
+ "epoch": 13.0,
+ "eval_loss": 4.338420391082764,
+ "eval_runtime": 2.6292,
+ "eval_samples_per_second": 928.437,
+ "eval_steps_per_second": 14.834,
+ "step": 936
+ },
+ {
+ "epoch": 13.055846422338568,
+ "grad_norm": 0.260672390460968,
+ "learning_rate": 0.0006,
+ "loss": 5.363108062744141,
+ "step": 940
+ },
+ {
+ "epoch": 13.195462478184991,
+ "grad_norm": 0.2868488132953644,
+ "learning_rate": 0.0006,
+ "loss": 5.353092193603516,
+ "step": 950
+ },
+ {
+ "epoch": 13.335078534031414,
+ "grad_norm": 0.25650185346603394,
+ "learning_rate": 0.0006,
+ "loss": 5.392884826660156,
+ "step": 960
+ },
+ {
+ "epoch": 13.474694589877837,
+ "grad_norm": 0.22798645496368408,
+ "learning_rate": 0.0006,
+ "loss": 5.364498138427734,
+ "step": 970
+ },
+ {
+ "epoch": 13.614310645724258,
+ "grad_norm": 0.2200073003768921,
+ "learning_rate": 0.0006,
+ "loss": 5.345271301269531,
+ "step": 980
+ },
+ {
+ "epoch": 13.75392670157068,
+ "grad_norm": 0.22006338834762573,
+ "learning_rate": 0.0006,
+ "loss": 5.370860290527344,
+ "step": 990
+ },
+ {
+ "epoch": 13.893542757417103,
+ "grad_norm": 0.22214198112487793,
+ "learning_rate": 0.0006,
+ "loss": 5.3526771545410154,
+ "step": 1000
+ },
+ {
+ "epoch": 14.0,
+ "eval_loss": 4.27860689163208,
+ "eval_runtime": 2.6447,
+ "eval_samples_per_second": 922.97,
+ "eval_steps_per_second": 14.746,
+ "step": 1008
+ },
+ {
+ "epoch": 14.027923211169284,
+ "grad_norm": 0.2005612701177597,
+ "learning_rate": 0.0006,
+ "loss": 5.336740112304687,
+ "step": 1010
+ },
+ {
+ "epoch": 14.167539267015707,
+ "grad_norm": 0.3388651907444,
+ "learning_rate": 0.0006,
+ "loss": 5.382823181152344,
+ "step": 1020
+ },
+ {
+ "epoch": 14.30715532286213,
+ "grad_norm": 0.2352493852376938,
+ "learning_rate": 0.0006,
+ "loss": 5.344551849365234,
+ "step": 1030
+ },
+ {
+ "epoch": 14.44677137870855,
+ "grad_norm": 0.256282776594162,
+ "learning_rate": 0.0006,
+ "loss": 5.339216613769532,
+ "step": 1040
+ },
+ {
+ "epoch": 14.586387434554974,
+ "grad_norm": 0.1897788941860199,
+ "learning_rate": 0.0006,
+ "loss": 5.299592590332031,
+ "step": 1050
+ },
+ {
+ "epoch": 14.726003490401396,
+ "grad_norm": 0.2235531359910965,
+ "learning_rate": 0.0006,
+ "loss": 5.294806671142578,
+ "step": 1060
+ },
+ {
+ "epoch": 14.86561954624782,
+ "grad_norm": 0.21937283873558044,
+ "learning_rate": 0.0006,
+ "loss": 5.328453826904297,
+ "step": 1070
+ },
+ {
+ "epoch": 15.0,
+ "grad_norm": 0.256804496049881,
+ "learning_rate": 0.0006,
+ "loss": 5.346640777587891,
+ "step": 1080
+ },
+ {
+ "epoch": 15.0,
+ "eval_loss": 4.231722831726074,
+ "eval_runtime": 2.629,
+ "eval_samples_per_second": 928.501,
+ "eval_steps_per_second": 14.835,
+ "step": 1080
+ },
+ {
+ "epoch": 15.139616055846423,
+ "grad_norm": 0.288019597530365,
+ "learning_rate": 0.0006,
+ "loss": 5.282761764526367,
+ "step": 1090
+ },
+ {
+ "epoch": 15.279232111692846,
+ "grad_norm": 0.23798708617687225,
+ "learning_rate": 0.0006,
+ "loss": 5.24670639038086,
+ "step": 1100
+ },
+ {
+ "epoch": 15.418848167539267,
+ "grad_norm": 0.23103944957256317,
+ "learning_rate": 0.0006,
+ "loss": 5.311689758300782,
+ "step": 1110
+ },
+ {
+ "epoch": 15.55846422338569,
+ "grad_norm": 0.2084844708442688,
+ "learning_rate": 0.0006,
+ "loss": 5.265097045898438,
+ "step": 1120
+ },
+ {
+ "epoch": 15.698080279232112,
+ "grad_norm": 0.20497052371501923,
+ "learning_rate": 0.0006,
+ "loss": 5.321389770507812,
+ "step": 1130
+ },
+ {
+ "epoch": 15.837696335078533,
+ "grad_norm": 0.22159650921821594,
+ "learning_rate": 0.0006,
+ "loss": 5.222319030761719,
+ "step": 1140
+ },
+ {
+ "epoch": 15.977312390924956,
+ "grad_norm": 0.27039533853530884,
+ "learning_rate": 0.0006,
+ "loss": 5.222488784790039,
+ "step": 1150
+ },
+ {
+ "epoch": 16.0,
+ "eval_loss": 4.197320938110352,
+ "eval_runtime": 2.631,
+ "eval_samples_per_second": 927.792,
+ "eval_steps_per_second": 14.823,
+ "step": 1152
+ },
+ {
+ "epoch": 16.111692844677137,
+ "grad_norm": 0.238840714097023,
+ "learning_rate": 0.0006,
+ "loss": 5.225128555297852,
+ "step": 1160
+ },
+ {
+ "epoch": 16.25130890052356,
+ "grad_norm": 0.20235341787338257,
+ "learning_rate": 0.0006,
+ "loss": 5.273032379150391,
+ "step": 1170
+ },
+ {
+ "epoch": 16.390924956369982,
+ "grad_norm": 0.21497970819473267,
+ "learning_rate": 0.0006,
+ "loss": 5.272190093994141,
+ "step": 1180
+ },
+ {
+ "epoch": 16.530541012216403,
+ "grad_norm": 0.2179366946220398,
+ "learning_rate": 0.0006,
+ "loss": 5.232486343383789,
+ "step": 1190
+ },
+ {
+ "epoch": 16.670157068062828,
+ "grad_norm": 0.2536570727825165,
+ "learning_rate": 0.0006,
+ "loss": 5.222574996948242,
+ "step": 1200
+ },
+ {
+ "epoch": 16.80977312390925,
+ "grad_norm": 0.2684154808521271,
+ "learning_rate": 0.0006,
+ "loss": 5.242881011962891,
+ "step": 1210
+ },
+ {
+ "epoch": 16.949389179755673,
+ "grad_norm": 0.23101234436035156,
+ "learning_rate": 0.0006,
+ "loss": 5.193366241455078,
+ "step": 1220
+ },
+ {
+ "epoch": 17.0,
+ "eval_loss": 4.160302639007568,
+ "eval_runtime": 2.6454,
+ "eval_samples_per_second": 922.721,
+ "eval_steps_per_second": 14.742,
+ "step": 1224
+ },
+ {
+ "epoch": 17.083769633507853,
+ "grad_norm": 0.24604123830795288,
+ "learning_rate": 0.0006,
+ "loss": 5.168920516967773,
+ "step": 1230
+ },
+ {
+ "epoch": 17.223385689354277,
+ "grad_norm": 0.20689134299755096,
+ "learning_rate": 0.0006,
+ "loss": 5.190824127197265,
+ "step": 1240
+ },
+ {
+ "epoch": 17.363001745200698,
+ "grad_norm": 0.23793166875839233,
+ "learning_rate": 0.0006,
+ "loss": 5.154198837280274,
+ "step": 1250
+ },
+ {
+ "epoch": 17.50261780104712,
+ "grad_norm": 0.24281951785087585,
+ "learning_rate": 0.0006,
+ "loss": 5.153511047363281,
+ "step": 1260
+ },
+ {
+ "epoch": 17.642233856893544,
+ "grad_norm": 0.2628422677516937,
+ "learning_rate": 0.0006,
+ "loss": 5.219535827636719,
+ "step": 1270
+ },
+ {
+ "epoch": 17.781849912739965,
+ "grad_norm": 0.20748260617256165,
+ "learning_rate": 0.0006,
+ "loss": 5.2032112121582035,
+ "step": 1280
+ },
+ {
+ "epoch": 17.921465968586386,
+ "grad_norm": 0.2066967934370041,
+ "learning_rate": 0.0006,
+ "loss": 5.217464447021484,
+ "step": 1290
+ },
+ {
+ "epoch": 18.0,
+ "eval_loss": 4.1286940574646,
+ "eval_runtime": 2.646,
+ "eval_samples_per_second": 922.518,
+ "eval_steps_per_second": 14.739,
+ "step": 1296
+ },
+ {
+ "epoch": 18.05584642233857,
+ "grad_norm": 0.20843403041362762,
+ "learning_rate": 0.0006,
+ "loss": 5.18970832824707,
+ "step": 1300
+ },
+ {
+ "epoch": 18.195462478184993,
+ "grad_norm": 0.19469091296195984,
+ "learning_rate": 0.0006,
+ "loss": 5.18148193359375,
+ "step": 1310
+ },
+ {
+ "epoch": 18.335078534031414,
+ "grad_norm": 0.24746035039424896,
+ "learning_rate": 0.0006,
+ "loss": 5.152855682373047,
+ "step": 1320
+ },
+ {
+ "epoch": 18.474694589877835,
+ "grad_norm": 0.28262102603912354,
+ "learning_rate": 0.0006,
+ "loss": 5.1487586975097654,
+ "step": 1330
+ },
+ {
+ "epoch": 18.61431064572426,
+ "grad_norm": 0.25076258182525635,
+ "learning_rate": 0.0006,
+ "loss": 5.137223815917968,
+ "step": 1340
+ },
+ {
+ "epoch": 18.75392670157068,
+ "grad_norm": 0.2132120132446289,
+ "learning_rate": 0.0006,
+ "loss": 5.164315032958984,
+ "step": 1350
+ },
+ {
+ "epoch": 18.8935427574171,
+ "grad_norm": 0.20220761001110077,
+ "learning_rate": 0.0006,
+ "loss": 5.187496948242187,
+ "step": 1360
+ },
+ {
+ "epoch": 19.0,
+ "eval_loss": 4.094858169555664,
+ "eval_runtime": 2.6328,
+ "eval_samples_per_second": 927.153,
+ "eval_steps_per_second": 14.813,
+ "step": 1368
+ },
+ {
+ "epoch": 19.027923211169284,
+ "grad_norm": 0.2218279391527176,
+ "learning_rate": 0.0006,
+ "loss": 5.1239776611328125,
+ "step": 1370
+ },
+ {
+ "epoch": 19.167539267015705,
+ "grad_norm": 0.2772565484046936,
+ "learning_rate": 0.0006,
+ "loss": 5.137679672241211,
+ "step": 1380
+ },
+ {
+ "epoch": 19.30715532286213,
+ "grad_norm": 0.23273946344852448,
+ "learning_rate": 0.0006,
+ "loss": 5.0904380798339846,
+ "step": 1390
+ },
+ {
+ "epoch": 19.44677137870855,
+ "grad_norm": 0.22909092903137207,
+ "learning_rate": 0.0006,
+ "loss": 5.213278961181641,
+ "step": 1400
+ },
+ {
+ "epoch": 19.586387434554975,
+ "grad_norm": 0.238947793841362,
+ "learning_rate": 0.0006,
+ "loss": 5.159337997436523,
+ "step": 1410
+ },
+ {
+ "epoch": 19.726003490401396,
+ "grad_norm": 0.2493923306465149,
+ "learning_rate": 0.0006,
+ "loss": 5.145233154296875,
+ "step": 1420
+ },
+ {
+ "epoch": 19.865619546247817,
+ "grad_norm": 0.1983357071876526,
+ "learning_rate": 0.0006,
+ "loss": 5.16369400024414,
+ "step": 1430
+ },
+ {
+ "epoch": 20.0,
+ "grad_norm": 0.25833871960639954,
+ "learning_rate": 0.0006,
+ "loss": 5.1294403076171875,
+ "step": 1440
+ },
+ {
+ "epoch": 20.0,
+ "eval_loss": 4.07643461227417,
+ "eval_runtime": 2.6421,
+ "eval_samples_per_second": 923.895,
+ "eval_steps_per_second": 14.761,
+ "step": 1440
+ },
+ {
+ "epoch": 20.13961605584642,
+ "grad_norm": 0.22790855169296265,
+ "learning_rate": 0.0006,
+ "loss": 5.157700729370117,
+ "step": 1450
+ },
+ {
+ "epoch": 20.279232111692846,
+ "grad_norm": 0.24981844425201416,
+ "learning_rate": 0.0006,
+ "loss": 5.0888629913330075,
+ "step": 1460
+ },
+ {
+ "epoch": 20.418848167539267,
+ "grad_norm": 0.24376538395881653,
+ "learning_rate": 0.0006,
+ "loss": 5.083840942382812,
+ "step": 1470
+ },
+ {
+ "epoch": 20.55846422338569,
+ "grad_norm": 0.2585503160953522,
+ "learning_rate": 0.0006,
+ "loss": 5.089398574829102,
+ "step": 1480
+ },
+ {
+ "epoch": 20.698080279232112,
+ "grad_norm": 0.21763764321804047,
+ "learning_rate": 0.0006,
+ "loss": 5.101925659179687,
+ "step": 1490
+ },
+ {
+ "epoch": 20.837696335078533,
+ "grad_norm": 0.22620759904384613,
+ "learning_rate": 0.0006,
+ "loss": 5.117087554931641,
+ "step": 1500
+ },
+ {
+ "epoch": 20.977312390924958,
+ "grad_norm": 0.19539913535118103,
+ "learning_rate": 0.0006,
+ "loss": 5.101424789428711,
+ "step": 1510
+ },
+ {
+ "epoch": 21.0,
+ "eval_loss": 4.052947998046875,
+ "eval_runtime": 2.6292,
+ "eval_samples_per_second": 928.416,
+ "eval_steps_per_second": 14.833,
+ "step": 1512
+ },
+ {
+ "epoch": 21.111692844677137,
+ "grad_norm": 0.28258487582206726,
+ "learning_rate": 0.0006,
+ "loss": 5.0637969970703125,
+ "step": 1520
+ },
+ {
+ "epoch": 21.25130890052356,
+ "grad_norm": 0.2839506268501282,
+ "learning_rate": 0.0006,
+ "loss": 5.060111999511719,
+ "step": 1530
+ },
+ {
+ "epoch": 21.390924956369982,
+ "grad_norm": 0.2183990776538849,
+ "learning_rate": 0.0006,
+ "loss": 5.083434295654297,
+ "step": 1540
+ },
+ {
+ "epoch": 21.530541012216403,
+ "grad_norm": 0.23024298250675201,
+ "learning_rate": 0.0006,
+ "loss": 5.043057250976562,
+ "step": 1550
+ },
+ {
+ "epoch": 21.670157068062828,
+ "grad_norm": 0.23482301831245422,
+ "learning_rate": 0.0006,
+ "loss": 5.11565933227539,
+ "step": 1560
+ },
+ {
+ "epoch": 21.80977312390925,
+ "grad_norm": 0.21137303113937378,
+ "learning_rate": 0.0006,
+ "loss": 5.15265007019043,
+ "step": 1570
+ },
+ {
+ "epoch": 21.949389179755673,
+ "grad_norm": 0.261631041765213,
+ "learning_rate": 0.0006,
+ "loss": 5.08914794921875,
+ "step": 1580
+ },
+ {
+ "epoch": 22.0,
+ "eval_loss": 4.037649154663086,
+ "eval_runtime": 2.6317,
+ "eval_samples_per_second": 927.533,
+ "eval_steps_per_second": 14.819,
+ "step": 1584
+ },
+ {
+ "epoch": 22.083769633507853,
+ "grad_norm": 0.27435752749443054,
+ "learning_rate": 0.0006,
+ "loss": 5.111014938354492,
+ "step": 1590
+ },
+ {
+ "epoch": 22.223385689354277,
+ "grad_norm": 3.312088966369629,
+ "learning_rate": 0.0006,
+ "loss": 5.088009643554687,
+ "step": 1600
+ },
+ {
+ "epoch": 22.363001745200698,
+ "grad_norm": 1.054248571395874,
+ "learning_rate": 0.0006,
+ "loss": 5.481085205078125,
+ "step": 1610
+ },
+ {
+ "epoch": 22.50261780104712,
+ "grad_norm": 1.023116111755371,
+ "learning_rate": 0.0006,
+ "loss": 5.609699249267578,
+ "step": 1620
+ },
+ {
+ "epoch": 22.642233856893544,
+ "grad_norm": 0.28160881996154785,
+ "learning_rate": 0.0006,
+ "loss": 5.532663726806641,
+ "step": 1630
+ },
+ {
+ "epoch": 22.781849912739965,
+ "grad_norm": 0.1903880536556244,
+ "learning_rate": 0.0006,
+ "loss": 5.424597549438476,
+ "step": 1640
+ },
+ {
+ "epoch": 22.921465968586386,
+ "grad_norm": 0.1674928218126297,
+ "learning_rate": 0.0006,
+ "loss": 5.31115951538086,
+ "step": 1650
+ },
+ {
+ "epoch": 23.0,
+ "eval_loss": 4.148776054382324,
+ "eval_runtime": 2.6348,
+ "eval_samples_per_second": 926.436,
+ "eval_steps_per_second": 14.802,
+ "step": 1656
+ },
+ {
+ "epoch": 23.05584642233857,
+ "grad_norm": 0.16569462418556213,
+ "learning_rate": 0.0006,
+ "loss": 5.2249397277832035,
+ "step": 1660
+ },
+ {
+ "epoch": 23.195462478184993,
+ "grad_norm": 0.17030368745326996,
+ "learning_rate": 0.0006,
+ "loss": 5.133450317382812,
+ "step": 1670
+ },
+ {
+ "epoch": 23.335078534031414,
+ "grad_norm": 0.14305879175662994,
+ "learning_rate": 0.0006,
+ "loss": 5.167530822753906,
+ "step": 1680
+ },
+ {
+ "epoch": 23.474694589877835,
+ "grad_norm": 0.13189075887203217,
+ "learning_rate": 0.0006,
+ "loss": 5.109375,
+ "step": 1690
+ },
+ {
+ "epoch": 23.61431064572426,
+ "grad_norm": 0.13836157321929932,
+ "learning_rate": 0.0006,
+ "loss": 5.059216690063477,
+ "step": 1700
+ },
+ {
+ "epoch": 23.75392670157068,
+ "grad_norm": 0.14915993809700012,
+ "learning_rate": 0.0006,
+ "loss": 5.095584106445313,
+ "step": 1710
+ },
+ {
+ "epoch": 23.8935427574171,
+ "grad_norm": 0.1517930030822754,
+ "learning_rate": 0.0006,
+ "loss": 5.058177947998047,
+ "step": 1720
+ },
+ {
+ "epoch": 24.0,
+ "eval_loss": 4.040567398071289,
+ "eval_runtime": 2.6464,
+ "eval_samples_per_second": 922.383,
+ "eval_steps_per_second": 14.737,
+ "step": 1728
+ },
+ {
+ "epoch": 24.027923211169284,
+ "grad_norm": 0.1622583568096161,
+ "learning_rate": 0.0006,
+ "loss": 5.135578918457031,
+ "step": 1730
+ },
+ {
+ "epoch": 24.167539267015705,
+ "grad_norm": 0.16430628299713135,
+ "learning_rate": 0.0006,
+ "loss": 5.094981384277344,
+ "step": 1740
+ },
+ {
+ "epoch": 24.30715532286213,
+ "grad_norm": 0.17117135226726532,
+ "learning_rate": 0.0006,
+ "loss": 5.083906555175782,
+ "step": 1750
+ },
+ {
+ "epoch": 24.44677137870855,
+ "grad_norm": 0.1558687388896942,
+ "learning_rate": 0.0006,
+ "loss": 5.08638916015625,
+ "step": 1760
+ },
+ {
+ "epoch": 24.586387434554975,
+ "grad_norm": 0.16037550568580627,
+ "learning_rate": 0.0006,
+ "loss": 5.095936965942383,
+ "step": 1770
+ },
+ {
+ "epoch": 24.726003490401396,
+ "grad_norm": 0.17815271019935608,
+ "learning_rate": 0.0006,
+ "loss": 5.09428596496582,
+ "step": 1780
+ },
+ {
+ "epoch": 24.865619546247817,
+ "grad_norm": 0.1615881621837616,
+ "learning_rate": 0.0006,
+ "loss": 5.10687026977539,
+ "step": 1790
+ },
+ {
+ "epoch": 25.0,
+ "grad_norm": 0.19104330241680145,
+ "learning_rate": 0.0006,
+ "loss": 5.078309631347656,
+ "step": 1800
+ },
+ {
+ "epoch": 25.0,
+ "eval_loss": 4.012660026550293,
+ "eval_runtime": 2.6305,
+ "eval_samples_per_second": 927.973,
+ "eval_steps_per_second": 14.826,
+ "step": 1800
+ },
+ {
+ "epoch": 25.13961605584642,
+ "grad_norm": 0.16317501664161682,
+ "learning_rate": 0.0006,
+ "loss": 5.0187629699707035,
+ "step": 1810
+ },
+ {
+ "epoch": 25.279232111692846,
+ "grad_norm": 0.17901495099067688,
+ "learning_rate": 0.0006,
+ "loss": 5.041153335571289,
+ "step": 1820
+ },
+ {
+ "epoch": 25.418848167539267,
+ "grad_norm": 0.16743682324886322,
+ "learning_rate": 0.0006,
+ "loss": 4.966635894775391,
+ "step": 1830
+ },
+ {
+ "epoch": 25.55846422338569,
+ "grad_norm": 0.18324463069438934,
+ "learning_rate": 0.0006,
+ "loss": 4.9602313995361325,
+ "step": 1840
+ },
+ {
+ "epoch": 25.698080279232112,
+ "grad_norm": 0.15934979915618896,
+ "learning_rate": 0.0006,
+ "loss": 5.030778121948242,
+ "step": 1850
+ },
+ {
+ "epoch": 25.837696335078533,
+ "grad_norm": 0.16859354078769684,
+ "learning_rate": 0.0006,
+ "loss": 4.965367126464844,
+ "step": 1860
+ },
+ {
+ "epoch": 25.977312390924958,
+ "grad_norm": 0.18080615997314453,
+ "learning_rate": 0.0006,
+ "loss": 5.056902313232422,
+ "step": 1870
+ },
+ {
+ "epoch": 26.0,
+ "eval_loss": 3.993298053741455,
+ "eval_runtime": 2.628,
+ "eval_samples_per_second": 928.833,
+ "eval_steps_per_second": 14.84,
+ "step": 1872
+ },
+ {
+ "epoch": 26.111692844677137,
+ "grad_norm": 0.19646763801574707,
+ "learning_rate": 0.0006,
+ "loss": 4.9817146301269535,
+ "step": 1880
+ },
+ {
+ "epoch": 26.25130890052356,
+ "grad_norm": 0.17267638444900513,
+ "learning_rate": 0.0006,
+ "loss": 5.046759414672851,
+ "step": 1890
+ },
+ {
+ "epoch": 26.390924956369982,
+ "grad_norm": 0.17786990106105804,
+ "learning_rate": 0.0006,
+ "loss": 4.959859466552734,
+ "step": 1900
+ },
+ {
+ "epoch": 26.530541012216403,
+ "grad_norm": 0.1823761910200119,
+ "learning_rate": 0.0006,
+ "loss": 4.985243225097657,
+ "step": 1910
+ },
+ {
+ "epoch": 26.670157068062828,
+ "grad_norm": 0.18100738525390625,
+ "learning_rate": 0.0006,
+ "loss": 5.024576950073242,
+ "step": 1920
+ },
+ {
+ "epoch": 26.80977312390925,
+ "grad_norm": 0.20121173560619354,
+ "learning_rate": 0.0006,
+ "loss": 5.037619400024414,
+ "step": 1930
+ },
+ {
+ "epoch": 26.949389179755673,
+ "grad_norm": 0.18984580039978027,
+ "learning_rate": 0.0006,
+ "loss": 4.998384857177735,
+ "step": 1940
+ },
+ {
+ "epoch": 27.0,
+ "eval_loss": 3.9770328998565674,
+ "eval_runtime": 2.6448,
+ "eval_samples_per_second": 922.927,
+ "eval_steps_per_second": 14.746,
+ "step": 1944
+ },
+ {
+ "epoch": 27.083769633507853,
+ "grad_norm": 0.17435336112976074,
+ "learning_rate": 0.0006,
+ "loss": 4.963232803344726,
+ "step": 1950
+ },
+ {
+ "epoch": 27.223385689354277,
+ "grad_norm": 0.20194946229457855,
+ "learning_rate": 0.0006,
+ "loss": 5.022874450683593,
+ "step": 1960
+ },
+ {
+ "epoch": 27.363001745200698,
+ "grad_norm": 0.18803757429122925,
+ "learning_rate": 0.0006,
+ "loss": 5.009196472167969,
+ "step": 1970
+ },
+ {
+ "epoch": 27.50261780104712,
+ "grad_norm": 0.17181430757045746,
+ "learning_rate": 0.0006,
+ "loss": 5.040612411499024,
+ "step": 1980
+ },
+ {
+ "epoch": 27.642233856893544,
+ "grad_norm": 0.19058096408843994,
+ "learning_rate": 0.0006,
+ "loss": 4.995831298828125,
+ "step": 1990
+ },
+ {
+ "epoch": 27.781849912739965,
+ "grad_norm": 0.17448312044143677,
+ "learning_rate": 0.0006,
+ "loss": 4.996364974975586,
+ "step": 2000
+ },
+ {
+ "epoch": 27.921465968586386,
+ "grad_norm": 0.18749794363975525,
+ "learning_rate": 0.0006,
+ "loss": 4.976995849609375,
+ "step": 2010
+ },
+ {
+ "epoch": 28.0,
+ "eval_loss": 3.9625418186187744,
+ "eval_runtime": 2.6381,
+ "eval_samples_per_second": 925.294,
+ "eval_steps_per_second": 14.783,
+ "step": 2016
+ },
+ {
+ "epoch": 28.05584642233857,
+ "grad_norm": 0.18358758091926575,
+ "learning_rate": 0.0006,
+ "loss": 4.970183563232422,
+ "step": 2020
+ },
+ {
+ "epoch": 28.195462478184993,
+ "grad_norm": 0.20860563218593597,
+ "learning_rate": 0.0006,
+ "loss": 4.961148452758789,
+ "step": 2030
+ },
+ {
+ "epoch": 28.335078534031414,
+ "grad_norm": 0.18246056139469147,
+ "learning_rate": 0.0006,
+ "loss": 4.967233276367187,
+ "step": 2040
+ },
+ {
+ "epoch": 28.474694589877835,
+ "grad_norm": 0.2207871824502945,
+ "learning_rate": 0.0006,
+ "loss": 4.978598022460938,
+ "step": 2050
+ },
+ {
+ "epoch": 28.61431064572426,
+ "grad_norm": 0.18436560034751892,
+ "learning_rate": 0.0006,
+ "loss": 5.009732818603515,
+ "step": 2060
+ },
+ {
+ "epoch": 28.75392670157068,
+ "grad_norm": 0.18158754706382751,
+ "learning_rate": 0.0006,
+ "loss": 4.965972900390625,
+ "step": 2070
+ },
+ {
+ "epoch": 28.8935427574171,
+ "grad_norm": 0.18567092716693878,
+ "learning_rate": 0.0006,
+ "loss": 5.002320098876953,
+ "step": 2080
+ },
+ {
+ "epoch": 29.0,
+ "eval_loss": 3.950843334197998,
+ "eval_runtime": 2.6432,
+ "eval_samples_per_second": 923.493,
+ "eval_steps_per_second": 14.755,
+ "step": 2088
+ },
+ {
+ "epoch": 29.027923211169284,
+ "grad_norm": 0.20220407843589783,
+ "learning_rate": 0.0006,
+ "loss": 4.907463073730469,
+ "step": 2090
+ },
+ {
+ "epoch": 29.167539267015705,
+ "grad_norm": 0.17813152074813843,
+ "learning_rate": 0.0006,
+ "loss": 4.934707260131836,
+ "step": 2100
+ },
+ {
+ "epoch": 29.30715532286213,
+ "grad_norm": 0.19826142489910126,
+ "learning_rate": 0.0006,
+ "loss": 4.96417007446289,
+ "step": 2110
+ },
+ {
+ "epoch": 29.44677137870855,
+ "grad_norm": 0.17958831787109375,
+ "learning_rate": 0.0006,
+ "loss": 4.923752212524414,
+ "step": 2120
+ },
+ {
+ "epoch": 29.586387434554975,
+ "grad_norm": 0.2165687084197998,
+ "learning_rate": 0.0006,
+ "loss": 4.941191864013672,
+ "step": 2130
+ },
+ {
+ "epoch": 29.726003490401396,
+ "grad_norm": 0.1816084384918213,
+ "learning_rate": 0.0006,
+ "loss": 5.001884841918946,
+ "step": 2140
+ },
+ {
+ "epoch": 29.865619546247817,
+ "grad_norm": 0.19192010164260864,
+ "learning_rate": 0.0006,
+ "loss": 4.968138122558594,
+ "step": 2150
+ },
+ {
+ "epoch": 30.0,
+ "grad_norm": 0.20404483377933502,
+ "learning_rate": 0.0006,
+ "loss": 4.947370529174805,
+ "step": 2160
+ },
+ {
+ "epoch": 30.0,
+ "eval_loss": 3.9439475536346436,
+ "eval_runtime": 2.6286,
+ "eval_samples_per_second": 928.614,
+ "eval_steps_per_second": 14.837,
+ "step": 2160
+ },
+ {
+ "epoch": 30.13961605584642,
+ "grad_norm": 0.21564748883247375,
+ "learning_rate": 0.0006,
+ "loss": 4.904407501220703,
+ "step": 2170
+ },
+ {
+ "epoch": 30.279232111692846,
+ "grad_norm": 0.20033442974090576,
+ "learning_rate": 0.0006,
+ "loss": 4.97376708984375,
+ "step": 2180
+ },
+ {
+ "epoch": 30.418848167539267,
+ "grad_norm": 0.21858379244804382,
+ "learning_rate": 0.0006,
+ "loss": 4.951438140869141,
+ "step": 2190
+ },
+ {
+ "epoch": 30.55846422338569,
+ "grad_norm": 0.19368436932563782,
+ "learning_rate": 0.0006,
+ "loss": 4.92889633178711,
+ "step": 2200
+ },
+ {
+ "epoch": 30.698080279232112,
+ "grad_norm": 0.2785509526729584,
+ "learning_rate": 0.0006,
+ "loss": 4.868854522705078,
+ "step": 2210
+ },
+ {
+ "epoch": 30.837696335078533,
+ "grad_norm": 0.1883997619152069,
+ "learning_rate": 0.0006,
+ "loss": 4.8884521484375,
+ "step": 2220
+ },
+ {
+ "epoch": 30.977312390924958,
+ "grad_norm": 0.18403448164463043,
+ "learning_rate": 0.0006,
+ "loss": 4.935443496704101,
+ "step": 2230
+ },
+ {
+ "epoch": 31.0,
+ "eval_loss": 3.9323782920837402,
+ "eval_runtime": 2.649,
+ "eval_samples_per_second": 921.468,
+ "eval_steps_per_second": 14.722,
+ "step": 2232
+ },
+ {
+ "epoch": 31.111692844677137,
+ "grad_norm": 0.20197638869285583,
+ "learning_rate": 0.0006,
+ "loss": 4.939474487304688,
+ "step": 2240
+ },
+ {
+ "epoch": 31.25130890052356,
+ "grad_norm": 0.19043312966823578,
+ "learning_rate": 0.0006,
+ "loss": 4.877633666992187,
+ "step": 2250
+ },
+ {
+ "epoch": 31.390924956369982,
+ "grad_norm": 0.23196810483932495,
+ "learning_rate": 0.0006,
+ "loss": 4.893428039550781,
+ "step": 2260
+ },
+ {
+ "epoch": 31.530541012216403,
+ "grad_norm": 0.19379320740699768,
+ "learning_rate": 0.0006,
+ "loss": 4.867322540283203,
+ "step": 2270
+ },
+ {
+ "epoch": 31.670157068062828,
+ "grad_norm": 0.1862214058637619,
+ "learning_rate": 0.0006,
+ "loss": 4.8929389953613285,
+ "step": 2280
+ },
+ {
+ "epoch": 31.80977312390925,
+ "grad_norm": 0.18675987422466278,
+ "learning_rate": 0.0006,
+ "loss": 4.945796966552734,
+ "step": 2290
+ },
+ {
+ "epoch": 31.949389179755673,
+ "grad_norm": 0.1910662204027176,
+ "learning_rate": 0.0006,
+ "loss": 4.919340515136719,
+ "step": 2300
+ },
+ {
+ "epoch": 32.0,
+ "eval_loss": 3.9239120483398438,
+ "eval_runtime": 2.6438,
+ "eval_samples_per_second": 923.298,
+ "eval_steps_per_second": 14.752,
+ "step": 2304
+ },
+ {
+ "epoch": 32.083769633507856,
+ "grad_norm": 0.2327888011932373,
+ "learning_rate": 0.0006,
+ "loss": 4.900206756591797,
+ "step": 2310
+ },
+ {
+ "epoch": 32.223385689354274,
+ "grad_norm": 0.20494481921195984,
+ "learning_rate": 0.0006,
+ "loss": 4.88743896484375,
+ "step": 2320
+ },
+ {
+ "epoch": 32.3630017452007,
+ "grad_norm": 0.1939210593700409,
+ "learning_rate": 0.0006,
+ "loss": 4.908115005493164,
+ "step": 2330
+ },
+ {
+ "epoch": 32.50261780104712,
+ "grad_norm": 0.21181504428386688,
+ "learning_rate": 0.0006,
+ "loss": 4.849565505981445,
+ "step": 2340
+ },
+ {
+ "epoch": 32.64223385689354,
+ "grad_norm": 0.20313170552253723,
+ "learning_rate": 0.0006,
+ "loss": 4.955375671386719,
+ "step": 2350
+ },
+ {
+ "epoch": 32.781849912739965,
+ "grad_norm": 0.18007895350456238,
+ "learning_rate": 0.0006,
+ "loss": 4.935146331787109,
+ "step": 2360
+ },
+ {
+ "epoch": 32.92146596858639,
+ "grad_norm": 0.21850770711898804,
+ "learning_rate": 0.0006,
+ "loss": 4.939716339111328,
+ "step": 2370
+ },
+ {
+ "epoch": 33.0,
+ "eval_loss": 3.919341802597046,
+ "eval_runtime": 2.6302,
+ "eval_samples_per_second": 928.068,
+ "eval_steps_per_second": 14.828,
+ "step": 2376
+ },
+ {
+ "epoch": 33.05584642233857,
+ "grad_norm": 0.2179473489522934,
+ "learning_rate": 0.0006,
+ "loss": 4.972684478759765,
+ "step": 2380
+ },
+ {
+ "epoch": 33.19546247818499,
+ "grad_norm": 0.1947934925556183,
+ "learning_rate": 0.0006,
+ "loss": 4.8848731994628904,
+ "step": 2390
+ },
+ {
+ "epoch": 33.33507853403141,
+ "grad_norm": 0.2406018078327179,
+ "learning_rate": 0.0006,
+ "loss": 4.886126708984375,
+ "step": 2400
+ },
+ {
+ "epoch": 33.474694589877835,
+ "grad_norm": 0.18514129519462585,
+ "learning_rate": 0.0006,
+ "loss": 4.904386138916015,
+ "step": 2410
+ },
+ {
+ "epoch": 33.61431064572426,
+ "grad_norm": 0.2080937623977661,
+ "learning_rate": 0.0006,
+ "loss": 4.925667572021484,
+ "step": 2420
+ },
+ {
+ "epoch": 33.753926701570684,
+ "grad_norm": 0.20405733585357666,
+ "learning_rate": 0.0006,
+ "loss": 4.861196136474609,
+ "step": 2430
+ },
+ {
+ "epoch": 33.8935427574171,
+ "grad_norm": 0.19720759987831116,
+ "learning_rate": 0.0006,
+ "loss": 4.867739868164063,
+ "step": 2440
+ },
+ {
+ "epoch": 34.0,
+ "eval_loss": 3.9056503772735596,
+ "eval_runtime": 2.6327,
+ "eval_samples_per_second": 927.192,
+ "eval_steps_per_second": 14.814,
+ "step": 2448
+ },
+ {
+ "epoch": 34.02792321116929,
+ "grad_norm": 0.1980334222316742,
+ "learning_rate": 0.0006,
+ "loss": 4.905588531494141,
+ "step": 2450
+ },
+ {
+ "epoch": 34.167539267015705,
+ "grad_norm": 0.2271888107061386,
+ "learning_rate": 0.0006,
+ "loss": 4.817427062988282,
+ "step": 2460
+ },
+ {
+ "epoch": 34.30715532286213,
+ "grad_norm": 0.1933271735906601,
+ "learning_rate": 0.0006,
+ "loss": 4.862692260742188,
+ "step": 2470
+ },
+ {
+ "epoch": 34.446771378708554,
+ "grad_norm": 0.19238920509815216,
+ "learning_rate": 0.0006,
+ "loss": 4.861783218383789,
+ "step": 2480
+ },
+ {
+ "epoch": 34.58638743455497,
+ "grad_norm": 0.17851795256137848,
+ "learning_rate": 0.0006,
+ "loss": 4.84129638671875,
+ "step": 2490
+ },
+ {
+ "epoch": 34.726003490401396,
+ "grad_norm": 0.21409454941749573,
+ "learning_rate": 0.0006,
+ "loss": 4.92855224609375,
+ "step": 2500
+ },
+ {
+ "epoch": 34.86561954624782,
+ "grad_norm": 0.19701579213142395,
+ "learning_rate": 0.0006,
+ "loss": 4.923699188232422,
+ "step": 2510
+ },
+ {
+ "epoch": 35.0,
+ "grad_norm": 0.24597975611686707,
+ "learning_rate": 0.0006,
+ "loss": 4.860978698730468,
+ "step": 2520
+ },
+ {
+ "epoch": 35.0,
+ "eval_loss": 3.904244899749756,
+ "eval_runtime": 2.6256,
+ "eval_samples_per_second": 929.703,
+ "eval_steps_per_second": 14.854,
+ "step": 2520
+ },
+ {
+ "epoch": 35.139616055846425,
+ "grad_norm": 0.1964741051197052,
+ "learning_rate": 0.0006,
+ "loss": 4.799929428100586,
+ "step": 2530
+ },
+ {
+ "epoch": 35.27923211169284,
+ "grad_norm": 0.20848998427391052,
+ "learning_rate": 0.0006,
+ "loss": 4.8812416076660154,
+ "step": 2540
+ },
+ {
+ "epoch": 35.41884816753927,
+ "grad_norm": 0.23481105268001556,
+ "learning_rate": 0.0006,
+ "loss": 4.8670204162597654,
+ "step": 2550
+ },
+ {
+ "epoch": 35.55846422338569,
+ "grad_norm": 0.20906837284564972,
+ "learning_rate": 0.0006,
+ "loss": 4.8502655029296875,
+ "step": 2560
+ },
+ {
+ "epoch": 35.69808027923211,
+ "grad_norm": 0.21247205138206482,
+ "learning_rate": 0.0006,
+ "loss": 4.895785522460938,
+ "step": 2570
+ },
+ {
+ "epoch": 35.83769633507853,
+ "grad_norm": 0.19133782386779785,
+ "learning_rate": 0.0006,
+ "loss": 4.8910667419433596,
+ "step": 2580
+ },
+ {
+ "epoch": 35.97731239092496,
+ "grad_norm": 0.1973542422056198,
+ "learning_rate": 0.0006,
+ "loss": 4.909728240966797,
+ "step": 2590
+ },
+ {
+ "epoch": 36.0,
+ "eval_loss": 3.895230531692505,
+ "eval_runtime": 2.6247,
+ "eval_samples_per_second": 929.998,
+ "eval_steps_per_second": 14.859,
+ "step": 2592
+ },
+ {
+ "epoch": 36.11169284467714,
+ "grad_norm": 0.19480381906032562,
+ "learning_rate": 0.0006,
+ "loss": 4.856618499755859,
+ "step": 2600
+ },
+ {
+ "epoch": 36.25130890052356,
+ "grad_norm": 0.23110529780387878,
+ "learning_rate": 0.0006,
+ "loss": 4.835741424560547,
+ "step": 2610
+ },
+ {
+ "epoch": 36.390924956369986,
+ "grad_norm": 0.1905093640089035,
+ "learning_rate": 0.0006,
+ "loss": 4.864842224121094,
+ "step": 2620
+ },
+ {
+ "epoch": 36.5305410122164,
+ "grad_norm": 0.18635043501853943,
+ "learning_rate": 0.0006,
+ "loss": 4.878033447265625,
+ "step": 2630
+ },
+ {
+ "epoch": 36.67015706806283,
+ "grad_norm": 0.22753402590751648,
+ "learning_rate": 0.0006,
+ "loss": 4.830284881591797,
+ "step": 2640
+ },
+ {
+ "epoch": 36.80977312390925,
+ "grad_norm": 0.2048637866973877,
+ "learning_rate": 0.0006,
+ "loss": 4.808927154541015,
+ "step": 2650
+ },
+ {
+ "epoch": 36.94938917975567,
+ "grad_norm": 0.1901310235261917,
+ "learning_rate": 0.0006,
+ "loss": 4.857204818725586,
+ "step": 2660
+ },
+ {
+ "epoch": 37.0,
+ "eval_loss": 3.887256383895874,
+ "eval_runtime": 2.6241,
+ "eval_samples_per_second": 930.206,
+ "eval_steps_per_second": 14.862,
+ "step": 2664
+ },
+ {
+ "epoch": 37.083769633507856,
+ "grad_norm": 0.2037169337272644,
+ "learning_rate": 0.0006,
+ "loss": 4.846020126342774,
+ "step": 2670
+ },
+ {
+ "epoch": 37.223385689354274,
+ "grad_norm": 0.20483948290348053,
+ "learning_rate": 0.0006,
+ "loss": 4.782030487060547,
+ "step": 2680
+ },
+ {
+ "epoch": 37.3630017452007,
+ "grad_norm": 0.202740877866745,
+ "learning_rate": 0.0006,
+ "loss": 4.840346908569336,
+ "step": 2690
+ },
+ {
+ "epoch": 37.50261780104712,
+ "grad_norm": 0.21044015884399414,
+ "learning_rate": 0.0006,
+ "loss": 4.878639984130859,
+ "step": 2700
+ },
+ {
+ "epoch": 37.64223385689354,
+ "grad_norm": 0.18962226808071136,
+ "learning_rate": 0.0006,
+ "loss": 4.884788894653321,
+ "step": 2710
+ },
+ {
+ "epoch": 37.781849912739965,
+ "grad_norm": 0.1964670568704605,
+ "learning_rate": 0.0006,
+ "loss": 4.898389053344727,
+ "step": 2720
+ },
+ {
+ "epoch": 37.92146596858639,
+ "grad_norm": 0.2292124330997467,
+ "learning_rate": 0.0006,
+ "loss": 4.875354385375976,
+ "step": 2730
+ },
+ {
+ "epoch": 38.0,
+ "eval_loss": 3.8863890171051025,
+ "eval_runtime": 2.6382,
+ "eval_samples_per_second": 925.256,
+ "eval_steps_per_second": 14.783,
+ "step": 2736
+ },
+ {
+ "epoch": 38.05584642233857,
+ "grad_norm": 0.21509435772895813,
+ "learning_rate": 0.0006,
+ "loss": 4.792652130126953,
+ "step": 2740
+ },
+ {
+ "epoch": 38.19546247818499,
+ "grad_norm": 0.20651887357234955,
+ "learning_rate": 0.0006,
+ "loss": 4.815517425537109,
+ "step": 2750
+ },
+ {
+ "epoch": 38.33507853403141,
+ "grad_norm": 0.21954692900180817,
+ "learning_rate": 0.0006,
+ "loss": 4.847898483276367,
+ "step": 2760
+ },
+ {
+ "epoch": 38.474694589877835,
+ "grad_norm": 0.20071472227573395,
+ "learning_rate": 0.0006,
+ "loss": 4.84312858581543,
+ "step": 2770
+ },
+ {
+ "epoch": 38.61431064572426,
+ "grad_norm": 0.21303021907806396,
+ "learning_rate": 0.0006,
+ "loss": 4.786624145507813,
+ "step": 2780
+ },
+ {
+ "epoch": 38.753926701570684,
+ "grad_norm": 0.20232032239437103,
+ "learning_rate": 0.0006,
+ "loss": 4.8715869903564455,
+ "step": 2790
+ },
+ {
+ "epoch": 38.8935427574171,
+ "grad_norm": 0.1989263892173767,
+ "learning_rate": 0.0006,
+ "loss": 4.863502502441406,
+ "step": 2800
+ },
+ {
+ "epoch": 39.0,
+ "eval_loss": 3.8790223598480225,
+ "eval_runtime": 2.6313,
+ "eval_samples_per_second": 927.691,
+ "eval_steps_per_second": 14.822,
+ "step": 2808
+ },
+ {
+ "epoch": 39.02792321116929,
+ "grad_norm": 0.20564888417720795,
+ "learning_rate": 0.0006,
+ "loss": 4.810267639160156,
+ "step": 2810
+ },
+ {
+ "epoch": 39.167539267015705,
+ "grad_norm": 0.20790626108646393,
+ "learning_rate": 0.0006,
+ "loss": 4.78194580078125,
+ "step": 2820
+ },
+ {
+ "epoch": 39.30715532286213,
+ "grad_norm": 0.19991527497768402,
+ "learning_rate": 0.0006,
+ "loss": 4.867174911499023,
+ "step": 2830
+ },
+ {
+ "epoch": 39.446771378708554,
+ "grad_norm": 0.20038843154907227,
+ "learning_rate": 0.0006,
+ "loss": 4.79072265625,
+ "step": 2840
+ },
+ {
+ "epoch": 39.58638743455497,
+ "grad_norm": 0.2362481653690338,
+ "learning_rate": 0.0006,
+ "loss": 4.83765983581543,
+ "step": 2850
+ },
+ {
+ "epoch": 39.726003490401396,
+ "grad_norm": 0.23314230144023895,
+ "learning_rate": 0.0006,
+ "loss": 4.8610984802246096,
+ "step": 2860
+ },
+ {
+ "epoch": 39.86561954624782,
+ "grad_norm": 0.21623562276363373,
+ "learning_rate": 0.0006,
+ "loss": 4.854366302490234,
+ "step": 2870
+ },
+ {
+ "epoch": 40.0,
+ "grad_norm": 0.2093755155801773,
+ "learning_rate": 0.0006,
+ "loss": 4.864644622802734,
+ "step": 2880
+ },
+ {
+ "epoch": 40.0,
+ "eval_loss": 3.87245512008667,
+ "eval_runtime": 2.6231,
+ "eval_samples_per_second": 930.566,
+ "eval_steps_per_second": 14.868,
+ "step": 2880
+ },
+ {
+ "epoch": 40.139616055846425,
+ "grad_norm": 0.20051318407058716,
+ "learning_rate": 0.0006,
+ "loss": 4.837277221679687,
+ "step": 2890
+ },
+ {
+ "epoch": 40.27923211169284,
+ "grad_norm": 0.20958660542964935,
+ "learning_rate": 0.0006,
+ "loss": 4.749915695190429,
+ "step": 2900
+ },
+ {
+ "epoch": 40.41884816753927,
+ "grad_norm": 0.19435004889965057,
+ "learning_rate": 0.0006,
+ "loss": 4.797109222412109,
+ "step": 2910
+ },
+ {
+ "epoch": 40.55846422338569,
+ "grad_norm": 0.1970185488462448,
+ "learning_rate": 0.0006,
+ "loss": 4.810147094726562,
+ "step": 2920
+ },
+ {
+ "epoch": 40.69808027923211,
+ "grad_norm": 0.20284049212932587,
+ "learning_rate": 0.0006,
+ "loss": 4.828338241577148,
+ "step": 2930
+ },
+ {
+ "epoch": 40.83769633507853,
+ "grad_norm": 0.19942472875118256,
+ "learning_rate": 0.0006,
+ "loss": 4.880989074707031,
+ "step": 2940
+ },
+ {
+ "epoch": 40.97731239092496,
+ "grad_norm": 0.2194322645664215,
+ "learning_rate": 0.0006,
+ "loss": 4.8256378173828125,
+ "step": 2950
+ },
+ {
+ "epoch": 41.0,
+ "eval_loss": 3.8675320148468018,
+ "eval_runtime": 2.6336,
+ "eval_samples_per_second": 926.867,
+ "eval_steps_per_second": 14.809,
+ "step": 2952
+ },
+ {
+ "epoch": 41.11169284467714,
+ "grad_norm": 0.23923835158348083,
+ "learning_rate": 0.0006,
+ "loss": 4.791503143310547,
+ "step": 2960
+ },
+ {
+ "epoch": 41.25130890052356,
+ "grad_norm": 0.19268065690994263,
+ "learning_rate": 0.0006,
+ "loss": 4.827136611938476,
+ "step": 2970
+ },
+ {
+ "epoch": 41.390924956369986,
+ "grad_norm": 0.2224045991897583,
+ "learning_rate": 0.0006,
+ "loss": 4.744147491455078,
+ "step": 2980
+ },
+ {
+ "epoch": 41.5305410122164,
+ "grad_norm": 0.2124592363834381,
+ "learning_rate": 0.0006,
+ "loss": 4.865887451171875,
+ "step": 2990
+ },
+ {
+ "epoch": 41.67015706806283,
+ "grad_norm": 0.21108582615852356,
+ "learning_rate": 0.0006,
+ "loss": 4.811237335205078,
+ "step": 3000
+ },
+ {
+ "epoch": 41.80977312390925,
+ "grad_norm": 0.2068081945180893,
+ "learning_rate": 0.0006,
+ "loss": 4.8243568420410154,
+ "step": 3010
+ },
+ {
+ "epoch": 41.94938917975567,
+ "grad_norm": 0.20204509794712067,
+ "learning_rate": 0.0006,
+ "loss": 4.811293029785157,
+ "step": 3020
+ },
+ {
+ "epoch": 42.0,
+ "eval_loss": 3.8697407245635986,
+ "eval_runtime": 2.6412,
+ "eval_samples_per_second": 924.19,
+ "eval_steps_per_second": 14.766,
+ "step": 3024
+ },
+ {
+ "epoch": 42.083769633507856,
+ "grad_norm": 0.20372816920280457,
+ "learning_rate": 0.0006,
+ "loss": 4.795223999023437,
+ "step": 3030
+ },
+ {
+ "epoch": 42.223385689354274,
+ "grad_norm": 0.22295735776424408,
+ "learning_rate": 0.0006,
+ "loss": 4.790752410888672,
+ "step": 3040
+ },
+ {
+ "epoch": 42.3630017452007,
+ "grad_norm": 0.2060958743095398,
+ "learning_rate": 0.0006,
+ "loss": 4.753063583374024,
+ "step": 3050
+ },
+ {
+ "epoch": 42.50261780104712,
+ "grad_norm": 0.24282215535640717,
+ "learning_rate": 0.0006,
+ "loss": 4.830035400390625,
+ "step": 3060
+ },
+ {
+ "epoch": 42.64223385689354,
+ "grad_norm": 0.20753711462020874,
+ "learning_rate": 0.0006,
+ "loss": 4.818644714355469,
+ "step": 3070
+ },
+ {
+ "epoch": 42.781849912739965,
+ "grad_norm": 0.22238077223300934,
+ "learning_rate": 0.0006,
+ "loss": 4.821812438964844,
+ "step": 3080
+ },
+ {
+ "epoch": 42.92146596858639,
+ "grad_norm": 0.2211921066045761,
+ "learning_rate": 0.0006,
+ "loss": 4.833587646484375,
+ "step": 3090
+ },
+ {
+ "epoch": 43.0,
+ "eval_loss": 3.8600575923919678,
+ "eval_runtime": 2.6447,
+ "eval_samples_per_second": 922.976,
+ "eval_steps_per_second": 14.746,
+ "step": 3096
+ },
+ {
+ "epoch": 43.05584642233857,
+ "grad_norm": 0.25350046157836914,
+ "learning_rate": 0.0006,
+ "loss": 4.804128265380859,
+ "step": 3100
+ },
+ {
+ "epoch": 43.19546247818499,
+ "grad_norm": 0.21559758484363556,
+ "learning_rate": 0.0006,
+ "loss": 4.8152313232421875,
+ "step": 3110
+ },
+ {
+ "epoch": 43.33507853403141,
+ "grad_norm": 0.20546399056911469,
+ "learning_rate": 0.0006,
+ "loss": 4.808087158203125,
+ "step": 3120
+ },
+ {
+ "epoch": 43.474694589877835,
+ "grad_norm": 0.2224588692188263,
+ "learning_rate": 0.0006,
+ "loss": 4.792206573486328,
+ "step": 3130
+ },
+ {
+ "epoch": 43.61431064572426,
+ "grad_norm": 0.19151629507541656,
+ "learning_rate": 0.0006,
+ "loss": 4.784809494018555,
+ "step": 3140
+ },
+ {
+ "epoch": 43.753926701570684,
+ "grad_norm": 0.2352108359336853,
+ "learning_rate": 0.0006,
+ "loss": 4.8115684509277346,
+ "step": 3150
+ },
+ {
+ "epoch": 43.8935427574171,
+ "grad_norm": 0.20950759947299957,
+ "learning_rate": 0.0006,
+ "loss": 4.785184860229492,
+ "step": 3160
+ },
+ {
+ "epoch": 44.0,
+ "eval_loss": 3.8647193908691406,
+ "eval_runtime": 2.6264,
+ "eval_samples_per_second": 929.42,
+ "eval_steps_per_second": 14.849,
+ "step": 3168
+ },
+ {
+ "epoch": 44.02792321116929,
+ "grad_norm": 0.24875704944133759,
+ "learning_rate": 0.0006,
+ "loss": 4.851061248779297,
+ "step": 3170
+ },
+ {
+ "epoch": 44.167539267015705,
+ "grad_norm": 0.23177148401737213,
+ "learning_rate": 0.0006,
+ "loss": 4.778269195556641,
+ "step": 3180
+ },
+ {
+ "epoch": 44.30715532286213,
+ "grad_norm": 0.20917125046253204,
+ "learning_rate": 0.0006,
+ "loss": 4.766267776489258,
+ "step": 3190
+ },
+ {
+ "epoch": 44.446771378708554,
+ "grad_norm": 0.23103971779346466,
+ "learning_rate": 0.0006,
+ "loss": 4.766683197021484,
+ "step": 3200
+ },
+ {
+ "epoch": 44.58638743455497,
+ "grad_norm": 0.1903815120458603,
+ "learning_rate": 0.0006,
+ "loss": 4.766812896728515,
+ "step": 3210
+ },
+ {
+ "epoch": 44.726003490401396,
+ "grad_norm": 0.23820948600769043,
+ "learning_rate": 0.0006,
+ "loss": 4.779206085205078,
+ "step": 3220
+ },
+ {
+ "epoch": 44.86561954624782,
+ "grad_norm": 0.19525264203548431,
+ "learning_rate": 0.0006,
+ "loss": 4.835651016235351,
+ "step": 3230
+ },
+ {
+ "epoch": 45.0,
+ "grad_norm": 0.2601972222328186,
+ "learning_rate": 0.0006,
+ "loss": 4.695940399169922,
+ "step": 3240
+ },
+ {
+ "epoch": 45.0,
+ "eval_loss": 3.8567686080932617,
+ "eval_runtime": 2.6307,
+ "eval_samples_per_second": 927.892,
+ "eval_steps_per_second": 14.825,
+ "step": 3240
+ },
+ {
+ "epoch": 45.139616055846425,
+ "grad_norm": 0.203897625207901,
+ "learning_rate": 0.0006,
+ "loss": 4.711742782592774,
+ "step": 3250
+ },
+ {
+ "epoch": 45.27923211169284,
+ "grad_norm": 0.21360500156879425,
+ "learning_rate": 0.0006,
+ "loss": 4.8410179138183596,
+ "step": 3260
+ },
+ {
+ "epoch": 45.41884816753927,
+ "grad_norm": 0.1932840645313263,
+ "learning_rate": 0.0006,
+ "loss": 4.78133544921875,
+ "step": 3270
+ },
+ {
+ "epoch": 45.55846422338569,
+ "grad_norm": 0.22860406339168549,
+ "learning_rate": 0.0006,
+ "loss": 4.814667510986328,
+ "step": 3280
+ },
+ {
+ "epoch": 45.69808027923211,
+ "grad_norm": 0.2251984030008316,
+ "learning_rate": 0.0006,
+ "loss": 4.801509857177734,
+ "step": 3290
+ },
+ {
+ "epoch": 45.83769633507853,
+ "grad_norm": 0.2596592307090759,
+ "learning_rate": 0.0006,
+ "loss": 4.7847412109375,
+ "step": 3300
+ },
+ {
+ "epoch": 45.97731239092496,
+ "grad_norm": 0.20352135598659515,
+ "learning_rate": 0.0006,
+ "loss": 4.774466705322266,
+ "step": 3310
+ },
+ {
+ "epoch": 46.0,
+ "eval_loss": 3.850667953491211,
+ "eval_runtime": 2.6282,
+ "eval_samples_per_second": 928.773,
+ "eval_steps_per_second": 14.839,
+ "step": 3312
+ },
+ {
+ "epoch": 46.11169284467714,
+ "grad_norm": 0.24113047122955322,
+ "learning_rate": 0.0006,
+ "loss": 4.762670516967773,
+ "step": 3320
+ },
+ {
+ "epoch": 46.25130890052356,
+ "grad_norm": 0.19837063550949097,
+ "learning_rate": 0.0006,
+ "loss": 4.738725280761718,
+ "step": 3330
+ },
+ {
+ "epoch": 46.390924956369986,
+ "grad_norm": 0.1964500993490219,
+ "learning_rate": 0.0006,
+ "loss": 4.761268615722656,
+ "step": 3340
+ },
+ {
+ "epoch": 46.5305410122164,
+ "grad_norm": 0.21460574865341187,
+ "learning_rate": 0.0006,
+ "loss": 4.804327011108398,
+ "step": 3350
+ },
+ {
+ "epoch": 46.67015706806283,
+ "grad_norm": 0.2273215800523758,
+ "learning_rate": 0.0006,
+ "loss": 4.832907485961914,
+ "step": 3360
+ },
+ {
+ "epoch": 46.80977312390925,
+ "grad_norm": 0.2531464695930481,
+ "learning_rate": 0.0006,
+ "loss": 4.797586059570312,
+ "step": 3370
+ },
+ {
+ "epoch": 46.94938917975567,
+ "grad_norm": 0.22594818472862244,
+ "learning_rate": 0.0006,
+ "loss": 4.813489532470703,
+ "step": 3380
+ },
+ {
+ "epoch": 47.0,
+ "eval_loss": 3.8478593826293945,
+ "eval_runtime": 2.6328,
+ "eval_samples_per_second": 927.144,
+ "eval_steps_per_second": 14.813,
+ "step": 3384
+ },
+ {
+ "epoch": 47.083769633507856,
+ "grad_norm": 0.21628274023532867,
+ "learning_rate": 0.0006,
+ "loss": 4.764807510375976,
+ "step": 3390
+ },
+ {
+ "epoch": 47.223385689354274,
+ "grad_norm": 0.21843400597572327,
+ "learning_rate": 0.0006,
+ "loss": 4.71702880859375,
+ "step": 3400
+ },
+ {
+ "epoch": 47.3630017452007,
+ "grad_norm": 0.18754912912845612,
+ "learning_rate": 0.0006,
+ "loss": 4.76793327331543,
+ "step": 3410
+ },
+ {
+ "epoch": 47.50261780104712,
+ "grad_norm": 0.20255476236343384,
+ "learning_rate": 0.0006,
+ "loss": 4.796073913574219,
+ "step": 3420
+ },
+ {
+ "epoch": 47.64223385689354,
+ "grad_norm": 0.21224115788936615,
+ "learning_rate": 0.0006,
+ "loss": 4.723152542114258,
+ "step": 3430
+ },
+ {
+ "epoch": 47.781849912739965,
+ "grad_norm": 0.2122431844472885,
+ "learning_rate": 0.0006,
+ "loss": 4.828311157226563,
+ "step": 3440
+ },
+ {
+ "epoch": 47.92146596858639,
+ "grad_norm": 0.18767400085926056,
+ "learning_rate": 0.0006,
+ "loss": 4.808477783203125,
+ "step": 3450
+ },
+ {
+ "epoch": 48.0,
+ "eval_loss": 3.842599391937256,
+ "eval_runtime": 2.6445,
+ "eval_samples_per_second": 923.042,
+ "eval_steps_per_second": 14.747,
+ "step": 3456
+ },
+ {
+ "epoch": 48.05584642233857,
+ "grad_norm": 0.2661328911781311,
+ "learning_rate": 0.0006,
+ "loss": 4.6859130859375,
+ "step": 3460
+ },
+ {
+ "epoch": 48.19546247818499,
+ "grad_norm": 0.23248155415058136,
+ "learning_rate": 0.0006,
+ "loss": 4.771257781982422,
+ "step": 3470
+ },
+ {
+ "epoch": 48.33507853403141,
+ "grad_norm": 0.2126832902431488,
+ "learning_rate": 0.0006,
+ "loss": 4.725955963134766,
+ "step": 3480
+ },
+ {
+ "epoch": 48.474694589877835,
+ "grad_norm": 0.22934485971927643,
+ "learning_rate": 0.0006,
+ "loss": 4.7511474609375,
+ "step": 3490
+ },
+ {
+ "epoch": 48.61431064572426,
+ "grad_norm": 0.20873117446899414,
+ "learning_rate": 0.0006,
+ "loss": 4.7393756866455075,
+ "step": 3500
+ },
+ {
+ "epoch": 48.753926701570684,
+ "grad_norm": 0.21192826330661774,
+ "learning_rate": 0.0006,
+ "loss": 4.723600769042969,
+ "step": 3510
+ },
+ {
+ "epoch": 48.8935427574171,
+ "grad_norm": 0.2070503532886505,
+ "learning_rate": 0.0006,
+ "loss": 4.7907569885253904,
+ "step": 3520
+ },
+ {
+ "epoch": 49.0,
+ "eval_loss": 3.8490540981292725,
+ "eval_runtime": 2.6325,
+ "eval_samples_per_second": 927.254,
+ "eval_steps_per_second": 14.815,
+ "step": 3528
+ },
+ {
+ "epoch": 49.02792321116929,
+ "grad_norm": 0.27866652607917786,
+ "learning_rate": 0.0006,
+ "loss": 4.766044616699219,
+ "step": 3530
+ },
+ {
+ "epoch": 49.167539267015705,
+ "grad_norm": 0.27155712246894836,
+ "learning_rate": 0.0006,
+ "loss": 4.720732879638672,
+ "step": 3540
+ },
+ {
+ "epoch": 49.30715532286213,
+ "grad_norm": 0.1931651383638382,
+ "learning_rate": 0.0006,
+ "loss": 4.732492828369141,
+ "step": 3550
+ },
+ {
+ "epoch": 49.446771378708554,
+ "grad_norm": 0.2052825689315796,
+ "learning_rate": 0.0006,
+ "loss": 4.713016510009766,
+ "step": 3560
+ },
+ {
+ "epoch": 49.58638743455497,
+ "grad_norm": 0.19564181566238403,
+ "learning_rate": 0.0006,
+ "loss": 4.7052757263183596,
+ "step": 3570
+ },
+ {
+ "epoch": 49.726003490401396,
+ "grad_norm": 0.24012479186058044,
+ "learning_rate": 0.0006,
+ "loss": 4.781451416015625,
+ "step": 3580
+ },
+ {
+ "epoch": 49.86561954624782,
+ "grad_norm": 0.21026678383350372,
+ "learning_rate": 0.0006,
+ "loss": 4.773517608642578,
+ "step": 3590
+ },
+ {
+ "epoch": 50.0,
+ "grad_norm": 0.2199603021144867,
+ "learning_rate": 0.0006,
+ "loss": 4.755486297607422,
+ "step": 3600
+ },
+ {
+ "epoch": 50.0,
+ "eval_loss": 3.837543249130249,
+ "eval_runtime": 2.6414,
+ "eval_samples_per_second": 924.149,
+ "eval_steps_per_second": 14.765,
+ "step": 3600
+ },
+ {
+ "epoch": 50.139616055846425,
+ "grad_norm": 0.22548460960388184,
+ "learning_rate": 0.0006,
+ "loss": 4.663722991943359,
+ "step": 3610
+ },
+ {
+ "epoch": 50.27923211169284,
+ "grad_norm": 0.239980548620224,
+ "learning_rate": 0.0006,
+ "loss": 4.755983352661133,
+ "step": 3620
+ },
+ {
+ "epoch": 50.41884816753927,
+ "grad_norm": 0.2253708392381668,
+ "learning_rate": 0.0006,
+ "loss": 4.7740226745605465,
+ "step": 3630
+ },
+ {
+ "epoch": 50.55846422338569,
+ "grad_norm": 0.19401922821998596,
+ "learning_rate": 0.0006,
+ "loss": 4.686423110961914,
+ "step": 3640
+ },
+ {
+ "epoch": 50.69808027923211,
+ "grad_norm": 0.2387697845697403,
+ "learning_rate": 0.0006,
+ "loss": 4.771978759765625,
+ "step": 3650
+ },
+ {
+ "epoch": 50.83769633507853,
+ "grad_norm": 0.2253815084695816,
+ "learning_rate": 0.0006,
+ "loss": 4.749351501464844,
+ "step": 3660
+ },
+ {
+ "epoch": 50.97731239092496,
+ "grad_norm": 0.19555051624774933,
+ "learning_rate": 0.0006,
+ "loss": 4.824957275390625,
+ "step": 3670
+ },
+ {
+ "epoch": 51.0,
+ "eval_loss": 3.8389782905578613,
+ "eval_runtime": 2.6213,
+ "eval_samples_per_second": 931.2,
+ "eval_steps_per_second": 14.878,
+ "step": 3672
+ },
+ {
+ "epoch": 51.11169284467714,
+ "grad_norm": 0.2489580363035202,
+ "learning_rate": 0.0006,
+ "loss": 4.6748615264892575,
+ "step": 3680
+ },
+ {
+ "epoch": 51.25130890052356,
+ "grad_norm": 0.21283455193042755,
+ "learning_rate": 0.0006,
+ "loss": 4.763482666015625,
+ "step": 3690
+ },
+ {
+ "epoch": 51.390924956369986,
+ "grad_norm": 0.22551512718200684,
+ "learning_rate": 0.0006,
+ "loss": 4.777619171142578,
+ "step": 3700
+ },
+ {
+ "epoch": 51.5305410122164,
+ "grad_norm": 0.2295781821012497,
+ "learning_rate": 0.0006,
+ "loss": 4.705105209350586,
+ "step": 3710
+ },
+ {
+ "epoch": 51.67015706806283,
+ "grad_norm": 0.22100993990898132,
+ "learning_rate": 0.0006,
+ "loss": 4.777925109863281,
+ "step": 3720
+ },
+ {
+ "epoch": 51.80977312390925,
+ "grad_norm": 0.1953113079071045,
+ "learning_rate": 0.0006,
+ "loss": 4.743285369873047,
+ "step": 3730
+ },
+ {
+ "epoch": 51.94938917975567,
+ "grad_norm": 0.20731113851070404,
+ "learning_rate": 0.0006,
+ "loss": 4.746595001220703,
+ "step": 3740
+ },
+ {
+ "epoch": 52.0,
+ "eval_loss": 3.8366425037384033,
+ "eval_runtime": 2.6235,
+ "eval_samples_per_second": 930.423,
+ "eval_steps_per_second": 14.865,
+ "step": 3744
+ },
+ {
+ "epoch": 52.083769633507856,
+ "grad_norm": 0.22595086693763733,
+ "learning_rate": 0.0006,
+ "loss": 4.712912368774414,
+ "step": 3750
+ },
+ {
+ "epoch": 52.223385689354274,
+ "grad_norm": 0.24032500386238098,
+ "learning_rate": 0.0006,
+ "loss": 4.720623779296875,
+ "step": 3760
+ },
+ {
+ "epoch": 52.3630017452007,
+ "grad_norm": 0.2121676653623581,
+ "learning_rate": 0.0006,
+ "loss": 4.751589965820313,
+ "step": 3770
+ },
+ {
+ "epoch": 52.50261780104712,
+ "grad_norm": 0.21700821816921234,
+ "learning_rate": 0.0006,
+ "loss": 4.716579437255859,
+ "step": 3780
+ },
+ {
+ "epoch": 52.64223385689354,
+ "grad_norm": 0.21031205356121063,
+ "learning_rate": 0.0006,
+ "loss": 4.728091812133789,
+ "step": 3790
+ },
+ {
+ "epoch": 52.781849912739965,
+ "grad_norm": 0.2244471162557602,
+ "learning_rate": 0.0006,
+ "loss": 4.712991714477539,
+ "step": 3800
+ },
+ {
+ "epoch": 52.92146596858639,
+ "grad_norm": 0.25091612339019775,
+ "learning_rate": 0.0006,
+ "loss": 4.7021934509277346,
+ "step": 3810
+ },
+ {
+ "epoch": 53.0,
+ "eval_loss": 3.8360655307769775,
+ "eval_runtime": 2.6233,
+ "eval_samples_per_second": 930.517,
+ "eval_steps_per_second": 14.867,
+ "step": 3816
+ },
+ {
+ "epoch": 53.05584642233857,
+ "grad_norm": 0.21437247097492218,
+ "learning_rate": 0.0006,
+ "loss": 4.7370155334472654,
+ "step": 3820
+ },
+ {
+ "epoch": 53.19546247818499,
+ "grad_norm": 0.24725942313671112,
+ "learning_rate": 0.0006,
+ "loss": 4.731758499145508,
+ "step": 3830
+ },
+ {
+ "epoch": 53.33507853403141,
+ "grad_norm": 0.20370157063007355,
+ "learning_rate": 0.0006,
+ "loss": 4.7096405029296875,
+ "step": 3840
+ },
+ {
+ "epoch": 53.474694589877835,
+ "grad_norm": 0.26552197337150574,
+ "learning_rate": 0.0006,
+ "loss": 4.779978942871094,
+ "step": 3850
+ },
+ {
+ "epoch": 53.61431064572426,
+ "grad_norm": 0.2116101086139679,
+ "learning_rate": 0.0006,
+ "loss": 4.697503662109375,
+ "step": 3860
+ },
+ {
+ "epoch": 53.753926701570684,
+ "grad_norm": 0.2282920479774475,
+ "learning_rate": 0.0006,
+ "loss": 4.7148277282714846,
+ "step": 3870
+ },
+ {
+ "epoch": 53.8935427574171,
+ "grad_norm": 0.2052813321352005,
+ "learning_rate": 0.0006,
+ "loss": 4.686869049072266,
+ "step": 3880
+ },
+ {
+ "epoch": 54.0,
+ "eval_loss": 3.827383279800415,
+ "eval_runtime": 2.6262,
+ "eval_samples_per_second": 929.495,
+ "eval_steps_per_second": 14.851,
+ "step": 3888
+ },
+ {
+ "epoch": 54.02792321116929,
+ "grad_norm": 0.2415243536233902,
+ "learning_rate": 0.0006,
+ "loss": 4.6765998840332035,
+ "step": 3890
+ },
+ {
+ "epoch": 54.167539267015705,
+ "grad_norm": 0.2280266135931015,
+ "learning_rate": 0.0006,
+ "loss": 4.710375213623047,
+ "step": 3900
+ },
+ {
+ "epoch": 54.30715532286213,
+ "grad_norm": 0.24030883610248566,
+ "learning_rate": 0.0006,
+ "loss": 4.7074226379394535,
+ "step": 3910
+ },
+ {
+ "epoch": 54.446771378708554,
+ "grad_norm": 0.22903414070606232,
+ "learning_rate": 0.0006,
+ "loss": 4.696738433837891,
+ "step": 3920
+ },
+ {
+ "epoch": 54.58638743455497,
+ "grad_norm": 0.2566727101802826,
+ "learning_rate": 0.0006,
+ "loss": 4.74979248046875,
+ "step": 3930
+ },
+ {
+ "epoch": 54.726003490401396,
+ "grad_norm": 0.23190483450889587,
+ "learning_rate": 0.0006,
+ "loss": 4.714979934692383,
+ "step": 3940
+ },
+ {
+ "epoch": 54.86561954624782,
+ "grad_norm": 0.23417457938194275,
+ "learning_rate": 0.0006,
+ "loss": 4.748725128173828,
+ "step": 3950
+ },
+ {
+ "epoch": 55.0,
+ "grad_norm": 0.2296326756477356,
+ "learning_rate": 0.0006,
+ "loss": 4.722671127319336,
+ "step": 3960
+ },
+ {
+ "epoch": 55.0,
+ "eval_loss": 3.8337364196777344,
+ "eval_runtime": 2.6267,
+ "eval_samples_per_second": 929.29,
+ "eval_steps_per_second": 14.847,
+ "step": 3960
+ },
+ {
+ "epoch": 55.139616055846425,
+ "grad_norm": 0.24937421083450317,
+ "learning_rate": 0.0006,
+ "loss": 4.702664184570312,
+ "step": 3970
+ },
+ {
+ "epoch": 55.27923211169284,
+ "grad_norm": 0.23149779438972473,
+ "learning_rate": 0.0006,
+ "loss": 4.6620033264160154,
+ "step": 3980
+ },
+ {
+ "epoch": 55.41884816753927,
+ "grad_norm": 0.20848318934440613,
+ "learning_rate": 0.0006,
+ "loss": 4.703611755371094,
+ "step": 3990
+ },
+ {
+ "epoch": 55.55846422338569,
+ "grad_norm": 0.20032435655593872,
+ "learning_rate": 0.0006,
+ "loss": 4.696936798095703,
+ "step": 4000
+ },
+ {
+ "epoch": 55.69808027923211,
+ "grad_norm": 0.2041357010602951,
+ "learning_rate": 0.0006,
+ "loss": 4.701785659790039,
+ "step": 4010
+ },
+ {
+ "epoch": 55.83769633507853,
+ "grad_norm": 0.22746939957141876,
+ "learning_rate": 0.0006,
+ "loss": 4.714264678955078,
+ "step": 4020
+ },
+ {
+ "epoch": 55.97731239092496,
+ "grad_norm": 0.24209359288215637,
+ "learning_rate": 0.0006,
+ "loss": 4.727444839477539,
+ "step": 4030
+ },
+ {
+ "epoch": 56.0,
+ "eval_loss": 3.8306937217712402,
+ "eval_runtime": 2.6311,
+ "eval_samples_per_second": 927.762,
+ "eval_steps_per_second": 14.823,
+ "step": 4032
+ },
+ {
+ "epoch": 56.11169284467714,
+ "grad_norm": 0.22774094343185425,
+ "learning_rate": 0.0006,
+ "loss": 4.684212493896484,
+ "step": 4040
+ },
+ {
+ "epoch": 56.25130890052356,
+ "grad_norm": 0.2309621274471283,
+ "learning_rate": 0.0006,
+ "loss": 4.686544036865234,
+ "step": 4050
+ },
+ {
+ "epoch": 56.390924956369986,
+ "grad_norm": 0.24732667207717896,
+ "learning_rate": 0.0006,
+ "loss": 4.674255752563477,
+ "step": 4060
+ },
+ {
+ "epoch": 56.5305410122164,
+ "grad_norm": 0.24624750018119812,
+ "learning_rate": 0.0006,
+ "loss": 4.70080337524414,
+ "step": 4070
+ },
+ {
+ "epoch": 56.67015706806283,
+ "grad_norm": 0.20607540011405945,
+ "learning_rate": 0.0006,
+ "loss": 4.7326210021972654,
+ "step": 4080
+ },
+ {
+ "epoch": 56.80977312390925,
+ "grad_norm": 0.20070451498031616,
+ "learning_rate": 0.0006,
+ "loss": 4.726334762573242,
+ "step": 4090
+ },
+ {
+ "epoch": 56.94938917975567,
+ "grad_norm": 0.26721078157424927,
+ "learning_rate": 0.0006,
+ "loss": 4.666836547851562,
+ "step": 4100
+ },
+ {
+ "epoch": 57.0,
+ "eval_loss": 3.829385280609131,
+ "eval_runtime": 2.6217,
+ "eval_samples_per_second": 931.081,
+ "eval_steps_per_second": 14.876,
+ "step": 4104
+ },
+ {
+ "epoch": 57.083769633507856,
+ "grad_norm": 0.24299316108226776,
+ "learning_rate": 0.0006,
+ "loss": 4.620271301269531,
+ "step": 4110
+ },
+ {
+ "epoch": 57.223385689354274,
+ "grad_norm": 0.21570660173892975,
+ "learning_rate": 0.0006,
+ "loss": 4.6212913513183596,
+ "step": 4120
+ },
+ {
+ "epoch": 57.3630017452007,
+ "grad_norm": 0.22026455402374268,
+ "learning_rate": 0.0006,
+ "loss": 4.694970703125,
+ "step": 4130
+ },
+ {
+ "epoch": 57.50261780104712,
+ "grad_norm": 0.23656229674816132,
+ "learning_rate": 0.0006,
+ "loss": 4.723772430419922,
+ "step": 4140
+ },
+ {
+ "epoch": 57.64223385689354,
+ "grad_norm": 0.21035727858543396,
+ "learning_rate": 0.0006,
+ "loss": 4.703632736206055,
+ "step": 4150
+ },
+ {
+ "epoch": 57.781849912739965,
+ "grad_norm": 0.20874467492103577,
+ "learning_rate": 0.0006,
+ "loss": 4.690879058837891,
+ "step": 4160
+ },
+ {
+ "epoch": 57.92146596858639,
+ "grad_norm": 0.21504655480384827,
+ "learning_rate": 0.0006,
+ "loss": 4.739430236816406,
+ "step": 4170
+ },
+ {
+ "epoch": 58.0,
+ "eval_loss": 3.826014280319214,
+ "eval_runtime": 2.6445,
+ "eval_samples_per_second": 923.04,
+ "eval_steps_per_second": 14.747,
+ "step": 4176
+ },
+ {
+ "epoch": 58.05584642233857,
+ "grad_norm": 0.24393349885940552,
+ "learning_rate": 0.0006,
+ "loss": 4.6598876953125,
+ "step": 4180
+ },
+ {
+ "epoch": 58.19546247818499,
+ "grad_norm": 0.21980427205562592,
+ "learning_rate": 0.0006,
+ "loss": 4.607953643798828,
+ "step": 4190
+ },
+ {
+ "epoch": 58.33507853403141,
+ "grad_norm": 0.21421635150909424,
+ "learning_rate": 0.0006,
+ "loss": 4.738935852050782,
+ "step": 4200
+ },
+ {
+ "epoch": 58.474694589877835,
+ "grad_norm": 0.23230312764644623,
+ "learning_rate": 0.0006,
+ "loss": 4.688872528076172,
+ "step": 4210
+ },
+ {
+ "epoch": 58.61431064572426,
+ "grad_norm": 0.21955081820487976,
+ "learning_rate": 0.0006,
+ "loss": 4.701759719848633,
+ "step": 4220
+ },
+ {
+ "epoch": 58.753926701570684,
+ "grad_norm": 0.22231166064739227,
+ "learning_rate": 0.0006,
+ "loss": 4.6752769470214846,
+ "step": 4230
+ },
+ {
+ "epoch": 58.8935427574171,
+ "grad_norm": 0.2188156396150589,
+ "learning_rate": 0.0006,
+ "loss": 4.685034942626953,
+ "step": 4240
+ },
+ {
+ "epoch": 59.0,
+ "eval_loss": 3.825808525085449,
+ "eval_runtime": 2.6318,
+ "eval_samples_per_second": 927.49,
+ "eval_steps_per_second": 14.819,
+ "step": 4248
+ },
+ {
+ "epoch": 59.02792321116929,
+ "grad_norm": 0.2534332573413849,
+ "learning_rate": 0.0006,
+ "loss": 4.666550064086914,
+ "step": 4250
+ },
+ {
+ "epoch": 59.167539267015705,
+ "grad_norm": 0.24862292408943176,
+ "learning_rate": 0.0006,
+ "loss": 4.668922424316406,
+ "step": 4260
+ },
+ {
+ "epoch": 59.30715532286213,
+ "grad_norm": 0.22129274904727936,
+ "learning_rate": 0.0006,
+ "loss": 4.691058349609375,
+ "step": 4270
+ },
+ {
+ "epoch": 59.446771378708554,
+ "grad_norm": 0.23209738731384277,
+ "learning_rate": 0.0006,
+ "loss": 4.618634414672852,
+ "step": 4280
+ },
+ {
+ "epoch": 59.58638743455497,
+ "grad_norm": 0.23824480175971985,
+ "learning_rate": 0.0006,
+ "loss": 4.672640991210938,
+ "step": 4290
+ },
+ {
+ "epoch": 59.726003490401396,
+ "grad_norm": 0.2508269250392914,
+ "learning_rate": 0.0006,
+ "loss": 4.705478286743164,
+ "step": 4300
+ },
+ {
+ "epoch": 59.86561954624782,
+ "grad_norm": 0.22110970318317413,
+ "learning_rate": 0.0006,
+ "loss": 4.707386016845703,
+ "step": 4310
+ },
+ {
+ "epoch": 60.0,
+ "grad_norm": 0.2587730884552002,
+ "learning_rate": 0.0006,
+ "loss": 4.765533447265625,
+ "step": 4320
+ },
+ {
+ "epoch": 60.0,
+ "eval_loss": 3.8281450271606445,
+ "eval_runtime": 2.6279,
+ "eval_samples_per_second": 928.889,
+ "eval_steps_per_second": 14.841,
+ "step": 4320
+ },
+ {
+ "epoch": 60.139616055846425,
+ "grad_norm": 0.22804045677185059,
+ "learning_rate": 0.0006,
+ "loss": 4.631346130371094,
+ "step": 4330
+ },
+ {
+ "epoch": 60.27923211169284,
+ "grad_norm": 0.22687020897865295,
+ "learning_rate": 0.0006,
+ "loss": 4.642991256713867,
+ "step": 4340
+ },
+ {
+ "epoch": 60.41884816753927,
+ "grad_norm": 0.21560311317443848,
+ "learning_rate": 0.0006,
+ "loss": 4.652992248535156,
+ "step": 4350
+ },
+ {
+ "epoch": 60.55846422338569,
+ "grad_norm": 0.3004172146320343,
+ "learning_rate": 0.0006,
+ "loss": 4.663480377197265,
+ "step": 4360
+ },
+ {
+ "epoch": 60.69808027923211,
+ "grad_norm": 0.20911958813667297,
+ "learning_rate": 0.0006,
+ "loss": 4.6610462188720705,
+ "step": 4370
+ },
+ {
+ "epoch": 60.83769633507853,
+ "grad_norm": 0.21075640618801117,
+ "learning_rate": 0.0006,
+ "loss": 4.732083892822265,
+ "step": 4380
+ },
+ {
+ "epoch": 60.97731239092496,
+ "grad_norm": 0.22661499679088593,
+ "learning_rate": 0.0006,
+ "loss": 4.659018707275391,
+ "step": 4390
+ },
+ {
+ "epoch": 61.0,
+ "eval_loss": 3.819471597671509,
+ "eval_runtime": 2.6206,
+ "eval_samples_per_second": 931.454,
+ "eval_steps_per_second": 14.882,
+ "step": 4392
+ },
+ {
+ "epoch": 61.11169284467714,
+ "grad_norm": 0.2081788182258606,
+ "learning_rate": 0.0006,
+ "loss": 4.633404541015625,
+ "step": 4400
+ },
+ {
+ "epoch": 61.25130890052356,
+ "grad_norm": 0.22514978051185608,
+ "learning_rate": 0.0006,
+ "loss": 4.648419189453125,
+ "step": 4410
+ },
+ {
+ "epoch": 61.390924956369986,
+ "grad_norm": 0.25621235370635986,
+ "learning_rate": 0.0006,
+ "loss": 4.614633941650391,
+ "step": 4420
+ },
+ {
+ "epoch": 61.5305410122164,
+ "grad_norm": 0.25171685218811035,
+ "learning_rate": 0.0006,
+ "loss": 4.706722259521484,
+ "step": 4430
+ },
+ {
+ "epoch": 61.67015706806283,
+ "grad_norm": 0.24947360157966614,
+ "learning_rate": 0.0006,
+ "loss": 4.701406097412109,
+ "step": 4440
+ },
+ {
+ "epoch": 61.80977312390925,
+ "grad_norm": 0.20666071772575378,
+ "learning_rate": 0.0006,
+ "loss": 4.707528305053711,
+ "step": 4450
+ },
+ {
+ "epoch": 61.94938917975567,
+ "grad_norm": 0.21224376559257507,
+ "learning_rate": 0.0006,
+ "loss": 4.685092163085938,
+ "step": 4460
+ },
+ {
+ "epoch": 62.0,
+ "eval_loss": 3.822190523147583,
+ "eval_runtime": 2.6288,
+ "eval_samples_per_second": 928.571,
+ "eval_steps_per_second": 14.836,
+ "step": 4464
+ },
+ {
+ "epoch": 62.083769633507856,
+ "grad_norm": 0.26400279998779297,
+ "learning_rate": 0.0006,
+ "loss": 4.702872848510742,
+ "step": 4470
+ },
+ {
+ "epoch": 62.223385689354274,
+ "grad_norm": 0.23932775855064392,
+ "learning_rate": 0.0006,
+ "loss": 4.6926933288574215,
+ "step": 4480
+ },
+ {
+ "epoch": 62.3630017452007,
+ "grad_norm": 0.2537601590156555,
+ "learning_rate": 0.0006,
+ "loss": 4.650541305541992,
+ "step": 4490
+ },
+ {
+ "epoch": 62.50261780104712,
+ "grad_norm": 0.21822933852672577,
+ "learning_rate": 0.0006,
+ "loss": 4.680432510375977,
+ "step": 4500
+ },
+ {
+ "epoch": 62.64223385689354,
+ "grad_norm": 0.20359112322330475,
+ "learning_rate": 0.0006,
+ "loss": 4.725475311279297,
+ "step": 4510
+ },
+ {
+ "epoch": 62.781849912739965,
+ "grad_norm": 0.2199525535106659,
+ "learning_rate": 0.0006,
+ "loss": 4.718374633789063,
+ "step": 4520
+ },
+ {
+ "epoch": 62.92146596858639,
+ "grad_norm": 0.23217497766017914,
+ "learning_rate": 0.0006,
+ "loss": 4.721165466308594,
+ "step": 4530
+ },
+ {
+ "epoch": 63.0,
+ "eval_loss": 3.8256478309631348,
+ "eval_runtime": 2.631,
+ "eval_samples_per_second": 927.777,
+ "eval_steps_per_second": 14.823,
+ "step": 4536
+ },
+ {
+ "epoch": 63.05584642233857,
+ "grad_norm": 0.27743974328041077,
+ "learning_rate": 0.0006,
+ "loss": 4.656081008911133,
+ "step": 4540
+ },
+ {
+ "epoch": 63.19546247818499,
+ "grad_norm": 0.2188839614391327,
+ "learning_rate": 0.0006,
+ "loss": 4.617463684082031,
+ "step": 4550
+ },
+ {
+ "epoch": 63.33507853403141,
+ "grad_norm": 0.22164078056812286,
+ "learning_rate": 0.0006,
+ "loss": 4.662931823730469,
+ "step": 4560
+ },
+ {
+ "epoch": 63.474694589877835,
+ "grad_norm": 0.23727895319461823,
+ "learning_rate": 0.0006,
+ "loss": 4.669344329833985,
+ "step": 4570
+ },
+ {
+ "epoch": 63.61431064572426,
+ "grad_norm": 0.28194838762283325,
+ "learning_rate": 0.0006,
+ "loss": 4.693380737304688,
+ "step": 4580
+ },
+ {
+ "epoch": 63.753926701570684,
+ "grad_norm": 0.21883027255535126,
+ "learning_rate": 0.0006,
+ "loss": 4.637828063964844,
+ "step": 4590
+ },
+ {
+ "epoch": 63.8935427574171,
+ "grad_norm": 0.20807673037052155,
+ "learning_rate": 0.0006,
+ "loss": 4.6886749267578125,
+ "step": 4600
+ },
+ {
+ "epoch": 64.0,
+ "eval_loss": 3.8195042610168457,
+ "eval_runtime": 2.6169,
+ "eval_samples_per_second": 932.789,
+ "eval_steps_per_second": 14.903,
+ "step": 4608
+ },
+ {
+ "epoch": 64.02792321116928,
+ "grad_norm": 0.2278340756893158,
+ "learning_rate": 0.0006,
+ "loss": 4.609199523925781,
+ "step": 4610
+ },
+ {
+ "epoch": 64.16753926701571,
+ "grad_norm": 0.21193623542785645,
+ "learning_rate": 0.0006,
+ "loss": 4.6198688507080075,
+ "step": 4620
+ },
+ {
+ "epoch": 64.30715532286213,
+ "grad_norm": 0.24349398910999298,
+ "learning_rate": 0.0006,
+ "loss": 4.667620086669922,
+ "step": 4630
+ },
+ {
+ "epoch": 64.44677137870855,
+ "grad_norm": 0.21560658514499664,
+ "learning_rate": 0.0006,
+ "loss": 4.651830673217773,
+ "step": 4640
+ },
+ {
+ "epoch": 64.58638743455498,
+ "grad_norm": 0.24399498105049133,
+ "learning_rate": 0.0006,
+ "loss": 4.677852630615234,
+ "step": 4650
+ },
+ {
+ "epoch": 64.7260034904014,
+ "grad_norm": 0.21118076145648956,
+ "learning_rate": 0.0006,
+ "loss": 4.6762855529785154,
+ "step": 4660
+ },
+ {
+ "epoch": 64.86561954624781,
+ "grad_norm": 0.22841718792915344,
+ "learning_rate": 0.0006,
+ "loss": 4.697907257080078,
+ "step": 4670
+ },
+ {
+ "epoch": 65.0,
+ "grad_norm": 0.2611827254295349,
+ "learning_rate": 0.0006,
+ "loss": 4.68389663696289,
+ "step": 4680
+ },
+ {
+ "epoch": 65.0,
+ "eval_loss": 3.822420358657837,
+ "eval_runtime": 2.6418,
+ "eval_samples_per_second": 923.975,
+ "eval_steps_per_second": 14.762,
+ "step": 4680
+ },
+ {
+ "epoch": 65.13961605584642,
+ "grad_norm": 0.23498481512069702,
+ "learning_rate": 0.0006,
+ "loss": 4.647805786132812,
+ "step": 4690
+ },
+ {
+ "epoch": 65.27923211169285,
+ "grad_norm": 0.26314952969551086,
+ "learning_rate": 0.0006,
+ "loss": 4.626142120361328,
+ "step": 4700
+ },
+ {
+ "epoch": 65.41884816753927,
+ "grad_norm": 0.22402924299240112,
+ "learning_rate": 0.0006,
+ "loss": 4.6020454406738285,
+ "step": 4710
+ },
+ {
+ "epoch": 65.55846422338568,
+ "grad_norm": 0.24825657904148102,
+ "learning_rate": 0.0006,
+ "loss": 4.666808700561523,
+ "step": 4720
+ },
+ {
+ "epoch": 65.69808027923212,
+ "grad_norm": 0.22201190888881683,
+ "learning_rate": 0.0006,
+ "loss": 4.602214050292969,
+ "step": 4730
+ },
+ {
+ "epoch": 65.83769633507853,
+ "grad_norm": 0.25751712918281555,
+ "learning_rate": 0.0006,
+ "loss": 4.747171783447266,
+ "step": 4740
+ },
+ {
+ "epoch": 65.97731239092495,
+ "grad_norm": 0.2735569477081299,
+ "learning_rate": 0.0006,
+ "loss": 4.738280487060547,
+ "step": 4750
+ },
+ {
+ "epoch": 66.0,
+ "eval_loss": 3.8257622718811035,
+ "eval_runtime": 2.6313,
+ "eval_samples_per_second": 927.696,
+ "eval_steps_per_second": 14.822,
+ "step": 4752
+ },
+ {
+ "epoch": 66.11169284467714,
+ "grad_norm": 0.21869422495365143,
+ "learning_rate": 0.0006,
+ "loss": 4.643059158325196,
+ "step": 4760
+ },
+ {
+ "epoch": 66.25130890052355,
+ "grad_norm": 0.23635639250278473,
+ "learning_rate": 0.0006,
+ "loss": 4.6363269805908205,
+ "step": 4770
+ },
+ {
+ "epoch": 66.39092495636999,
+ "grad_norm": 0.22751888632774353,
+ "learning_rate": 0.0006,
+ "loss": 4.654495620727539,
+ "step": 4780
+ },
+ {
+ "epoch": 66.5305410122164,
+ "grad_norm": 0.23424245417118073,
+ "learning_rate": 0.0006,
+ "loss": 4.654530334472656,
+ "step": 4790
+ },
+ {
+ "epoch": 66.67015706806282,
+ "grad_norm": 0.2226940393447876,
+ "learning_rate": 0.0006,
+ "loss": 4.656908798217773,
+ "step": 4800
+ },
+ {
+ "epoch": 66.80977312390925,
+ "grad_norm": 0.21243661642074585,
+ "learning_rate": 0.0006,
+ "loss": 4.59869155883789,
+ "step": 4810
+ },
+ {
+ "epoch": 66.94938917975567,
+ "grad_norm": 0.210525244474411,
+ "learning_rate": 0.0006,
+ "loss": 4.657334899902343,
+ "step": 4820
+ },
+ {
+ "epoch": 67.0,
+ "eval_loss": 3.814976930618286,
+ "eval_runtime": 2.6215,
+ "eval_samples_per_second": 931.135,
+ "eval_steps_per_second": 14.877,
+ "step": 4824
+ },
+ {
+ "epoch": 67.08376963350786,
+ "grad_norm": 0.2847031056880951,
+ "learning_rate": 0.0006,
+ "loss": 4.612943267822265,
+ "step": 4830
+ },
+ {
+ "epoch": 67.22338568935427,
+ "grad_norm": 0.23910482227802277,
+ "learning_rate": 0.0006,
+ "loss": 4.666780853271485,
+ "step": 4840
+ },
+ {
+ "epoch": 67.36300174520069,
+ "grad_norm": 0.2791076600551605,
+ "learning_rate": 0.0006,
+ "loss": 4.618587493896484,
+ "step": 4850
+ },
+ {
+ "epoch": 67.50261780104712,
+ "grad_norm": 0.2432953417301178,
+ "learning_rate": 0.0006,
+ "loss": 4.593872833251953,
+ "step": 4860
+ },
+ {
+ "epoch": 67.64223385689354,
+ "grad_norm": 0.21688038110733032,
+ "learning_rate": 0.0006,
+ "loss": 4.639530944824219,
+ "step": 4870
+ },
+ {
+ "epoch": 67.78184991273997,
+ "grad_norm": 0.22923719882965088,
+ "learning_rate": 0.0006,
+ "loss": 4.636190414428711,
+ "step": 4880
+ },
+ {
+ "epoch": 67.92146596858639,
+ "grad_norm": 0.22068427503108978,
+ "learning_rate": 0.0006,
+ "loss": 4.612844085693359,
+ "step": 4890
+ },
+ {
+ "epoch": 68.0,
+ "eval_loss": 3.817354440689087,
+ "eval_runtime": 2.6293,
+ "eval_samples_per_second": 928.39,
+ "eval_steps_per_second": 14.833,
+ "step": 4896
+ },
+ {
+ "epoch": 68.05584642233858,
+ "grad_norm": 0.2689245641231537,
+ "learning_rate": 0.0006,
+ "loss": 4.601172637939453,
+ "step": 4900
+ },
+ {
+ "epoch": 68.195462478185,
+ "grad_norm": 0.24516913294792175,
+ "learning_rate": 0.0006,
+ "loss": 4.608419799804688,
+ "step": 4910
+ },
+ {
+ "epoch": 68.33507853403141,
+ "grad_norm": 0.21611154079437256,
+ "learning_rate": 0.0006,
+ "loss": 4.599663925170899,
+ "step": 4920
+ },
+ {
+ "epoch": 68.47469458987784,
+ "grad_norm": 0.2209528535604477,
+ "learning_rate": 0.0006,
+ "loss": 4.675615692138672,
+ "step": 4930
+ },
+ {
+ "epoch": 68.61431064572426,
+ "grad_norm": 0.26648378372192383,
+ "learning_rate": 0.0006,
+ "loss": 4.687446594238281,
+ "step": 4940
+ },
+ {
+ "epoch": 68.75392670157068,
+ "grad_norm": 0.23374858498573303,
+ "learning_rate": 0.0006,
+ "loss": 4.651599884033203,
+ "step": 4950
+ },
+ {
+ "epoch": 68.89354275741711,
+ "grad_norm": 0.228072851896286,
+ "learning_rate": 0.0006,
+ "loss": 4.590849304199219,
+ "step": 4960
+ },
+ {
+ "epoch": 69.0,
+ "eval_loss": 3.8172595500946045,
+ "eval_runtime": 2.6438,
+ "eval_samples_per_second": 923.309,
+ "eval_steps_per_second": 14.752,
+ "step": 4968
+ },
+ {
+ "epoch": 69.02792321116928,
+ "grad_norm": 0.2559078633785248,
+ "learning_rate": 0.0006,
+ "loss": 4.6411384582519535,
+ "step": 4970
+ },
+ {
+ "epoch": 69.16753926701571,
+ "grad_norm": 0.23780183494091034,
+ "learning_rate": 0.0006,
+ "loss": 4.612059020996094,
+ "step": 4980
+ },
+ {
+ "epoch": 69.30715532286213,
+ "grad_norm": 0.24188317358493805,
+ "learning_rate": 0.0006,
+ "loss": 4.570114517211914,
+ "step": 4990
+ },
+ {
+ "epoch": 69.44677137870855,
+ "grad_norm": 0.2245684117078781,
+ "learning_rate": 0.0006,
+ "loss": 4.595469284057617,
+ "step": 5000
+ },
+ {
+ "epoch": 69.58638743455498,
+ "grad_norm": 0.258552610874176,
+ "learning_rate": 0.0006,
+ "loss": 4.69404296875,
+ "step": 5010
+ },
+ {
+ "epoch": 69.7260034904014,
+ "grad_norm": 0.2237822264432907,
+ "learning_rate": 0.0006,
+ "loss": 4.620712661743164,
+ "step": 5020
+ },
+ {
+ "epoch": 69.86561954624781,
+ "grad_norm": 0.22875608503818512,
+ "learning_rate": 0.0006,
+ "loss": 4.648826599121094,
+ "step": 5030
+ },
+ {
+ "epoch": 70.0,
+ "grad_norm": 0.2256709784269333,
+ "learning_rate": 0.0006,
+ "loss": 4.591566467285157,
+ "step": 5040
+ },
+ {
+ "epoch": 70.0,
+ "eval_loss": 3.818233013153076,
+ "eval_runtime": 2.6185,
+ "eval_samples_per_second": 932.213,
+ "eval_steps_per_second": 14.894,
+ "step": 5040
+ },
+ {
+ "epoch": 70.13961605584642,
+ "grad_norm": 0.22915765643119812,
+ "learning_rate": 0.0006,
+ "loss": 4.569583129882813,
+ "step": 5050
+ },
+ {
+ "epoch": 70.27923211169285,
+ "grad_norm": 0.28833693265914917,
+ "learning_rate": 0.0006,
+ "loss": 4.599281692504883,
+ "step": 5060
+ },
+ {
+ "epoch": 70.41884816753927,
+ "grad_norm": 0.25256142020225525,
+ "learning_rate": 0.0006,
+ "loss": 4.673934555053711,
+ "step": 5070
+ },
+ {
+ "epoch": 70.55846422338568,
+ "grad_norm": 0.27048975229263306,
+ "learning_rate": 0.0006,
+ "loss": 4.650726318359375,
+ "step": 5080
+ },
+ {
+ "epoch": 70.69808027923212,
+ "grad_norm": 0.23120802640914917,
+ "learning_rate": 0.0006,
+ "loss": 4.623807525634765,
+ "step": 5090
+ },
+ {
+ "epoch": 70.83769633507853,
+ "grad_norm": 0.2553875148296356,
+ "learning_rate": 0.0006,
+ "loss": 4.683557510375977,
+ "step": 5100
+ },
+ {
+ "epoch": 70.97731239092495,
+ "grad_norm": 0.23946985602378845,
+ "learning_rate": 0.0006,
+ "loss": 4.671528244018555,
+ "step": 5110
+ },
+ {
+ "epoch": 71.0,
+ "eval_loss": 3.816521406173706,
+ "eval_runtime": 2.62,
+ "eval_samples_per_second": 931.678,
+ "eval_steps_per_second": 14.885,
+ "step": 5112
+ },
+ {
+ "epoch": 71.11169284467714,
+ "grad_norm": 0.2694677412509918,
+ "learning_rate": 0.0006,
+ "loss": 4.649788665771484,
+ "step": 5120
+ },
+ {
+ "epoch": 71.25130890052355,
+ "grad_norm": 0.26981931924819946,
+ "learning_rate": 0.0006,
+ "loss": 4.632007217407226,
+ "step": 5130
+ },
+ {
+ "epoch": 71.39092495636999,
+ "grad_norm": 0.2278425395488739,
+ "learning_rate": 0.0006,
+ "loss": 4.630437850952148,
+ "step": 5140
+ },
+ {
+ "epoch": 71.5305410122164,
+ "grad_norm": 0.2156454622745514,
+ "learning_rate": 0.0006,
+ "loss": 4.621918487548828,
+ "step": 5150
+ },
+ {
+ "epoch": 71.67015706806282,
+ "grad_norm": 0.24177463352680206,
+ "learning_rate": 0.0006,
+ "loss": 4.643003845214844,
+ "step": 5160
+ },
+ {
+ "epoch": 71.80977312390925,
+ "grad_norm": 0.24491575360298157,
+ "learning_rate": 0.0006,
+ "loss": 4.66778564453125,
+ "step": 5170
+ },
+ {
+ "epoch": 71.94938917975567,
+ "grad_norm": 0.21527987718582153,
+ "learning_rate": 0.0006,
+ "loss": 4.608951568603516,
+ "step": 5180
+ },
+ {
+ "epoch": 72.0,
+ "eval_loss": 3.8198604583740234,
+ "eval_runtime": 2.6236,
+ "eval_samples_per_second": 930.407,
+ "eval_steps_per_second": 14.865,
+ "step": 5184
+ },
+ {
+ "epoch": 72.08376963350786,
+ "grad_norm": 0.294766902923584,
+ "learning_rate": 0.0006,
+ "loss": 4.628131103515625,
+ "step": 5190
+ },
+ {
+ "epoch": 72.22338568935427,
+ "grad_norm": 0.2562844753265381,
+ "learning_rate": 0.0006,
+ "loss": 4.574409866333008,
+ "step": 5200
+ },
+ {
+ "epoch": 72.36300174520069,
+ "grad_norm": 0.23244014382362366,
+ "learning_rate": 0.0006,
+ "loss": 4.605133819580078,
+ "step": 5210
+ },
+ {
+ "epoch": 72.50261780104712,
+ "grad_norm": 0.22150513529777527,
+ "learning_rate": 0.0006,
+ "loss": 4.609692764282227,
+ "step": 5220
+ },
+ {
+ "epoch": 72.64223385689354,
+ "grad_norm": 0.23824742436408997,
+ "learning_rate": 0.0006,
+ "loss": 4.654159164428711,
+ "step": 5230
+ },
+ {
+ "epoch": 72.78184991273997,
+ "grad_norm": 0.24856653809547424,
+ "learning_rate": 0.0006,
+ "loss": 4.716836929321289,
+ "step": 5240
+ },
+ {
+ "epoch": 72.92146596858639,
+ "grad_norm": 0.24326862394809723,
+ "learning_rate": 0.0006,
+ "loss": 4.560251998901367,
+ "step": 5250
+ },
+ {
+ "epoch": 73.0,
+ "eval_loss": 3.8162474632263184,
+ "eval_runtime": 2.6218,
+ "eval_samples_per_second": 931.052,
+ "eval_steps_per_second": 14.875,
+ "step": 5256
+ },
+ {
+ "epoch": 73.05584642233858,
+ "grad_norm": 0.22982092201709747,
+ "learning_rate": 0.0006,
+ "loss": 4.61522216796875,
+ "step": 5260
+ },
+ {
+ "epoch": 73.195462478185,
+ "grad_norm": 0.23327194154262543,
+ "learning_rate": 0.0006,
+ "loss": 4.681885528564453,
+ "step": 5270
+ },
+ {
+ "epoch": 73.33507853403141,
+ "grad_norm": 0.23683564364910126,
+ "learning_rate": 0.0006,
+ "loss": 4.5400390625,
+ "step": 5280
+ },
+ {
+ "epoch": 73.47469458987784,
+ "grad_norm": 0.23709233105182648,
+ "learning_rate": 0.0006,
+ "loss": 4.607470703125,
+ "step": 5290
+ },
+ {
+ "epoch": 73.61431064572426,
+ "grad_norm": 0.23746120929718018,
+ "learning_rate": 0.0006,
+ "loss": 4.663421249389648,
+ "step": 5300
+ },
+ {
+ "epoch": 73.75392670157068,
+ "grad_norm": 0.25837641954421997,
+ "learning_rate": 0.0006,
+ "loss": 4.606998443603516,
+ "step": 5310
+ },
+ {
+ "epoch": 73.89354275741711,
+ "grad_norm": 0.25555235147476196,
+ "learning_rate": 0.0006,
+ "loss": 4.602525329589843,
+ "step": 5320
+ },
+ {
+ "epoch": 74.0,
+ "eval_loss": 3.812615156173706,
+ "eval_runtime": 2.6448,
+ "eval_samples_per_second": 922.93,
+ "eval_steps_per_second": 14.746,
+ "step": 5328
+ },
+ {
+ "epoch": 74.02792321116928,
+ "grad_norm": 0.2339109629392624,
+ "learning_rate": 0.0006,
+ "loss": 4.6449943542480465,
+ "step": 5330
+ },
+ {
+ "epoch": 74.16753926701571,
+ "grad_norm": 0.23981855809688568,
+ "learning_rate": 0.0006,
+ "loss": 4.60916748046875,
+ "step": 5340
+ },
+ {
+ "epoch": 74.30715532286213,
+ "grad_norm": 0.25061002373695374,
+ "learning_rate": 0.0006,
+ "loss": 4.670564270019531,
+ "step": 5350
+ },
+ {
+ "epoch": 74.44677137870855,
+ "grad_norm": 0.25055378675460815,
+ "learning_rate": 0.0006,
+ "loss": 4.641884613037109,
+ "step": 5360
+ },
+ {
+ "epoch": 74.58638743455498,
+ "grad_norm": 0.21823881566524506,
+ "learning_rate": 0.0006,
+ "loss": 4.598336029052734,
+ "step": 5370
+ },
+ {
+ "epoch": 74.7260034904014,
+ "grad_norm": 0.230833038687706,
+ "learning_rate": 0.0006,
+ "loss": 4.574833679199219,
+ "step": 5380
+ },
+ {
+ "epoch": 74.86561954624781,
+ "grad_norm": 0.22848647832870483,
+ "learning_rate": 0.0006,
+ "loss": 4.572506332397461,
+ "step": 5390
+ },
+ {
+ "epoch": 75.0,
+ "grad_norm": 0.3436893820762634,
+ "learning_rate": 0.0006,
+ "loss": 4.678738403320312,
+ "step": 5400
+ },
+ {
+ "epoch": 75.0,
+ "eval_loss": 3.813210964202881,
+ "eval_runtime": 2.6322,
+ "eval_samples_per_second": 927.369,
+ "eval_steps_per_second": 14.817,
+ "step": 5400
+ },
+ {
+ "epoch": 75.13961605584642,
+ "grad_norm": 0.233714759349823,
+ "learning_rate": 0.0006,
+ "loss": 4.534310150146484,
+ "step": 5410
+ },
+ {
+ "epoch": 75.27923211169285,
+ "grad_norm": 0.2582712471485138,
+ "learning_rate": 0.0006,
+ "loss": 4.613925933837891,
+ "step": 5420
+ },
+ {
+ "epoch": 75.41884816753927,
+ "grad_norm": 0.24095703661441803,
+ "learning_rate": 0.0006,
+ "loss": 4.607852172851563,
+ "step": 5430
+ },
+ {
+ "epoch": 75.55846422338568,
+ "grad_norm": 0.246512308716774,
+ "learning_rate": 0.0006,
+ "loss": 4.680973815917969,
+ "step": 5440
+ },
+ {
+ "epoch": 75.69808027923212,
+ "grad_norm": 0.2395925670862198,
+ "learning_rate": 0.0006,
+ "loss": 4.6067626953125,
+ "step": 5450
+ },
+ {
+ "epoch": 75.83769633507853,
+ "grad_norm": 0.23765592277050018,
+ "learning_rate": 0.0006,
+ "loss": 4.644847106933594,
+ "step": 5460
+ },
+ {
+ "epoch": 75.97731239092495,
+ "grad_norm": 0.24615119397640228,
+ "learning_rate": 0.0006,
+ "loss": 4.620822906494141,
+ "step": 5470
+ },
+ {
+ "epoch": 76.0,
+ "eval_loss": 3.8163974285125732,
+ "eval_runtime": 2.6416,
+ "eval_samples_per_second": 924.062,
+ "eval_steps_per_second": 14.764,
+ "step": 5472
+ },
+ {
+ "epoch": 76.11169284467714,
+ "grad_norm": 0.24472106993198395,
+ "learning_rate": 0.0006,
+ "loss": 4.602330017089844,
+ "step": 5480
+ },
+ {
+ "epoch": 76.25130890052355,
+ "grad_norm": 0.21911656856536865,
+ "learning_rate": 0.0006,
+ "loss": 4.612461090087891,
+ "step": 5490
+ },
+ {
+ "epoch": 76.39092495636999,
+ "grad_norm": 0.2760387659072876,
+ "learning_rate": 0.0006,
+ "loss": 4.591555786132813,
+ "step": 5500
+ },
+ {
+ "epoch": 76.5305410122164,
+ "grad_norm": 0.2392030507326126,
+ "learning_rate": 0.0006,
+ "loss": 4.541796875,
+ "step": 5510
+ },
+ {
+ "epoch": 76.67015706806282,
+ "grad_norm": 0.25309398770332336,
+ "learning_rate": 0.0006,
+ "loss": 4.5635730743408205,
+ "step": 5520
+ },
+ {
+ "epoch": 76.80977312390925,
+ "grad_norm": 0.26336631178855896,
+ "learning_rate": 0.0006,
+ "loss": 4.649578094482422,
+ "step": 5530
+ },
+ {
+ "epoch": 76.94938917975567,
+ "grad_norm": 0.25335443019866943,
+ "learning_rate": 0.0006,
+ "loss": 4.648662567138672,
+ "step": 5540
+ },
+ {
+ "epoch": 77.0,
+ "eval_loss": 3.814716339111328,
+ "eval_runtime": 2.6451,
+ "eval_samples_per_second": 922.843,
+ "eval_steps_per_second": 14.744,
+ "step": 5544
+ },
+ {
+ "epoch": 77.08376963350786,
+ "grad_norm": 0.25959572196006775,
+ "learning_rate": 0.0006,
+ "loss": 4.5998588562011715,
+ "step": 5550
+ },
+ {
+ "epoch": 77.22338568935427,
+ "grad_norm": 0.2660267949104309,
+ "learning_rate": 0.0006,
+ "loss": 4.5999900817871096,
+ "step": 5560
+ },
+ {
+ "epoch": 77.36300174520069,
+ "grad_norm": 0.2847590446472168,
+ "learning_rate": 0.0006,
+ "loss": 4.622273254394531,
+ "step": 5570
+ },
+ {
+ "epoch": 77.50261780104712,
+ "grad_norm": 0.2606430649757385,
+ "learning_rate": 0.0006,
+ "loss": 4.624336242675781,
+ "step": 5580
+ },
+ {
+ "epoch": 77.64223385689354,
+ "grad_norm": 0.21769173443317413,
+ "learning_rate": 0.0006,
+ "loss": 4.628031539916992,
+ "step": 5590
+ },
+ {
+ "epoch": 77.78184991273997,
+ "grad_norm": 0.2316882461309433,
+ "learning_rate": 0.0006,
+ "loss": 4.6515556335449215,
+ "step": 5600
+ },
+ {
+ "epoch": 77.92146596858639,
+ "grad_norm": 0.22509649395942688,
+ "learning_rate": 0.0006,
+ "loss": 4.636623382568359,
+ "step": 5610
+ },
+ {
+ "epoch": 78.0,
+ "eval_loss": 3.8160510063171387,
+ "eval_runtime": 2.6282,
+ "eval_samples_per_second": 928.781,
+ "eval_steps_per_second": 14.839,
+ "step": 5616
+ },
+ {
+ "epoch": 78.05584642233858,
+ "grad_norm": 0.25643405318260193,
+ "learning_rate": 0.0006,
+ "loss": 4.623127746582031,
+ "step": 5620
+ },
+ {
+ "epoch": 78.195462478185,
+ "grad_norm": 0.24310752749443054,
+ "learning_rate": 0.0006,
+ "loss": 4.598170471191406,
+ "step": 5630
+ },
+ {
+ "epoch": 78.33507853403141,
+ "grad_norm": 0.22656871378421783,
+ "learning_rate": 0.0006,
+ "loss": 4.604681396484375,
+ "step": 5640
+ },
+ {
+ "epoch": 78.47469458987784,
+ "grad_norm": 0.2452905923128128,
+ "learning_rate": 0.0006,
+ "loss": 4.577301788330078,
+ "step": 5650
+ },
+ {
+ "epoch": 78.61431064572426,
+ "grad_norm": 0.29254090785980225,
+ "learning_rate": 0.0006,
+ "loss": 4.642354965209961,
+ "step": 5660
+ },
+ {
+ "epoch": 78.75392670157068,
+ "grad_norm": 0.25913918018341064,
+ "learning_rate": 0.0006,
+ "loss": 4.633901977539063,
+ "step": 5670
+ },
+ {
+ "epoch": 78.89354275741711,
+ "grad_norm": 0.23167496919631958,
+ "learning_rate": 0.0006,
+ "loss": 4.613640213012696,
+ "step": 5680
+ },
+ {
+ "epoch": 79.0,
+ "eval_loss": 3.8133344650268555,
+ "eval_runtime": 2.6298,
+ "eval_samples_per_second": 928.213,
+ "eval_steps_per_second": 14.83,
+ "step": 5688
+ },
+ {
+ "epoch": 79.02792321116928,
+ "grad_norm": 0.2598387897014618,
+ "learning_rate": 0.0006,
+ "loss": 4.632302856445312,
+ "step": 5690
+ },
+ {
+ "epoch": 79.16753926701571,
+ "grad_norm": 0.27564752101898193,
+ "learning_rate": 0.0006,
+ "loss": 4.541592407226562,
+ "step": 5700
+ },
+ {
+ "epoch": 79.30715532286213,
+ "grad_norm": 0.2656550705432892,
+ "learning_rate": 0.0006,
+ "loss": 4.595349884033203,
+ "step": 5710
+ },
+ {
+ "epoch": 79.44677137870855,
+ "grad_norm": 0.24405606091022491,
+ "learning_rate": 0.0006,
+ "loss": 4.658450317382813,
+ "step": 5720
+ },
+ {
+ "epoch": 79.58638743455498,
+ "grad_norm": 0.2523824870586395,
+ "learning_rate": 0.0006,
+ "loss": 4.630407333374023,
+ "step": 5730
+ },
+ {
+ "epoch": 79.7260034904014,
+ "grad_norm": 0.2422277331352234,
+ "learning_rate": 0.0006,
+ "loss": 4.654317855834961,
+ "step": 5740
+ },
+ {
+ "epoch": 79.86561954624781,
+ "grad_norm": 0.2784562408924103,
+ "learning_rate": 0.0006,
+ "loss": 4.608974456787109,
+ "step": 5750
+ },
+ {
+ "epoch": 80.0,
+ "grad_norm": 0.26027077436447144,
+ "learning_rate": 0.0006,
+ "loss": 4.631936645507812,
+ "step": 5760
+ },
+ {
+ "epoch": 80.0,
+ "eval_loss": 3.8133418560028076,
+ "eval_runtime": 2.6389,
+ "eval_samples_per_second": 925.003,
+ "eval_steps_per_second": 14.779,
+ "step": 5760
+ },
+ {
+ "epoch": 80.13961605584642,
+ "grad_norm": 0.24270443618297577,
+ "learning_rate": 0.0006,
+ "loss": 4.556245803833008,
+ "step": 5770
+ },
+ {
+ "epoch": 80.27923211169285,
+ "grad_norm": 0.25798848271369934,
+ "learning_rate": 0.0006,
+ "loss": 4.581745910644531,
+ "step": 5780
+ },
+ {
+ "epoch": 80.41884816753927,
+ "grad_norm": 0.28767889738082886,
+ "learning_rate": 0.0006,
+ "loss": 4.561468505859375,
+ "step": 5790
+ },
+ {
+ "epoch": 80.55846422338568,
+ "grad_norm": 0.22843380272388458,
+ "learning_rate": 0.0006,
+ "loss": 4.691260147094726,
+ "step": 5800
+ },
+ {
+ "epoch": 80.69808027923212,
+ "grad_norm": 0.2304852455854416,
+ "learning_rate": 0.0006,
+ "loss": 4.611994934082031,
+ "step": 5810
+ },
+ {
+ "epoch": 80.83769633507853,
+ "grad_norm": 0.25335612893104553,
+ "learning_rate": 0.0006,
+ "loss": 4.6227272033691404,
+ "step": 5820
+ },
+ {
+ "epoch": 80.97731239092495,
+ "grad_norm": 0.2381243109703064,
+ "learning_rate": 0.0006,
+ "loss": 4.608074951171875,
+ "step": 5830
+ },
+ {
+ "epoch": 81.0,
+ "eval_loss": 3.81557559967041,
+ "eval_runtime": 2.6436,
+ "eval_samples_per_second": 923.365,
+ "eval_steps_per_second": 14.753,
+ "step": 5832
+ },
+ {
+ "epoch": 81.11169284467714,
+ "grad_norm": 0.24145108461380005,
+ "learning_rate": 0.0006,
+ "loss": 4.598822402954101,
+ "step": 5840
+ },
+ {
+ "epoch": 81.25130890052355,
+ "grad_norm": 0.23835121095180511,
+ "learning_rate": 0.0006,
+ "loss": 4.610679626464844,
+ "step": 5850
+ },
+ {
+ "epoch": 81.39092495636999,
+ "grad_norm": 0.303744912147522,
+ "learning_rate": 0.0006,
+ "loss": 4.6151683807373045,
+ "step": 5860
+ },
+ {
+ "epoch": 81.5305410122164,
+ "grad_norm": 0.24035553634166718,
+ "learning_rate": 0.0006,
+ "loss": 4.55999641418457,
+ "step": 5870
+ },
+ {
+ "epoch": 81.67015706806282,
+ "grad_norm": 0.2239544838666916,
+ "learning_rate": 0.0006,
+ "loss": 4.626630783081055,
+ "step": 5880
+ },
+ {
+ "epoch": 81.80977312390925,
+ "grad_norm": 0.2187451422214508,
+ "learning_rate": 0.0006,
+ "loss": 4.559969329833985,
+ "step": 5890
+ },
+ {
+ "epoch": 81.94938917975567,
+ "grad_norm": 0.2428794652223587,
+ "learning_rate": 0.0006,
+ "loss": 4.567893981933594,
+ "step": 5900
+ },
+ {
+ "epoch": 82.0,
+ "eval_loss": 3.815549612045288,
+ "eval_runtime": 2.6269,
+ "eval_samples_per_second": 929.215,
+ "eval_steps_per_second": 14.846,
+ "step": 5904
+ },
+ {
+ "epoch": 82.08376963350786,
+ "grad_norm": 0.263607382774353,
+ "learning_rate": 0.0006,
+ "loss": 4.57384147644043,
+ "step": 5910
+ },
+ {
+ "epoch": 82.22338568935427,
+ "grad_norm": 0.29102781414985657,
+ "learning_rate": 0.0006,
+ "loss": 4.569207763671875,
+ "step": 5920
+ },
+ {
+ "epoch": 82.36300174520069,
+ "grad_norm": 0.2425755113363266,
+ "learning_rate": 0.0006,
+ "loss": 4.519753265380859,
+ "step": 5930
+ },
+ {
+ "epoch": 82.50261780104712,
+ "grad_norm": 0.2448194921016693,
+ "learning_rate": 0.0006,
+ "loss": 4.6185657501220705,
+ "step": 5940
+ },
+ {
+ "epoch": 82.64223385689354,
+ "grad_norm": 0.23335236310958862,
+ "learning_rate": 0.0006,
+ "loss": 4.583773040771485,
+ "step": 5950
+ },
+ {
+ "epoch": 82.78184991273997,
+ "grad_norm": 0.23363573849201202,
+ "learning_rate": 0.0006,
+ "loss": 4.565946197509765,
+ "step": 5960
+ },
+ {
+ "epoch": 82.92146596858639,
+ "grad_norm": 0.2610592246055603,
+ "learning_rate": 0.0006,
+ "loss": 4.620330047607422,
+ "step": 5970
+ },
+ {
+ "epoch": 83.0,
+ "eval_loss": 3.818089246749878,
+ "eval_runtime": 2.6221,
+ "eval_samples_per_second": 930.944,
+ "eval_steps_per_second": 14.874,
+ "step": 5976
+ },
+ {
+ "epoch": 83.05584642233858,
+ "grad_norm": 0.28086763620376587,
+ "learning_rate": 0.0006,
+ "loss": 4.620018386840821,
+ "step": 5980
+ },
+ {
+ "epoch": 83.195462478185,
+ "grad_norm": 0.24167688190937042,
+ "learning_rate": 0.0006,
+ "loss": 4.547782897949219,
+ "step": 5990
+ },
+ {
+ "epoch": 83.33507853403141,
+ "grad_norm": 0.2810423970222473,
+ "learning_rate": 0.0006,
+ "loss": 4.554836654663086,
+ "step": 6000
+ },
+ {
+ "epoch": 83.47469458987784,
+ "grad_norm": 0.27146396040916443,
+ "learning_rate": 0.0006,
+ "loss": 4.595916748046875,
+ "step": 6010
+ },
+ {
+ "epoch": 83.61431064572426,
+ "grad_norm": 0.2764207422733307,
+ "learning_rate": 0.0006,
+ "loss": 4.586863708496094,
+ "step": 6020
+ },
+ {
+ "epoch": 83.75392670157068,
+ "grad_norm": 0.26348456740379333,
+ "learning_rate": 0.0006,
+ "loss": 4.57434196472168,
+ "step": 6030
+ },
+ {
+ "epoch": 83.89354275741711,
+ "grad_norm": 0.24266548454761505,
+ "learning_rate": 0.0006,
+ "loss": 4.543191528320312,
+ "step": 6040
+ },
+ {
+ "epoch": 84.0,
+ "eval_loss": 3.814624309539795,
+ "eval_runtime": 2.6312,
+ "eval_samples_per_second": 927.714,
+ "eval_steps_per_second": 14.822,
+ "step": 6048
+ },
+ {
+ "epoch": 84.02792321116928,
+ "grad_norm": 0.26932990550994873,
+ "learning_rate": 0.0006,
+ "loss": 4.63446159362793,
+ "step": 6050
+ },
+ {
+ "epoch": 84.16753926701571,
+ "grad_norm": 0.2514069080352783,
+ "learning_rate": 0.0006,
+ "loss": 4.596696472167968,
+ "step": 6060
+ },
+ {
+ "epoch": 84.30715532286213,
+ "grad_norm": 0.25114378333091736,
+ "learning_rate": 0.0006,
+ "loss": 4.566738891601562,
+ "step": 6070
+ },
+ {
+ "epoch": 84.44677137870855,
+ "grad_norm": 0.235788032412529,
+ "learning_rate": 0.0006,
+ "loss": 4.534809112548828,
+ "step": 6080
+ },
+ {
+ "epoch": 84.58638743455498,
+ "grad_norm": 0.22468598186969757,
+ "learning_rate": 0.0006,
+ "loss": 4.6085655212402346,
+ "step": 6090
+ },
+ {
+ "epoch": 84.7260034904014,
+ "grad_norm": 0.2624860107898712,
+ "learning_rate": 0.0006,
+ "loss": 4.567503356933594,
+ "step": 6100
+ },
+ {
+ "epoch": 84.86561954624781,
+ "grad_norm": 0.2914150357246399,
+ "learning_rate": 0.0006,
+ "loss": 4.591632843017578,
+ "step": 6110
+ },
+ {
+ "epoch": 85.0,
+ "grad_norm": 0.2853717803955078,
+ "learning_rate": 0.0006,
+ "loss": 4.612430572509766,
+ "step": 6120
+ },
+ {
+ "epoch": 85.0,
+ "eval_loss": 3.819162607192993,
+ "eval_runtime": 2.6434,
+ "eval_samples_per_second": 923.415,
+ "eval_steps_per_second": 14.753,
+ "step": 6120
+ },
+ {
+ "epoch": 85.13961605584642,
+ "grad_norm": 0.30460819602012634,
+ "learning_rate": 0.0006,
+ "loss": 4.510371398925781,
+ "step": 6130
+ },
+ {
+ "epoch": 85.27923211169285,
+ "grad_norm": 0.29113221168518066,
+ "learning_rate": 0.0006,
+ "loss": 4.57672119140625,
+ "step": 6140
+ },
+ {
+ "epoch": 85.41884816753927,
+ "grad_norm": 0.2544858157634735,
+ "learning_rate": 0.0006,
+ "loss": 4.567066955566406,
+ "step": 6150
+ },
+ {
+ "epoch": 85.55846422338568,
+ "grad_norm": 0.2419157475233078,
+ "learning_rate": 0.0006,
+ "loss": 4.6029613494873045,
+ "step": 6160
+ },
+ {
+ "epoch": 85.69808027923212,
+ "grad_norm": 0.2361719012260437,
+ "learning_rate": 0.0006,
+ "loss": 4.58320198059082,
+ "step": 6170
+ },
+ {
+ "epoch": 85.83769633507853,
+ "grad_norm": 0.24084147810935974,
+ "learning_rate": 0.0006,
+ "loss": 4.628932189941406,
+ "step": 6180
+ },
+ {
+ "epoch": 85.97731239092495,
+ "grad_norm": 0.24800293147563934,
+ "learning_rate": 0.0006,
+ "loss": 4.626763153076172,
+ "step": 6190
+ },
+ {
+ "epoch": 86.0,
+ "eval_loss": 3.81020450592041,
+ "eval_runtime": 2.6306,
+ "eval_samples_per_second": 927.916,
+ "eval_steps_per_second": 14.825,
+ "step": 6192
+ },
+ {
+ "epoch": 86.11169284467714,
+ "grad_norm": 3.9235308170318604,
+ "learning_rate": 0.0006,
+ "loss": 4.591695785522461,
+ "step": 6200
+ },
+ {
+ "epoch": 86.25130890052355,
+ "grad_norm": 2.242257595062256,
+ "learning_rate": 0.0006,
+ "loss": 4.664269256591797,
+ "step": 6210
+ },
+ {
+ "epoch": 86.39092495636999,
+ "grad_norm": 0.9056429862976074,
+ "learning_rate": 0.0006,
+ "loss": 4.7139404296875,
+ "step": 6220
+ },
+ {
+ "epoch": 86.5305410122164,
+ "grad_norm": 0.7155718803405762,
+ "learning_rate": 0.0006,
+ "loss": 4.772116851806641,
+ "step": 6230
+ },
+ {
+ "epoch": 86.67015706806282,
+ "grad_norm": 0.280024915933609,
+ "learning_rate": 0.0006,
+ "loss": 4.753601455688477,
+ "step": 6240
+ },
+ {
+ "epoch": 86.80977312390925,
+ "grad_norm": 0.2076989859342575,
+ "learning_rate": 0.0006,
+ "loss": 4.706963348388672,
+ "step": 6250
+ },
+ {
+ "epoch": 86.94938917975567,
+ "grad_norm": 0.2784584164619446,
+ "learning_rate": 0.0006,
+ "loss": 4.7035774230957035,
+ "step": 6260
+ },
+ {
+ "epoch": 87.0,
+ "eval_loss": 3.830559015274048,
+ "eval_runtime": 2.6299,
+ "eval_samples_per_second": 928.159,
+ "eval_steps_per_second": 14.829,
+ "step": 6264
+ },
+ {
+ "epoch": 87.08376963350786,
+ "grad_norm": 0.22300276160240173,
+ "learning_rate": 0.0006,
+ "loss": 4.6537017822265625,
+ "step": 6270
+ },
+ {
+ "epoch": 87.22338568935427,
+ "grad_norm": 0.20326067507266998,
+ "learning_rate": 0.0006,
+ "loss": 4.5924560546875,
+ "step": 6280
+ },
+ {
+ "epoch": 87.36300174520069,
+ "grad_norm": 0.18952086567878723,
+ "learning_rate": 0.0006,
+ "loss": 4.59033432006836,
+ "step": 6290
+ },
+ {
+ "epoch": 87.50261780104712,
+ "grad_norm": 0.20353400707244873,
+ "learning_rate": 0.0006,
+ "loss": 4.564707946777344,
+ "step": 6300
+ },
+ {
+ "epoch": 87.64223385689354,
+ "grad_norm": 0.1963331252336502,
+ "learning_rate": 0.0006,
+ "loss": 4.6279338836669925,
+ "step": 6310
+ },
+ {
+ "epoch": 87.78184991273997,
+ "grad_norm": 0.1816292256116867,
+ "learning_rate": 0.0006,
+ "loss": 4.602526092529297,
+ "step": 6320
+ },
+ {
+ "epoch": 87.92146596858639,
+ "grad_norm": 0.19760827720165253,
+ "learning_rate": 0.0006,
+ "loss": 4.582072830200195,
+ "step": 6330
+ },
+ {
+ "epoch": 88.0,
+ "eval_loss": 3.8116214275360107,
+ "eval_runtime": 2.6283,
+ "eval_samples_per_second": 928.734,
+ "eval_steps_per_second": 14.838,
+ "step": 6336
+ },
+ {
+ "epoch": 88.05584642233858,
+ "grad_norm": 0.2092871069908142,
+ "learning_rate": 0.0006,
+ "loss": 4.620587921142578,
+ "step": 6340
+ },
+ {
+ "epoch": 88.195462478185,
+ "grad_norm": 0.18753844499588013,
+ "learning_rate": 0.0006,
+ "loss": 4.595518112182617,
+ "step": 6350
+ },
+ {
+ "epoch": 88.33507853403141,
+ "grad_norm": 0.19224905967712402,
+ "learning_rate": 0.0006,
+ "loss": 4.604162979125976,
+ "step": 6360
+ },
+ {
+ "epoch": 88.47469458987784,
+ "grad_norm": 0.19662661850452423,
+ "learning_rate": 0.0006,
+ "loss": 4.525022888183594,
+ "step": 6370
+ },
+ {
+ "epoch": 88.61431064572426,
+ "grad_norm": 0.24526594579219818,
+ "learning_rate": 0.0006,
+ "loss": 4.60724105834961,
+ "step": 6380
+ },
+ {
+ "epoch": 88.75392670157068,
+ "grad_norm": 0.23190777003765106,
+ "learning_rate": 0.0006,
+ "loss": 4.587713623046875,
+ "step": 6390
+ },
+ {
+ "epoch": 88.89354275741711,
+ "grad_norm": 0.20585203170776367,
+ "learning_rate": 0.0006,
+ "loss": 4.6002037048339846,
+ "step": 6400
+ },
+ {
+ "epoch": 89.0,
+ "eval_loss": 3.8122479915618896,
+ "eval_runtime": 2.6358,
+ "eval_samples_per_second": 926.082,
+ "eval_steps_per_second": 14.796,
+ "step": 6408
+ },
+ {
+ "epoch": 89.02792321116928,
+ "grad_norm": 0.23852229118347168,
+ "learning_rate": 0.0006,
+ "loss": 4.5794422149658205,
+ "step": 6410
+ },
+ {
+ "epoch": 89.16753926701571,
+ "grad_norm": 0.2129506766796112,
+ "learning_rate": 0.0006,
+ "loss": 4.527211380004883,
+ "step": 6420
+ },
+ {
+ "epoch": 89.30715532286213,
+ "grad_norm": 0.2272108942270279,
+ "learning_rate": 0.0006,
+ "loss": 4.583210372924805,
+ "step": 6430
+ },
+ {
+ "epoch": 89.44677137870855,
+ "grad_norm": 0.23771491646766663,
+ "learning_rate": 0.0006,
+ "loss": 4.543557357788086,
+ "step": 6440
+ },
+ {
+ "epoch": 89.58638743455498,
+ "grad_norm": 0.2311193197965622,
+ "learning_rate": 0.0006,
+ "loss": 4.611130142211914,
+ "step": 6450
+ },
+ {
+ "epoch": 89.7260034904014,
+ "grad_norm": 0.194376602768898,
+ "learning_rate": 0.0006,
+ "loss": 4.608315658569336,
+ "step": 6460
+ },
+ {
+ "epoch": 89.86561954624781,
+ "grad_norm": 0.2020266205072403,
+ "learning_rate": 0.0006,
+ "loss": 4.600410079956054,
+ "step": 6470
+ },
+ {
+ "epoch": 90.0,
+ "grad_norm": 0.22890925407409668,
+ "learning_rate": 0.0006,
+ "loss": 4.548514175415039,
+ "step": 6480
+ },
+ {
+ "epoch": 90.0,
+ "eval_loss": 3.809403419494629,
+ "eval_runtime": 2.6315,
+ "eval_samples_per_second": 927.591,
+ "eval_steps_per_second": 14.82,
+ "step": 6480
+ },
+ {
+ "epoch": 90.13961605584642,
+ "grad_norm": 0.23181113600730896,
+ "learning_rate": 0.0006,
+ "loss": 4.502600479125976,
+ "step": 6490
+ },
+ {
+ "epoch": 90.27923211169285,
+ "grad_norm": 0.24220430850982666,
+ "learning_rate": 0.0006,
+ "loss": 4.603355789184571,
+ "step": 6500
+ },
+ {
+ "epoch": 90.41884816753927,
+ "grad_norm": 0.2229602187871933,
+ "learning_rate": 0.0006,
+ "loss": 4.5686088562011715,
+ "step": 6510
+ },
+ {
+ "epoch": 90.55846422338568,
+ "grad_norm": 0.2281477302312851,
+ "learning_rate": 0.0006,
+ "loss": 4.564717102050781,
+ "step": 6520
+ },
+ {
+ "epoch": 90.69808027923212,
+ "grad_norm": 0.23453843593597412,
+ "learning_rate": 0.0006,
+ "loss": 4.567852783203125,
+ "step": 6530
+ },
+ {
+ "epoch": 90.83769633507853,
+ "grad_norm": 0.2270478755235672,
+ "learning_rate": 0.0006,
+ "loss": 4.643734741210937,
+ "step": 6540
+ },
+ {
+ "epoch": 90.97731239092495,
+ "grad_norm": 0.2520331144332886,
+ "learning_rate": 0.0006,
+ "loss": 4.588864517211914,
+ "step": 6550
+ },
+ {
+ "epoch": 91.0,
+ "eval_loss": 3.8090758323669434,
+ "eval_runtime": 2.6279,
+ "eval_samples_per_second": 928.861,
+ "eval_steps_per_second": 14.84,
+ "step": 6552
+ }
+ ],
+ "logging_steps": 10,
+ "max_steps": 7200,
+ "num_input_tokens_seen": 0,
+ "num_train_epochs": 100,
+ "save_steps": 500,
+ "stateful_callbacks": {
+ "TrainerControl": {
+ "args": {
+ "should_epoch_stop": false,
+ "should_evaluate": false,
+ "should_log": false,
+ "should_save": false,
+ "should_training_stop": false
+ },
+ "attributes": {}
+ }
+ },
+ "total_flos": 2.795787773183787e+18,
+ "train_batch_size": 8,
+ "trial_name": null,
+ "trial_params": null
+}
diff --git a/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/train.log b/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/train.log
new file mode 100644
index 0000000000000000000000000000000000000000..0f768984f440c2843e5d8dae25091e5b9917e1db
--- /dev/null
+++ b/runs/t1a/t1a_S2_C3_rand5_r2l50_i5exp/train.log
@@ -0,0 +1,5030 @@
+W0818 06:43:40.036000 2597794 torch/distributed/run.py:851]
+W0818 06:43:40.036000 2597794 torch/distributed/run.py:851] *****************************************
+W0818 06:43:40.036000 2597794 torch/distributed/run.py:851] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
+W0818 06:43:40.036000 2597794 torch/distributed/run.py:851] *****************************************
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] loading configuration file config.json from cache at /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-Embedding-0.6B/snapshots/97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3/config.json
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+[transformers] Model config Qwen3Config {
+ "architectures": [
+ "Qwen3ForCausalLM"
+ ],
+ "attention_bias": false,
+ "attention_dropout": 0.0,
+ "bos_token_id": 151643,
+ "dtype": "bfloat16",
+ "eos_token_id": 151643,
+ "head_dim": 128,
+ "hidden_act": "silu",
+ "hidden_size": 1024,
+ "initializer_range": 0.02,
+ "intermediate_size": 3072,
+ "layer_types": [
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention",
+ "full_attention"
+ ],
+ "max_position_embeddings": 32768,
+ "max_window_layers": 28,
+ "model_type": "qwen3",
+ "num_attention_heads": 16,
+ "num_hidden_layers": 28,
+ "num_key_value_heads": 8,
+ "pad_token_id": null,
+ "rms_norm_eps": 1e-06,
+ "rope_parameters": {
+ "rope_theta": 1000000,
+ "rope_type": "default"
+ },
+ "sliding_window": null,
+ "tie_word_embeddings": true,
+ "transformers_version": "5.15.0",
+ "use_cache": true,
+ "use_sliding_window": false,
+ "vocab_size": 151669
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+EOS token: <|im_end|>
+EOS token: <|im_end|>
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+[transformers] Generate config GenerationConfig {
+ "eos_token_id": 151645,
+ "output_attentions": false,
+ "output_hidden_states": false,
+ "pad_token_id": 151645,
+ "use_cache": true
+}
+
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 145836544
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 145836544
+total params: 145836544
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+total params: 145836544
+total params: 145836544
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+total params: 145836544
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[dataset] loaded packed cache: 36614 train / 2441 eval blocks
+[transformers] PyTorch: setting up devices
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] Setting TF32 in CUDA backends to speedup torch compile, you won't see any improvement otherwise.
+[transformers] ***** Running training *****
+[transformers] Num examples = 36,614
+[transformers] Num Epochs = 100
+[transformers] Num update steps per epoch = 72
+[transformers] Instantaneous batch size per device = 8
+[transformers] Total train batch size (w. parallel, distributed & accumulation) = 512
+[transformers] Gradient Accumulation steps = 8
+[transformers] Total optimization steps = 7,200
+[transformers] Number of trainable parameters = 145,836,544
+wandb: WARNING Changes to your `wandb` environment variables will be ignored because your `wandb` session has already started. For more information on how to modify your settings with `wandb.init()` arguments, please refer to https://wandb.me/wandb-init.
+wandb: setting up run 7rc94bqn
+wandb: Tracking run with wandb version 0.28.2
+wandb: Run data is saved locally in /home/ubuntu/workspace/data-augmentations-for-pretraining/wandb/run-20260818_064359-7rc94bqn
+wandb: Run `wandb offline` to turn off syncing.
+wandb: Syncing run t1a_S2_C3_rand5_r2l50_i5exp
+wandb: ⭐️ View project at https://wandb.ai/slowrun-wankers/data-aug-pretraining
+wandb: 🚀 View run at https://wandb.ai/slowrun-wankers/data-aug-pretraining/runs/7rc94bqn
+
0%| | 0/7200 [00:00 will be ignored
+[rank2]:W0818 06:44:23.282000 2597918 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank1]:W0818 06:44:23.308000 2597917 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank6]:W0818 06:44:23.327000 2597922 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank3]:W0818 06:44:23.329000 2597919 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+[rank7]:W0818 06:44:23.446000 2597923 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+[rank5]:W0818 06:44:23.667000 2597921 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+[rank0]:W0818 06:44:24.221000 2597916 torch/_logging/_internal.py:1279] [0/0] Profiler record function will be ignored
+
0%| | 1/7200 [00:40<81:50:10, 40.92s/it]
{'loss': '12.04', 'grad_norm': '2.488', 'learning_rate': '0', 'epoch': '0.01396'}
+
0%| | 1/7200 [00:40<81:50:10, 40.92s/it]
0%| | 2/7200 [00:41<34:28:45, 17.24s/it]
0%| | 3/7200 [00:42<19:18:36, 9.66s/it]
0%| | 4/7200 [00:42<12:10:55, 6.09s/it]
0%| | 5/7200 [00:43<8:14:35, 4.12s/it]
0%| | 6/7200 [00:44<5:52:08, 2.94s/it]
0%| | 7/7200 [00:44<4:21:44, 2.18s/it]
0%| | 8/7200 [00:45<3:22:31, 1.69s/it]
0%| | 9/7200 [00:46<2:42:53, 1.36s/it]
0%| | 10/7200 [00:46<2:15:55, 1.13s/it]
{'loss': '11.71', 'grad_norm': '1.678', 'learning_rate': '5.4e-05', 'epoch': '0.1396'}
+
0%| | 10/7200 [00:46<2:15:55, 1.13s/it]
0%| | 11/7200 [00:47<1:57:30, 1.02it/s]
0%| | 12/7200 [00:47<1:44:44, 1.14it/s]
0%| | 13/7200 [00:48<1:35:53, 1.25it/s]
0%| | 14/7200 [00:49<1:29:45, 1.33it/s]
0%| | 15/7200 [00:49<1:25:30, 1.40it/s]
0%| | 16/7200 [00:50<1:22:29, 1.45it/s]
0%| | 17/7200 [00:51<1:20:23, 1.49it/s]
0%| | 18/7200 [00:51<1:18:54, 1.52it/s]
0%| | 19/7200 [00:52<1:17:52, 1.54it/s]
0%| | 20/7200 [00:52<1:17:07, 1.55it/s]
{'loss': '11.06', 'grad_norm': '1.6', 'learning_rate': '0.000114', 'epoch': '0.2792'}
+
0%| | 20/7200 [00:52<1:17:07, 1.55it/s]
0%| | 21/7200 [00:53<1:16:35, 1.56it/s]
0%| | 22/7200 [00:54<1:16:17, 1.57it/s]
0%| | 23/7200 [00:54<1:16:01, 1.57it/s]
0%| | 24/7200 [00:55<1:15:48, 1.58it/s]
0%| | 25/7200 [00:56<1:15:43, 1.58it/s]
0%| | 26/7200 [00:56<1:15:40, 1.58it/s]
0%| | 27/7200 [00:57<1:15:32, 1.58it/s]
0%| | 28/7200 [00:58<1:15:24, 1.59it/s]
0%| | 29/7200 [00:58<1:15:21, 1.59it/s]
0%| | 30/7200 [00:59<1:15:22, 1.59it/s]
{'loss': '10.34', 'grad_norm': '1.541', 'learning_rate': '0.000174', 'epoch': '0.4188'}
+
0%| | 30/7200 [00:59<1:15:22, 1.59it/s]
0%| | 31/7200 [00:59<1:15:20, 1.59it/s]
0%| | 32/7200 [01:00<1:15:14, 1.59it/s]
0%| | 33/7200 [01:01<1:15:16, 1.59it/s]
0%| | 34/7200 [01:01<1:15:17, 1.59it/s]
0%| | 35/7200 [01:02<1:15:13, 1.59it/s]
0%| | 36/7200 [01:03<1:15:10, 1.59it/s]
1%| | 37/7200 [01:03<1:15:08, 1.59it/s]
1%| | 38/7200 [01:04<1:15:09, 1.59it/s]
1%| | 39/7200 [01:04<1:15:12, 1.59it/s]
1%| | 40/7200 [01:05<1:15:13, 1.59it/s]
{'loss': '9.387', 'grad_norm': '1.367', 'learning_rate': '0.000234', 'epoch': '0.5585'}
+
1%| | 40/7200 [01:05<1:15:13, 1.59it/s]
1%| | 41/7200 [01:06<1:15:12, 1.59it/s]
1%| | 42/7200 [01:06<1:15:08, 1.59it/s]
1%| | 43/7200 [01:07<1:15:04, 1.59it/s]
1%| | 44/7200 [01:08<1:15:02, 1.59it/s]
1%| | 45/7200 [01:08<1:15:03, 1.59it/s]
1%| | 46/7200 [01:09<1:15:07, 1.59it/s]
1%| | 47/7200 [01:09<1:15:06, 1.59it/s]
1%| | 48/7200 [01:10<1:15:04, 1.59it/s]
1%| | 49/7200 [01:11<1:15:03, 1.59it/s]
1%| | 50/7200 [01:11<1:15:04, 1.59it/s]
{'loss': '8.494', 'grad_norm': '0.8155', 'learning_rate': '0.000294', 'epoch': '0.6981'}
+
1%| | 50/7200 [01:11<1:15:04, 1.59it/s]
1%| | 51/7200 [01:12<1:15:05, 1.59it/s]
1%| | 52/7200 [01:13<1:15:05, 1.59it/s]
1%| | 53/7200 [01:13<1:15:06, 1.59it/s]
1%| | 54/7200 [01:14<1:15:04, 1.59it/s]
1%| | 55/7200 [01:15<1:15:01, 1.59it/s]
1%| | 56/7200 [01:15<1:15:00, 1.59it/s]
1%| | 57/7200 [01:16<1:15:06, 1.59it/s]
1%| | 58/7200 [01:16<1:15:06, 1.58it/s]
1%| | 59/7200 [01:17<1:15:05, 1.58it/s]
1%| | 60/7200 [01:18<1:15:08, 1.58it/s]
{'loss': '8.025', 'grad_norm': '0.2593', 'learning_rate': '0.000354', 'epoch': '0.8377'}
+
1%| | 60/7200 [01:18<1:15:08, 1.58it/s]
1%| | 61/7200 [01:18<1:15:07, 1.58it/s]
1%| | 62/7200 [01:19<1:15:04, 1.58it/s]
1%| | 63/7200 [01:20<1:15:04, 1.58it/s]
1%| | 64/7200 [01:20<1:15:02, 1.58it/s]
1%| | 65/7200 [01:21<1:15:01, 1.58it/s]
1%| | 66/7200 [01:21<1:15:04, 1.58it/s]
1%| | 67/7200 [01:22<1:15:08, 1.58it/s]
1%| | 68/7200 [01:23<1:15:08, 1.58it/s]
1%| | 69/7200 [01:23<1:15:12, 1.58it/s]
1%| | 70/7200 [01:24<1:15:16, 1.58it/s]
{'loss': '7.909', 'grad_norm': '0.2733', 'learning_rate': '0.000414', 'epoch': '0.9773'}
+
1%| | 70/7200 [01:24<1:15:16, 1.58it/s]
1%| | 71/7200 [01:25<1:15:13, 1.58it/s]
1%| | 72/7200 [01:25<1:06:50, 1.78it/s][transformers]
+***** Running Evaluation *****
+[transformers] Num examples = 2441
+[transformers] Batch size = 8
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: [wandb.login()] Using explicit session credentials for https://api.wandb.ai.
+wandb: Appending key for api.wandb.ai to your netrc file: /home/ubuntu/.netrc
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+wandb: Currently logged in as: michaelckj (slowrun-wankers) to https://api.wandb.ai. Use `wandb login --relogin` to force relogin
+
+
0%| | 0/39 [00:00