Hugging Face's logo

sparsetrace
/
d32ft

Model card Files
xet
 Community
sparsetrace commited on
Commit
2fe40d9
·
verified ·
1 Parent(s): 9f91853

d32ft mirror latest (40 files)

Browse files
Files changed (40) hide show
  1. latest/meta_000600.json +68 -0
  2. latest/meta_000700.json +68 -0
  3. latest/meta_000800.json +68 -0
  4. latest/meta_000900.json +68 -0
  5. latest/model_000600.pt +3 -0
  6. latest/model_000700.pt +3 -0
  7. latest/model_000800.pt +3 -0
  8. latest/model_000900.pt +3 -0
  9. latest/optim_000600_rank0.pt +3 -0
  10. latest/optim_000600_rank1.pt +3 -0
  11. latest/optim_000600_rank2.pt +3 -0
  12. latest/optim_000600_rank3.pt +3 -0
  13. latest/optim_000600_rank4.pt +3 -0
  14. latest/optim_000600_rank5.pt +3 -0
  15. latest/optim_000600_rank6.pt +3 -0
  16. latest/optim_000600_rank7.pt +3 -0
  17. latest/optim_000700_rank0.pt +3 -0
  18. latest/optim_000700_rank1.pt +3 -0
  19. latest/optim_000700_rank2.pt +3 -0
  20. latest/optim_000700_rank3.pt +3 -0
  21. latest/optim_000700_rank4.pt +3 -0
  22. latest/optim_000700_rank5.pt +3 -0
  23. latest/optim_000700_rank6.pt +3 -0
  24. latest/optim_000700_rank7.pt +3 -0
  25. latest/optim_000800_rank0.pt +3 -0
  26. latest/optim_000800_rank1.pt +3 -0
  27. latest/optim_000800_rank2.pt +3 -0
  28. latest/optim_000800_rank3.pt +3 -0
  29. latest/optim_000800_rank4.pt +3 -0
  30. latest/optim_000800_rank5.pt +3 -0
  31. latest/optim_000800_rank6.pt +3 -0
  32. latest/optim_000800_rank7.pt +3 -0
  33. latest/optim_000900_rank0.pt +3 -0
  34. latest/optim_000900_rank1.pt +3 -0
  35. latest/optim_000900_rank2.pt +3 -0
  36. latest/optim_000900_rank3.pt +3 -0
  37. latest/optim_000900_rank4.pt +3 -0
  38. latest/optim_000900_rank5.pt +3 -0
  39. latest/optim_000900_rank6.pt +3 -0
  40. latest/optim_000900_rank7.pt +3 -0
latest/meta_000600.json ADDED
@@ -0,0 +1,68 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "step": 600,
3
+ "val_bpb": 0.726215773156184,
4
+ "model_config": {
5
+ "sequence_len": 2048,
6
+ "vocab_size": 65536,
7
+ "n_layer": 32,
8
+ "n_head": 16,
9
+ "n_kv_head": 16,
10
+ "n_embd": 2048,
11
+ "window_pattern": "L",
12
+ "attn_variant": "hmap",
13
+ "hmap_alpha": 0.0,
14
+ "witten": false,
15
+ "bidirectional": false
16
+ },
17
+ "user_config": {
18
+ "run": "dummy",
19
+ "device_type": "",
20
+ "fp8": false,
21
+ "fp8_recipe": "tensorwise",
22
+ "depth": 32,
23
+ "aspect_ratio": 64,
24
+ "head_dim": 128,
25
+ "max_seq_len": 2048,
26
+ "window_pattern": "L",
27
+ "attn_variant": "hmap",
28
+ "hmap_alpha": 0.0,
29
+ "witten": false,
30
+ "objective": "ar",
31
+ "mlm_mask_prob": 0.15,
32
+ "num_iterations": 1600,
33
+ "target_flops": -1.0,
34
+ "target_param_data_ratio": 12,
35
+ "device_batch_size": 4,
36
+ "total_batch_size": 524288,
37
+ "embedding_lr": 0.03,
38
+ "unembedding_lr": 0.0008,
39
+ "weight_decay": 0.0,
40
+ "matrix_lr": 0.002,
41
+ "scalar_lr": 0.05,
42
+ "warmup_steps": 40,
43
+ "warmdown_ratio": 0.0,
44
+ "final_lr_frac": 1.0,
45
+ "resume_from_step": 100,
46
+ "init_from_model": "",
47
+ "eval_every": 100,
48
+ "eval_tokens": 41943040,
49
+ "core_metric_every": -1,
50
+ "core_metric_max_per_task": 500,
51
+ "sample_every": 250,
52
+ "save_every": 100,
53
+ "model_tag": "d32ft-amap"
54
+ },
55
+ "device_batch_size": 4,
56
+ "max_seq_len": 2048,
57
+ "total_batch_size": 524288,
58
+ "dataloader_state_dict": {
59
+ "pq_idx": 7,
60
+ "rg_idx": 8,
61
+ "epoch": 1
62
+ },
63
+ "loop_state": {
64
+ "min_val_bpb": 0.726215773156184,
65
+ "smooth_train_loss": 2.3654931810643007,
66
+ "total_training_time": 1499.3194689750671
67
+ }
68
+ }
latest/meta_000700.json ADDED
@@ -0,0 +1,68 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "step": 700,
3
+ "val_bpb": 0.7228979720861637,
4
+ "model_config": {
5
+ "sequence_len": 2048,
6
+ "vocab_size": 65536,
7
+ "n_layer": 32,
8
+ "n_head": 16,
9
+ "n_kv_head": 16,
10
+ "n_embd": 2048,
11
+ "window_pattern": "L",
12
+ "attn_variant": "hmap",
13
+ "hmap_alpha": 0.0,
14
+ "witten": false,
15
+ "bidirectional": false
16
+ },
17
+ "user_config": {
18
+ "run": "dummy",
19
+ "device_type": "",
20
+ "fp8": false,
21
+ "fp8_recipe": "tensorwise",
22
+ "depth": 32,
23
+ "aspect_ratio": 64,
24
+ "head_dim": 128,
25
+ "max_seq_len": 2048,
26
+ "window_pattern": "L",
27
+ "attn_variant": "hmap",
28
+ "hmap_alpha": 0.0,
29
+ "witten": false,
30
+ "objective": "ar",
31
+ "mlm_mask_prob": 0.15,
32
+ "num_iterations": 1600,
33
+ "target_flops": -1.0,
34
+ "target_param_data_ratio": 12,
35
+ "device_batch_size": 4,
36
+ "total_batch_size": 524288,
37
+ "embedding_lr": 0.03,
38
+ "unembedding_lr": 0.0008,
39
+ "weight_decay": 0.0,
40
+ "matrix_lr": 0.002,
41
+ "scalar_lr": 0.05,
42
+ "warmup_steps": 40,
43
+ "warmdown_ratio": 0.0,
44
+ "final_lr_frac": 1.0,
45
+ "resume_from_step": 100,
46
+ "init_from_model": "",
47
+ "eval_every": 100,
48
+ "eval_tokens": 41943040,
49
+ "core_metric_every": -1,
50
+ "core_metric_max_per_task": 500,
51
+ "sample_every": 250,
52
+ "save_every": 100,
53
+ "model_tag": "d32ft-amap"
54
+ },
55
+ "device_batch_size": 4,
56
+ "max_seq_len": 2048,
57
+ "total_batch_size": 524288,
58
+ "dataloader_state_dict": {
59
+ "pq_idx": 8,
60
+ "rg_idx": 24,
61
+ "epoch": 1
62
+ },
63
+ "loop_state": {
64
+ "min_val_bpb": 0.7228979720861637,
65
+ "smooth_train_loss": 2.433283090455347,
66
+ "total_training_time": 1734.1685729026794
67
+ }
68
+ }
latest/meta_000800.json ADDED
@@ -0,0 +1,68 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "step": 800,
3
+ "val_bpb": 0.7203195348042326,
4
+ "model_config": {
5
+ "sequence_len": 2048,
6
+ "vocab_size": 65536,
7
+ "n_layer": 32,
8
+ "n_head": 16,
9
+ "n_kv_head": 16,
10
+ "n_embd": 2048,
11
+ "window_pattern": "L",
12
+ "attn_variant": "hmap",
13
+ "hmap_alpha": 0.0,
14
+ "witten": false,
15
+ "bidirectional": false
16
+ },
17
+ "user_config": {
18
+ "run": "dummy",
19
+ "device_type": "",
20
+ "fp8": false,
21
+ "fp8_recipe": "tensorwise",
22
+ "depth": 32,
23
+ "aspect_ratio": 64,
24
+ "head_dim": 128,
25
+ "max_seq_len": 2048,
26
+ "window_pattern": "L",
27
+ "attn_variant": "hmap",
28
+ "hmap_alpha": 0.0,
29
+ "witten": false,
30
+ "objective": "ar",
31
+ "mlm_mask_prob": 0.15,
32
+ "num_iterations": 1600,
33
+ "target_flops": -1.0,
34
+ "target_param_data_ratio": 12,
35
+ "device_batch_size": 4,
36
+ "total_batch_size": 524288,
37
+ "embedding_lr": 0.03,
38
+ "unembedding_lr": 0.0008,
39
+ "weight_decay": 0.0,
40
+ "matrix_lr": 0.002,
41
+ "scalar_lr": 0.05,
42
+ "warmup_steps": 40,
43
+ "warmdown_ratio": 0.0,
44
+ "final_lr_frac": 1.0,
45
+ "resume_from_step": 100,
46
+ "init_from_model": "",
47
+ "eval_every": 100,
48
+ "eval_tokens": 41943040,
49
+ "core_metric_every": -1,
50
+ "core_metric_max_per_task": 500,
51
+ "sample_every": 250,
52
+ "save_every": 100,
53
+ "model_tag": "d32ft-amap"
54
+ },
55
+ "device_batch_size": 4,
56
+ "max_seq_len": 2048,
57
+ "total_batch_size": 524288,
58
+ "dataloader_state_dict": {
59
+ "pq_idx": 9,
60
+ "rg_idx": 32,
61
+ "epoch": 1
62
+ },
63
+ "loop_state": {
64
+ "min_val_bpb": 0.7203195348042326,
65
+ "smooth_train_loss": 2.3653833405434943,
66
+ "total_training_time": 1969.0279605388641
67
+ }
68
+ }
latest/meta_000900.json ADDED
@@ -0,0 +1,68 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "step": 900,
3
+ "val_bpb": 0.7181737972569234,
4
+ "model_config": {
5
+ "sequence_len": 2048,
6
+ "vocab_size": 65536,
7
+ "n_layer": 32,
8
+ "n_head": 16,
9
+ "n_kv_head": 16,
10
+ "n_embd": 2048,
11
+ "window_pattern": "L",
12
+ "attn_variant": "hmap",
13
+ "hmap_alpha": 0.0,
14
+ "witten": false,
15
+ "bidirectional": false
16
+ },
17
+ "user_config": {
18
+ "run": "dummy",
19
+ "device_type": "",
20
+ "fp8": false,
21
+ "fp8_recipe": "tensorwise",
22
+ "depth": 32,
23
+ "aspect_ratio": 64,
24
+ "head_dim": 128,
25
+ "max_seq_len": 2048,
26
+ "window_pattern": "L",
27
+ "attn_variant": "hmap",
28
+ "hmap_alpha": 0.0,
29
+ "witten": false,
30
+ "objective": "ar",
31
+ "mlm_mask_prob": 0.15,
32
+ "num_iterations": 1600,
33
+ "target_flops": -1.0,
34
+ "target_param_data_ratio": 12,
35
+ "device_batch_size": 4,
36
+ "total_batch_size": 524288,
37
+ "embedding_lr": 0.03,
38
+ "unembedding_lr": 0.0008,
39
+ "weight_decay": 0.0,
40
+ "matrix_lr": 0.002,
41
+ "scalar_lr": 0.05,
42
+ "warmup_steps": 40,
43
+ "warmdown_ratio": 0.0,
44
+ "final_lr_frac": 1.0,
45
+ "resume_from_step": 100,
46
+ "init_from_model": "",
47
+ "eval_every": 100,
48
+ "eval_tokens": 41943040,
49
+ "core_metric_every": -1,
50
+ "core_metric_max_per_task": 500,
51
+ "sample_every": 250,
52
+ "save_every": 100,
53
+ "model_tag": "d32ft-amap"
54
+ },
55
+ "device_batch_size": 4,
56
+ "max_seq_len": 2048,
57
+ "total_batch_size": 524288,
58
+ "dataloader_state_dict": {
59
+ "pq_idx": 10,
60
+ "rg_idx": 40,
61
+ "epoch": 1
62
+ },
63
+ "loop_state": {
64
+ "min_val_bpb": 0.7181737972569234,
65
+ "smooth_train_loss": 2.3197431898366494,
66
+ "total_training_time": 2203.8475222587585
67
+ }
68
+ }
latest/model_000600.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dcfb55c016127413e4398c00a5426fef8d2bd7553b381573c3c6154c3f0a9cff
3
+ size 11542830898
latest/model_000700.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c504def02ca9818885bca4af5284d0cfe23ace7f1296ac30a49e0a6f420b4925
3
+ size 11542830898
latest/model_000800.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59a667f3fb9675623b0ba98b79269fd7fc54059ba963ecafaafd446a2c3c1919
3
+ size 11542830898
latest/model_000900.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79830cf2316e227a11cd1c828a35d2796f43df1614e1f564f9383736ed8d983c
3
+ size 11542830898
latest/optim_000600_rank0.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54c606dbd80932b19cca2ac64673fd01c8cd046c9c8acc6a35b103323f8cdf3b
3
+ size 2080789141
latest/optim_000600_rank1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2ad184d2192ae9b12a928bea438b2d5189c6af0fb604fb851e09b41c2cbc267
3
+ size 2080789141
latest/optim_000600_rank2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca670afbd6d85be120d9682e45bdafdfd4bb3b6ac6a1266f37f30feacc22ca37
3
+ size 2080789141
latest/optim_000600_rank3.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a02c7d857d2cbd62e603c19e70d458138346d43329ed05ee713b959342c74c5
3
+ size 2080789141
latest/optim_000600_rank4.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e26726d26080b5614d2dafc3386a139b48010b17c276dbaf1677fefa4bfc5e4d
3
+ size 2080789141
latest/optim_000600_rank5.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a4de5e1f9ff4a5a9c5d226147910cad3aaf5c612381fe0da79e09b86cbf5716d
3
+ size 2080789141
latest/optim_000600_rank6.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4328ead40e8d6560f92ba5b098b32ac0952142c7db544ffe0289350594395a1c
3
+ size 2080789141
latest/optim_000600_rank7.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:56534bde8b050078b3e58d4b0eedac5347aef738b75a2326fdd1e52833e61d3a
3
+ size 2080789141
latest/optim_000700_rank0.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b6df1ec22792f7cda04f06c8c73b60056d15351097ea3bc37e234e9f09716b5e
3
+ size 2080789141
latest/optim_000700_rank1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:69bf4fcf7a371ad02ce1e2c5dbf5bc39a50a03dc28777e78ec3e4203596afdd7
3
+ size 2080789141
latest/optim_000700_rank2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a1f58968f188a578af5db33cfee338fad862b098e2363194a8988409dc85e75f
3
+ size 2080789141
latest/optim_000700_rank3.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98514966b5d879c4de6c8bf5f46b5cc1a8c84a3422f8e272dbce12701296161f
3
+ size 2080789141
latest/optim_000700_rank4.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:acca6d7c2402fe8b187c9afcfc927ff9d41c338369062585ef0f0ba480de6479
3
+ size 2080789141
latest/optim_000700_rank5.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41e0530804651793b98b367dd2fe165f25fffb41e294e15055c577c5c56c6e2a
3
+ size 2080789141
latest/optim_000700_rank6.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d9a29d8943c72bd5786fbd3a68de7d611d31550080c2a94598986fef43add367
3
+ size 2080789141
latest/optim_000700_rank7.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f34175f821b8cd80c653c0c1bd7a3021311fb3541d85a4b2386f4c0ab2147d37
3
+ size 2080789141
latest/optim_000800_rank0.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24671289b047898b6ae5c24d2d2b24bacf0d7a4576d5d5c653b03c76e059daa1
3
+ size 2080789141
latest/optim_000800_rank1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b50d470a604929c687e8516c93bda0f8e13bfafc457716bef000596be1b98665
3
+ size 2080789141
latest/optim_000800_rank2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f7e33b1f34269bc59a24c16c6e1bd34b350422544b07d59fe7e9b65947782e5
3
+ size 2080789141
latest/optim_000800_rank3.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:066d09b485acefabdd26aa82b7e90e81f0cb776f86a0e4fcb50d61a0cc0ab5dd
3
+ size 2080789141
latest/optim_000800_rank4.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d0f062811a8477623deb04d193ac649424538f2d13393b1b6639088b788403c
3
+ size 2080789141
latest/optim_000800_rank5.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d1dc93f6f82f6bbf23752c15ad29bcff05342965d53e2911eb83d4960df5b69
3
+ size 2080789141
latest/optim_000800_rank6.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c5fe0618a63cce2331c0e476c758430369834ac0a261ca48cce5d316b1a0b93
3
+ size 2080789141
latest/optim_000800_rank7.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bdb752c660a487523f34ea56344a7ea979dd3c85608f61578b1c7efac9fe050b
3
+ size 2080789141
latest/optim_000900_rank0.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a3b351ac769f4cdad10a8057baa4b748fa1b38f44b9564fce33d3dab7f0075dc
3
+ size 2080789141
latest/optim_000900_rank1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db7d83d7d38cbcbeba76638c5041c5034eb7abd80bd2041a3e7814a9b18660d8
3
+ size 2080789141
latest/optim_000900_rank2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c12d735cd90cd2f1be29dcca5e7997790f637fad6e29a29e48e00ade75d5568
3
+ size 2080789141
latest/optim_000900_rank3.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f07837221c270c9836c095a173e304d073385db0cc07e7e77dd9c6576ba288c
3
+ size 2080789141
latest/optim_000900_rank4.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c952a28f9b528075bfc23749035836fb35f61b3002207be4130480116ba53bbe
3
+ size 2080789141
latest/optim_000900_rank5.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2efb37ca91306e2e05eee9e9d5e5a2ea1cc7dd752e9c3b4c899fed8fd172b35
3
+ size 2080789141
latest/optim_000900_rank6.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94e84e90a6fdbebc7ef06886ffc153a8493e812f71912db20214819401649438
3
+ size 2080789141
latest/optim_000900_rank7.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13ec169b31deb126dc82c382fa839ed92b8237e36dbe158ea820593b048a6360
3
+ size 2080789141