Hugging Face's logo

Trelis
/
nanochat-recursive

English
Model card Files
xet
 Community
RonanMcGovern commited on
Commit
0bfc9d7
·
verified ·
1 Parent(s): b651223

Upload via push_to_hf.py

Browse files
Files changed (50) hide show
  1. base/d20/meta_005000.json +62 -0
  2. base/d20/meta_010000.json +62 -0
  3. base/d20/meta_015000.json +62 -0
  4. base/d20/meta_020000.json +62 -0
  5. base/d20/meta_021292.json +62 -0
  6. base/d20/model_005000.pt +3 -0
  7. base/d20/model_010000.pt +3 -0
  8. base/d20/model_015000.pt +3 -0
  9. base/d20/model_020000.pt +3 -0
  10. base/d20/model_021292.pt +3 -0
  11. base/d20/optim_005000_rank0.pt +3 -0
  12. base/d20/optim_005000_rank1.pt +3 -0
  13. base/d20/optim_005000_rank2.pt +3 -0
  14. base/d20/optim_005000_rank3.pt +3 -0
  15. base/d20/optim_005000_rank4.pt +3 -0
  16. base/d20/optim_005000_rank5.pt +3 -0
  17. base/d20/optim_005000_rank6.pt +3 -0
  18. base/d20/optim_005000_rank7.pt +3 -0
  19. base/d20/optim_010000_rank0.pt +3 -0
  20. base/d20/optim_010000_rank1.pt +3 -0
  21. base/d20/optim_010000_rank2.pt +3 -0
  22. base/d20/optim_010000_rank3.pt +3 -0
  23. base/d20/optim_010000_rank4.pt +3 -0
  24. base/d20/optim_010000_rank5.pt +3 -0
  25. base/d20/optim_010000_rank6.pt +3 -0
  26. base/d20/optim_010000_rank7.pt +3 -0
  27. base/d20/optim_015000_rank0.pt +3 -0
  28. base/d20/optim_015000_rank1.pt +3 -0
  29. base/d20/optim_015000_rank2.pt +3 -0
  30. base/d20/optim_015000_rank3.pt +3 -0
  31. base/d20/optim_015000_rank4.pt +3 -0
  32. base/d20/optim_015000_rank5.pt +3 -0
  33. base/d20/optim_015000_rank6.pt +3 -0
  34. base/d20/optim_015000_rank7.pt +3 -0
  35. base/d20/optim_020000_rank0.pt +3 -0
  36. base/d20/optim_020000_rank1.pt +3 -0
  37. base/d20/optim_020000_rank2.pt +3 -0
  38. base/d20/optim_020000_rank3.pt +3 -0
  39. base/d20/optim_020000_rank4.pt +3 -0
  40. base/d20/optim_020000_rank5.pt +3 -0
  41. base/d20/optim_020000_rank6.pt +3 -0
  42. base/d20/optim_020000_rank7.pt +3 -0
  43. base/d20/optim_021292_rank0.pt +3 -0
  44. base/d20/optim_021292_rank1.pt +3 -0
  45. base/d20/optim_021292_rank2.pt +3 -0
  46. base/d20/optim_021292_rank3.pt +3 -0
  47. base/d20/optim_021292_rank4.pt +3 -0
  48. base/d20/optim_021292_rank5.pt +3 -0
  49. base/d20/optim_021292_rank6.pt +3 -0
  50. base/d20/optim_021292_rank7.pt +3 -0
base/d20/meta_005000.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "step": 5000,
3
+ "val_bpb": 0.9190692005880656,
4
+ "model_config": {
5
+ "sequence_len": 2048,
6
+ "vocab_size": 65536,
7
+ "n_layer": 20,
8
+ "n_head": 10,
9
+ "n_kv_head": 10,
10
+ "n_embd": 1280,
11
+ "n_prelude": 2,
12
+ "n_recur_block": 4,
13
+ "n_coda": 2,
14
+ "train_recur_mean": 4.0,
15
+ "train_recur_max": 16,
16
+ "bptt_k": 4
17
+ },
18
+ "user_config": {
19
+ "run": "recursive-d20",
20
+ "device_type": "",
21
+ "depth": 20,
22
+ "max_seq_len": 2048,
23
+ "n_prelude": 2,
24
+ "n_recur_block": 4,
25
+ "n_coda": 2,
26
+ "train_recur_mean": 4.0,
27
+ "train_recur_max": 16,
28
+ "bptt_k": 4,
29
+ "num_iterations": -1,
30
+ "target_flops": -1.0,
31
+ "target_param_data_ratio": 34,
32
+ "device_batch_size": 32,
33
+ "total_batch_size": 524288,
34
+ "embedding_lr": 0.2,
35
+ "unembedding_lr": 0.004,
36
+ "weight_decay": 0.0,
37
+ "matrix_lr": 0.02,
38
+ "grad_clip": 1.0,
39
+ "warmup_ratio": 0.0,
40
+ "warmdown_ratio": 0.2,
41
+ "final_lr_frac": 0.0,
42
+ "resume_from_step": -1,
43
+ "eval_every": 250,
44
+ "eval_tokens": 10485760,
45
+ "core_metric_every": 2000,
46
+ "core_metric_max_per_task": 500,
47
+ "sample_every": 2000,
48
+ "save_every": 5000,
49
+ "model_tag": ""
50
+ },
51
+ "device_batch_size": 32,
52
+ "max_seq_len": 2048,
53
+ "dataloader_state_dict": {
54
+ "pq_idx": 46,
55
+ "rg_idx": 24
56
+ },
57
+ "loop_state": {
58
+ "min_val_bpb": 0.9190692005880656,
59
+ "smooth_train_loss": 3.102558805513702,
60
+ "total_training_time": 4112.87543463707
61
+ }
62
+ }
base/d20/meta_010000.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "step": 10000,
3
+ "val_bpb": 0.8901832528290349,
4
+ "model_config": {
5
+ "sequence_len": 2048,
6
+ "vocab_size": 65536,
7
+ "n_layer": 20,
8
+ "n_head": 10,
9
+ "n_kv_head": 10,
10
+ "n_embd": 1280,
11
+ "n_prelude": 2,
12
+ "n_recur_block": 4,
13
+ "n_coda": 2,
14
+ "train_recur_mean": 4.0,
15
+ "train_recur_max": 16,
16
+ "bptt_k": 4
17
+ },
18
+ "user_config": {
19
+ "run": "recursive-d20",
20
+ "device_type": "",
21
+ "depth": 20,
22
+ "max_seq_len": 2048,
23
+ "n_prelude": 2,
24
+ "n_recur_block": 4,
25
+ "n_coda": 2,
26
+ "train_recur_mean": 4.0,
27
+ "train_recur_max": 16,
28
+ "bptt_k": 4,
29
+ "num_iterations": -1,
30
+ "target_flops": -1.0,
31
+ "target_param_data_ratio": 34,
32
+ "device_batch_size": 32,
33
+ "total_batch_size": 524288,
34
+ "embedding_lr": 0.2,
35
+ "unembedding_lr": 0.004,
36
+ "weight_decay": 0.0,
37
+ "matrix_lr": 0.02,
38
+ "grad_clip": 1.0,
39
+ "warmup_ratio": 0.0,
40
+ "warmdown_ratio": 0.2,
41
+ "final_lr_frac": 0.0,
42
+ "resume_from_step": -1,
43
+ "eval_every": 250,
44
+ "eval_tokens": 10485760,
45
+ "core_metric_every": 2000,
46
+ "core_metric_max_per_task": 500,
47
+ "sample_every": 2000,
48
+ "save_every": 5000,
49
+ "model_tag": ""
50
+ },
51
+ "device_batch_size": 32,
52
+ "max_seq_len": 2048,
53
+ "dataloader_state_dict": {
54
+ "pq_idx": 92,
55
+ "rg_idx": 40
56
+ },
57
+ "loop_state": {
58
+ "min_val_bpb": 0.8901832528290349,
59
+ "smooth_train_loss": 3.036366580615659,
60
+ "total_training_time": 7203.946924448013
61
+ }
62
+ }
base/d20/meta_015000.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "step": 15000,
3
+ "val_bpb": 0.8763460640535932,
4
+ "model_config": {
5
+ "sequence_len": 2048,
6
+ "vocab_size": 65536,
7
+ "n_layer": 20,
8
+ "n_head": 10,
9
+ "n_kv_head": 10,
10
+ "n_embd": 1280,
11
+ "n_prelude": 2,
12
+ "n_recur_block": 4,
13
+ "n_coda": 2,
14
+ "train_recur_mean": 4.0,
15
+ "train_recur_max": 16,
16
+ "bptt_k": 4
17
+ },
18
+ "user_config": {
19
+ "run": "recursive-d20",
20
+ "device_type": "",
21
+ "depth": 20,
22
+ "max_seq_len": 2048,
23
+ "n_prelude": 2,
24
+ "n_recur_block": 4,
25
+ "n_coda": 2,
26
+ "train_recur_mean": 4.0,
27
+ "train_recur_max": 16,
28
+ "bptt_k": 4,
29
+ "num_iterations": -1,
30
+ "target_flops": -1.0,
31
+ "target_param_data_ratio": 34,
32
+ "device_batch_size": 32,
33
+ "total_batch_size": 524288,
34
+ "embedding_lr": 0.2,
35
+ "unembedding_lr": 0.004,
36
+ "weight_decay": 0.0,
37
+ "matrix_lr": 0.02,
38
+ "grad_clip": 1.0,
39
+ "warmup_ratio": 0.0,
40
+ "warmdown_ratio": 0.2,
41
+ "final_lr_frac": 0.0,
42
+ "resume_from_step": -1,
43
+ "eval_every": 250,
44
+ "eval_tokens": 10485760,
45
+ "core_metric_every": 2000,
46
+ "core_metric_max_per_task": 500,
47
+ "sample_every": 2000,
48
+ "save_every": 5000,
49
+ "model_tag": ""
50
+ },
51
+ "device_batch_size": 32,
52
+ "max_seq_len": 2048,
53
+ "dataloader_state_dict": {
54
+ "pq_idx": 139,
55
+ "rg_idx": 16
56
+ },
57
+ "loop_state": {
58
+ "min_val_bpb": 0.8763460640535932,
59
+ "smooth_train_loss": 2.9514984572688956,
60
+ "total_training_time": 10300.430169820786
61
+ }
62
+ }
base/d20/meta_020000.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "step": 20000,
3
+ "val_bpb": 0.8461040896385207,
4
+ "model_config": {
5
+ "sequence_len": 2048,
6
+ "vocab_size": 65536,
7
+ "n_layer": 20,
8
+ "n_head": 10,
9
+ "n_kv_head": 10,
10
+ "n_embd": 1280,
11
+ "n_prelude": 2,
12
+ "n_recur_block": 4,
13
+ "n_coda": 2,
14
+ "train_recur_mean": 4.0,
15
+ "train_recur_max": 16,
16
+ "bptt_k": 4
17
+ },
18
+ "user_config": {
19
+ "run": "recursive-d20",
20
+ "device_type": "",
21
+ "depth": 20,
22
+ "max_seq_len": 2048,
23
+ "n_prelude": 2,
24
+ "n_recur_block": 4,
25
+ "n_coda": 2,
26
+ "train_recur_mean": 4.0,
27
+ "train_recur_max": 16,
28
+ "bptt_k": 4,
29
+ "num_iterations": -1,
30
+ "target_flops": -1.0,
31
+ "target_param_data_ratio": 34,
32
+ "device_batch_size": 32,
33
+ "total_batch_size": 524288,
34
+ "embedding_lr": 0.2,
35
+ "unembedding_lr": 0.004,
36
+ "weight_decay": 0.0,
37
+ "matrix_lr": 0.02,
38
+ "grad_clip": 1.0,
39
+ "warmup_ratio": 0.0,
40
+ "warmdown_ratio": 0.2,
41
+ "final_lr_frac": 0.0,
42
+ "resume_from_step": -1,
43
+ "eval_every": 250,
44
+ "eval_tokens": 10485760,
45
+ "core_metric_every": 2000,
46
+ "core_metric_max_per_task": 500,
47
+ "sample_every": 2000,
48
+ "save_every": 5000,
49
+ "model_tag": ""
50
+ },
51
+ "device_batch_size": 32,
52
+ "max_seq_len": 2048,
53
+ "dataloader_state_dict": {
54
+ "pq_idx": 185,
55
+ "rg_idx": 40
56
+ },
57
+ "loop_state": {
58
+ "min_val_bpb": 0.8461040896385207,
59
+ "smooth_train_loss": 2.85765343997204,
60
+ "total_training_time": 13392.44127035141
61
+ }
62
+ }
base/d20/meta_021292.json ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "step": 21292,
3
+ "val_bpb": 0.8379230356842381,
4
+ "model_config": {
5
+ "sequence_len": 2048,
6
+ "vocab_size": 65536,
7
+ "n_layer": 20,
8
+ "n_head": 10,
9
+ "n_kv_head": 10,
10
+ "n_embd": 1280,
11
+ "n_prelude": 2,
12
+ "n_recur_block": 4,
13
+ "n_coda": 2,
14
+ "train_recur_mean": 4.0,
15
+ "train_recur_max": 16,
16
+ "bptt_k": 4
17
+ },
18
+ "user_config": {
19
+ "run": "recursive-d20",
20
+ "device_type": "",
21
+ "depth": 20,
22
+ "max_seq_len": 2048,
23
+ "n_prelude": 2,
24
+ "n_recur_block": 4,
25
+ "n_coda": 2,
26
+ "train_recur_mean": 4.0,
27
+ "train_recur_max": 16,
28
+ "bptt_k": 4,
29
+ "num_iterations": -1,
30
+ "target_flops": -1.0,
31
+ "target_param_data_ratio": 34,
32
+ "device_batch_size": 32,
33
+ "total_batch_size": 524288,
34
+ "embedding_lr": 0.2,
35
+ "unembedding_lr": 0.004,
36
+ "weight_decay": 0.0,
37
+ "matrix_lr": 0.02,
38
+ "grad_clip": 1.0,
39
+ "warmup_ratio": 0.0,
40
+ "warmdown_ratio": 0.2,
41
+ "final_lr_frac": 0.0,
42
+ "resume_from_step": -1,
43
+ "eval_every": 250,
44
+ "eval_tokens": 10485760,
45
+ "core_metric_every": 2000,
46
+ "core_metric_max_per_task": 500,
47
+ "sample_every": 2000,
48
+ "save_every": 5000,
49
+ "model_tag": ""
50
+ },
51
+ "device_batch_size": 32,
52
+ "max_seq_len": 2048,
53
+ "dataloader_state_dict": {
54
+ "pq_idx": 197,
55
+ "rg_idx": 48
56
+ },
57
+ "loop_state": {
58
+ "min_val_bpb": 0.8379230356842381,
59
+ "smooth_train_loss": 2.8288418270440188,
60
+ "total_training_time": 14194.468224525452
61
+ }
62
+ }
base/d20/model_005000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f098a518e2bba851d9272c7671747a25f0d99999dfd0f8d4266fbd4363ad937c
3
+ size 1145590951
base/d20/model_010000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a41547939ddce95dee45e2804ebbe3bba7b788a424bfb9ae4eddf8126682057d
3
+ size 1145590951
base/d20/model_015000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:058bbe149e981444906c23b09e26ad2dd9a43c4b7b5752cd8649b015c79dbfc5
3
+ size 1145590951
base/d20/model_020000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f0f17e617bb00854f4050543d9f99dc4b9cd11b796e35629b7bace15794c15b3
3
+ size 1145590951
base/d20/model_021292.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bf77e6743d4892ee00e831a384daeb42b8d3e0e2c9727044fcec959824b3d80
3
+ size 1145590951
base/d20/optim_005000_rank0.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4bffae7b46081201e15546c25b1bb16f1699fdaca5dfb48426125d00a0ded0ee
3
+ size 289675885
base/d20/optim_005000_rank1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16cd11e4f8e40a79eb00e390d09ca3cf33ef968d3dd648b1e18345540cd92915
3
+ size 276568421
base/d20/optim_005000_rank2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:44ad9989bd1244676437a01882eaadcfed5b15c3231d6f9b0fe28a89367ce986
3
+ size 276568421
base/d20/optim_005000_rank3.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:20072563d46d376849db1992e9c6fed56884590d7a24e0dc2770f53d2cfc20d7
3
+ size 276568421
base/d20/optim_005000_rank4.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96691085c32cc7a2fe9c850599e814e991698fd2c84df7dd3e3678c109935d97
3
+ size 276568421
base/d20/optim_005000_rank5.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6a9bdbf87c0655cc83052777e8d40d443b9926f148cd31283e88759220fc6d5
3
+ size 276568421
base/d20/optim_005000_rank6.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d4426e5bbde33c355eeb2c7f6fce34d0f609c8402fa67a25c057ca8c9f8cd0b3
3
+ size 276568421
base/d20/optim_005000_rank7.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a61fd55c1cb7ba54a129f017f5711c7f1d796539f132304c9b33821b4e7f703d
3
+ size 276568421
base/d20/optim_010000_rank0.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba4a00b25641f2399ec7fcc2488f38d294192d244e19d479926c6301cdc1e7fb
3
+ size 289675885
base/d20/optim_010000_rank1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:307b8964dfef0fdb28f9a3c8f0ea003bdf4d2f125abe8707d7eef663aa492630
3
+ size 276568421
base/d20/optim_010000_rank2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:337080b8ea5504cc84f73bebc5c6268289ba1fb67a873b238b7ccb77ce2195b7
3
+ size 276568421
base/d20/optim_010000_rank3.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5df734bb26dd6e526a1c9146eed6dc9d29814396703d64d215b69dfbd255ce20
3
+ size 276568421
base/d20/optim_010000_rank4.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6bec5236ca0c80a078e327890fafad7ada3f5a6d611b0c5d90e1560b962e776
3
+ size 276568421
base/d20/optim_010000_rank5.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b56ad3f5107c22d7302adae3fe6bc1bbdd6c8f71a791bd5343b20b852aec88be
3
+ size 276568421
base/d20/optim_010000_rank6.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:112a1e482e0a15c870373d4693f0c1ff6ec167490a577cdc59d9f8295eba38cc
3
+ size 276568421
base/d20/optim_010000_rank7.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f124393483830e6080d19a0871ff717b5828be8431629aaafa56d177da696cc
3
+ size 276568421
base/d20/optim_015000_rank0.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3ff3c93f324a2888da6d2a01811e5786bd5fe9ede80e19af463c4e68ee244fd0
3
+ size 289675885
base/d20/optim_015000_rank1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ace2695da0d91e51b759462e4d8a2ce49b7a0b166dc07637317131f8a842f595
3
+ size 276568421
base/d20/optim_015000_rank2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6376a758e0aea1e66db977bd12bf02b8b917d3d74d20e22316c778c2e5604b07
3
+ size 276568421
base/d20/optim_015000_rank3.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:daecd3e09ebd8c8c9f8a6156fbc446bbc618cb9376e32d660e26a2cbc3c3d847
3
+ size 276568421
base/d20/optim_015000_rank4.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b73ec256e76dadcf430fa8f30eacd705e30c706b2c0d90217b0ce07173bc0e77
3
+ size 276568421
base/d20/optim_015000_rank5.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:753c30c59b1763daed145a62669867f62ce0198a5e30d380bb9be6694c713932
3
+ size 276568421
base/d20/optim_015000_rank6.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8f7f334b8a7b8bfa51ec495a5c1bbdfd734509d00092c07a76acc803f9e018e
3
+ size 276568421
base/d20/optim_015000_rank7.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53c22701e9a0d659c25f506664bfe4f314c6fd0a64f5113a1c26ec38ad24c5a8
3
+ size 276568421
base/d20/optim_020000_rank0.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f65ef8b8c7240ea0453a6607028730b4b9796f18667bb99a4e75ba4d0a4b15d
3
+ size 289675885
base/d20/optim_020000_rank1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e56a973116f6d2aba5864137853c4c739e7cb374eff908d70ac290deba893e6
3
+ size 276568421
base/d20/optim_020000_rank2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:477127897f8848092180bd6499e15c5317764d90da738c5cfa3a77e58b2fe268
3
+ size 276568421
base/d20/optim_020000_rank3.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5da1fd4729ee3959da84560d88ddbcba2030a60cf20f65a342a966017cbf4e54
3
+ size 276568421
base/d20/optim_020000_rank4.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:68dba864bbc16f42cf9204142e6528cb2266c6682ba17621906b59828395c270
3
+ size 276568421
base/d20/optim_020000_rank5.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:425e67f583c6609d7b29fe1c255789c044ee61ba679a4ef5dd63e846823cf741
3
+ size 276568421
base/d20/optim_020000_rank6.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33b1b6e33fb035dc359ad1134e3f79005450884fab016ffb3ad8b25cb2fad317
3
+ size 276568421
base/d20/optim_020000_rank7.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:384ffc9d428e32b9ef51ae47cfaeb985a945dd99f5c4f003ea1727931f1c2997
3
+ size 276568421
base/d20/optim_021292_rank0.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21aa6de00bf839abd9a86b49cef872305b6e930f4c1f1757a6a36c2f999421f5
3
+ size 289675885
base/d20/optim_021292_rank1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:95db85874c7a9f4cee71a913ddd12c525a66138b47f5fe8794cafc9e535b105b
3
+ size 276568421
base/d20/optim_021292_rank2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f24672d324cc2db0b931fe64eceb7e4070df61c5affc40d4e98e930a136bbf3
3
+ size 276568421
base/d20/optim_021292_rank3.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b634a324d5eed6f36d6818391a62aae959d315fef6d4d6632bae6ebc58f6d34
3
+ size 276568421
base/d20/optim_021292_rank4.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:114a5e91627a915de8b98d89a0ff89de40ca89a7978de3e7b6f5f80a64c34523
3
+ size 276568421
base/d20/optim_021292_rank5.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f1272bbcfb493632c59dbe142ce8182ccdcb4cc8956c7e49716e12757b529c8
3
+ size 276568421
base/d20/optim_021292_rank6.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d24245b5cfbb635258c2ab67bd0b7f17609edef33c3110665633219e055a7f5c
3
+ size 276568421
base/d20/optim_021292_rank7.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e667a93da49091bfd61df4cd53cfbbbaed63451b10005d789f2123eebdc4f285
3
+ size 276568421