Hugging Face's logo

andrew-healey
/
context-compression

Model card Files
xet
 Community
andrew-healey commited on
Commit
006a2cc
·
verified ·
1 Parent(s): deb202b

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. attention_kindselective_n_heads4_seed1340/args.json +1 -1
  2. attention_kindselective_n_heads4_seed1340/dataloader_15000.pt +3 -0
  3. attention_kindselective_n_heads4_seed1340/dataloader_17500.pt +3 -0
  4. attention_kindselective_n_heads4_seed1340/dataloader_20000.pt +3 -0
  5. attention_kindselective_n_heads4_seed1340/dataloader_22500.pt +3 -0
  6. attention_kindselective_n_heads4_seed1340/dataloader_25000.pt +3 -0
  7. attention_kindselective_n_heads4_seed1340/dataloader_27500.pt +3 -0
  8. attention_kindselective_n_heads4_seed1340/dataloader_30000.pt +3 -0
  9. attention_kindselective_n_heads4_seed1340/dataloader_32500.pt +3 -0
  10. attention_kindselective_n_heads4_seed1340/dataloader_35000.pt +3 -0
  11. attention_kindselective_n_heads4_seed1340/dataloader_37500.pt +3 -0
  12. attention_kindselective_n_heads4_seed1340/dataloader_40000.pt +3 -0
  13. attention_kindselective_n_heads4_seed1340/log2.txt +0 -0
  14. attention_kindselective_n_heads4_seed1340/model_02500.pt +1 -1
  15. attention_kindselective_n_heads4_seed1340/model_05000.pt +1 -1
  16. attention_kindselective_n_heads4_seed1340/model_07500.pt +1 -1
  17. attention_kindselective_n_heads4_seed1340/model_10000.pt +1 -1
  18. attention_kindselective_n_heads4_seed1340/model_12500.pt +1 -1
  19. attention_kindselective_n_heads4_seed1340/model_15000.pt +3 -0
  20. attention_kindselective_n_heads4_seed1340/model_17500.pt +3 -0
  21. attention_kindselective_n_heads4_seed1340/model_20000.pt +3 -0
  22. attention_kindselective_n_heads4_seed1340/model_22500.pt +3 -0
  23. attention_kindselective_n_heads4_seed1340/model_25000.pt +3 -0
  24. attention_kindselective_n_heads4_seed1340/model_27500.pt +3 -0
  25. attention_kindselective_n_heads4_seed1340/model_30000.pt +3 -0
  26. attention_kindselective_n_heads4_seed1340/model_32500.pt +3 -0
  27. attention_kindselective_n_heads4_seed1340/model_35000.pt +3 -0
  28. attention_kindselective_n_heads4_seed1340/model_37500.pt +3 -0
  29. attention_kindselective_n_heads4_seed1340/model_40000.pt +3 -0
  30. attention_kindselective_n_heads4_seed1340/model_42500.pt +1 -1
  31. attention_kindselective_n_heads4_seed1340/model_45000.pt +1 -1
  32. attention_kindselective_n_heads4_seed1340/model_47500.pt +1 -1
  33. attention_kindselective_n_heads4_seed1340/model_49999.pt +1 -1
  34. attention_kindselective_n_heads4_seed1340/optimizer_02500.pt +1 -1
  35. attention_kindselective_n_heads4_seed1340/optimizer_05000.pt +1 -1
  36. attention_kindselective_n_heads4_seed1340/optimizer_07500.pt +1 -1
  37. attention_kindselective_n_heads4_seed1340/optimizer_10000.pt +1 -1
  38. attention_kindselective_n_heads4_seed1340/optimizer_12500.pt +1 -1
  39. attention_kindselective_n_heads4_seed1340/optimizer_15000.pt +3 -0
  40. attention_kindselective_n_heads4_seed1340/optimizer_17500.pt +3 -0
  41. attention_kindselective_n_heads4_seed1340/optimizer_20000.pt +3 -0
  42. attention_kindselective_n_heads4_seed1340/optimizer_22500.pt +3 -0
  43. attention_kindselective_n_heads4_seed1340/optimizer_25000.pt +3 -0
  44. attention_kindselective_n_heads4_seed1340/optimizer_27500.pt +3 -0
  45. attention_kindselective_n_heads4_seed1340/optimizer_30000.pt +3 -0
  46. attention_kindselective_n_heads4_seed1340/optimizer_32500.pt +3 -0
  47. attention_kindselective_n_heads4_seed1340/optimizer_35000.pt +3 -0
  48. attention_kindselective_n_heads4_seed1340/optimizer_37500.pt +3 -0
  49. attention_kindselective_n_heads4_seed1340/optimizer_40000.pt +3 -0
  50. attention_kindselective_n_heads4_seed1340/optimizer_42500.pt +1 -1
attention_kindselective_n_heads4_seed1340/args.json CHANGED
@@ -1 +1 @@
1
- {"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_7/attention_kindselective_n_heads4_seed1340", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 4, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 50000, "warmup_steps": 200, "group": "wider_is_better_7", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1340, "memory_penalty_epsilon": 0.1, "selection_head_linear_combo": "none", "selection_head_linear_combo_scale": 1.0, "protection_kind": "none", "leaky_relu_alpha": null, "leaky_relu_bias": null, "use_compile": true, "use_mini_model": false, "upload_to_hf": true, "seq_len": 256, "batch_size": 40, "total_batch_size": 10240, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "n_latent_masks": null, "mask_layernorm": false, "residual_attention_masks": false, "compute_base_shapes": false, "base_shapes_savefile": null, "mup": true, "disable_selection": false, "mup_enable_coord_check_logging": false, "max_lr": 7e-05, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "7e-5_10240_4_1340", "n_embd": 256}
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_7/attention_kindselective_n_heads4_seed1340", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 4, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 50000, "warmup_steps": 200, "group": "wider_is_better_7", "use_wandb": true, "kill_self_after_run": false, "random_seed": 1340, "memory_penalty_epsilon": 0.1, "selection_head_linear_combo": "none", "selection_head_linear_combo_scale": 1.0, "protection_kind": "none", "leaky_relu_alpha": null, "leaky_relu_bias": null, "use_compile": true, "use_mini_model": false, "upload_to_hf": true, "seq_len": 256, "batch_size": 40, "total_batch_size": 10240, "protection_head_scaling_factor": 1.0, "protection_head_bias": 0.0, "n_sliced_masks": null, "n_latent_masks": null, "mask_layernorm": false, "residual_attention_masks": false, "compute_base_shapes": false, "base_shapes_savefile": null, "mup": true, "disable_selection": false, "mup_enable_coord_check_logging": false, "max_lr": 2e-05, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "2e-5_10240_4_1340", "n_embd": 256}
attention_kindselective_n_heads4_seed1340/dataloader_15000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:450020c7c306c73e5c07c463518ab937102d657515ea5a38da6f2e7291f20324
3
+ size 964
attention_kindselective_n_heads4_seed1340/dataloader_17500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0696a655d7c8a9d7d275c7489f74a5a948ee029ac3941b045d6abaf12544a5b1
3
+ size 964
attention_kindselective_n_heads4_seed1340/dataloader_20000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4eb226bdcee777fd1ca493533704dae226c077ef79c842fc9dc59a534d5381c1
3
+ size 964
attention_kindselective_n_heads4_seed1340/dataloader_22500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:88833bfd821adac4edd1dd0772083ae007c7b8d33041f66e53a679e1fa8993e0
3
+ size 964
attention_kindselective_n_heads4_seed1340/dataloader_25000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:554555a425bac43d626c36f1c81c2b0aba51eda3281dab27a9cb56b61f413354
3
+ size 964
attention_kindselective_n_heads4_seed1340/dataloader_27500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a52940b3b45414e6cdbec0dbaeba848f52d681c2daf78c269027057332d7fbd
3
+ size 964
attention_kindselective_n_heads4_seed1340/dataloader_30000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:775519ea26122cb70d533c496bcdbbc19f759f3d096e83e98cca1dc10275fe8e
3
+ size 964
attention_kindselective_n_heads4_seed1340/dataloader_32500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab3422c8205fe45210246fed3dd6c317b75df02228cd8b75fba669574ce3b2d9
3
+ size 964
attention_kindselective_n_heads4_seed1340/dataloader_35000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:199ed3be67b88981f686112c97a2261729a37e0be3d4b0f4a289985a95d3cdf1
3
+ size 964
attention_kindselective_n_heads4_seed1340/dataloader_37500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f77178b3be9dd3f8cd03c935236251f73fde6da7948ba9feda0c888fb8912dfe
3
+ size 964
attention_kindselective_n_heads4_seed1340/dataloader_40000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:84f58ba3b4a1f9be7da4e697e06782f0e1ce4d3aca49f1997087fc83aa466dd9
3
+ size 964
attention_kindselective_n_heads4_seed1340/log2.txt CHANGED
The diff for this file is too large to render. See raw diff
 
attention_kindselective_n_heads4_seed1340/model_02500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e856700b2ba8c1c78fec371f511980a9165c3a304e65b1e2ecf19f9a9ba2c4f5
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d40b44be7123ddcbe8c8270c95b84e3da6036f8ad4a89f29bd78174771f5ad1d
3
  size 92843394
attention_kindselective_n_heads4_seed1340/model_05000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6ce0f91188a61cf2a20045bdc8d3dc3070278a30ac99975978bf3d232dd03f70
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64114b9e470f27de31f0af97063b74e5cb302fb1c4b95c1bab4f07ef2f27c50e
3
  size 92843394
attention_kindselective_n_heads4_seed1340/model_07500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:21edaf2540d7212d65d0bd210419f7d6a0f9281cde1d06f864c3ec1d63fff124
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e359763ad74be085ba664e365dd25cebe3d59184a9895012e9a3e79d28815537
3
  size 92843394
attention_kindselective_n_heads4_seed1340/model_10000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9e28d809d0626f302c7278753b8c13072c86cc80bea7caeb91264050622dd5e7
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:acfc53c9a7cd4ecd4786fe1dbfd12b3665d56e5218f464f0a7377d3c19cc0571
3
  size 92843394
attention_kindselective_n_heads4_seed1340/model_12500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:69a6cd3908266914a289ba86b8a7d2af8ce51bf4034f6b670c6730502d6f65d4
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:869376a3b0bb3fde527f196c220c30e7f871fd914115e19e80c0b37bf4c442fd
3
  size 92843394
attention_kindselective_n_heads4_seed1340/model_15000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:915e146b8111649e6a6db269b7778691b7cbe651ee1e60dec23d29627b76fd80
3
+ size 92843394
attention_kindselective_n_heads4_seed1340/model_17500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cce67e01256c5cce1cde6f7cc2d691f251068070c6622f2e4497418d105cfdf7
3
+ size 92843394
attention_kindselective_n_heads4_seed1340/model_20000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e58a9162de1095f0dc094f589088a4a286f14df34fde95304844bd261b137114
3
+ size 92843394
attention_kindselective_n_heads4_seed1340/model_22500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6cb8425cc4b92ac619610aa095caad0181fb1e64a22bdf07e0f8a478d4fbc637
3
+ size 92843394
attention_kindselective_n_heads4_seed1340/model_25000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a54319fac2ccb5018c7a6c9d11a35d9c18c453909cbc138ac59131dc99fb576
3
+ size 92843394
attention_kindselective_n_heads4_seed1340/model_27500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:178b2ef1de395feb11e65b597ad37d5e8925e9d5a1e95e955a6707aea00e122f
3
+ size 92843394
attention_kindselective_n_heads4_seed1340/model_30000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f23f9d5154f322670c4460e9fe3c45b108edee73a84595627941c922acd3bf4
3
+ size 92843394
attention_kindselective_n_heads4_seed1340/model_32500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c72a4c4302438f826476f9fd4f4472b9054a5b02e723684d14f8c5a41b51e5d
3
+ size 92843394
attention_kindselective_n_heads4_seed1340/model_35000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f29591ab1e09c43505e5fdc647dd4e9d21abfe6a0ab7f50ab47f9fe4207045ac
3
+ size 92843394
attention_kindselective_n_heads4_seed1340/model_37500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06a37d0898b5b62e808aaefe9b89f988cf90d4fb9c10d5fd38410275bfecbc21
3
+ size 92843394
attention_kindselective_n_heads4_seed1340/model_40000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03aa4bc431bade8c18c5c59922094723549f04f741f4366931664738dd354a17
3
+ size 92843394
attention_kindselective_n_heads4_seed1340/model_42500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7e937e5bde413b5f1d5cd2e7a03428c77cca6ac8ef7547cc1d8aa0565c33ea0d
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5aa2c435ba305fa32a8e5282696e70255f511cfaf0c5b519f809a0a2368b0f9c
3
  size 92843394
attention_kindselective_n_heads4_seed1340/model_45000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:03ef3fa0f39e905fced4c8e9f6d4c4c36d3dfdec8a8589b65a5f4c6363469015
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d0d999a95806ce0df25259e77ae207370f9d5fbc520455ebc8c6df91d2a1049
3
  size 92843394
attention_kindselective_n_heads4_seed1340/model_47500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:58579aca525c6640a50ed6de3f5bc997ccf568b1c74ac74080a8413181c72725
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8c612bd311cf5bf36b7cd0eaf2d390b96cbd61422cc6c461817195a65724ea8
3
  size 92843394
attention_kindselective_n_heads4_seed1340/model_49999.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ddd2671bc53d3d47c2a3597b04a352fe6e56e1aa82a19bed3ee6f4e205890842
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:329cc7c07d336b08babec9f25a79e6cb3f45e98b7201a91059aa2e8f0a57113b
3
  size 92843394
attention_kindselective_n_heads4_seed1340/optimizer_02500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:71d5d80bc8621c01c7517a9c8966efcf316b98161adb2e069c5201828111ea9b
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e96f244321b4ceb32e937deaaff4052c2e16226a5c21490d87da1ae07a2e463
3
  size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_05000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:4cf0986c29b0869c7e9016f8a369414e67e321e3e73344a91dc6a3451b88ff99
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d2eb29f56578c3cff4293ded7486dacf3c591b28bf309ee22630b302d36f8f24
3
  size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_07500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f45a3b4647bd40c9e04b365a1e5c1703f4207fa3698dc385fc2cfd5bc0ea4bd5
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f46662c30e0c586b62614381c40c4d40361fe89d16f64503cefe582b93844dd2
3
  size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_10000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:06bc9c8845c72cf18f14e45b43f7c4c19b17aa50a69cff163f981ea53f88744f
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07246a91a4608bf5ab9f139643924fd088b34009b22bdfeb77ec5a8dc3d6e94b
3
  size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_12500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:61803c8687b0d4b6de778e759324ef063faecd26bcfb6e193e45505fe759b8d9
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc0fdd1cddbb185439cc502fcc422c882464781999d991dc77c352104ede18c0
3
  size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_15000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d994c5254e5c42075db07a54cd8b3196eeed538d59eb3f014786b64ec3da70c
3
+ size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_17500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d2670345760eecd01305e061ae5a4ee5dfeaa24cf41a44adf5919d03ea63db3
3
+ size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_20000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:72b225f844a81f011bef5ad885e1f90ef3eb824869267a3703f1a6434b680fc2
3
+ size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_22500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6f83f09e13afe18b673705f672eb4c982a0acdcd440f50922885b335c734953
3
+ size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_25000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6eeb58ca059298e41a9d34cf2a5fcfc8f342f052ea97c57e3106bc36533f5bfe
3
+ size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_27500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18ce8d19d6111ae0bd7475511f2adab74b2164eb9aacff4ab75e722b6f73a311
3
+ size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_30000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:42c27ce04b2b34ff778efa00ae992b9cfaa0b12cbaf53df09c38f5745821a6a0
3
+ size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_32500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:38c3b9a8f3ab7c52f68a25fcbe1dd87ab4bb869b6d9d5e7d9cf7698080629f9e
3
+ size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_35000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f3bc2abb7ee0d8214d323a2d1e726e5b56e1bbf13ce1fc078bd5ad0738eb44c9
3
+ size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_37500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:115d76cf5edc7fdc26fe7491f3cf57823c30944da97f62912cfd04b9b8cdb858
3
+ size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_40000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba7f05d99562a7da60349b27638be67608c000adbd7c48fd923b87f1e0542277
3
+ size 179406214
attention_kindselective_n_heads4_seed1340/optimizer_42500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:dffbcca299af84d6d052050b7c9c7230f9052fbb489bbc7104607f111ccd656e
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a626934e4e9e095eecfd4bd018cf53ce8e61b807d9c1512a49275dea6dc52a73
3
  size 179406214