Hugging Face's logo

andrew-healey
/
context-compression

Model card Files
xet
 Community
andrew-healey commited on
Commit
619b5a3
·
verified ·
1 Parent(s): 9f82b89

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_seed1339/args.json +1 -1
  2. attention_kindselective_n_heads4_seed1339/dataloader_12500.pt +3 -0
  3. attention_kindselective_n_heads4_seed1339/dataloader_15000.pt +3 -0
  4. attention_kindselective_n_heads4_seed1339/dataloader_17500.pt +3 -0
  5. attention_kindselective_n_heads4_seed1339/dataloader_20000.pt +3 -0
  6. attention_kindselective_n_heads4_seed1339/dataloader_22500.pt +3 -0
  7. attention_kindselective_n_heads4_seed1339/dataloader_25000.pt +3 -0
  8. attention_kindselective_n_heads4_seed1339/dataloader_27500.pt +3 -0
  9. attention_kindselective_n_heads4_seed1339/dataloader_30000.pt +3 -0
  10. attention_kindselective_n_heads4_seed1339/dataloader_32500.pt +3 -0
  11. attention_kindselective_n_heads4_seed1339/dataloader_35000.pt +3 -0
  12. attention_kindselective_n_heads4_seed1339/dataloader_37500.pt +3 -0
  13. attention_kindselective_n_heads4_seed1339/log2.txt +0 -0
  14. attention_kindselective_n_heads4_seed1339/model_02500.pt +1 -1
  15. attention_kindselective_n_heads4_seed1339/model_05000.pt +1 -1
  16. attention_kindselective_n_heads4_seed1339/model_07500.pt +1 -1
  17. attention_kindselective_n_heads4_seed1339/model_10000.pt +1 -1
  18. attention_kindselective_n_heads4_seed1339/model_12500.pt +3 -0
  19. attention_kindselective_n_heads4_seed1339/model_15000.pt +3 -0
  20. attention_kindselective_n_heads4_seed1339/model_17500.pt +3 -0
  21. attention_kindselective_n_heads4_seed1339/model_20000.pt +3 -0
  22. attention_kindselective_n_heads4_seed1339/model_22500.pt +3 -0
  23. attention_kindselective_n_heads4_seed1339/model_25000.pt +3 -0
  24. attention_kindselective_n_heads4_seed1339/model_27500.pt +3 -0
  25. attention_kindselective_n_heads4_seed1339/model_30000.pt +3 -0
  26. attention_kindselective_n_heads4_seed1339/model_32500.pt +3 -0
  27. attention_kindselective_n_heads4_seed1339/model_35000.pt +3 -0
  28. attention_kindselective_n_heads4_seed1339/model_37500.pt +3 -0
  29. attention_kindselective_n_heads4_seed1339/model_40000.pt +1 -1
  30. attention_kindselective_n_heads4_seed1339/model_42500.pt +1 -1
  31. attention_kindselective_n_heads4_seed1339/model_45000.pt +1 -1
  32. attention_kindselective_n_heads4_seed1339/model_47500.pt +1 -1
  33. attention_kindselective_n_heads4_seed1339/model_49999.pt +1 -1
  34. attention_kindselective_n_heads4_seed1339/optimizer_02500.pt +1 -1
  35. attention_kindselective_n_heads4_seed1339/optimizer_05000.pt +1 -1
  36. attention_kindselective_n_heads4_seed1339/optimizer_07500.pt +1 -1
  37. attention_kindselective_n_heads4_seed1339/optimizer_10000.pt +1 -1
  38. attention_kindselective_n_heads4_seed1339/optimizer_12500.pt +3 -0
  39. attention_kindselective_n_heads4_seed1339/optimizer_15000.pt +3 -0
  40. attention_kindselective_n_heads4_seed1339/optimizer_17500.pt +3 -0
  41. attention_kindselective_n_heads4_seed1339/optimizer_20000.pt +3 -0
  42. attention_kindselective_n_heads4_seed1339/optimizer_22500.pt +3 -0
  43. attention_kindselective_n_heads4_seed1339/optimizer_25000.pt +3 -0
  44. attention_kindselective_n_heads4_seed1339/optimizer_27500.pt +3 -0
  45. attention_kindselective_n_heads4_seed1339/optimizer_30000.pt +3 -0
  46. attention_kindselective_n_heads4_seed1339/optimizer_32500.pt +3 -0
  47. attention_kindselective_n_heads4_seed1339/optimizer_35000.pt +3 -0
  48. attention_kindselective_n_heads4_seed1339/optimizer_37500.pt +3 -0
  49. attention_kindselective_n_heads4_seed1339/optimizer_40000.pt +1 -1
  50. attention_kindselective_n_heads4_seed1339/optimizer_42500.pt +1 -1
attention_kindselective_n_heads4_seed1339/args.json CHANGED
@@ -1 +1 @@
1
- {"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_7/attention_kindselective_n_heads4_seed1339", "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": 1339, "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": 0.0001, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "10e-5_10240_4_1339", "n_embd": 256}
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_7/attention_kindselective_n_heads4_seed1339", "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": 1339, "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_1339", "n_embd": 256}
attention_kindselective_n_heads4_seed1339/dataloader_12500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab3779d33c2e0a7873fcd8c39402e44260740665950323ad1445480ec339965a
3
+ size 964
attention_kindselective_n_heads4_seed1339/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_seed1339/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_seed1339/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_seed1339/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_seed1339/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_seed1339/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_seed1339/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_seed1339/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_seed1339/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_seed1339/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_seed1339/log2.txt CHANGED
The diff for this file is too large to render. See raw diff
 
attention_kindselective_n_heads4_seed1339/model_02500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2ec7853f39c020bf0817d8a9424f0786ad4b93fc0839e1f8734dc36c9b628b3c
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d5fec9045af2708a11b1103c045ec7a59a9b1e48b4bc322a7fb0d864f4fe143
3
  size 92843394
attention_kindselective_n_heads4_seed1339/model_05000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e7a3ea4733ac190d2c44f57c3fc804dc59f7571ad4f028f02fe4775d0b66df1b
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6425c5316fad4e1ed4ca5670f21cacaad297f0b0c03153ad6258b436a6c27271
3
  size 92843394
attention_kindselective_n_heads4_seed1339/model_07500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5d9cba9958a987856e63371998d573139f4d7e12bf83d1dc6741e08c1656ae1e
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe251dc7a96a35307a4218cddf07ddef72e020cd1dcd16c06fd8a07f476d2b3d
3
  size 92843394
attention_kindselective_n_heads4_seed1339/model_10000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:75c9a83ca6838f6643993a01ade7f926d7dccd9aaabbaa9d1cc052a66d45ad6a
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5fd919e371e6ea4001f6469599129e20ba9c35d2917a86faebec778aa5af7d62
3
  size 92843394
attention_kindselective_n_heads4_seed1339/model_12500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e90652aa199388014ee309da42595a880d3c574892a0112cc9949e2b8638ce26
3
+ size 92843394
attention_kindselective_n_heads4_seed1339/model_15000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51bc84a3106488453a5165f744f925a378d18cfbb260c8d6172c8542270e41b8
3
+ size 92843394
attention_kindselective_n_heads4_seed1339/model_17500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b65470ea1dbc614d43c8967a5647ec824d520c4dcbc8bf8bb997eba0e8a8588
3
+ size 92843394
attention_kindselective_n_heads4_seed1339/model_20000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cdd979d4e03f11a0456610a83b22b5e55182f12ee739d1b7f61c723f42c42f48
3
+ size 92843394
attention_kindselective_n_heads4_seed1339/model_22500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a871162895775d404e04532e5cd853fb0824250f90d37bb1d22bd0f3c24629e
3
+ size 92843394
attention_kindselective_n_heads4_seed1339/model_25000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a83e359f8b0060fada80673e3c5ff586fda7a9daab15b2effb38085e33a9b42b
3
+ size 92843394
attention_kindselective_n_heads4_seed1339/model_27500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ffc6f7f6a861696e3c363d8c66639412eb48e3b6182b3c073248470ea7b0ffa
3
+ size 92843394
attention_kindselective_n_heads4_seed1339/model_30000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d07b7c5cc6f74f1959a7fb18d77e008f804277c518ea7b34db9812b6e50c9907
3
+ size 92843394
attention_kindselective_n_heads4_seed1339/model_32500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:12d6178d05edaf2390b6f89ec425795f99ea95c3ed1e2b33f4a97309d1531997
3
+ size 92843394
attention_kindselective_n_heads4_seed1339/model_35000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e3b75d837f318b1c757629de47cc740ccfa4f287f25ac4e0c4333eb9c78a5ee
3
+ size 92843394
attention_kindselective_n_heads4_seed1339/model_37500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca6ccb8895713bd20d9694a58e9172efa70ca0fedb143bf06d7407e5e1e38a5e
3
+ size 92843394
attention_kindselective_n_heads4_seed1339/model_40000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:78a8fd9d6a750808245efe7c72bed034116984b1a3d236604a1e986d60e7ed58
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a533c25bdb31ab0d17973befa3c8c3dcdcb5d641b8659f6aa9f3b8facc3b2780
3
  size 92843394
attention_kindselective_n_heads4_seed1339/model_42500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c3363e83947e45c3c64503ecbdcc0b270aa74e74aa9ec5438a596bd3c8156913
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a19c01365e93545b970a342c34ae9c9fa346575249ff72f728858556a727cb65
3
  size 92843394
attention_kindselective_n_heads4_seed1339/model_45000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:652267ce4a3c3eb24b3fdd1182cbeaad71435a28d6533b8d65f60dfd783e5411
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5a557feb627ced00d55c191957ef9d595e11f01512e3b8d6ca00283170fe997
3
  size 92843394
attention_kindselective_n_heads4_seed1339/model_47500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d198491e4dd117689476940ce9f230be59d99c41399afc7f16477ab0abe13f73
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2518192db956074cf791c8857f72dae7cb8006abf2e21173b4251ce253896af9
3
  size 92843394
attention_kindselective_n_heads4_seed1339/model_49999.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:bb07662f40aa23921523b0c43b25a496ce763a42d93ca236f9ab6e69e0f9407d
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:40c1c361a207f5c416fef30c79e06b4edc3315c07f7d7cb1c6deccac65695913
3
  size 92843394
attention_kindselective_n_heads4_seed1339/optimizer_02500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6d0daaea20dfea170f7f2a3011fad8c68fd0d34d7837cfc7e1b82e10c8afac2e
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd90610bdfd7254edfb1fbed624dad17d4f7fb821d7e8099ff353419a048f679
3
  size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_05000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d6b399697378d1367510c9f4104e498e29bd1617ea27a7878564f423ee87160d
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:418298de7332b6ca0f7c05bced21479834ce4b6d543e9849da08709fe1de4027
3
  size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_07500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:827fa057c0c4bae82594e7d86c915d39ef2208d8da842bda634d1d07f8a6f14c
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6cff77e448d0e70d4643a97f8c9274780bb2ed5a0a6cce01879bf1ecf907295f
3
  size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_10000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:bdb56e2388c4d739ff044d95f0f99d11044dc7064e262ea9e0da4677e4b7369f
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa0a87a2a81beb289e8e65a4ab7fc951d274bd263127fca5d51aa21d13888550
3
  size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_12500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d7a0ba73b14ef416ec076c5d37e507cfede5fbb34783c0f7b600aea887abb29
3
+ size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_15000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32d28686675b5b55348834047127817f0e2c3dc674e7bd10f22d92b6f149a217
3
+ size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_17500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03b5777fddf8f8c914472aa5eafe22ceb2cfb59f26d567878cdd4884ab50c479
3
+ size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_20000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8292f83f4d2bc0af27f2aa4f429fcb06e230b506e186df8de062a95dfb5dfeb3
3
+ size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_22500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dfaac188309cc159a685e295d76251abe6a9406bfdd79ba55f666a22afd78108
3
+ size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_25000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:606eede0cf315db37d8147a825918ca71a1293450df3a0536665e7f7d7229a74
3
+ size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_27500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c87bdc27376baa7605f19f576e1e3b4cd128c4df6940579b3c3c3d083cea83f
3
+ size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_30000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3effc948bf127f1c27e7f4cb83d15688cd556614fc32ec8f5f3a2e192a0c8f97
3
+ size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_32500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87f585f51d45083438bd55a44b333c2af592eebc7d797aaf73d8aa401c1082f9
3
+ size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_35000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78f5449827bf9b6cf3ccb135222d294071d320102418282e681f5cb76d9b8932
3
+ size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_37500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fe8ff6437f7738f84d63aea8ab7b27705ca04186f5bf66c11d59cdcbb788622
3
+ size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_40000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7ecc8125e65f3bc259ac039c34b488e49892d0778ea59a14ffd86b25c567c0ac
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4cdb59a59ed901dc5c16544ba72a3310a0fade3e062facba26115f016a6d092f
3
  size 179406214
attention_kindselective_n_heads4_seed1339/optimizer_42500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a22bc9fe743fba8fc4b4560471a341de420ff16d5fb5457b184c26c331fefcdb
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa8fb26bf4fe7759cd2b76c5ed8febaccf16c0a57fb589c13a33a961029cde2e
3
  size 179406214