Hugging Face's logo

andrew-healey
/
context-compression

Model card Files
xet
 Community
andrew-healey commited on
Commit
3bb0718
·
verified ·
1 Parent(s): 1fee985

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_seed1341/args.json +1 -1
  2. attention_kindselective_n_heads4_seed1341/dataloader_10000.pt +3 -0
  3. attention_kindselective_n_heads4_seed1341/dataloader_12500.pt +3 -0
  4. attention_kindselective_n_heads4_seed1341/dataloader_15000.pt +3 -0
  5. attention_kindselective_n_heads4_seed1341/dataloader_17500.pt +3 -0
  6. attention_kindselective_n_heads4_seed1341/dataloader_20000.pt +3 -0
  7. attention_kindselective_n_heads4_seed1341/dataloader_22500.pt +3 -0
  8. attention_kindselective_n_heads4_seed1341/dataloader_25000.pt +3 -0
  9. attention_kindselective_n_heads4_seed1341/dataloader_27500.pt +3 -0
  10. attention_kindselective_n_heads4_seed1341/dataloader_30000.pt +3 -0
  11. attention_kindselective_n_heads4_seed1341/dataloader_32500.pt +3 -0
  12. attention_kindselective_n_heads4_seed1341/dataloader_35000.pt +3 -0
  13. attention_kindselective_n_heads4_seed1341/dataloader_37500.pt +3 -0
  14. attention_kindselective_n_heads4_seed1341/dataloader_40000.pt +3 -0
  15. attention_kindselective_n_heads4_seed1341/dataloader_42500.pt +3 -0
  16. attention_kindselective_n_heads4_seed1341/dataloader_45000.pt +3 -0
  17. attention_kindselective_n_heads4_seed1341/log2.txt +0 -0
  18. attention_kindselective_n_heads4_seed1341/model_02500.pt +1 -1
  19. attention_kindselective_n_heads4_seed1341/model_05000.pt +1 -1
  20. attention_kindselective_n_heads4_seed1341/model_07500.pt +1 -1
  21. attention_kindselective_n_heads4_seed1341/model_10000.pt +3 -0
  22. attention_kindselective_n_heads4_seed1341/model_12500.pt +3 -0
  23. attention_kindselective_n_heads4_seed1341/model_15000.pt +3 -0
  24. attention_kindselective_n_heads4_seed1341/model_17500.pt +3 -0
  25. attention_kindselective_n_heads4_seed1341/model_20000.pt +3 -0
  26. attention_kindselective_n_heads4_seed1341/model_22500.pt +3 -0
  27. attention_kindselective_n_heads4_seed1341/model_25000.pt +3 -0
  28. attention_kindselective_n_heads4_seed1341/model_27500.pt +3 -0
  29. attention_kindselective_n_heads4_seed1341/model_30000.pt +3 -0
  30. attention_kindselective_n_heads4_seed1341/model_32500.pt +3 -0
  31. attention_kindselective_n_heads4_seed1341/model_35000.pt +3 -0
  32. attention_kindselective_n_heads4_seed1341/model_37500.pt +3 -0
  33. attention_kindselective_n_heads4_seed1341/model_40000.pt +3 -0
  34. attention_kindselective_n_heads4_seed1341/model_42500.pt +3 -0
  35. attention_kindselective_n_heads4_seed1341/model_45000.pt +3 -0
  36. attention_kindselective_n_heads4_seed1341/model_47500.pt +1 -1
  37. attention_kindselective_n_heads4_seed1341/model_49999.pt +1 -1
  38. attention_kindselective_n_heads4_seed1341/optimizer_02500.pt +1 -1
  39. attention_kindselective_n_heads4_seed1341/optimizer_05000.pt +1 -1
  40. attention_kindselective_n_heads4_seed1341/optimizer_07500.pt +1 -1
  41. attention_kindselective_n_heads4_seed1341/optimizer_10000.pt +3 -0
  42. attention_kindselective_n_heads4_seed1341/optimizer_12500.pt +3 -0
  43. attention_kindselective_n_heads4_seed1341/optimizer_15000.pt +3 -0
  44. attention_kindselective_n_heads4_seed1341/optimizer_17500.pt +3 -0
  45. attention_kindselective_n_heads4_seed1341/optimizer_20000.pt +3 -0
  46. attention_kindselective_n_heads4_seed1341/optimizer_22500.pt +3 -0
  47. attention_kindselective_n_heads4_seed1341/optimizer_25000.pt +3 -0
  48. attention_kindselective_n_heads4_seed1341/optimizer_27500.pt +3 -0
  49. attention_kindselective_n_heads4_seed1341/optimizer_30000.pt +3 -0
  50. attention_kindselective_n_heads4_seed1341/optimizer_32500.pt +3 -0
attention_kindselective_n_heads4_seed1341/args.json CHANGED
@@ -1 +1 @@
1
- {"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_7/attention_kindselective_n_heads4_seed1341", "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": 1341, "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_1341", "n_embd": 256}
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_7/attention_kindselective_n_heads4_seed1341", "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": 1341, "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_1341", "n_embd": 256}
attention_kindselective_n_heads4_seed1341/dataloader_10000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f3858f6c832feea78a674d8c5c384061cc7d4f22cddbd0a2be6de33bc91e2c72
3
+ size 964
attention_kindselective_n_heads4_seed1341/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_seed1341/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_seed1341/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_seed1341/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_seed1341/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_seed1341/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_seed1341/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_seed1341/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_seed1341/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_seed1341/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_seed1341/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_seed1341/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_seed1341/dataloader_42500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf6d24c78d89100d146bce9f26be940db3d71092473d9b55db97d6b35531eac2
3
+ size 964
attention_kindselective_n_heads4_seed1341/dataloader_45000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22bb90b43d81f3da5454f91a70e1ed29aeb2f470a727ce38390ff8a5c4924889
3
+ size 964
attention_kindselective_n_heads4_seed1341/log2.txt CHANGED
The diff for this file is too large to render. See raw diff
 
attention_kindselective_n_heads4_seed1341/model_02500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c484aa037e9b115403161e2fe304d0c7a89dc6b79d216e3cdfd363f6ca13d25d
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9be0ae6a1500ccdd6fb25290b75c4aa5cad5df4ff202df8d30bf64fb32830a49
3
  size 92843394
attention_kindselective_n_heads4_seed1341/model_05000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e58ae40f92461f191f2a2c7ae689b02053d1241f805b9a567e39cb8e5277a851
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ed8c056540906461d7125570140a49c884941a998e834c885f919773ed5a7f5
3
  size 92843394
attention_kindselective_n_heads4_seed1341/model_07500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d3bfda09d2b22fdab50914f273bc586412124529cdfa64474e214c7816f9e627
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c38e2493098f850413b9bf630b581247017cd6934de174dd3a28f6665ba47c9
3
  size 92843394
attention_kindselective_n_heads4_seed1341/model_10000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59bddcb60af0e23b41bda959258fd176c4b8766360510deff267004e57a4e0de
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_12500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:875176bfdf881b5fc9ff2f8f606bc10d742f2e775b58bd7e1944d9959e2f90d4
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_15000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d6fbafd23666846f3d49b9d6780b4c38f54201cb1490921243e9fe4f416756c
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_17500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f65285a4ec29e037c8b40770cb285d59d90c1bb9419f3a115e9f50dfff9e08e3
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_20000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:86bb24316fbd6b5f53481c0404e322d0a292473e3c8b601e89258863c07923a8
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_22500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:330f600120c8929427e542292dabccaba368ed2d02fbea4f5313f57356164b97
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_25000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c85eae63babcc7b9606b1d62018b1d2237424a7b2edfc5fe5baf38bd7280c72c
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_27500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a875e21c8b8935d625b5bb183f5c1a2f0295ce88fce2d9132933cb0639c9f61f
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_30000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:397c8ae3f6fb7d1c08b98c981d7f7d0659e1859fad7eec236476dcc7a0176dc3
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_32500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd9829080b4f893600ed7acd641838505198b9bf3545a8db6d7dcf4a3646c6d0
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_35000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0326841882dbaa3f7541f4074e0a56cc4d87c35e1c36fa8a319f4a69566f7b6f
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_37500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53e08e519e33fb6431360809b1849e6e3eccba88f6aaf88fa80953369f374ed7
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_40000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:952d7e8d936be733c6adbeae4cda1c8ad6f6eb1f5cfa924c171a1c4c9ac105ca
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_42500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2ea54fef770a7d9d826f6492ef69a9edbd289cddb8a966556901e0a42a142f4
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_45000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f61615cd5739917fc974acc450587866f3e3c8a6255c17ea22349e1c274b1d6
3
+ size 92843394
attention_kindselective_n_heads4_seed1341/model_47500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6aeb2a7a2351f8f16532e3f96f1535852348a874b6e53498f54dbb2dbea87f2e
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9eae84098b96e876b98aac32a8c36f77fc8790952a7e382ce8909bb0c616ec76
3
  size 92843394
attention_kindselective_n_heads4_seed1341/model_49999.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0a27f8d94a91205f44b4097397cbe96f34c7125f7acaf12a27ed77f3f225c47c
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f43b466896f202b1c512cc7966ed31153489bec9e66730115b4391d2c528b648
3
  size 92843394
attention_kindselective_n_heads4_seed1341/optimizer_02500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c5a91ef2bc89f32f183c058e5eeb515108d458e3939109aed168281f8270a000
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4451193db27fcfd259ca86985ca50079097afcda3de8ef12899aad3deb7b3a8b
3
  size 179406214
attention_kindselective_n_heads4_seed1341/optimizer_05000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a4eaac3f7ff71b93997f325d4f277c4c0ec68fbf025793ee4565dbfdba779a6e
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16e9f5de4822a6da4a7c5ebc4b1be114be696f40cb71ad5094f4a5036edbbc1f
3
  size 179406214
attention_kindselective_n_heads4_seed1341/optimizer_07500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:553854895d5cc504a01b0c2b72a8825e26e1f903273fbf8c71f350fa796a32d4
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3aa6a192b836fd2fea2915b9f3875f2f27ae191db2f3b5fa7b11949e0699cac9
3
  size 179406214
attention_kindselective_n_heads4_seed1341/optimizer_10000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04fab138ca5dab326197fa18fbc6866b5d180a3a5cc824e9e21a5e775ff66c78
3
+ size 179406214
attention_kindselective_n_heads4_seed1341/optimizer_12500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a9a870a04c9b0968802417299b0b927cb8f166659f0b572f3d98536c7a518d9
3
+ size 179406214
attention_kindselective_n_heads4_seed1341/optimizer_15000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1522c09d3659aff31223a5f7791712dc2a8da4083f4fbef02a4f8c835ac2f58f
3
+ size 179406214
attention_kindselective_n_heads4_seed1341/optimizer_17500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2130ab0509e9b1b7a7a739dfc5060d46378b7a1646a02bd49fcd0502934dfdd
3
+ size 179406214
attention_kindselective_n_heads4_seed1341/optimizer_20000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98450d8f0814e9228347e84c6a7a1df848dc332264a30a45534a5b82df90a106
3
+ size 179406214
attention_kindselective_n_heads4_seed1341/optimizer_22500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70de55ca9f452f3956a15f9e4603ea849c6cdd46d4515dd312c1be347f6f27bf
3
+ size 179406214
attention_kindselective_n_heads4_seed1341/optimizer_25000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93781efa827024d34e0380279a80a98957319bd1b649af26527709d924a2fbf0
3
+ size 179406214
attention_kindselective_n_heads4_seed1341/optimizer_27500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0bd709741029562525a20e6b381301456b6720804a9654b9a4a9f7565e7fd60
3
+ size 179406214
attention_kindselective_n_heads4_seed1341/optimizer_30000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:291fe6b9b42cd58551655044bd0bbf4d8f5c81ce97898d00c74cec51fd6bc16c
3
+ size 179406214
attention_kindselective_n_heads4_seed1341/optimizer_32500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c5433712b10d9d0791248973c2582f820c5d035a55dc1a26d4626664ffbdaed7
3
+ size 179406214