Hugging Face's logo

andrew-healey
/
context-compression

Model card Files
xet
Community
andrew-healey commited on
Commit
fbc30dd
·
verified ·
1 Parent(s): 3865c47

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_heads2_seed1340/args.json +1 -1
  2. attention_kindselective_n_heads2_seed1340/dataloader_10000.pt +3 -0
  3. attention_kindselective_n_heads2_seed1340/dataloader_12500.pt +3 -0
  4. attention_kindselective_n_heads2_seed1340/dataloader_15000.pt +3 -0
  5. attention_kindselective_n_heads2_seed1340/dataloader_17500.pt +3 -0
  6. attention_kindselective_n_heads2_seed1340/dataloader_20000.pt +3 -0
  7. attention_kindselective_n_heads2_seed1340/dataloader_22500.pt +3 -0
  8. attention_kindselective_n_heads2_seed1340/dataloader_25000.pt +3 -0
  9. attention_kindselective_n_heads2_seed1340/dataloader_27500.pt +3 -0
  10. attention_kindselective_n_heads2_seed1340/dataloader_30000.pt +3 -0
  11. attention_kindselective_n_heads2_seed1340/dataloader_32500.pt +3 -0
  12. attention_kindselective_n_heads2_seed1340/dataloader_35000.pt +3 -0
  13. attention_kindselective_n_heads2_seed1340/dataloader_37500.pt +3 -0
  14. attention_kindselective_n_heads2_seed1340/dataloader_40000.pt +3 -0
  15. attention_kindselective_n_heads2_seed1340/dataloader_42500.pt +3 -0
  16. attention_kindselective_n_heads2_seed1340/dataloader_45000.pt +3 -0
  17. attention_kindselective_n_heads2_seed1340/dataloader_47500.pt +3 -0
  18. attention_kindselective_n_heads2_seed1340/dataloader_49999.pt +3 -0
  19. attention_kindselective_n_heads2_seed1340/log2.txt +0 -0
  20. attention_kindselective_n_heads2_seed1340/model_02500.pt +1 -1
  21. attention_kindselective_n_heads2_seed1340/model_05000.pt +1 -1
  22. attention_kindselective_n_heads2_seed1340/model_07500.pt +1 -1
  23. attention_kindselective_n_heads2_seed1340/model_10000.pt +3 -0
  24. attention_kindselective_n_heads2_seed1340/model_12500.pt +3 -0
  25. attention_kindselective_n_heads2_seed1340/model_15000.pt +3 -0
  26. attention_kindselective_n_heads2_seed1340/model_17500.pt +3 -0
  27. attention_kindselective_n_heads2_seed1340/model_20000.pt +3 -0
  28. attention_kindselective_n_heads2_seed1340/model_22500.pt +3 -0
  29. attention_kindselective_n_heads2_seed1340/model_25000.pt +3 -0
  30. attention_kindselective_n_heads2_seed1340/model_27500.pt +3 -0
  31. attention_kindselective_n_heads2_seed1340/model_30000.pt +3 -0
  32. attention_kindselective_n_heads2_seed1340/model_32500.pt +3 -0
  33. attention_kindselective_n_heads2_seed1340/model_35000.pt +3 -0
  34. attention_kindselective_n_heads2_seed1340/model_37500.pt +3 -0
  35. attention_kindselective_n_heads2_seed1340/model_40000.pt +3 -0
  36. attention_kindselective_n_heads2_seed1340/model_42500.pt +3 -0
  37. attention_kindselective_n_heads2_seed1340/model_45000.pt +3 -0
  38. attention_kindselective_n_heads2_seed1340/model_47500.pt +3 -0
  39. attention_kindselective_n_heads2_seed1340/model_49999.pt +3 -0
  40. attention_kindselective_n_heads2_seed1340/optimizer_02500.pt +1 -1
  41. attention_kindselective_n_heads2_seed1340/optimizer_05000.pt +1 -1
  42. attention_kindselective_n_heads2_seed1340/optimizer_07500.pt +1 -1
  43. attention_kindselective_n_heads2_seed1340/optimizer_10000.pt +3 -0
  44. attention_kindselective_n_heads2_seed1340/optimizer_12500.pt +3 -0
  45. attention_kindselective_n_heads2_seed1340/optimizer_15000.pt +3 -0
  46. attention_kindselective_n_heads2_seed1340/optimizer_17500.pt +3 -0
  47. attention_kindselective_n_heads2_seed1340/optimizer_20000.pt +3 -0
  48. attention_kindselective_n_heads2_seed1340/optimizer_22500.pt +3 -0
  49. attention_kindselective_n_heads2_seed1340/optimizer_25000.pt +3 -0
  50. attention_kindselective_n_heads2_seed1340/optimizer_27500.pt +3 -0
attention_kindselective_n_heads2_seed1340/args.json CHANGED
@@ -1 +1 @@
1
- {"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_6/attention_kindselective_n_heads2_seed1340", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 2, "protect_bos_token": true, "prevent_from_masking_myself": true, "max_steps": 10000, "warmup_steps": 200, "group": "wider_is_better_6", "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": 0.00015, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "15e-5_10240_2_1340", "n_embd": 128}
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_7/attention_kindselective_n_heads2_seed1340", "resume_checkpoint": null, "resume_optimizer": false, "add_a_head": false, "add_head_to_start": true, "new_head_init": "normal", "n_heads": 2, "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": 0.0001, "decay_lr": true, "readout_zero_init": false, "query_zero_init": false, "l1_loss": false, "debugpy": false, "key": "10e-5_10240_2_1340", "n_embd": 128}
attention_kindselective_n_heads2_seed1340/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_heads2_seed1340/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_heads2_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_heads2_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_heads2_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_heads2_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_heads2_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_heads2_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_heads2_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_heads2_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_heads2_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_heads2_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_heads2_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_heads2_seed1340/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_heads2_seed1340/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_heads2_seed1340/dataloader_47500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:55507725e6988f190e4963078652fafa6b68e8d4f79221387612612babf3e1c1
3
+ size 964
attention_kindselective_n_heads2_seed1340/dataloader_49999.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:47776cddb8021172f048a950b83f25b692cb340214b800ce3837c15ceb58907c
3
+ size 964
attention_kindselective_n_heads2_seed1340/log2.txt CHANGED
The diff for this file is too large to render. See raw diff
 
attention_kindselective_n_heads2_seed1340/model_02500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:db5e71e2651a7e4c04adf1b1bdd2fd67898b9838ff7ce2857cd735f4766b5532
3
  size 38587970
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a43c3fea68b44f6e94a64fabd4bec96f4af33541fb79fc8ab291548d61ad3ba8
3
  size 38587970
attention_kindselective_n_heads2_seed1340/model_05000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:87a3a4e1a720bf19b1c7f850b83a50505c837892b3964a6f17b36ac335b8a606
3
  size 38587970
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cedd4c5d0d207b268fe96d32662f008ce85caf9422332a5f205480009a0bae8a
3
  size 38587970
attention_kindselective_n_heads2_seed1340/model_07500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:933638ce43638ef9bd5900408f7bbcb30a32be65bd1c3e240216a1265da81d33
3
  size 38587970
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cf3ae491c351dd98c03ca4270a47dadf6b7299816fd4d4d28e612be080c8eaf6
3
  size 38587970
attention_kindselective_n_heads2_seed1340/model_10000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d3917438735ab535c976c8cf747154e5ff137b3226377c17494de2db52a5dbc9
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_12500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82f39c782dc0ecfa52b440f257e68f9fbd5d1c95cb1327b31ba1a4a90fddbd14
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_15000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c9a999d0f39f00dac906e79883b2c60cacea327110d17ce2aeb2ca063f6dd405
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_17500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:77730d4395b567ca5bcbb5ee491c47a6347ea335cdafbcaa57606e056f99d5be
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_20000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c1b0f211ab6773a5b61676a5550d5e97c97a33cb69ba778acf90934dbc1edd6b
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_22500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:704498bc3ab665573a16170d899809e4f5b2e0433ee9568c75b1a299e86e63d0
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_25000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:abad2d98112c3e440e66c5dbf182d0a97537a774c60116df1cc4e601a19b6d31
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_27500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:231c011ec9ed5d97d25c7481c92c73d78b16474dea2f6e86258c8ef2055ce77f
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_30000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c5fa0b4c0e1c96c7546e72b0608230b2b1be0609b274c5f83c59202f5463a12
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_32500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7db2e6312252b187d62e6ebdbc4915b0d541396bd5728579d3b5167316a4611
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_35000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:49399b8ac5d522d2716a77e73e6a22da180c585b1c865fdc6647f0186d199c0a
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_37500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5cee4edf4b83913b1bfc2b1150a9871f06b8205e075d2edd1d1f43cc015cd3ef
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_40000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e12690a1fa5e37625d4caa7ed93f085e14857ea2fd92fde5d9c686eaadf988cd
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_42500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8fbfa201aad46b6a3b9db8b8a6e3ffe19b2b0bf7135c941a10352a0f659dbd2f
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_45000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9dd2d44c2ef7a35bb4a5a6a4caae579ea0b5bbefab52fcc539b46efb5e230c44
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_47500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:671e09d821aad0b05acfd58c4cb5943479619d40c185ec2b6e6081321e58589a
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/model_49999.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ade59c1a7b1f27604ba9d522da69929f07674b7e48af67cb6f07699745ecf35
3
+ size 38587970
attention_kindselective_n_heads2_seed1340/optimizer_02500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ad182381a14e875a3c04a9d0fce780b0641e040b738f8454b13df6255a740e34
3
  size 70895430
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f8b2588354f55480e0234fb6a453af90f8dce086ada918b3cfb81acfd1ade067
3
  size 70895430
attention_kindselective_n_heads2_seed1340/optimizer_05000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:89e51f2fe0afb3163bdb42b8d6c2d3968fb0c1d3a51b33ea7ae5061524b2b719
3
  size 70895430
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a0fd46e04dbb546ef5f7f70b8792710a6b2c3d746613f6b100e414a89e996728
3
  size 70895430
attention_kindselective_n_heads2_seed1340/optimizer_07500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2c475776587ad7676fd824bc7c9c9570e94a3ffb32776966605d3ccf70784164
3
  size 70895430
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06a62406b060c40f3fc2ef5b420b68e53741a74ba8388e186bc8ea3757ec7472
3
  size 70895430
attention_kindselective_n_heads2_seed1340/optimizer_10000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fca96eb06f8fcae983adc5b5fd192a2b465e885c3ce28a93a16dff13711aed2a
3
+ size 70895430
attention_kindselective_n_heads2_seed1340/optimizer_12500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:976e03738be53dea6d55c73ba3ec3c146d8423d50974c2d70bf10f0f5942be6a
3
+ size 70895430
attention_kindselective_n_heads2_seed1340/optimizer_15000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ccb9e25a713df0525fc8663aa13e51a23de1821fe91a03628167abe36ea38647
3
+ size 70895430
attention_kindselective_n_heads2_seed1340/optimizer_17500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f09d15b8a367be854baf5c8608688081f0710680a6364d6cfac81c9d751acf67
3
+ size 70895430
attention_kindselective_n_heads2_seed1340/optimizer_20000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e238640eaa708aa5d306f56d91790321b5bfa8e6c0dc770b07dea34c1eb7de1f
3
+ size 70895430
attention_kindselective_n_heads2_seed1340/optimizer_22500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b84fcade25645820cc712b0b50986685542b78af0f11c9d5779617e41883b28
3
+ size 70895430
attention_kindselective_n_heads2_seed1340/optimizer_25000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0345fb665f4a9124835ba1fa36906d37e6402ad0cc34129b35cf08f1f03a412
3
+ size 70895430
attention_kindselective_n_heads2_seed1340/optimizer_27500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a7b48a02e56eaa60ecd353f6f6ca16dabbe3aca0844593838434ead44e59fa4
3
+ size 70895430