Hugging Face's logo

andrew-healey
/
context-compression

Model card Files
xet
 Community
andrew-healey commited on
Commit
b615303
·
verified ·
1 Parent(s): 619b5a3

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_seed1338/args.json +1 -1
  2. attention_kindselective_n_heads4_seed1338/dataloader_10000.pt +3 -0
  3. attention_kindselective_n_heads4_seed1338/dataloader_12500.pt +3 -0
  4. attention_kindselective_n_heads4_seed1338/dataloader_15000.pt +3 -0
  5. attention_kindselective_n_heads4_seed1338/dataloader_17500.pt +3 -0
  6. attention_kindselective_n_heads4_seed1338/dataloader_20000.pt +3 -0
  7. attention_kindselective_n_heads4_seed1338/dataloader_22500.pt +3 -0
  8. attention_kindselective_n_heads4_seed1338/dataloader_25000.pt +3 -0
  9. attention_kindselective_n_heads4_seed1338/dataloader_27500.pt +3 -0
  10. attention_kindselective_n_heads4_seed1338/dataloader_30000.pt +3 -0
  11. attention_kindselective_n_heads4_seed1338/dataloader_32500.pt +3 -0
  12. attention_kindselective_n_heads4_seed1338/dataloader_35000.pt +3 -0
  13. attention_kindselective_n_heads4_seed1338/dataloader_37500.pt +3 -0
  14. attention_kindselective_n_heads4_seed1338/dataloader_40000.pt +3 -0
  15. attention_kindselective_n_heads4_seed1338/dataloader_42500.pt +3 -0
  16. attention_kindselective_n_heads4_seed1338/dataloader_45000.pt +3 -0
  17. attention_kindselective_n_heads4_seed1338/dataloader_47500.pt +3 -0
  18. attention_kindselective_n_heads4_seed1338/log2.txt +0 -0
  19. attention_kindselective_n_heads4_seed1338/model_02500.pt +1 -1
  20. attention_kindselective_n_heads4_seed1338/model_05000.pt +1 -1
  21. attention_kindselective_n_heads4_seed1338/model_07500.pt +1 -1
  22. attention_kindselective_n_heads4_seed1338/model_10000.pt +3 -0
  23. attention_kindselective_n_heads4_seed1338/model_12500.pt +3 -0
  24. attention_kindselective_n_heads4_seed1338/model_15000.pt +3 -0
  25. attention_kindselective_n_heads4_seed1338/model_17500.pt +3 -0
  26. attention_kindselective_n_heads4_seed1338/model_20000.pt +3 -0
  27. attention_kindselective_n_heads4_seed1338/model_22500.pt +3 -0
  28. attention_kindselective_n_heads4_seed1338/model_25000.pt +3 -0
  29. attention_kindselective_n_heads4_seed1338/model_27500.pt +3 -0
  30. attention_kindselective_n_heads4_seed1338/model_30000.pt +3 -0
  31. attention_kindselective_n_heads4_seed1338/model_32500.pt +3 -0
  32. attention_kindselective_n_heads4_seed1338/model_35000.pt +3 -0
  33. attention_kindselective_n_heads4_seed1338/model_37500.pt +3 -0
  34. attention_kindselective_n_heads4_seed1338/model_40000.pt +3 -0
  35. attention_kindselective_n_heads4_seed1338/model_42500.pt +3 -0
  36. attention_kindselective_n_heads4_seed1338/model_45000.pt +3 -0
  37. attention_kindselective_n_heads4_seed1338/model_47500.pt +3 -0
  38. attention_kindselective_n_heads4_seed1338/model_49999.pt +1 -1
  39. attention_kindselective_n_heads4_seed1338/optimizer_02500.pt +1 -1
  40. attention_kindselective_n_heads4_seed1338/optimizer_05000.pt +1 -1
  41. attention_kindselective_n_heads4_seed1338/optimizer_07500.pt +1 -1
  42. attention_kindselective_n_heads4_seed1338/optimizer_10000.pt +3 -0
  43. attention_kindselective_n_heads4_seed1338/optimizer_12500.pt +3 -0
  44. attention_kindselective_n_heads4_seed1338/optimizer_15000.pt +3 -0
  45. attention_kindselective_n_heads4_seed1338/optimizer_17500.pt +3 -0
  46. attention_kindselective_n_heads4_seed1338/optimizer_20000.pt +3 -0
  47. attention_kindselective_n_heads4_seed1338/optimizer_22500.pt +3 -0
  48. attention_kindselective_n_heads4_seed1338/optimizer_25000.pt +3 -0
  49. attention_kindselective_n_heads4_seed1338/optimizer_27500.pt +3 -0
  50. attention_kindselective_n_heads4_seed1338/optimizer_30000.pt +3 -0
attention_kindselective_n_heads4_seed1338/args.json CHANGED
@@ -1 +1 @@
1
- {"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_7/attention_kindselective_n_heads4_seed1338", "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": 1338, "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_1338", "n_embd": 256}
 
1
+ {"hellaswag": true, "attention_kind": "selective", "log_dir": "wider_is_better_7/attention_kindselective_n_heads4_seed1338", "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": 1338, "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_1338", "n_embd": 256}
attention_kindselective_n_heads4_seed1338/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_seed1338/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_seed1338/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_seed1338/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_seed1338/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_seed1338/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_seed1338/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_seed1338/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_seed1338/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_seed1338/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_seed1338/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_seed1338/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_seed1338/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_seed1338/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_seed1338/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_seed1338/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_heads4_seed1338/log2.txt CHANGED
The diff for this file is too large to render. See raw diff
 
attention_kindselective_n_heads4_seed1338/model_02500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:3c6648672e2ce46ef2c72f27a7e35646a9f331dc839e0c3de86954ef8d7d8291
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:790d8004b55731aa2b1c6522245d6f3fda4f98a78e194fa55d16517cbb1f99d2
3
  size 92843394
attention_kindselective_n_heads4_seed1338/model_05000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9d9ea43744d5ba5a11950db60d43a1c72d13cd6b254504e009b672276c992e55
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b54cb26b05bce7c2e240458a6b2f6c2144f58c4a1d70bc00ca7be280d70d48f6
3
  size 92843394
attention_kindselective_n_heads4_seed1338/model_07500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9cbae5f64034c45aa6d3b77fa65fb10c81552911bd4741998e006320b1711f0c
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:47effc0fbefd294579b455970fee71a436a551f65eefc471f43fe7dd6c6333c3
3
  size 92843394
attention_kindselective_n_heads4_seed1338/model_10000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0abd31abe114c67eebf67710863d9d2ef2abab4f519bf658d0c8ba3e4768171
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_12500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c82b824457a3811b8a356168b6f3651cda934ec8fd53753d67f47f071f267a5
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_15000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b573c71590e026d9a3f418a3c3df5335a37dd481df74868e4066669cd1190beb
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_17500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe75973680a6adf97e27812f43059ca8c25d1611f953621ab059bafcd9d61aa6
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_20000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24924842a4f1e5dd0d646bb70492c485c9fc9b4c7ebf7fffbfb371ce9d5d017f
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_22500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4f31f97107a4c02b5ed5086a7ee97c68673fd430ba182e1810c94f4576dc3e4
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_25000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:578a6ab03a0ca67608b08e918f8224e03636f96947ee4f860a8caa41ade14a75
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_27500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:38860ef4cda4f80c82cf36a4ccf7bea2abf0880d0b47f022279c8bb5d167d61e
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_30000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7cc8f2c0d12c7c88f3044df3fd23cbe37962746b43c69aa071335d97039f270f
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_32500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f555ffef2ec4b0b8e0beab33d831089603e18695955921257fcc6792e47a6c92
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_35000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33db2a07514b7f302c8cf8c4fc6c5309adad2da2e86d1d4d347710084d5feb96
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_37500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a086cf7871c2f9363b70c1816d92662eddb127be1a502a00c4e03d130292fc0
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_40000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a23488d4836b0831fe5a10e10a4305491fd2e1b4a8a0491dd3c70625384e459e
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_42500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67c4bcf2764f145829f47ee4bf02d9962335e720c41dcf85b0dab0338304eb10
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_45000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:65b0a47e1e172bd780bf541bc2401f16cb332c1a1ec84a269ce908d40936489c
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_47500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46d09cb6fc51a4e2e38caf65eb2495f110025f01ba7f4f78c3e9973a5e614764
3
+ size 92843394
attention_kindselective_n_heads4_seed1338/model_49999.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a666cbfe8b8d81c87e55ca0a40ddd37c3e00a2d57171edc5c86e7430b2847152
3
  size 92843394
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f69b2d8cb063137a1fcd4d0dfcc6f6e7ec6e4801ded02dc82fb6b6e6b29c13dd
3
  size 92843394
attention_kindselective_n_heads4_seed1338/optimizer_02500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b5ca883a2c1f5af0edfbb3484142a9889ad2c189ad3100e539e5ec66ad649251
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8c4269ef7b46216196a93f784b32257150ff29810a6c7a943c3d1b51c5a8668
3
  size 179406214
attention_kindselective_n_heads4_seed1338/optimizer_05000.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8e0bd99e08f05d48b4d1f16e815c7e377d6e3544aa099fbd16cb50ee3b4209b5
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2e996f73f5598028ee25500503e610252838124be0c5f0dadb02dc8cec9c34f
3
  size 179406214
attention_kindselective_n_heads4_seed1338/optimizer_07500.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a66617492cb186ff6414f31cd844ed707e6b267b7da4cfc8937de0d3fc3386ee
3
  size 179406214
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:172715e5ec904e580f8358e455be568bc6552aaac5a58405b1c855198515a59b
3
  size 179406214
attention_kindselective_n_heads4_seed1338/optimizer_10000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4c200fd9b08bb420751bb584a60885152f5e0d95f5a3367ebded82b642c634b3
3
+ size 179406214
attention_kindselective_n_heads4_seed1338/optimizer_12500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e0fa46c8e932849a25aca9956f1f3a21c268b082dcd796d683a3998dcdbfaee6
3
+ size 179406214
attention_kindselective_n_heads4_seed1338/optimizer_15000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:256361b42d3b6c290e1a4a1ac8da8c9490cadfca6f3d0e57741e178878600fb4
3
+ size 179406214
attention_kindselective_n_heads4_seed1338/optimizer_17500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a94d13392e04ec23ab5e6a3cfe15c3d368140ee51506294fbd93948b32cdcf7
3
+ size 179406214
attention_kindselective_n_heads4_seed1338/optimizer_20000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e13f474764fc361f163e9fc83362f44e3cc184b3fa5fc4edf62c4bc63332f8bc
3
+ size 179406214
attention_kindselective_n_heads4_seed1338/optimizer_22500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a794e30adc8c4d26a847c60ab6fc31a0727e933ba2454fff46702a3b9e75c013
3
+ size 179406214
attention_kindselective_n_heads4_seed1338/optimizer_25000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:168daadb88ed2d896c0fd76a1713046c69ba4cce786aa853e02b2e63cb1a09ed
3
+ size 179406214
attention_kindselective_n_heads4_seed1338/optimizer_27500.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18e9629944e2b0992c11938d7110eba27f311a6d3bb7d78bdf16a6597cdb4e68
3
+ size 179406214
attention_kindselective_n_heads4_seed1338/optimizer_30000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a1dcbfead3ebc718cbb13f0e8cc6882d63d3c92a949d6851c7866dd45680927
3
+ size 179406214