Hugging Face's logo

Batnini
/
knowledge-agent-write-back

Safetensors
Model card Files
xet
 Community
Batnini commited on
Commit
40a707f
·
verified ·
1 Parent(s): c727534

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. adapter/adapter_config.json +5 -5
  2. adapter/adapter_model.safetensors +1 -1
  3. adapter/checkpoint-1/adapter_config.json +5 -5
  4. adapter/checkpoint-10/adapter_config.json +5 -5
  5. adapter/checkpoint-10/adapter_model.safetensors +1 -1
  6. adapter/checkpoint-10/optimizer.pt +1 -1
  7. adapter/checkpoint-10/trainer_state.json +4 -4
  8. adapter/checkpoint-11/adapter_config.json +5 -5
  9. adapter/checkpoint-11/adapter_model.safetensors +1 -1
  10. adapter/checkpoint-11/optimizer.pt +1 -1
  11. adapter/checkpoint-11/trainer_state.json +4 -4
  12. adapter/checkpoint-12/adapter_config.json +5 -5
  13. adapter/checkpoint-12/adapter_model.safetensors +1 -1
  14. adapter/checkpoint-12/optimizer.pt +1 -1
  15. adapter/checkpoint-12/trainer_state.json +4 -4
  16. adapter/checkpoint-13/adapter_config.json +5 -5
  17. adapter/checkpoint-13/adapter_model.safetensors +1 -1
  18. adapter/checkpoint-13/optimizer.pt +1 -1
  19. adapter/checkpoint-13/trainer_state.json +4 -4
  20. adapter/checkpoint-14/adapter_config.json +5 -5
  21. adapter/checkpoint-14/adapter_model.safetensors +1 -1
  22. adapter/checkpoint-14/optimizer.pt +1 -1
  23. adapter/checkpoint-14/trainer_state.json +4 -4
  24. adapter/checkpoint-15/adapter_config.json +5 -5
  25. adapter/checkpoint-15/adapter_model.safetensors +1 -1
  26. adapter/checkpoint-15/optimizer.pt +1 -1
  27. adapter/checkpoint-15/trainer_state.json +4 -4
  28. adapter/checkpoint-2/adapter_config.json +5 -5
  29. adapter/checkpoint-2/adapter_model.safetensors +1 -1
  30. adapter/checkpoint-2/optimizer.pt +1 -1
  31. adapter/checkpoint-3/adapter_config.json +5 -5
  32. adapter/checkpoint-3/adapter_model.safetensors +1 -1
  33. adapter/checkpoint-3/optimizer.pt +1 -1
  34. adapter/checkpoint-4/adapter_config.json +5 -5
  35. adapter/checkpoint-4/adapter_model.safetensors +1 -1
  36. adapter/checkpoint-4/optimizer.pt +1 -1
  37. adapter/checkpoint-5/adapter_config.json +5 -5
  38. adapter/checkpoint-5/adapter_model.safetensors +1 -1
  39. adapter/checkpoint-5/optimizer.pt +1 -1
  40. adapter/checkpoint-6/adapter_config.json +5 -5
  41. adapter/checkpoint-6/adapter_model.safetensors +1 -1
  42. adapter/checkpoint-6/optimizer.pt +1 -1
  43. adapter/checkpoint-7/adapter_config.json +5 -5
  44. adapter/checkpoint-7/adapter_model.safetensors +1 -1
  45. adapter/checkpoint-7/optimizer.pt +1 -1
  46. adapter/checkpoint-8/adapter_config.json +5 -5
  47. adapter/checkpoint-8/adapter_model.safetensors +1 -1
  48. adapter/checkpoint-8/optimizer.pt +1 -1
  49. adapter/checkpoint-9/adapter_config.json +5 -5
  50. adapter/checkpoint-9/adapter_model.safetensors +1 -1
adapter/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e3be2c8518433696a6aa5b088a43823facc099b20c7beab385e9352ec3023f2e
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a595d437f4fcb0c6030840d9e70e7ff6818a5e490943aabe19f00c3dee6ba9c5
3
  size 239536776
adapter/checkpoint-1/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-10/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-10/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:05d17ae14cc83cb1bacf66ddbf8a731aa95e7ee88c27331d712b7f05316ab93b
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2bf31fa16c7c88660017f477ad7d02c7780ab77c6958e8963d96b8136b997444
3
  size 239536776
adapter/checkpoint-10/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5f101ceedd36da51a74e912ba21ac9744d4bcf49d3acf3677c9a6a4f2f45e9b5
3
  size 243818437
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3ba1ae4d12685af72d6883a2ab8c658110ab1baf47b7c5129877df055c8f519f
3
  size 243818437
adapter/checkpoint-10/trainer_state.json CHANGED
@@ -10,12 +10,12 @@
10
  "is_world_process_zero": true,
11
  "log_history": [
12
  {
13
- "entropy": 0.8513693287968636,
14
  "epoch": 10.0,
15
- "grad_norm": 1.34375,
16
  "learning_rate": 8e-05,
17
- "loss": 1.310697078704834,
18
- "mean_token_accuracy": 0.8319162212312221,
19
  "num_tokens": 6910.0,
20
  "step": 10
21
  }
 
10
  "is_world_process_zero": true,
11
  "log_history": [
12
  {
13
+ "entropy": 3.8962098330259325,
14
  "epoch": 10.0,
15
+ "grad_norm": 368.0,
16
  "learning_rate": 8e-05,
17
+ "loss": 5.190721893310547,
18
+ "mean_token_accuracy": 0.37155905868858097,
19
  "num_tokens": 6910.0,
20
  "step": 10
21
  }
adapter/checkpoint-11/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-11/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9d9681efd88617c5414ad8a2f16a9fe33ff44ea2cff0324cee219e52ebabff7c
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1349bcf09b314d4e7c6011d14578a8ed965f03e7221a72299920aad5ff281fe3
3
  size 239536776
adapter/checkpoint-11/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2fa7a3a67810ffd51f4413bb310c5e0b07749a8aef580fb8d5694adec4318e1a
3
  size 243818437
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d81b388a1cf32781d57740a88bc414884439ca3d4c067b097af482971c8c036e
3
  size 243818437
adapter/checkpoint-11/trainer_state.json CHANGED
@@ -10,12 +10,12 @@
10
  "is_world_process_zero": true,
11
  "log_history": [
12
  {
13
- "entropy": 0.8513693287968636,
14
  "epoch": 10.0,
15
- "grad_norm": 1.34375,
16
  "learning_rate": 8e-05,
17
- "loss": 1.310697078704834,
18
- "mean_token_accuracy": 0.8319162212312221,
19
  "num_tokens": 6910.0,
20
  "step": 10
21
  }
 
10
  "is_world_process_zero": true,
11
  "log_history": [
12
  {
13
+ "entropy": 3.8962098330259325,
14
  "epoch": 10.0,
15
+ "grad_norm": 368.0,
16
  "learning_rate": 8e-05,
17
+ "loss": 5.190721893310547,
18
+ "mean_token_accuracy": 0.37155905868858097,
19
  "num_tokens": 6910.0,
20
  "step": 10
21
  }
adapter/checkpoint-12/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-12/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6e2df4368fe55f5a79123103fc8d9321193f5a1c9b4771867f8c8a0e1d214108
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:595eef2f37979c4a21aa3468807a944d4ef916f631a812a27c6d807dd1e44d7b
3
  size 239536776
adapter/checkpoint-12/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:effe09760ff16f3e0e212c42ab9a3bf3df4644871b0cd48be8e746b423d3045a
3
  size 243818437
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a77a4ed360343bc275a374d0f7647552aa5145e9ae1d8312a8af34f8a4fd404e
3
  size 243818437
adapter/checkpoint-12/trainer_state.json CHANGED
@@ -10,12 +10,12 @@
10
  "is_world_process_zero": true,
11
  "log_history": [
12
  {
13
- "entropy": 0.8513693287968636,
14
  "epoch": 10.0,
15
- "grad_norm": 1.34375,
16
  "learning_rate": 8e-05,
17
- "loss": 1.310697078704834,
18
- "mean_token_accuracy": 0.8319162212312221,
19
  "num_tokens": 6910.0,
20
  "step": 10
21
  }
 
10
  "is_world_process_zero": true,
11
  "log_history": [
12
  {
13
+ "entropy": 3.8962098330259325,
14
  "epoch": 10.0,
15
+ "grad_norm": 368.0,
16
  "learning_rate": 8e-05,
17
+ "loss": 5.190721893310547,
18
+ "mean_token_accuracy": 0.37155905868858097,
19
  "num_tokens": 6910.0,
20
  "step": 10
21
  }
adapter/checkpoint-13/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-13/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:90c8bf5cc5e3a9080536acd4e22d9b395f3b96ac285a926ec195764e9a3ac920
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f4f5dcf4854154b3748d6241b60e69d6cf4d81671433917d80649ae3e7ebbec
3
  size 239536776
adapter/checkpoint-13/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8de3a29fd07f2ee3595d0ac98721df096041694f03ff5aa11649d155f5ecb5ca
3
  size 243818437
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e88af799967be423526730277d7cde331772ddc7f14bfca26f3bd4bbb34ca4e8
3
  size 243818437
adapter/checkpoint-13/trainer_state.json CHANGED
@@ -10,12 +10,12 @@
10
  "is_world_process_zero": true,
11
  "log_history": [
12
  {
13
- "entropy": 0.8513693287968636,
14
  "epoch": 10.0,
15
- "grad_norm": 1.34375,
16
  "learning_rate": 8e-05,
17
- "loss": 1.310697078704834,
18
- "mean_token_accuracy": 0.8319162212312221,
19
  "num_tokens": 6910.0,
20
  "step": 10
21
  }
 
10
  "is_world_process_zero": true,
11
  "log_history": [
12
  {
13
+ "entropy": 3.8962098330259325,
14
  "epoch": 10.0,
15
+ "grad_norm": 368.0,
16
  "learning_rate": 8e-05,
17
+ "loss": 5.190721893310547,
18
+ "mean_token_accuracy": 0.37155905868858097,
19
  "num_tokens": 6910.0,
20
  "step": 10
21
  }
adapter/checkpoint-14/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-14/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d50aa8b388ec8733ac44bd472b9465ce8dd25750a07b47bb921b8abe78e93449
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61f5c3b0328c37da03c486ade9c2f7c128cefbbb64a64d30331445806c06066c
3
  size 239536776
adapter/checkpoint-14/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b435012f854d460c7c66d1e22a7d951f7b7fc22aa9f957bc2f757111325d4c16
3
  size 243818437
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d40565a8090ccbd941560ea637ff66f1a1d5f370441a578ceede47830bff632d
3
  size 243818437
adapter/checkpoint-14/trainer_state.json CHANGED
@@ -10,12 +10,12 @@
10
  "is_world_process_zero": true,
11
  "log_history": [
12
  {
13
- "entropy": 0.8513693287968636,
14
  "epoch": 10.0,
15
- "grad_norm": 1.34375,
16
  "learning_rate": 8e-05,
17
- "loss": 1.310697078704834,
18
- "mean_token_accuracy": 0.8319162212312221,
19
  "num_tokens": 6910.0,
20
  "step": 10
21
  }
 
10
  "is_world_process_zero": true,
11
  "log_history": [
12
  {
13
+ "entropy": 3.8962098330259325,
14
  "epoch": 10.0,
15
+ "grad_norm": 368.0,
16
  "learning_rate": 8e-05,
17
+ "loss": 5.190721893310547,
18
+ "mean_token_accuracy": 0.37155905868858097,
19
  "num_tokens": 6910.0,
20
  "step": 10
21
  }
adapter/checkpoint-15/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-15/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e3be2c8518433696a6aa5b088a43823facc099b20c7beab385e9352ec3023f2e
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a595d437f4fcb0c6030840d9e70e7ff6818a5e490943aabe19f00c3dee6ba9c5
3
  size 239536776
adapter/checkpoint-15/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b4935ec169e25391ef800c9705cd37a3a5b35a3c6cde2dd9395cae3ff03c3ee2
3
  size 243818437
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cfea4b6d476419e34dcd29293fede2338729433dce76a040899df509fcb32b5a
3
  size 243818437
adapter/checkpoint-15/trainer_state.json CHANGED
@@ -10,12 +10,12 @@
10
  "is_world_process_zero": true,
11
  "log_history": [
12
  {
13
- "entropy": 0.8513693287968636,
14
  "epoch": 10.0,
15
- "grad_norm": 1.34375,
16
  "learning_rate": 8e-05,
17
- "loss": 1.310697078704834,
18
- "mean_token_accuracy": 0.8319162212312221,
19
  "num_tokens": 6910.0,
20
  "step": 10
21
  }
 
10
  "is_world_process_zero": true,
11
  "log_history": [
12
  {
13
+ "entropy": 3.8962098330259325,
14
  "epoch": 10.0,
15
+ "grad_norm": 368.0,
16
  "learning_rate": 8e-05,
17
+ "loss": 5.190721893310547,
18
+ "mean_token_accuracy": 0.37155905868858097,
19
  "num_tokens": 6910.0,
20
  "step": 10
21
  }
adapter/checkpoint-2/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-2/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:43e7a048bea5c9b88eae829f5a6ea192d63a9a9ee3be127d8931054344774a0b
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5f28440ff3f267bc7a6e3529f9ba86a1da9431cea7748c7014fd80838bae088
3
  size 239536776
adapter/checkpoint-2/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b7b530d3378e9cbb1e3f4ffbcd10e7d058234260cffaeab63775acb44a4da560
3
  size 243818437
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:73cbb9ecaa22640c0fffda414474bb65a07691b0d054549c3c56b7d87a77c80b
3
  size 243818437
adapter/checkpoint-3/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-3/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:63a2df1a70456844c689f40525f9c3ac54f76c0275316566a7b60d88668ba86d
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd6f8f22893645472442d2ac34a88966a12206dc4b8fe844e92a46c2d7841f14
3
  size 239536776
adapter/checkpoint-3/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ea813c605e804352ba8ff66dd96e9f8683da294ca935813cb507520dff866fd7
3
  size 243818437
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:35edc07011d9a778a6f92cddb94e16b5cf91466abd22943cf379ef288ac6c4a5
3
  size 243818437
adapter/checkpoint-4/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-4/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:0b759651321e274d2fac4ede5eae09ec051c644ddc3da7cd711b4fee072ff7ab
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07c9721da7ae65cbc73b276b059a18f46de2579d51182898a55b6a81b25a3b70
3
  size 239536776
adapter/checkpoint-4/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d70fbb46583d0982619b54aa0647e44b07513ed9d5bbe96c792d690c4cd82385
3
  size 243818437
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c087c0ee39c0a3dbf77220bb7a45e3e49d2f7ad1c07b92960d3903c21220d29
3
  size 243818437
adapter/checkpoint-5/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-5/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:29e56d014bad376378d3b34998bcb3b3b9fb51b51b3f5db99fc3198c9852018d
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0cc0ebb8aeefb322a48df1be8b5363be8a5a8f982e41070c3061e67b753c6ddd
3
  size 239536776
adapter/checkpoint-5/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c01e814587fc44eb23dc47e7d9794565fa2c4a0982fc0e8ba8e92e6933f5bf3e
3
  size 243818437
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd38cc2cb9cc376bd47c7b1eeabc584bd125abb715b1ad3b67cf702869dfe5ad
3
  size 243818437
adapter/checkpoint-6/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-6/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2d520e27667af63bf65784a8245076234ec27b98f7e084552481b2c0ac118a13
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac2b4923da7aa08825a2ef8e36bd4e2eb438de607846aaa3edcf5eeddbce4031
3
  size 239536776
adapter/checkpoint-6/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2fa67e6c425f93c8b24ac0d40e17dd9d2e7902edc418a3ff7636da87ffc1da8f
3
  size 243818437
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06ad60be8ae0e218d0edace7e76e29afe97d9474c0108b82bd962d94e08ce99f
3
  size 243818437
adapter/checkpoint-7/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-7/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6c8b39faf3a2c769ccbcd0f4bf97596484347460e0b96cdc227c636c8f7e523d
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ffc9e1a2fb5f52dba2d02e58e21b841caf69d5cc3db65765036878a332b29617
3
  size 239536776
adapter/checkpoint-7/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b94dbe7064b764155bb9e2b3d791bb4f4a5fa064cd4d2ffbbd4fbb5fdcd4b70b
3
  size 243818437
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:618eb287cd639948b42b7172333b50379df12661b16c7ebd1f944d841f833f81
3
  size 243818437
adapter/checkpoint-8/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-8/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:00f650b9434b1e9710dbda8ecdf0913aabff56e76246b504a909c8bf1cfd5820
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:20fe7d37644a3c5e3936894db23065925af1124fbea2f3deae0ac82efdb88916
3
  size 239536776
adapter/checkpoint-8/optimizer.pt CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:16011f2abf7e734e798897529fdb53e299bc3ae0e164decc6417556b83bc625f
3
  size 243818437
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:befc8812f05ed59f776319cb845c37271e189ef00b30bccb0a62754bba7fa75e
3
  size 243818437
adapter/checkpoint-9/adapter_config.json CHANGED
@@ -31,13 +31,13 @@
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
34
- "o_proj",
35
  "up_proj",
36
- "k_proj",
37
  "gate_proj",
38
- "down_proj",
39
  "v_proj",
40
- "q_proj"
 
 
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
@@ -45,6 +45,6 @@
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
- "use_rslora": false,
49
  "velora_config": null
50
  }
 
31
  "rank_pattern": {},
32
  "revision": null,
33
  "target_modules": [
 
34
  "up_proj",
 
35
  "gate_proj",
36
+ "o_proj",
37
  "v_proj",
38
+ "down_proj",
39
+ "q_proj",
40
+ "k_proj"
41
  ],
42
  "target_parameters": null,
43
  "task_type": "CAUSAL_LM",
 
45
  "use_bdlora": null,
46
  "use_dora": false,
47
  "use_qalora": false,
48
+ "use_rslora": true,
49
  "velora_config": null
50
  }
adapter/checkpoint-9/adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:034d5337cafb6b34232eb6a40723e4d87caba75cdc73b10f25fa97eefb8a99a5
3
  size 239536776
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba07341a810f242ae89c79e431175a1eb5ca4db3a3a5da5d13cc4c33410d1f76
3
  size 239536776