Hugging Face's logo

yakazimir
/
qwen_cpo_entropy_0_1

Text Generation
Transformers
Safetensors
qwen2
alignment-handbook
trl
simpo
Generated from Trainer
conversational
text-generation-inference
Inference Endpoints
Model card Files Community
yakazimir commited on
Commit
72c8015
•
1 Parent(s): b1ccc59

Model save

Browse files
Files changed (4) hide show
  1. README.md +25 -25
  2. all_results.json +4 -4
  3. train_results.json +4 -4
  4. trainer_state.json +0 -0
README.md CHANGED
@@ -18,16 +18,16 @@ should probably proofread and complete it, then remove this comment. -->
18
 
19
  This model is a fine-tuned version of [trl-lib/qwen1.5-0.5b-sft](https://huggingface.co/trl-lib/qwen1.5-0.5b-sft) on an unknown dataset.
20
  It achieves the following results on the evaluation set:
21
- - Loss: 0.7448
22
- - Sft Loss: 1.9881
23
- - Rewards/chosen: -2.0736
24
- - Rewards/rejected: -3.1334
25
- - Rewards/accuracies: 0.7070
26
- - Rewards/margins: 1.0598
27
- - Logps/rejected: -3.1334
28
- - Logps/chosen: -2.0736
29
- - Logits/rejected: 0.5813
30
- - Logits/chosen: 0.4340
31
 
32
  ## Model description
33
 
@@ -46,7 +46,7 @@ More information needed
46
  ### Training hyperparameters
47
 
48
  The following hyperparameters were used during training:
49
- - learning_rate: 3e-06
50
  - train_batch_size: 2
51
  - eval_batch_size: 4
52
  - seed: 42
@@ -62,20 +62,20 @@ The following hyperparameters were used during training:
62
 
63
  | Training Loss | Epoch | Step | Validation Loss | Sft Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
64
  |:-------------:|:------:|:----:|:---------------:|:--------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
65
- | 0.7978 | 0.2141 | 400 | 0.7943 | 1.4613 | -1.4694 | -1.7232 | 0.6046 | 0.2538 | -1.7232 | -1.4694 | 0.3545 | 0.2666 |
66
- | 0.7634 | 0.4282 | 800 | 0.7630 | 1.6008 | -1.6434 | -2.1177 | 0.6461 | 0.4743 | -2.1177 | -1.6434 | 0.2223 | 0.1296 |
67
- | 0.7941 | 0.6422 | 1200 | 0.7446 | 1.6244 | -1.6029 | -2.1044 | 0.6751 | 0.5015 | -2.1044 | -1.6029 | 0.4172 | 0.3041 |
68
- | 0.7152 | 0.8563 | 1600 | 0.7451 | 1.6680 | -1.6570 | -2.1579 | 0.6795 | 0.5009 | -2.1579 | -1.6570 | 0.6005 | 0.4663 |
69
- | 0.7358 | 1.0704 | 2000 | 0.7325 | 1.6955 | -1.6992 | -2.3333 | 0.6825 | 0.6341 | -2.3333 | -1.6992 | 0.5037 | 0.3711 |
70
- | 0.6698 | 1.2845 | 2400 | 0.7332 | 1.8308 | -1.8658 | -2.6575 | 0.7018 | 0.7917 | -2.6575 | -1.8658 | 0.5028 | 0.3708 |
71
- | 0.6975 | 1.4986 | 2800 | 0.7278 | 1.7287 | -1.7409 | -2.4820 | 0.6892 | 0.7411 | -2.4820 | -1.7409 | 0.8810 | 0.7225 |
72
- | 0.7081 | 1.7127 | 3200 | 0.7276 | 1.6777 | -1.6914 | -2.3420 | 0.6773 | 0.6506 | -2.3420 | -1.6914 | 0.5738 | 0.4414 |
73
- | 0.6451 | 1.9267 | 3600 | 0.7215 | 1.7451 | -1.7517 | -2.4919 | 0.6914 | 0.7402 | -2.4919 | -1.7517 | 0.4252 | 0.3011 |
74
- | 0.5342 | 2.1408 | 4000 | 0.7366 | 1.9275 | -1.9957 | -2.9730 | 0.6966 | 0.9773 | -2.9730 | -1.9957 | 0.6010 | 0.4547 |
75
- | 0.5733 | 2.3549 | 4400 | 0.7454 | 1.9592 | -2.0642 | -3.1108 | 0.7055 | 1.0466 | -3.1108 | -2.0642 | 0.5969 | 0.4500 |
76
- | 0.5581 | 2.5690 | 4800 | 0.7417 | 1.9637 | -2.0442 | -3.0719 | 0.7018 | 1.0277 | -3.0719 | -2.0442 | 0.6573 | 0.5046 |
77
- | 0.5281 | 2.7831 | 5200 | 0.7447 | 1.9814 | -2.0666 | -3.1212 | 0.7033 | 1.0546 | -3.1212 | -2.0666 | 0.6008 | 0.4524 |
78
- | 0.5395 | 2.9972 | 5600 | 0.7448 | 1.9881 | -2.0736 | -3.1334 | 0.7070 | 1.0598 | -3.1334 | -2.0736 | 0.5813 | 0.4340 |
79
 
80
 
81
  ### Framework versions
 
18
 
19
  This model is a fine-tuned version of [trl-lib/qwen1.5-0.5b-sft](https://huggingface.co/trl-lib/qwen1.5-0.5b-sft) on an unknown dataset.
20
  It achieves the following results on the evaluation set:
21
+ - Loss: 0.7405
22
+ - Sft Loss: 1.6848
23
+ - Rewards/chosen: -1.7146
24
+ - Rewards/rejected: -2.3727
25
+ - Rewards/accuracies: 0.6773
26
+ - Rewards/margins: 0.6581
27
+ - Logps/rejected: -2.3727
28
+ - Logps/chosen: -1.7146
29
+ - Logits/rejected: 0.3000
30
+ - Logits/chosen: 0.1875
31
 
32
  ## Model description
33
 
 
46
  ### Training hyperparameters
47
 
48
  The following hyperparameters were used during training:
49
+ - learning_rate: 1e-06
50
  - train_batch_size: 2
51
  - eval_batch_size: 4
52
  - seed: 42
 
62
 
63
  | Training Loss | Epoch | Step | Validation Loss | Sft Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
64
  |:-------------:|:------:|:----:|:---------------:|:--------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
65
+ | 0.8248 | 0.2141 | 400 | 0.8255 | 1.3905 | -1.3850 | -1.5360 | 0.5645 | 0.1510 | -1.5360 | -1.3850 | 0.3069 | 0.2210 |
66
+ | 0.7884 | 0.4282 | 800 | 0.7811 | 1.4857 | -1.5199 | -1.8625 | 0.6113 | 0.3426 | -1.8625 | -1.5199 | 0.4914 | 0.3895 |
67
+ | 0.8073 | 0.6422 | 1200 | 0.7653 | 1.5452 | -1.5531 | -1.9756 | 0.6298 | 0.4226 | -1.9756 | -1.5531 | 0.5229 | 0.4111 |
68
+ | 0.7417 | 0.8563 | 1600 | 0.7599 | 1.5652 | -1.5632 | -1.9862 | 0.6484 | 0.4230 | -1.9862 | -1.5632 | 0.5072 | 0.3924 |
69
+ | 0.8212 | 1.0704 | 2000 | 0.7518 | 1.5561 | -1.5506 | -2.0302 | 0.6543 | 0.4796 | -2.0302 | -1.5506 | 0.4351 | 0.3208 |
70
+ | 0.7326 | 1.2845 | 2400 | 0.7455 | 1.6027 | -1.6077 | -2.1582 | 0.6632 | 0.5505 | -2.1582 | -1.6077 | 0.4993 | 0.3799 |
71
+ | 0.7742 | 1.4986 | 2800 | 0.7444 | 1.6196 | -1.6148 | -2.1590 | 0.6632 | 0.5442 | -2.1590 | -1.6148 | 0.4611 | 0.3432 |
72
+ | 0.7597 | 1.7127 | 3200 | 0.7438 | 1.6039 | -1.6049 | -2.1441 | 0.6632 | 0.5392 | -2.1441 | -1.6049 | 0.3926 | 0.2796 |
73
+ | 0.7128 | 1.9267 | 3600 | 0.7399 | 1.6368 | -1.6446 | -2.2337 | 0.6780 | 0.5891 | -2.2337 | -1.6446 | 0.3607 | 0.2486 |
74
+ | 0.6636 | 2.1408 | 4000 | 0.7399 | 1.6738 | -1.6828 | -2.3162 | 0.6780 | 0.6334 | -2.3162 | -1.6828 | 0.3064 | 0.1955 |
75
+ | 0.6929 | 2.3549 | 4400 | 0.7421 | 1.7043 | -1.7385 | -2.4029 | 0.6795 | 0.6644 | -2.4029 | -1.7385 | 0.3030 | 0.1902 |
76
+ | 0.6939 | 2.5690 | 4800 | 0.7411 | 1.6769 | -1.7078 | -2.3536 | 0.6758 | 0.6458 | -2.3536 | -1.7078 | 0.1986 | 0.0944 |
77
+ | 0.6831 | 2.7831 | 5200 | 0.7409 | 1.6830 | -1.7130 | -2.3694 | 0.6766 | 0.6564 | -2.3694 | -1.7130 | 0.3256 | 0.2110 |
78
+ | 0.6951 | 2.9972 | 5600 | 0.7405 | 1.6848 | -1.7146 | -2.3727 | 0.6773 | 0.6581 | -2.3727 | -1.7146 | 0.3000 | 0.1875 |
79
 
80
 
81
  ### Framework versions
all_results.json CHANGED
@@ -1,9 +1,9 @@
1
  {
2
  "epoch": 2.999297541394882,
3
  "total_flos": 0.0,
4
- "train_loss": 0.6688217556451066,
5
- "train_runtime": 34631.8364,
6
  "train_samples": 59790,
7
- "train_samples_per_second": 5.179,
8
- "train_steps_per_second": 0.162
9
  }
 
1
  {
2
  "epoch": 2.999297541394882,
3
  "total_flos": 0.0,
4
+ "train_loss": 0.7456102068299996,
5
+ "train_runtime": 34974.0603,
6
  "train_samples": 59790,
7
+ "train_samples_per_second": 5.129,
8
+ "train_steps_per_second": 0.16
9
  }
train_results.json CHANGED
@@ -1,9 +1,9 @@
1
  {
2
  "epoch": 2.999297541394882,
3
  "total_flos": 0.0,
4
- "train_loss": 0.6688217556451066,
5
- "train_runtime": 34631.8364,
6
  "train_samples": 59790,
7
- "train_samples_per_second": 5.179,
8
- "train_steps_per_second": 0.162
9
  }
 
1
  {
2
  "epoch": 2.999297541394882,
3
  "total_flos": 0.0,
4
+ "train_loss": 0.7456102068299996,
5
+ "train_runtime": 34974.0603,
6
  "train_samples": 59790,
7
+ "train_samples_per_second": 5.129,
8
+ "train_steps_per_second": 0.16
9
  }
trainer_state.json CHANGED
The diff for this file is too large to render. See raw diff