Hugging Face's logo

amirabdullah19852020
/
gpt-neo-125m_hh_reward

Text Generation
Transformers
Safetensors
gpt_neo
trl
dpo
Generated from Trainer
Inference Endpoints
Model card Files Community
amirabdullah19852020 commited on
Commit
8b94c3a
•
1 Parent(s): 9ca671f

End of training

Browse files
Files changed (3) hide show
  1. README.md +20 -28
  2. generation_config.json +1 -1
  3. model.safetensors +1 -1
README.md CHANGED
@@ -15,17 +15,17 @@ should probably proofread and complete it, then remove this comment. -->
15
 
16
  # gpt-neo-125m_hh_reward
17
 
18
- This model is a DPO fine-tuned version of [EleutherAI/gpt-neo-125m](https://huggingface.co/EleutherAI/gpt-neo-125m) on Anthropics HH dataset.
19
  It achieves the following results on the evaluation set:
20
- - Loss: 0.7764
21
- - Rewards/chosen: -1.0726
22
- - Rewards/rejected: -1.1588
23
- - Rewards/accuracies: 0.5592
24
- - Rewards/margins: 0.0861
25
- - Logps/rejected: -127.6699
26
- - Logps/chosen: -107.5507
27
- - Logits/rejected: -15.9221
28
- - Logits/chosen: -15.7406
29
 
30
  ## Model description
31
 
@@ -44,34 +44,26 @@ More information needed
44
  ### Training hyperparameters
45
 
46
  The following hyperparameters were used during training:
47
- - learning_rate: 1e-05
48
- - train_batch_size: 8
49
  - eval_batch_size: 8
50
  - seed: 42
51
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
52
  - lr_scheduler_type: linear
53
  - lr_scheduler_warmup_steps: 150
54
- - training_steps: 20050
55
 
56
  ### Training results
57
 
58
- | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
59
- |:-------------:|:-----:|:-----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
60
- | 0.7685 | 0.1 | 2000 | 0.7585 | -0.6757 | -0.7643 | 0.5066 | 0.0886 | -126.6837 | -106.5583 | -16.3107 | -16.1234 |
61
- | 0.795 | 0.2 | 4000 | 0.7531 | -0.9330 | -1.0793 | 0.5625 | 0.1463 | -127.4711 | -107.2016 | -16.6020 | -16.4468 |
62
- | 0.7366 | 0.3 | 6000 | 0.7961 | -1.0751 | -1.1163 | 0.5033 | 0.0412 | -127.5638 | -107.5568 | -16.3735 | -16.2226 |
63
- | 0.8288 | 0.4 | 8000 | 0.7860 | -1.0546 | -1.1457 | 0.5559 | 0.0911 | -127.6372 | -107.5056 | -16.4345 | -16.2588 |
64
- | 0.7272 | 0.5 | 10000 | 0.7767 | -1.1017 | -1.2133 | 0.5658 | 0.1116 | -127.8062 | -107.6233 | -16.2325 | -16.0800 |
65
- | 0.8404 | 0.6 | 12000 | 0.8029 | -1.2133 | -1.2948 | 0.5329 | 0.0815 | -128.0099 | -107.9024 | -15.9207 | -15.7572 |
66
- | 0.8224 | 0.7 | 14000 | 0.7702 | -1.0578 | -1.1817 | 0.5625 | 0.1239 | -127.7272 | -107.5137 | -15.9578 | -15.7879 |
67
- | 0.7267 | 0.8 | 16000 | 0.7929 | -1.1534 | -1.2255 | 0.5559 | 0.0721 | -127.8368 | -107.7526 | -16.0551 | -15.8740 |
68
- | 0.6329 | 0.9 | 18000 | 0.8052 | -1.0222 | -1.0692 | 0.5428 | 0.0471 | -127.4461 | -107.4246 | -15.8842 | -15.7008 |
69
- | 0.7247 | 1.0 | 20000 | 0.7764 | -1.0726 | -1.1588 | 0.5592 | 0.0861 | -127.6699 | -107.5507 | -15.9221 | -15.7406 |
70
 
71
 
72
  ### Framework versions
73
 
74
- - Transformers 4.36.2
75
  - Pytorch 2.0.1+cu117
76
- - Datasets 2.14.5
77
- - Tokenizers 0.15.0
 
15
 
16
  # gpt-neo-125m_hh_reward
17
 
18
+ This model is a fine-tuned version of [EleutherAI/gpt-neo-125m](https://huggingface.co/EleutherAI/gpt-neo-125m) on an unknown dataset.
19
  It achieves the following results on the evaluation set:
20
+ - Loss: 0.7503
21
+ - Rewards/chosen: -4.2523
22
+ - Rewards/rejected: -4.3731
23
+ - Rewards/accuracies: 0.5625
24
+ - Rewards/margins: 0.1208
25
+ - Logps/rejected: -168.5040
26
+ - Logps/chosen: -147.3926
27
+ - Logits/rejected: -11.6528
28
+ - Logits/chosen: -11.5062
29
 
30
  ## Model description
31
 
 
44
  ### Training hyperparameters
45
 
46
  The following hyperparameters were used during training:
47
+ - learning_rate: 0.0001
48
+ - train_batch_size: 16
49
  - eval_batch_size: 8
50
  - seed: 42
51
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
52
  - lr_scheduler_type: linear
53
  - lr_scheduler_warmup_steps: 150
54
+ - training_steps: 4050
55
 
56
  ### Training results
57
 
58
+ | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
59
+ |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
60
+ | 0.8022 | 0.2 | 2000 | 0.7737 | -4.8718 | -5.0523 | 0.5724 | 0.1805 | -175.2956 | -153.5872 | -11.7730 | -11.6673 |
61
+ | 0.7336 | 0.4 | 4000 | 0.7503 | -4.2523 | -4.3731 | 0.5625 | 0.1208 | -168.5040 | -147.3926 | -11.6528 | -11.5062 |
 
 
 
 
 
 
 
 
62
 
63
 
64
  ### Framework versions
65
 
66
+ - Transformers 4.38.2
67
  - Pytorch 2.0.1+cu117
68
+ - Datasets 2.17.1
69
+ - Tokenizers 0.15.2
generation_config.json CHANGED
@@ -2,5 +2,5 @@
2
  "_from_model_config": true,
3
  "bos_token_id": 50256,
4
  "eos_token_id": 50256,
5
- "transformers_version": "4.36.2"
6
  }
 
2
  "_from_model_config": true,
3
  "bos_token_id": 50256,
4
  "eos_token_id": 50256,
5
+ "transformers_version": "4.38.2"
6
  }
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7a272fd5c1346c4832f6e28703ac625e58b1486f2731305e52f8e4abf42e8571
3
  size 500811336
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a79b11394c73f336dcf58ed736a6c6ac84d4272b257781ecf159009957210fc0
3
  size 500811336