Minbyul commited on
Commit
1e93d07
1 Parent(s): 9a67d7f

End of training

Browse files
Files changed (4) hide show
  1. README.md +15 -9
  2. all_results.json +13 -0
  3. config.json +1 -1
  4. eval_results.json +16 -0
README.md CHANGED
@@ -2,9 +2,15 @@
2
  license: llama2
3
  base_model: epfl-llm/meditron-7b
4
  tags:
 
5
  - trl
6
  - dpo
7
  - generated_from_trainer
 
 
 
 
 
8
  model-index:
9
  - name: meditron-7b-dpo-full-wo-live_qa-ep3
10
  results: []
@@ -15,17 +21,17 @@ should probably proofread and complete it, then remove this comment. -->
15
 
16
  # meditron-7b-dpo-full-wo-live_qa-ep3
17
 
18
- This model is a fine-tuned version of [epfl-llm/meditron-7b](https://huggingface.co/epfl-llm/meditron-7b) on an unknown dataset.
19
  It achieves the following results on the evaluation set:
20
  - Loss: 0.5356
21
- - Rewards/chosen: -0.2875
22
- - Rewards/rejected: -0.7751
23
- - Rewards/accuracies: 0.7019
24
- - Rewards/margins: 0.4876
25
- - Logps/rejected: -1205.2603
26
- - Logps/chosen: -986.1399
27
- - Logits/rejected: -0.8883
28
- - Logits/chosen: -0.8904
29
 
30
  ## Model description
31
 
 
2
  license: llama2
3
  base_model: epfl-llm/meditron-7b
4
  tags:
5
+ - alignment-handbook
6
  - trl
7
  - dpo
8
  - generated_from_trainer
9
+ - trl
10
+ - dpo
11
+ - generated_from_trainer
12
+ datasets:
13
+ - HuggingFaceH4/ultrafeedback_binarized
14
  model-index:
15
  - name: meditron-7b-dpo-full-wo-live_qa-ep3
16
  results: []
 
21
 
22
  # meditron-7b-dpo-full-wo-live_qa-ep3
23
 
24
+ This model is a fine-tuned version of [epfl-llm/meditron-7b](https://huggingface.co/epfl-llm/meditron-7b) on the HuggingFaceH4/ultrafeedback_binarized dataset.
25
  It achieves the following results on the evaluation set:
26
  - Loss: 0.5356
27
+ - Rewards/chosen: -0.2871
28
+ - Rewards/rejected: -0.7760
29
+ - Rewards/accuracies: 0.6923
30
+ - Rewards/margins: 0.4889
31
+ - Logps/rejected: -1205.3544
32
+ - Logps/chosen: -986.1032
33
+ - Logits/rejected: -0.8878
34
+ - Logits/chosen: -0.8900
35
 
36
  ## Model description
37
 
all_results.json CHANGED
@@ -1,5 +1,18 @@
1
  {
2
  "epoch": 1.0,
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  "train_loss": 0.539328022330415,
4
  "train_runtime": 3313.1548,
5
  "train_samples": 13051,
 
1
  {
2
  "epoch": 1.0,
3
+ "eval_logits/chosen": -0.8900327086448669,
4
+ "eval_logits/rejected": -0.8877933025360107,
5
+ "eval_logps/chosen": -986.1032104492188,
6
+ "eval_logps/rejected": -1205.3543701171875,
7
+ "eval_loss": 0.5355749130249023,
8
+ "eval_rewards/accuracies": 0.692307710647583,
9
+ "eval_rewards/chosen": -0.28710055351257324,
10
+ "eval_rewards/margins": 0.4889349639415741,
11
+ "eval_rewards/rejected": -0.7760356068611145,
12
+ "eval_runtime": 47.8688,
13
+ "eval_samples": 416,
14
+ "eval_samples_per_second": 8.69,
15
+ "eval_steps_per_second": 0.272,
16
  "train_loss": 0.539328022330415,
17
  "train_runtime": 3313.1548,
18
  "train_samples": 13051,
config.json CHANGED
@@ -23,6 +23,6 @@
23
  "tie_word_embeddings": false,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.39.0.dev0",
26
- "use_cache": false,
27
  "vocab_size": 32017
28
  }
 
23
  "tie_word_embeddings": false,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.39.0.dev0",
26
+ "use_cache": true,
27
  "vocab_size": 32017
28
  }
eval_results.json ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "eval_logits/chosen": -0.8900327086448669,
4
+ "eval_logits/rejected": -0.8877933025360107,
5
+ "eval_logps/chosen": -986.1032104492188,
6
+ "eval_logps/rejected": -1205.3543701171875,
7
+ "eval_loss": 0.5355749130249023,
8
+ "eval_rewards/accuracies": 0.692307710647583,
9
+ "eval_rewards/chosen": -0.28710055351257324,
10
+ "eval_rewards/margins": 0.4889349639415741,
11
+ "eval_rewards/rejected": -0.7760356068611145,
12
+ "eval_runtime": 47.8688,
13
+ "eval_samples": 416,
14
+ "eval_samples_per_second": 8.69,
15
+ "eval_steps_per_second": 0.272
16
+ }