Minbyul commited on
Commit
6d7651b
1 Parent(s): 2f6c378

End of training

Browse files
Files changed (4) hide show
  1. README.md +17 -1
  2. all_results.json +13 -0
  3. config.json +1 -1
  4. eval_results.json +16 -0
README.md CHANGED
@@ -1,9 +1,15 @@
1
  ---
2
  base_model: Minbyul/selfbiorag-7b-wo-healthsearch_qa-sft
3
  tags:
 
4
  - trl
5
  - dpo
6
  - generated_from_trainer
 
 
 
 
 
7
  model-index:
8
  - name: selfbiorag-7b-dpo-full-sft-wo-healthsearch_qa
9
  results: []
@@ -14,7 +20,17 @@ should probably proofread and complete it, then remove this comment. -->
14
 
15
  # selfbiorag-7b-dpo-full-sft-wo-healthsearch_qa
16
 
17
- This model is a fine-tuned version of [Minbyul/selfbiorag-7b-wo-healthsearch_qa-sft](https://huggingface.co/Minbyul/selfbiorag-7b-wo-healthsearch_qa-sft) on an unknown dataset.
 
 
 
 
 
 
 
 
 
 
18
 
19
  ## Model description
20
 
 
1
  ---
2
  base_model: Minbyul/selfbiorag-7b-wo-healthsearch_qa-sft
3
  tags:
4
+ - alignment-handbook
5
  - trl
6
  - dpo
7
  - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ - generated_from_trainer
11
+ datasets:
12
+ - HuggingFaceH4/ultrafeedback_binarized
13
  model-index:
14
  - name: selfbiorag-7b-dpo-full-sft-wo-healthsearch_qa
15
  results: []
 
20
 
21
  # selfbiorag-7b-dpo-full-sft-wo-healthsearch_qa
22
 
23
+ This model is a fine-tuned version of [Minbyul/selfbiorag-7b-wo-healthsearch_qa-sft](https://huggingface.co/Minbyul/selfbiorag-7b-wo-healthsearch_qa-sft) on the HuggingFaceH4/ultrafeedback_binarized dataset.
24
+ It achieves the following results on the evaluation set:
25
+ - Loss: 0.4465
26
+ - Rewards/chosen: -0.5929
27
+ - Rewards/rejected: -1.6772
28
+ - Rewards/accuracies: 0.7846
29
+ - Rewards/margins: 1.0843
30
+ - Logps/rejected: -1480.8429
31
+ - Logps/chosen: -984.8102
32
+ - Logits/rejected: -3.4642
33
+ - Logits/chosen: -2.6475
34
 
35
  ## Model description
36
 
all_results.json CHANGED
@@ -1,5 +1,18 @@
1
  {
2
  "epoch": 1.0,
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  "train_loss": 0.2915988182410216,
4
  "train_runtime": 1026.489,
5
  "train_samples": 4992,
 
1
  {
2
  "epoch": 1.0,
3
+ "eval_logits/chosen": -2.6475114822387695,
4
+ "eval_logits/rejected": -3.464193344116211,
5
+ "eval_logps/chosen": -984.8101806640625,
6
+ "eval_logps/rejected": -1480.8428955078125,
7
+ "eval_loss": 0.4465055763721466,
8
+ "eval_rewards/accuracies": 0.7845849990844727,
9
+ "eval_rewards/chosen": -0.5929392576217651,
10
+ "eval_rewards/margins": 1.0842887163162231,
11
+ "eval_rewards/rejected": -1.6772277355194092,
12
+ "eval_runtime": 1610.6329,
13
+ "eval_samples": 16161,
14
+ "eval_samples_per_second": 10.034,
15
+ "eval_steps_per_second": 0.314,
16
  "train_loss": 0.2915988182410216,
17
  "train_runtime": 1026.489,
18
  "train_samples": 4992,
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": 32016
28
  }
 
23
  "tie_word_embeddings": false,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.39.0.dev0",
26
+ "use_cache": true,
27
  "vocab_size": 32016
28
  }
eval_results.json ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "eval_logits/chosen": -2.6475114822387695,
4
+ "eval_logits/rejected": -3.464193344116211,
5
+ "eval_logps/chosen": -984.8101806640625,
6
+ "eval_logps/rejected": -1480.8428955078125,
7
+ "eval_loss": 0.4465055763721466,
8
+ "eval_rewards/accuracies": 0.7845849990844727,
9
+ "eval_rewards/chosen": -0.5929392576217651,
10
+ "eval_rewards/margins": 1.0842887163162231,
11
+ "eval_rewards/rejected": -1.6772277355194092,
12
+ "eval_runtime": 1610.6329,
13
+ "eval_samples": 16161,
14
+ "eval_samples_per_second": 10.034,
15
+ "eval_steps_per_second": 0.314
16
+ }