lewtun HF staff commited on
Commit
135dcd0
1 Parent(s): 1467c8f

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
@@ -2,9 +2,15 @@
2
  license: other
3
  base_model: lewtun/gemma-7b-sft-full-openhermes-v0
4
  tags:
 
5
  - trl
6
  - dpo
7
  - generated_from_trainer
 
 
 
 
 
8
  model-index:
9
  - name: gemma-7b-dpo-full-openhermes-mix1-beta-0.2
10
  results: []
@@ -15,7 +21,17 @@ should probably proofread and complete it, then remove this comment. -->
15
 
16
  # gemma-7b-dpo-full-openhermes-mix1-beta-0.2
17
 
18
- This model is a fine-tuned version of [lewtun/gemma-7b-sft-full-openhermes-v0](https://huggingface.co/lewtun/gemma-7b-sft-full-openhermes-v0) on the None dataset.
 
 
 
 
 
 
 
 
 
 
19
 
20
  ## Model description
21
 
 
2
  license: other
3
  base_model: lewtun/gemma-7b-sft-full-openhermes-v0
4
  tags:
5
+ - alignment-handbook
6
  - trl
7
  - dpo
8
  - generated_from_trainer
9
+ - trl
10
+ - dpo
11
+ - generated_from_trainer
12
+ datasets:
13
+ - argilla/dpo-mix-7k
14
  model-index:
15
  - name: gemma-7b-dpo-full-openhermes-mix1-beta-0.2
16
  results: []
 
21
 
22
  # gemma-7b-dpo-full-openhermes-mix1-beta-0.2
23
 
24
+ This model is a fine-tuned version of [lewtun/gemma-7b-sft-full-openhermes-v0](https://huggingface.co/lewtun/gemma-7b-sft-full-openhermes-v0) on the argilla/dpo-mix-7k dataset.
25
+ It achieves the following results on the evaluation set:
26
+ - Loss: 0.9486
27
+ - Rewards/chosen: 1.9255
28
+ - Rewards/rejected: -1.4494
29
+ - Rewards/accuracies: 0.7708
30
+ - Rewards/margins: 3.3749
31
+ - Logps/rejected: -814.3776
32
+ - Logps/chosen: -782.6418
33
+ - Logits/rejected: 101.8832
34
+ - Logits/chosen: 110.1703
35
 
36
  ## Model description
37
 
all_results.json CHANGED
@@ -1,5 +1,18 @@
1
  {
2
  "epoch": 0.99,
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  "train_loss": 1.1432483746455266,
4
  "train_runtime": 572.1827,
5
  "train_samples": 6750,
 
1
  {
2
  "epoch": 0.99,
3
+ "eval_logits/chosen": 110.1702880859375,
4
+ "eval_logits/rejected": 101.88317108154297,
5
+ "eval_logps/chosen": -782.641845703125,
6
+ "eval_logps/rejected": -814.3776245117188,
7
+ "eval_loss": 0.9485784769058228,
8
+ "eval_rewards/accuracies": 0.7708333134651184,
9
+ "eval_rewards/chosen": 1.925491213798523,
10
+ "eval_rewards/margins": 3.3749351501464844,
11
+ "eval_rewards/rejected": -1.4494438171386719,
12
+ "eval_runtime": 50.7482,
13
+ "eval_samples": 750,
14
+ "eval_samples_per_second": 14.779,
15
+ "eval_steps_per_second": 0.473,
16
  "train_loss": 1.1432483746455266,
17
  "train_runtime": 572.1827,
18
  "train_samples": 6750,
config.json CHANGED
@@ -23,6 +23,6 @@
23
  "rope_theta": 10000.0,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.39.0.dev0",
26
- "use_cache": false,
27
  "vocab_size": 256000
28
  }
 
23
  "rope_theta": 10000.0,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.39.0.dev0",
26
+ "use_cache": true,
27
  "vocab_size": 256000
28
  }
eval_results.json ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 0.99,
3
+ "eval_logits/chosen": 110.1702880859375,
4
+ "eval_logits/rejected": 101.88317108154297,
5
+ "eval_logps/chosen": -782.641845703125,
6
+ "eval_logps/rejected": -814.3776245117188,
7
+ "eval_loss": 0.9485784769058228,
8
+ "eval_rewards/accuracies": 0.7708333134651184,
9
+ "eval_rewards/chosen": 1.925491213798523,
10
+ "eval_rewards/margins": 3.3749351501464844,
11
+ "eval_rewards/rejected": -1.4494438171386719,
12
+ "eval_runtime": 50.7482,
13
+ "eval_samples": 750,
14
+ "eval_samples_per_second": 14.779,
15
+ "eval_steps_per_second": 0.473
16
+ }