jbjeong91 commited on
Commit
9d983d9
1 Parent(s): b7e74d3

End of training

Browse files
README.md CHANGED
@@ -3,9 +3,15 @@ library_name: transformers
3
  license: llama3.1
4
  base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
5
  tags:
 
6
  - trl
7
  - cpo
8
  - generated_from_trainer
 
 
 
 
 
9
  model-index:
10
  - name: llama3.1-cpo_j-full-0912
11
  results: []
@@ -16,18 +22,18 @@ should probably proofread and complete it, then remove this comment. -->
16
 
17
  # llama3.1-cpo_j-full-0912
18
 
19
- This model is a fine-tuned version of [meta-llama/Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct) on an unknown dataset.
20
  It achieves the following results on the evaluation set:
21
- - Loss: 1.4429
22
- - Rewards/chosen: -16.1577
23
- - Rewards/rejected: -16.9120
24
- - Rewards/accuracies: 0.6304
25
- - Rewards/margins: 0.7544
26
- - Logps/rejected: -169.1204
27
- - Logps/chosen: -161.5765
28
- - Logits/rejected: -0.3509
29
- - Logits/chosen: -0.3812
30
- - Nll Loss: 0.2845
31
 
32
  ## Model description
33
 
 
3
  license: llama3.1
4
  base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
5
  tags:
6
+ - alignment-handbook
7
  - trl
8
  - cpo
9
  - generated_from_trainer
10
+ - trl
11
+ - cpo
12
+ - generated_from_trainer
13
+ datasets:
14
+ - princeton-nlp/llama3-ultrafeedback
15
  model-index:
16
  - name: llama3.1-cpo_j-full-0912
17
  results: []
 
22
 
23
  # llama3.1-cpo_j-full-0912
24
 
25
+ This model is a fine-tuned version of [meta-llama/Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct) on the princeton-nlp/llama3-ultrafeedback dataset.
26
  It achieves the following results on the evaluation set:
27
+ - Loss: 1.4395
28
+ - Rewards/chosen: -16.1609
29
+ - Rewards/rejected: -16.9344
30
+ - Rewards/accuracies: 0.6326
31
+ - Rewards/margins: 0.7735
32
+ - Logps/rejected: -169.3439
33
+ - Logps/chosen: -161.6093
34
+ - Logits/rejected: -0.3578
35
+ - Logits/chosen: -0.3883
36
+ - Nll Loss: 0.2841
37
 
38
  ## Model description
39
 
all_results.json CHANGED
@@ -1,5 +1,19 @@
1
  {
2
  "epoch": 0.9985553308292401,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  "total_flos": 0.0,
4
  "train_loss": 1.6653077276768509,
5
  "train_runtime": 9934.0195,
 
1
  {
2
  "epoch": 0.9985553308292401,
3
+ "eval_logits/chosen": -0.3882608115673065,
4
+ "eval_logits/rejected": -0.3577967882156372,
5
+ "eval_logps/chosen": -161.60928344726562,
6
+ "eval_logps/rejected": -169.34393310546875,
7
+ "eval_loss": 1.4395084381103516,
8
+ "eval_nll_loss": 0.28411635756492615,
9
+ "eval_rewards/accuracies": 0.6326087117195129,
10
+ "eval_rewards/chosen": -16.160930633544922,
11
+ "eval_rewards/margins": 0.773465096950531,
12
+ "eval_rewards/rejected": -16.93439483642578,
13
+ "eval_runtime": 76.7453,
14
+ "eval_samples": 1826,
15
+ "eval_samples_per_second": 23.793,
16
+ "eval_steps_per_second": 1.498,
17
  "total_flos": 0.0,
18
  "train_loss": 1.6653077276768509,
19
  "train_runtime": 9934.0195,
config.json CHANGED
@@ -34,6 +34,6 @@
34
  "tie_word_embeddings": false,
35
  "torch_dtype": "bfloat16",
36
  "transformers_version": "4.44.2",
37
- "use_cache": false,
38
  "vocab_size": 128256
39
  }
 
34
  "tie_word_embeddings": false,
35
  "torch_dtype": "bfloat16",
36
  "transformers_version": "4.44.2",
37
+ "use_cache": true,
38
  "vocab_size": 128256
39
  }
eval_results.json ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 0.9985553308292401,
3
+ "eval_logits/chosen": -0.3882608115673065,
4
+ "eval_logits/rejected": -0.3577967882156372,
5
+ "eval_logps/chosen": -161.60928344726562,
6
+ "eval_logps/rejected": -169.34393310546875,
7
+ "eval_loss": 1.4395084381103516,
8
+ "eval_nll_loss": 0.28411635756492615,
9
+ "eval_rewards/accuracies": 0.6326087117195129,
10
+ "eval_rewards/chosen": -16.160930633544922,
11
+ "eval_rewards/margins": 0.773465096950531,
12
+ "eval_rewards/rejected": -16.93439483642578,
13
+ "eval_runtime": 76.7453,
14
+ "eval_samples": 1826,
15
+ "eval_samples_per_second": 23.793,
16
+ "eval_steps_per_second": 1.498
17
+ }
runs/Sep12_20-48-52_jjb_prism_dev2/events.out.tfevents.1726152785.jjb_prism_dev2.76248.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:345d4080688067c750928ef1cd4c34643384ae6017626d4ae5791eba670373ae
3
+ size 880