End of training
Browse files
README.md
CHANGED
@@ -2,9 +2,15 @@
|
|
2 |
license: other
|
3 |
base_model: deepseek-ai/deepseek-math-7b-base
|
4 |
tags:
|
|
|
5 |
- trl
|
6 |
- orpo
|
7 |
- generated_from_trainer
|
|
|
|
|
|
|
|
|
|
|
8 |
model-index:
|
9 |
- name: deepseek-8b-orpo-full
|
10 |
results: []
|
@@ -16,7 +22,20 @@ should probably proofread and complete it, then remove this comment. -->
|
|
16 |
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/thuzfz1/huggingface/runs/dtbyb3v7)
|
17 |
# deepseek-8b-orpo-full
|
18 |
|
19 |
-
This model is a fine-tuned version of [deepseek-ai/deepseek-math-7b-base](https://huggingface.co/deepseek-ai/deepseek-math-7b-base) on
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
|
21 |
## Model description
|
22 |
|
|
|
2 |
license: other
|
3 |
base_model: deepseek-ai/deepseek-math-7b-base
|
4 |
tags:
|
5 |
+
- alignment-handbook
|
6 |
- trl
|
7 |
- orpo
|
8 |
- generated_from_trainer
|
9 |
+
- trl
|
10 |
+
- orpo
|
11 |
+
- generated_from_trainer
|
12 |
+
datasets:
|
13 |
+
- zfz1/my_preference_gsm8k_deepseek
|
14 |
model-index:
|
15 |
- name: deepseek-8b-orpo-full
|
16 |
results: []
|
|
|
22 |
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/thuzfz1/huggingface/runs/dtbyb3v7)
|
23 |
# deepseek-8b-orpo-full
|
24 |
|
25 |
+
This model is a fine-tuned version of [deepseek-ai/deepseek-math-7b-base](https://huggingface.co/deepseek-ai/deepseek-math-7b-base) on the zfz1/my_preference_gsm8k_deepseek dataset.
|
26 |
+
It achieves the following results on the evaluation set:
|
27 |
+
- Loss: 0.0797
|
28 |
+
- Rewards/chosen: -0.0089
|
29 |
+
- Rewards/rejected: -0.2196
|
30 |
+
- Rewards/accuracies: 1.0
|
31 |
+
- Rewards/margins: 0.2107
|
32 |
+
- Logps/rejected: -2.1960
|
33 |
+
- Logps/chosen: -0.0885
|
34 |
+
- Logits/rejected: 13.8906
|
35 |
+
- Logits/chosen: 8.5947
|
36 |
+
- Nll Loss: 0.0786
|
37 |
+
- Log Odds Ratio: -0.0204
|
38 |
+
- Log Odds Chosen: 5.3533
|
39 |
|
40 |
## Model description
|
41 |
|
all_results.json
CHANGED
@@ -1,5 +1,21 @@
|
|
1 |
{
|
2 |
"epoch": 2.0,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
"total_flos": 0.0,
|
4 |
"train_loss": 0.2130171504020691,
|
5 |
"train_runtime": 11716.0982,
|
|
|
1 |
{
|
2 |
"epoch": 2.0,
|
3 |
+
"eval_log_odds_chosen": 5.353264808654785,
|
4 |
+
"eval_log_odds_ratio": -0.020422911271452904,
|
5 |
+
"eval_logits/chosen": 8.594698905944824,
|
6 |
+
"eval_logits/rejected": 13.890603065490723,
|
7 |
+
"eval_logps/chosen": -0.08852054923772812,
|
8 |
+
"eval_logps/rejected": -2.195981502532959,
|
9 |
+
"eval_loss": 0.07974232733249664,
|
10 |
+
"eval_nll_loss": 0.07859200984239578,
|
11 |
+
"eval_rewards/accuracies": 1.0,
|
12 |
+
"eval_rewards/chosen": -0.008852055296301842,
|
13 |
+
"eval_rewards/margins": 0.21074610948562622,
|
14 |
+
"eval_rewards/rejected": -0.2195981740951538,
|
15 |
+
"eval_runtime": 336.9846,
|
16 |
+
"eval_samples": 5398,
|
17 |
+
"eval_samples_per_second": 16.019,
|
18 |
+
"eval_steps_per_second": 1.003,
|
19 |
"total_flos": 0.0,
|
20 |
"train_loss": 0.2130171504020691,
|
21 |
"train_runtime": 11716.0982,
|
config.json
CHANGED
@@ -24,6 +24,6 @@
|
|
24 |
"tie_word_embeddings": false,
|
25 |
"torch_dtype": "bfloat16",
|
26 |
"transformers_version": "4.42.3",
|
27 |
-
"use_cache":
|
28 |
"vocab_size": 102400
|
29 |
}
|
|
|
24 |
"tie_word_embeddings": false,
|
25 |
"torch_dtype": "bfloat16",
|
26 |
"transformers_version": "4.42.3",
|
27 |
+
"use_cache": true,
|
28 |
"vocab_size": 102400
|
29 |
}
|
eval_results.json
ADDED
@@ -0,0 +1,19 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 2.0,
|
3 |
+
"eval_log_odds_chosen": 5.353264808654785,
|
4 |
+
"eval_log_odds_ratio": -0.020422911271452904,
|
5 |
+
"eval_logits/chosen": 8.594698905944824,
|
6 |
+
"eval_logits/rejected": 13.890603065490723,
|
7 |
+
"eval_logps/chosen": -0.08852054923772812,
|
8 |
+
"eval_logps/rejected": -2.195981502532959,
|
9 |
+
"eval_loss": 0.07974232733249664,
|
10 |
+
"eval_nll_loss": 0.07859200984239578,
|
11 |
+
"eval_rewards/accuracies": 1.0,
|
12 |
+
"eval_rewards/chosen": -0.008852055296301842,
|
13 |
+
"eval_rewards/margins": 0.21074610948562622,
|
14 |
+
"eval_rewards/rejected": -0.2195981740951538,
|
15 |
+
"eval_runtime": 336.9846,
|
16 |
+
"eval_samples": 5398,
|
17 |
+
"eval_samples_per_second": 16.019,
|
18 |
+
"eval_steps_per_second": 1.003
|
19 |
+
}
|
runs/Jul15_19-44-09_phe108-jieyuzhao-01/events.out.tfevents.1721110827.phe108-jieyuzhao-01.232196.1
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bf97d2bac05783325fe82fbf6438b7fb3ee4844a180c37697700a19087006a60
|
3 |
+
size 997
|