metadata
license: apache-2.0
base_model: hZzy/qwen2.5-0.5b-sft-news-IFT
tags:
- alignment-handbook
- ndcg
- trl
- expo
- generated_from_trainer
- trl
- expo
- generated_from_trainer
datasets:
- hZzy/train_pairwise
model-index:
- name: qwen2.5-0.5b-expo-DPO-ES-1
results: []
qwen2.5-0.5b-expo-DPO-ES-1
This model is a fine-tuned version of hZzy/qwen2.5-0.5b-sft-news-IFT on the hZzy/train_pairwise dataset. It achieves the following results on the evaluation set:
- Loss: 2.3243
- Logps: -83.2882
- Logits: -0.6651
- Objective: 2.2471
- Dpo Loss: 2.2471
- Regularize: 2.2471
- Ranking Simple: 0.5378
- Ranking Idealized: 0.5295
- Ranking Idealized Expo: 0.5212
- Wo Beta: 6.6815
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-06
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- distributed_type: multi-GPU
- num_devices: 3
- gradient_accumulation_steps: 12
- total_train_batch_size: 144
- total_eval_batch_size: 12
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 5
Training results
| Training Loss | Epoch | Step | Validation Loss | Logps | Logits | Objective | Dpo Loss | Regularize | Ranking Simple | Ranking Idealized | Ranking Idealized Expo | Wo Beta |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.7017 | 0.1417 | 50 | 0.8470 | -93.0243 | -1.4582 | 0.8570 | 0.8570 | 0.8570 | 0.5238 | 0.5295 | 0.5212 | 7.8507 |
| 0.8112 | 0.2834 | 100 | 1.0529 | -86.6835 | -1.4382 | 1.0273 | 1.0273 | 1.0273 | 0.5285 | 0.5295 | 0.5212 | 7.4982 |
| 1.0895 | 0.4251 | 150 | 1.4497 | -84.4337 | -1.2965 | 1.4010 | 1.4010 | 1.4010 | 0.5321 | 0.5295 | 0.5212 | 7.2692 |
| 1.2363 | 0.5668 | 200 | 1.7035 | -77.7201 | -1.2956 | 1.6116 | 1.6116 | 1.6116 | 0.5321 | 0.5295 | 0.5212 | 7.2264 |
| 1.3152 | 0.7085 | 250 | 1.9222 | -92.7241 | -1.2565 | 1.8319 | 1.8319 | 1.8319 | 0.5311 | 0.5295 | 0.5212 | 7.1856 |
| 1.1899 | 0.8503 | 300 | 2.0298 | -90.9373 | -0.9785 | 1.9588 | 1.9588 | 1.9588 | 0.5367 | 0.5295 | 0.5212 | 6.9336 |
| 1.1443 | 0.9920 | 350 | 2.1654 | -82.1414 | -1.0214 | 2.0541 | 2.0541 | 2.0541 | 0.5435 | 0.5295 | 0.5212 | 7.0024 |
| 0.725 | 1.1337 | 400 | 2.2884 | -84.2526 | -0.7535 | 2.2360 | 2.2360 | 2.2360 | 0.5336 | 0.5295 | 0.5212 | 7.1525 |
| 0.7629 | 1.2754 | 450 | 2.1606 | -80.4165 | -0.8866 | 2.0671 | 2.0671 | 2.0671 | 0.5321 | 0.5295 | 0.5212 | 6.7949 |
| 0.8044 | 1.4171 | 500 | 2.2094 | -82.3927 | -0.7503 | 2.0981 | 2.0981 | 2.0981 | 0.5347 | 0.5295 | 0.5212 | 6.8050 |
| 0.7105 | 1.5588 | 550 | 2.1697 | -84.9780 | -0.6734 | 2.0733 | 2.0733 | 2.0733 | 0.5321 | 0.5295 | 0.5212 | 6.8722 |
| 0.6925 | 1.7005 | 600 | 2.1957 | -81.5342 | -0.7411 | 2.0558 | 2.0558 | 2.0558 | 0.5357 | 0.5295 | 0.5212 | 6.7186 |
| 0.6883 | 1.8422 | 650 | 2.2080 | -82.7303 | -0.6908 | 2.1330 | 2.1330 | 2.1330 | 0.5383 | 0.5295 | 0.5212 | 6.8081 |
| 0.6486 | 1.9839 | 700 | 2.3243 | -83.2882 | -0.6651 | 2.2471 | 2.2471 | 2.2471 | 0.5378 | 0.5295 | 0.5212 | 6.6815 |
| 0.3793 | 2.1256 | 750 | 2.2675 | -84.2296 | -0.7879 | 2.1825 | 2.1825 | 2.1825 | 0.5409 | 0.5295 | 0.5212 | 6.8794 |
| 0.3314 | 2.2674 | 800 | 2.2106 | -84.3675 | -0.6651 | 2.1041 | 2.1041 | 2.1041 | 0.5414 | 0.5295 | 0.5212 | 6.7463 |
| 0.3301 | 2.4091 | 850 | 2.2964 | -84.8913 | -0.6177 | 2.2221 | 2.2221 | 2.2221 | 0.5388 | 0.5295 | 0.5212 | 6.8020 |
| 0.3509 | 2.5508 | 900 | 2.2796 | -84.3833 | -0.6097 | 2.2099 | 2.2099 | 2.2099 | 0.5393 | 0.5295 | 0.5212 | 6.7934 |
| 0.321 | 2.6925 | 950 | 2.3403 | -83.2967 | -0.7158 | 2.2649 | 2.2649 | 2.2649 | 0.5331 | 0.5295 | 0.5212 | 6.8864 |
Framework versions
- Transformers 4.42.0
- Pytorch 2.3.0+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1