|
--- |
|
license: apache-2.0 |
|
base_model: hZzy/qwen2.5-0.5b-sft-news-IFT |
|
tags: |
|
- trl |
|
- expo |
|
- generated_from_trainer |
|
model-index: |
|
- name: qwen2.5-0.5b-expo-L2EXPO-ES-1 |
|
results: [] |
|
--- |
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You |
|
should probably proofread and complete it, then remove this comment. --> |
|
|
|
[<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/zhiyuzha-university-of-florida/huggingface/runs/72yqfbdr) |
|
# qwen2.5-0.5b-expo-L2EXPO-ES-1 |
|
|
|
This model is a fine-tuned version of [hZzy/qwen2.5-0.5b-sft-news-IFT](https://huggingface.co/hZzy/qwen2.5-0.5b-sft-news-IFT) on an unknown dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 5.0286 |
|
- Logps: -83.8820 |
|
- Logits: -0.4938 |
|
- Objective: 5.0013 |
|
- Dpo Loss: 2.6194 |
|
- Regularize: 5.0013 |
|
- Ranking Simple: 0.5197 |
|
- Ranking Idealized: 0.5295 |
|
- Ranking Idealized Expo: 0.5212 |
|
- Wo Beta: 14.2504 |
|
|
|
## 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 | |
|
|:-------------:|:------:|:----:|:---------------:|:--------:|:-------:|:---------:|:--------:|:----------:|:--------------:|:-----------------:|:----------------------:|:-------:| |
|
| 0.6418 | 0.1417 | 50 | 0.7369 | -89.5788 | -1.4384 | 0.7343 | 0.7480 | 0.7343 | 0.5248 | 0.5295 | 0.5212 | 16.0414 | |
|
| 1.7208 | 0.2834 | 100 | 1.7082 | -87.8064 | -1.3168 | 1.6950 | 1.0867 | 1.6950 | 0.5228 | 0.5295 | 0.5212 | 15.5148 | |
|
| 2.841 | 0.4251 | 150 | 2.9302 | -83.1791 | -1.1086 | 2.8768 | 1.6352 | 2.8768 | 0.5300 | 0.5295 | 0.5212 | 15.0680 | |
|
| 3.5072 | 0.5668 | 200 | 4.2317 | -80.2960 | -0.8688 | 4.2210 | 2.3120 | 4.2210 | 0.5155 | 0.5295 | 0.5212 | 14.5319 | |
|
| 3.7707 | 0.7085 | 250 | 4.3648 | -80.5389 | -0.7639 | 4.3627 | 2.2988 | 4.3627 | 0.5212 | 0.5295 | 0.5212 | 14.5663 | |
|
| 3.5773 | 0.8503 | 300 | 4.3904 | -83.8565 | -0.5388 | 4.3972 | 2.2955 | 4.3972 | 0.5238 | 0.5295 | 0.5212 | 14.3098 | |
|
| 3.359 | 0.9920 | 350 | 4.6868 | -82.1212 | -0.5555 | 4.6293 | 2.4176 | 4.6293 | 0.5264 | 0.5295 | 0.5212 | 14.3177 | |
|
| 3.0892 | 1.1337 | 400 | 4.8991 | -80.1851 | -0.4846 | 4.9208 | 2.5732 | 4.9208 | 0.5238 | 0.5295 | 0.5212 | 14.1271 | |
|
| 3.001 | 1.2754 | 450 | 4.8651 | -82.0773 | -0.5097 | 4.8038 | 2.4966 | 4.8038 | 0.5233 | 0.5295 | 0.5212 | 14.2309 | |
|
| 2.8358 | 1.4171 | 500 | 4.8734 | -81.9592 | -0.4937 | 4.8544 | 2.5685 | 4.8544 | 0.5243 | 0.5295 | 0.5212 | 14.2662 | |
|
| 2.6622 | 1.5588 | 550 | 4.8760 | -81.5020 | -0.5513 | 4.9098 | 2.5441 | 4.9098 | 0.5243 | 0.5295 | 0.5212 | 14.2522 | |
|
| 2.5417 | 1.7005 | 600 | 5.0324 | -83.9181 | -0.5043 | 5.0251 | 2.5863 | 5.0251 | 0.5259 | 0.5295 | 0.5212 | 14.2325 | |
|
| 2.435 | 1.8422 | 650 | 5.0286 | -83.8820 | -0.4938 | 5.0013 | 2.6194 | 5.0013 | 0.5197 | 0.5295 | 0.5212 | 14.2504 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.42.0 |
|
- Pytorch 2.3.0+cu121 |
|
- Datasets 2.19.1 |
|
- Tokenizers 0.19.1 |
|
|
