|
--- |
|
tags: |
|
- trl |
|
- dpo |
|
- generated_from_trainer |
|
model-index: |
|
- name: zephyr-7b-dpo-full |
|
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. --> |
|
|
|
# zephyr-7b-dpo-full |
|
|
|
This model was trained from scratch on the None dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.3183 |
|
- Rewards/chosen: -0.6032 |
|
- Rewards/rejected: -2.1160 |
|
- Rewards/accuracies: 0.8711 |
|
- Rewards/margins: 1.5128 |
|
- Logps/rejected: -584.2130 |
|
- Logps/chosen: -439.6992 |
|
- Logits/rejected: -5.8852 |
|
- Logits/chosen: -5.4031 |
|
|
|
## 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-07 |
|
- train_batch_size: 8 |
|
- eval_batch_size: 8 |
|
- seed: 42 |
|
- distributed_type: multi-GPU |
|
- num_devices: 8 |
|
- gradient_accumulation_steps: 2 |
|
- total_train_batch_size: 128 |
|
- total_eval_batch_size: 64 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: cosine |
|
- lr_scheduler_warmup_ratio: 0.1 |
|
- num_epochs: 1 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | |
|
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:| |
|
| 0.5118 | 0.1151 | 100 | 0.5923 | -0.1120 | -0.4506 | 0.7070 | 0.3386 | -417.6701 | -390.5766 | -2.1984 | -2.2213 | |
|
| 0.4206 | 0.2303 | 200 | 0.5055 | -0.2913 | -1.0785 | 0.8008 | 0.7872 | -480.4641 | -408.5089 | -3.2280 | -3.1644 | |
|
| 0.4144 | 0.3454 | 300 | 0.4504 | -0.3084 | -1.2736 | 0.7773 | 0.9651 | -499.9700 | -410.2218 | -4.0963 | -3.8861 | |
|
| 0.4011 | 0.4606 | 400 | 0.4135 | -0.4247 | -1.5332 | 0.8086 | 1.1086 | -525.9362 | -421.8441 | -4.8370 | -4.5018 | |
|
| 0.3915 | 0.5757 | 500 | 0.3740 | -0.3892 | -1.7143 | 0.8516 | 1.3251 | -544.0394 | -418.2938 | -5.1877 | -4.7675 | |
|
| 0.3726 | 0.6908 | 600 | 0.3468 | -0.4807 | -1.8892 | 0.8438 | 1.4085 | -561.5286 | -427.4439 | -5.6248 | -5.1461 | |
|
| 0.3522 | 0.8060 | 700 | 0.3249 | -0.5431 | -2.0476 | 0.8789 | 1.5044 | -577.3692 | -433.6906 | -5.6819 | -5.2107 | |
|
| 0.3643 | 0.9211 | 800 | 0.3183 | -0.6032 | -2.1160 | 0.8711 | 1.5128 | -584.2130 | -439.6992 | -5.8852 | -5.4031 | |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.41.1 |
|
- Pytorch 2.1.2+cu118 |
|
- Datasets 2.16.1 |
|
- Tokenizers 0.19.1 |
|
|