|
--- |
|
license: mit |
|
library_name: peft |
|
tags: |
|
- trl |
|
- kto |
|
- KTO |
|
- WeniGPT |
|
- generated_from_trainer |
|
base_model: HuggingFaceH4/zephyr-7b-beta |
|
model-index: |
|
- name: kto-test |
|
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. --> |
|
|
|
# kto-test |
|
|
|
This model is a fine-tuned version of [HuggingFaceH4/zephyr-7b-beta](https://huggingface.co/HuggingFaceH4/zephyr-7b-beta) on the None dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 0.0147 |
|
- Rewards/chosen: 5.6143 |
|
- Rewards/rejected: -31.0540 |
|
- Rewards/margins: 36.6683 |
|
- Kl: 0.0 |
|
- Logps/chosen: -130.3461 |
|
- Logps/rejected: -503.4655 |
|
|
|
## 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: 0.0002 |
|
- train_batch_size: 2 |
|
- eval_batch_size: 2 |
|
- seed: 42 |
|
- distributed_type: multi-GPU |
|
- num_devices: 2 |
|
- gradient_accumulation_steps: 8 |
|
- total_train_batch_size: 32 |
|
- total_eval_batch_size: 4 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: linear |
|
- lr_scheduler_warmup_ratio: 0.03 |
|
- training_steps: 786 |
|
- mixed_precision_training: Native AMP |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/margins | Kl | Logps/chosen | Logps/rejected | |
|
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:---------------:|:------:|:------------:|:--------------:| |
|
| 371.0873 | 0.38 | 50 | 0.0440 | 4.6808 | -9.5189 | 14.1997 | 0.0150 | -139.6814 | -288.1148 | |
|
| 57.9834 | 0.76 | 100 | 0.0275 | 5.1394 | -31.8945 | 37.0339 | 0.0 | -135.0947 | -511.8704 | |
|
| 37.3685 | 1.14 | 150 | 0.0196 | 5.2556 | -27.1934 | 32.4491 | 0.0 | -133.9325 | -464.8599 | |
|
| 3.6561 | 1.52 | 200 | 0.0162 | 5.4306 | -22.6310 | 28.0615 | 0.0 | -132.1833 | -419.2354 | |
|
| 59.5367 | 1.9 | 250 | 0.0143 | 5.7355 | -31.1619 | 36.8974 | 0.0 | -129.1339 | -504.5448 | |
|
| 13.1891 | 2.29 | 300 | 0.0147 | 5.6143 | -31.0540 | 36.6683 | 0.0 | -130.3461 | -503.4655 | |
|
| 3.8532 | 2.67 | 350 | 0.0131 | 5.8860 | -26.4154 | 32.3014 | 0.0 | -127.6289 | -457.0801 | |
|
| 3.7678 | 3.05 | 400 | 0.0162 | 5.9318 | -26.7524 | 32.6841 | 0.0 | -127.1711 | -460.4493 | |
|
| 49.3456 | 3.43 | 450 | 0.0167 | 5.9252 | -28.7033 | 34.6286 | 0.0 | -127.2365 | -479.9590 | |
|
| 12.2886 | 3.81 | 500 | 0.0164 | 6.0009 | -29.4493 | 35.4501 | 0.0 | -126.4803 | -487.4185 | |
|
| 2.3745 | 4.19 | 550 | 0.0173 | 6.0124 | -29.9808 | 35.9932 | 0.0 | -126.3649 | -492.7338 | |
|
| 0.46 | 4.57 | 600 | 0.0173 | 6.0060 | -30.4606 | 36.4666 | 0.0 | -126.4293 | -497.5318 | |
|
| 7.7723 | 4.95 | 650 | 0.0180 | 6.0079 | -30.7030 | 36.7109 | 0.0 | -126.4096 | -499.9554 | |
|
| 4.1333 | 5.33 | 700 | 0.0184 | 6.0037 | -30.8948 | 36.8984 | 0.0 | -126.4521 | -501.8734 | |
|
| 1.6938 | 5.71 | 750 | 0.0183 | 6.0119 | -30.9672 | 36.9791 | 0.0 | -126.3704 | -502.5979 | |
|
|
|
|
|
### Framework versions |
|
|
|
- PEFT 0.10.0 |
|
- Transformers 4.39.1 |
|
- Pytorch 2.1.0+cu118 |
|
- Datasets 2.18.0 |
|
- Tokenizers 0.15.1 |
