|
--- |
|
license: apache-2.0 |
|
base_model: TheBloke/Mistral-7B-Instruct-v0.2-GPTQ |
|
tags: |
|
- trl |
|
- sft |
|
- generated_from_trainer |
|
model-index: |
|
- name: v3 |
|
results: [] |
|
datasets: |
|
- hieunguyenminh/roleplay |
|
--- |
|
|
|
<!-- 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. --> |
|
|
|
# v3 |
|
|
|
This model is a fine-tuned version of [TheBloke/Mistral-7B-Instruct-v0.2-GPTQ](https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.2-GPTQ) on the [hieunguyenminh/roleplay](hieunguyenminh/roleplay) dataset. |
|
|
|
## Model description |
|
|
|
This model can adapt to any type of characters and provide answer that personalize that character. |
|
|
|
## Training and evaluation data |
|
|
|
It is trained with supervised learning and will be trained with DPO in the future. |
|
|
|
## Training procedure |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 0.0002 |
|
- train_batch_size: 16 |
|
- eval_batch_size: 8 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: cosine |
|
- training_steps: 400 |
|
- mixed_precision_training: Native AMP |
|
|
|
### Training results |
|
|
|
Loss after 1 epochs: 0.6 |
|
|
|
### Framework versions |
|
|
|
- Transformers 4.35.2 |
|
- Pytorch 2.1.0+cu121 |
|
- Datasets 2.16.1 |
|
- Tokenizers 0.15.0 |
