---
license: apache-2.0
base_model: mistralai/Mistral-7B-v0.1
tags:
- alignment-handbook
- trl
- orpo
- generated_from_trainer
datasets:
- argilla/dpo-mix-7k
model-index:
- name: mistral-orpo-mix-7k
  results: []
language:
- en
---
# mistral-orpo-mix-7k

This model is a ORPO full fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the argilla/dpo-mix-7k dataset with the [huggingface/alignment-handbook](https://github.com/huggingface/alignment-handbook).

## Training procedure

Trained for 4.5 hours on 1xA100

### Aligment Handbook recipe

```yaml
# Model arguments
model_name_or_path: mistralai/Mistral-7B-v0.1
model_revision: main
torch_dtype: bfloat16
use_flash_attention_2: true
trust_remote_code: true

# Data training arguments
chat_template: "{% for message in messages %}\n{% if message['role'] == 'user' %}\n{{ '<|user|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'system' %}\n{{ '<|system|>\n' + message['content'] + eos_token }}\n{% elif message['role'] == 'assistant' %}\n{{ '<|assistant|>\n'  + message['content'] + eos_token }}\n{% endif %}\n{% if loop.last and add_generation_prompt %}\n{{ '<|assistant|>' }}\n{% endif %}\n{% endfor %}"
dataset_mixer:
  argilla/dpo-mix-7k: 1.0
dataset_splits:
- train
- test
preprocessing_num_workers: 8

# ORPOTrainer arguments
bf16: true
beta: 0.05
gradient_accumulation_steps: 8
gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: true
hub_model_id: mistral-orpo-mix-7k
hub_private_repo: true
learning_rate: 5.0e-6
log_level: info
logging_steps: 10
lr_scheduler_type: inverse_sqrt
max_length: 2048
max_prompt_length: 1792
num_train_epochs: 3
optim: adamw_bnb_8bit
output_dir: data/mistral-orpo-mix-7k
per_device_train_batch_size: 4
push_to_hub: true
report_to:
- tensorboard
- wandb
save_strategy: "no"
seed: 42
warmup_steps: 100
```

### Framework versions

- Transformers 4.41.0.dev0
- Pytorch 2.1.2
- Datasets 2.19.0
- Tokenizers 0.19.1