| | ---
|
| | license: apache-2.0
|
| | datasets:
|
| | - liyucheng/zhihu_rlhf_3k
|
| | language:
|
| | - zh
|
| | base_model:
|
| | - lucky2me/Dorami-Instruct
|
| | ---
|
| |
|
| | # Dorami-Chat
|
| |
|
| | Dorami-Chat is a Direct Preference Optimization(DPO) model based on the Supervised Fine-tuning(SFT) model lucky2me/Dorami-Instruct
|
| |
|
| | ## Model description
|
| |
|
| | ### Training data
|
| |
|
| | - [liyucheng/zhihu_rlhf_3k](https://huggingface.co/datasets/liyucheng/zhihu_rlhf_3k)
|
| |
|
| | ### Training code
|
| |
|
| | - [dorami](https://github.com/6zeus/dorami.git)
|
| |
|
| | ## How to use
|
| |
|
| | ### 1. Download model from Hugging Face Hub to local
|
| |
|
| | ```
|
| | git lfs install
|
| | git clone https://huggingface.co/lucky2me/Dorami-Chat
|
| | ```
|
| |
|
| | ### 2. Use the model downloaded above
|
| | ```python
|
| | from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig
|
| | model_path = "The path of the model downloaded above"
|
| | tokenizer = AutoTokenizer.from_pretrained(model_path)
|
| | model = AutoModelForCausalLM.from_pretrained(model_path)
|
| | prompt="fill in any prompt you like."
|
| | inputs = tokenizer(prompt, return_tensors="pt")
|
| | generation_config = GenerationConfig(max_new_tokens=64, do_sample=True, top_k=2, eos_token_id=model.config.eos_token_id)
|
| | outputs = model.generate(**inputs, generation_config=generation_config)
|
| | decoded_text = tokenizer.batch_decode(outputs, skip_special_tokens=True)
|
| | print(decoded_text)
|
| | ``` |
