---
license: openrail
language:
- zh
pipeline_tag: text-generation
library_name: transformers
---


## Original model card 

Buy me a coffee if you like this project ;)
<a href="https://www.buymeacoffee.com/s3nh"><img src="https://www.buymeacoffee.com/assets/img/guidelines/download-assets-sm-1.svg" alt=""></a>

#### Description 

GGML Format model files for [This project](https://huggingface.co/hiyouga/baichuan-13b-sft/tree/main).


### inference 


```python

import ctransformers

from ctransformers import AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained(output_dir, ggml_file,
gpu_layers=32, model_type="llama")

manual_input: str = "Tell me about your last dream, please."


llm(manual_input, 
      max_new_tokens=256, 
      temperature=0.9, 
      top_p= 0.7)

```

# Original model card

A bilingual instruction-tuned LoRA model of https://huggingface.co/baichuan-inc/Baichuan-13B-Base

- Instruction-following datasets used: alpaca-en, alpaca-zh, sharegpt, open assistant, lima, refgpt
- Training framework: https://github.com/hiyouga/LLaMA-Efficient-Tuning

Usage:

```python
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer

tokenizer = AutoTokenizer.from_pretrained("hiyouga/baichuan-13b-sft", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("hiyouga/baichuan-13b-sft", trust_remote_code=True).cuda()
streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)

query = "晚上睡不着怎么办"
template = (
    "A chat between a curious user and an artificial intelligence assistant. "
    "The assistant gives helpful, detailed, and polite answers to the user's questions.\n"
    "Human: {}\nAssistant: "
)

inputs = tokenizer([template.format(query)], return_tensors="pt")
inputs = inputs.to("cuda")
generate_ids = model.generate(**inputs, max_new_tokens=256, streamer=streamer)
```

You could also alternatively launch a CLI demo by using the script in https://github.com/hiyouga/LLaMA-Efficient-Tuning

```bash
python src/cli_demo.py --template default --model_name_or_path hiyouga/baichuan-13b-sft
```

---

You can reproduce our results by visiting the following step-by-step (Chinese) guide:

https://zhuanlan.zhihu.com/p/645010851

or using the following scripts in [LLaMA-Efficient-Tuning](https://github.com/hiyouga/LLaMA-Efficient-Tuning):

```bash
CUDA_VISIBLE_DEVICES=0 python src/train_bash.py \
    --stage sft \
    --model_name_or_path baichuan-inc/Baichuan-13B-Base \
    --do_train \
    --dataset alpaca_gpt4_en,alpaca_gpt4_zh,self_cognition,oaast_sft,sharegpt_zh,refgpt_zh_p1,lima,self_cognition \
    --template default \
    --finetuning_type lora \
    --lora_rank 32 \
    --lora_target W_pack,o_proj,down_proj,up_proj,gate_proj \
    --output_dir baichuan_13b_lora \
    --per_device_train_batch_size 4 \
    --gradient_accumulation_steps 8 \
    --preprocessing_num_workers 16 \
    --max_source_length 1024 \
    --max_target_length 512 \
    --optim paged_adamw_32bit \
    --lr_scheduler_type cosine \
    --logging_steps 10 \
    --save_steps 100 \
    --eval_steps 100 \
    --warmup_steps 100 \
    --learning_rate 3e-5 \
    --max_grad_norm 0.5 \
    --num_train_epochs 2.0 \
    --quantization_bit 4 \
    --plot_loss \
    --fp16
```

Loss curve:

![loss](loss.svg)

---

Examples:

| Question | Baichuan-13b-sft | ChatGPT |
| -------- | ---------------- | ------- |
| 你是谁 | 我是 \<NAME\>，一款由 \<AUTHOR\> 开发的人工智能助手。我可以回答各种问题，提供帮助和支持。 | 我是ChatGPT，一个由OpenAI开发的人工智能语言模型。我基于GPT-3.5架构训练，旨在回答各种问题和提供帮助。有什么我可以帮助你的吗？ |
| 中国第三代战斗机的代表是什么？ | 中