Lora微调Qwen-1_8B-Chat模型实践（魔搭社区）

一、准备工作

1，克隆qwen仓库代码：

git clone https://github.com/QwenLM/Qwen.git

2，跳转至Qwen项目的目录下：

cd Qwen

3，安装依赖包：

pip install deepspeed transformers==4.32.0 peft pydantic==1.10.13 transformers_stream_generator einops tiktoken

4，训练数据准备：

样本放入列表存为json文件，命名为chat.json（可以先下载再上传）

5，下载模型文件：

git clone https://modelscope.cn/qwen/Qwen-1_8B-Chat.git

二、单机单卡微调

1，export CUDA_DEVICE_MAX_CONNECTIONS=1
      export CUDA_VISIBLE_DEVICES=0

2，执行微调python文件：

python finetune.py --model_name_or_path Qwen-1_8B-Chat --data_path chat.json --fp16 True --output_dir output_qwen --num_train_epochs 5 --per_device_train_batch_size 2 --per_device_eval_batch_size 1 --gradient_accumulation_steps 8 --evaluation_strategy "no" --save_strategy "steps" --save_steps 1000 --save_total_limit 10 --learning_rate 3e-4 --weight_decay 0.1 --adam_beta2 0.95 --warmup_ratio 0.01 --lr_scheduler_type "cosine" --logging_steps 1 --report_to "none" --model_max_length 512 --lazy_preprocess True --gradient_checkpointing --use_lora

 此处容易出现的问题-

（1）相关依赖包不匹配：根据错误提示更新即可

1‘No module named 'transformers.models.gemma

解决：更新transformers库pip install transformers --upgrade

2’Your currently installed version of Keras is Keras 3, but this is not yet supported in Transformers. Please install the backwards-compatible tf-keras package with `pip install tf-keras`.

解决：pip install tf-keras

（2）fp16精度问题：True改为False

三、保存微调后模型并测试

1，保存：

from peft import AutoPeftModelForCausalLM 
from transformers import AutoTokenizer 
model = AutoPeftModelForCausalLM.from_pretrained( "output_qwen", device_map="auto", trust_remote_code=True ).eval() 
merged_model = model.merge_and_unload() 
merged_model.save_pretrained("qwen-1_8b-finetune", max_shard_size="2048MB", safe_serialization=True) # 最大分片2g
tokenizer = AutoTokenizer.from_pretrained( "output_qwen", trust_remote_code=True ) 
 
tokenizer.save_pretrained("qwen-1_8b-finetune")

结果展示：

2，测试：

from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig 
tokenizer = AutoTokenizer.from_pretrained("qwen-1_8b-finetune", trust_remote_code=True)
 
model = AutoModelForCausalLM.from_pretrained("qwen-1_8b-finetune", device_map="auto", 
trust_remote_code=True).eval() 
response, history = model.chat(tokenizer, "xxxx", history=None) 
print(response)

结果展示：