llama-factory微调工具使用入门_llama-factory中断后能继续吗

一、定义

1. 环境配置
2. 案例： https://zhuanlan.zhihu.com/p/695287607
3. chatglm3 案例
4. 多卡训练deepspeed
5. llama factory 案例Qwen1.5
6. 报错

二、实现

1. 环境配置

git clone https://github.com/hiyouga/LLaMA-Factory.git
conda create -n llama_factory python=3.10
conda activate llama_factory
cd LLaMA-Factory
pip install -e '.[torch,metrics]'
如果发生冲突：    pip install --no-deps -e .  
1
2
3
4
5
6

同时对本库的基础安装做一下校验，输入以下命令获取训练相关的参数指导, 否则说明库还没有安装成功

llamafactory-cli train -h
1

模型下载与可用性校对

git clone https://www.modelscope.cn/LLM-Research/Meta-Llama-3-8B-Instruct.git
1

import transformers
import torch

# 切换为你下载的模型文件目录, 这里的demo是Llama-3-8B-Instruct
# 如果是其他模型，比如qwen，chatglm，请使用其对应的官方demo
model_id = "/home/Meta-Llama-3-8B-Instruct"
 
pipeline = transformers.pipeline(
    "text-generation",
    model=model_id,
    model_kwargs={"torch_dtype": torch.bfloat16},
    device_map="auto",
)

messages = [
    {"role": "system", "content": "You are a pirate chatbot who always responds in pirate speak!"},
    {"role": "user", "content": "Who are you?"},
]

prompt = pipeline.tokenizer.apply_chat_template(
        messages,
        tokenize=False,
        add_generation_prompt=True
)

terminators = [
    pipeline.tokenizer.eos_token_id,
    pipeline.tokenizer.convert_tokens_to_ids("<|eot_id|>")
]

outputs = pipeline(
    prompt,
    max_new_tokens=256,
    eos_token_id=terminators,
    do_sample=True,
    temperature=0.6,
    top_p=0.9,
)
print(outputs[0]["generated_text"][len(prompt):])
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39

2. 案例： https://zhuanlan.zhihu.com/p/695287607

2.1 数据准备
将该自定义数据集放到我们的系统中使用，则需要进行如下两步操作
a 复制该数据集到 data目录下
b 修改 data/dataset_info.json 新加内容完成注册, 该注册同时完成了3件事
b1 自定义数据集的名称为adgen_local，后续训练的时候就使用这个名称来找到该数据集
b2 指定了数据集具体文件位置
b3 定义了原数据集的输入输出和我们所需要的格式之间的映射关系

2. 微调：
下载模型
>> git clone https://www.modelscope.cn/LLM-Research/Meta-Llama-3-8B-Instruct.git
微调

CUDA_VISIBLE_DEVICES=0 llamafactory-cli train \
    --stage sft \
    --do_train \
    --model_name_or_path /home/Meta-Llama-3-8B-Instruct \
    --dataset alpaca_gpt4_zh,identity,adgen_local \
    --dataset_dir ./data \
    --template llama3 \
    --finetuning_type lora \
    --output_dir ./saves/LLaMA3-8B/lora/sft \
    --overwrite_cache \
    --overwrite_output_dir \
    --cutoff_len 1024 \
    --preprocessing_num_workers 16 \
    --per_device_train_batch_size 2 \
    --per_device_eval_batch_size 1 \
    --gradient_accumulation_steps 8 \
    --lr_scheduler_type cosine \
    --logging_steps 50 \
    --warmup_steps 20 \
    --save_steps 100 \
    --eval_steps 50 \
    --evaluation_strategy steps \
    --load_best_model_at_end \
    --learning_rate 5e-5 \
    --num_train_epochs 5.0 \
    --max_samples 1000 \
    --val_size 0.1 \
    --plot_loss \
    --fp16
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

或者：

CUDA_VISIBLE_DEVICES=0 llamafactory-cli train ./examples/train_lora/llama3_lora_sft.yaml
1

3. 推理

CUDA_VISIBLE_DEVICES=0 llamafactory-cli chat ./examples/inferce/llama3_lora_sft.yaml
1

或

CUDA_VISIBLE_DEVICES=0 llamafactory-cli chat \
    --model_name_or_path /home/Meta-Llama-3-8B-Instruct \
    --adapter_name_or_path ./saves/LLaMA3-8B/lora/sft  \
    --template llama3 \
    --finetuning_type lora
1
2
3
4
5

4. 批量预测与训练效果评估

CUDA_VISIBLE_DEVICES=0 llamafactory-cli chat ./examples/train/llama3_lora_predict.yaml
1

或

CUDA_VISIBLE_DEVICES=0 llamafactory-cli train \
    --stage sft \
    --do_predict \
    --model_name_or_path /media/codingma/LLM/llama3/Meta-Llama-3-8B-Instruct \
    --adapter_name_or_path ./saves/LLaMA3-8B/lora/sft  \
    --dataset alpaca_gpt4_zh,identity,adgen_local \
    --dataset_dir ./data \
    --template llama3 \
    --finetuning_type lora \
    --output_dir ./saves/LLaMA3-8B/lora/predict \
    --overwrite_cache \
    --overwrite_output_dir \
    --cutoff_len 1024 \
    --preprocessing_num_workers 16 \
    --per_device_eval_batch_size 1 \
    --max_samples 20 \
    --predict_with_generate
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

5. LoRA模型合并导出

CUDA_VISIBLE_DEVICES=0 llamafactory-cli export \
    --model_name_or_path /home/Meta-Llama-3-8B-Instruct \
    --adapter_name_or_path ./saves/LLaMA3-8B/lora/sft  \
    --template llama3 \
    --finetuning_type lora \
    --export_dir megred-model-path \
    --export_size 2 \
    --export_device cpu \
    --export_legacy_format False
1
2
3
4
5
6
7
8
9

CUDA_VISIBLE_DEVICES=0 llamafactory-cli export ./examples/merge_lora/llama3_lora_sft.yaml
1

6. api 调用

CUDA_VISIBLE_DEVICES=0 API_PORT=8000 nohup llamafactory-cli api \
    --model_name_or_path /media/codingma/LLM/llama3/Meta-Llama-3-8B-Instruct \
    --adapter_name_or_path ./saves/LLaMA3-8B/lora/sft \
    --template llama3 \
    --finetuning_type lora
1
2
3
4
5

项目也支持了基于vllm 的推理后端，但是这里由于一些限制，需要提前将LoRA 模型进行merge，使用merge后的完整版模型目录或者训练前的模型原始目录都可。

CUDA_VISIBLE_DEVICES=0 API_PORT=8000 nohup llamafactory-cli api \
    --model_name_or_path megred-model-path \
    --template llama3 \
    --infer_backend vllm \
    --vllm_enforce_eager>output.log 2>&1 &
1
2
3
4
5

import os
from openai import OpenAI
from transformers.utils.versions import require_version

require_version("openai>=1.5.0", "To fix: pip install openai>=1.5.0")

if __name__ == '__main__':
    # change to your custom port
    port = 8000
    client = OpenAI(
        api_key="0",
        base_url="http://localhost:{}/v1".format(os.environ.get("API_PORT", 8000)),
    )
    messages = []
    messages.append({"role": "user", "content": "hello, where is USA"})
    result = client.chat.completions.create(messages=messages, model="test")
    print(result.choices[0].message)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

7. 测试

CUDA_VISIBLE_DEVICES=0 llamafactory-cli eval ./examples/train/llama3_lora_eval.yaml
1

或

CUDA_VISIBLE_DEVICES=0 llamafactory-cli eval \
--model_name_or_path /media/codingma/LLM/llama3/Meta-Llama-3-8B-Instruct \
--template llama3 \
--task mmlu \
--split validation \
--lang en \
--n_shot 5 \
--batch_size 1
1
2
3
4
5
6
7
8

3. chatglm3 案例
   见专题模块

4. 多卡训练deepspeed
   多卡看llama3_lora_sft_ds0.yaml

5. 报错

   1，RuntimeError: Failed to import trl.trainer.dpo_trainer because of the following error (look up to see its traceback):
   ‘FieldInfo’ object has no attribute ‘required’
   解决：换干净的环境，重新安装。