深度探索 DeepSeek 微调：LoRA 与全参数微调实战指南

pip install torch transformers peft accelerate

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "deepseek-ai/deepseek-mistral-7b"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

from peft import LoraConfig, get_peft_model

# 配置 LoRA 训练参数
lora_config = LoraConfig(
    r=8,  # 低秩矩阵的秩
    lora_alpha=32,  # LoRA 缩放因子
    lora_dropout=0.1,  # dropout 率
    bias="none",
    target_modules=["q_proj", "v_proj"],  # 仅对部分层进行微调
)

# 应用 LoRA
model = get_peft_model(model, lora_config)
model.print_trainable_parameters()

from transformers import Trainer, TrainingArguments

training_args = TrainingArguments(
    output_dir="./lora_model",
    per_device_train_batch_size=4,
    num_train_epochs=3,
    save_steps=100,
    logging_dir="./logs",
)

trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=my_train_dataset,  # 替换为你的数据集
)
trainer.train()

pip install deepspeed transformers torch

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "deepseek-ai/deepseek-mistral-7b"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

from transformers import TrainingArguments

training_args = TrainingArguments(
    output_dir="./full_finetune",
    per_device_train_batch_size=2,
    num_train_epochs=3,
    save_strategy="epoch",
    report_to="tensorboard",
    logging_dir="./logs",
    deepspeed="./ds_config.json"  # DeepSpeed 加速
)

from transformers import Trainer

trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=my_train_dataset,  # 替换为你的数据集
)
trainer.train()

from peft import PeftModel

# 加载微调后的模型
fine_tuned_model = PeftModel.from_pretrained(model, "./lora_model")
fine_tuned_model.eval()

input_text = "DeepSeek 在 NLP 领域的应用有哪些？"
inputs = tokenizer(input_text, return_tensors="pt")

output = fine_tuned_model.generate(**inputs)
print(tokenizer.decode(output[0], skip_special_tokens=True))

{
  "zero_optimization": {
    "stage": 2,
    "offload_optimizer": "cpu",
    "offload_param": "none"
  }
}

training_args = TrainingArguments(deepspeed="./ds_config.json")