在Codelab对llama3做Lora Fine tune微调

Unsloth 高效微调大模型的工具，通过Unsloth微调Llama3, Mistral, Gemma 速度提升2-5倍，内存减少70%！

Codelab 创建一个jupyter notebook

选择 T4 GPU

安装Fine tune 相关的lib

%%capture
import torch
major_version, minor_version= torch.cuda.get_device_capability()
# Must install separately since Colab has torch 2.2.1, which breaks packages
!pip install "unsloth[colab-new] @ git+https://github.com/unslothai/unsloth.git"
if major_version >= 8:# Use this for new GPs like Ampere, Hopper GPUs(RTX 30xx. RIX 40xx, A100. H100. L40)!pip install -no-deps packaging ninja einops flash-attn xformers trl peft accelerate bitsandbytes
else:# Use this for older GPUs (V100, Tesla T4, RTX 20xx)!pip install --no-deps xformers trl peft accelerate bitsandbytes
pass

下载llama3

from unsloth import FastLanguageModel
import torch
max_seq_length = 2048 # Choose any! We auto support RoPE Scaling internally!
dtype = None # None for auto detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+
load_in_4bit = True # Use 4bit quantization to reduce memory usage. Can be False# 4bit pre quantized models we support for 4x faster downloading + no OOMs
fourbit_models = ["unsloth/mistral-7b-bnb-4bit","unsloth/mistral-7b-instruct-bnb-4bit","unsloth/llama-2-7b-bnb-4bit","unsloth/gemma-7b-bnb-4bit","unsloth/gemma-7b-it-bnb-4bit","unsloth/gemma-2b-bnb-4bit","unsloth/gemma-2b-it-bnb-4bit","unsloth/llama-3-8b-bnb-4bit",
] # More models at https://huggingface.co/unslothmodel, tokenizer = FastLanguageModel.from_pretrained(model_name = "unsloth/llama-3-8b-bnb-4bit",max_seq_length = max_seq_length,dtype = dtype,load_in_4bit = load_in_4bit# token = "hf_...", # use one if using gated models like meta-llama/Llama-2-7b-hf)

model = FastLanguageModel.get_peft_model(model,r = 16, # Choose any number > 0 ! Suggested 8, 16, 32, 64, 128target_modules = ["q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj",],lora_alpha = 16,lora_dropout = 0, # Supports any, but = 0 is optimizedbias = "none", # Supports any, but = "none" is optimized# [NEW] "unsloth" uses 30% less VRAM, fits 2x larger batch sizes!use_gradient_checkpointing = "unsloth", # True or "unsloth" for very long contextrandom_state = 3407,use_rslora = False, # We support rank stabilized LoRAloftq_config = None # And LoftQ
)

加载hugging face数据集

alpaca_prompt = """Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
### Instruction:
{}
### Input:
{}
### Response:
{}
"""EOS_TOKEN = tokenizer.eos_token # Must add EOS_TOKEN
def formatting_prompts_func(examples):instructions = examples["instruction"]inputs = examples["input"]outputs = examples["output"]texts = []for instruction, input, output in zip(instructions, inputs, outputs):# Must add EOS_TOKEN, otherwise your generation will go on forever!text = alpaca_prompt.format(instruction, input, output) + EOS_TOKENtexts.append(text)return { "text": texts, }
passfrom datasets import load_dataset
dataset = load_dataset("pinzhenchen/alpaca-cleaned-zh", split="train")
dataset = dataset.map(formatting_prompts_func, batched=True,)

HuggingFace 官网， 点击数据集 Datasets

搜索数据集 alpaca-cleaned-zh

复制数据集的名字 pinzhenchen/alpaca-cleaned-zh

定义training 方法

from trl import SFTTrainer
from transformers import TrainingArgumentstrainer = SFTTrainer(model = model,tokenizer = tokenizer,train_dataset = dataset,dataset_text_field = "text",max_seq_length = max_seq_length,dataset_num_proc = 2,packing = False, # Can make training 5x faster for short sequences.args = TrainingArguments(per_device_train_batch_size = 2,gradient_accumulation_steps = 4,warmup_steps = 5,max_steps = 60,learning_rate = 2e-4,fp16 = not torch.cuda.is_bf16_supported(),bf16 = torch.cuda.is_bf16_supported(),logging_steps = 1,optim = "adamw_8bit",weight_decay = 0.01,lr_scheduler_type = "linear",seed = 3407,output_dir = "outputs",),
)

打印显存使用情况

#@title Show current memory stats
gpu_stats = torch.cuda.get_device_properties(0)
start_gpu_memory = round(torch.cuda.max_memory_reserved() / 1024 / 1024 / 1024, 3)
max_memory = round(gpu_stats.total_memory / 1024 / 1024 / 1024, 3)
print(f"GPU = (gpu_stats.name). Max memory = (max_memory) GB.")
print(f"(start_gpu_memory) GB of memory reserved.")

开始FineTune

trainer_stats = trainer.train()

#@title Show final memory and time stats
used_memory = round(torch.cuda.max_memory_reserved() / 1024 / 1024 / 1024, 3)
used_memory_for_lora = round(used_memory - start_gpu_memory, 3)
used_percentage = round(used_memory / max_memory*100, 3)
lora_percentage = round(used_memory_for_lora / max_memory*100, 3)
print(f"{trainer_stats.metrics['train_runtime']} seconds used for training.")
print(f"{round(trainer_stats.metrics['train_runtime']/60, 2)} minutes used for training.")
print(f"Peak reserved memory = {used_memory} GB.")
print(f"Peak reserved memory for training = {used_memory_for_lora} GB.")
print(f"Peak reserved memory % of max memory = {used_percentage} %.")
print(f"Peak reserved memory for training % of max memory = {lora_percentage} GB.")

用fineTune 过的model，做问答

# alpaca_prompt = Copied from above
FastLanguageModel.for_inference(model) # Enable native 2x faster inference
inputs = tokenizer([alpaca_prompt.format("如何保持健康", # instruction"", # input"", # output - leave this blank for generation!)], return_tensors = "pt"
).to("cuda")outputs = model.generate(**inputs, max_new_tokens = 64, use_cache=True)
tokenizer.batch_decode(outputs)

TextStreamer 流式一个字一个字地打印结果

# alpaca_prompt = Copied from above
FastLanguageModel.for_inference(model) # Enable native 2x faster inference
inputs = tokenizer([alpaca_prompt.format("续写这段话", # instruction"天天向上，好好学习", # input"", # output - leave this blank for generation!)], return_tensors = "pt"
).to("cuda")from transformers import TextStreamer
text_streamer = TextStreamer(tokenizer)
_ = model.generate(**inputs, streamer = text_streamer, max_new_tokens=128)

保存model到google drive 和 HuggingFace

model.save_pretrained("lora_model") # local saving
model.push_to_hub("zgpeace/lora_model", token="####") # online saving

google drive