QLoRA+百万数据对baichun-7b模型进行高效指令微调
更多详情请查看Github项目: Firefly(流萤): 中文对话式大语言模型(全量微调+QLoRA)
单轮对话脚本:
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
model_name = 'YeungNLP/firefly-baichuan-7b-qlora-sft-merge'
max_new_tokens = 500
top_p = 0.9
temperature = 0.35
repetition_penalty = 1.0
device = 'cuda'
input_pattern = '<s>{}</s>'
model = AutoModelForCausalLM.from_pretrained(
model_name,
trust_remote_code=True,
low_cpu_mem_usage=True,
torch_dtype=torch.float16,
device_map='auto'
)
model.eval()
model = model.to(device)
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
text = input('User：')
while True:
text = input_pattern.format(text)
input_ids = tokenizer(text, return_tensors="pt").input_ids
input_ids = input_ids.to(device)
outputs = model.generate(
input_ids=input_ids, max_new_tokens=max_new_tokens, do_sample=True,
top_p=top_p, temperature=temperature, repetition_penalty=repetition_penalty,
eos_token_id=tokenizer.eos_token_id
)
rets = tokenizer.batch_decode(outputs)
output = rets[0].strip().replace(text, "").replace('</s>', "")
print("Firefly：{}".format(output))
text = input('User：')
多轮对话脚本:
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
device = 'cuda'
model_name = 'YeungNLP/firefly-baichuan-7b1-qlora-sft-merge'
max_new_tokens = 500
top_p = 0.9
temperature = 0.35
repetition_penalty = 1.0
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
model_name,
trust_remote_code=True,
low_cpu_mem_usage=True,
torch_dtype=torch.float16,
device_map='auto'
)
model.eval()
model = model.to(device)
# 记录所有历史记录
history_token_ids = tokenizer('<s>', return_tensors="pt").input_ids
# 输入模型的最大长度
history_max_len = 1000
user_input = input('User：')
while True:
user_input = '{}</s>'.format(user_input)
user_input_ids = tokenizer(user_input, return_tensors="pt").input_ids
history_token_ids = torch.concat((history_token_ids, user_input_ids), dim=1)
model_input_ids = history_token_ids[:, -history_max_len:].to(device)
outputs = model.generate(
input_ids=model_input_ids, max_new_tokens=max_new_tokens, do_sample=True, top_p=top_p,
temperature=temperature, repetition_penalty=repetition_penalty, eos_token_id=tokenizer.eos_token_id
)
model_input_ids_len = model_input_ids.size(1)
response_ids = outputs[:, model_input_ids_len:]
history_token_ids = torch.concat((history_token_ids, response_ids.cpu()), dim=1)
response = tokenizer.batch_decode(response_ids)
print("Firefly：" + response[0].strip().replace('</s>', ""))
user_input = input('User：')