Message='BaiChuanTokenizer' object has no attribute 'sp_model'

[toniedeng]

Message='BaiChuanTokenizer' object has no attribute 'sp_model'
Source=C:\Users\Administrator.cache\huggingface\modules\transformers_modules\Sunsimiao\tokenization_baichuan.py
StackTrace:
File "C:\Users\Administrator.cache\huggingface\modules\transformers_modules\Sunsimiao\tokenization_baichuan.py", line 104, in vocab_size
return self.sp_model.get_piece_size()
File "C:\Users\Administrator.cache\huggingface\modules\transformers_modules\Sunsimiao\tokenization_baichuan.py", line 108, in get_vocab (Current frame)
vocab = {self.convert_ids_to_tokens(i): i for i in range(self.vocab_size)}
File "C:\Users\Administrator.cache\huggingface\modules\transformers_modules\Sunsimiao\tokenization_baichuan.py", line 74, in init
super().init(
File "C:\Users\Administrator.cache\modelscope\modelscope_modules\Sunsimiao\ms_wrapper.py", line 41, in init
self.tokenizer = AutoTokenizer.from_pretrained(model_dir, trust_remote_code=True)
File "C:\Users\Administrator.cache\modelscope\modelscope_modules\Sunsimiao\ms_wrapper.py", line 20, in init
model = SunsimiaoTextGeneration(model) if isinstance(model, str) else model
File "C:\Users\Administrator\source\repos\Sunsimiao\scripts\inference_ms.py", line 4, in
pipe = pipeline(task=Tasks.text_generation,

how can i deal?

[toniedeng]

deal with change transformers==4.33.1

[jingnant]

尝试降级transformers==4.33.3 或者修改tokenization_baichuan.py，super() 修改到最后执行

       self.vocab_file = vocab_file
       self.add_bos_token = add_bos_token
       self.add_eos_token = add_eos_token
       self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
       self.sp_model.Load(vocab_file)
       super().__init__(
           bos_token=bos_token,
           eos_token=eos_token,
           unk_token=unk_token,
           pad_token=pad_token,
           add_bos_token=add_bos_token,
           add_eos_token=add_eos_token,
           sp_model_kwargs=self.sp_model_kwargs,
           clean_up_tokenization_spaces=clean_up_tokenization_spaces,
           **kwargs,
       )
       # self.vocab_file = vocab_file
       # self.add_bos_token = add_bos_token
       # self.add_eos_token = add_eos_token
       # self.sp_model = spm.SentencePieceProcessor(**self.sp_model_kwargs)
       # self.sp_model.Load(vocab_file)

[Robtzz]

我仍然抛出这个异常

[Robtzz]

我仍然抛出这个异常

在魔塔降低版本之后需要重启免费的服务器内核，来保证生效，我重启后可以运行

[jingnant]

尝试新模型Sunsimiao-7B的推理代码：

from modelscope import AutoModelForCausalLM, AutoTokenizer
device = "cuda" # the device to load the model onto

model = AutoModelForCausalLM.from_pretrained(
    "X-D-Lab/Sunsimiao-Qwen2-7B",
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("qwen/Qwen2-1.5B-Instruct")

prompt = "Give me a short introduction to large language model."
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(device)

generated_ids = model.generate(
    model_inputs.input_ids,
    max_new_tokens=512
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)

[Robtzz]

尝试新模型Sunsimiao-7B的推理代码：

from modelscope import AutoModelForCausalLM, AutoTokenizer
device = "cuda" # the device to load the model onto

model = AutoModelForCausalLM.from_pretrained(
    "X-D-Lab/Sunsimiao-Qwen2-7B",
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("qwen/Qwen2-1.5B-Instruct")

prompt = "Give me a short introduction to large language model."
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(device)

generated_ids = model.generate(
    model_inputs.input_ids,
    max_new_tokens=512
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)

新的代码运行不但没有错误，推理的速度也快了很多，但是我看不出他和平常的LLM的区别，类似 GPT-4o

[Robtzz]

尝试新模型Sunsimiao-7B的推理代码：

from modelscope import AutoModelForCausalLM, AutoTokenizer
device = "cuda" # the device to load the model onto

model = AutoModelForCausalLM.from_pretrained(
    "X-D-Lab/Sunsimiao-Qwen2-7B",
    torch_dtype="auto",
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("qwen/Qwen2-1.5B-Instruct")

prompt = "Give me a short introduction to large language model."
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(device)

generated_ids = model.generate(
    model_inputs.input_ids,
    max_new_tokens=512
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)

新的代码运行不但没有错误，推理的速度也快了很多，但是我看不出他和平常的LLM的区别，类似 GPT-4o

甚至GPT-4o的返回更加令人容易理解和接受