运行Phi-4-mini-instruct模型遇RuntimeError：张量维度2与3不匹配如何解决？

问题描述 #

在Google Colab T4 Notebook中严格复制Phi-4-mini-instruct官方示例代码执行，出现以下错误：

RuntimeError: Tensors must have same number of dimensions: got 2 and 3

同时伴随警告：do_sample设为False但temperature设为0.0，该参数仅在采样生成模式下生效。

报错原因 #

1. transformers版本不兼容：Phi-4系列模型依赖较新的transformers库版本，Colab默认安装的版本可能无法正确处理模型的对话格式输入，导致张量维度计算出错。
2. 生成参数冲突：do_sample=False（贪心搜索模式）下设置temperature=0.0属于无效配置，虽然是警告，但可能触发后续生成逻辑中的维度异常。

解决方案 #

方案1：升级依赖库并修正参数 #

首先在Colab中执行以下命令升级依赖：

!pip install --upgrade transformers accelerate torch

然后修改生成参数，移除temperature=0.0（因为do_sample=False时该参数无意义）：

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline

torch.random.manual_seed(0)

model_path = "microsoft/Phi-4-mini-instruct"

model = AutoModelForCausalLM.from_pretrained(
    model_path,
    device_map="auto",
    torch_dtype="auto",
    trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained(model_path)

messages = [
    {"role": "system", "content": "You are a helpful AI assistant."},
    {"role": "user", "content": "Can you provide ways to eat combinations of bananas and dragonfruits?"},
    {"role": "assistant", "content": "Sure! Here are some ways to eat bananas and dragonfruits together: 1. Banana and dragonfruit smoothie: Blend bananas and dragonfruits together with some milk and honey. 2. Banana and dragonfruit salad: Mix sliced bananas and dragonfruits together with some lemon juice and honey."},
    {"role": "user", "content": "What about solving an 2x + 3 = 7 equation?"},
]

pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
)

generation_args = {
    "max_new_tokens": 500,
    "return_full_text": False,
    "do_sample": False,  # 贪心搜索模式，无需temperature
}

output = pipe(messages, **generation_args)
print(output[0]['generated_text'])

方案2：手动构建对话Prompt（避免pipeline自动处理的兼容性问题） #

如果升级依赖后仍有问题，可以手动使用tokenizer的对话模板构建输入文本：

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

torch.random.manual_seed(0)

model_path = "microsoft/Phi-4-mini-instruct"

model = AutoModelForCausalLM.from_pretrained(
    model_path,
    device_map="auto",
    torch_dtype="auto",
    trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained(model_path)

messages = [
    {"role": "system", "content": "You are a helpful AI assistant."},
    {"role": "user", "content": "Can you provide ways to eat combinations of bananas and dragonfruits?"},
    {"role": "assistant", "content": "Sure! Here are some ways to eat bananas and dragonfruits together: 1. Banana and dragonfruit smoothie: Blend bananas and dragonfruits together with some milk and honey. 2. Banana and dragonfruit salad: Mix sliced bananas and dragonfruits together with some lemon juice and honey."},
    {"role": "user", "content": "What about solving an 2x + 3 = 7 equation?"},
]

# 手动应用对话模板
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

generation_args = {
    "max_new_tokens": 500,
    "return_full_text": False,
    "do_sample": False,
}

outputs = model.generate(**inputs, **generation_args)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

内容的提问来源于stack exchange，提问作者최영진