Python调用AI实战指南

一、环境准备

1. 基础工具链

```bash

pip install pandas numpy matplotlib scikit-learn

```

2. AI专用库

二、调用云端AI API

1. OpenAI GPT系列

```python

from openai import OpenAI

client = OpenAI(api_key="YOUR_KEY")

response = client.pletions.create(

model="gpt-4",

messages=[{"role": "user", "content": "解释量子力学"}]

```

2. 百度文心一言

```python

import requests

url = "

headers = {"Content-Type": "application/json"}

data = {"messages": [{"role": "user", "content": "写一首诗"}]}

response = requests.post(url, headers=headers, json=data)

```

3. KimiGPT

```python

from openai import OpenAI

client = OpenAI(

api_key="YOUR_KEY",

base_url="

response = client.pletions.create(

model="moonshot-v1-128k",

messages=[{"role": "user", "content": "生成Python代码"}]

```

三、本地模型调用

1. Hugging Face Transformers

```python

from transformers import pipeline

generator = pipeline("text-generation", model="gpt2")

print(generator("AI的未来是"))

```

2. 自定义PyTorch模型

```python

import torch

model = torch.load('model.pth')

inputs = torch.tensor([[1, 174, 33, 900, 1]]) 示例输入

outputs = model(inputs)

```

四、实战场景

1. 智能数据清洗

```python

from sklearn.impute import KNNImputer

import pandas as pd

df = pd.read_excel("data.xlsx")

imputer = KNNImputer(n_neighbors=3)

cleaned = imputer.fit_transform(df)

```

2. 图像分类（30分钟入门）

```python

import tensorflow as tf

model = tf.keras.Sequential([

tf.keras.layers.Conv2D(32, (3,3), activation='relu'),

tf.keras.layers.MaxPooling2D((2,2))

])

```

五、注意事项

本地模型部署要考虑硬件加速（CUDA/MPS）