from transformers import pipeline

# 1. 情感分析
classifier = pipeline("sentiment-analysis")
result = classifier("我非常喜欢这个产品！")
# 输出: [{'label': 'POSITIVE', 'score': 0.9998}]

# 2. 文本生成
generator = pipeline("text-generation")
text = generator("人工智能正在改变世界", max_length=50)
# 输出: 生成的续写文本

# 3. 翻译（中译英）
translator = pipeline("translation", model="Helsinki-NLP/opus-mt-zh-en")
result = translator("大语言模型正在改变我们与计算机的交互方式。")
# 输出: [{'translation_text': 'Large language models are changing the way we interact with computers.'}]

# 4. 命名实体识别
ner = pipeline("ner", grouped_entities=True)
result = ner("Hugging Face公司总部位于纽约布鲁克林区。")
# 输出: [{'entity_group': 'ORG', 'word': 'Hugging Face'}, {'entity_group': 'LOC', 'word': '纽约布鲁克林区'}]

from transformers import AutoTokenizer, AutoModelForSequenceClassification

# 1. 加载模型和分词器
model_name = "bert-base-chinese"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

# 2. 文本预处理
text = "这是一个测试文本"
inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True)

# 3. 模型推理
outputs = model(**inputs)
predictions = outputs.logits.argmax(-1)

from transformers import BitsAndBytesConfig

# 4位量化，大幅减少显存占用
model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-2-7b-hf",
    load_in_4bit=True
)

# 自动分配到GPU
model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-2-7b-hf",
    device_map="auto",
    torch_dtype="auto"
)

from transformers import pipeline

# 创建翻译pipeline
zh_to_en = pipeline("translation", model="Helsinki-NLP/opus-mt-zh-en")
en_to_zh = pipeline("translation", model="Helsinki-NLP/opus-mt-en-zh")

# 中译英
chinese_text = "大语言模型正在改变我们与计算机的交互方式。"
english = zh_to_en(chinese_text)
print(f"英文: {english[0]['translation_text']}")

# 英译中
english_text = "Large language models are revolutionizing natural language processing."
chinese = en_to_zh(english_text)
print(f"中文: {chinese[0]['translation_text']}")

from transformers import AutoModel

# 自动识别并加载任意架构的模型
model = AutoModel.from_pretrained("bert-base-chinese")  # BERT
model = AutoModel.from_pretrained("gpt2")               # GPT-2
model = AutoModel.from_pretrained("t5-small")           # T5

pip install huggingface_hub requests

import requests

API_URL = "https://api-inference.huggingface.co/models/gpt2"
headers = {"Authorization": "Bearer YOUR_HF_TOKEN"}

response = requests.post(
    API_URL,
    headers=headers,
    json={"inputs": "Hello, I am a language model,"}
)
print(response.json())

from huggingface_hub import InferenceClient

client = InferenceClient(token="YOUR_HF_TOKEN")

# 文本生成
response = client.text_generation(
    prompt="解释什么是 Transformer：",
    model="google/gemma-2-2b-it",
    max_new_tokens=50
)

# 对话补全
chat = client.chat_completion(
    messages=[{"role": "user", "content": "什么是注意力机制？"}],
    model="meta-llama/Meta-Llama-3-8B-Instruct",
    max_tokens=100
)
print(chat.choices[0].message.content)

class SmartDocAssistant:
    """智能文档助手"""
    
    def classify_document(self, text):
        """零样本分类 - 无需训练即可识别文档类型"""
        # 使用 BART-large-MNLI
        # candidate_labels 定义候选类别
        result = self.api.query(
            "facebook/bart-large-mnli",
            text,
            {"candidate_labels": ["技术文档", "新闻", "商业报告"]}
        )
        return result["labels"][0], result["scores"][0]
    
    def extract_keywords(self, text):
        """命名实体识别 - 提取人名/地名/机构名"""
        # 使用 BERT-NER
        result = self.api.query("dslim/bert-base-NER", text)
        return [{"word": e["word"], "type": e["entity_group"]} for e in result]
    
    def generate_summary(self, text):
        """摘要生成"""
        # 使用 BART-CNN
        result = self.api.query("facebook/bart-large-cnn", text)
        return result[0]["summary_text"]
    
    def translate(self, text, direction="zh_en"):
        """中英互译"""
        model = "Helsinki-NLP/opus-mt-zh-en" if direction == "zh_en" else "Helsinki-NLP/opus-mt-en-zh"
        result = self.api.query(model, text)
        return result[0]["translation_text"]

def process_document(text, translate_to=None):
    """
    一次性执行所有分析任务
    """
    # 1. 分类
    doc_type = classify_document(text)
    
    # 2. 关键词
    keywords = extract_keywords(text)
    
    # 3. 摘要
    summary = generate_summary(text)
    
    # 4. 翻译（可选）
    translation = translate(text) if translate_to else None
    
    return {
        "doc_type": doc_type,
        "keywords": keywords,
        "summary": summary,
        "translation": translation
    }

HF_TOKEN = "hf_xxxxxxxxxxxxx"  # 从 https://huggingface.co/settings/tokens 获取

# 演示模式（使用内置示例）
python smart_doc_assistant.py

# 交互模式（自定义输入）
python smart_doc_assistant.py --interactive