用openai库 请求时，流式请求时缺stream_options={"include_usage": True}的处理，用于计算流式tokens

[sasicDHH]

Reminder

• I have read the README and searched the existing issues.

Reproduction

response = client.chat.completions.create(
        model=model_name,
        messages=[
            # {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": msg_clean},
        ],
        temperature=0.8,
        top_p=0.8,
        max_tokens=2048,
        stream=True,
        stream_options={"include_usage": True},
    )

参考
建议在llamafactory/api/chat.py 第144行左右create_stream_chat_completion_response 函数下
_create_stream_chat_completion_chunk 对stream_options 处理

Expected behavior

计算steam=True时，通过stream_options={"include_usage": True}, 获取流式tokens
详细请参考

System Info

No response

Others

No response