跳到主要内容

Cerebras

https://inference-docs.cerebras.ai/api-reference/chat-completions

提示

我们支持所有 Cerebras 模型，只需在发送 litellm 请求时将 model=cerebras/<any-model-on-cerebras> 设置为前缀即可。

API 密钥​

# env variable
os.environ['CEREBRAS_API_KEY']

使用示例​

from litellm import completion
import os

os.environ['CEREBRAS_API_KEY'] = ""
response = completion(
    model="cerebras/llama3-70b-instruct",
    messages=[
        {
            "role": "user",
            "content": "What's the weather like in Boston today in Fahrenheit? (Write in JSON)",
        }
    ],
    max_tokens=10,
        
    # The prompt should include JSON if 'json_object' is selected; otherwise, you will get error code 400.
    response_format={ "type": "json_object" },
    seed=123,
    stop=["\n\n"],
    temperature=0.2,
    top_p=0.9,
    tool_choice="auto",
    tools=[],
    user="user",
)
print(response)

使用示例 - 流式​

from litellm import completion
import os

os.environ['CEREBRAS_API_KEY'] = ""
response = completion(
    model="cerebras/llama3-70b-instruct",
    messages=[
        {
            "role": "user",
            "content": "What's the weather like in Boston today in Fahrenheit? (Write in JSON)",
        }
    ],
    stream=True,
    max_tokens=10,

    # The prompt should include JSON if 'json_object' is selected; otherwise, you will get error code 400.
    response_format={ "type": "json_object" }, 
    seed=123,
    stop=["\n\n"],
    temperature=0.2,
    top_p=0.9,
    tool_choice="auto",
    tools=[],
    user="user",
)

for chunk in response:
    print(chunk)

与 LiteLLM 代理服务器一起使用​

以下是如何使用 LiteLLM 代理服务器调用 Cerebras 模型

  1. 修改 config.yaml

    model_list:
      - model_name: my-model
        litellm_params:
          model: cerebras/<your-model-name>  # add cerebras/ prefix to route as Cerebras provider
          api_key: api-key                 # api key to send your model

  1. 启动代理

    $ litellm --config /path/to/config.yaml

  2. 发送请求到 LiteLLM 代理服务器

    import openai
    client = openai.OpenAI(
        api_key="sk-1234",             # pass litellm proxy key, if you're using virtual keys
        base_url="http://0.0.0.0:4000" # litellm-proxy-base url
    )

    response = client.chat.completions.create(
        model="my-model",
        messages = [
            {
                "role": "user",
                "content": "what llm are you"
            }
        ],
    )

    print(response)