Predibase

Predibase

LiteLLM에서 Predibase의 모든 모델을 사용하는 방법을 알아봐요.

출처: 문서

본문

LiteLLM은 Predibase의 모든 모델을 지원해요.

사용법

API 키

import os
os.environ["PREDIBASE_API_KEY"] = ""

호출 예시

from litellm import completion
import os
## set ENV variables
os.environ["PREDIBASE_API_KEY"] = "predibase key"
os.environ["PREDIBASE_TENANT_ID"] = "predibase tenant id"

# predibase llama-3 call
response = completion(
    model="predibase/llama-3-8b-instruct",
    messages = [{ "content": "Hello, how are you?","role": "user"}]
)

config.yaml에 모델 추가:

model_list:
  - model_name: llama-3
    litellm_params:
      model: predibase/llama-3-8b-instruct
      api_key: os.environ/PREDIBASE_API_KEY
      tenant_id: os.environ/PREDIBASE_TENANT_ID

Proxy 시작:

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

OpenAI Python SDK로 요청:

import openai
client = openai.OpenAI(
    api_key="sk-",             # 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="llama-3",
    messages = [
      {
          "role": "system",
          "content": "Be a good human!"
      },
      {
          "role": "user",
          "content": "What do you know about earth?"
      }
  ]
)

print(response)

curl --location 'http://0.0.0.0:4000/chat/completions' \
    --header "Authorization: Bearer ***" \
    --header 'Content-Type: application/json' \
    --data '{
    "model": "llama-3",
    "messages": [
      {
          "role": "system",
          "content": "Be a good human!"
      },
      {
          "role": "user",
          "content": "What do you know about earth?"
      }
      ],
}'

고급 사용법 - 프롬프트 포맷팅

LiteLLM은 모든 meta-llama llama3 instruct 모델에 대한 프롬프트 템플릿 매핑을 제공해요. 코드 보기

사용자 정의 프롬프트 템플릿을 적용하려면:

import litellm

import os
os.environ["PREDIBASE_API_KEY"] = ""

# Create your own custom prompt template
litellm.register_prompt_template(
	    model="togethercomputer/LLaMA-2-7B-32K",
        initial_prompt_value="You are a good assistant", # [OPTIONAL]
	    roles={
            "system": {
                "pre_message": "[INST] >\n", # [OPTIONAL]
                "post_message": "\n>\n [/INST]\n" # [OPTIONAL]
            },
            "user": {
                "pre_message": "[INST] ", # [OPTIONAL]
                "post_message": " [/INST]" # [OPTIONAL]
            },
            "assistant": {
                "pre_message": "\n", # [OPTIONAL]
                "post_message": "\n" # [OPTIONAL]
            }
        },
        final_prompt_value="Now answer as best you can:" # [OPTIONAL]
)

def predibase_custom_model():
    model = "predibase/togethercomputer/LLaMA-2-7B-32K"
    response = completion(model=model, messages=messages)
    print(response['choices'][0]['message']['content'])
    return response

predibase_custom_model()

# Model-specific parameters
model_list:
  - model_name: mistral-7b # model alias
    litellm_params: # actual params for litellm.completion()
      model: "predibase/mistralai/Mistral-7B-Instruct-v0.1"
      api_key: os.environ/PREDIBASE_API_KEY
      initial_prompt_value: "\n"
      roles: {"system":{"pre_message":"system\n", "post_message":""}, "assistant":{"pre_message":"assistant\n","post_message":""}, "user":{"pre_message":"user\n","post_message":""}}
      final_prompt_value: "\n"
      bos_token: ""
      eos_token: ""
      max_tokens: 4096

추가 파라미터 전달 - max_tokens, temperature

litellm.completion 지원 파라미터 전체 목록은 여기를 참고해요.

# !uv add litellm
from litellm import completion
import os
## set ENV variables
os.environ["PREDIBASE_API_KEY"] = "predibase key"

# predibae llama-3 call
response = completion(
    model="predibase/llama3-8b-instruct",
    messages = [{ "content": "Hello, how are you?","role": "user"}],
    max_tokens=20,
    temperature=0.5
)

프록시:

  model_list:
    - model_name: llama-3
      litellm_params:
        model: predibase/llama-3-8b-instruct
        api_key: os.environ/PREDIBASE_API_KEY
        max_tokens: 20
        temperature: 0.5

Predibase 전용 파라미터 전달 - adapter_id, adapter_source

litellm.completion()이 지원하지 않지만 Predibase가 지원하는 파라미터는 litellm.completion에 직접 전달해요. 예를 들어 adapter_id, adapter_source는 Predibase 전용 파라미터예요.

# !uv add litellm
from litellm import completion
import os
## set ENV variables
os.environ["PREDIBASE_API_KEY"] = "predibase key"

# predibase llama3 call
response = completion(
    model="predibase/llama-3-8b-instruct",
    messages = [{ "content": "Hello, how are you?","role": "user"}],
    adapter_id="my_repo/3",
    adapter_source="pbase",
)

프록시:

  model_list:
    - model_name: llama-3
      litellm_params:
        model: predibase/llama-3-8b-instruct
        api_key: os.environ/PREDIBASE_API_KEY
        adapter_id: my_repo/3
        adapter_source: pbase

더 알아보기 (Learn more)

• Predibase 공식 문서
• LiteLLM 컴플리션 API