Triton Inference Server

LiteLLM supports Embedding Models on Triton Inference Servers

Usage

SDK

Example Call

Use the triton/ prefix to route to triton server

from litellm import embedding
import os

response = await litellm.aembedding(
    model="triton/<your-triton-model>",                                                       
    api_base="https://your-triton-api-base/triton/embeddings", # /embeddings endpoint you want litellm to call on your server
    input=["good morning from litellm"],
)

PROXY

1. Add models to your config.yaml

   model_list:
     - model_name: my-triton-model
       litellm_params:
         model: triton/<your-triton-model>"
         api_base: https://your-triton-api-base/triton/embeddings

2. Start the proxy

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

3. Send Request to LiteLLM Proxy Server

   OpenAI Python v1.0.0+

   import openai
   from openai import OpenAI
   
   # set base_url to your proxy server
   # set api_key to send to proxy server
   client = OpenAI(api_key="<proxy-api-key>", base_url="http://0.0.0.0:4000")
   
   response = client.embeddings.create(
       input=["hello from litellm"],
       model="my-triton-model"
   )
   
   print(response)
   

   curl

   --header is optional, only required if you're using litellm proxy with Virtual Keys

   curl --location 'http://0.0.0.0:4000/embeddings' \
   --header 'Content-Type: application/json' \
   --header 'Authorization: Bearer sk-1234' \
   --data ' {
   "model": "my-triton-model",
   "input": ["write a litellm poem"]
   }'