diff --git a/litellm/proxy/proxy_config.yaml b/litellm/proxy/proxy_config.yaml index 0378efced..b1cbf2e81 100644 --- a/litellm/proxy/proxy_config.yaml +++ b/litellm/proxy/proxy_config.yaml @@ -8,6 +8,10 @@ model_list: litellm_params: model: openai/* api_key: os.environ/OPENAI_API_KEY + - model_name: my-triton-model + litellm_params: + model: triton/any" + api_base: https://exampleopenaiendpoint-production.up.railway.app/triton/embeddings general_settings: store_model_in_db: true diff --git a/litellm/utils.py b/litellm/utils.py index 0babe2f0b..b4cb8a04b 100644 --- a/litellm/utils.py +++ b/litellm/utils.py @@ -4814,6 +4814,12 @@ def get_optional_params_embeddings( status_code=500, message=f"Setting dimensions is not supported for OpenAI `text-embedding-3` and later models. To drop it from the call, set `litellm.drop_params = True`.", ) + if custom_llm_provider == "triton": + keys = list(non_default_params.keys()) + for k in keys: + non_default_params.pop(k, None) + final_params = {**non_default_params, **kwargs} + return final_params if custom_llm_provider == "vertex_ai": if len(non_default_params.keys()) > 0: if litellm.drop_params is True: # drop the unsupported non-default values