phoenix-oss/llama-stack
1
0
Fork 0
forked from phoenix-oss/llama-stack-mirror
Code Pull requests Releases Packages 2 Activity Actions

chore: remove llama_models.llama3.api imports from providers (#1107)

Browse source 
There should be a choke-point for llama3.api imports -- this is the
prompt adapter. Creating a ChatFormat() object on demand is inexpensive.
The underlying Tokenizer is a singleton anyway.
This commit is contained in:
Ashwin Bharambe 2025-02-19 19:01:29 -08:00 • committed by GitHub
parent e9b8259cf9
commit cdcbeb005b
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
13 changed files with 77 additions and 113 deletions
Download patch file Download diff file Expand all files Collapse all files

17
llama_stack/providers/remote/inference/cerebras/cerebras.py
View file

@ -7,8 +7,6 @@
from typing import AsyncGenerator, List, Optional, Union
from cerebras.cloud.sdk import AsyncCerebras
from llama_models.llama3.api.chat_format import ChatFormat
from llama_models.llama3.api.tokenizer import Tokenizer
from llama_stack.apis.common.content_types import InterleavedContent
from llama_stack.apis.inference import (
@ -53,7 +51,6 @@ class CerebrasInferenceAdapter(ModelRegistryHelper, Inference):
model_aliases=model_aliases,
)
self.config = config
self.formatter = ChatFormat(Tokenizer.get_instance())
self.client = AsyncCerebras(
base_url=self.config.base_url,
@ -96,14 +93,14 @@ class CerebrasInferenceAdapter(ModelRegistryHelper, Inference):
r = await self.client.completions.create(**params)
return process_completion_response(r, self.formatter)
return process_completion_response(r)
async def _stream_completion(self, request: CompletionRequest) -> AsyncGenerator:
params = await self._get_params(request)
stream = await self.client.completions.create(**params)
async for chunk in process_completion_stream_response(stream, self.formatter):
async for chunk in process_completion_stream_response(stream):
yield chunk
async def chat_completion(
@ -143,14 +140,14 @@ class CerebrasInferenceAdapter(ModelRegistryHelper, Inference):
r = await self.client.completions.create(**params)
return process_chat_completion_response(r, self.formatter, request)
return process_chat_completion_response(r, request)
async def _stream_chat_completion(self, request: CompletionRequest) -> AsyncGenerator:
params = await self._get_params(request)
stream = await self.client.completions.create(**params)
async for chunk in process_chat_completion_stream_response(stream, self.formatter, request):
async for chunk in process_chat_completion_stream_response(stream, request):
yield chunk
async def _get_params(self, request: Union[ChatCompletionRequest, CompletionRequest]) -> dict:
@ -159,11 +156,9 @@ class CerebrasInferenceAdapter(ModelRegistryHelper, Inference):
prompt = ""
if isinstance(request, ChatCompletionRequest):
prompt = await chat_completion_request_to_prompt(
request, self.get_llama_model(request.model), self.formatter
)
prompt = await chat_completion_request_to_prompt(request, self.get_llama_model(request.model))
elif isinstance(request, CompletionRequest):
prompt = await completion_request_to_prompt(request, self.formatter)
prompt = await completion_request_to_prompt(request)
else:
raise ValueError(f"Unknown request type {type(request)}")