xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-14 06:25:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

[V1] Fix Detokenizer loading in AsyncLLM (#10997)

Browse Source 
Signed-off-by: Roger Wang <ywang@roblox.com>
This commit is contained in:
Roger Wang 2024-12-09 08:27:10 -08:00 committed by GitHub
parent d1c2e15eb3
commit c690357928
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 6 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
vllm/v1/engine/async_llm.py
View File

@ -65,7 +65,12 @@ class AsyncLLM(EngineClient):
input_registry) input_registry)
# Detokenizer (converts EngineCoreOutputs --> RequestOutput). # Detokenizer (converts EngineCoreOutputs --> RequestOutput).
self.detokenizer = Detokenizer(vllm_config.model_config.tokenizer) self.detokenizer = Detokenizer(
tokenizer_name=vllm_config.model_config.tokenizer,
tokenizer_mode=vllm_config.model_config.tokenizer_mode,
trust_remote_code=vllm_config.model_config.trust_remote_code,
revision=vllm_config.model_config.tokenizer_revision,
)
# EngineCore (starts the engine in background process). # EngineCore (starts the engine in background process).
self.engine_core = EngineCoreClient.make_client( self.engine_core = EngineCoreClient.make_client(