Skip to content

GitLab

Unverified Commit c40f3525 authored by jh-nv's avatar jh-nv Committed by GitHub
Browse files

chore: use config.modality directly for multimodal init and remove incorrect fallback value (#6471)

parent 038b50d2
Show whitespace changes
Inline Side-by-side
Showing with 1 addition and 2 deletions
components/src/dynamo/trtllm/workers/llm_worker.py
View file @ c40f3525
...@@ -166,7 +166,6 @@ async def init_llm_worker( ...@@ -166,7 +166,6 @@ async def init_llm_worker(
) )
kv_connector_config = build_kv_connector_config(config) kv_connector_config = build_kv_connector_config(config)
modality = getattr(config, "modality", None) or "text"
arg_map = { arg_map = {
"model": model_path, "model": model_path,
"scheduler_config": scheduler_config, "scheduler_config": scheduler_config,
...@@ -295,7 +294,7 @@ async def init_llm_worker( ...@@ -295,7 +294,7 @@ async def init_llm_worker(
# This overrides the skip_tokenizer_init=True set earlier # This overrides the skip_tokenizer_init=True set earlier
engine_args["skip_tokenizer_init"] = False engine_args["skip_tokenizer_init"] = False
if modality == Modality.MULTIMODAL: if config.modality == Modality.MULTIMODAL:
engine_args["skip_tokenizer_init"] = False engine_args["skip_tokenizer_init"] = False
model_config = AutoConfig.from_pretrained(config.model, trust_remote_code=True) model_config = AutoConfig.from_pretrained(config.model, trust_remote_code=True)
multimodal_processor = MultimodalRequestProcessor( multimodal_processor = MultimodalRequestProcessor(
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment