kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-01-14 06:27:45 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

fix: Pass local dir to processor creation (#4018)

Browse Source 
Signed-off-by: Alexandre Milesi <30204471+milesial@users.noreply.github.com>
Co-authored-by: Alexandre Milesi <30204471+milesial@users.noreply.github.com>
Co-authored-by: Haohang Huang <31998628+symphonylyh@users.noreply.github.com>
This commit is contained in:
milesial 2025-05-06 12:25:04 -07:00 committed by GitHub
parent cba1793cda
commit 001e666fc5
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
tensorrt_llm/llmapi/llm.py
View File

@ -546,7 +546,7 @@ class LLM:
# Multimodal special handling:
# 1. Default load_tokenizer may fail because MM has different tokenizer configuration. Hence we initialize it inside input processor
# 2. May need to modify model weights for MM (e.g., resize vocab embedding). We must do such operation via input processor's __init__
self.input_processor = create_input_processor(self.args.model,
self.input_processor = create_input_processor(self._hf_model_dir,
self.tokenizer)
self.tokenizer = self.input_processor.tokenizer