turboderp-org/exllamav2
1
0
Fork 0
mirror of https://github.com/turboderp-org/exllamav2.git synced 2026-03-15 00:07:26 +00:00
Code Issues Packages Projects Releases Wiki Activity

Added Fix for Loading unquantized Huggingface models. (#771)

Browse Source 
* Added banned_strings parameter to the generator.

* Added Fix for loading unquantized huggingface models.

---------

Co-authored-by: RaahimSiddiqi <Raahim.siddiqi@vidizmo.com>
This commit is contained in:
RaahimSiddiqi
2025-04-24 03:38:10 +05:00
committed by GitHub
parent 3a90264940
commit ea6cc68ac1
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
exllamav2/config.py
View File

@@ -584,7 +584,7 @@ class ExLlamaV2Config:
self.vision_intermediate_size = read(read_config, int, ["vision_config->intermediate_size"], no_default)
self.vision_fullatt_block_indexes = read(read_config, list, ["vision_config->fullatt_block_indexes", None])
self.vision_window_size = read(read_config, int, ["vision_config->window_size", None])
assert image_processor_type == "Qwen2_5_VLImageProcessor", \
assert image_processor_type == "Qwen2_5_VLImageProcessor" or image_processor_type == "Qwen2VLImageProcessor", \
f"Wrong image processor type: {image_processor_type}"
self.vision_merger_intermediate_size = 5120 # TODO: This doesn't seem to appear in the config anywhere?