theroyallab/tabbyAPI
1
0
Fork 0
mirror of https://github.com/theroyallab/tabbyAPI.git synced 2026-05-11 08:20:08 +00:00
Code Issues Packages Projects Releases Wiki Activity

ExLlamaV3: Respect device split when loading draft model

Browse Source
This commit is contained in:
turboderp
2026-04-25 01:51:46 +02:00
parent 6aa842a1b2
commit e909f7ecdb
1 changed files with 2 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
backends/exllamav3/model.py
View File

@@ -500,6 +500,7 @@ class ExllamaV3Container(BaseModelContainer):
if self.use_vision:
for value in self.vision_model.load_gen(
reserve_per_device=self.autosplit_reserve,
use_per_device=self.gpu_split,
callback=progress_callback,
):
if value:
@@ -508,6 +509,7 @@ class ExllamaV3Container(BaseModelContainer):
if self.use_draft_model:
for value in self.draft_model.load_gen(
reserve_per_device=self.autosplit_reserve,
use_per_device=self.gpu_split,
callback=progress_callback,
):
if value: