tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

skip set_quant_config if quantization not given

Browse Source
This commit is contained in:
Sunny Liu
2025-04-21 17:17:41 -04:00
parent fcef8c95fe
commit 266ef3f479
1 changed files with 3 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/axolotl/utils/models.py
View File

@@ -834,6 +834,8 @@ class ModelLoader:
del self.model_kwargs["device_map"]
def set_quantization_config(self) -> None:
if not self.cfg.quantization:
return
self.model_kwargs["load_in_8bit"] = self.cfg.load_in_8bit
self.model_kwargs["load_in_4bit"] = self.cfg.load_in_4bit
@@ -887,7 +889,7 @@ class ModelLoader:
# but deepspeed needs this still in bfloat16
bnb_config["bnb_4bit_quant_storage"] = torch.float32
if self.cfg.quantization and self.cfg.quantization.bnb_config_kwargs:
if self.cfg.quantization.bnb_config_kwargs:
bnb_config.update(self.cfg.quantization.bnb_config_kwargs)
self.model_kwargs["quantization_config"] = BitsAndBytesConfig(