chore: add extra empty cache
This commit is contained in:
@@ -197,6 +197,7 @@ class ModelLoader:
|
|||||||
if get_moe_quantized_count() > 0:
|
if get_moe_quantized_count() > 0:
|
||||||
self.model._moe_experts_quantized = True
|
self.model._moe_experts_quantized = True
|
||||||
patch_peft_target_parameters_matching()
|
patch_peft_target_parameters_matching()
|
||||||
|
torch.cuda.empty_cache()
|
||||||
|
|
||||||
PLUGIN_MANAGER.post_model_build(self.cfg, self.model)
|
PLUGIN_MANAGER.post_model_build(self.cfg, self.model)
|
||||||
|
|
||||||
|
|||||||
Reference in New Issue
Block a user