Added missing token kwarg in Peft model loading (huggingface#1825)

mkopecki · web-flow · commit b873095a87d9 · 2024-07-10T19:11:13.000+02:00
diff --git a/trl/models/modeling_base.py b/trl/models/modeling_base.py
@@ -215,7 +215,7 @@ class and the arguments that are specific to trl models. The kwargs
 
                 # Wrap the pretrained model with the trained peft adapter
                 pretrained_model = PeftModel.from_pretrained(
-                    pretrained_model, pretrained_model_name_or_path, is_trainable=is_trainable
+                    pretrained_model, pretrained_model_name_or_path, is_trainable=is_trainable, token=token
                 )
                 logging.info("Trained peft adapter loaded")
             else:

Original file line number	Diff line number	Diff line change
`@@ -215,7 +215,7 @@ class and the arguments that are specific to trl models. The kwargs`
`215`	`215`
`216`	`216`	`# Wrap the pretrained model with the trained peft adapter`
`217`	`217`	`pretrained_model = PeftModel.from_pretrained(`
`218`		`- pretrained_model, pretrained_model_name_or_path, is_trainable=is_trainable`
	`218`	`+ pretrained_model, pretrained_model_name_or_path, is_trainable=is_trainable, token=token`
`219`	`219`	`)`
`220`	`220`	`logging.info("Trained peft adapter loaded")`
`221`	`221`	`else:`