billbai0102
/

qwen-vl-int4

Text Generation

4-bit precision

Model card Files Files and versions

billbai0102 commited on Sep 14, 2023

Commit

2e6f82c

·

1 Parent(s): 1587fa5

Update modeling_qwen.py

Files changed (1) hide show

modeling_qwen.py +2 -0

modeling_qwen.py CHANGED Viewed

@@ -537,6 +537,8 @@ class QWenModel(QWenPreTrainedModel):
                 images.append(bytes(image).decode('utf-8'))
             images = self.visual.encode(images)
         return images

                 images.append(bytes(image).decode('utf-8'))
             images = self.visual.encode(images)
+        torch.cuda.empty_cache()
         return images