update chat

multimodal/open_flamingo/chat/conversation.py

@@ -368,16 +368,26 @@ class Chat:
         image_nums = [1] * len(input_ids)
         # and torch.cuda.amp.autocast(dtype=torch.float16)
         with torch.no_grad():
-            outputs = self.model.generate(
-                batch_images,
-                input_ids,
+            outputs = model(
+                vision_x=vision_x,
+                lang_x=lang_x,
                 attention_mask=attention_mask,
-                max_new_tokens=100,
-                # min_new_tokens=8,
-                num_beams=1,
-                image_start_index_list=image_start_index_list,
                 image_nums=image_nums,
+                image_start_index_list=image_start_index_list,
+                added_bbox_list=None,
+                add_box=False,
             )
+        # with torch.no_grad():
+        #     outputs = self.model.generate(
+        #         batch_images,
+        #         input_ids,
+        #         attention_mask=attention_mask,
+        #         max_new_tokens=100,
+        #         # min_new_tokens=8,
+        #         num_beams=1,
+        #         image_start_index_list=image_start_index_list,
+        #         image_nums=image_nums,
+        #     )
         boxes = outputs["boxes"]
         scores = outputs["scores"]
         if len(scores) > 0: