Upload modeling_mplug_owl2.py with huggingface_hub

modeling_mplug_owl2.py

@@ -282,9 +282,10 @@ class MPLUGOwl2LlamaForCausalLM(LlamaForCausalLM, MPLUGOwl2MetaForCausalLM):
                 return torch.softmax(output_logits, -1) @ self.weight_tensor
         else:
             video = [[expand2square(frame, tuple(int(x*255) for x in self.image_processor.image_mean)) for frame in vid] for vid in images]
+            input_ids = tokenizer_image_token(prompt, self.tokenizer, IMAGE_TOKEN_INDEX, return_tensors='pt').unsqueeze(0).to(self.device)
             with torch.inference_mode():
                 video_tensors = [self.image_processor.preprocess(vid, return_tensors="pt")["pixel_values"].half().to(self.model.device) for vid in video]
-                output_logits = self(self.input_ids.repeat(len(video_tensors), 1),
+                output_logits = self(input_ids.repeat(len(video_tensors), 1),
                             images=video_tensors)["logits"][:,-1, self.preferential_ids_]
                 return torch.softmax(output_logits, -1) @ self.weight_tensor