Update modeling_skywork_chat.py

modeling_skywork_chat.py

@@ -212,10 +212,7 @@ class SkyworkChatModel(PreTrainedModel):
 
         img_context_token_id = tokenizer.convert_tokens_to_ids(IMG_CONTEXT_TOKEN)
         self.img_context_token_id = img_context_token_id
-        # print("##############1################")
-        # print(self.img_context_token_id)
-        # print("##############1################")
-        # exit()
+       
 
         if verbose and pixel_values is not None:
             image_bs = pixel_values.shape[0]
@@ -265,18 +262,13 @@ class SkyworkChatModel(PreTrainedModel):
 
         img_context_token_id = tokenizer.convert_tokens_to_ids(IMG_CONTEXT_TOKEN)
         self.img_context_token_id = img_context_token_id
-        # print("##############2################")
-        # print(self.img_context_token_id)
-        # print("##############2################")
+       
 
         template = get_conv_template(self.template)
         template.system_message = self.system_message
         eos_token_id = tokenizer.convert_tokens_to_ids(template.sep.strip())
-        # print("##############2.5################")
-        # print(template.sep.strip())
-        # print(eos_token_id)
-        # print("##############2.5################")
-
+     
+       
         history = [] if history is None else history
         for (old_question, old_answer) in history:
             template.append_message(template.roles[0], old_question)
@@ -284,13 +276,6 @@ class SkyworkChatModel(PreTrainedModel):
         template.append_message(template.roles[0], question)
         template.append_message(template.roles[1], None)
         query = template.get_prompt()
-#         print("##############3################")
-#         print(query)
-#         print("##############3################")
-#         query = """<｜begin▁of▁sentence｜>user
-# <image>
-# 图片内容是什么？<｜end▁of▁sentence｜>
-# <｜begin▁of▁sentence｜>assistant"""
 
         if verbose and pixel_values is not None:
             image_bs = pixel_values.shape[0]
@@ -299,9 +284,7 @@ class SkyworkChatModel(PreTrainedModel):
         for num_patches in num_patches_list:
             image_tokens = IMG_START_TOKEN + IMG_CONTEXT_TOKEN * self.num_image_token * num_patches + IMG_END_TOKEN
             query = query.replace('<image>', image_tokens, 1)
-        # print("##############4################")
-        # # print(query)
-        # print("##############4################")
+       
 
         model_inputs = tokenizer(query, return_tensors='pt')
         input_ids = model_inputs['input_ids'].to(self.device)
@@ -316,7 +299,7 @@ class SkyworkChatModel(PreTrainedModel):
         response = tokenizer.batch_decode(generation_output, skip_special_tokens=True)[0]
         response = response.split(template.sep.strip())[0].strip()
         history.append((question, response))
-        # print("###" + str(response))
+       
         if return_history:
             return response, history
         else:
@@ -350,12 +333,7 @@ class SkyworkChatModel(PreTrainedModel):
 
             input_ids = input_ids.reshape(B * N)
             selected = (input_ids == self.img_context_token_id)
-            # print("#######################5####################")
-            # print(self.img_context_token_id)
-            # print(selected)
-            # print(selected.sum())
-            # print("#######################5####################")
-            # exit()
+          
             assert selected.sum() != 0
             input_embeds[selected] = vit_embeds.reshape(-1, C).to(input_embeds.device)
 
@@ -363,11 +341,7 @@ class SkyworkChatModel(PreTrainedModel):
         else:
             input_embeds = self.language_model.get_input_embeddings()(input_ids)
 
-        # print("#######################6####################")
-        # print(attention_mask)
-        # print(attention_mask.sum())
-        # print(output_hidden_states)
-        # print("#######################6####################")
+       
 
         outputs = self.language_model.generate(
             inputs_embeds=input_embeds,