add stream params for chat and char_crop

modeling_GOT.py

@@ -484,7 +484,7 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
         setattr(torch.nn.Linear, "reset_parameters", lambda self: None)
         setattr(torch.nn.LayerNorm, "reset_parameters", lambda self: None)
 
-    def chat(self, tokenizer, image_file, ocr_type, ocr_box='', ocr_color='', render=False, save_render_file=None, print_prompt=False, gradio_input=False, stream_flag = False):
+    def chat(self, tokenizer, image_file, ocr_type, ocr_box='', ocr_color='', render=False, save_render_file=None, print_prompt=False, gradio_input=False, stream_flag = False, streamer=None):
 
         self.disable_torch_init()
 
@@ -563,8 +563,8 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
         stop_str = conv.sep if conv.sep_style != SeparatorStyle.TWO else conv.sep2
         keywords = [stop_str]
         stopping_criteria = KeywordsStoppingCriteria(keywords, tokenizer, input_ids)
-        streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
-
+        streamer = streamer if streamer else TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
+        
         if stream_flag:
             with torch.autocast("cuda", dtype=torch.bfloat16):
                 output_ids = self.generate(
@@ -728,7 +728,7 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
         return processed_images
 
 
-    def chat_crop(self, tokenizer, image_file, ocr_type, render=False, save_render_file=None, print_prompt=False, gradio_input=False, stream_flag = False):
+    def chat_crop(self, tokenizer, image_file, ocr_type, render=False, save_render_file=None, print_prompt=False, gradio_input=False, stream_flag = False, streamer=None):
         # Model
         self.disable_torch_init()
         multi_page=False
@@ -817,7 +817,7 @@ class GOTQwenForCausalLM(Qwen2ForCausalLM):
         stop_str = conv.sep if conv.sep_style != SeparatorStyle.TWO else conv.sep2
         keywords = [stop_str]
         stopping_criteria = KeywordsStoppingCriteria(keywords, tokenizer, input_ids)
-        streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
+        streamer = streamer if streamer else TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
 
         if stream_flag:
             with torch.autocast("cuda", dtype=torch.bfloat16):