add streamer for infer method when not eval mode

modeling_deepseekocr.py

@@ -700,7 +700,7 @@ class DeepseekOCRForCausalLM(DeepseekV2ForCausalLM):
 
 
 
-    def infer(self, tokenizer, prompt='', image_file='', output_path = '', base_size=1024, image_size=640, crop_mode=True, test_compress=False, save_results=False, eval_mode=False):
+    def infer(self, tokenizer, prompt='', image_file='', output_path = '', base_size=1024, image_size=640, crop_mode=True, test_compress=False, save_results=False, eval_mode=False, streamer=None):
         self.disable_torch_init()
 
         os.makedirs(output_path, exist_ok=True)
@@ -910,7 +910,7 @@ class DeepseekOCRForCausalLM(DeepseekV2ForCausalLM):
 
 
         if not eval_mode:
-            streamer = NoEOSTextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=False)
+            streamer = streamer or NoEOSTextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=False)
             with torch.autocast("cuda", dtype=torch.bfloat16):
                 with torch.no_grad():
                     output_ids = self.generate(