Update app.py

app.py

@@ -55,7 +55,7 @@ def downsample_video(video_path):
     return frames
 
 MODEL_ID = "kaamd/chtvctr"  # Alternatively: "Qwen/Qwen2.5-VL-3B-Instruct" 
-processor = AutoProcessor.from_pretrained(MODEL_ID, trust_remote_code=True)
+processor = AutoProcessor.from_pretrained(MODEL_ID, trust_remote_code=True, min_pixels=256*28*28, max_pixels=1280*28*28)
 model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     MODEL_ID,
     trust_remote_code=True,
@@ -102,7 +102,7 @@ def model_inference(input_dict, history):
         ).to("cuda")
         # Set up streaming generation.
         streamer = TextIteratorStreamer(processor, skip_prompt=True, skip_special_tokens=True)
-        generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=256, no_repeat_ngram_size=20, temperature=2, num_beams=5, repetition_penalty=1.5)
+        generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=256, no_repeat_ngram_size=20, temperature=2, top_p=0.6, top_k=5, num_beams=5, repetition_penalty=1.5)
         thread = Thread(target=model.generate, kwargs=generation_kwargs)
         thread.start()
         buffer = ""