Update app.py

app.py

@@ -21,13 +21,12 @@ from transformers import (
 from transformers.image_utils import load_image
 
 # Constants for text generation
-MAX_MAX_NEW_TOKENS = 2048
-DEFAULT_MAX_NEW_TOKENS = 1024
+MAX_MAX_NEW_TOKENS = 4096
+DEFAULT_MAX_NEW_TOKENS = 1536
 MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
-# Load multimodal processor and model (Callisto OCR3)
 MODEL_ID = "nvidia/Cosmos-Reason1-7B"
 processor = AutoProcessor.from_pretrained(MODEL_ID, trust_remote_code=True)
 model_m = Qwen2_5_VLForConditionalGeneration.from_pretrained(
@@ -207,4 +206,4 @@ with gr.Blocks() as demo:
     )
 
 if __name__ == "__main__":
-    demo.queue(max_size=20).launch(share=True)
+    demo.queue(max_size=30).launch(share=True, ssr_mode=False, show_error=True)