rm unused models in app.py (#1)

- rm unused models in app.py (4deb69a55c5ae5112ce0d1b37b0e8a553c4704f0)


Co-authored-by: Prithiv Sakthi <[email protected]>

app.py

@@ -30,28 +30,11 @@ import subprocess
 
 subprocess.run(shlex.split("pip install flash-attn  --no-build-isolation"), env=os.environ | {"FLASH_ATTENTION_SKIP_CUDA_BUILD": "TRUE"}, check=True)
 
-
 MAX_MAX_NEW_TOKENS = 4096
 DEFAULT_MAX_NEW_TOKENS = 1024
 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 Qwen2.5-VL-7B-Instruct
-# MODEL_ID_M = "Qwen/Qwen2.5-VL-7B-Instruct"
-# processor_m = AutoProcessor.from_pretrained(MODEL_ID_M, trust_remote_code=True)
-# model_m = Qwen2_5_VLForConditionalGeneration.from_pretrained(
-#     MODEL_ID_M,
-#     trust_remote_code=True,
-#     torch_dtype=torch.float16).to(device).eval()
-
-# # Load Qwen2.5-VL-3B-Instruct
-# MODEL_ID_X = "Qwen/Qwen2.5-VL-3B-Instruct"
-# processor_x = AutoProcessor.from_pretrained(MODEL_ID_X, trust_remote_code=True)
-# model_x = Qwen2_5_VLForConditionalGeneration.from_pretrained(
-#     MODEL_ID_X,
-#     trust_remote_code=True,
-#     torch_dtype=torch.float16).to(device).eval()
-
 # Load Qwen3-VL-4B-Instruct
 MODEL_ID_Q = "Qwen/Qwen3-VL-4B-Instruct"
 processor_q = AutoProcessor.from_pretrained(MODEL_ID_Q, trust_remote_code=True)
@@ -68,14 +51,6 @@ model_y = Qwen3VLForConditionalGeneration.from_pretrained(
     trust_remote_code=True,
     torch_dtype=torch.bfloat16).to(device).eval()
 
-# # Load Qwen3-VL-8B-Thinking
-# MODEL_ID_Z = "Qwen/Qwen3-VL-8B-Thinking"
-# processor_z = AutoProcessor.from_pretrained(MODEL_ID_Z, trust_remote_code=True)
-# model_z = Qwen3VLForConditionalGeneration.from_pretrained(
-#     MODEL_ID_Z,
-#     trust_remote_code=True,
-#     torch_dtype=torch.bfloat16).to(device).eval()
-
 # Load Qwen3-VL-2B-Instruct
 MODEL_ID_L = "Qwen/Qwen3-VL-2B-Instruct"
 processor_l = AutoProcessor.from_pretrained(MODEL_ID_L, trust_remote_code=True)
@@ -183,16 +158,10 @@ def generate_image(model_name: str, text: str, image: Image.Image,
     """
     Generates responses using the selected model for image input.
     """
-    # if model_name == "Qwen2.5-VL-7B-Instruct":
-    #     processor, model = processor_m, model_m
-    # elif model_name == "Qwen2.5-VL-3B-Instruct":
-    #     processor, model = processor_x, model_x
     if model_name == "Qwen3-VL-4B-Instruct":
         processor, model = processor_q, model_q
     elif model_name == "Qwen3-VL-8B-Instruct":
         processor, model = processor_y, model_y
-    # elif model_name == "Qwen3-VL-8B-Thinking":
-    #     processor, model = processor_z, model_z
     elif model_name == "Qwen3-VL-4B-Thinking":
         processor, model = processor_t, model_t
     elif model_name == "Qwen3-VL-2B-Instruct":
@@ -229,16 +198,10 @@ def generate_video(model_name: str, text: str, video_path: str,
     """
     Generates responses using the selected model for video input.
     """
-    # if model_name == "Qwen2.5-VL-7B-Instruct":
-    #     processor, model = processor_m, model_m
-    # elif model_name == "Qwen2.5-VL-3B-Instruct":
-    #     processor, model = processor_x, model_x
     if model_name == "Qwen3-VL-4B-Instruct":
         processor, model = processor_q, model_q
     elif model_name == "Qwen3-VL-8B-Instruct":
         processor, model = processor_y, model_y
-    # elif model_name == "Qwen3-VL-8B-Thinking":
-    #     processor, model = processor_z, model_z
     elif model_name == "Qwen3-VL-4B-Thinking":
         processor, model = processor_t, model_t
     elif model_name == "Qwen3-VL-2B-Instruct":
@@ -419,7 +382,7 @@ with gr.Blocks(css=css, theme=gr.themes.Soft()) as demo:
                             ])
             
             model_choice = gr.Radio(
-                choices=["Qwen3-VL-4B-Instruct", "Qwen3-VL-8B-Instruct", "Qwen3-VL-2B-Instruct", "Qwen3-VL-2B-Thinking", "Qwen3-VL-4B-Thinking"], #"Qwen2.5-VL-3B-Instruct", "Qwen2.5-VL-7B-Instruct"],
+                choices=["Qwen3-VL-4B-Instruct", "Qwen3-VL-8B-Instruct", "Qwen3-VL-2B-Instruct", "Qwen3-VL-2B-Thinking", "Qwen3-VL-4B-Thinking"],
                 label="Select Model",
                 value="Qwen3-VL-4B-Instruct"
             )