Update app.py

app.py

@@ -29,6 +29,15 @@ 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 Behemoth-3B-070225-post0.1
+MODEL_ID_N = "prithivMLmods/Behemoth-3B-070225-post0.1"
+processor_n = AutoProcessor.from_pretrained(MODEL_ID_N, trust_remote_code=True)
+model_n = Qwen2_5_VLForConditionalGeneration.from_pretrained(
+    MODEL_ID_N,
+    trust_remote_code=True,
+    torch_dtype=torch.float16
+).to(device).eval()
+
 # Load SkyCaptioner-V1
 MODEL_ID_M = "Skywork/SkyCaptioner-V1"
 processor_m = AutoProcessor.from_pretrained(MODEL_ID_M, trust_remote_code=True)
@@ -65,7 +74,7 @@ model_y = Qwen2_5_VLForConditionalGeneration.from_pretrained(
     torch_dtype=torch.float16
 ).to(device).eval()
 
-
+#video sampling
 def downsample_video(video_path):
     """
     Downsamples the video to evenly spaced frames.
@@ -100,6 +109,9 @@ def generate_image(model_name: str, text: str, image: Image.Image,
     if model_name == "SkyCaptioner-V1":
         processor = processor_m
         model = model_m
+    elif model_name == "Behemoth-3B-070225-post0.1":
+        processor = processor_n
+        model = model_n
     elif model_name == "SpaceThinker-3B":
         processor = processor_z
         model = model_z
@@ -157,6 +169,9 @@ def generate_video(model_name: str, text: str, video_path: str,
     if model_name == "SkyCaptioner-V1":
         processor = processor_m
         model = model_m
+    elif model_name == "Behemoth-3B-070225-post0.1":
+        processor = processor_n
+        model = model_n
     elif model_name == "SpaceThinker-3B":
         processor = processor_z
         model = model_z
@@ -267,7 +282,7 @@ with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
         with gr.Column():
             output = gr.Textbox(label="Output", interactive=False, lines=2, scale=2)
             model_choice = gr.Radio(
-                choices=["SkyCaptioner-V1", "SpaceThinker-3B", "coreOCR-7B-050325-preview", "SpaceOm-3B"],
+                choices=["SkyCaptioner-V1", "Behemoth-3B-070225-post0.1", "SpaceThinker-3B", "coreOCR-7B-050325-preview", "SpaceOm-3B"],
                 label="Select Model",
                 value="SkyCaptioner-V1"
             )