Upload folder using huggingface_hub

README.md

@@ -1,8 +1,8 @@
 ---
 title: CLIP Interrogator
 emoji: 🕵️‍♂️
-colorFrom: green
-colorTo: purple
+colorFrom: blue
+colorTo: red
 sdk: gradio
 sdk_version: 4.8.0
 app_file: app.py

app.py

@@ -1,16 +1,8 @@
 import spaces
-import torch
 import gradio as gr
-from clip_interrogator import Config, Interrogator
+from panna import CLIPInterrogator
 
-
-config = Config()
-config.device = 'cuda' if torch.cuda.is_available() else 'cpu'
-config.blip_offload = False if torch.cuda.is_available() else True
-config.chunk_size = 2048
-config.flavor_intermediate_count = 512
-config.blip_num_beams = 64
-ci = Interrogator(config)
+model = CLIPInterrogator()
 css = """
 #col-container {
     margin: 0 auto;
@@ -20,31 +12,18 @@ css = """
 
 
 @spaces.GPU
-def infer(image, mode, best_max_flavors):
+def infer(image, best_max_flavors):
     image = image.convert('RGB')
-    if mode == 'best':
-        prompt_result = ci.interrogate(image, max_flavors=int(best_max_flavors))
-    elif mode == 'classic':
-        prompt_result = ci.interrogate_classic(image)
-    else:
-        prompt_result = ci.interrogate_fast(image)
-    return prompt_result
+    return model.image2text([image], best_max_flavors=best_max_flavors)[0]
 
 
 with gr.Blocks(css=css) as demo:
     with gr.Column(elem_id="col-container"):
         gr.Markdown("# CLIP Interrogator")
-        input_image = gr.Image(type='pil', elem_id="input-img")
+        input_image = gr.Image(type='pil')
         with gr.Row():
-            mode_input = gr.Radio(['best', 'classic', 'fast'], label='Select mode', value='best')
             flavor_input = gr.Slider(minimum=2, maximum=48, step=2, value=32, label='best mode max flavors')
         run_button = gr.Button("Submit")
         output_text = gr.Textbox(label="Description Output")
-    run_button.click(
-        fn=infer,
-        inputs=[input_image, mode_input, flavor_input],
-        outputs=[output_text],
-        concurrency_limit=10
-    )
-
-demo.queue().launch()
+    run_button.click(fn=infer, inputs=[input_image, flavor_input], outputs=[output_text], concurrency_limit=10)
+demo.launch()

requirements.txt

@@ -3,3 +3,4 @@ torch
 torchvision
 spaces
 clip-interrogator
+panna