Upload folder using huggingface_hub

app.py

@@ -1,13 +1,13 @@
-import torch
+# import torch
 import gradio as gr
 from diffusers.utils import load_image
 import spaces
-# from panna.pipeline import PipelineDepth2ImageV2
-from panna import Depth2Image, DepthAnythingV2
+from panna.pipeline import PipelineDepth2ImageV2
+# from panna import Depth2Image, DepthAnythingV2
 
-# model = PipelineDepth2ImageV2()
-model_depth = DepthAnythingV2("depth-anything/Depth-Anything-V2-Large-hf", torch_dtype=torch.float32)
-model_image = Depth2Image("stabilityai/stable-diffusion-2-depth")
+model = PipelineDepth2ImageV2()
+# model_depth = DepthAnythingV2()
+# model_image = Depth2Image()
 title = ("# [Depth2Image](https://huggingface.co/stabilityai/stable-diffusion-2-depth) with [DepthAnythingV2](https://huggingface.co/depth-anything/Depth-Anything-V2-Large-hf)\n"
          "Depth2Image with depth map predicted by DepthAnything V2. The demo is part of [panna](https://github.com/abacws-abacus/panna) project.")
 example_files = []
@@ -16,34 +16,34 @@ for n in range(1, 10):
     example_files.append(f"demo{n:0>2}.jpg")
 
 
-@spaces.GPU
-def infer(init_image, prompt, negative_prompt, seed, width, height, guidance_scale, num_inference_steps):
-    depth = model_depth.image2depth([init_image], return_tensor=True)
-    return model_image.text2image(
-        [init_image],
-        depth_maps=depth,
-        prompt=[prompt],
-        negative_prompt=[negative_prompt],
-        guidance_scale=guidance_scale,
-        num_inference_steps=num_inference_steps,
-        height=height,
-        width=width,
-        seed=seed
-    )[0]
-
-
 # @spaces.GPU
 # def infer(init_image, prompt, negative_prompt, seed, width, height, guidance_scale, num_inference_steps):
-#     return model(
-#         init_image,
-#         prompt=prompt,
-#         negative_prompt=negative_prompt,
+#     depth = model_depth.image2depth([init_image], return_tensor=True)
+#     return model_image.text2image(
+#         [init_image],
+#         depth_maps=depth,
+#         prompt=[prompt],
+#         negative_prompt=[negative_prompt],
 #         guidance_scale=guidance_scale,
 #         num_inference_steps=num_inference_steps,
 #         height=height,
 #         width=width,
 #         seed=seed
-#     )
+#     )[0]
+
+
+@spaces.GPU
+def infer(init_image, prompt, negative_prompt, seed, width, height, guidance_scale, num_inference_steps):
+    return model(
+        init_image,
+        prompt=prompt,
+        negative_prompt=negative_prompt,
+        guidance_scale=guidance_scale,
+        num_inference_steps=num_inference_steps,
+        height=height,
+        width=width,
+        seed=seed
+    )
 
 
 with gr.Blocks() as demo: