import os
import shutil 
import spaces
import gradio as gr 
from gradio_litmodel3d import LitModel3D 
# TMP_DIR = os.path.join(
#     os.path.dirname(os.path.abspath(__file__)), "sessions/imageto3d"
# )
# os.makedirs(TMP_DIR, exist_ok=True)

# RBG_REMOVER = RembgRemover()
# SAM_PREDICTOR = SAMPredictor(model_type="vit_h")
# DELIGHT = DelightingModel()
# IMAGESR_MODEL = ImageRealESRGAN(outscale=4)
# PIPELINE = TrellisImageTo3DPipeline.from_pretrained(
#     "JeffreyXiang/TRELLIS-image-large"
# )
# # PIPELINE.cuda()

# IMAGE_BUFFER = {}
# SEG_CHECKER = ImageSegChecker(GPT_CLIENT)
# GEO_CHECKER = MeshGeoChecker(GPT_CLIENT)
# AESTHETIC_CHECKER = ImageAestheticChecker()
# CHECKERS = [GEO_CHECKER, SEG_CHECKER, AESTHETIC_CHECKER]
# URDF_CONVERTOR = URDFGenerator(GPT_CLIENT, render_view_num=4)

@spaces.GPU
def greet(n):
    print(zero.device) # <-- 'cuda:0' 🤗
    return f"Hello {zero + n} Tensor"

with gr.Blocks(
) as demo:
    with gr.Column():
        # video_output = gr.Video(
        #     label="Generated 3D Asset",
        #     autoplay=True,
        #     loop=True,
        #     height=300,
        #     interactive=False
        # )
        # model_output_gs = gr.Model3D(
        #     label="Gaussian Representation", height=300, interactive=False
        # )
        # aligned_gs = gr.Textbox(visible=False)
    
        # model_output_mesh = LitModel3D(
        #     # label="Mesh Representation",
        #     # height=300,
        #     # exposure=10,
        #     # interactive=False
        # )


        # model_output_mesh = LitModel3D(label="Extracted GLB/Gaussian", exposure=10.0, height=300)
        gr.Model3D(
            clear_color=[0.9, 0.9, 0.9, 1.0],
        )
        # gr.Markdown(
        #     """ The rendering of `Gaussian Representation` takes additional 10s. """  # noqa
        # )


if __name__ == "__main__":
    demo.launch()