import torch

print(torch.__version__)
print(torch.version.cuda)
print(torch.cuda.is_available())

import os, subprocess
import uuid, tempfile
from glob import glob

env_list = os.environ['PATH'].split(':')
env_list.append('/usr/local/cuda/bin')
os.environ['PATH'] = ':'.join(env_list)
os.environ['TORCH_CUDA_ARCH_LIST'] = '8.6'

import gradio as gr
from huggingface_hub import snapshot_download

os.makedirs("pretrained", exist_ok=True)
snapshot_download(
    repo_id = "jiawei011/L4GM",
    local_dir = "./pretrained"
)

# Folder containing example images
examples_folder = "data_test"

# Retrieve all file paths in the folder
video_examples = [
    os.path.join(examples_folder, file)
    for file in os.listdir(examples_folder)
    if os.path.isfile(os.path.join(examples_folder, file))
]


def generate(input_video):

    unique_id = str(uuid.uuid4())
    
    workdir = f"results_{unique_id}"
    recon_model = "pretrained/recon.safetensors"
    interp_model = "pretrained/interp.safetensors"
    num_frames = 16
    test_path = input_video
    
    try:
        # Run the inference command
        subprocess.run(
            [
                "python", "infer_3d.py", "big",
                "--workspace", f"{workdir}",
                "--resume", f"{recon_model}",
                "--num_frames", f"1",
                "--test_path", f"{test_path}",
            ],
            check=True
        )

        subprocess.run(
            [
                "python", "infer_4d.py", "big",
                "--workspace", f"{workdir}",
                "--resume", f"{recon_model}",
                "--interpresume", f"{interp_model}",
                "--num_frames", f"{num_frames}",
                "--test_path", f"{test_path}",
            ],
            check=True
        )

        output_videos = glob(os.path.join(f"{workdir}", "*.mp4"))
        print(output_videos)
        return output_videos[4]
   
    except subprocess.CalledProcessError as e:
        raise gr.Error(f"Error during inference: {str(e)}")

with gr.Blocks() as demo:
    with gr.Column():
        with gr.Row():
            with gr.Column():
                input_video = gr.Video(label="Input Video")
                submit_btn = gr.Button("Submit")
            with gr.Column():
    
                output_result_4 = gr.Video(label="Result")

        gr.Examples(
            examples = video_examples,
            inputs = [input_video]
        )

    submit_btn.click(
        fn = generate,
        inputs = [input_video],
        outputs = [
            output_result_4
        ]
    )

demo.queue().launch(show_api=False, show_error=True)