apps/third_party/Wonder3D/render_codes/distributed.py

# multiprocessing render
import json
import multiprocessing
import subprocess
from dataclasses import dataclass
from typing import Optional
import os

import boto3

import argparse

parser = argparse.ArgumentParser(description='distributed rendering')

parser.add_argument('--workers_per_gpu', type=int,
                    help='number of workers per gpu.')
parser.add_argument('--input_models_path', type=str,
                    help='Path to a json file containing a list of 3D object files.')
parser.add_argument('--upload_to_s3', type=bool, default=False,
                    help='Whether to upload the rendered images to S3.')
parser.add_argument('--log_to_wandb', type=bool, default=False,
                    help='Whether to log the progress to wandb.')
parser.add_argument('--num_gpus', type=int, default=-1,
                    help='number of gpus to use. -1 means all available gpus.')
parser.add_argument('--gpu_list',nargs='+', type=int, 
                    help='the avalaible gpus')

parser.add_argument('--mode', type=str, default='render', 
                choices=['render_ortho', 'render_persp'],
                    help='use orthogonal camera or perspective camera')

parser.add_argument('--start_i', type=int, default=0,
                    help='the index of first object to be rendered.')

parser.add_argument('--end_i', type=int, default=-1,
                    help='the index of the last object to be rendered.')

parser.add_argument('--objaverse_root', type=str, default='/ghome/l5/xxlong/.objaverse/hf-objaverse-v1',
                    help='Path to a json file containing a list of 3D object files.')

parser.add_argument('--save_folder', type=str, default=None,
                    help='Path to a json file containing a list of 3D object files.')

parser.add_argument('--blender_install_path', type=str, default=None,
                    help='blender path.')

parser.add_argument('--view_idx', type=int, default=2,
                    help='the number of render views.')

parser.add_argument('--ortho_scale', type=float, default=1.25,
                    help='ortho rendering usage; how large the object is')

parser.add_argument('--random_pose', action='store_true',
                    help='whether randomly rotate the poses to be rendered')

args = parser.parse_args()


view_idx = args.view_idx

VIEWS = ["front", "back", "right", "left", "front_right", "front_left", "back_right", "back_left"]

def check_task_finish(render_dir, view_index):
    files_type = ['rgb', 'normals']
    flag = True
    view_index = "%03d" % view_index
    if os.path.exists(render_dir):
        for t in files_type:
            for face in VIEWS:
                fpath = os.path.join(render_dir, f'{t}_{view_index}_{face}.webp')
                # print(fpath)
                if not os.path.exists(fpath):
                    flag = False
    else:
        flag = False

    return flag

def worker(
    queue: multiprocessing.JoinableQueue,
    count: multiprocessing.Value,
    gpu: int,
    s3: Optional[boto3.client],
) -> None:
    while True:
        item = queue.get()
        if item is None:
            break

        view_path = os.path.join(args.save_folder, item.split('/')[-1][:2], item.split('/')[-1][:-4])
        print(view_path)
        if  'render' in args.mode:
            if check_task_finish(view_path, view_idx):
                queue.task_done()
                print('========', item, 'rendered', '========')
                
                continue
            else:
                os.makedirs(view_path, exist_ok = True)

        # Perform some operation on the item
        print(item, gpu)

        if args.mode == 'render_ortho':
            command = (
                f" CUDA_VISIBLE_DEVICES={gpu} "
                f" blenderproc run --blender-install-path {args.blender_install_path} blenderProc_ortho.py"
                f" --object_path {item} --view {view_idx}"
                f" --output_folder {args.save_folder}"
                f" --ortho_scale {args.ortho_scale} "
            )
            if args.random_pose:
                print("random pose to render")
                command += f" --random_pose"
        elif args.mode == 'render_persp':
            command = (
                f" CUDA_VISIBLE_DEVICES={gpu} "
                f" blenderproc run --blender-install-path {args.blender_install_path} blenderProc_persp.py"
                f" --object_path {item} --view {view_idx}"
                f" --output_folder {args.save_folder}"
            )
            if args.random_pose:
                print("random pose to render")
                command += f" --random_pose"
    
        print(command)
        subprocess.run(command, shell=True)

        with count.get_lock():
            count.value += 1

        queue.task_done()


if __name__ == "__main__":
    # args = tyro.cli(Args)

    s3 = boto3.client("s3") if args.upload_to_s3 else None
    queue = multiprocessing.JoinableQueue()
    count = multiprocessing.Value("i", 0)

    # Start worker processes on each of the GPUs
    for gpu_i in range(args.num_gpus):
        for worker_i in range(args.workers_per_gpu):
            worker_i = gpu_i * args.workers_per_gpu + worker_i
            process = multiprocessing.Process(
                target=worker, args=(queue, count, args.gpu_list[gpu_i], s3)
            )
            process.daemon = True
            process.start()
        
    # Add items to the queue
    if args.input_models_path is not None:
        with open(args.input_models_path, "r") as f:
            model_paths = json.load(f)

    args.end_i = len(model_paths) if args.end_i > len(model_paths) else args.end_i

    for item in model_paths[args.start_i:args.end_i]:

        if os.path.exists(os.path.join(args.objaverse_root, os.path.basename(item))):
            obj_path = os.path.join(args.objaverse_root, os.path.basename(item))
        elif os.path.exists(os.path.join(args.objaverse_root, item)):
            obj_path = os.path.join(args.objaverse_root, item)
        else:
            obj_path = os.path.join(args.objaverse_root, item[:2], item+".glb")
        queue.put(obj_path)

    # Wait for all tasks to be completed
    queue.join()

    # Add sentinels to the queue to stop the worker processes
    for i in range(args.num_gpus * args.workers_per_gpu):
        queue.put(None)