scripts/triplane_rec_inference.py

"""
Train a diffusion model on images.
"""
import sys
import os
sys.path.append('.')

import torch as th
import torch.multiprocessing as mp

import argparse
import dnnlib
from guided_diffusion import dist_util, logger
from guided_diffusion.script_util import (
    args_to_dict,
    add_dict_to_argparser,
)
from nsr.train_util import TrainLoop3DRec as TrainLoop
from nsr.script_util import create_3DAE_model, encoder_and_nsr_defaults, create_Triplane, loss_defaults
from datasets.shapenet import load_eval_rays, load_data, load_eval_data
from nsr.losses.builder import E3DGELossClass

from pdb import set_trace as st

def inference_loop(rank, master_addr, master_port, args):
    dist_util.setup_dist(rank, master_addr, master_port, args.gpus)

    logger.configure(dir=args.logdir)

    logger.log("creating eval rays...")
    # TODO, load shapenet data
    eval_data = load_eval_data(
        file_path=args.data_dir,
        batch_size=args.batch_size,
        reso=args.image_size,
        reso_encoder=args.image_size_encoder, # 224 -> 128
        num_workers=args.num_workers,
        load_depth=args.depth_lambda > 0
    )
    # c_list = load_eval_rays(
    #     file_path=args.data_dir,
    #     reso=args.image_size,
    #     reso_encoder=args.image_size_encoder, # 224 -> 128
    # )

    # try dry run
    # batch = next(data)
    # batch = None

    # logger.log("creating model and diffusion...")

    logger.log("loading encoder and NSR decoder...") 
    auto_encoder = create_Triplane( # basically overfitting tirplane
        **args_to_dict(args,
                       encoder_and_nsr_defaults().keys()))

    # auto_encoder = create_3DAE_model(
    #     **args_to_dict(args,
    #                    encoder_and_nsr_defaults().keys()))
    auto_encoder.to(dist_util.dev())
    auto_encoder.eval()

    # schedule_sampler = create_named_schedule_sampler(args.schedule_sampler, diffusion)


    opt = dnnlib.EasyDict(args_to_dict(args, loss_defaults().keys()) )
    loss_class = E3DGELossClass(dist_util.dev(), opt).to(dist_util.dev())

    logger.log("training...")
    TrainLoop(
        model=auto_encoder,
        # encoder,
        # decoder
        loss_class=loss_class,
        # diffusion=diffusion,
        data=eval_data, # TODO
        # data=batch,
        batch_size=args.batch_size,
        microbatch=args.microbatch,
        lr=args.lr,
        ema_rate=args.ema_rate,
        log_interval=args.log_interval,
        save_interval=args.save_interval,
        resume_checkpoint=args.resume_checkpoint,
        use_fp16=args.use_fp16,
        fp16_scale_growth=args.fp16_scale_growth,
        weight_decay=args.weight_decay,
        lr_anneal_steps=args.lr_anneal_steps,
    ).eval_loop() # ! overfitting


def create_argparser(**kwargs):
    # defaults.update(model_and_diffusion_defaults())

    defaults = dict(
        num_workers=4,
        local_rank=0,
        gpus=1,
        image_size=128,
        image_size_encoder=224,
        iterations=150000,
        anneal_lr=False,
        lr=5e-5,
        weight_decay=0.0,
        lr_anneal_steps=0,
        batch_size=1,
        microbatch=-1,  # -1 disables microbatches
        ema_rate="0.9999",  # comma-separated list of EMA values
        log_interval=10,
        save_interval=10000,
        resume_checkpoint="",
        use_fp16=False,
        fp16_scale_growth=1e-3,
        data_dir="",
        # load_depth=False, # TODO
        logdir="/mnt/lustre/yslan/logs/nips23/eval",
    )

    defaults.update(encoder_and_nsr_defaults())  # type: ignore
    defaults.update(loss_defaults())

    parser = argparse.ArgumentParser()
    add_dict_to_argparser(parser, defaults)

    return parser


if __name__ == "__main__":
    # os.environ[
    #     "TORCH_DISTRIBUTED_DEBUG"
    # ] = "DETAIL"  # set to DETAIL for runtime logging.
    # os.environ["TORCH_CPP_LOG_LEVEL"]="INFO"

    args = create_argparser().parse_args()
    # st()

    master_addr = '127.0.0.1'
    master_port = dist_util._find_free_port()

    # Launch processes.
    print('Launching processes...')
    th.multiprocessing.set_start_method('spawn')

    subprocess_fn = inference_loop

    # launch using torch.multiprocessing.spawn
    if args.gpus == 1:
        subprocess_fn(rank=0, master_addr=master_addr, master_port=master_port, args=args)
    else:
        th.multiprocessing.spawn(fn=subprocess_fn,
                                    args=(master_addr, master_port,args),
                                    nprocs=args.gpus)