train.py

import logging
import math
import random
from datetime import timedelta
from pathlib import Path

import hydra
import numpy as np
import torch
import torch.distributed as distributed
from hydra import compose
from hydra.core.hydra_config import HydraConfig
from omegaconf import DictConfig, open_dict
from torch.distributed.elastic.multiprocessing.errors import record

from mmaudio.data.data_setup import setup_training_datasets, setup_val_datasets
from mmaudio.model.sequence_config import CONFIG_16K, CONFIG_44K
from mmaudio.runner import Runner
from mmaudio.sample import sample
from mmaudio.utils.dist_utils import info_if_rank_zero, local_rank, world_size
from mmaudio.utils.logger import TensorboardLogger
from mmaudio.utils.synthesize_ema import synthesize_ema

torch.backends.cuda.matmul.allow_tf32 = True
torch.backends.cudnn.allow_tf32 = True

log = logging.getLogger()


def distributed_setup():
    distributed.init_process_group(backend="nccl", timeout=timedelta(hours=2))
    log.info(f'Initialized: local_rank={local_rank}, world_size={world_size}')
    return local_rank, world_size


@record
@hydra.main(version_base='1.3.2', config_path='config', config_name='train_config.yaml')
def train(cfg: DictConfig):
    # initial setup
    torch.cuda.set_device(local_rank)
    torch.backends.cudnn.benchmark = cfg.cudnn_benchmark
    distributed_setup()
    num_gpus = world_size
    run_dir = HydraConfig.get().run.dir

    # compose early such that it does not rely on future hard disk reading
    eval_cfg = compose('eval_config', overrides=[f'exp_id={cfg.exp_id}'])

    # patch data dim
    if cfg.model.endswith('16k'):
        seq_cfg = CONFIG_16K
    elif cfg.model.endswith('44k'):
        seq_cfg = CONFIG_44K
    else:
        raise ValueError(f'Unknown model: {cfg.model}')
    with open_dict(cfg):
        cfg.data_dim.latent_seq_len = seq_cfg.latent_seq_len
        cfg.data_dim.clip_seq_len = seq_cfg.clip_seq_len
        cfg.data_dim.sync_seq_len = seq_cfg.sync_seq_len

    # wrap python logger with a tensorboard logger
    log = TensorboardLogger(cfg.exp_id,
                            run_dir,
                            logging.getLogger(),
                            is_rank0=(local_rank == 0),
                            enable_email=cfg.enable_email and not cfg.debug)

    info_if_rank_zero(log, f'All configuration: {cfg}')
    info_if_rank_zero(log, f'Number of GPUs detected: {num_gpus}')

    # number of dataloader workers
    info_if_rank_zero(log, f'Number of dataloader workers (per GPU): {cfg.num_workers}')

    # Set seeds to ensure the same initialization
    torch.manual_seed(cfg.seed)
    np.random.seed(cfg.seed)
    random.seed(cfg.seed)

    # setting up configurations
    info_if_rank_zero(log, f'Training configuration: {cfg}')
    cfg.batch_size //= num_gpus
    info_if_rank_zero(log, f'Batch size (per GPU): {cfg.batch_size}')

    # determine time to change max skip
    total_iterations = cfg['num_iterations']

    # setup datasets
    dataset, sampler, loader = setup_training_datasets(cfg)
    info_if_rank_zero(log, f'Number of training samples: {len(dataset)}')
    info_if_rank_zero(log, f'Number of training batches: {len(loader)}')

    val_dataset, val_loader, eval_loader = setup_val_datasets(cfg)
    info_if_rank_zero(log, f'Number of val samples: {len(val_dataset)}')
    val_cfg = cfg.data.ExtractedVGG_val

    # compute and set mean and std
    latent_mean, latent_std = dataset.compute_latent_stats()

    # construct the trainer
    trainer = Runner(cfg,
                     log=log,
                     run_path=run_dir,
                     for_training=True,
                     latent_mean=latent_mean,
                     latent_std=latent_std).enter_train()
    eval_rng_clone = trainer.rng.graphsafe_get_state()

    # load previous checkpoint if needed
    if cfg['checkpoint'] is not None:
        curr_iter = trainer.load_checkpoint(cfg['checkpoint'])
        cfg['checkpoint'] = None
        info_if_rank_zero(log, 'Model checkpoint loaded!')
    else:
        # if run_dir exists, load the latest checkpoint
        checkpoint = trainer.get_latest_checkpoint_path()
        if checkpoint is not None:
            curr_iter = trainer.load_checkpoint(checkpoint)
            info_if_rank_zero(log, 'Latest checkpoint loaded!')
        else:
            # load previous network weights if needed
            curr_iter = 0
            if cfg['weights'] is not None:
                info_if_rank_zero(log, 'Loading weights from the disk')
                trainer.load_weights(cfg['weights'])
                cfg['weights'] = None

    # determine max epoch
    total_epoch = math.ceil(total_iterations / len(loader))
    current_epoch = curr_iter // len(loader)
    info_if_rank_zero(log, f'We will approximately use {total_epoch} epochs.')

    # training loop
    try:
        # Need this to select random bases in different workers
        np.random.seed(np.random.randint(2**30 - 1) + local_rank * 1000)
        while curr_iter < total_iterations:
            # Crucial for randomness!
            sampler.set_epoch(current_epoch)
            current_epoch += 1
            log.debug(f'Current epoch: {current_epoch}')

            trainer.enter_train()
            trainer.log.data_timer.start()
            for data in loader:
                trainer.train_pass(data, curr_iter)

                if (curr_iter + 1) % cfg.val_interval == 0:
                    # swap into a eval rng state, i.e., use the same seed for every validation pass
                    train_rng_snapshot = trainer.rng.graphsafe_get_state()
                    trainer.rng.graphsafe_set_state(eval_rng_clone)
                    info_if_rank_zero(log, f'Iteration {curr_iter}: validating')
                    for data in val_loader:
                        trainer.validation_pass(data, curr_iter)
                    distributed.barrier()
                    trainer.val_integrator.finalize('val', curr_iter, ignore_timer=True)
                    trainer.rng.graphsafe_set_state(train_rng_snapshot)

                if (curr_iter + 1) % cfg.eval_interval == 0:
                    save_eval = (curr_iter + 1) % cfg.save_eval_interval == 0
                    train_rng_snapshot = trainer.rng.graphsafe_get_state()
                    trainer.rng.graphsafe_set_state(eval_rng_clone)
                    info_if_rank_zero(log, f'Iteration {curr_iter}: validating')
                    for data in eval_loader:
                        audio_path = trainer.inference_pass(data,
                                                            curr_iter,
                                                            val_cfg,
                                                            save_eval=save_eval)
                    distributed.barrier()
                    trainer.rng.graphsafe_set_state(train_rng_snapshot)
                    trainer.eval(audio_path, curr_iter, val_cfg)

                curr_iter += 1

                if curr_iter >= total_iterations:
                    break
    except Exception as e:
        log.error(f'Error occurred at iteration {curr_iter}!')
        log.critical(e.message if hasattr(e, 'message') else str(e))
        raise
    finally:
        if not cfg.debug:
            trainer.save_checkpoint(curr_iter)
            trainer.save_weights(curr_iter)

    # Inference pass
    del trainer
    torch.cuda.empty_cache()

    # Synthesize EMA
    if local_rank == 0:
        log.info(f'Synthesizing EMA with sigma={cfg.ema.default_output_sigma}')
        ema_sigma = cfg.ema.default_output_sigma
        state_dict = synthesize_ema(cfg, ema_sigma, step=None)
        save_dir = Path(run_dir) / f'{cfg.exp_id}_ema_final.pth'
        torch.save(state_dict, save_dir)
        log.info(f'Synthesized EMA saved to {save_dir}!')
    distributed.barrier()

    log.info(f'Evaluation: {eval_cfg}')
    sample(eval_cfg)

    # clean-up
    log.complete()
    distributed.barrier()
    distributed.destroy_process_group()


if __name__ == '__main__':
    train()