isaac/cv/defom-stereo/utils/utils.py

import logging
import sys

import torch
import torch.optim as optim
from torch.utils.tensorboard import SummaryWriter

import numpy as np
import random

def seed_everything(seed):
    torch.manual_seed(seed)       # Current CPU
    torch.cuda.manual_seed(seed)  # Current GPU
    np.random.seed(seed)          # Numpy module
    random.seed(seed)             # Python random module
    torch.backends.cudnn.benchmark = False    # Close optimization
    torch.backends.cudnn.deterministic = True # Close optimization
    torch.cuda.manual_seed_all(seed) # All GPU (Optional)


def sequence_loss(flow_preds, flow_gt, valid, loss_gamma=0.9, max_flow=700):
    """ Loss function defined over sequence of flow predictions """

    n_predictions = len(flow_preds)
    assert n_predictions >= 1
    flow_loss = 0.0

    # exlude invalid pixels and extremely large diplacements
    mag = torch.sum(flow_gt ** 2, dim=1, keepdim=True).sqrt()

    # exclude extremly large displacements
    valid = ((valid >= 0.5) & (mag < max_flow))
    assert valid.shape == flow_gt.shape, [valid.shape, flow_gt.shape]
    assert not torch.isinf(flow_gt[valid.bool()]).any()

    for i in range(n_predictions):
        assert not torch.isnan(flow_preds[i]).any() and not torch.isinf(flow_preds[i]).any()
        # We adjust the loss_gamma so it is consistent for any number of RAFT-Stereo iterations
        adjusted_loss_gamma = loss_gamma ** (15 / (n_predictions))
        i_weight = adjusted_loss_gamma ** (n_predictions - i)
        i_loss = (flow_preds[i] - flow_gt).abs()
        assert i_loss.shape == valid.shape, [i_loss.shape, valid.shape, flow_gt.shape, flow_preds[i].shape]
        flow_loss += i_weight * i_loss[valid.bool()].mean()

    epe = torch.sum((flow_preds[-1] - flow_gt) ** 2, dim=1).sqrt()
    epe = epe.view(-1)[valid.view(-1)]

    metrics = {
        'epe': epe.mean().item(),
        '1px': (epe < 1).float().mean().item(),
        '3px': (epe < 3).float().mean().item(),
        '5px': (epe < 5).float().mean().item(),
    }

    return flow_loss, metrics


def fetch_optimizer(args, model, last_epoch=-1, checkpoint=None):
    """ Create the optimizer and learning rate scheduler """
    trainable_params = filter(lambda p: p.requires_grad, model.parameters())
    optimizer = optim.AdamW(trainable_params, lr=args.lr, weight_decay=args.wdecay, eps=1e-8)
    if checkpoint is not None:
        optimizer.load_state_dict(checkpoint['optimizer'])

    scheduler = optim.lr_scheduler.OneCycleLR(optimizer, args.lr, args.num_steps + 100, pct_start=0.01,
                                              cycle_momentum=False, anneal_strategy='linear', last_epoch=last_epoch)

    return optimizer, scheduler


class Logger:
    SUM_FREQ = 100

    def __init__(self, model, scheduler, name):
        self.model = model
        self.scheduler = scheduler
        self.total_steps = 0
        self.running_loss = {}
        self.log_dir = 'runs/' + name
        self.writer = SummaryWriter(log_dir=self.log_dir)

    def _print_training_status(self):
        metrics_data = [self.running_loss[k] / Logger.SUM_FREQ for k in sorted(self.running_loss.keys())]
        training_str = "[{:6d}, {:10.7f}] ".format(self.total_steps + 1, self.scheduler.get_last_lr()[0])
        metrics_str = ("{:10.4f}, " * len(metrics_data)).format(*metrics_data)

        # print the training status
        logging.info(f"Training Metrics ({self.total_steps}): {training_str + metrics_str}")

        if self.writer is None:
            self.writer = SummaryWriter(log_dir=self.log_dir)

        for k in self.running_loss:
            self.writer.add_scalar("train/" + k, self.running_loss[k] / Logger.SUM_FREQ, self.total_steps)
            self.running_loss[k] = 0.0

    def push(self, metrics):
        self.total_steps += 1

        for key in metrics:
            if key not in self.running_loss:
                self.running_loss[key] = 0.0

            self.running_loss[key] += metrics[key]

        if self.total_steps % Logger.SUM_FREQ == Logger.SUM_FREQ - 1:
            self._print_training_status()
            self.running_loss = {}

    def write_dict(self, results):
        if self.writer is None:
            self.writer = SummaryWriter(log_dir=self.log_dir)

        for key in results:
            self.writer.add_scalar("valid/" + key, results[key], self.total_steps)

    def close(self):
        self.writer.close()