engine.py

# Copyright (c) 2015-present, Facebook, Inc.
# All rights reserved.
"""
Train and eval functions used in main.py
"""
import math
import sys
from typing import Iterable, Optional

import torch

from timm.data import Mixup
from timm.utils import accuracy, ModelEma

from losses import DistillationLoss
import utils

from pruning_utils import * 

def get_tau(max_tau, min_tau, ite, total):
    tau = min_tau + (max_tau - min_tau) * ite / total 
    return tau 

def get_sparsity(s_start, s_end, t, t_0, pruning_times, delta_t):
    coef = (1-(t-t_0)/(pruning_times*delta_t))**3
    sparsity = s_end + (s_start - s_end)*coef 
    return sparsity


ite_step = 0
def train_one_epoch(model: torch.nn.Module, criterion: DistillationLoss,
                    data_loader: Iterable, optimizer: torch.optim.Optimizer,
                    device: torch.device, epoch: int, loss_scaler, max_norm: float = 0,
                    model_ema: Optional[ModelEma] = None, mixup_fn: Optional[Mixup] = None,
                    set_training_mode=True, args=None):
    model.train(set_training_mode)
    metric_logger = utils.MetricLogger(delimiter="  ")
    metric_logger.add_meter('lr', utils.SmoothedValue(window_size=1, fmt='{value:.6f}'))
    header = 'Epoch: [{}]'.format(epoch)
    print_freq = 10
    
    total_iteration = len(data_loader) * (args.epochs)
    for samples, targets in metric_logger.log_every(data_loader, print_freq, header):
        samples = samples.to(device, non_blocking=True)
        targets = targets.to(device, non_blocking=True)

        global ite_step
        if mixup_fn is not None:
            samples, targets = mixup_fn(samples, targets)
        
        if args.token_selection:
            tau = get_tau(10, 0.1, ite_step, total_iteration)
        else:
            tau = -1 

        with torch.cuda.amp.autocast():
            try:
                outputs = model(samples, tau=tau, number=args.token_number)
            except:
                outputs = model(samples)
            loss = criterion(samples, outputs, targets)

        loss_value = loss.item()

        if not math.isfinite(loss_value):
            print("Loss is {}, stopping training".format(loss_value))
            sys.exit(1)

        optimizer.zero_grad()
        ite_step += 1

        # this attribute is added by timm on one optimizer (adahessian)
        is_second_order = hasattr(optimizer, 'is_second_order') and optimizer.is_second_order
        loss_scaler(loss, optimizer, clip_grad=max_norm,
                    parameters=model.parameters(), create_graph=is_second_order)

        torch.cuda.synchronize()
        if model_ema is not None:
            model_ema.update(model)

        metric_logger.update(loss=loss_value)
        metric_logger.update(lr=optimizer.param_groups[0]["lr"])
    # gather the stats from all processes
    metric_logger.synchronize_between_processes()
    print("Averaged stats:", metric_logger)
    print("iteration step = {}".format(ite_step))
    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}

ite_step = 0
state = 0
def train_one_epoch_GMP(model: torch.nn.Module, criterion: DistillationLoss,
                    data_loader: Iterable, optimizer: torch.optim.Optimizer,
                    device: torch.device, epoch: int, loss_scaler, max_norm: float = 0,
                    model_ema: Optional[ModelEma] = None, mixup_fn: Optional[Mixup] = None,
                    set_training_mode=True, args=None):
    model.train(set_training_mode)
    metric_logger = utils.MetricLogger(delimiter="  ")
    metric_logger.add_meter('lr', utils.SmoothedValue(window_size=1, fmt='{value:.6f}'))
    header = 'Epoch: [{}]'.format(epoch)
    print_freq = 10

    for samples, targets in metric_logger.log_every(data_loader, print_freq, header):
        samples = samples.to(device, non_blocking=True)
        targets = targets.to(device, non_blocking=True)

        if mixup_fn is not None:
            samples, targets = mixup_fn(samples, targets)

        with torch.cuda.amp.autocast():
            outputs = model(samples)
            loss = criterion(samples, outputs, targets)

        loss_value = loss.item()

        if not math.isfinite(loss_value):
            print("Loss is {}, stopping training".format(loss_value))
            sys.exit(1)

        optimizer.zero_grad()

        # this attribute is added by timm on one optimizer (adahessian)
        is_second_order = hasattr(optimizer, 'is_second_order') and optimizer.is_second_order
        loss_scaler(loss, optimizer, clip_grad=max_norm,
                    parameters=model.parameters(), create_graph=is_second_order)

        torch.cuda.synchronize()
        if model_ema is not None:
            model_ema.update(model)

        metric_logger.update(loss=loss_value)
        metric_logger.update(lr=optimizer.param_groups[0]["lr"])

        # Gradual pruning
        global ite_step, state
        ite_step += 1
        if ite_step>args.t_start and (ite_step-args.t_start) % args.delta_t == 0 and state < args.pruning_times:
            print('start pruning for the {} times'.format(state+1))
            sparsity = get_sparsity(s_start=0, s_end=args.sparsity, t=ite_step, t_0=args.t_start, 
                                    pruning_times=args.pruning_times, delta_t=args.delta_t)
            pruner = Mag(masked_parameters(model))
            pruner.score(model, loss=None, dataloader=None, device=None)
            pruner.mask(1-sparsity, 'global')
            check_sparsity_dict(model.state_dict())
            state+=1

    # gather the stats from all processes
    metric_logger.synchronize_between_processes()
    print("Averaged stats:", metric_logger)
    print("iteration step = {}".format(ite_step))
    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}


@torch.no_grad()
def evaluate(data_loader, model, device):
    criterion = torch.nn.CrossEntropyLoss()

    metric_logger = utils.MetricLogger(delimiter="  ")
    header = 'Test:'

    # switch to evaluation mode
    model.eval()

    for images, target in metric_logger.log_every(data_loader, 10, header):
        images = images.to(device, non_blocking=True)
        target = target.to(device, non_blocking=True)

        # compute output
        with torch.cuda.amp.autocast():
            output = model(images)
            loss = criterion(output, target)

        acc1, acc5 = accuracy(output, target, topk=(1, 5))

        batch_size = images.shape[0]
        metric_logger.update(loss=loss.item())
        metric_logger.meters['acc1'].update(acc1.item(), n=batch_size)
        metric_logger.meters['acc5'].update(acc5.item(), n=batch_size)
    # gather the stats from all processes
    metric_logger.synchronize_between_processes()
    print('* Acc@1 {top1.global_avg:.3f} Acc@5 {top5.global_avg:.3f} loss {losses.global_avg:.3f}'
            .format(top1=metric_logger.acc1, top5=metric_logger.acc5, losses=metric_logger.loss))

    return {k: meter.global_avg for k, meter in metric_logger.meters.items()}