deep-learning-with-pytorch/p2ch11/training.py

import argparse
import datetime
import os
import sys

import numpy as np

from torch.utils.tensorboard import SummaryWriter

import torch
import torch.nn as nn
from torch.optim import SGD, Adan
from torch.utils.data import DataLoader

from util.util import enumerateWithEstimate
from .dsets import LunaDataset
from util.logconf import logging
from .model import LunaModel

log = logging.getLogger(__name__)
# log.setLevel(logging.WARN)
log.setLevel(logging.INFO)
# log.setLevel(logging.DEBUG)

# Used for computeBatchLoss and logMetrics to index into metrics_t/metrics_a
METRICS_LABEL_NDX=0
METRICS_PRED_NDX=1
METRICS_LOSS_NDX=2
METRICS_SIZE = 3

class LunaTrainingApp(object):
    def __init__(self, sys_argv=None):
        if sys_argv is None:
            sys_argv = sys.argv[1:]

        parser = argparse.ArgumentParser()
        parser.add_argument('--num-workers',
            help='Number of worker processes for background data loading',
            default=8,
            type=int,
        )
        parser.add_argument('--batch-size',
            help='Batch size to use for training',
            default=32,
            type=int,
        )
        parser.add_argument('--epochs',
            help='Number of epochs to train for',
            default=1,
            type=int,
        )

        parser.add_argument('--tb-prefix',
            default='p2ch11',
            help="Data prefix to use for Tensorboard run. Defaults to chapter.",
        )

        parser.add_argument('comment',
            help="Comment suffix for Tensorboard run.",
            nargs='?',
            default='dwlpt',
        )
        self.cli_args = parser.parse_args(sys_argv)
        self.time_str = datetime.datetime.now().strftime('%Y-%m-%d_%H.%M.%S')

        self.trn_writer = None
        self.val_writer = None
        self.totalTrainingSamples_count = 0

        self.use_cuda = torch.cuda.is_available()
        self.device = torch.device("cuda" if self.use_cuda else "cpu")

        self.model = self.initModel()
        self.optimizer = self.initOptimizer()

    def initModel(self):
        model = LunaModel()
        if self.use_cuda:
            log.info("Using CUDA with {} devices.".format(torch.cuda.device_count()))
            if torch.cuda.device_count() > 1:
                model = nn.DataParallel(model)
            model = model.to(self.device)
        return model

    def initOptimizer(self):
        return SGD(self.model.parameters(), lr=0.001, momentum=0.99)
        # return Adam(self.model.parameters())

    def initTrainDl(self):
        train_ds = LunaDataset(
            val_stride=10,
            isValSet_bool=False,
        )

        train_dl = DataLoader(
            train_ds,
            batch_size=self.cli_args.batch_size * (torch.cuda.device_count() if self.use_cuda else 1),
            num_workers=self.cli_args.num_workers,
            pin_memory=self.use_cuda,
        )

        return train_dl

    def initValDl(self):
        val_ds = LunaDataset(
            val_stride=10,
            isValSet_bool=True,
        )

        val_dl = DataLoader(
            val_ds,
            batch_size=self.cli_args.batch_size * (torch.cuda.device_count() if self.use_cuda else 1),
            num_workers=self.cli_args.num_workers,
            pin_memory=self.use_cuda,
        )

        return val_dl

    def initTensorboardWriters(self):
        if self.trn_writer is None:
            log_dir = os.path.join('runs', self.cli_args.tb_prefix, self.time_str)

            self.trn_writer = SummaryWriter(log_dir=log_dir + '-trn_cls-' + self.cli_args.comment)
            self.val_writer = SummaryWriter(log_dir=log_dir + '-val_cls-' + self.cli_args.comment)


    def main(self):
        log.info("Starting {}, {}".format(type(self).__name__, self.cli_args))

        train_dl = self.initTrainDl()
        val_dl = self.initValDl()

        self.initTensorboardWriters()

        for epoch_ndx in range(1, self.cli_args.epochs + 1):

            log.info("Epoch {} of {}, {}/{} batches of size {}*{}".format(
                epoch_ndx,
                self.cli_args.epochs,
                len(train_dl),
                len(val_dl),
                self.cli_args.batch_size,
                (torch.cuda.device_count() if self.use_cuda else 1),
            ))

            trnMetrics_t = self.doTraining(epoch_ndx, train_dl)
            self.logMetrics(epoch_ndx, 'trn', trnMetrics_t)

            valMetrics_t = self.doValidation(epoch_ndx, val_dl)
            self.logMetrics(epoch_ndx, 'val', valMetrics_t)

        if hasattr(self, 'trn_writer'):
            self.trn_writer.close()
            self.val_writer.close()


    def doTraining(self, epoch_ndx, train_dl):
        self.model.train()
        trnMetrics_g = torch.zeros(METRICS_SIZE, len(train_dl.dataset)).to(self.device)
        batch_iter = enumerateWithEstimate(
            train_dl,
            "E{} Training".format(epoch_ndx),
            start_ndx=train_dl.num_workers,
        )
        for batch_ndx, batch_tup in batch_iter:
            self.optimizer.zero_grad()

            loss_var = self.computeBatchLoss(
                batch_ndx,
                batch_tup,
                train_dl.batch_size,
                trnMetrics_g
            )

            loss_var.backward()
            self.optimizer.step()
            del loss_var

            # # This is for adding the model graph to TensorBoard.
            # if epoch_ndx == 1 and batch_ndx == 0:
            #     with torch.no_grad():
            #         model = LunaModel()
            #         self.trn_writer.add_graph(model, batch_tup[0], verbose=True)
            #         self.trn_writer.close()

        self.totalTrainingSamples_count += len(train_dl.dataset)

        return trnMetrics_g.to('cpu')


    def doValidation(self, epoch_ndx, val_dl):
        with torch.no_grad():
            self.model.eval()
            valMetrics_g = torch.zeros(METRICS_SIZE, len(val_dl.dataset)).to(self.device)
            batch_iter = enumerateWithEstimate(
                val_dl,
                "E{} Validation ".format(epoch_ndx),
                start_ndx=val_dl.num_workers,
            )
            for batch_ndx, batch_tup in batch_iter:
                self.computeBatchLoss(batch_ndx, batch_tup, val_dl.batch_size, valMetrics_g)

        return valMetrics_g.to('cpu')



    def computeBatchLoss(self, batch_ndx, batch_tup, batch_size, metrics_g):
        input_t, label_t, _series_list, _center_list = batch_tup

        input_g = input_t.to(self.device, non_blocking=True)
        label_g = label_t.to(self.device, non_blocking=True)

        logits_g, probability_g = self.model(input_g)

        loss_func = nn.CrossEntropyLoss(reduction='none')
        loss_g = loss_func(
            logits_g,
            label_g[:,1],
        )
        start_ndx = batch_ndx * batch_size
        end_ndx = start_ndx + label_t.size(0)

        metrics_g[METRICS_LABEL_NDX, start_ndx:end_ndx] = label_g[:,1]
        metrics_g[METRICS_PRED_NDX, start_ndx:end_ndx] = probability_g[:,1]
        metrics_g[METRICS_LOSS_NDX, start_ndx:end_ndx] = loss_g

        return loss_g.mean()


    def logMetrics(
            self,
            epoch_ndx,
            mode_str,
            metrics_t,
    ):
        self.initTensorboardWriters()
        log.info("E{} {}".format(
            epoch_ndx,
            type(self).__name__,
        ))

        benLabel_mask = metrics_t[METRICS_LABEL_NDX] <= 0.5
        benPred_mask = metrics_t[METRICS_PRED_NDX] <= 0.5

        malLabel_mask = ~benLabel_mask
        malPred_mask = ~benPred_mask

        ben_count = benLabel_mask.sum()
        mal_count = malLabel_mask.sum()

        ben_correct = (benLabel_mask & benPred_mask).sum()
        mal_correct = (malLabel_mask & malPred_mask).sum()

        # trueNeg_count = ben_correct = (benLabel_mask & benPred_mask).sum()
        # truePos_count = mal_correct = (malLabel_mask & malPred_mask).sum()
        #
        # falsePos_count = ben_count - ben_correct
        # falseNeg_count = malLabel_count - mal_correct

        # log.info(['min loss', metrics_a[METRICS_LOSS_NDX, benLabel_mask].min(), metrics_a[METRICS_LOSS_NDX, malLabel_mask].min()])
        # log.info(['max loss', metrics_a[METRICS_LOSS_NDX, benLabel_mask].max(), metrics_a[METRICS_LOSS_NDX, malLabel_mask].max()])


        metrics_dict = {}
        metrics_dict['loss/all'] = metrics_t[METRICS_LOSS_NDX].mean()
        metrics_dict['loss/ben'] = metrics_t[METRICS_LOSS_NDX, benLabel_mask].mean()
        metrics_dict['loss/mal'] = metrics_t[METRICS_LOSS_NDX, malLabel_mask].mean()

        metrics_dict['correct/all'] = (mal_correct + ben_correct) / metrics_t.shape[1] * 100
        metrics_dict['correct/ben'] = (ben_correct) / ben_count * 100
        metrics_dict['correct/mal'] = (mal_correct) / mal_count * 100

        log.info(
            ("E{} {:8} {loss/all:.4f} loss, "
                 + "{correct/all:-5.1f}% correct, "
            ).format(
                epoch_ndx,
                mode_str,
                **metrics_dict,
            )
        )
        log.info(
            ("E{} {:8} {loss/ben:.4f} loss, "
                 + "{correct/ben:-5.1f}% correct ({ben_correct:} of {ben_count:})"
            ).format(
                epoch_ndx,
                mode_str + '_ben',
                ben_correct=ben_correct,
                ben_count=ben_count,
                **metrics_dict,
            )
        )
        log.info(
            ("E{} {:8} {loss/mal:.4f} loss, "
                 + "{correct/mal:-5.1f}% correct ({mal_correct:} of {mal_count:})"
            ).format(
                epoch_ndx,
                mode_str + '_mal',
                mal_correct=mal_correct,
                mal_count=mal_count,
                **metrics_dict,
            )
        )

        writer = getattr(self, mode_str + '_writer')

        for key, value in metrics_dict.items():
            writer.add_scalar(key, value, self.totalTrainingSamples_count)

        writer.add_pr_curve(
            'pr',
            metrics_t[METRICS_LABEL_NDX],
            metrics_t[METRICS_PRED_NDX],
            self.totalTrainingSamples_count,
        )

        bins = [x/50.0 for x in range(51)]

        benHist_mask = benLabel_mask & (metrics_t[METRICS_PRED_NDX] > 0.01)
        malHist_mask = malLabel_mask & (metrics_t[METRICS_PRED_NDX] < 0.99)

        if benHist_mask.any():
            writer.add_histogram(
                'is_ben',
                metrics_t[METRICS_PRED_NDX, benHist_mask],
                self.totalTrainingSamples_count,
                bins=bins,
            )
        if malHist_mask.any():
            writer.add_histogram(
                'is_mal',
                metrics_t[METRICS_PRED_NDX, malHist_mask],
                self.totalTrainingSamples_count,
                bins=bins,
            )

        # score = 1 \
        #     + metrics_dict['pr/f1_score'] \
        #     - metrics_dict['loss/mal'] * 0.01 \
        #     - metrics_dict['loss/all'] * 0.0001
        #
        # return score

    # def logModelMetrics(self, model):
    #     writer = getattr(self, 'trn_writer')
    #
    #     model = getattr(model, 'module', model)
    #
    #     for name, param in model.named_parameters():
    #         if param.requires_grad:
    #             min_data = float(param.data.min())
    #             max_data = float(param.data.max())
    #             max_extent = max(abs(min_data), abs(max_data))
    #
    #             # bins = [x/50*max_extent for x in range(-50, 51)]
    #
    #             try:
    #                 writer.add_histogram(
    #                     name.rsplit('.', 1)[-1] + '/' + name,
    #                     param.data.cpu().numpy(),
    #                     # metrics_a[METRICS_PRED_NDX, benHist_mask],
    #                     self.totalTrainingSamples_count,
    #                     # bins=bins,
    #                 )
    #             except Exception as e:
    #                 log.error([min_data, max_data])
    #                 raise


if __name__ == '__main__':
    LunaTrainingApp().main()