import warnings warnings.filterwarnings("ignore", category=UserWarning, module="torch.nn.functional") from comet_ml import Experiment from fastai.vision import * import torch import argparse import os from image_classification.arguments import get_args from image_classification.datasets.dataset import get_dataset from image_classification.utils.utils import * from image_classification.models.custom_resnet import * from trainer import * args = get_args(description='Flow of Solution Procedure KD', mode='train') expt = 'fsp-kd' torch.manual_seed(args.seed) if args.gpu != 'cpu': args.gpu = int(args.gpu) torch.cuda.set_device(args.gpu) torch.cuda.manual_seed(args.seed) hyper_params = { "dataset": args.dataset, "model": args.model, "stage": 0, "num_classes": 10, "batch_size": 64, "num_epochs": args.epoch,
import warnings warnings.filterwarnings("ignore", category=UserWarning, module="torch.nn.functional") from comet_ml import Experiment from fastai.vision import * import torch import argparse import os from image_classification.arguments import get_args from image_classification.datasets.dataset import get_dataset from image_classification.utils.utils import * from image_classification.models.custom_resnet import * from trainer import * args = get_args(description='Simultaneous KD', mode='train') expt = 'simultaneous-kd' torch.manual_seed(args.seed) if args.gpu != 'cpu': args.gpu = int(args.gpu) torch.cuda.set_device(args.gpu) torch.cuda.manual_seed(args.seed) hyper_params = { "dataset": args.dataset, "model": args.model, "stage": 0, "num_classes": 10, "batch_size": 64, "num_epochs": args.epoch,
warnings.filterwarnings("ignore", category=UserWarning, module="torch.nn.functional") from comet_ml import Experiment from fastai.vision import * import torch import argparse import os from image_classification.arguments import get_args from image_classification.datasets.dataset import get_dataset from image_classification.utils.utils import * from image_classification.models.custom_resnet import * from trainer import * args = get_args(description='Attention Transfer KD', mode='train') expt = 'attention-kd' torch.manual_seed(args.seed) if args.gpu != 'cpu': args.gpu = int(args.gpu) torch.cuda.set_device(args.gpu) torch.cuda.manual_seed(args.seed) hyper_params = { "dataset": args.dataset, "model": args.model, "stage": 0, "num_classes": 10, "batch_size": 64, "num_epochs": args.epoch,
import warnings warnings.filterwarnings("ignore", category=UserWarning, module="torch.nn.functional") from comet_ml import Experiment from fastai.vision import * import torch import argparse import os from image_classification.arguments import get_args from image_classification.datasets.dataset import get_dataset from image_classification.utils.utils import * from image_classification.models.custom_resnet import * from trainer import * args = get_args(description='No Teacher', mode='train') expt = 'no-teacher' torch.manual_seed(args.seed) if args.gpu != 'cpu': args.gpu = int(args.gpu) torch.cuda.set_device(args.gpu) torch.cuda.manual_seed(args.seed) hyper_params = { "dataset": args.dataset, "model": args.model, "stage": 0, "num_classes": 10, "batch_size": 64, "num_epochs": args.epoch,
import warnings warnings.filterwarnings("ignore", category=UserWarning, module="torch.nn.functional") from comet_ml import Experiment from fastai.vision import * import torch import argparse import os from image_classification.arguments import get_args from image_classification.datasets.dataset import get_dataset from image_classification.utils.utils import * from image_classification.models.custom_resnet import * from trainer import * args = get_args(description='Hinton KD', mode='train') expt = 'hinton-kd' torch.manual_seed(args.seed) if args.gpu != 'cpu': args.gpu = int(args.gpu) torch.cuda.set_device(args.gpu) torch.cuda.manual_seed(args.seed) hyper_params = { "dataset": args.dataset, "model": args.model, "num_classes": 10, "batch_size": 64, "num_epochs": args.epoch, "learning_rate": 1e-4,
import os from pathlib import Path from fastai.vision import * import shutil import random import argparse import sys from image_classification.arguments import get_args # parser = argparse.ArgumentParser(description = 'Creating dataset for less data') # parser.add_argument('-d', choices = ['imagenette', 'imagewoof', 'cifar10'], help = 'Give the dataset name from the choices') # parser.add_argument('-p', type = int, help = 'Give percentage of dataset') # args = parser.parse_args() args = get_args(description='Creating dataset for less data experiments', mode='data') random.seed(args.seed) if args.dataset == 'imagenette' or args.dataset == 'imagewoof': NUM_ = int(args.percentage * 13) if args.dataset == 'imagenette': data = untar_data( 'https://s3.amazonaws.com/fast-ai-imageclas/imagenette') else: data = untar_data( 'https://s3.amazonaws.com/fast-ai-imageclas/imagewoof') elif args.dataset == 'cifar10': NUM_ = int(args.percentage * 50) data = untar_data(URLs.CIFAR) else: print('Give dataset from choices only')
import warnings warnings.filterwarnings("ignore", category=UserWarning, module="torch.nn.functional") from comet_ml import Experiment from fastai.vision import * import torch import argparse import os from image_classification.arguments import get_args from image_classification.datasets.dataset import get_dataset from image_classification.utils.utils import * from image_classification.models.custom_resnet import * from trainer import * args = get_args(description='Traditional KD', mode='train') expt = 'traditional-kd' torch.manual_seed(args.seed) if args.gpu != 'cpu': args.gpu = int(args.gpu) torch.cuda.set_device(args.gpu) torch.cuda.manual_seed(args.seed) hyper_params = { "dataset": args.dataset, "model": args.model, "stage": 0, "num_classes": 10, "batch_size": 64, "num_epochs": args.epoch, "learning_rate": 1e-4,
import warnings warnings.filterwarnings("ignore", category=UserWarning, module="torch.nn.functional") from comet_ml import Experiment from fastai.vision import * import torch import argparse import os from image_classification.arguments import get_args from image_classification.datasets.dataset import get_dataset from image_classification.utils.utils import * from image_classification.models.custom_resnet import * args = get_args(description='Evaluation Script', mode='eval') torch.manual_seed(args.seed) if args.gpu != 'cpu': torch.cuda.set_device(args.gpu) torch.cuda.manual_seed(args.seed) hyper_params = { "dataset": args.dataset, "model": args.model, "stage": 0, "num_classes": 10, "batch_size": 64, "learning_rate": 1e-4, "seed": args.seed, "percentage": args.percentage, "gpu": args.gpu,
import warnings warnings.filterwarnings("ignore", category=UserWarning, module="torch.nn.functional") from comet_ml import Experiment from fastai.vision import * import torch import argparse import os from image_classification.arguments import get_args from image_classification.datasets.dataset import get_dataset from image_classification.utils.utils import * from image_classification.models.custom_resnet import * from trainer import * args = get_args(description='Stagewise KD', mode='train') expt = 'stagewise-kd' torch.manual_seed(args.seed) if args.gpu != 'cpu': args.gpu = int(args.gpu) torch.cuda.set_device(args.gpu) torch.cuda.manual_seed(args.seed) hyper_params = { "dataset": args.dataset, "model": args.model, "stage": 0, "num_classes": 10, "batch_size": 64, "num_epochs": args.epoch,