def cnn_process():
use_model = args.cnn_use_model
util.topic_log(use_model)
if use_model == 'resnet50':
pretrain_model = torch_models.resnet50(pretrained=True)
elif use_model == 'resnet101':
pretrain_model = torch_models.resnet101(pretrained=True)
elif use_model == 'resnet152':
pretrain_model = torch_models.resnet152(pretrained=True)
elif use_model == 'alexnet':
pretrain_model = torch_models.alexnet(pretrained=True)
else:
raise Exception
model = models.FineTuneModel(
pretrain_model,
'resnet' if 'resnet' in use_model else use_model).to(args.device)
optimizer = optim.Adam(filter(lambda p: p.requires_grad,
model.parameters()),
lr=args.lr,
weight_decay=0.0001)
best_top1 = 0
if args.load_model:
best_top1 = util.load_model(model, optimizer, args,
args.save_model_path)
print(f'load_model: {args.load_model} ({best_top1})')
util.model_fit(model, optimizer, args, train_loader, test_loader,
best_top1)
_, predicts = util.test_epoch(model, args, test_loader, get_predicts=True)
util.evaluate_log(predicts, test_labels)
def encoding_process():
util.topic_log("Accuracy before huffman encoding")
util.print_nonzeros(model, args.log_file_path)
top1_acc, top5_acc = util.val_epoch(val_loader, model, args, topk=(1, 5))
util.log(args.log_file_path,
f"accuracy before huffman encoding\t{top1_acc} ({top5_acc})")
util.topic_log("encoding")
mesa2_huffman_encode_model(model, args)
def part1():
util.topic_log('part1', 80)
first_image_path, _ = train_set.image_paths_with_labels[0]
image_ds = dcmread(first_image_path)
ori_image_data = image_ds.pixel_array.astype(np.float32)
trans_image_data = train_set.transform(ori_image_data /
np.max(ori_image_data)).numpy()
print(f'csv-field: {train_set.csv_df.columns}')
print(
f'original image: size ({ori_image_data.shape}), min ({np.min(ori_image_data)}), max ({np.max(ori_image_data)}), mean ({np.mean(ori_image_data)})'
)
print(
f'transform image: size ({trans_image_data.shape}), min ({np.min(trans_image_data)}), max ({np.max(trans_image_data)}), mean ({np.mean(trans_image_data)})'
)
def pruning_process():
util.topic_log("Before pruning")
top1_acc, top5_acc = util.val_epoch(val_loader, model, args, topk=(1, 5))
util.log(args.log_file_path,
f"before pruning accuracy\t{top1_acc} ({top5_acc})")
util.topic_log("Pruning CNN")
model.prune(args)
util.topic_log("After prune CNN")
util.print_nonzeros(model, args.log_file_path)
top1_acc, top5_acc = util.val_epoch(val_loader, model, args, topk=(1, 5))
util.log(args.log_file_path,
f"after pruning accuracy\t{top1_acc} ({top5_acc})")
util.topic_log("Start retrain after prune CNN")
util.train(model, args, train_loader, val_loader, 'prune_retrain')
util.topic_log("After retraining")
top1_acc, top5_acc = util.val_epoch(val_loader, model, args, topk=(1, 5))
util.log(args.log_file_path,
f"after pruning and retrain accuracy\t{top1_acc} ({top5_acc})")
def quantize_process():
util.topic_log("Accuracy before weight sharing")
top1_acc, top5_acc = util.val_epoch(val_loader, model, args, topk=(1, 5))
util.log(args.log_file_path,
f"accuracy before weight sharing\t{top1_acc} ({top5_acc})")
util.topic_log("Accuracy after weight sharing")
layer_mame_to_quan_indices = apply_weight_sharing(model, args)
top1_acc, top5_acc = util.val_epoch(val_loader, model, args, topk=(1, 5))
util.save_masked_checkpoint(model, "quantized", top1_acc, "initial", args)
util.log(
args.log_file_path,
f"accuracy after weight sharing {args.bits}bits\t{top1_acc} ({top5_acc})"
)
util.topic_log("Quantize retraining")
util.quantized_retrain(model, args, layer_mame_to_quan_indices,
train_loader, val_loader)
top1_acc, top5_acc = util.val_epoch(val_loader, model, args, topk=(1, 5))
util.save_masked_checkpoint(model, "quantized", top1_acc, "end", args)
util.log(args.log_file_path,
f"accuracy after qauntize and retrain\t{top1_acc} ({top5_acc})")
def initial_process():
util.topic_log("Initial training")
util.train(model, args, train_loader, val_loader, 'initial_train')
top1_acc, top5_acc = util.val_epoch(val_loader, model, args, topk=(1, 5))
util.log(args.log_file_path, f"initial_accuracy\t{top1_acc} ({top5_acc})")
def part2():
util.topic_log('part2', 80)
svm_process()
kmeans_process()
random_foreset_process()
cnn_process()
def random_foreset_process():
util.topic_log('Random Forest')
model = RandomForestClassifier()
model.fit(train_data2d, train_labels)
predicts = model.predict(test_data2d)
util.evaluate_log(predicts, test_labels)
def kmeans_process():
util.topic_log('Kmeans')
model = KMeans(n_clusters=2, random_state=0)
model.fit(train_data2d)
predicts = model.predict(test_data2d)
util.evaluate_log(predicts, test_labels)
def svm_process():
util.topic_log('SVM')
model = svm.SVC()
model.fit(train_data2d, train_labels)
predicts = model.predict(test_data2d)
util.evaluate_log(predicts, test_labels)
