모델 정리2

Author: gellston

ppt/시대별 딥러닝 모델 특징.pptx

@@ -0,0 +1,122 @@
+-import torch
+import torch.nn as nn
+import random
+
+
+from ptflops import get_model_complexity_info
+from torchsummary import summary
+from torch.utils.data import DataLoader
+
+
+from model.CSPDenseNet import CSPDenseNet
+from util.FIATClassificationDataset import FIATClassificationDataset
+
+
+USE_CUDA = torch.cuda.is_available() # GPU를 사용가능하면 True, 아니라면 False를 리턴
+device = torch.device("cuda" if USE_CUDA else "cpu") # GPU 사용 가능하면 사용하고 아니면 CPU 사용
+print("다음 기기로 학습합니다:", device)
+
+
+# for reproducibility
+random.seed(777)
+torch.manual_seed(777)
+if device == 'cuda':
+    torch.cuda.manual_seed_all(777)
+
+
+## Hyper parameter
+training_epochs = 40
+batch_size = 5
+target_accuracy = 0.99
+learning_rate = 0.0003
+accuracy_threshold = 0.5
+## Hyper parameter
+
+
+model = CSPDenseNet(class_num=4,
+                    block_config=(6, 12, 24, 16),
+                    expansion_rate=4,                       ##Bottleneck exansion size
+                    growth_rate=12,
+                    activation=torch.nn.SiLU).to(device)
+
+
+
+"""
+model = DenseNet(class_num=4,
+                 num_init_features=12,
+                 block_config=(6, 6, 6),
+                 expansion_rate=4,                       ##Bottleneck exansion size
+                 growth_rate=40).to(device)
+"""
+print('==== model info ====')
+summary(model, (3, 224, 224))
+print('====================')
+
+macs, params = get_model_complexity_info(model,
+                                         (3, 224, 224),
+                                         as_strings=True,
+                                         print_per_layer_stat=True, verbose=True)
+print('{:<30}  {:<8}'.format('Computational complexity: ', macs))
+print('{:<30}  {:<8}'.format('Number of parameters: ', params))
+
+
+## no Train Model Save
+
+model.eval()
+compiled_model = torch.jit.script(model)
+torch.jit.save(compiled_model, "C://Github//DeepLearningStudy//trained_model//FIAT(CSPDenseNet).pt")
+
+trace_input = torch.rand(1, 3, 224, 224).to(device, dtype=torch.float32)
+trace_model = torch.jit.trace(model, trace_input)
+torch.jit.save(trace_model, "C://Github//DeepLearningStudy//trained_model//FIAT(CSPDenseNet)_Trace.pt")
+
+## no Train Model Save
+
+
+datasets = FIATClassificationDataset('C://Github//DeepLearningStudy//dataset//FIAT_dataset_food//',
+                                     label_height=224,
+                                     label_width=224,
+                                     isColor=True,
+                                     isNorm=False)
+data_loader = DataLoader(datasets, batch_size=batch_size, shuffle=True)
+
+
+model.train()
+criterion = nn.BCELoss()
+optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)
+
+
+for epoch in range(training_epochs): # 앞서 training_epochs의 값은 15로 지정함.
+    avg_cost = 0
+    avg_acc = 0
+    total_batch = len(data_loader)
+
+    for X, Y in data_loader:
+        gpu_X = X.to(device)
+        gpu_Y = Y.to(device)
+
+        model.train()
+        optimizer.zero_grad()
+        hypothesis = model(gpu_X)
+        cost = criterion(hypothesis, gpu_Y)
+        cost.backward()
+        avg_cost += (cost / total_batch)
+        optimizer.step()
+
+        model.eval()
+        prediction = model(gpu_X)
+        correct_prediction = torch.argmax(prediction, 1) == torch.argmax(gpu_Y, 1)
+        accuracy = correct_prediction.float().mean()
+        avg_acc += (accuracy / total_batch)
+
+    print('Epoch:', '%04d' % (epoch + 1), 'cost =', '{:.9f}'.format(avg_cost), 'acc =', '{:.9f}'.format(avg_acc))
+    if avg_acc > target_accuracy:
+        break;
+
+## no Train Model Save
+model.eval()
+compiled_model = torch.jit.script(model)
+torch.jit.save(compiled_model, "C://Github//DeepLearningStudy//trained_model//TRAIN_FIAT(CSPDenseNet).pt")
+## no Train Model Save
+
+print('Learning finished')