697 lines
103 KiB
Plaintext
697 lines
103 KiB
Plaintext
{
|
||
"cells": [
|
||
{
|
||
"metadata": {
|
||
"ExecuteTime": {
|
||
"end_time": "2025-06-24T17:37:28.323528Z",
|
||
"start_time": "2025-06-24T17:37:28.321407Z"
|
||
}
|
||
},
|
||
"cell_type": "code",
|
||
"source": "# Modify.py\n",
|
||
"id": "98da35e8f6af6b7a",
|
||
"outputs": [],
|
||
"execution_count": 9
|
||
},
|
||
{
|
||
"metadata": {
|
||
"ExecuteTime": {
|
||
"end_time": "2025-06-24T17:37:28.355899Z",
|
||
"start_time": "2025-06-24T17:37:28.352650Z"
|
||
}
|
||
},
|
||
"cell_type": "code",
|
||
"source": [
|
||
"import os\n",
|
||
"import random\n",
|
||
"\n",
|
||
"import numpy as np\n",
|
||
"import pandas as pd\n",
|
||
"import deepquantum as dq\n",
|
||
"import matplotlib.pyplot as plt\n",
|
||
"import torch\n",
|
||
"import torch.nn as nn\n",
|
||
"import torch.optim as optim\n",
|
||
"import torchvision.transforms as transforms\n",
|
||
"from torchvision.datasets import FashionMNIST\n",
|
||
"from tqdm import tqdm\n",
|
||
"from torch.utils.data import DataLoader\n",
|
||
"from multiprocessing import freeze_support\n"
|
||
],
|
||
"id": "fba02718b5ce470f",
|
||
"outputs": [],
|
||
"execution_count": 10
|
||
},
|
||
{
|
||
"metadata": {
|
||
"ExecuteTime": {
|
||
"end_time": "2025-06-24T17:37:28.376643Z",
|
||
"start_time": "2025-06-24T17:37:28.373463Z"
|
||
}
|
||
},
|
||
"cell_type": "code",
|
||
"source": [
|
||
"def seed_torch(seed=1024):\n",
|
||
" random.seed(seed)\n",
|
||
" os.environ['PYTHONHASHSEED'] = str(seed)\n",
|
||
" np.random.seed(seed)\n",
|
||
" torch.manual_seed(seed)\n",
|
||
" torch.cuda.manual_seed(seed)\n",
|
||
" torch.cuda.manual_seed_all(seed)\n",
|
||
" torch.backends.cudnn.benchmark = False\n",
|
||
" torch.backends.cudnn.deterministic = True\n"
|
||
],
|
||
"id": "e21c1ebc100b5079",
|
||
"outputs": [],
|
||
"execution_count": 11
|
||
},
|
||
{
|
||
"metadata": {
|
||
"ExecuteTime": {
|
||
"end_time": "2025-06-24T17:37:28.406416Z",
|
||
"start_time": "2025-06-24T17:37:28.403413Z"
|
||
}
|
||
},
|
||
"cell_type": "code",
|
||
"source": [
|
||
"def calculate_score(y_true, y_preds):\n",
|
||
" preds_prob = torch.softmax(y_preds, dim=1)\n",
|
||
" preds_class = torch.argmax(preds_prob, dim=1)\n",
|
||
" correct = (preds_class == y_true).float()\n",
|
||
" return (correct.sum() / len(correct)).cpu().numpy()\n"
|
||
],
|
||
"id": "f70cc647d264747",
|
||
"outputs": [],
|
||
"execution_count": 12
|
||
},
|
||
{
|
||
"metadata": {
|
||
"ExecuteTime": {
|
||
"end_time": "2025-06-24T17:37:28.439020Z",
|
||
"start_time": "2025-06-24T17:37:28.433485Z"
|
||
}
|
||
},
|
||
"cell_type": "code",
|
||
"source": [
|
||
"def train_model(model, criterion, optimizer, scheduler, train_loader, valid_loader, num_epochs, device, save_path):\n",
|
||
" model.to(device)\n",
|
||
" best_acc = 0.0\n",
|
||
" metrics = {'epoch': [], 'train_acc': [], 'valid_acc': [], 'train_loss': [], 'valid_loss': []}\n",
|
||
"\n",
|
||
" for epoch in range(1, num_epochs + 1):\n",
|
||
" # --- 训练阶段 ---\n",
|
||
" model.train()\n",
|
||
" running_loss, running_acc = 0.0, 0.0\n",
|
||
" for imgs, labels in train_loader:\n",
|
||
" imgs, labels = imgs.to(device), labels.to(device)\n",
|
||
" optimizer.zero_grad()\n",
|
||
" outputs = model(imgs)\n",
|
||
" loss = criterion(outputs, labels)\n",
|
||
" loss.backward()\n",
|
||
" torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=1.0)\n",
|
||
" optimizer.step()\n",
|
||
" running_loss += loss.item()\n",
|
||
" running_acc += calculate_score(labels, outputs)\n",
|
||
"\n",
|
||
" train_loss = running_loss / len(train_loader)\n",
|
||
" train_acc = running_acc / len(train_loader)\n",
|
||
" scheduler.step()\n",
|
||
"\n",
|
||
" # --- 验证阶段 ---\n",
|
||
" model.eval()\n",
|
||
" val_loss, val_acc = 0.0, 0.0\n",
|
||
" with torch.no_grad():\n",
|
||
" for imgs, labels in valid_loader:\n",
|
||
" imgs, labels = imgs.to(device), labels.to(device)\n",
|
||
" outputs = model(imgs)\n",
|
||
" loss = criterion(outputs, labels)\n",
|
||
" val_loss += loss.item()\n",
|
||
" val_acc += calculate_score(labels, outputs)\n",
|
||
"\n",
|
||
" valid_loss = val_loss / len(valid_loader)\n",
|
||
" valid_acc = val_acc / len(valid_loader)\n",
|
||
"\n",
|
||
" metrics['epoch'].append(epoch)\n",
|
||
" metrics['train_loss'].append(train_loss)\n",
|
||
" metrics['valid_loss'].append(valid_loss)\n",
|
||
" metrics['train_acc'].append(train_acc)\n",
|
||
" metrics['valid_acc'].append(valid_acc)\n",
|
||
"\n",
|
||
" tqdm.write(f\"[{save_path}] Epoch {epoch}/{num_epochs} \"\n",
|
||
" f\"Train Acc: {train_acc:.4f} Valid Acc: {valid_acc:.4f}\")\n",
|
||
"\n",
|
||
" if valid_acc > best_acc:\n",
|
||
" best_acc = valid_acc\n",
|
||
" torch.save(model.state_dict(), save_path)\n",
|
||
"\n",
|
||
" return model, metrics\n"
|
||
],
|
||
"id": "7cfe89d63ab0e68e",
|
||
"outputs": [],
|
||
"execution_count": 13
|
||
},
|
||
{
|
||
"metadata": {
|
||
"ExecuteTime": {
|
||
"end_time": "2025-06-24T17:37:28.467170Z",
|
||
"start_time": "2025-06-24T17:37:28.463678Z"
|
||
}
|
||
},
|
||
"cell_type": "code",
|
||
"source": [
|
||
"def test_model(model, test_loader, device):\n",
|
||
" model.to(device).eval()\n",
|
||
" acc = 0.0\n",
|
||
" with torch.no_grad():\n",
|
||
" for imgs, labels in test_loader:\n",
|
||
" imgs, labels = imgs.to(device), labels.to(device)\n",
|
||
" outputs = model(imgs)\n",
|
||
" acc += calculate_score(labels, outputs)\n",
|
||
" acc /= len(test_loader)\n",
|
||
" print(f\"Test Accuracy: {acc:.4f}\")\n",
|
||
" return acc\n"
|
||
],
|
||
"id": "235ac16bd786e65f",
|
||
"outputs": [],
|
||
"execution_count": 14
|
||
},
|
||
{
|
||
"metadata": {
|
||
"ExecuteTime": {
|
||
"end_time": "2025-06-24T17:37:28.506986Z",
|
||
"start_time": "2025-06-24T17:37:28.496003Z"
|
||
}
|
||
},
|
||
"cell_type": "code",
|
||
"source": [
|
||
"singlegate_list = ['rx','ry','rz','s','t','p','u3']\n",
|
||
"doublegate_list = ['rxx','ryy','rzz','swap','cnot','cp','ch','cu','ct','cz']\n",
|
||
"\n",
|
||
"class RandomQuantumConvolutionalLayer(nn.Module):\n",
|
||
" def __init__(self, nqubit, num_circuits, seed=1024):\n",
|
||
" super().__init__()\n",
|
||
" random.seed(seed)\n",
|
||
" self.nqubit = nqubit\n",
|
||
" self.cirs = nn.ModuleList([self.circuit(nqubit) for _ in range(num_circuits)])\n",
|
||
" def circuit(self, nqubit):\n",
|
||
" cir = dq.QubitCircuit(nqubit)\n",
|
||
" cir.rxlayer(encode=True); cir.barrier()\n",
|
||
" for _ in range(3):\n",
|
||
" for i in range(nqubit):\n",
|
||
" getattr(cir, random.choice(singlegate_list))(i)\n",
|
||
" c,t = random.sample(range(nqubit),2)\n",
|
||
" gate = random.choice(doublegate_list)\n",
|
||
" if gate[0] in ['r','s']:\n",
|
||
" getattr(cir, gate)([c,t])\n",
|
||
" else:\n",
|
||
" getattr(cir, gate)(c,t)\n",
|
||
" cir.barrier()\n",
|
||
" cir.observable(0)\n",
|
||
" return cir\n",
|
||
" def forward(self, x):\n",
|
||
" k,s = 2,2\n",
|
||
" x_unf = x.unfold(2,k,s).unfold(3,k,s)\n",
|
||
" w = (x.shape[-1]-k)//s + 1\n",
|
||
" x_r = x_unf.reshape(-1, self.nqubit)\n",
|
||
" exps = []\n",
|
||
" for cir in self.cirs:\n",
|
||
" cir(x_r)\n",
|
||
" exps.append(cir.expectation())\n",
|
||
" exps = torch.stack(exps,1).reshape(x.size(0), len(self.cirs), w, w)\n",
|
||
" return exps\n",
|
||
"\n",
|
||
"class RandomQCCNN(nn.Module):\n",
|
||
" def __init__(self):\n",
|
||
" super().__init__()\n",
|
||
" self.conv = nn.Sequential(\n",
|
||
" RandomQuantumConvolutionalLayer(4,3,seed=1024),\n",
|
||
" nn.ReLU(), nn.MaxPool2d(2,1),\n",
|
||
" nn.Conv2d(3,6,2,1), nn.ReLU(), nn.MaxPool2d(2,1)\n",
|
||
" )\n",
|
||
" self.fc = nn.Sequential(\n",
|
||
" nn.Linear(6*6*6,1024), nn.Dropout(0.4),\n",
|
||
" nn.Linear(1024,10)\n",
|
||
" )\n",
|
||
" def forward(self,x):\n",
|
||
" x = self.conv(x)\n",
|
||
" x = x.view(x.size(0),-1)\n",
|
||
" return self.fc(x)\n",
|
||
"\n",
|
||
"class ParameterizedQuantumConvolutionalLayer(nn.Module):\n",
|
||
" def __init__(self,nqubit,num_circuits):\n",
|
||
" super().__init__()\n",
|
||
" self.nqubit = nqubit\n",
|
||
" self.cirs = nn.ModuleList([self.circuit(nqubit) for _ in range(num_circuits)])\n",
|
||
" def circuit(self,nqubit):\n",
|
||
" cir = dq.QubitCircuit(nqubit)\n",
|
||
" cir.rxlayer(encode=True); cir.barrier()\n",
|
||
" for _ in range(4):\n",
|
||
" cir.rylayer(); cir.cnot_ring(); cir.barrier()\n",
|
||
" cir.observable(0)\n",
|
||
" return cir\n",
|
||
" def forward(self,x):\n",
|
||
" k,s = 2,2\n",
|
||
" x_unf = x.unfold(2,k,s).unfold(3,k,s)\n",
|
||
" w = (x.shape[-1]-k)//s +1\n",
|
||
" x_r = x_unf.reshape(-1,self.nqubit)\n",
|
||
" exps = []\n",
|
||
" for cir in self.cirs:\n",
|
||
" cir(x_r); exps.append(cir.expectation())\n",
|
||
" exps = torch.stack(exps,1).reshape(x.size(0),len(self.cirs),w,w)\n",
|
||
" return exps\n",
|
||
"\n",
|
||
"class QCCNN(nn.Module):\n",
|
||
" def __init__(self):\n",
|
||
" super().__init__()\n",
|
||
" self.conv = nn.Sequential(\n",
|
||
" ParameterizedQuantumConvolutionalLayer(4,3),\n",
|
||
" nn.ReLU(), nn.MaxPool2d(2,1)\n",
|
||
" )\n",
|
||
" self.fc = nn.Sequential(\n",
|
||
" nn.Linear(8*8*3,128), nn.Dropout(0.4), nn.ReLU(),\n",
|
||
" nn.Linear(128,10)\n",
|
||
" )\n",
|
||
" def forward(self,x):\n",
|
||
" x = self.conv(x); x = x.view(x.size(0),-1)\n",
|
||
" return self.fc(x)\n",
|
||
"\n",
|
||
"def vgg_block(in_c,out_c,n_convs):\n",
|
||
" layers = [nn.Conv2d(in_c,out_c,3,padding=1), nn.ReLU()]\n",
|
||
" for _ in range(n_convs-1):\n",
|
||
" layers += [nn.Conv2d(out_c,out_c,3,padding=1), nn.ReLU()]\n",
|
||
" layers.append(nn.MaxPool2d(2,2))\n",
|
||
" return nn.Sequential(*layers)\n",
|
||
"\n",
|
||
"VGG = nn.Sequential(\n",
|
||
" vgg_block(1,10,3),\n",
|
||
" vgg_block(10,16,3),\n",
|
||
" nn.Flatten(),\n",
|
||
" nn.Linear(16*4*4,120), nn.Sigmoid(),\n",
|
||
" nn.Linear(120,84), nn.Sigmoid(),\n",
|
||
" nn.Linear(84,10), nn.Softmax(dim=-1)\n",
|
||
")\n"
|
||
],
|
||
"id": "c996822b3d5f8305",
|
||
"outputs": [],
|
||
"execution_count": 15
|
||
},
|
||
{
|
||
"metadata": {
|
||
"ExecuteTime": {
|
||
"end_time": "2025-06-24T18:51:44.622880Z",
|
||
"start_time": "2025-06-24T17:37:28.529442Z"
|
||
}
|
||
},
|
||
"cell_type": "code",
|
||
"source": [
|
||
"if __name__ == '__main__':\n",
|
||
" freeze_support()\n",
|
||
"\n",
|
||
" # 数据增广与加载\n",
|
||
" train_transform = transforms.Compose([\n",
|
||
" transforms.Resize((18, 18)),\n",
|
||
" transforms.RandomRotation(15),\n",
|
||
" transforms.RandomHorizontalFlip(),\n",
|
||
" transforms.RandomVerticalFlip(0.3),\n",
|
||
" transforms.ToTensor(),\n",
|
||
" transforms.Normalize((0.5,), (0.5,))\n",
|
||
" ])\n",
|
||
" eval_transform = transforms.Compose([\n",
|
||
" transforms.Resize((18, 18)),\n",
|
||
" transforms.ToTensor(),\n",
|
||
" transforms.Normalize((0.5,), (0.5,))\n",
|
||
" ])\n",
|
||
"\n",
|
||
" full_train = FashionMNIST(root='./data/notebook2', train=True, transform=train_transform, download=True)\n",
|
||
" test_dataset = FashionMNIST(root='./data/notebook2', train=False, transform=eval_transform, download=True)\n",
|
||
" train_size = int(0.8 * len(full_train))\n",
|
||
" valid_size = len(full_train) - train_size\n",
|
||
" train_ds, valid_ds = torch.utils.data.random_split(full_train, [train_size, valid_size])\n",
|
||
" valid_ds.dataset.transform = eval_transform\n",
|
||
"\n",
|
||
" batch_size = 128\n",
|
||
" train_loader = DataLoader(train_ds, batch_size=batch_size, shuffle=True, drop_last=True, num_workers=4)\n",
|
||
" valid_loader = DataLoader(valid_ds, batch_size=batch_size, shuffle=False, drop_last=True, num_workers=4)\n",
|
||
" test_loader = DataLoader(test_dataset, batch_size=batch_size, shuffle=False, drop_last=False, num_workers=4)\n",
|
||
"\n",
|
||
" # 三种模型配置\n",
|
||
" device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
|
||
" models = {\n",
|
||
" 'random_qccnn': (RandomQCCNN(), 1e-3, './data/notebook2/random_qccnn_best.pt'),\n",
|
||
" 'qccnn': (QCCNN(), 1e-4, './data/notebook2/qccnn_best.pt'),\n",
|
||
" 'vgg': (VGG, 1e-4, './data/notebook2/vgg_best.pt')\n",
|
||
" }\n",
|
||
"\n",
|
||
" all_metrics = {}\n",
|
||
" for name, (model, lr, save_path) in models.items():\n",
|
||
" seed_torch(1024)\n",
|
||
" model = model.to(device)\n",
|
||
" criterion = nn.CrossEntropyLoss()\n",
|
||
" optimizer = optim.AdamW(model.parameters(), lr=lr, weight_decay=1e-4)\n",
|
||
" scheduler = optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=50)\n",
|
||
"\n",
|
||
" print(f\"\\n=== Training {name} ===\")\n",
|
||
" _, metrics = train_model(\n",
|
||
" model, criterion, optimizer, scheduler,\n",
|
||
" train_loader, valid_loader,\n",
|
||
" num_epochs=50, device=device, save_path=save_path\n",
|
||
" )\n",
|
||
" all_metrics[name] = metrics\n",
|
||
" pd.DataFrame(metrics).to_csv(f'./data/notebook2/{name}_metrics.csv', index=False)\n",
|
||
"\n",
|
||
" # 测试与可视化\n",
|
||
" plt.figure(figsize=(12,5))\n",
|
||
" for i,(name,metrics) in enumerate(all_metrics.items(),1):\n",
|
||
" model, _, save_path = models[name]\n",
|
||
" best_model = model.to(device)\n",
|
||
" best_model.load_state_dict(torch.load(save_path))\n",
|
||
" print(f\"\\n--- Testing {name} ---\")\n",
|
||
" test_model(best_model, test_loader, device)\n",
|
||
"\n",
|
||
" plt.subplot(1,3,i)\n",
|
||
" plt.plot(metrics['epoch'], metrics['valid_acc'], label=f'{name} Val Acc')\n",
|
||
" plt.xlabel('Epoch'); plt.ylabel('Valid Acc')\n",
|
||
" plt.title(name); plt.legend()\n",
|
||
"\n",
|
||
" plt.tight_layout(); plt.show()\n",
|
||
"\n",
|
||
" # 参数量统计\n",
|
||
" def count_parameters(m):\n",
|
||
" return sum(p.numel() for p in m.parameters() if p.requires_grad)\n",
|
||
"\n",
|
||
" print(\"\\nParameter Counts:\")\n",
|
||
" for name,(model,_,_) in models.items():\n",
|
||
" print(f\"{name}: {count_parameters(model)}\")\n"
|
||
],
|
||
"id": "2d6b93bb78001086",
|
||
"outputs": [
|
||
{
|
||
"name": "stdout",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"\n",
|
||
"=== Training random_qccnn ===\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 1/50 Train Acc: 0.6377 Valid Acc: 0.7426\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 2/50 Train Acc: 0.7589 Valid Acc: 0.7799\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 3/50 Train Acc: 0.7830 Valid Acc: 0.7955\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 4/50 Train Acc: 0.7928 Valid Acc: 0.8010\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 5/50 Train Acc: 0.7997 Valid Acc: 0.8065\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 6/50 Train Acc: 0.8044 Valid Acc: 0.8140\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 7/50 Train Acc: 0.8097 Valid Acc: 0.8157\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 8/50 Train Acc: 0.8155 Valid Acc: 0.8163\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 9/50 Train Acc: 0.8162 Valid Acc: 0.8159\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 10/50 Train Acc: 0.8169 Valid Acc: 0.8160\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 11/50 Train Acc: 0.8210 Valid Acc: 0.8269\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 12/50 Train Acc: 0.8210 Valid Acc: 0.8266\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 13/50 Train Acc: 0.8241 Valid Acc: 0.8212\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 14/50 Train Acc: 0.8240 Valid Acc: 0.8264\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 15/50 Train Acc: 0.8245 Valid Acc: 0.8231\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 16/50 Train Acc: 0.8270 Valid Acc: 0.8291\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 17/50 Train Acc: 0.8274 Valid Acc: 0.8297\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 18/50 Train Acc: 0.8281 Valid Acc: 0.8338\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 19/50 Train Acc: 0.8295 Valid Acc: 0.8291\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 20/50 Train Acc: 0.8306 Valid Acc: 0.8304\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 21/50 Train Acc: 0.8320 Valid Acc: 0.8280\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 22/50 Train Acc: 0.8316 Valid Acc: 0.8293\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 23/50 Train Acc: 0.8315 Valid Acc: 0.8298\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 24/50 Train Acc: 0.8329 Valid Acc: 0.8282\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 25/50 Train Acc: 0.8321 Valid Acc: 0.8313\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 26/50 Train Acc: 0.8350 Valid Acc: 0.8311\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 27/50 Train Acc: 0.8343 Valid Acc: 0.8335\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 28/50 Train Acc: 0.8347 Valid Acc: 0.8320\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 29/50 Train Acc: 0.8354 Valid Acc: 0.8333\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 30/50 Train Acc: 0.8359 Valid Acc: 0.8314\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 31/50 Train Acc: 0.8380 Valid Acc: 0.8348\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 32/50 Train Acc: 0.8369 Valid Acc: 0.8330\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 33/50 Train Acc: 0.8375 Valid Acc: 0.8367\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 34/50 Train Acc: 0.8373 Valid Acc: 0.8315\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 35/50 Train Acc: 0.8381 Valid Acc: 0.8352\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 36/50 Train Acc: 0.8393 Valid Acc: 0.8374\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 37/50 Train Acc: 0.8384 Valid Acc: 0.8348\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 38/50 Train Acc: 0.8398 Valid Acc: 0.8355\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 39/50 Train Acc: 0.8402 Valid Acc: 0.8365\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 40/50 Train Acc: 0.8400 Valid Acc: 0.8346\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 41/50 Train Acc: 0.8411 Valid Acc: 0.8374\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 42/50 Train Acc: 0.8409 Valid Acc: 0.8373\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 43/50 Train Acc: 0.8415 Valid Acc: 0.8364\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 44/50 Train Acc: 0.8414 Valid Acc: 0.8359\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 45/50 Train Acc: 0.8409 Valid Acc: 0.8364\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 46/50 Train Acc: 0.8419 Valid Acc: 0.8371\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 47/50 Train Acc: 0.8422 Valid Acc: 0.8369\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 48/50 Train Acc: 0.8421 Valid Acc: 0.8360\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 49/50 Train Acc: 0.8417 Valid Acc: 0.8369\n",
|
||
"[./data/notebook2/random_qccnn_best.pt] Epoch 50/50 Train Acc: 0.8408 Valid Acc: 0.8369\n",
|
||
"\n",
|
||
"=== Training qccnn ===\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 1/50 Train Acc: 0.3113 Valid Acc: 0.5076\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 2/50 Train Acc: 0.4621 Valid Acc: 0.5548\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 3/50 Train Acc: 0.5201 Valid Acc: 0.5892\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 4/50 Train Acc: 0.5642 Valid Acc: 0.6311\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 5/50 Train Acc: 0.6056 Valid Acc: 0.6599\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 6/50 Train Acc: 0.6317 Valid Acc: 0.6804\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 7/50 Train Acc: 0.6514 Valid Acc: 0.6929\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 8/50 Train Acc: 0.6678 Valid Acc: 0.7008\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 9/50 Train Acc: 0.6809 Valid Acc: 0.7096\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 10/50 Train Acc: 0.6877 Valid Acc: 0.7186\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 11/50 Train Acc: 0.6976 Valid Acc: 0.7247\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 12/50 Train Acc: 0.7031 Valid Acc: 0.7303\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 13/50 Train Acc: 0.7119 Valid Acc: 0.7317\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 14/50 Train Acc: 0.7164 Valid Acc: 0.7404\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 15/50 Train Acc: 0.7211 Valid Acc: 0.7438\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 16/50 Train Acc: 0.7249 Valid Acc: 0.7481\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 17/50 Train Acc: 0.7294 Valid Acc: 0.7500\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 18/50 Train Acc: 0.7345 Valid Acc: 0.7518\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 19/50 Train Acc: 0.7350 Valid Acc: 0.7550\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 20/50 Train Acc: 0.7391 Valid Acc: 0.7587\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 21/50 Train Acc: 0.7434 Valid Acc: 0.7608\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 22/50 Train Acc: 0.7443 Valid Acc: 0.7634\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 23/50 Train Acc: 0.7481 Valid Acc: 0.7654\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 24/50 Train Acc: 0.7498 Valid Acc: 0.7683\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 25/50 Train Acc: 0.7529 Valid Acc: 0.7696\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 26/50 Train Acc: 0.7547 Valid Acc: 0.7708\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 27/50 Train Acc: 0.7547 Valid Acc: 0.7723\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 28/50 Train Acc: 0.7580 Valid Acc: 0.7736\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 29/50 Train Acc: 0.7571 Valid Acc: 0.7749\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 30/50 Train Acc: 0.7602 Valid Acc: 0.7760\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 31/50 Train Acc: 0.7610 Valid Acc: 0.7767\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 32/50 Train Acc: 0.7618 Valid Acc: 0.7764\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 33/50 Train Acc: 0.7630 Valid Acc: 0.7784\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 34/50 Train Acc: 0.7632 Valid Acc: 0.7791\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 35/50 Train Acc: 0.7627 Valid Acc: 0.7786\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 36/50 Train Acc: 0.7653 Valid Acc: 0.7803\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 37/50 Train Acc: 0.7640 Valid Acc: 0.7811\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 38/50 Train Acc: 0.7674 Valid Acc: 0.7799\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 39/50 Train Acc: 0.7649 Valid Acc: 0.7816\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 40/50 Train Acc: 0.7661 Valid Acc: 0.7823\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 41/50 Train Acc: 0.7668 Valid Acc: 0.7818\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 42/50 Train Acc: 0.7662 Valid Acc: 0.7818\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 43/50 Train Acc: 0.7668 Valid Acc: 0.7824\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 44/50 Train Acc: 0.7678 Valid Acc: 0.7825\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 45/50 Train Acc: 0.7677 Valid Acc: 0.7826\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 46/50 Train Acc: 0.7666 Valid Acc: 0.7827\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 47/50 Train Acc: 0.7689 Valid Acc: 0.7828\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 48/50 Train Acc: 0.7675 Valid Acc: 0.7827\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 49/50 Train Acc: 0.7678 Valid Acc: 0.7828\n",
|
||
"[./data/notebook2/qccnn_best.pt] Epoch 50/50 Train Acc: 0.7677 Valid Acc: 0.7826\n",
|
||
"\n",
|
||
"=== Training vgg ===\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 1/50 Train Acc: 0.2536 Valid Acc: 0.4195\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 2/50 Train Acc: 0.4692 Valid Acc: 0.5073\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 3/50 Train Acc: 0.5270 Valid Acc: 0.5266\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 4/50 Train Acc: 0.5354 Valid Acc: 0.5351\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 5/50 Train Acc: 0.5935 Valid Acc: 0.6175\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 6/50 Train Acc: 0.6366 Valid Acc: 0.6626\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 7/50 Train Acc: 0.6866 Valid Acc: 0.7303\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 8/50 Train Acc: 0.7415 Valid Acc: 0.7512\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 9/50 Train Acc: 0.7542 Valid Acc: 0.7582\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 10/50 Train Acc: 0.7608 Valid Acc: 0.7646\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 11/50 Train Acc: 0.7677 Valid Acc: 0.7719\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 12/50 Train Acc: 0.7705 Valid Acc: 0.7731\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 13/50 Train Acc: 0.7716 Valid Acc: 0.7736\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 14/50 Train Acc: 0.7748 Valid Acc: 0.7749\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 15/50 Train Acc: 0.7763 Valid Acc: 0.7751\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 16/50 Train Acc: 0.7790 Valid Acc: 0.7768\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 17/50 Train Acc: 0.7801 Valid Acc: 0.7802\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 18/50 Train Acc: 0.7818 Valid Acc: 0.7828\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 19/50 Train Acc: 0.7827 Valid Acc: 0.7816\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 20/50 Train Acc: 0.7827 Valid Acc: 0.7840\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 21/50 Train Acc: 0.7849 Valid Acc: 0.7858\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 22/50 Train Acc: 0.7877 Valid Acc: 0.7849\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 23/50 Train Acc: 0.7880 Valid Acc: 0.7849\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 24/50 Train Acc: 0.7891 Valid Acc: 0.7860\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 25/50 Train Acc: 0.7903 Valid Acc: 0.7884\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 26/50 Train Acc: 0.7910 Valid Acc: 0.7900\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 27/50 Train Acc: 0.7919 Valid Acc: 0.7886\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 28/50 Train Acc: 0.7937 Valid Acc: 0.7906\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 29/50 Train Acc: 0.7934 Valid Acc: 0.7876\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 30/50 Train Acc: 0.7946 Valid Acc: 0.7902\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 31/50 Train Acc: 0.7959 Valid Acc: 0.7933\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 32/50 Train Acc: 0.7956 Valid Acc: 0.7933\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 33/50 Train Acc: 0.7971 Valid Acc: 0.7920\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 34/50 Train Acc: 0.7974 Valid Acc: 0.7932\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 35/50 Train Acc: 0.7980 Valid Acc: 0.7948\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 36/50 Train Acc: 0.7985 Valid Acc: 0.7950\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 37/50 Train Acc: 0.7990 Valid Acc: 0.7959\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 38/50 Train Acc: 0.7992 Valid Acc: 0.7949\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 39/50 Train Acc: 0.8001 Valid Acc: 0.7960\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 40/50 Train Acc: 0.8006 Valid Acc: 0.7957\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 41/50 Train Acc: 0.8007 Valid Acc: 0.7963\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 42/50 Train Acc: 0.8015 Valid Acc: 0.7959\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 43/50 Train Acc: 0.8014 Valid Acc: 0.7962\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 44/50 Train Acc: 0.8016 Valid Acc: 0.7965\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 45/50 Train Acc: 0.8018 Valid Acc: 0.7958\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 46/50 Train Acc: 0.8021 Valid Acc: 0.7965\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 47/50 Train Acc: 0.8025 Valid Acc: 0.7966\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 48/50 Train Acc: 0.8026 Valid Acc: 0.7962\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 49/50 Train Acc: 0.8025 Valid Acc: 0.7970\n",
|
||
"[./data/notebook2/vgg_best.pt] Epoch 50/50 Train Acc: 0.8025 Valid Acc: 0.7970\n",
|
||
"\n",
|
||
"--- Testing random_qccnn ---\n",
|
||
"Test Accuracy: 0.8249\n",
|
||
"\n",
|
||
"--- Testing qccnn ---\n",
|
||
"Test Accuracy: 0.7720\n",
|
||
"\n",
|
||
"--- Testing vgg ---\n",
|
||
"Test Accuracy: 0.7899\n"
|
||
]
|
||
},
|
||
{
|
||
"data": {
|
||
"text/plain": [
|
||
"<Figure size 1200x500 with 3 Axes>"
|
||
],
|
||
"image/png": "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"
|
||
},
|
||
"metadata": {},
|
||
"output_type": "display_data"
|
||
},
|
||
{
|
||
"name": "stdout",
|
||
"output_type": "stream",
|
||
"text": [
|
||
"\n",
|
||
"Parameter Counts:\n",
|
||
"random_qccnn: 232581\n",
|
||
"qccnn: 26042\n",
|
||
"vgg: 49870\n"
|
||
]
|
||
}
|
||
],
|
||
"execution_count": 16
|
||
},
|
||
{
|
||
"metadata": {},
|
||
"cell_type": "markdown",
|
||
"source": [
|
||
"下面是新版本脚本相对于原始代码的主要改进点及对应说明，已结合关键代码片段进行标注：\n",
|
||
"\n",
|
||
"1. **数据增强（Data Augmentation）**\n",
|
||
"\n",
|
||
" * **原始**：训练和验证都只做了 `Resize` + `ToTensor`，数据量固定；\n",
|
||
" * **改进**：对训练集添加了 `RandomRotation`、`RandomHorizontalFlip`、`RandomVerticalFlip` 等操作，大幅增加了样本多样性，有助于提升模型泛化能力。\n",
|
||
"\n",
|
||
" ```python\n",
|
||
" train_transform = transforms.Compose([\n",
|
||
" transforms.Resize((18, 18)),\n",
|
||
" transforms.RandomRotation(15), # 随机旋转\n",
|
||
" transforms.RandomHorizontalFlip(), # 随机水平翻转\n",
|
||
" transforms.RandomVerticalFlip(0.3), # 随机垂直翻转\n",
|
||
" transforms.ToTensor(),\n",
|
||
" transforms.Normalize((0.5,), (0.5,))\n",
|
||
" ])\n",
|
||
" ```\n",
|
||
"\n",
|
||
"2. **优化器与学习率调度**\n",
|
||
"\n",
|
||
" * **原始**：`SGD(lr=0.01, weight_decay=0.001)` 或 `Adam(lr=1e-5)`，无学习率变化；\n",
|
||
" * **改进**：统一使用 `AdamW`（带权重衰减的 Adam），更稳定；添加 `CosineAnnealingLR`，能在训练中动态调整学习率，促进更快收敛、更高精度。\n",
|
||
"\n",
|
||
" ```python\n",
|
||
" optimizer = optim.AdamW(model.parameters(), lr=1e-3, weight_decay=1e-4)\n",
|
||
" scheduler = optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=50)\n",
|
||
" ```\n",
|
||
"\n",
|
||
"3. **批大小与训练轮次（Batch Size & Epochs）**\n",
|
||
"\n",
|
||
" * **原始**：`batch_size=64`，`num_epochs=300`；\n",
|
||
" * **改进**：增大到 `batch_size=128`，减少到 `num_epochs=50`。\n",
|
||
"\n",
|
||
" * **批量更大**：更好利用 GPU 并行能力；\n",
|
||
" * **轮次更少**：缩短训练时间，同时在学习率调度下仍能达到高准确度。\n",
|
||
"\n",
|
||
"4. **梯度裁剪（Gradient Clipping）**\n",
|
||
"\n",
|
||
" * 增加了 `torch.nn.utils.clip_grad_norm_(…)`，避免量子网络中可能出现的梯度爆炸，保障训练稳定性。\n",
|
||
"\n",
|
||
" ```python\n",
|
||
" torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=1.0)\n",
|
||
" ```\n",
|
||
"\n",
|
||
"5. **统一训练框架与多模型支持**\n",
|
||
"\n",
|
||
" * 将原来三段几乎重复的训练逻辑，抽象成 `train_model(...)` 函数，并通过一个字典 `models` 循环一次性训练：\n",
|
||
"\n",
|
||
" * `random_qccnn`、`qccnn`、`vgg` 三种架构统一流程；\n",
|
||
" * 每种模型均保存最优权重（`save_path`）与训练曲线（`metrics.csv`）。\n",
|
||
"\n",
|
||
" ```python\n",
|
||
" models = {\n",
|
||
" 'random_qccnn': (RandomQCCNN(), 1e-3, './random_qccnn_best.pt'),\n",
|
||
" 'qccnn': (QCCNN(), 1e-4, './qccnn_best.pt'),\n",
|
||
" 'vgg': (VGG, 1e-4, './vgg_best.pt')\n",
|
||
" }\n",
|
||
" for name, (model, lr, save_path) in models.items():\n",
|
||
" trained, metrics = train_model(\n",
|
||
" model, criterion, optimizer, scheduler,\n",
|
||
" train_loader, valid_loader,\n",
|
||
" num_epochs=50, device=device, save_path=save_path\n",
|
||
" )\n",
|
||
" pd.DataFrame(metrics).to_csv(f'./{name}_metrics.csv', index=False)\n",
|
||
" ```\n",
|
||
"\n",
|
||
"6. **最佳模型保存和测试分离**\n",
|
||
"\n",
|
||
" * `train_model` 内部自动监控验证集准确率并保存最佳参数；\n",
|
||
" * 训练结束后，再统一加载最佳权重进行测试，确保测试结果与最优状态对应。\n",
|
||
"\n",
|
||
"7. **可视化对比**\n",
|
||
"\n",
|
||
" * 最后一个代码块中，通过 `matplotlib` 并排绘制三种模型的验证准确率曲线，直观比较不同模型的收敛速度和最终性能。\n",
|
||
"\n",
|
||
" ```python\n",
|
||
" plt.figure(figsize=(12,5))\n",
|
||
" for i,(name,metrics) in enumerate(all_metrics.items(),1):\n",
|
||
" plt.subplot(1,3,i)\n",
|
||
" plt.plot(metrics['epoch'], metrics['valid_acc'], label=f'{name} Val Acc')\n",
|
||
" plt.title(name); plt.legend()\n",
|
||
" plt.tight_layout(); plt.show()\n",
|
||
" ```\n",
|
||
"\n",
|
||
"8. **参数量统计**\n",
|
||
"\n",
|
||
" * 在脚本末尾增加 `count_parameters` 函数，打印三种模型的可训练参数量，帮助评估模型复杂度与性能的权衡。\n",
|
||
"\n",
|
||
" ```python\n",
|
||
" def count_parameters(m):\n",
|
||
" return sum(p.numel() for p in m.parameters() if p.requires_grad)\n",
|
||
" for name,(model,_,_) in models.items():\n",
|
||
" print(f\"{name}: {count_parameters(model)}\")\n",
|
||
" ```\n",
|
||
"\n",
|
||
"---\n",
|
||
"\n",
|
||
"**总体效果**：\n",
|
||
"\n",
|
||
"* **训练速度**：批量更大、轮次更少、学习率动态调度，整体训练时间显著缩短。\n",
|
||
"* **模型准确率**：数据增强 + 优化器 + 调度器 + 梯度裁剪等多项改进，显著提高了各模型在验证集和测试集上的准确率。\n",
|
||
"* **可维护性**：统一框架、函数抽象、循环训练及结果保存，大大简化了代码结构，便于后续扩展和调试。\n"
|
||
],
|
||
"id": "6a5e9602f481107e"
|
||
}
|
||
],
|
||
"metadata": {},
|
||
"nbformat": 4,
|
||
"nbformat_minor": 5
|
||
}
|