{"id":1057,"date":"2026-05-14T19:18:44","date_gmt":"2026-05-14T11:18:44","guid":{"rendered":"https:\/\/www.eutaboo.com\/index.php\/2026\/05\/14\/%e3%80%8annu-net-0%e5%9f%ba%e7%a1%80%e5%85%a5%e9%97%a8%ef%bc%889%ef%bc%89%ef%bc%9a%e4%bf%ae%e6%94%b9-nnu-net-trainer%ef%bc%8c%e4%bb%8e%e7%bb%a7%e6%89%bf-nnunettrainer-%e5%bc%80%e5%a7%8b%e3%80%8b\/"},"modified":"2026-05-14T19:18:44","modified_gmt":"2026-05-14T11:18:44","slug":"%e3%80%8annu-net-0%e5%9f%ba%e7%a1%80%e5%85%a5%e9%97%a8%ef%bc%889%ef%bc%89%ef%bc%9a%e4%bf%ae%e6%94%b9-nnu-net-trainer%ef%bc%8c%e4%bb%8e%e7%bb%a7%e6%89%bf-nnunettrainer-%e5%bc%80%e5%a7%8b%e3%80%8b","status":"publish","type":"post","link":"https:\/\/www.eutaboo.com\/index.php\/2026\/05\/14\/%e3%80%8annu-net-0%e5%9f%ba%e7%a1%80%e5%85%a5%e9%97%a8%ef%bc%889%ef%bc%89%ef%bc%9a%e4%bf%ae%e6%94%b9-nnu-net-trainer%ef%bc%8c%e4%bb%8e%e7%bb%a7%e6%89%bf-nnunettrainer-%e5%bc%80%e5%a7%8b%e3%80%8b\/","title":{"rendered":"\u300annU-Net 0\u57fa\u7840\u5165\u95e8\uff089\uff09\uff1a\u4fee\u6539 nnU-Net Trainer\uff0c\u4ece\u7ee7\u627f nnUNetTrainer \u5f00\u59cb\u300b"},"content":{"rendered":"<h2>\u672c\u7bc7\u5b66\u4e60\u76ee\u6807<\/h2>\n<p>\u8fd9\u662f\u300annU-Net 0\u57fa\u7840\u5165\u95e8\u300b\u7cfb\u5217\u7684\u7b2c 9 \u7bc7\u3002\u524d\u4e00\u7bc7\u6211\u4eec\u5efa\u7acb\u4e86 nnU-Net v2 \u7684\u5185\u90e8\u6846\u67b6\u5730\u56fe\u3002\u672c\u6587\u5f00\u59cb\u771f\u6b63\u4fee\u6539 nnU-Net\uff1a\u4ece\u81ea\u5b9a\u4e49 Trainer \u5165\u624b\u3002<\/p>\n<p>\u8bfb\u5b8c\u672c\u6587\uff0c\u4f60\u5e94\u8be5\u80fd\u591f\uff1a<\/p>\n<ol>\n<li>\u7406\u89e3\u4e3a\u4ec0\u4e48\u4fee\u6539\u8bad\u7ec3\u6d41\u7a0b\u65f6\u63a8\u8350\u7ee7\u627f <code>nnUNetTrainer<\/code>\u3002<\/li>\n<li>\u5199\u51fa\u4e00\u4e2a\u6700\u5c0f\u81ea\u5b9a\u4e49 Trainer\u3002<\/li>\n<li>\u7528 <code>-tr<\/code> \u53c2\u6570\u8ba9 <code>nnUNetv2_train<\/code> \u8c03\u7528\u4f60\u7684 Trainer\u3002<\/li>\n<li>\u77e5\u9053\u54ea\u4e9b\u6539\u52a8\u9002\u5408\u653e\u5728 Trainer\uff0c\u54ea\u4e9b\u4e0d\u5e94\u8be5\u76f4\u63a5\u786c\u6539\u6838\u5fc3\u6e90\u7801\u3002<\/li>\n<\/ol>\n<h2>1. \u4e3a\u4ec0\u4e48\u4ece Trainer \u5f00\u59cb\u6539<\/h2>\n<p>\u5b98\u65b9\u6269\u5c55\u6587\u6863\u660e\u786e\u5efa\u8bae\uff1a\u5982\u679c\u4f60\u60f3\u4fee\u6539 training procedure\uff0c\u4e5f\u5c31\u662f\u8bad\u7ec3\u6d41\u7a0b\uff0c\u4f8b\u5982 loss\u3001sampling\u3001data augmentation\u3001lr scheduler \u7b49\uff0c\u5e94\u5b9e\u73b0\u81ea\u5df1\u7684 trainer class\u3002\u6700\u4f73\u5b9e\u8df5\u662f\u521b\u5efa\u4e00\u4e2a\u7ee7\u627f <code>nnUNetTrainer<\/code> \u7684\u7c7b\uff0c\u5e76\u53ea\u8986\u76d6\u4f60\u9700\u8981\u4fee\u6539\u7684\u65b9\u6cd5\u3002<\/p>\n<p>\u8fd9\u6837\u505a\u6709\u4e09\u4e2a\u597d\u5904\uff1a<\/p>\n<ul>\n<li>\u4fdd\u7559 nnU-Net v2 \u5df2\u7ecf\u5b9e\u73b0\u597d\u7684\u6570\u636e\u52a0\u8f7d\u3001\u65e5\u5fd7\u3001checkpoint\u3001\u9a8c\u8bc1\u548c\u63a8\u7406\u517c\u5bb9\u903b\u8f91\u3002<\/li>\n<li>\u4f60\u7684\u5b9e\u9a8c\u6539\u52a8\u96c6\u4e2d\u5728\u4e00\u4e2a\u65b0\u7c7b\u91cc\uff0c\u4fbf\u4e8e\u56de\u6eda\u548c\u590d\u73b0\u3002<\/li>\n<li>\u8bad\u7ec3\u547d\u4ee4\u4e2d\u53ef\u4ee5\u901a\u8fc7 <code>-tr<\/code> \u660e\u786e\u8bb0\u5f55\u4f7f\u7528\u4e86\u54ea\u4e2a Trainer\u3002<\/li>\n<\/ul>\n<h2>2. \u4e0d\u63a8\u8350\u76f4\u63a5\u6539\u6838\u5fc3\u6e90\u7801<\/h2>\n<p>\u5f88\u591a\u521d\u5b66\u8005\u4f1a\u76f4\u63a5\u6253\u5f00 <code>nnUNetTrainer.py<\/code> \u4fee\u6539\u51e0\u884c\u4ee3\u7801\u3002\u8fd9\u79cd\u65b9\u5f0f\u77ed\u671f\u770b\u6700\u5feb\uff0c\u4f46\u95ee\u9898\u5f88\u5927\uff1a<\/p>\n<ul>\n<li>\u4ee5\u540e\u66f4\u65b0 nnU-Net \u65f6\u5bb9\u6613\u88ab\u8986\u76d6\u3002<\/li>\n<li>\u65e0\u6cd5\u6e05\u695a\u533a\u5206\u201c\u5b98\u65b9\u9ed8\u8ba4\u884c\u4e3a\u201d\u548c\u201c\u4f60\u81ea\u5df1\u7684\u5b9e\u9a8c\u884c\u4e3a\u201d\u3002<\/li>\n<li>\u590d\u73b0\u5b9e\u9a8c\u65f6\u5f88\u96be\u8bf4\u660e\u5230\u5e95\u6539\u4e86\u54ea\u91cc\u3002<\/li>\n<li>\u548c\u522b\u4eba\u5171\u4eab checkpoint \u65f6\uff0c\u5bf9\u65b9\u53ef\u80fd\u65e0\u6cd5\u63a8\u7406\u6216\u7ee7\u7eed\u8bad\u7ec3\u3002<\/li>\n<\/ul>\n<p>\u66f4\u7a33\u59a5\u7684\u65b9\u5f0f\u662f\uff1a\u65b0\u5efa\u4e00\u4e2a Trainer \u7c7b\uff0c\u7ee7\u627f\u5b98\u65b9 <code>nnUNetTrainer<\/code>\uff0c\u53ea\u8986\u76d6\u4f60\u9700\u8981\u7684\u90e8\u5206\u3002<\/p>\n<h2>3. Trainer \u4fee\u6539\u5165\u53e3\u56fe<\/h2>\n<pre><code class=\"language-mermaid\">flowchart TD\n    A[nnUNetv2_train] --> B[-tr \u6307\u5b9a Trainer \u540d\u79f0]\n    B --> C[\u67e5\u627e Trainer \u7c7b]\n    C --> D[\u5b9e\u4f8b\u5316\u81ea\u5b9a\u4e49 Trainer]\n    D --> E[initialize]\n    E --> F[build_network_architecture]\n    E --> G[_build_loss]\n    E --> H[configure_optimizers]\n    E --> I[get_training_transforms]\n    D --> J[run_training]\n<\/code><\/pre>\n<p>\u8fd9\u5f20\u56fe\u8bf4\u660e\uff1aTrainer \u4e0d\u662f\u53ea\u8d1f\u8d23\u4e00\u4e2a\u51fd\u6570\uff0c\u800c\u662f\u8bad\u7ec3\u6d41\u7a0b\u7684\u7ec4\u7ec7\u8005\u3002\u4f60\u53ef\u4ee5\u8986\u76d6 loss\u3001optimizer\u3001augmentation\u3001network architecture \u7b49\u4e0d\u540c\u5165\u53e3\uff0c\u4f46\u6bcf\u6b21\u6539\u52a8\u90fd\u5e94\u5c3d\u91cf\u5c0f\u3002<\/p>\n<h2>4. \u4e00\u4e2a\u6700\u5c0f\u81ea\u5b9a\u4e49 Trainer\uff1a\u628a\u4f18\u5316\u5668\u6362\u6210 AdamW<\/h2>\n<p>\u4e0b\u9762\u793a\u4f8b\u6f14\u793a\u5982\u4f55\u5199\u4e00\u4e2a\u81ea\u5b9a\u4e49 Trainer\uff0c\u628a\u9ed8\u8ba4\u4f18\u5316\u5668\u66ff\u6362\u4e3a AdamW\u3002\u8fd9\u4e2a\u4f8b\u5b50\u53ea\u662f\u6559\u5b66\u7528\uff0c\u4e0d\u80fd\u4fdd\u8bc1\u6bd4\u9ed8\u8ba4 SGD \u66f4\u597d\u3002\u533b\u5b66\u56fe\u50cf\u5206\u5272\u5b9e\u9a8c\u4e2d\uff0c\u4efb\u4f55\u4f18\u5316\u5668\u6539\u52a8\u90fd\u5fc5\u987b\u7528\u9a8c\u8bc1\u96c6\u7ed3\u679c\u8bc1\u660e\u3002<\/p>\n<pre><code class=\"language-python\">import torch\n\nfrom nnunetv2.training.lr_scheduler.polylr import PolyLRScheduler\nfrom nnunetv2.training.nnUNetTrainer.nnUNetTrainer import nnUNetTrainer\n\n\nclass nnUNetTrainerMyAdamW(nnUNetTrainer):\n    def __init__(\n        self,\n        plans: dict,\n        configuration: str,\n        fold: int,\n        dataset_json: dict,\n        device: torch.device = torch.device(\"cuda\"),\n    ):\n        super().__init__(plans, configuration, fold, dataset_json, device)\n        self.initial_lr = 3e-4\n        self.weight_decay = 1e-4\n\n    def configure_optimizers(self):\n        optimizer = torch.optim.AdamW(\n            self.network.parameters(),\n            lr=self.initial_lr,\n            weight_decay=self.weight_decay,\n        )\n        lr_scheduler = PolyLRScheduler(\n            optimizer,\n            self.initial_lr,\n            self.num_epochs,\n        )\n        return optimizer, lr_scheduler<\/code><\/pre>\n<p>\u8fd9\u91cc\u8986\u76d6\u7684\u662f <code>configure_optimizers<\/code>\u3002\u5b98\u65b9\u5f53\u524d <code>nnUNetTrainer<\/code> \u9ed8\u8ba4\u4f7f\u7528 SGD\u3001momentum\u3001Nesterov \u548c poly learning rate scheduler\u3002\u6211\u4eec\u6ca1\u6709\u6539\u6570\u636e\u52a0\u8f7d\u3001loss\u3001augmentation \u548c checkpoint \u884c\u4e3a\u3002<\/p>\n<h2>5. \u628a Trainer \u653e\u5728\u54ea\u91cc<\/h2>\n<p>\u5b98\u65b9\u8bad\u7ec3\u5165\u53e3\u4f1a\u6839\u636e <code>-tr<\/code> \u53c2\u6570\u67e5\u627e Trainer \u7c7b\u3002\u5f53\u524d\u6e90\u7801\u4e2d\uff0c\u67e5\u627e\u987a\u5e8f\u5305\u62ec\uff1a<\/p>\n<ol>\n<li><code>nnunetv2\/training\/nnUNetTrainer<\/code> \u4e0b\u7684 Python \u6587\u4ef6\u3002<\/li>\n<li>\u5982\u679c\u8bbe\u7f6e\u4e86 <code>nnUNet_extTrainer<\/code> \u73af\u5883\u53d8\u91cf\uff0c\u4e5f\u4f1a\u641c\u7d22\u8be5\u5916\u90e8\u8def\u5f84\u3002<\/li>\n<\/ol>\n<p>\u5982\u679c\u4f60\u4f7f\u7528 editable install\uff0c\u53ef\u4ee5\u628a\u6587\u4ef6\u653e\u5230\u6e90\u7801\u76ee\u5f55\u4e2d\uff0c\u4f8b\u5982\uff1a<\/p>\n<pre><code class=\"language-text\">nnUNet\/\n\u2514\u2500\u2500 nnunetv2\/\n    \u2514\u2500\u2500 training\/\n        \u2514\u2500\u2500 nnUNetTrainer\/\n            \u2514\u2500\u2500 variants\/\n                \u2514\u2500\u2500 optimizer\/\n                    \u2514\u2500\u2500 nnUNetTrainerMyAdamW.py<\/code><\/pre>\n<p>\u4e5f\u53ef\u4ee5\u653e\u5230\u5916\u90e8\u76ee\u5f55\uff0c\u7136\u540e\u8bbe\u7f6e\u73af\u5883\u53d8\u91cf\uff1a<\/p>\n<pre><code class=\"language-bash\">export nnUNet_extTrainer=\"\/home\/you\/nnunet_custom_trainers\"<\/code><\/pre>\n<p>\u5916\u90e8\u76ee\u5f55\u4e2d\u653e\u5165\uff1a<\/p>\n<pre><code class=\"language-text\">\/home\/you\/nnunet_custom_trainers\/\n\u2514\u2500\u2500 nnUNetTrainerMyAdamW.py<\/code><\/pre>\n<p>\u5bf9\u4e8e\u521d\u5b66\u8005\uff0c\u6211\u5efa\u8bae\u5148\u4f7f\u7528 editable install \u5e76\u628a trainer \u653e\u5728\u6e90\u7801\u6811\u7684 <code>variants<\/code> \u76ee\u5f55\u4e0b\u3002\u7b49\u4f60\u8981\u957f\u671f\u7ef4\u62a4\u591a\u4e2a\u79c1\u6709 Trainer\uff0c\u518d\u8003\u8651 <code>nnUNet_extTrainer<\/code>\u3002<\/p>\n<h2>6. \u7528 -tr \u8c03\u7528\u81ea\u5b9a\u4e49 Trainer<\/h2>\n<p>\u8bad\u7ec3\u65f6\u4f7f\u7528 <code>-tr<\/code> \u6307\u5b9a\u7c7b\u540d\uff1a<\/p>\n<pre><code class=\"language-bash\">nnUNetv2_train 1 3d_fullres 0 -tr nnUNetTrainerMyAdamW --npz<\/code><\/pre>\n<p>\u5982\u679c\u4f60\u8fd8\u4f7f\u7528\u81ea\u5b9a\u4e49 plans identifier\uff0c\u53ef\u4ee5\u540c\u65f6\u7528 <code>-p<\/code> \u6307\u5b9a\uff1a<\/p>\n<pre><code class=\"language-bash\">nnUNetv2_train 1 3d_fullres 0 -tr nnUNetTrainerMyAdamW -p nnUNetPlans --npz<\/code><\/pre>\n<p>\u8bad\u7ec3\u8f93\u51fa\u76ee\u5f55\u4f1a\u5305\u542b Trainer \u540d\u79f0\uff0c\u4f8b\u5982\uff1a<\/p>\n<pre><code class=\"language-text\">nnUNet_results\/\n\u2514\u2500\u2500 Dataset001_Liver\/\n    \u2514\u2500\u2500 nnUNetTrainerMyAdamW__nnUNetPlans__3d_fullres\/\n        \u2514\u2500\u2500 fold_0\/<\/code><\/pre>\n<p>\u8fd9\u975e\u5e38\u91cd\u8981\u3002\u76ee\u5f55\u540d\u672c\u8eab\u5c31\u8bb0\u5f55\u4e86\u4f60\u4f7f\u7528\u7684 Trainer\u3001plans \u548c configuration\uff0c\u6709\u5229\u4e8e\u5b9e\u9a8c\u7ba1\u7406\u3002<\/p>\n<h2>7. \u8c03\u8bd5\u81ea\u5b9a\u4e49 Trainer \u7684\u63a8\u8350\u987a\u5e8f<\/h2>\n<p>\u4e0d\u8981\u4e00\u4e0a\u6765\u5c31\u8bad\u7ec3\u5b8c\u6574 1000 epochs\u3002\u5efa\u8bae\u5148\u7528\u5b98\u65b9\u5df2\u6709\u77ed\u8bad\u7ec3 Trainer \u6216\u4f60\u81ea\u5df1\u7684 debug Trainer \u8dd1\u6700\u5c0f\u9a8c\u8bc1\u3002\u4f8b\u5982\u5b98\u65b9\u4ed3\u5e93\u91cc\u6709 <code>nnUNetTrainer_5epochs<\/code> \u8fd9\u7c7b\u8bad\u7ec3\u957f\u5ea6\u53d8\u4f53\uff0c\u9002\u5408\u5feb\u901f\u68c0\u67e5\u6d41\u7a0b\u3002<\/p>\n<p>\u5bf9\u81ea\u5df1\u7684 Trainer\uff0c\u53ef\u4ee5\u5148\u8fd9\u6837\u505a\uff1a<\/p>\n<pre><code class=\"language-bash\"># \u5148\u786e\u8ba4\u7c7b\u80fd\u88ab\u627e\u5230\nnnUNetv2_train 1 3d_fullres 0 -tr nnUNetTrainerMyAdamW --disable_checkpointing\n\n# \u6b63\u5f0f\u8bad\u7ec3\u65f6\u518d\u52a0 --npz\nnnUNetv2_train 1 3d_fullres 0 -tr nnUNetTrainerMyAdamW --npz<\/code><\/pre>\n<p><code>--disable_checkpointing<\/code> \u9002\u5408\u6d4b\u8bd5\uff0c\u907f\u514d\u8c03\u8bd5\u65f6\u4ea7\u751f\u5927\u91cf checkpoint\u3002\u6b63\u5f0f\u5b9e\u9a8c\u4e0d\u8981\u968f\u610f\u7981\u7528 checkpoint\uff0c\u5426\u5219\u4e2d\u65ad\u540e\u5f88\u96be\u6062\u590d\u3002<\/p>\n<h2>8. \u5e38\u89c1\u9519\u8bef<\/h2>\n<table>\n<thead>\n<tr>\n<th>\u9519\u8bef\u73b0\u8c61<\/th>\n<th>\u5e38\u89c1\u539f\u56e0<\/th>\n<th>\u89e3\u51b3\u65b9\u5f0f<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>\u627e\u4e0d\u5230 Trainer<\/td>\n<td>\u7c7b\u540d\u548c <code>-tr<\/code> \u4e0d\u4e00\u81f4\uff0c\u6216\u6587\u4ef6\u4e0d\u5728\u641c\u7d22\u8def\u5f84\u4e2d<\/td>\n<td>\u68c0\u67e5\u7c7b\u540d\u3001\u6587\u4ef6\u540d\u3001editable install \u548c <code>nnUNet_extTrainer<\/code><\/td>\n<\/tr>\n<tr>\n<td>Trainer \u4e0d\u662f\u5b50\u7c7b<\/td>\n<td>\u6ca1\u6709\u7ee7\u627f <code>nnUNetTrainer<\/code><\/td>\n<td>\u786e\u8ba4 <code>class MyTrainer(nnUNetTrainer)<\/code><\/td>\n<\/tr>\n<tr>\n<td>\u8bad\u7ec3\u80fd\u8dd1\uff0c\u63a8\u7406\u62a5\u9519<\/td>\n<td>\u63a8\u7406\u73af\u5883\u627e\u4e0d\u5230\u540c\u4e00\u4e2a\u81ea\u5b9a\u4e49 Trainer<\/td>\n<td>\u4fdd\u8bc1\u63a8\u7406\u673a\u5668\u4e5f\u80fd import \u8be5 Trainer<\/td>\n<\/tr>\n<tr>\n<td>\u6539\u4e86\u5f88\u591a\u5730\u65b9\u540e\u4e0d\u77e5\u9053\u54ea\u91cc\u574f\u4e86<\/td>\n<td>\u4e00\u6b21\u6539\u52a8\u8303\u56f4\u592a\u5927<\/td>\n<td>\u6bcf\u6b21\u53ea\u6539\u4e00\u4e2a\u5165\u53e3\uff0c\u4f8b\u5982\u5148\u53ea\u6539 optimizer<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h2>9. \u54ea\u4e9b\u5185\u5bb9\u9002\u5408\u5728 Trainer \u4e2d\u6539<\/h2>\n<table>\n<thead>\n<tr>\n<th>\u76ee\u6807<\/th>\n<th>\u63a8\u8350\u5165\u53e3<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>\u6362 optimizer \u6216 lr scheduler<\/td>\n<td><code>configure_optimizers<\/code><\/td>\n<\/tr>\n<tr>\n<td>\u6362 loss<\/td>\n<td><code>_build_loss<\/code><\/td>\n<\/tr>\n<tr>\n<td>\u6539 augmentation<\/td>\n<td><code>get_training_transforms<\/code> \u6216\u76f8\u5173 transform \u6784\u5efa\u903b\u8f91<\/td>\n<\/tr>\n<tr>\n<td>\u6539\u7f51\u7edc\u7ed3\u6784<\/td>\n<td><code>build_network_architecture<\/code>\uff0c\u5e76\u5904\u7406 deep supervision \u517c\u5bb9<\/td>\n<\/tr>\n<tr>\n<td>\u6539\u8bad\u7ec3 epoch\u3001\u5b66\u4e60\u7387\u7b49\u8d85\u53c2\u6570<\/td>\n<td><code>__init__<\/code> \u4e2d\u4fee\u6539\u5bf9\u5e94\u5c5e\u6027<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h2>10. \u5b98\u65b9\u8d44\u6599\u5165\u53e3<\/h2>\n<p>\u672c\u6587\u4e3b\u8981\u53c2\u8003\uff1a<\/p>\n<ul>\n<li><a href=\"https:\/\/github.com\/MIC-DKFZ\/nnUNet\/blob\/master\/documentation\/extending_nnunet.md\" target=\"_blank\"  rel=\"nofollow\" >Extending nnU-Net<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/MIC-DKFZ\/nnUNet\/blob\/master\/documentation\/how-to\/train-models.md\" target=\"_blank\"  rel=\"nofollow\" >Train Models<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/MIC-DKFZ\/nnUNet\/blob\/master\/nnunetv2\/run\/run_training.py\" target=\"_blank\"  rel=\"nofollow\" >run_training.py \u5b98\u65b9\u6e90\u7801<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/MIC-DKFZ\/nnUNet\/blob\/master\/nnunetv2\/utilities\/find_objects.py\" target=\"_blank\"  rel=\"nofollow\" >find_objects.py \u5b98\u65b9\u6e90\u7801<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/MIC-DKFZ\/nnUNet\/blob\/master\/nnunetv2\/training\/nnUNetTrainer\/nnUNetTrainer.py\" target=\"_blank\"  rel=\"nofollow\" >nnUNetTrainer.py \u5b98\u65b9\u6e90\u7801<\/a><\/li>\n<li><a href=\"https:\/\/github.com\/MIC-DKFZ\/nnUNet\/blob\/master\/nnunetv2\/training\/nnUNetTrainer\/variants\/training_length\/nnUNetTrainer_Xepochs.py\" target=\"_blank\"  rel=\"nofollow\" >\u5b98\u65b9 Trainer \u53d8\u4f53\u793a\u4f8b<\/a><\/li>\n<\/ul>\n<h2>\u672c\u7bc7\u603b\u7ed3<\/h2>\n<p>\u4fee\u6539 nnU-Net v2 \u8bad\u7ec3\u6d41\u7a0b\u7684\u63a8\u8350\u65b9\u5f0f\u662f\u7ee7\u627f <code>nnUNetTrainer<\/code>\uff0c\u521b\u5efa\u81ea\u5df1\u7684 Trainer \u7c7b\uff0c\u7136\u540e\u5728\u8bad\u7ec3\u547d\u4ee4\u4e2d\u7528 <code>-tr<\/code> \u6307\u5b9a\u3002\u8fd9\u6837\u65e2\u80fd\u590d\u7528\u5b98\u65b9\u5b8c\u6574\u8bad\u7ec3\u6846\u67b6\uff0c\u53c8\u80fd\u8ba9\u5b9e\u9a8c\u6539\u52a8\u6e05\u6670\u3001\u53ef\u590d\u73b0\u3001\u53ef\u56de\u6eda\u3002\u672c\u6587\u7528 AdamW \u793a\u4f8b\u6f14\u793a\u4e86\u6700\u5c0f\u4fee\u6539\u65b9\u5f0f\uff0c\u540e\u7eed\u6211\u4eec\u4f1a\u7ee7\u7eed\u7528 Trainer \u4fee\u6539 loss \u548c augmentation\u3002<\/p>\n<h2>\u4e0b\u4e00\u7bc7\u9884\u544a<\/h2>\n<p>\u4e0b\u4e00\u7bc7\u6211\u4eec\u4f1a\u4e13\u95e8\u4fee\u6539 loss\uff1a\u7406\u89e3 nnU-Net \u9ed8\u8ba4\u7684 Dice + CE \/ Dice + BCE \u7ec4\u5408\u3001deep supervision wrapper\uff0c\u4ee5\u53ca\u600e\u6837\u5728\u81ea\u5b9a\u4e49 Trainer \u4e2d\u66ff\u6362\u6216\u7ec4\u5408\u81ea\u5df1\u7684 loss\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u672c\u7bc7\u4ece\u81ea\u5b9a\u4e49 Trainer \u5f00\u59cb\uff0c\u6f14\u793a\u5982\u4f55\u7ee7\u627f nnUNetTrainer\u3001\u8986\u76d6\u4f18\u5316\u5668\u914d\u7f6e\uff0c\u5e76\u7528 -tr \u8c03\u7528\u81ea\u5df1\u7684 Trainer\u3002<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"emotion":"","emotion_color":"","title_style":"","license":"","footnotes":""},"categories":[83],"tags":[],"class_list":["post-1057","post","type-post","status-publish","format-standard","hentry","category-83"],"views":2,"_links":{"self":[{"href":"https:\/\/www.eutaboo.com\/index.php\/wp-json\/wp\/v2\/posts\/1057","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.eutaboo.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.eutaboo.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.eutaboo.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.eutaboo.com\/index.php\/wp-json\/wp\/v2\/comments?post=1057"}],"version-history":[{"count":0,"href":"https:\/\/www.eutaboo.com\/index.php\/wp-json\/wp\/v2\/posts\/1057\/revisions"}],"wp:attachment":[{"href":"https:\/\/www.eutaboo.com\/index.php\/wp-json\/wp\/v2\/media?parent=1057"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.eutaboo.com\/index.php\/wp-json\/wp\/v2\/categories?post=1057"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.eutaboo.com\/index.php\/wp-json\/wp\/v2\/tags?post=1057"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}