https://elated-raman-42e0c2.netlify.app/tags/deep_learning/Recent content in Deep_learning on Sawyer Zheng's BlogHugozh-cnFri, 20 Jun 2025 16:04:15 +0800https://elated-raman-42e0c2.netlify.app/post/notes/ai/deeplearning/Sun, 04 Feb 2024 00:00:00 +0000https://elated-raman-42e0c2.netlify.app/post/notes/ai/deeplearning/<div id="outline-container-headline-1" class="outline-2"> <h2 id="headline-1"> 激活函数 </h2> <div id="outline-text-headline-1" class="outline-text-2"> <p>参考：</p> <ul> <li><a href="https://towardsdatascience.com/sigmoid-and-softmax-functions-in-5-minutes-f516c80ea1f9">Sigmoid and SoftMax Functions in 5 minutes | by Gabriel Furnieles | Towards D…</a></li> </ul> <p>类型：</p> <ul> <li> <p>线性激活函数</p> <ul> <li>易训练</li> <li>不适合复杂任务</li> </ul> </li> <li> <p>非线性激活函数</p> <ul> <li>sigmoid, tanh, ReLU</li> <li>难训练</li> <li>适合复杂任务</li> </ul> </li> </ul> <div id="outline-container-headline-2" class="outline-3"> <h3 id="headline-2"> sigmoid, tanh, ReLU 比较 </h3> <div id="outline-text-headline-2" class="outline-text-3"> <p>参考：</p> <ul> <li><a href="https://zhuanlan.zhihu.com/p/427473958?utm_id=0">原来ReLU这么好用！一文带你深度了解ReLU激活函数！ - 知乎</a></li> </ul> <ol> <li> <p>易训练性：</p>https://elated-raman-42e0c2.netlify.app/post/notes/ai/train/Wed, 09 Aug 2023 00:00:00 +0000https://elated-raman-42e0c2.netlify.app/post/notes/ai/train/<div id="outline-container-headline-1" class="outline-2"> <h2 id="headline-1"> 关于拆分数据多机器 </h2> </div>https://elated-raman-42e0c2.netlify.app/post/notes/python/pytorch/Tue, 08 Aug 2023 00:00:00 +0000https://elated-raman-42e0c2.netlify.app/post/notes/python/pytorch/<div id="outline-container-headline-1" class="outline-2"> <h2 id="headline-1"> 安装 </h2> <div id="outline-text-headline-1" class="outline-text-2"> <ul> <li><a href="https://pytorch.org/get-started/locally/">Start Locally | PyTorch</a></li> </ul> </div> </div> <div id="outline-container-headline-2" class="outline-2"> <h2 id="headline-2"> torchrun 训练 </h2> <div id="outline-text-headline-2" class="outline-text-2"> <div id="outline-container-headline-3" class="outline-3"> <h3 id="headline-3"> 错误问题 </h3> <div id="outline-text-headline-3" class="outline-text-3"> <div id="outline-container-headline-4" class="outline-4"> <h4 id="headline-4"> 突然中断 sighup 1 </h4> <div id="outline-text-headline-4" class="outline-text-4"> <ol> <li> <p>原因</p> <ul> <li>使用 nohup + torchrun 后台运行导致</li> </ul> </li> <li> <p>解决方法</p> <ul> <li>参考： <a href="https://discuss.pytorch.org/t/ddp-error-torch-distributed-elastic-agent-server-api-received-1-death-signal-shutting-down-workers/135720/7">DDP Error: torch.distributed.elastic.agent.server.api:Received 1 death signal…</a></li> <li>使用 tmux 或者 screen 命令把 torchrun 后台运行</li> </ul> </li> </ol> </div> </div> </div> </div> </div> </div> <div id="outline-container-headline-5" class="outline-2"> <h2 id="headline-5"> torch.Tensor </h2> <div id="outline-text-headline-5" class="outline-text-2"> <div id="outline-container-headline-6" class="outline-3"> <h3 id="headline-6"> Tensor.scatter_ 用法 </h3> <div id="outline-text-headline-6" class="outline-text-3"> <p>参考：</p>