hydraは階層構造を持つ設定を扱うためのPythonフレームワーク このサンプルが分かりやすい 設定を外部ファイル（yaml形式）として保持 メインプログラムは設定ファイルを読み込んで，インスタンスを生成 このjupyter notebookを見ると，設定ファイルを読み込…

改めてWindowsはクソ． 手順としては，以下のようにインストールしていく． cuDNNのインストールディレクトリを環境変数に設定しなくてもPyTorchのサンプルプログラムは動いたけど，本当に大丈夫なのかは謎． NVIDIA Driverのインストール NVIDIA Driverを最…

ここを参考にして，以下の構成でCUDA, gpytorch, botorch, axのインストールに成功 nvidia-driver: 470 CUDA: 11.1 cuDNN: 8.20 PyTorch: 1.9.0の組み合わせ GPyTorch: botorch: 0. バックアップ timeshiftをインストールして，システムバックアップを取る．…

Summary about Anaconda's channel priority to avoid unexpected version change of some packages manage channel priority by conda config --add/--prepend/--append channels new_channel -. You can check the official information here The problem …

This post describes how to setup PyTorch 1.0 with CUDA 10.0 and cuDNN 7.5 on Ubuntu 18.04. Step 0 uninstall CUDA 10.1 related components# uninstall PyTorchconda uninstall pytorch torchvision cudatoolkit# uninstall cuDNNsudo dpkg -r libcudn…

The official site provides conda package with the combination of CUDA toolkit 10.0 and cuDNN 7.4. My installation is CUDA toolkit 10.1 and cuDNN 7.5.After the installation, a sample program worked, however I may need to downgrade the toolk…

I just wrote a simple code to visualize trained filters and feature maps of pytorch. For simplicity, the below code uses pretrained AlexNet but the code must work with any network with Conv2d layers. [code lang="python"] !/usr/bin/env pyth…

See the tutorial. Important note: torchvision provides several pre-trained models with their trained parameters. AlexNet, DenseNet, Inception, ResNet, VGG are available, see here. With pretrained=True, pre-trained parameters are available.…

I got a runtime error at evaluation phase of my simple siamese network as RuntimeError: cuda runtime error (2) : out of memory at the line of executing forward pass. Some comments on stackoverflow suggests to define x and y with volatile=T…

See this tutorial. Set requires_grad=True for all tensors we compute gradients w.r.t. them. [code lang='python'] input and output x = torch.randn(N, D_in, device=device, dtype=dtype) y = torch.randn(N, D_out, device=device, dtype=dtype) pa…

See the tutorial. It's quite simple (Keras provides simpler interface though). A neural network is defined as a class that has a forward function. The forward function defines how input data is processed through the network. PyTorch automa…

Super easy installation GUI on the web as below