๋ฐ˜์‘ํ˜•

๐Ÿ’ฉ ์—๋Ÿฌ ํ•ด๊ฒฐ 7

VSCode Remote-SSH ์œˆ๋„์šฐ ์„œ๋ฒ„ ๋ฌดํ•œ๋กœ๋”ฉ, downloading server

1. ๋ฌธ์ œ VSCode์—์„œ Remote-SSH๋ฅผ ์ด์šฉํ•ด ์œˆ๋„์šฐ ์„œ๋ฒ„๋กœ ์›๊ฒฉ์—ฐ๊ฒฐ์„ ํ•˜๋ ค๊ณ  ํ•˜๋Š”๋ฐ downloading server์—์„œ ๊ณ„์† ๋ฉˆ์ถฐ์žˆ๊ณ  ์ง„ํ–‰์ด ์•ˆ๋๋‹ค. ์ด์ƒํ•œ๊ฑด ๋ฆฌ๋ˆ…์Šค ์„œ๋ฒ„๋Š” ์—ฐ๊ฒฐ์ด ์ž˜ ๋˜๊ณ , ๋˜ ํ„ฐ๋ฏธ๋„์—์„œ ssh๋กœ ์œˆ๋„์šฐ ์„œ๋ฒ„์— ์ ‘์†ํ•˜๋ฉด ์ž˜๋˜๋Š”๋ฐ Remote-SSH๋กœ ์œˆ๋„์šฐ์— ์ ‘์†ํ•  ๋•Œ๋งŒ ์ ‘์†ํ•  ๋•Œ๋งŒ ์•ˆ๋˜๋Š” ๊ฒƒ.. ๊ตฌ๊ธ€์— ๊ฒ€์ƒ‰ํ•ด๋ณด๋‹ˆ Remote-SSH ๋ฒ„์ „์„ ๋‚ฎ์ถ”๋ผ๋Š” ์–˜๊ธฐ๋„ ์žˆ์–ด์„œ ํ•ด๋ดค๋Š”๋ฐ ๋˜‘๊ฐ™์•˜๋‹ค. 2. ํ•ด๊ฒฐ๊ณผ์ • ์œˆ๋„์šฐ ์„œ๋ฒ„์˜ vscode-server์—์„œ ๋ญ”๊ฐ€ ๋ฌธ์ œ๊ฐ€ ์žˆ์–ด๋ณด์—ฌ์„œ ํฌ๋กฌ ์›๊ฒฉ ๋ฐ์Šคํฌํƒ‘์œผ๋กœ ๋ณด๋ฉด์„œ ๋‹ค์‹œ Remote-SSH๋ฅผ ์‹คํ–‰ํ•ด ๋ดค๋‹ค. V3์—์„œ vscode-server์„ ์„ค์น˜ํ•˜๋ ค๋Š” powershell์„ ์•…์„ฑ ํ–‰์œ„๋ผ๊ณ  ํƒ์ง€ํ•ด์„œ ๊ฐ•์ œ์ข…๋ฃŒ์‹œ์ผฐ๋‹ค. ๊ทธ๋ž˜์„œ ๊ณ„์† downloading์—์„œ..

[PyTorch/์—๋Ÿฌ ํ•ด๊ฒฐ] Dataparallel์ด complex tensor์„ real view๋กœ ์ „ํ™˜์‹œํ‚ค๋Š” ๋ฌธ์ œ

๋ฌธ์ œ: complex tensor์„ input์œผ๋กœ ๋ฐ›๋Š” ๋ชจ๋ธ์„ ์‚ฌ์šฉ ์ค‘์ด์—ˆ๊ณ , forward method๋ฅผ ํ…Œ์ŠคํŠธ ํ•  ๋•Œ๋Š” ์ž˜ ๋Œ์•„๊ฐ€๋‹ค๊ฐ€ ์ „์ฒด train ์ฝ”๋“œ๋ฅผ ๋Œ๋ ธ๋”๋‹ˆ tensor ์ฐจ์›์ด ์•ˆ๋งž๋Š”๋‹ค๋Š” ์—๋Ÿฌ๋ฅผ ๋‚ด๋ฑ‰์—ˆ๋‹ค.. RuntimeError: The size of tensor a (2) must match the size of tensor b (232) at non-singleton dimension 3 ๋ฐ”๋กœ ์ด๋ ‡๊ฒŒ.. ๋””๋ฒ„๊น…์„ ํ•ด๋ณด๋‹ˆ complex tensor๊ฐ€ model ๋‚ด๋ถ€๋กœ ๋“ค์–ด๊ฐ€๋ฉด float์œผ๋กœ ๋ณ€ํ™˜๋˜๋ฉด์„œ real-imag part๊ฐ€ ๋ถ„๋ฆฌ๋˜๋Š” ๊ฒƒ์ด์—ˆ๋‹ค,, ๋”ฐ๋กœ model forward ์ฝ”๋“œ๋งŒ ๋Œ๋ฆด๋•Œ๋Š” ์ž˜๋งŒ ๋Œ์•„๊ฐ”๋Š”๋ฐ ? ์›์ธ: ๊ฒฐ๋ก ์€.. nn.DataParallel์ด ๋ฌธ์ œ์˜€๋‹ค nn.DataPa..

[Tensorflow/์—๋Ÿฌ ํ•ด๊ฒฐ] Memory growth cannot differ between GPU devices

์ถœ์ฒ˜ : https://fixexception.com/tensorflow/memory-growth-cannot-differ-between-gpu-devices/ ๋ฉ”๋ชจ๋ฆฌ ์ฆ๊ฐ€๋ฅผ ํ—ˆ์šฉ์„ ๊ฒฐ์ •ํ•˜๋Š” ํ•จ์ˆ˜์ธ tf.config.experimental.set_memory_growth๋ฅผ ์ด์šฉํ•  ๋•Œ ๋‹ค์Œ๊ณผ ๊ฐ™์€ ์—๋Ÿฌ๊ฐ€ ๋ฐœ์ƒํ•  ์ˆ˜ ์žˆ๋‹ค. ValueError: Memory growth cannot differ between GPU devices ์ด๋Š” gpu๊ฐ€ ์—ฌ๋Ÿฌ ๊ฐœ ์žˆ์„ ๋•Œ, ๋ชจ๋“  gpu์— ๋Œ€ํ•ด memory growth๋ฅผ ๋™์ผํ•˜๊ฒŒ ์„ธํŒ…ํ•ด์ฃผ์ง€ ์•Š์•„์„œ ๋ฐœ์ƒํ•˜๋Š” ์—๋Ÿฌ์ด๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด gpu๊ฐ€ ๋‘ ๋Œ€ ์žˆ๋Š”๋ฐ ํ•˜๋‚˜์˜ gpu์—๋งŒ set_memory_growth๋ฅผ True๋กœ ์„ค์ •ํ•ด์ฃผ์—ˆ์„ ๊ฒฝ์šฐ์ด๋‹ค. ์ด ๊ฒฝ์šฐ์—๋Š” ๋ชจ๋“  gpu์— ๋Œ€ํ•ด se..

[Tensorflow/์—๋Ÿฌ ํ•ด๊ฒฐ] Tensorflow 1.x ๋ฒ„์ „์„ ์‚ฌ์šฉ ์ค‘์ž„์—๋„ "deprecated" warning์ด ๋‚˜ํƒ€๋‚˜๋Š” ๋ฌธ์ œ

Tensorflow 1.x ๋ฒ„์ „์„ ์‚ฌ์šฉ ์ค‘์ž„์—๋„ ๋ถˆ๊ตฌํ•˜๊ณ , Tensorflow 2์—์„œ ์‚ฌ๋ผ์ง„ ํ•จ์ˆ˜๋“ค์„ ํ˜ธ์ถœํ•  ๋•Œ warning ๋ฉ”์„ธ์ง€๊ฐ€ ๋–ด๋‹ค. ์ฒจ์—๋Š” ๊ทธ๋ƒฅ ๋ฌด์‹œํ•˜๊ณ  ์ง„ํ–‰ํ•˜๋ ค ํ–ˆ์ง€๋งŒ.. ๋„ˆ๋ฌด ๊ฑฐ์Šฌ๋ฆฐ๋‹ค ๋‹ค์Œ Tensorflow issue์— ์—ฌ๋Ÿฌ ๊ฐ€์ง€ ํ•ด๊ฒฐ ๋ฐฉ๋ฒ•๋“ค์ด ๋‚˜์™€ ์žˆ๋‹ค (๊ทธ๋ƒฅ TF2.0์„ ์‚ฌ์šฉํ•˜๋ผ๋Š” ๋‹ต๋ณ€์ด ๋น„์ถ” 72๊ฐœ๋ฅผ ๋ฐ›์•˜๋‹ค) https://github.com/tensorflow/tensorflow/issues/27023 TF 1.x: remove the "deprecated" warning messages · Issue #27023 · tensorflow/tensorflow I know the functional APIs, such as tf.layers.dense, will disappear in..

[pip/์—๋Ÿฌ ํ•ด๊ฒฐ] inplace-abn ์„ค์น˜ ์‹œ ์˜ค๋ฅ˜

pip๋ฅผ ์ด์šฉํ•ด inplace-abn์„ ์„ค์น˜ํ•˜๋ ค๊ณ  ํ–ˆ๋Š”๋ฐ ๋‹ค์Œ๊ณผ ๊ฐ™์€ ์˜ค๋ฅ˜๊ฐ€ ๋ฐœ์ƒํ–ˆ๋‹ค. >> pip install inplace-abn Collecting inplace-abn Using cached inplace-abn-1.1.0.tar.gz (137 kB) Building wheels for collected packages: inplace-abn Building wheel for inplace-abn (setup.py) ... error ERROR: Command errored out with exit status 1: command: /home/user/anaconda3/bin/python -u -c 'import sys, setuptools, tokenize; sys.argv[0] = '"'"'..

[PyTorch/์—๋Ÿฌ ํ•ด๊ฒฐ] torchaudio.transforms.SpectralCentroid output์ด nan์ด ๋˜๋Š” ๋ฌธ์ œ

ํ›ˆ๋ จ์„ ์‹œํ‚ค๋‹ค๊ฐ€ loss๊ฐ€ ์ž๊พธ nan์ด ๋˜๋Š” ํ˜„์ƒ์ด ๋ฐœ์ƒํ–ˆ๋‹ค. ๋””๋ฒ„๊น…์„ ํ•ด๋ณด๋‹ˆ loss๊ฐ€ ์ž˜ ์ค„์–ด๋“ค๋‹ค๊ฐ€ ๊ฐ‘์ž๊ธฐ nan์œผ๋กœ ๋ฐ”๋€Œ์–ด์„œ ํ•™์Šต ๊ณผ์ • ์ž์ฒด์˜ ๋ฌธ์ œ๋Š” ์•„๋‹Œ ๊ฒƒ ๊ฐ™์•˜๊ณ , ์‚ดํŽด๋ณด๋‹ˆ ๋ชจ๋ธ ๋‚ด๋ถ€์—์„œ ๊ฐ’์ด nan์ด ๋˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ์žˆ์—ˆ๋‹ค. ๋ชจ๋ธ์—์„œ torchaudio.transforms.SpectralCentroid๋ฅผ ์ด์šฉํ•ด ๋ฐ์ดํ„ฐ์˜ spectral centroid๋ฅผ ์ถ”์ถœํ•˜๋Š” ๊ณผ์ •์ด ์žˆ์—ˆ๋Š”๋ฐ, ์ผ๋ถ€ ๋ฐ์ดํ„ฐ์˜ spectral centroid๊ฐ€ nan์„ ํฌํ•จํ•˜๊ณ  ์žˆ์—ˆ๋‹ค. ์ฝ”๋“œ๋ฅผ ์‚ดํŽด๋ณด๋‹ˆ spectral centroid๋ฅผ ๊ณ„์‚ฐํ•˜๋Š” ๊ณผ์ •์—์„œ spectrogram์˜ sum์œผ๋กœ ๊ฐ’์„ ๋‚˜๋ˆ ์ฃผ๋Š” ๊ณผ์ •์ด ์žˆ์—ˆ๋Š”๋ฐ, (๋งˆ์ง€๋ง‰ ์ค„) ์ด ๊ฐ’์ด 0์ด ๋˜์–ด์„œ nan๊ฐ’์ด ๋‚˜์˜จ ๊ฒƒ์ด์—ˆ๋‹ค. Solution torch.nan_to_num์„ ..

[PyTorch/์—๋Ÿฌ ํ•ด๊ฒฐ] PyTorch-Tensorboard์™€ libprotobuf ๋ฌธ์ œ

๋ฉ€์ฉกํ•˜๊ฒŒ ์ž˜ ๋Œ์•„๊ฐ€๋˜ ์ฝ”๋“œ๊ฐ€ ๊ฐ‘์ž๊ธฐ ์•Œ ์ˆ˜ ์—†๋Š” ์—๋Ÿฌ๊ฐ€ ๋‚˜๋ฉด์„œ ์‹คํ–‰์ด ์•ˆ๋๋‹ค.. ============================================================================ [libprotobuf FATAL ../3rdparty/protobuf/src/google/protobuf/stubs/common.cc:68] This program requires version 3.8.0 of the Protocol Buffer runtime library, but the installed version is 3.5.1. Please update your library. If you compiled the program yourself, make sure that you..

๋ฐ˜์‘ํ˜•