jph00

amaarora

amaarora

jph00

from fastcore.utils import *

host = 8888,'localhost'
sock = start_server(*host)
print(f'Serving on {host}...')
while True:
    conn,addr = sock.accept()
    with conn:
        data = conn.recv(1024)
        print(data.decode())

jph00

All of the python 3.9 standard library

For a version without the collapsible details sections (so you can search the whole thing in your browser), click here.

abc

• abc.ABC
• abc.ABCMeta

jph00

#!/usr/bin/env bash
set -e

echo
if ! [[ $(id -u) = 0 ]]; then
  echo "Please run 'sudo ./install-wireguard.sh'" >&2
  exit 1
fi

read -e -p "Use VPN for *all* internet traffic? [y/n] " -i n ROUTE_ALL

jph00

Using this software requires agreeing to the license agreement.

cuDNN	CUDA9.0	CUDA9.2	CUDA10.0	CUDA10.1	CUDA10.2	CUDA11.0
8.0.2	-	-	-	10.1-8.0.2 (991.32MB)	10.2-8.0.2 (1017.23MB)	11.0-8.0.2 (1.32GB)
7.6.1	9.0-7.6.1 (392.82MB)	9.2-7.6.1 (396.89MB)	[10.0-7.6.1](https://developer.download.nvidia.com/compute/redist/cudnn/v7.6.1/cudnn-10.0-linux-x64-v7.6.1.34.tg

TengdaHan

Multi-node-training on slurm with PyTorch

What's this?

• A simple note for how to start multi-node-training on slurm scheduler with PyTorch.
• Useful especially when scheduler is too busy that you cannot get multiple GPUs allocated, or you need more than 4 GPUs for a single job.
• Requirement: Have to use PyTorch DistributedDataParallel(DDP) for this purpose.
• Warning: might need to re-factor your own code.
• Warning: might be secretly condemned by your colleagues because using too many GPUs.

alexandre01

import torch
x = torch.randn(1, 3, 800, 1200)
n, _, h, w, = x.shape
rnd = torch.rand(2, n, 1, 2).sort(-1).values
r, c = torch.linspace(0, 1, h+2)[None, None], torch.linspace(0, 1, w+2)[None, None]
mask = (((r > rnd[0, :, :, :1]) & (r < rnd[0, :, :, 1:])).unsqueeze(-1) * 
        ((c > rnd[1, :, :, :1]) & (c < rnd[1, :, :, 1:])).unsqueeze(-2))[:, :, 1:-1, 1:-1].expand_as(x)

matthewwardrop

# (c) Matthew Wardrop 2019; Licensed under the MIT license
#
# This script provides the ability to run a notebook in the same Python
# process as this script, allowing it to access to variables created
# by the notebook for other purposes. In most cases, this is of limited
# utility and not a best-practice, but there are some limited cases in
# which this capability is valuable, and this script was created for
# such cases. For all other cases, you are better off using the
# `nbconvert` execution API found @:
# https://nbconvert.readthedocs.io/en/latest/execute_api.html