Ryuichiro Hataya moskomule

ツイートを消す

#pipx install hatch
hatch run delete_tweets.py API_KEY API_KEY_SECRET ACCESS_TOKEN ACCESS_TOKEN_SECRET PATH_TO_TWEETS.JS

API_KEY, API_KEY_SECRET, ACCESS_TOKEN, ACCESS_TOKEN_SECRETはhttps://zenn.dev/shomtsm/articles/52864a4b7b5eecを参考に　https://developer.twitter.com/en/portal/petition/essential/basic-info から取得．
PATH_TO_TWEETS.JS はTwitterからダウンロードしたアーカイブ内の data/tweets.js または data/tweet-headers.js までのパス．

以下をuBlacklistに加える

https://gist.githubusercontent.com/moskomule/3bddfc5deb5845fccc0d7c2519c600ce/raw/uBlacklist.txt

raytracer

A simple Python raytracer that supports spheres with configurable "material" properties (base color and a bunch of light coefficients). To generate a raytraced image of the pre-defined scene, run: python raytracer.py and open image.ppm with a PPM-compatible viewer (eog works fine on Linux):

acknowledgements

I found the following resources extremely helpful:

utility for downloading webvision2018 dataset

Requires gnu parallel and aria2c. Both can be downloaded via brew

wget -O grive https://docs.google.com/uc?id=0B3X9GlR6EmbnQ0FtZmJJUXEyRTA
chmod 755 grive # not sure if it's necessary
./gdrive upload [--recursive] NAME

	import torch.utils.data as data
	import os
	import re
	import torch
	import tarfile
	from PIL import Image


	IMG_EXTENSIONS = ['.png', '.jpg', '.jpeg']

	cat results//.json \| grep -o -E 'args": "[^"]+' \| grep -o -E '\-\-loss.*' >f1.txt
	cat results//.json \| grep -o -E 'args": "[^"]+' \| grep -o -E '\-\-dataset[^\-]+' > f2.txt
	cat results//.json \| grep -o -E 'miou_test": \[\[[^"]+' \| grep -o -E '29, [^]]+' > f3.txt
	paste f1.txt f2.txt f3.txt \| sed -e 's/--loss/\|/g' \| sed -e 's/--batch_size/\|/g' \| sed -e 's/29,/\|/g'\| sed -e 's/--dataset/\|/g' \| sort

	#
	# \| ce \| 1 \| sbd \| 0.7214421629905701
	# \| ce \| 1 \| voc \| 0.6946691870689392
	# \| ce \| 2 \| sbd \| 0.7422053217887878
	# \| ce \| 2 \| voc \| 0.7283321619033813

	def Rop(y, x, v):
	"""Computes an Rop.

	Arguments:
	y (Variable): output of differentiated function
	x (Variable): differentiated input
	v (Variable): vector to be multiplied with Jacobian from the right
	"""
	w = torch.ones_like(y, requires_grad=True)
	return torch.autograd.grad(torch.autograd.grad(y, x, w), w, v)