Ayaka ayaka14732

Tensorflow 2.10.0 wheel, built for Python 3.10 with TPU support, compactible with JAX v0.3.6, jaxlib v0.3.5 and libtpu-nightly v0.1.dev20220412.

Get the required Tensorflow commit of jaxlib v0.3.5 from https://github.com/google/jax/blob/jaxlib-v0.3.5/WORKSPACE#L25.

Build: Run the build script.

Visualise a nested Python dictionary by Graphviz

Suppose we have a nested Python dictionary like this:

d = {
    'fc1': {'kernel': (768, 3072), 'bias': (3072,)},
    'fc2': {'kernel': (3072, 768), 'bias': (768,)},
    'self_attn': {
        'k_proj': {'kernel': (768, 768), 'bias': (768,)},

小韻號	小韻首字	上字	下字	被切字音韻描述們	上字音韻描述們	下字音韻描述們
1	東	德	紅	端一東平	端開登入	匣一東平
2	同	徒	紅	定一東平	定模平	匣一東平
3	中	陟	弓	知三東平	知開蒸入	見三東平
4	蟲	直	弓	澄三東平	澄開蒸入	見三東平
5	終	職	戎	章三東平	章開蒸入	日三東平
6	忡	敕	中	徹三東平	徹開蒸入	知三東平
7	崇	鋤	弓	崇三東平	崇魚平	見三東平
8	嵩	息	弓	心三東平	心開蒸入	見三東平
9	戎	如	融	日三東平	日魚平	以三東平

The original links (broken)

https://az412801.vo.msecnd.net/vhd/VMBuild_20141027/VirtualBox/IE7/Linux/IE7.Vista.For.Linux.VirtualBox.zip.001
https://az412801.vo.msecnd.net/vhd/VMBuild_20141027/VirtualBox/IE7/Linux/IE7.Vista.For.Linux.VirtualBox.zip.002
https://az412801.vo.msecnd.net/vhd/VMBuild_20141027/VirtualBox/IE7/Linux/IE7.Vista.For.Linux.VirtualBox.zip.003
https://az412801.vo.msecnd.net/vhd/VMBuild_20141027/VirtualBox/IE7/Linux/IE7.Vista.For.Linux.VirtualBox.zip.004
https://az412801.vo.msecnd.net/vhd/VMBuild_20141027/VirtualBox/IE7/Linux/IE7.Vista.For.Linux.VirtualBox.zip.005

	def is_tpu_in_use() -> bool:
	import fcntl
	from fcntl import LOCK_EX, LOCK_NB
	import posix
	from posix import O_CREAT, O_RDWR
	import sys

	# https://github.com/tensorflow/tensorflow/blob/1a05bad57a2eb870a561daba0255761e0a472d1d/tensorflow/core/tpu/tpu_initializer_helper.cc#L167-L171
	fd = posix.open('/tmp/libtpu_lockfile', O_CREAT \| O_RDWR, 0o644)
	try:

	# monitor the memory profile with `watch --color -n1 go tool pprof -tags /dev/shm/memory.prof`

	import functools
	import jax
	import jax.numpy as np
	import random
	import threading

	devices = jax.devices()
	n_devices = jax.device_count()

	/zh_en.linux
	/char.csv
	/词表.txt
	/结果.txt

	# Generate texts by only selecting the words that starts with the letter s

	import jax
	import jax.nn as nn
	import jax.numpy as np
	import jax.random as rand
	import string
	from transformers import GPT2Tokenizer, FlaxGPT2LMHeadModel

	# filters

	小韻號,韻部原貌,ID,原小韻號,頁號,行號,字號,字頭,上字,下字,被切字音韻描述,被切字母,被切字呼,被切字等,被切字重紐,被切字韻,被切字聲,被切字清濁,被切字組,被切字音,被切字攝,上字音韻描述們,上字母們,上字呼們,上字等們,上字重紐們,上字韻們,上字聲們,上字清濁們,上字組們,上字音們,上字攝們,下字音韻描述們,下字母們,下字呼們,下字等們,下字重紐們,下字韻們,下字聲們,下字清濁們,下字組們,下字音們,下字攝們,備註
	1,東,1077,w10131,2,27,6,蓬,薄,紅,並一東平,並,x,一,x,東,平,全濁,幫,脣,通,並唐入,並,x,一,x,唐,入,全濁,幫,脣,宕,匣一東平,匣,x,一,x,東,平,全濁,影,喉,通,
	2,東,590,w10123,2,20,4,蒙,莫,紅,明一東平,明,x,一,x,東,平,次濁,幫,脣,通,明唐入,明,x,一,x,唐,入,次濁,幫,脣,宕,匣一東平,匣,x,一,x,東,平,全濁,影,喉,通,
	3,東,3,w10101,2,11,1,東,德,紅,端一東平,端,x,一,x,東,平,全清,端,舌,通,端開登入,端,開,一,x,登,入,全清,端,舌,曾,匣一東平,匣,x,一,x,東,平,全濁,影,喉,通,
	4,東,949,w10129,2,25,8,通,他,紅,透一東平,透,x,一,x,東,平,次清,端,舌,通,透開一歌平,透,開,一,x,歌,平,次清,端,舌,果,匣一東平,匣,x,一,x,東,平,全濁,影,喉,通,
	5,東,42,w10102,2,11,3,同,徒,紅,定一東平,定,x,一,x,東,平,全濁,端,舌,通,定模平,定,x,一,x,模,平,全濁,端,舌,遇,匣一東平,匣,x,一,x,東,平,全濁,影,喉,通,
	6,東,698,w10124,2,21,8,籠,盧,紅,來一東平,來,x,一,x,東,平,次濁,x,舌,通,來模平,來,x,一,x,模,平,次濁,x,舌,遇,匣一東平,匣,x,一,x,東,平,全濁,影,喉,通,
	7,東,995,w10130,2,25,12,葼,子,紅,精一東平,精,x,一,x,東,平,全清,精,齒,通,精之上,精,開,三,x,之,上,全清,精,齒,止,匣一東平,匣,x,一,x,東,平,全濁,影,喉,通,
	8,東,893,w10128,2,24,7,怱,倉,紅,清一東平,清,x,一,x,東,平,次清,精,齒,通,清開唐平,清,開