Skip to content

Instantly share code, notes, and snippets.

View jonnyli1125's full-sized avatar

Jonny Li jonnyli1125

39 followers · 26 following
View GitHub Profile
@jonnyli1125
jonnyli1125 / neural_network_from_scratch.ipynb
Last active February 20, 2024 07:46
Implementing and training a neural network from scratch in numpy only
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
@jonnyli1125
jonnyli1125 / rnn_transducer.py
Last active October 9, 2023 17:56
RNN Transducer in ~100 lines of NumPy code. Paper: https://arxiv.org/abs/1211.3711
from dataclasses import dataclass
import numpy as np
vocab = [None, 'a', 'b', 'c']
null_idx = 0
V = len(vocab)
@dataclass
class LSTMWeights:
xi: list[list[float]] # (V-1, V)
@jonnyli1125
jonnyli1125 / .spark-cluster.md
Last active September 16, 2023 20:28
One node Hadoop, Spark, Livy cluster setup with Docker compose that uses the official images for apache/hadoop and apache/spark.

One-node Spark cluster

This is a Docker compose definition of a Hadoop + Spark + Livy cluster with one node only. You can use it for development or testing purposes.

Example usage:

docker compose build
docker compose up -d

docker exec -it  /bin/bash
@jonnyli1125
jonnyli1125 / preprocess_lang8.py
Last active October 20, 2021 02:04
Preprocessing script for Japanese entries in NAIST Lang8 Corpus
import os
import argparse
import json
import re
import unicodedata
invalid_bytes_re = re.compile(r'[\x00-\x1F]+')
sline_re = re.compile(r'\[sline\].*?\[/sline\]')
color_tags = ['[f-blue]','[/f-blue]',
@jonnyli1125
jonnyli1125 / stockfish_selfplay.py
Created June 21, 2021 04:23
Stockfish self-play game generation script
import argparse
import chess
import chess.engine
import chess.pgn
def main(stockfish_path, output_path):
engine = chess.engine.SimpleEngine.popen_uci(stockfish_path)
@jonnyli1125
jonnyli1125 / weighted_sparse_categorical_crossentropy.py
Last active June 11, 2024 08:33
SparseCategoricalCrossentropy with class weights for Keras/Tensorflow 2
"""
Since Model.fit doesn't support class_weight when using multiple outputs,
this custom loss subclass may be useful.
Relevant issues:
https://github.com/keras-team/keras/issues/11735
https://github.com/tensorflow/tensorflow/issues/40457
https://github.com/tensorflow/tensorflow/issues/41448
"""
@jonnyli1125
jonnyli1125 / custom.css
Last active June 19, 2020 04:50
japanese font solution for discord with betterdiscord custom css
div>*:not(code) {
/* using the default discord theme's fonts */
font-family:
Whitney,
'Helvetica Neue',
Helvetica,
Arial,
'Meiryo', /* you can change this to whatever japanese font you prefer */
sans-serif !important;
}
@jonnyli1125
jonnyli1125 / jonnyeclipse.epf
Last active December 14, 2015 03:48
My Eclipse color theme.
file_export_version=3.0
/instance/org.codehaus.groovy.eclipse.ui/groovy.editor.groovyDoc.keyword.enabled=true
/instance/org.codehaus.groovy.eclipse.ui/groovy.editor.groovyDoc.link.enabled=true
/instance/org.codehaus.groovy.eclipse.ui/groovy.editor.groovyDoc.tag.enabled=true
/instance/org.eclipse.ui.editors/AbstractTextEditor.Color.SelectionForeground.SystemDefault=false
/instance/org.eclipse.ui.editors/AbstractTextEditor.Color.SelectionBackground.SystemDefault=false
/instance/org.eclipse.ui.editors/AbstractTextEditor.Color.Background.SystemDefault=false
/instance/org.eclipse.ui.editors/AbstractTextEditor.Color.Foreground.SystemDefault=false
/instance/org.epic.perleditor/AbstractTextEditor.Color.Background.SystemDefault=false
/instance/org.epic.perleditor/AbstractTextEditor.Color.Foreground.SystemDefault=false
@jonnyli1125
jonnyli1125 / BinarySearch.cs
Last active December 12, 2015 01:18
A little test of using a binary search algorithm to find indexes of values in a sorted list. Also just a little comparison between binary and linear/sequential searching.
using System;
using System.Collections.Generic;
namespace BinarySearch
{
class Program
{
static void Main(string[] args)
{
List<int> list = new List<int>() { 5, 8, 900, 1, 22, 52, 567, 666, 123, 256, 765, 0, 500 }; // example list
@jonnyli1125
jonnyli1125 / convert.txt
Created January 27, 2013 19:26
How to convert Minecraft player position and look packets to more readable values.
x = (player.x / 32)
y = (player.y / 32)
z = (player.z / 32)
pitch = round((double)(player.pitch * (360 / 256)))
yaw = round((double)((-(sbyte)(player.yaw)) * (360 / 256)))