kastnerkyle

import gzip

def gzip_search(query: str, candidate_chunks: list[str], top_k: int=1):
    """
    文字列ベースで類似したテキストチャンクを推定するアルゴリズム.
    `query`, `chunk`, および`query + " " + chunk`をそれぞれgzipで圧縮し、編集距離のようなものをベースに評価する.

    Parameters:
        query (str): 検索クエリとして使用する文字列.
        top_k (int, optional): 返される類似チャンクの上位k個を指定する (default: 1).

kastnerkyle

# Author: Kyle Kastner
# BSD 3-Clause

# Thanks to jakevdp for the nice blog post on FFT
# https://jakevdp.github.io/blog/2013/08/28/understanding-the-fft/
# Summary
# http://www.arazim-project.com/sites/default/files/public/lesson_sums/1fft.pdf
# Details on hartley and many xforms
# https://caxapa.ru/thumbs/455725/algorithms.pdf
# pg 332 http://sep.stanford.edu/data/media/public/oldreports/sep38/38_29.pdf

kastnerkyle

import torch
import k_diffusion as K

from PIL import Image
from torch import autocast
from einops import rearrange, repeat

def pil_img_to_latent(model, img, batch_size=1, device='cuda', half=True):
    init_image = pil_img_to_torch(img, half=half).to(device)
    init_image = repeat(init_image, '1 ... -> b ...', b=batch_size)

kastnerkyle

import math
import torch
from torch import optim


class AdamWFinetune(optim.Optimizer):
    r"""Implements AdamW algorithm with optional weight decay toward the starting value, to
    prevent overfitting to the new dataset during fine-tuning.

    The original Adam algorithm was proposed in `Adam: A Method for Stochastic Optimization`_.

kastnerkyle

def typical_top_k_filtering(logits, top_k=0, top_p=0.0, temperature=1.0, min_tokens_to_keep=1, filter_value=-1E12):
    """ Filter a distribution of logits using typicality, with optional top-k and/or nucleus (top-p) filtering
        Meister et. al. https://arxiv.org/abs/2202.00666
        Args:
            logits: logits distribution shape (..., vocabulary size)
            top_k >0: keep top k tokens with highest prob (top-k filtering).
            top_p >0.0: keep the top p tokens which compose cumulative probability mass top_p (nucleus filtering).
            min_tokens_to_keep >=1: always keep at least this many tokens through the top_p / nucleus sampling
    """
    # https://arxiv.org/abs/2202.00666

kastnerkyle

{
    "seconds_per_quarter": 0.5, 
    "parts_names": [
        "Soprano", 
        "Alto", 
        "Tenor", 
        "Bass"
    ], 
    "parts_cumulative_times": [
        [

kastnerkyle

import numpy as np

# make a minibatch of time, batch, features
# time length 7 
# batch size 2
# feature dimension 4: 
# 1:4, 10:14, 20:24, 30:34, etc for first minibatch element
# 5:8, 15:18, etc second minibatch el
n_features = 4
n_timesteps = 7

kastnerkyle

kastnerkyle

#!/usr/bin/env python3
"""
To use:
1. install/set-up the google cloud api and dependencies listed on https://github.com/GoogleCloudPlatform/python-docs-samples/tree/master/texttospeech/cloud-client
2. install pandoc and pypandoc, also tqdm
3. create and download a service_account.json ("Service account key") from https://console.cloud.google.com/apis/credentials
4. run GOOGLE_APPLICATION_CREDENTIALS=service_account.json python make_audiobook.py book_name.epub
"""
import re
import sys

kastnerkyle

#!/usr/bin/env python3
"""
To use:
1. install/set-up the google cloud api and dependencies listed on https://github.com/GoogleCloudPlatform/python-docs-samples/tree/master/texttospeech/cloud-client
2. install pandoc and pypandoc, also tqdm
3. create and download a service_account.json ("Service account key") from https://console.cloud.google.com/apis/credentials
4. run GOOGLE_APPLICATION_CREDENTIALS=service_account.json python make_audiobook.py book_name.epub
"""
import re
import sys