echogarden 1.4.4 → 1.6.0

package/src/text-translation/NLLBTextTranslation.ts

@@ -1,47 +1,80 @@
+import chalk from 'chalk'
+import { TranslationPair } from '../api/TextTranslation.js'
+import { splitToSentences } from '../nlp/Segmentation.js'
 import { Logger } from '../utilities/Logger.js'
 import { loadPackage } from '../utilities/PackageManager.js'
 
-export async function translateText(sourceText: string, sourceLanguage: string, targetLanguage: string) {
+export async function translateText(sourceText: string, sourceLanguage: string, targetLanguage: string): Promise<TranslationPair[]> {
 	const logger = new Logger()
 
-	const { AutoTokenizer,  M2M100ForConditionalGeneration } = await import('@echogarden/transformers-nodejs-lite')
+	//const languageNames = Object.keys(languageNameToNLLBCode)
 
-	const modelPath = await loadPackage(`xenova-nllb-200-distilled-600M-quantized`)
+	//logger.log(languageNames)
 
+	logger.start(`Load transformers.js module`)
+	const { AutoTokenizer, M2M100ForConditionalGeneration } = await import('@echogarden/transformers-nodejs-lite')
+
+	logger.start(`Load NLLB package`)
+	const modelPath = await loadPackage(`xenova-nllb-200-distilled-600M-q8`)
+
+	logger.start(`Load NLLB tokenizer`)
 	const tokenizer = await AutoTokenizer.from_pretrained(modelPath)
+
+	logger.start(`Load NLLB model`)
 	const model = await M2M100ForConditionalGeneration.from_pretrained(modelPath)
 
+	logger.start(`Split to sentences`)
+
 	const config = {
-		src_lang: 'eng_Latn',
-		tgt_lang: 'fra_Latn'
+		src_lang: 'spa_Latn',
+		tgt_lang: 'eng_Latn'
 	}
 
-	const inputs = (tokenizer as any)._build_translation_inputs(sourceText, {
-		padding: true,
-		truncation: true,
-	}, config)
+	const sentences = splitToSentences(sourceText, sourceLanguage)
+
+	logger.end()
 
-	const result = await model.generate(inputs.input_ids, config)
+	const translationPairs: TranslationPair[] = []
 
-	logger.log(tokenizer.model.convert_ids_to_tokens(result[0]))
+	for (let i = 0; i < sentences.length; i++) {
+		const sentence = sentences[i]
 
-	const inputTokens = tokenizer.model.convert_ids_to_tokens(Array.from(inputs.input_ids.data))
-	const embeddingResult = await model(inputs)
+		logger.logTitledMessage(`Translate sentence ${i + 1}/${sentences.length}`, `"${sentence.trim()}"`, chalk.magentaBright)
 
-	const lastHiddenState = embeddingResult.last_hidden_state
+		logger.start(`Tokenize sentence`)
+		const inputs = (tokenizer as any)._build_translation_inputs(sentence, {
+			padding: true,
+			truncation: true,
+		}, config)
 
-	const tokenCount = lastHiddenState.dims[1]
-	const embeddingSize = lastHiddenState.dims[2]
+		logger.start(`Translate sentence with NLLB model`)
+		const translationTokenIds = await model.generate(inputs.input_ids, config)
 
-	for (let i = 0; i < tokenCount; i++) {
-		const tokenEmbedding = lastHiddenState.data.slice(i * embeddingSize, (i + 1) * embeddingSize)
+		logger.start(`Extract tokens`)
 
-		const tokenId = inputTokens[i]
+		const translationTokens = tokenizer.model.convert_ids_to_tokens(translationTokenIds[0])
 
-		logger.log(`Token ${i} (${tokenId}):`, tokenEmbedding);
+		const translatedText = translationTokens
+			.slice(2, translationTokens.length - 1)
+			.map(token => {
+				if (token.startsWith('▁')) {
+					return token.replaceAll('▁', ' ')
+				}
+
+				return token
+			})
+			.join('')
+			.trim()
+
+		translationPairs.push({
+			sourceText: sentence,
+			translatedText
+		})
+
+		logger.end()
 	}
 
-	logger.log(inputTokens)
+	return translationPairs
 }
 
 const languageNameToNLLBCode: Record<string, string> = {
@@ -250,3 +283,210 @@ const languageNameToNLLBCode: Record<string, string> = {
 	'Yue Chinese': 'yue_Hant',
 	'Zulu': 'zul_Latn',
 }
+
+const languageNameToISO931: { [langName: string]: string } = {
+	'Acehnese (Arabic script)': 'unknown', // No ISO 639-1 code for Acehnese
+	'Acehnese (Latin script)': 'unknown', // No ISO 639-1 code for Acehnese
+	'Afrikaans': 'af',
+	'Akan': 'ak',
+	'Amharic': 'am',
+	'Armenian': 'hy',
+	'Assamese': 'as',
+	'Asturian': 'ast', // Approximate: ISO 639-3 code
+	'Awadhi': 'awa', // Approximate: ISO 639-3 code
+	'Ayacucho Quechua': 'qu', // Approximate: Quechua has many variants
+	'Balinese': 'ban', // Approximate: ISO 639-3 code
+	'Bambara': 'bm',
+	'Banjar (Arabic script)': 'unknown', // No ISO 639-1 code for Banjar
+	'Banjar (Latin script)': 'unknown', // No ISO 639-1 code for Banjar
+	'Bashkir': 'ba',
+	'Basque': 'eu',
+	'Belarusian': 'be',
+	'Bemba': 'bem', // Approximate: ISO 639-3 code
+	'Bengali': 'bn',
+	'Bhojpuri': 'bho',
+	'Bosnian': 'bs',
+	'Buginese': 'bug',
+	'Bulgarian': 'bg',
+	'Burmese': 'my',
+	'Catalan': 'ca',
+	'Cebuano': 'ceb', // Approximate: ISO 639-2 code
+	'Central Atlas Tamazight': 'tzm',
+	'Central Aymara': 'ay', // Approximate: Aymara has variants
+	'Central Kanuri (Arabic script)': 'kr', // Approximate: Kanuri uses multiple scripts
+	'Central Kanuri (Latin script)': 'kr', // Approximate: Kanuri uses multiple scripts
+	'Central Kurdish': 'ckb', // Approximate: Kurdish has several variants
+	'Chhattisgarhi': 'hne', // Approximate: ISO 639-3 code
+	'Chinese (Simplified)': 'zh', // Approximate: zh covers both Simplified and Traditional
+	'Chinese (Traditional)': 'zh', // Approximate: zh covers both Simplified and Traditional
+	'Chokwe': 'cjk', // Approximate: ISO 639-3 code
+	'Crimean Tatar': 'crh', // Approximate: ISO 639-3 code
+	'Croatian': 'hr',
+	'Czech': 'cs',
+	'Danish': 'da',
+	'Dari': 'prs', // Approximate: Dari is considered a dialect of Persian
+	'Dutch': 'nl',
+	'Dyula': 'dyu',
+	'Dzongkha': 'dz',
+	'Eastern Panjabi': 'pa', // Approximate: Panjabi has multiple writing systems
+	'Eastern Yiddish': 'yid', // Approximate: Yiddish has multiple variants
+	'Egyptian Arabic': 'arz',
+	'English': 'en',
+	'Esperanto': 'eo',
+	'Estonian': 'et',
+	'Ewe': 'ee',
+	'Faroese': 'fo',
+	'Fijian': 'fj',
+	'Finnish': 'fi',
+	'Fon': 'fon',
+	'French': 'fr',
+	'Friulian': 'fur', // Approximate: ISO 639-3 code
+	'Galician': 'gl',
+	'Ganda': 'lg',
+	'Georgian': 'ka',
+	'German': 'de',
+	'Greek': 'el',
+	'Guarani': 'gn',
+	'Gujarati': 'gu',
+	'Haitian Creole': 'ht',
+	'Halh Mongolian': 'mn', // Approximate: Mongolian encompasses several dialects
+	'Hausa': 'ha',
+	'Hebrew': 'he',
+	'Hindi': 'hi',
+	'Hungarian': 'hu',
+	'Icelandic': 'is',
+	'Igbo': 'ig',
+	'Ilocano': 'ilo',
+	'Indonesian': 'id',
+	'Irish': 'ga',
+	'Italian': 'it',
+	'Japanese': 'ja',
+	'Javanese': 'jv',
+	'Jingpho': 'kac', // Approximate: ISO 639-3 code
+	'Kabiyè': 'kbp', // Approximate: ISO 639-3 code
+	'Kabuverdianu': 'kea', // Approximate: ISO 639-3 code
+	'Kabyle': 'kab',
+	'Kamba': 'kam', // Approximate: ISO 639-3 code
+	'Kannada': 'kn',
+	'Kashmiri (Arabic script)': 'ks', // Approximate: Kashmiri uses multiple scripts
+	'Kashmiri (Devanagari script)': 'ks', // Approximate: Kashmiri uses multiple scripts
+	'Kazakh': 'kk',
+	'Khmer': 'km',
+	'Kikongo': 'kg',
+	'Kikuyu': 'ki',
+	'Kimbundu': 'kmb',
+	'Kinyarwanda': 'rw',
+	'Korean': 'ko',
+	'Kyrgyz': 'ky',
+	'Lao': 'lo',
+	'Latgalian': 'ltg', // Approximate: ISO 639-3 code
+	'Ligurian': 'lij', // Approximate: ISO 639-3 code
+	'Limburgish': 'li', // Approximate: ISO 639-1 code for Limburgish-Ripuarian
+	'Lingala': 'ln',
+	'Lithuanian': 'lt',
+	'Lombard': 'lmo', // Approximate: ISO 639-3 code
+	'Luba-Kasai': 'lua',
+	'Luo': 'luo',
+	'Luxembourgish': 'lb',
+	'Macedonian': 'mk',
+	'Magahi': 'mag', // Approximate: ISO 639-3 code
+	'Maithili': 'mai', // Approximate: ISO 639-3 code
+	'Malayalam': 'ml',
+	'Maltese': 'mt',
+	'Maori': 'mi',
+	'Marathi': 'mr',
+	'Meitei (Bengali script)': 'mni', // Approximate: Meitei uses multiple scripts
+	'Mesopotamian Arabic': 'acm', // Approximate: ISO 639-3 code
+	'Minangkabau (Arabic script)': 'min', // Approximate: Minangkabau uses multiple scripts
+	'Minangkabau (Latin script)': 'min', // Approximate: Minangkabau uses multiple scripts
+	'Mizo': 'lus',
+	'Modern Standard Arabic (Romanized)': 'ar', // Approximate: Modern Standard Arabic is a standardized form
+	'Modern Standard Arabic': 'ar', // Approximate: Modern Standard Arabic is a standardized form
+	'Moroccan Arabic': 'ary',
+	'Mossi': 'mos',
+	'Najdi Arabic': 'ars', // Approximate: ISO 639-3 code for Najdi Arabic
+	'Nepali': 'ne',
+	'Nigerian Fulfulde': 'fuv',
+	'North Azerbaijani': 'az', // Approximate: Azerbaijani is a pluricentric language
+	'North Levantine Arabic': 'apc', // Approximate: ISO 639-3 code for North Levantine Arabic
+	'Northern Kurdish': 'kmr', // Approximate: Kurdish has several variants
+	'Northern Sotho': 'nso',
+	'Northern Uzbek': 'uz', // Approximate: Uzbek is a pluricentric language
+	'Norwegian Bokmål': 'nb',
+	'Norwegian Nynorsk': 'nn',
+	'Nuer': 'nus', // Approximate: ISO 639-3 code
+	'Nyanja': 'ny',
+	'Occitan': 'oc',
+	'Odia': 'or',
+	'Pangasinan': 'pag',
+	'Papiamento': 'pap',
+	'Plateau Malagasy': 'plt', // Approximate: ISO 639-3 code for Plateau Malagasy
+	'Polish': 'pl',
+	'Portuguese': 'pt',
+	'Romanian': 'ro',
+	'Rundi': 'rn',
+	'Russian': 'ru',
+	'Samoan': 'sm',
+	'Sango': 'sg',
+	'Sanskrit': 'sa',
+	'Santali': 'sat',
+	'Sardinian': 'sc', // Approximate: Sardinian has several variants
+	'Scottish Gaelic': 'gd',
+	'Serbian': 'sr',
+	'Shan': 'shn', // Approximate: ISO 639-3 code
+	'Shona': 'sn',
+	'Sicilian': 'scn', // Approximate: ISO 639-3 code
+	'Silesian': 'szl', // Approximate: ISO 639-3 code
+	'Sindhi': 'sd',
+	'Sinhala': 'si',
+	'Slovak': 'sk',
+	'Slovenian': 'sl',
+	'Somali': 'so',
+	'South Azerbaijani': 'az', // Approximate: Azerbaijani is a pluricentric language
+	'South Levantine Arabic': 'ajp', // Approximate: ISO 639-3 code for South Levantine Arabic
+	'Southern Pashto': 'pst', // Approximate: Pashto has several variants
+	'Southern Sotho': 'st',
+	'Southwestern Dinka': 'dik', // Approximate: ISO 639-3 code for Southwestern Dinka
+	'Spanish': 'es',
+	'Standard Latvian': 'lv', // Approximate: Standard Latvian is the official form of Latvian
+	'Standard Malay': 'ms', // Approximate: Standard Malay is the basis for both Malaysian and Indonesian
+	'Standard Tibetan': 'bo', // Approximate: Tibetan has several variants
+	'Sundanese': 'su',
+	'Swahili': 'sw',
+	'Swati': 'ss',
+	'Swedish': 'sv',
+	'Tagalog': 'tl',
+	'Tajik': 'tg',
+	'Tamasheq (Latin script)': 'tmh', // Approximate: Tamasheq uses multiple scripts
+	'Tamasheq (Tifinagh script)': 'tmh', // Approximate: Tamasheq uses multiple scripts
+	'Tamil': 'ta',
+	'Tatar': 'tt',
+	'Ta’izzi-Adeni Arabic': 'acq', // Approximate: ISO 639-3 code for Ta'izzi-Adeni Arabic
+	'Telugu': 'te',
+	'Thai': 'th',
+	'Tigrinya': 'ti',
+	'Tok Pisin': 'tpi',
+	'Tosk Albanian': 'sq', // Approximate: Albanian encompasses both Tosk and Gheg dialects
+	'Tsonga': 'ts',
+	'Tswana': 'tn',
+	'Tumbuka': 'tum',
+	'Tunisian Arabic': 'aeb', // Approximate: ISO 639-3 code for Tunisian Arabic
+	'Turkish': 'tr',
+	'Turkmen': 'tk',
+	'Twi': 'tw',
+	'Ukrainian': 'uk',
+	'Umbundu': 'umb',
+	'Urdu': 'ur',
+	'Uyghur': 'ug',
+	'Venetian': 'vec', // Approximate: ISO 639-3 code
+	'Vietnamese': 'vi',
+	'Waray': 'war',
+	'Welsh': 'cy',
+	'West Central Oromo': 'om', // Approximate: Oromo has several variants
+	'Western Persian': 'fa', // Approximate: Western Persian is the most common variety of Persian
+	'Wolof': 'wo',
+	'Xhosa': 'xh',
+	'Yoruba': 'yo',
+	'Yue Chinese': 'yue', // Approximate: ISO 639-3 code
+	'Zulu': 'zu'
+};

package/src/typings/Fillers.d.ts

@@ -1,6 +1,25 @@
 declare module 'moving-median'
-declare module 'html-to-text'
-declare module 'cldr-segmentation'
+
+declare module 'html-to-text' {
+	export function htmlToText(html: string, options: {
+		wordwrap: boolean | number
+		selectors?: any[]
+	})
+}
+
+declare module 'cldr-segmentation' {
+	export function sentenceSplit(text: string, suppressions: any)
+	export function wordSplit(text: string, suppressions: any)
+	export const suppressions: Record<string, Suppressions>
+
+	export class Suppressions {
+		constructor(forwardTrie: any, backwardTrie: any, list: string)
+
+		merge(other: Suppressions)
+
+		static create(list: string[]): Suppressions
+	}
+}
 
 declare module 'html-escaper' {
 	export function escape(str: string): string
@@ -46,10 +65,12 @@ declare module 'onnxruntime-node' {
 
 	// From backend.d.ts
 	import { Backend, InferenceSession, SessionHandler, OnnxValue } from 'onnxruntime-common'
+
 	class OnnxruntimeBackend implements Backend {
 		init(): Promise<void>
 		createSessionHandler(pathOrBuffer: string | Uint8Array, options?: InferenceSession.SessionOptions): Promise<SessionHandler>
 	}
+
 	export const onnxruntimeBackend: OnnxruntimeBackend
 
 	// From binding.d.ts
@@ -76,8 +97,10 @@ declare module 'onnxruntime-node' {
 		interface InferenceSession {
 			loadModel(modelPath: string, options: SessionOptions): void
 			loadModel(buffer: ArrayBuffer, byteOffset: number, byteLength: number, options: SessionOptions): void
+
 			readonly inputNames: string[]
 			readonly outputNames: string[]
+
 			run(feeds: FeedsType, fetches: FetchesType, options: RunOptions): ReturnType
 		}
 		interface InferenceSessionConstructor {

package/src/utilities/BrowserRequestHeaders.ts

@@ -0,0 +1,59 @@
+export function getChromeOnWindowsHeaders(options: BrowserRequestHeadersOptions) {
+	const headers: Record<string, string> = {
+		'Sec-Ch-Ua': `"Google Chrome";v="125", "Chromium";v="125", "Not.A/Brand";v="24"`,
+		'X-Same-Domain': '1',
+		'Dnt': '1',
+		'Sec-Ch-Ua-Mobile': '?0',
+		'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/125.0.0.0 Safari/537.36',
+		'Sec-Ch-Ua-Arch': 'x86',
+		'Sec-Ch-Ua-Full-Version': '125.0.6422.114',
+		'Sec-Ch-Ua-Platform-Version': '"15.0.0"',
+		'Sec-Ch-Ua-Full-Version-List': `"Google Chrome";v="125.0.6422.114", "Chromium";v="125.0.6422.114", "Not.A/Brand";v="24.0.0.0"`,
+		'Sec-Ch-Ua-Bitness': '"64"',
+		'Sec-Ch-Ua-Model': '""',
+		'Sec-Ch-Ua-Platform': '"Windows"',
+		'Accept': '*/*',
+		'Origin': options.origin,
+		'Sec-Fetch-Site': 'same-origin',
+		'Sec-Fetch-Mode': 'cors',
+		'Sec-Fetch-Dest': 'empty',
+		'Referer': options.referrer,
+		'Accept-Encoding': 'gzip, deflate, br',
+		'Accept-Language': 'en-US,en;q=0.9',
+	}
+
+	return headers
+}
+
+export function getChromeOnAndroidHeaders(options: BrowserRequestHeadersOptions) {
+	const headers: Record<string, string> = {
+		'Sec-Ch-Ua': `"Not/A)Brand";v="8", "Chromium";v="126", "Google Chrome";v="126"`,
+		'Sec-Ch-Ua-Mobile': '?1',
+		'Sec-Ch-Ua-Full-Version': '"126.0.6478.122"',
+		'Sec-Ch-Ua-Arch': `""`,
+		'Sec-Ch-Ua-Platform': `"Android"`,
+		'Sec-Ch-Ua-Platform-Version': `"13"`,
+		'Sec-Ch-Ua-Model': `"Pixel 5"`,
+		'Sec-Ch-Ua-Bitness': `"64"`,
+		'Sec-Ch-Ua-Wow64': `?0`,
+		'Dnt': '1',
+		'Upgrade-Insecure-Requests': '1',
+		'User-Agent': `Mozilla/5.0 (Linux; Android 13; K) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/126.0.6478.122 Mobile Safari/537.36`,
+		'Accept': '*/*',
+		'Origin': options.origin,
+		'Sec-Fetch-Site': 'none',
+		'Sec-Fetch-Mode': 'navigate',
+		'Sec-Fetch-User': '?1',
+		'Sec-Fetch-Dest': 'document',
+		'Referer': options.referrer,
+		'Accept-Encoding': 'gzip, deflate, br',
+		'Accept-Language': 'en-US,en;q=0.9',
+	}
+
+	return headers
+}
+
+export interface BrowserRequestHeadersOptions {
+	origin: string
+	referrer: string
+}

package/src/utilities/DynamicUint8Array.ts

@@ -0,0 +1,39 @@
+export class DynamicUint8Array {
+	data = new Uint8Array(4)
+	length = 0
+
+	add(element: number) {
+		this.ensureCapacity(this.length + 1)
+
+		this.data[this.length] = element
+		this.length += 1
+	}
+
+	addMany(...elements: number[]) {
+		this.ensureCapacity(this.length + elements.length)
+
+		for (const element of elements) {
+			this.data[this.length] = element
+			this.length += 1
+		}
+	}
+
+	ensureCapacity(requiredCapacity: number) {
+		if (requiredCapacity > this.data.length) {
+			const newCapacity = requiredCapacity * 2
+
+			const newData = new Uint8Array(newCapacity)
+			newData.set(this.toUint8Array())
+
+			this.data = newData
+		}
+	}
+
+	toUint8Array() {
+		return this.data.subarray(0, this.length)
+	}
+
+	clear() {
+		this.length = 0
+	}
+}

package/src/utilities/Hashing.ts

@@ -1,170 +1,4 @@
-export function knuthMultiplicative(bytes: Buffer) {
-	let hash = 0
-
-	for (const byte of bytes) {
-		hash += Math.imul(byte, 2654435761)
-	}
-
-	return hash
-}
-
-export function xorShift32Hash(bytes: Buffer) {
-	let s = 0
-
-	for (const byte of bytes) {
-		s += byte
-
-		s ^= s << 13
-		s ^= s >> 17
-		s ^= s << 5
-	}
-
-	return s
-}
-
-export function jenkinsOneAtATime(bytes: Buffer) {
-	let hash = 0
-
-	for (const byte of bytes) {
-		hash += byte
-		hash += hash << 10
-		hash ^= hash >> 6
-	}
-
-	hash += hash << 3
-	hash ^= hash >> 11
-	hash += hash << 15
-
-	return hash >>> 0
-}
-
-export function FNV1a(bytes: Buffer) {
-	let hval = 2166136261 | 0
-
-	for (const byte of bytes) {
-		hval = Math.imul(hval ^ byte, 16777619)
-	}
-
-	return hval >>> 0
-}
-
-export function superFastHash(bytes: Buffer) {
-	let hash = bytes.length, tmp, p = 0
-	const len = bytes.length >>> 2
-
-	for (let i = 0; i < len; i++) {
-		hash += bytes[p] | bytes[p + 1] << 8
-		tmp = ((bytes[p + 2] | bytes[p + 3] << 8) << 11) ^ hash
-		hash = (hash << 16) ^ tmp
-		hash += hash >>> 11
-		p += 4
-	}
-
-	switch (bytes.length & 3) {
-		case 3:
-			hash += bytes[p] | bytes[p + 1] << 8
-			hash ^= hash << 16
-			hash ^= bytes[p + 2] << 18
-			hash += hash >>> 11
-			break
-		case 2:
-			hash += bytes[p] | bytes[p + 1] << 8
-			hash ^= hash << 11
-			hash += hash >>> 17
-			break
-		case 1:
-			hash += bytes[p]
-			hash ^= hash << 10
-			hash += hash >>> 1
-			break
-	}
-
-	hash ^= hash << 3
-	hash += hash >>> 5
-	hash ^= hash << 4
-	hash += hash >>> 17
-	hash ^= hash << 25
-	hash += hash >>> 6
-
-	return hash >>> 0
-}
-
-export function cyrb53Hash(bytes: Buffer, seed = 0) {
-	// https://github.com/bryc/code/blob/master/jshash/experimental/cyrb53.js
-
-	let h1 = 0xdeadbeef ^ seed
-	let h2 = 0x41c6ce57 ^ seed
-
-	for (const byte of bytes) {
-		h1 = Math.imul(h1 ^ byte, 2654435761)
-		h2 = Math.imul(h2 ^ byte, 1597334677)
-	}
-
-	h1 = Math.imul(h1 ^ (h1 >>> 16), 2246822507) ^ Math.imul(h2 ^ (h2 >>> 13), 3266489909)
-	h2 = Math.imul(h2 ^ (h2 >>> 16), 2246822507) ^ Math.imul(h1 ^ (h1 >>> 13), 3266489909)
-
-	return (4294967296 * (2097151 & h2)) + (h1 >>> 0)
-}
-
-export function djb2(bytes: Buffer) {
-	let hash = 5381
-
-	for (const byte of bytes) {
-		//hash += (hash << 5) + byte
-		hash += (hash * 33) + byte
-	}
-
-	return hash >>> 0
-}
-
-
-export function murmurHash1(bytes: Buffer, seed = 0) {
-	// https://github.com/bryc/code/blob/master/jshash/hashes/murmurhash1.js
-
-	const length = bytes.length
-
-	const multiplier = 3332679571
-	const intIterationMaxIndex = length & -4
-
-	let hash = seed ^ Math.imul(length, multiplier)
-	let index = 0
-
-	for (; index < intIterationMaxIndex; index += 4) {
-		hash += bytes[index + 3] << 24 |
-			bytes[index + 2] << 16 |
-			bytes[index + 1] << 8 |
-			bytes[index]
-
-		hash = Math.imul(hash, multiplier)
-		hash ^= hash >>> 16
-	}
-
-	switch (length & 3) {
-		case 3: {
-			hash += bytes[index + 2] << 16
-		}
-
-		case 2: {
-			hash += bytes[index + 1] << 8
-		}
-
-		case 1: {
-			hash += bytes[index]
-			hash = Math.imul(hash, multiplier)
-			hash ^= hash >>> 16
-		}
-	}
-
-	hash = Math.imul(hash, multiplier)
-	hash ^= hash >>> 10
-
-	hash = Math.imul(hash, multiplier)
-	hash ^= hash >>> 17
-
-	return hash >>> 0
-}
-
-export function MurmurHash3(bytes: Buffer, seed = 0) {
+export function murmurHash3(bytes: Uint8Array, seed = 0) {
 	// https://github.com/bryc/code/blob/master/jshash/hashes/murmurhash3.js
 
 	const p1 = 3432918353
@@ -228,3 +62,16 @@ export function MurmurHash3(bytes: Buffer, seed = 0) {
 
 	return hash >>> 0
 }
+
+export function convertToSingleInt32Hash(hash: (bytes: Uint8Array) => number) {
+	const bytes = new Uint8Array(4)
+	const ints = new Int32Array(bytes.buffer)
+
+	return (int32Val: number): number => {
+		ints[0] = int32Val
+
+		return hash(bytes)
+	}
+}
+
+export const murmurHash3_int32Input = convertToSingleInt32Hash(murmurHash3)