echogarden 1.7.0 → 1.8.1

package/src/source-separation/MDXNetSourceSeparation.ts

@@ -6,6 +6,7 @@ import { Logger } from '../utilities/Logger.js'
 import { OnnxExecutionProvider, dmlProviderAvailable, getOnnxSessionOptions } from '../utilities/OnnxUtilities.js'
 import chalk from 'chalk'
 import { WindowedList } from '../utilities/WindowedList.js'
+import { logLevelGreaterOrEqualTo } from '../api/API.js'
 
 const log = logToStderr
 
@@ -13,9 +14,9 @@ export async function isolate(
 	rawAudio: RawAudio,
 	modelFilePath: string,
 	modelProfile: MDXNetModelProfile,
-	executionProviders: OnnxExecutionProvider[]) {
+	options: MDXNetOptions) {
 
-	const model = new MDXNet(modelFilePath, modelProfile, executionProviders)
+	const model = new MDXNet(modelFilePath, modelProfile, options)
 
 	return model.processAudio(rawAudio)
 }
@@ -27,11 +28,11 @@ export class MDXNet {
 	constructor(
 		public readonly modelFilePath: string,
 		public readonly modelProfile: MDXNetModelProfile,
-		public readonly executionProviders: OnnxExecutionProvider[]) {
+		public readonly options: MDXNetOptions) {
 	}
 
 	async processAudio(rawAudio: RawAudio) {
-		if (rawAudio.audioChannels.length != 2) {
+		if (rawAudio.audioChannels.length !== 2) {
 			throw new Error(`Input audio must be stereo`)
 		}
 
@@ -43,12 +44,18 @@ export class MDXNet {
 			return getEmptyRawAudio(rawAudio.audioChannels.length, rawAudio.sampleRate)
 		}
 
+		const enableTraceLogging = logLevelGreaterOrEqualTo('trace')
+
 		const logger = new Logger()
 
-		await logger.startAsync(`Initialize MDXNet model`)
+		await logger.startAsync(`Initialize session for MDX-NET model '${this.options.model!}'`)
 
 		await this.initializeSessionIfNeeded()
 
+		logger.end()
+
+		logger.logTitledMessage(`Using ONNX execution provider`, `${this.onnxSessionOptions!.executionProviders!.join(', ')}`)
+
 		const Onnx = await import('onnxruntime-node')
 
 		const sampleRate = this.modelProfile.sampleRate
@@ -81,7 +88,11 @@ export class MDXNet {
 
 			const timePosition = segmentStartFrameOffset * (fftHopSize / sampleRate)
 
-			await logger.startAsync(`Compute STFT of segment at time position ${timePosition.toFixed(2)}`, undefined, chalk.magentaBright)
+			if (enableTraceLogging) {
+				await logger.startAsync(`Compute STFT of segment at time position ${timePosition.toFixed(2)}`, undefined, chalk.magentaBright)
+			} else {
+				await logger.startAsync(`Process segment at time position ${timePosition.toFixed(2)}`)
+			}
 
 			while (fftFramesLeftWindowedList.endOffset < segmentEndFrameOffset) {
 				const nextLeftFrameResult = await fftFramesLeftGenerator.next()
@@ -109,7 +120,9 @@ export class MDXNet {
 
 			const isLastSegment = segmentLength < segmentSize
 
-			await logger.startAsync(`Reshape STFT frames`)
+			if (enableTraceLogging) {
+				await logger.startAsync(`Reshape STFT frames`)
+			}
 
 			const flattenedInputTensor = new Float32Array(1 * 4 * binCount * segmentSize)
 
@@ -142,13 +155,17 @@ export class MDXNet {
 				}
 			}
 
-			await logger.startAsync(`Process with MDXNet model (ONNX provider: ${this.onnxSessionOptions!.executionProviders!.join(', ')})`)
+			if (enableTraceLogging) {
+				await logger.startAsync(`Process segment with MDXNet model`)
+			}
 
 			const inputTensor = new Onnx.Tensor('float32', flattenedInputTensor, [1, 4, binCount, segmentSize])
 
 			const { output: outputTensor } = await this.session!.run({ input: inputTensor })
 
-			await logger.startAsync('Reshape processed frames')
+			if (enableTraceLogging) {
+				await logger.startAsync('Reshape processed frames')
+			}
 
 			const flattenedOutputTensor = outputTensor.data as Float32Array
 
@@ -191,7 +208,9 @@ export class MDXNet {
 
 			const outputAudioChannels: Float32Array[] = []
 
-			await logger.startAsync(`Compute inverse STFT of model output for segment`)
+			if (enableTraceLogging) {
+				await logger.startAsync(`Compute inverse STFT of model output for segment`)
+			}
 
 			for (let channelIndex = 0; channelIndex < 2; channelIndex++) {
 				const samples = await stiftr(
@@ -268,13 +287,16 @@ export class MDXNet {
 
 		const Onnx = await import('onnxruntime-node')
 
-		this.onnxSessionOptions = getOnnxSessionOptions({ executionProviders: this.executionProviders })
+		const executionProviders: OnnxExecutionProvider[] =
+			this.options.provider ? [this.options.provider] : getDefaultMDXNetProviders()
+
+		this.onnxSessionOptions = getOnnxSessionOptions({ executionProviders })
 
 		this.session = await Onnx.InferenceSession.create(this.modelFilePath, this.onnxSessionOptions)
 	}
 }
 
-export function getDefaultMDXNetProviders() : OnnxExecutionProvider[] {
+export function getDefaultMDXNetProviders(): OnnxExecutionProvider[] {
 	if (dmlProviderAvailable()) {
 		return ['dml', 'cpu']
 	} else {
@@ -344,3 +366,13 @@ export type MDXNetModelName =
 	'UVR_MDXNET_Main' |
 	'Kim_Vocal_1' |
 	'Kim_Vocal_2'
+
+export interface MDXNetOptions {
+	model?: MDXNetModelName
+	provider?: OnnxExecutionProvider
+}
+
+export const defaultMDXNetOptions: MDXNetOptions = {
+	model: 'UVR_MDXNET_1_9703',
+	provider: undefined,
+}

package/src/typings/Fillers.d.ts

@@ -56,6 +56,8 @@ declare module '@echogarden/rnnoise-wasm'
 declare module '@echogarden/fvad-wasm'
 declare module '@echogarden/sonic-wasm'
 declare module '@echogarden/kissfft-wasm'
+declare module '@echogarden/pffft-wasm'
+declare module '@echogarden/pffft-wasm/simd'
 declare module '@echogarden/speex-resampler-wasm'
 declare module '@echogarden/vosk'
 

package/src/utilities/PackageManager.ts

@@ -192,13 +192,9 @@ const packageVersionTagResolutionLookup: { [packageName: string]: string } = {
 	'whisper.cpp-large-v3-turbo-q5_0': '20241003',
 
 	// whisper.cpp binaries
-	'whisper.cpp-binaries-linux-x64-cpu-1.6.2': '20241001',
-	'whisper.cpp-binaries-windows-x64-cpu-1.6.2': '20241001',
-	'whisper.cpp-binaries-windows-x64-cublas-12.4.0-1.6.2': '20241002',
-
-	'whisper.cpp-binaries-linux-x64-cpu-latest': '20241003',
-	'whisper.cpp-binaries-windows-x64-cpu-latest': '20241003',
-	'whisper.cpp-binaries-windows-x64-cublas-12.4.0-latest': '20241003',
+	'whisper.cpp-binaries-linux-x64-cpu-latest': '20241010',
+	'whisper.cpp-binaries-windows-x64-cpu-latest': '20241010',
+	'whisper.cpp-binaries-windows-x64-cublas-12.4.0-latest': '20241010',
 
 	// E5 models
 	'xenova-multilingual-e5-small-q8': '20240504',

package/src/utilities/WasmMemoryManager.ts

@@ -1,10 +1,35 @@
 export class WasmMemoryManager {
 	wasmModule: any
 
+	private wasmAlloc: WasmAllocMethod
+	private wasmFree: WasmFreeMethod
+
 	private allocatedReferences = new Set<WasmRef>()
 
-	constructor(wasmModule: any) {
+	constructor(wasmModule: any, options?: WasmMemoryManagerOptions) {
+		options = options ?? {}
+
 		this.wasmModule = wasmModule
+
+		if (options.wasmAlloc) {
+			this.wasmAlloc = options.wasmAlloc
+		} else {
+			if (!wasmModule._malloc) {
+				throw new Error(`Couldn't find a '_malloc' function in the module and no custom 'wasmAlloc' was provided in the options`)
+			}
+
+			this.wasmAlloc = wasmModule._malloc
+		}
+
+		if (options.wasmFree) {
+			this.wasmFree = options.wasmFree
+		} else {
+			if (!wasmModule._free) {
+				throw new Error(`Couldn't find a '_malloc' function in the module and no custom 'wasmFree' was provided in the options`)
+			}
+
+			this.wasmFree = wasmModule._free
+		}
 	}
 
 	allocInt8() {
@@ -209,8 +234,9 @@ export class WasmMemoryManager {
 	}
 
 	private alloc(size: number) {
-		const ptr = this.wasmModule._malloc(size)
-		return ptr as number
+		const ptr = this.wasmAlloc(size)
+
+		return ptr
 	}
 
 	free(wasmReference: WasmRef) {
@@ -218,7 +244,7 @@ export class WasmMemoryManager {
 			return
 		}
 
-		this.wasmModule._free(wasmReference.address)
+		this.wasmFree(wasmReference.address)
 
 		this.allocatedReferences.delete(wasmReference)
 		wasmReference.clearAddress()
@@ -514,3 +540,11 @@ export class Float64ArrayRef extends TypedArrayRef<Float64Array> {
 
 export type TypedArray = Int8Array | Uint8Array | Uint8ClampedArray | Int16Array | Uint16Array | Int32Array | Uint32Array | Float32Array | Float64Array
 export type WasmRef = ValueRef<number> | ValueRef<string> | TypedArrayRef<TypedArray>
+
+export interface WasmMemoryManagerOptions {
+	wasmAlloc?: WasmAllocMethod
+	wasmFree?: WasmFreeMethod
+}
+
+export type WasmAllocMethod = (size: number) => number
+export type WasmFreeMethod = (address: number) => void