Daniel Rothmann

Remove int4 variant which had too poor quality

5ed6654 7 months ago

1.9 kB

import Foundation

/// Configuration constants for the Plapre Pico TTS pipeline.
/// These values are fixed by the model architecture and must match the converted CoreML models.
enum PlapreConfig {
    // MARK: - Audio Parameters
    static let sampleRate = 24_000
    static let audioTokensPerSecond = 25

    // MARK: - Model Architecture
    static let vocabSize = 20_802
    static let maxContextLength = 512
    static let prefillSequenceLength = 512
    static let hiddenDimension = 576
    static let headDimension = 64
    static let numKVHeads = 3
    static let speakerEmbeddingDimension = 128

    // MARK: - Token Ranges
    static let eosToken: Int32 = 0
    static let textMarkerToken: Int32 = 8_000
    static let audioMarkerToken: Int32 = 8_001
    static let audioTokenOffset = 8_002
    static let audioTokenMax = 20_801

    // MARK: - Inference Parameters
    static let defaultTemperature: Float = 0.8
    static let defaultTopK = 50
    static let maxGenerationTokens = 500
    static let nonAudioStopThreshold = 10

    // MARK: - Vocoder Parameters
    static let kanadeChunkSize = 100  // 4 seconds at 25 tok/s

    // MARK: - Paths
    /// Computes the repository root directory from the current source file location.
    /// Traverses: Sources → swift-cli → repo root
    static var repoRoot: URL {
        URL(fileURLWithPath: #filePath)
            .deletingLastPathComponent()  // Sources
            .deletingLastPathComponent()  // swift-cli
            .deletingLastPathComponent()  // repo root
    }

    /// Returns the URL for a model package, accounting for quantization flags.
    static func modelURL(for name: String, useInt8: Bool) -> URL {
        if name == "PlaprePico" && useInt8 {
            return repoRoot.appendingPathComponent("PlaprePico_int8.mlpackage")
        }
        return repoRoot.appendingPathComponent("\(name).mlpackage")
    }
}

	import Foundation

	/// Configuration constants for the Plapre Pico TTS pipeline.
	/// These values are fixed by the model architecture and must match the converted CoreML models.
	enum PlapreConfig {
	// MARK: - Audio Parameters
	static let sampleRate = 24_000
	static let audioTokensPerSecond = 25

	// MARK: - Model Architecture
	static let vocabSize = 20_802
	static let maxContextLength = 512
	static let prefillSequenceLength = 512
	static let hiddenDimension = 576
	static let headDimension = 64
	static let numKVHeads = 3
	static let speakerEmbeddingDimension = 128

	// MARK: - Token Ranges
	static let eosToken: Int32 = 0
	static let textMarkerToken: Int32 = 8_000
	static let audioMarkerToken: Int32 = 8_001
	static let audioTokenOffset = 8_002
	static let audioTokenMax = 20_801

	// MARK: - Inference Parameters
	static let defaultTemperature: Float = 0.8
	static let defaultTopK = 50
	static let maxGenerationTokens = 500
	static let nonAudioStopThreshold = 10

	// MARK: - Vocoder Parameters
	static let kanadeChunkSize = 100 // 4 seconds at 25 tok/s

	// MARK: - Paths
	/// Computes the repository root directory from the current source file location.
	/// Traverses: Sources → swift-cli → repo root
	static var repoRoot: URL {
	URL(fileURLWithPath: #filePath)
	.deletingLastPathComponent() // Sources
	.deletingLastPathComponent() // swift-cli
	.deletingLastPathComponent() // repo root
	}

	/// Returns the URL for a model package, accounting for quantization flags.
	static func modelURL(for name: String, useInt8: Bool) -> URL {
	if name == "PlaprePico" && useInt8 {
	return repoRoot.appendingPathComponent("PlaprePico_int8.mlpackage")
	}
	return repoRoot.appendingPathComponent("\(name).mlpackage")
	}
	}