AIbase
EN
Home
Internal.wav2vec2 Base Superb Ks Int8 Structured79
I

Internal.wav2vec2 Base Superb Ks Int8 Structured79

Developed by yujiepan
This model is a fine-tuned version of wav2vec2-base-ft-keyword-spotting on the superb dataset for audio classification tasks, optimized through quantization and structured pruning.
Audio Classification
Transformers
Open Source License:Apache-2.0#Speech Keyword Recognition#Quantization and Pruning Optimization#High Accuracy Audio Classification
Downloads 16
Release Time : 3/14/2023

Model Overview

This is a fine-tuned and optimized audio classification model based on the wav2vec2 architecture, excelling in keyword recognition tasks.

Model Features

High Accuracy
Achieved 98.15% accuracy on the evaluation set
Optimization Processing
Optimized through quantization and structured pruning (80% sparsity in Transformer block linear layers)
Efficient Training
Training process optimized using mixed precision training and gradient accumulation techniques

Model Capabilities

Audio Classification
Keyword Recognition

Use Cases

Voice Interaction
Voice Command Recognition
Recognize keywords in device control commands
High accuracy recognition
Audio Analysis
Audio Content Classification
Classify audio clips
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
EnglishįŽ€äŊ“中文įšéĢ”中文ãĢãģんご
Š 2025AIbase