whisper-medium-he Open - source Hebrew Speech Recognition Model - Free Deployment for Accurate Speech - to

Home

Whisper Medium He

Developed by imvladikon

Hebrew automatic speech recognition (ASR) model based on OpenAI Whisper architecture, fine-tuned on Hebrew datasets

Speech Recognition

Transformers

OtherOpen Source License:Apache-2.0 #Hebrew speech recognition #Multilingual support #Low word error rate

Downloads 17

Release Time : 11/19/2023

Model Overview

This model is the Hebrew version of the Whisper-medium architecture, specifically designed for Hebrew speech recognition tasks, achieving a 13.4% word error rate on the evaluation set

Model Features

Hebrew optimization

Specially fine-tuned for Hebrew, offering better Hebrew recognition capabilities compared to the original Whisper model

Medium scale

Based on the Whisper-medium architecture, balancing performance and resource consumption

Low word error rate

Achieved a 13.4% word error rate on the evaluation set, demonstrating good performance

Model Capabilities

Hebrew speech-to-text

Automatic speech recognition

Audio transcription

Use Cases

Speech transcription

Hebrew meeting minutes

Automatically transcribe Hebrew meeting recordings into written records

Word error rate approximately 13.4%

Hebrew media subtitle generation

Automatically generate subtitles for Hebrew video content

🚀 whisper-medium-he[WIP]

This model is a fine - tuned version of imvladikon/whisper-medium-he, aiming to provide high - quality automatic speech recognition.

🚀 Quick Start

This model is a fine - tuned version of imvladikon/whisper-medium-he on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.2061
Wer: 13.4020

📚 Documentation

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

🔧 Technical Details

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e - 05
train_batch_size: 1
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 2
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
training_steps: 4000
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0983	0.1	1000	0.3072	16.4362
0.1219	0.2	2000	0.2923	15.6642
0.134	0.3	3000	0.2345	13.7450
0.2113	0.39	4000	0.2061	13.4020

💻 Usage Examples

Basic Usage

HF

from transformers import pipeline
pipe = pipeline("automatic-speech-recognition", model="imvladikon/whisper-medium-he", device_map="auto") # requires `pip install accelerate`
print(recognize("sample.mp3"))

whisper.cpp

Prepared : https://huggingface.co/imvladikon/whisper-medium-he/blob/main/ggml-hebrew.bin

But if need to convert:

git clone https://github.com/openai/whisper
git clone https://github.com/ggerganov/whisper.cpp
git clone https://huggingface.co/imvladikon/whisper-medium-he
python3 ./whisper.cpp/models/convert-h5-to-ggml.py ./whisper-medium-he/ ./whisper .

Then possible to check (if produced model is ggml-model.bin):

cd whisper.cpp && ./main -m ../ggml-model.bin -f ../sample.wav

Framework versions

Transformers 4.36.0.dev0
Pytorch 2.1.0+cu118
Datasets 2.15.0
Tokenizers 0.15.0

📄 License

This model is licensed under the Apache - 2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご