whisper.cpp/models/README.md
Georgi Gerganov abce28ea99
talk.wasm : move to https://whisper.ggerganov.com/talk
This way, we can share the same models across different WASM examples
and not have to download them for each page
2022-11-24 18:24:06 +02:00

3.2 KiB

Whisper model files in custom ggml format

The original Whisper PyTorch models provided by OpenAI have been converted to custom ggml format in order to be able to load them in C/C++. The conversion has been performed using the convert-pt-to-ggml.py script. You can either obtain the original models and generate the ggml files yourself using the conversion script, or you can use the download-ggml-model.sh script to download the already converted models. Currently, they are hosted on the following locations:

Sample usage:

$ ./download-ggml-model.sh base.en
Downloading ggml model base.en ...
models/ggml-base.en.bin          100%[=============================================>] 141.11M  5.41MB/s    in 22s
Done! Model 'base.en' saved in 'models/ggml-base.en.bin'
You can now use it like this:

  $ ./main -m models/ggml-base.en.bin -f samples/jfk.wav

A third option to obtain the model files is to download them from Hugging Face:

https://huggingface.co/datasets/ggerganov/whisper.cpp/tree/main

Available models

Model Disk Mem SHA
tiny 75 MB ~390 MB bd577a113a864445d4c299885e0cb97d4ba92b5f
tiny.en 75 MB ~390 MB c78c86eb1a8faa21b369bcd33207cc90d64ae9df
base 142 MB ~500 MB 465707469ff3a37a2b9b8d8f89f2f99de7299dac
base.en 142 MB ~500 MB 137c40403d78fd54d454da0f9bd998f78703390c
small 466 MB ~1.0 GB 55356645c2b361a969dfd0ef2c5a50d530afd8d5
small.en 466 MB ~1.0 GB db8a495a91d927739e50b3fc1cc4c6b8f6c2d022
medium 1.5 GB ~2.6 GB fd9727b6e1217c2f614f9b698455c4ffd82463b4
medium.en 1.5 GB ~2.6 GB 8c30f0e44ce9560643ebd10bbe50cd20eafd3723
large 2.9 GB ~4.7 GB b1caaf735c4cc1429223d5a74f0f4d0b9b59a299

Model files for testing purposes

The model files prefixed with for-tests- are empty (i.e. do not contain any weights) and are used by the CI for testing purposes. They are directly included in this repository for convenience and the Github Actions CI uses them to run various sanitizer tests.

Fine-tuned models

There are community efforts for creating fine-tuned Whisper models using extra training data. For example, this blog post describes a method for fine-tuning using Hugging Face (HF) Transformer implementation of Whisper. The produced models are in slightly different format compared to the original OpenAI format. To read the HF models you can use the convert-h5-to-ggml.py script like this:

git clone https://github.com/openai/whisper
git clone https://github.com/ggerganov/whisper.cpp

# clone HF fine-tuned model (this is just an example)
git clone https://huggingface.co/openai/whisper-base.en

# convert the model to ggml
python3 ./whisper.cpp/models/convert-h5-to-ggml.py ./whisper-medium/ ./whisper .