Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Adding Distil-Whisper - distilled version of Whisper that is 6 times faster, 49% smaller #126

Closed
strich opened this issue Nov 2, 2023 · 2 comments

Comments

@strich
Copy link

strich commented Nov 2, 2023

Per the HN link, it would be great to get Distil-Whisper as an option in this repo!

@sandrohanea
Copy link
Owner

Coming soon with 1.5.0
https://github.com/ggerganov/whisper.cpp/releases/tag/v1.5.0

sandrohanea added a commit that referenced this issue Nov 25, 2023
 - Improved cancellation using CancellationToken provided => it's using the onabort handler in native library to cancel the processing faster
    => addressing concerns from #23

  - Logging improvements: not logging to stderr by default anymore, but events can be registered (see [LogProvider](https://github.com/sandrohanea/whisper.net/tree/main/Whisper.net/Logger/LogProvider))
     => Added optional logging in simple samples + CoreML sample
     => Logging is now suppressed by default: #129

   - Added `useGpu` on WhisperFactory creation so that GPU usage can be disabled on specific runtimes (e.g. Cublas, CoreML, etc)

   - Upgraded dotnet targets to dotnet 8

   - Added support for whisper large v3 (including downloader + coreml model)
       => Old model Large in downloader should be renamed to LargeV2 see [GgmlType](https://github.com/sandrohanea/whisper.net/blob/main/Whisper.net/Ggml/GgmlType.cs)

   - Added linux-x64 support for Whisper.net.Runtime.Cublas

   - Added support for distil-whisper #126
       => No downloader added, you'll need to load the model yourself
       => How to build the model: https://github.com/ggerganov/whisper.cpp/tree/master/models#distilled-models

   - **Added automated builds for all runtimes using Github Actions**

   - Unified runtime builds with make for linux, macos, ios, tvos, wasm and powershell scripts for windows

   - Added Blazor with InteractivityAuto, InteractivityServer and InteractivityWebAssemly sample with Blazor updated in dotnet 8.

   - Unified Whisper.net.Runtime.Wasm with Whisper.net.Runtime
       => If previously you were using Whisper.net.Runtime.Wasm, just replace it with Whisper.net.Runtime

   - Added metal support for arm-based apple platforms (both Whisper.net.Runtime + Whisper.net.Runtime.CoreML)
sandrohanea added a commit that referenced this issue Nov 25, 2023
* - Upgraded to Whisper.cpp 1.5.1

 - Improved cancellation using CancellationToken provided => it's using the onabort handler in native library to cancel the processing faster
    => addressing concerns from #23

  - Logging improvements: not logging to stderr by default anymore, but events can be registered (see [LogProvider](https://github.com/sandrohanea/whisper.net/tree/main/Whisper.net/Logger/LogProvider))
     => Added optional logging in simple samples + CoreML sample
     => Logging is now suppressed by default: #129

   - Added `useGpu` on WhisperFactory creation so that GPU usage can be disabled on specific runtimes (e.g. Cublas, CoreML, etc)

   - Upgraded dotnet targets to dotnet 8

   - Added support for whisper large v3 (including downloader + coreml model)
       => Old model Large in downloader should be renamed to LargeV2 see [GgmlType](https://github.com/sandrohanea/whisper.net/blob/main/Whisper.net/Ggml/GgmlType.cs)

   - Added linux-x64 support for Whisper.net.Runtime.Cublas

   - Added support for distil-whisper #126
       => No downloader added, you'll need to load the model yourself
       => How to build the model: https://github.com/ggerganov/whisper.cpp/tree/master/models#distilled-models

   - **Added automated builds for all runtimes using Github Actions**

   - Unified runtime builds with make for linux, macos, ios, tvos, wasm and powershell scripts for windows

   - Added Blazor with InteractivityAuto, InteractivityServer and InteractivityWebAssemly sample with Blazor updated in dotnet 8.

   - Unified Whisper.net.Runtime.Wasm with Whisper.net.Runtime
       => If previously you were using Whisper.net.Runtime.Wasm, just replace it with Whisper.net.Runtime

   - Added metal support for arm-based apple platforms (both Whisper.net.Runtime + Whisper.net.Runtime.CoreML)

* Commented local_whisper nuget source + reset github actions to execute against PRs

* Fixed linux build + changed native github actions to work with PRs as well, not only branches

* Add native changes wasm

* Add native changes android

* Add native changes windows

* Add native changes macos

* Updated readme

* Add native changes linux

---------

Co-authored-by: sandrohanea <[email protected]>
@sandrohanea
Copy link
Owner

Distil whisper models are supported now, follow: https://github.com/ggerganov/whisper.cpp/tree/master/models#distilled-models to convert them to ggml and use them with whisper.net :)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants