Whisper VS codebase-visualizer-action

I've been using WhisperDesktop ( https://github.com/Const-me/Whisper ) with great success on a 3090 for fast & accurate transcription of often poor quality euro-english hours long multispeaker audio files. If there's an easy way to compare I'm certainly going to give this a try.

Why would you want OpenCL? Pretty sure D3D11 compute shaders gonna be adequate for a Torch backend, and they even work on Linux with Wine: https://github.com/Const-me/Whisper/issues/42 Native Vulkan compute shaders would be even better.

Why would you want unified address space? At least in my experience, it’s often too slow to be useful. DMA transfers (CopyResource in D3D11, copy command queue in D3D12, transfer queue in VK) are implemented by dedicated hardware inside GPUs, and are way more efficient.

https://github.com/ggerganov/whisper.cpp

https://github.com/Const-me/Whisper

I had fun with both of these. They will both do realtime transcription. Bit you will have to download the training data sets…

Gamers don’t care about FP64 performance, and it seems nVidia is using that for market segmentation. The FP64 performance for RTX 4090 is 1.142 TFlops, for RTX 3090 Ti 0.524 TFlops. AMD doesn’t do that, FP64 performance is consistently better there, and have been this way for quite a few years. For example, the figure for 3090 Ti (a $2000 card from 2022) is similar to Radeon RX Vega 56, a $400 card from 2017 which can do 0.518 TFlops.

And another thing: nVidia forbids usage of GeForce cards in data centers, while AMD allows that. I don’t know how specifically they define datacenter, whether it’s enforceable, or whether it’s tested in courts of various jurisdictions. I just don’t want to find out answers to these questions at the legal expenses of my employer. I believe they would prefer to not cut corners like that.

I think nVidia only beats AMD due to the ecosystem: for GPGPU that’s CUDA (and especially the included first-party libraries like BLAS, FFT, DNN and others), also due to the support in popular libraries like TensorFlow. However, it’s not that hard to ignore the ecosystem, and instead write some compute shaders in HLSL. Here’s a non-trivial open-source project unrelated to CAE, where I managed to do just that with decent results: https://github.com/Const-me/Whisper That software even works on Linux, probably due to Valve’s work on DXVK 2.0 (a compatibility layer which implements D3D11 on top of Vulkan).

Currently, I use a GUI for Whisper AI (https://github.com/Const-me/Whisper) to upload MP3s of interviews to get text transcripts. However, I'm hoping to find another tool that would recognize and split out the text per speaker.

Does such a thing exist?

I just used Whisper over the weekend to transcribe 5 hours of meeting, worked nicely and it can be run on a single GPU locally. https://github.com/ggerganov/whisper.cpp

There are a few wrappers available with GUI like https://github.com/Const-me/Whisper

Nice post - treemaps are great!

My friend and I made a codebase visualisation tool (https://www.codeatlas.dev/gallery) that's based on Voronoi treemaps, maybe of interest as an illustration of the aesthetics with a non-rectangular layout!

We've opted for zooming through double-clicks as the main method of navigating the map, because in deep codebases, the individual cells quickly get too small to accurately target with the cursor as shown in the key-path label approach!

If anyone's interested, this is also available as a Github Action to generate the treemap during CI: https://github.com/codeatlasHQ/codebase-visualizer-action

If you find this type of codebase visualisation useful, you might want to checkout codeatlas.dev and its Github Action (https://github.com/codeatlasHQ/codebase-visualizer-action). It doesn't animate the repo over time like gource (yet), but instead aims to give a beautiful interactive visual snapshot of a repo at a particular point in time. It also lets you zoom in on specific aspects like recent commit activity, programming language and hopefully in the future test coverage.

E.g. see here for a visualisation of the pytorch codebase we did a while ago: https://codeatlas.dev/gallery/pytorch/pytorch

(disclaimer: I'm the author)

If you think this is useful, you might also like codeatlas.dev and its Github Action (https://github.com/codeatlasHQ/codebase-visualizer-action). It currently does not support per-contributor activity, but we put a lot of effort into making the diagrams beautiful to look at and the basic approach of using treemaps for visualisation seems very similar. In fact, could be cool to collaborate on this, DM me if interested!

https://codeatlas.dev

https://codeatlas.dev - codebase visualisation tool

Takes your git repo and generates a beautiful visual representation of the code. Sort of an alternative navigation tool (in addition to IDEs) for large codebases. Can also run it as part of CI with our Github Action (https://github.com/codeatlasHQ/codebase-visualizer-action).

We made this because grokking complex software projects is really difficult and we've found that a visual overview of what's in a codebase can be quite helpful to get started.

E.g. checkout https://codeatlas.dev/gallery/kubernetes/kubernetes for the generated visualisation of the Kubernetes Github repo!

Currently making -10$/year to pay for the domain :D We slowed down active development after our initial attempts at dissemination didn't really go anywhere (bragging about side projects on the internet, ugh), but I'm still really keen on getting some feedback on whether this is actually useful to anyone else!

Note: The site works somewhat on mobile, but is much better on desktop!

Also, funny there's a post like this again, just like https://news.ycombinator.com/item?id=34531989 yesterday.

https://codeatlas.dev - codebase visualisation tool

It takes your git repo and generates a beautiful visual representation of the actual code that's in it. Sort of an alternative navigation tool (in addition to IDEs) for large codebases. You can run codeatlas as part of your CI with our Github Action (https://github.com/codeatlasHQ/codebase-visualizer-action).

We made this because grokking complex software projects is really difficult and we've found that a visual overview of what's in a codebase can be quite helpful to get started.

E.g. checkout https://codeatlas.dev/gallery/kubernetes/kubernetes for the generated visualisation of the Kubernetes Github repo!

We slowed down active development after our initial attempts at dissemination didn't really go anywhere (bragging about side projects on the internet, ugh), but would still love feedback on whether this is possibly useful to anyone else!

Note: The site works somewhat on mobile, but is much better on desktop!

I've already mentioned this on the other thread (https://news.ycombinator.com/item?id=31569646), but my friend and I have been working on [https://www.codeatlas.dev](https://www.codeatlas.dev/) as a sideproject - it's a tool for creating pretty (2D!) visualisations of codebases, while providing additional insights via overlays (e.g. commit density, programming language or other results from static analysis like dead code/test coverage/etc.). For example here's the Kubernetes codebase visualised using codeatlas: [https://www.codeatlas.dev/repo/kubernetes/kubernetes](https:....

At the moment, codeatlas is just the static gallery, but we're only a few weekends away from releasing a Github action that deploys this diagram on github pages for your own repos - if you're interested, feel free to watch this repo: https://github.com/codeatlasHQ/codebase-visualizer-action

OP, how close is this to what you had in mind in your question?

My friend and I have been working on https://www.codeatlas.dev in our spare time, which is a tool that creates pretty (2D!) visualisations of codebases, while providing additional insights via overlays (e.g. commit density, programming language). For example here's the Kubernetes codebase visualised using codeatlas: https://www.codeatlas.dev/repo/kubernetes/kubernetes.

At the moment, codeatlas is only a static gallery, but we're currently about 1-2 weekends away from releasing a Github action that deploys this diagram on github pages for your own repos - if you're interested, feel free to watch this repo: https://github.com/codeatlasHQ/codebase-visualizer-action