flash-attention VS whisper.cpp

If you're looking for an implementation, I highly recommend checking out fast attention [https://github.com/Dao-AILab/flash-attention]. It's my go-to, and far better than anything we could whip up here using just PyTorch or TensorFlow.

* Even if I bought a 3090, I would have to get a computer to go with it, along with a PSU and some cooling. Don't know where to start with that.

[1] https://github.com/Dao-AILab/flash-attention/issues/190

highly recommend using Tri's implementation https://github.com/Dao-AILab/flash-attention rotary should be built in, and some group overseas even contributed alibi

Doesn't seem so https://github.com/Dao-AILab/flash-attention/issues/542 No updates for a while.

I wonder how this compares to Flash Attention (https://github.com/HazyResearch/flash-attention), which is the other "memory aware" Attention project I'm aware of.

I guess Flash Attention is more about utilizing memory GPU SRam correctly, where this is more about using the OS/CPU memory better?

https://github.com/HazyResearch/flash-attention

After a very quick read, that's my understanding too: It's just KNN search. So I agree on points 1-3. When something works well, I don't care much about point 4.

I've had only mixed success with KNN search. Maybe I haven't done it right? Nothing seems to work quite as well for me as explicit token-token interactions by some form of attention, which as we all know is too costly for long sequences (O(n²)). Lately I've been playing with https://github.com/hazyresearch/safari , which uses a lot less compute and seems promising. Otherwise, for long sequences I've yet to find something better than https://github.com/HazyResearch/flash-attention for n×n interactions and https://github.com/glassroom/heinsen_routing for n×m interactions. If anyone here has other suggestions, I'd love to hear about them.

Longer sequence length in transformers is an active area of research (see e.g the great work from the Flash-attention team - https://github.com/HazyResearch/flash-attention), and I'm sure will improve things dramatically very soon.

Here's a list of tools for scaling up transformer context that have github repos:

* FlashAttention: In my experience, the current best solution for n² attention, but it's very hard to scale it beyond the low tens of thousands of tokens. Code: https://github.com/HazyResearch/flash-attention

* Heinsen Routing: In my experience, the current best solution for n×m attention. I've used it to pull up more than a million tokens as context. It's not a substitute for n² attention. Code: https://github.com/glassroom/heinsen_routing

* RWKV: A sort-of-recurrent model which claims to have performance comparable to n² attention in transformers. In my limited experience, it doesn't. Others agree: https://twitter.com/arankomatsuzaki/status/16390003799784038... . Code: https://github.com/BlinkDL/RWKV-LM

* RMT (this method): I'm skeptical that the recurrent connections will work as well as n² attention in practice, but I'm going to give it a try. Code: https://github.com/booydar/t5-experiments/tree/scaling-repor...

In addition, there's a group at Stanford working on state-space models that looks promising to me. The idea is to approximate n² attention dynamically using only O(n log n) compute. There's no code available, but here's a blog post about it: https://hazyresearch.stanford.edu/blog/2023-03-27-long-learn...

If anyone here has other suggestions for working with long sequences (hundreds of thousands to millions of tokens), I'd love to learn about them.

whisper.cpp [1] has a karaoke example that uses ffmpeg's drawtext filter to display rudimentary karaoke-like captions. It also supports diarisation. Perhaps it could be a starting point to create a better script that does what you need.

--

1: https://github.com/ggerganov/whisper.cpp/blob/master/README....

The ggml library is one of the first library for local LLM interference. It’s a pure C library that converts models to run on several devices, including desktops, laptops, and even mobile device - and therefore, it can also be considered as a tinkering tool, trying new optimizations, that will then be incorporated into other downstream projects. This tool is at the heart of several other projects, powering LLM interference on desktop or even mobile phones. Subprojects for running specific LLMs or LLM families exists, such as whisper.cpp.

I'm not sure if it is _fully_ openai compatible, but whispercpp has a server bundled that says it is "OAI-like": https://github.com/ggerganov/whisper.cpp/tree/master/example...

I don't have any direct experience with it... I've only played around with whisper locally, using scripts.

unless i'm misunderstanding `whisper.cpp` seems to support streaming & the repository includes a native example[0] and a WASM example[1] with a demo site[2].

[0]: https://github.com/ggerganov/whisper.cpp/tree/master/example...

Usage 1: Good to transcribe audio. An example use case could be to summarize YouTube videos or long courses. Usage 2: You talk with voice to your AI that responds with text (later with audio too). - https://github.com/ggerganov/whisper.cpp

🗣️🎙️ whisper.cpp (offline speech-to-text transcription, models trained by OpenAI, CLI based, browser based)

Has anyone played around with Whisper C++ for swift? I'm hitting a snag even on the demo. I've downloaded the github repo and everything matches up with this video [ https://youtu.be/b10OHCDHDQ4 ] but when he hits the transcribe button, it actually prints out the captioning. When I do it, it skips that part and just says "Done...". But it, does everything else - plays the audio, says it's transcribing.. just doesn't show me the transcription: and it's not in the debug window either. But the demo isn't throwing any errors, and I haven't messed with the code really so this is their example. https://github.com/ggerganov/whisper.cpp