lunixbochs

7 ms·

1.
by lunixbochs 29d ago
similar energy to my https://github.com/lunixbochs/crossldso
2.
by lunixbochs 2mo ago
I think Google's Conformer paper is SOTA at the <30M model size, where I think they put an incredible amount of flops into a 10M param model to reach around 2% lsc clean (the whole model and RNN decoder were trained domain specific
3.
by lunixbochs 4mo ago
Thank you.
4.
by lunixbochs 4mo ago
I don't think any of your points reflect what I was trying to communicate. > You're not interested in addressing customers' needs I would love to support Wayland, but it is my position that it is impossible to "suppor
5.
by lunixbochs 4mo ago
Hi, I'm the developer of Talon. It's possible to do the simple compositor specific hacks from Talon's scripting system to give yourself partial Wayland support at roughly the quality I'd be able to provide myself, and I
6.
by lunixbochs 8mo ago
naive c is just a memcpy. non-temporal uses the streaming instructions.
7.
by lunixbochs 8mo ago
your single core numbers seem way too low for peak throughput on one core, unless you stipulate that all cores are active and contending with each other for bandwidth e.g. dual channel zen 1 showing 25GB/s on a single core https:&#x2F
8.
by lunixbochs 9mo ago
I'm confused why they repeatedly call a slots class larger than a regular dict class, but don't count the size of the dict
9.
by lunixbochs 9mo ago
I'm not familiar with C# compile at runtime. Are you saying your change was to do an AOT compile locally?
10.
by lunixbochs 3y ago
I did a silly experiment to compress word embeddings with jpeg - to see how it collapses semantically as you decrease the quality. https://bochs.info/vec2jpg/ This was a very basic experiment. I expect you could perfor
11.
by lunixbochs 3y ago
Are you using https://github.com/EleutherAI/lm-evaluation-harness ?
12.
by lunixbochs 4y ago
> made gl4es Neverball was working in the original glshim project before ptitseb forked it to gl4es. (Not to discount the significant work he's put in since, including the ES2 backend)
13.
by lunixbochs 4y ago
The Talon model is fairly accurate, but it can be confusing for new users to use the command system correctly. I posted a sibling reply about this, but the most common reason for Talon users to complain about the recognition is that they ar
14.
by lunixbochs 4y ago
Fixed commands are fast, precise, and predictable. Assuming you mean speaking in natural language, that's slower to say, and likely less precise and predictable if you want to be able to just say "anything" any have a result.
15.
by lunixbochs 4y ago
Talon's eye tracking functions as a mouse replacement. Is there a specific demo you'd like to see? I can record one.
16.
by lunixbochs 4y ago
Depending on when that was: in 2018 the free model was the macOS speech engine, in 2019 it was a fast but relatively weak model, and as of late 2021 it's a much stronger model. I'm currently working on the next model series with a
17.
by lunixbochs 4y ago
A handful of the datasets I tested are fully held out (I have reason to believe none of the models have trained on them), and talon was trained on none of the dev or test data of any of the datasets in question. Due to whisper's weakly
18.
by lunixbochs 4y ago
Android voice access blogged about how they use a model to detect and classify buttons: https://ai.googleblog.com/2021/01/improving-mobile-app-acces...
19.
by lunixbochs 4y ago
Can you elaborate on how you see this working?
20.
by lunixbochs 4y ago
In large scale tests, I observed hallucinations from Whisper in speech regions of audio.
21.
by lunixbochs 4y ago
Nice catch. I'll run my test suite [1] on this and report back. [1] https://twitter.com/lunixbochs/status/1574848899897884672
22.
by lunixbochs 4y ago
If the target is the words "A A A" and you produce "B B B B", you have more errors than there were words in the target. 3 replacements and 1 insertion.
23.
by lunixbochs 4y ago
I used to work on a project called glshim (which ptitseb still maintains as gl4es, used by box86). glshim implements an ABI-compatible OpenGL 1.x fixed function API (as libGL.so.1) on top of an OpenGL ES 1.x driver. This allows you to accel
24.
by lunixbochs 4y ago
> To see ahead-of-time translated Rosetta code, I believe I had to disable SIP, compile a new x86 binary, give it a unique name, run it, and then run otool -tv /var/db/oah/ / /unique-name.aot (or use your to
25.
OpenAI Whisper Analysis(twitter.com)
10 points by lunixbochs 4y ago|2 comments
26.
by lunixbochs 4y ago
Just posted results here: https://twitter.com/lunixbochs/status/1574848899897884672
27.
by lunixbochs 4y ago
Do you have a demo audio clip for this? I'd be interested to see how it looks in practice.
28.
by lunixbochs 4y ago
If the Whisper models provide any benefits over the existing Talon models, and if it's possible to achieve any kind of reasonable interactive performance, I will likely integrate Whisper models into Talon. Talon's speech engine ba
29.
by lunixbochs 4y ago
For an offline (non-streaming) model, 1x realtime is actually kind of bad, because you need to wait for the audio to be available before you can start processing it. So if you wait 10 seconds for someone to finish speaking, you won't h
30.
by lunixbochs 4y ago
If you count the GPU component and memory bandwidth, the Apple M2 is slightly weaker on paper for 16-bit inference than the NVIDIA A2, if you manage to use the whole chip efficiently. The A16 is then slightly weaker than the M2. Sure, the W

More ›