doing something
@doing_something
712 subscribers
912 photos
36 videos
36 files
1.94K links
@smlkw doing something, just my notes to keep updates
Download Telegram
About
Blog
Apps
Platform
Join
doing something
712 subscribers
doing something
Tested Intel's quantized models of Whisper made by Neural Compressor:

https://github.com/egorsmkv/whisper-intel-optimized

#ai #asr #quantization
GitHub
GitHub - egorsmkv/optimized-whisper-intel: Run quantized Whisper models only on CPU with Intel hardware
Run quantized Whisper models only on CPU with Intel hardware - egorsmkv/optimized-whisper-intel
284 viewsedited  
doing something
doing something
HQQ https://github.com/mobiusml/hqq #ai #quantization
Their articles:

- https://mobiusml.github.io/hqq_blog/
- https://mobiusml.github.io/1bit_blog/

#quantization #ai
81 views
doing something
doing something
Their articles: - https://mobiusml.github.io/hqq_blog/ - https://mobiusml.github.io/1bit_blog/ #quantization #ai
https://mobiusml.github.io/whisper-static-cache-blog/

#asr #ai #whisper #quantization
mobiusml.github.io
Speeding up Whisper (ASR)
A support blog for the release of fast and small whisper model.
325 views
doing something
doing something
https://mobiusml.github.io/whisper-static-cache-blog/ #asr #ai #whisper #quantization
HQQ (4 bits) + Flash-Attention 2 + Whisper Large V3 on L4

#ai #asr
88 views
doing something
Whisper torch.compile go brrrrr

#ai #asr
254 viewsedited  
doing something
This media is not supported in your browser
VIEW IN TELEGRAM
Voice Activity Detection in the browser

https://github.com/wavey-ai/mel-spec

#rust #audio #wasm
316 views
doing something
https://huggingface.co/spaces/facebook/VFusion3D

image-to-mesh

#ai #3d
huggingface.co
VFusion3D - a Hugging Face Space by facebook
Upload an image to create a 3D model or video. The app processes the image to generate a 3D mesh or render it into a video, which you can then download.
107 views
doing something
https://bytez.com

#ai
Bytez
Bytez - open source AI
Keep up with AI. Discover, demo, and deploy open source models
108 viewsedited  
doing something
Made first Streamlit app (converting Gradio app using GitHub Copilot)

https://huggingface.co/spaces/Yehor/st-hubert-uk-demo

#ai #ui
huggingface.co
Streamlit Speech-to-Text for Ukrainian (HuBERT) - a Hugging Face Space by Yehor
Discover amazing ML apps made by the community
94 views
doing something
doing something
https://mobiusml.github.io/whisper-static-cache-blog/ #asr #ai #whisper #quantization
https://github.com/mobiusml/hqq/issues/108
GitHub
RuntimeError: Expected in.dtype() == at::kInt to be true, but got false. · Issue #108 · mobiusml/hqq
I want to reproduce Whisper + HQQ example but getting error: --------------------------------------------------------------------------- RuntimeError Traceback (most recent call last) [<ipython-...
98 views
doing something
Okay, I've made a working colab with these speeding mechanisms for Whisper:

- HQQ (quantize linear layers to 4-bit)
- torch.compile
- use static cache

Colab:

https://colab.research.google.com/drive/1zERp53F7gmE58OiqFR0iNV3gw3ybiwHM?usp=sharing

RTF: 0.0531 (batch size = 1) with large-v3

#ai #whisper #asr
Google
Nightly torch with HQQ (nbits=4) + torch.compile + static cache + Whisper.ipynb
Colab notebook
678 viewsedited  
doing something
Forwarded from Feed-Master
Why Rust?
rerun.io
Why Rust?
I've been a programmer for 20+ years, and few things excite me as much as Rust. My background is mostly in C++, though I have also worked in Python and Lua, and dabbled in many more languages. I started writing Rust around 2014, and since 2018 I've been writing…
87 views
doing something
bs=24, whisper large-v3

https://github.com/egorsmkv/optimized-whisper

#ai #asr
238 viewsedited  
doing something
doing something
bs=24, whisper large-v3 https://github.com/egorsmkv/optimized-whisper #ai #asr
A test, total time of the video: 1:51:45 (link)

Stats:

- Duration: 6480.020 seconds
- RTF: 0.0153
- Elapsed: 98.9704 seconds

Hardware:

- GPU: RTX 4000 Ada

#ai #asr
538 viewsedited  
doing something
doing something
A test, total time of the video: 1:51:45 (link) Stats: - Duration: 6480.020 seconds - RTF: 0.0153 - Elapsed: 98.9704 seconds Hardware: - GPU: RTX 4000 Ada #ai #asr
arestovich_ukrainian.txt
153.5 KB
517 views
doing something
https://github.com/mitsuhiko/minijinja

#rust
GitHub
GitHub - mitsuhiko/minijinja: MiniJinja is a powerful but minimal dependency template engine for Rust compatible with Jinja/Jinja2
MiniJinja is a powerful but minimal dependency template engine for Rust compatible with Jinja/Jinja2 - mitsuhiko/minijinja
96 views
doing something
Nice services:

- https://github.com/huridocs/pdf-document-layout-analysis

- https://github.com/huridocs/pdf-table-of-contents-extractor

- https://github.com/huridocs/pdf-text-extraction

#docker #ai #nlp #pdf #opensource
483 viewsedited  
doing something
Use it for good:

https://github.com/egorsmkv/vad-chunk-size

#ai #asr #vad
GitHub
GitHub - egorsmkv/vad-chunk-size: Extracted chunk-size VAD from WhisperX
Extracted chunk-size VAD from WhisperX. Contribute to egorsmkv/vad-chunk-size development by creating an account on GitHub.
91 views
doing something
Forwarded from Feed-Master
Burn 0.14.0 Released: The First Rust-Native Deep Learning Framework
burn.dev
Burn 0.14.0 Release Notes
This release marks the debut of our CubeCL integration, which brings cross-platform GPU programming capabilities directly to Rust.
As always, it also includes numerous bug fixes, performance enhancements, new tensor operations, and improved documentation.
92 views
doing something
https://github.com/esimov/pigo supports WASM

#go #cv #wasm
274 viewsedited  
doing something
Forwarded from Feed-Master
Rspack 1.0 – the next generation JavaScript bundler written in Rust
rspack.rs
Announcing Rspack 1.0 - Rspack
Today Rspack has reached a new milestone - 1.0. This means that Rspack is production-ready, covers most of webpack's APIs and features, and is now prepared to support more users.
89 views