Kirill Solodskikh (@GarchFather) / X

Kirill Solodskikh

303 posts

Kirill Solodskikh

@GarchFather

White chocolate @TheStageAI Co-founder, CEO, ex Huawei P50 AI cameras

Joined October 2022

Kirill Solodskikh
@GarchFather
Oct 29, 2025
AI builders: we're open-sourcing fine-tuned Whisper by @OpenAI with @TheStageAI optimized inference engine. Runs on @nvidia GPUs. 2 W power usage on @Apple devices via CoreML+MLX. Real-time streaming. Electron + ReactJS samples. Open-source weights on @huggingface.
Open-Source Inference Engine for AI Builders
From github.com
4.3M
Kirill Solodskikh
@GarchFather
Apr 30, 2023
If each person will train their #GPT4 based on their chats in messengers, then a result of specific dialogue can be predicted faster than it would occur in a real life. You can just approve the result of conversation. Future of conversation becomes more predictable.
11K
Kirill Solodskikh
@GarchFather
Aug 15, 2025
Can LLMs recognize ASCII art? Our tests show accelerated Elastic Models analyze line-by-line features and combine them using statistical patterns. Try it yourself with DeepSeek-Qwen-14B – 120 tok/s on H100, 40 tok/s on L40s, up to 3× faster. Free API token!
LLM vs ASCII Art — Tutorial | DeepSeek-Qwen-14B
From thestage.ai
220K
Kirill Solodskikh
@GarchFather
Aug 12, 2025
Our research team took @AIatMeta LLaMA-8B, quantized it with QLIP using post-training int8, applied SmoothQuant, and used pre-defined compiler-compatible NVIDIA configs. Why do this? Up to 2× fewer weights and 3.6× faster on one GPU. Try it with our simple Jupyter Notebook.
QLIP quantization tutorial with LLaMA 3.1 8B
From thestage.ai
228K
Kirill Solodskikh
@GarchFather
Jun 25, 2025
Meet Elastic MusicGen Large — our optimized fork of @metaai's MusicGen, powered by ANNA (@TheStageAI’s Automated Neural Network Accelerator): huggingface.co/TheStageAI/Ela… Ye @kanyewest used AI for vocals on "Bully," calling it the "next Auto-Tune." He switched up later, but tracks
TheStageAI/Elastic-musicgen-large · Hugging Face
From huggingface.co
242K
Kirill Solodskikh
@GarchFather
Jun 6, 2023
Current situation around top AI conferences like @NeurIPSConf, @CVPR, @iclr_conf, visas problems strongly motivates me to think wider and build proof-of-stake (PoS) conference based on AI + blockchain technology. What we need: 👇 1. Reviewers which got their stake based on their
21K
Kirill Solodskikh
@GarchFather
Jul 27, 2025
Been cooking up some audio tools. Made a quick playground on Hugging Face Spaces for easy testing. It’s Elastic MusicGen, our fork of Meta’s MusicGen Large by @TheStageAI. huggingface.co/spaces/TheStag… Drop prompts, get tracks — in seconds, right in your browser. 🚀 11× faster than
huggingface.co
Elastic Musicgen Large - a Hugging Face Space by TheStageAI
Playground for music generation using Elastic-musicgen-large
337K
Kirill Solodskikh
@GarchFather
Aug 20, 2025
Self-hosted text-to-image on H100 with @TheStageAI Elastic Models, accelerated from FLUX.1-schnell @bfl_ml. Our fastest model S generates a high-quality image in 0.5 s. Precompiled and ready-to-deploy – minimal cold start. Tutorial + access token inside if you want to try.
Tutorial: Accelerate FLUX.1-schnell Inference
From thestage.ai
138K
Kirill Solodskikh
@GarchFather
Jun 21, 2023
Replying to @GarchFather
Our work on #CVPR2023 "Integral Neural Networks" is a future of the knowledge extraction from large DNNs while reducing computational cost significantly! @TheStage_ai team will release python framework to build INNs in an efficient way. Project link: inn.thestage.ai
88K
Kirill Solodskikh
@GarchFather
Aug 27, 2025
Quantization delivers speedup but can reduce quality. Our researchers prepared a tutorial showing how ANNA automatically quantizes Flux and accelerates it 2× while keeping quality high. Orig. model latency: 6.4 s. Check the link. DM or comment for early access.
Quantizing Flux with ANNA: 2× Faster, Same Quality
From thestage.ai
110K
Kirill Solodskikh
@GarchFather
Sep 3, 2025
How to measure the quality of text-to-image models? Our research team @TheStageAI put together a comprehensive guide to check perceptual quality, sharpness, color, prompt alignment, and more. All the tricky image quality questions researchers usually ask are covered here↓
A Guide to Evaluating Text-to-Image Models
From thestage.ai
259K
Kirill Solodskikh
@GarchFather
Sep 9, 2025
MLPerf Inference v5.1 by @MLCommons is out – here’s what our team can do. We ran @StabilityAI SDXL on 8×H100 with our stack, ANNA, accelerating inference with high quality. 18.1 img/s Submitted alongside @Google, @NVIDIA, @nebiusai and more. Proud @TheStageAI made this ↓
MLPerf Inference v5.1 Benchmark Results
From mlcommons.org
96K
Kirill Solodskikh
@GarchFather
Jun 19, 2023
Yo Yo! #CVPR2023 participants! We are preparing friendly meetup 23.06 with drinks, food, talk about our award candidate paper INNs and plans with @TheStage_ai team on DNNs acceleration. ABOUT PROMO CODES FOR FREE TICKETS - WRITE ME!
eventbrite.com
Exploring Integral Neural Networks with CVPR2023 Award Candidates - Meetup
Integral Neural Networks - a new class of DNNs - are poised to transform AI!
243K
Kirill Solodskikh
@GarchFather
Oct 30, 2025
We believe that everyone will become a model builder! That's why we are creating an automated acceleration and deployment stack which undestands ai engineers needs
clem 🤗
@ClementDelangue
Oct 30, 2025
We’re finally reaching the era of everyone training their own models based on open-source (versus relying on black box generalist APIs) and it is glorious!
8.5K