modelscope's mass auto re-upload of content from HF is a bit annoying
OPEN SOURCE
-
Unsloth achieves 12× faster MoE model training with 35% less VRAM
By
–
You can now train MoE models 12× faster with 35% less VRAM via our new Triton kernels (no accuracy loss). Train gpt-oss locally on 12.8GB VRAM. In collab with @HuggingFace, Unsloth trains DeepSeek, Qwen3, GLM faster. Repo: github.com/unslothai/unsloth Blog: unsloth.ai/docs/new/faster-m…
→ View original post on X — @abhi1thakur, 2026-02-10 15:25 UTC
-
Auto-Tune vLLM Config Doubles Throughput and Halves Latency
By
–
3/5 The vertical fix: auto-tune vLLM config Our old settings were way too conservative. After tuning with Auto-Tune vLLM + GuideLLM we got: ~2× throughput, 2× lower latency, same GPU budget @VLLM @Openshift
-
HunyuanOCR Technical Report on Advanced Character Recognition
By
–
HunyuanOCR Technical Report https://
bit.ly/4bkGbyA
#AI #MachineLearning #DeepLearning #LLMs #DataScience -
The contradiction of open source projects relying on closed cloud APIs
By
–
"100% open source and local" but uses api keys for all the biggest cloud models … make it make sense !
-
Voxtral: Real-time audio streaming to text
By
–
Voxtral can now directly stream audio input into text output. Perfect for:
— Mistral AI for Developers (@MistralDevs) 6 février 2026
– Live subtitles
– Language learning apps
– Note-taking tools
– And more!
Made a demo for you to try directly on hugging face ! pic.twitter.com/kHLqGo41PqVoxtral can now directly stream audio input into text output. Perfect for:
– Live subtitles
– Language learning apps
– Note-taking tools
– And more! Made a demo for you to try directly on Hugging Face! [Translated from EN to English]→ View original post on X — @guillaumelample, 2026-02-06 19:45 UTC
-
EdgeFlowNet Optical Flow Deployment on OrangePix Plus Hardware
By
–
EdgeFlowNet optical flow running on @orangepixunlong 5 Plus + Metis M.2 Community member Enmin solved some gnarly deployment challenges:
→ Custom 6-channel DataAdapter for frame-pair calibration
→ Debug flags to hunt down shape mismatches
→ Non-symmetric padding fix, no -
Community Evals and Benchmark Datasets for AI Models
By
–
We have been shipping Community Evals & Benchmark Datasets: Benchmark datasets host benchmark leaderboards, you can now contribute eval results by opening a PR to model repositories, all PRs are fed to benchmark datasets Chat with datasets: agents live in Data
-
Voxtral Transcription Quality Improvements and FFT Fixes
By
–
Voxtral (from @MistralAI) transcription quality is quite incredible, the way it handles the punctation and all the rest, makes transcribed audio messages so much more understandable. I implemented a few fixes in the FFT and now there is no longer a skipped tokens issue in voxtral.c
→ View original post on X — @guillaumelample, 2026-02-06 10:11 UTC