AI Pulse by Inblix

AI Research

Cutting-edge AI research papers, breakthroughs, and academic developments.

601 articles

InstantMesh now converts raw 3D meshes into textured models in 10 seconds — Inblix summary
Research

InstantMesh now converts raw 3D meshes into textured models in 10 seconds

Hugging Face Blog · Sep 30, 2024 · 2 min read

Generative 3D tools like InstantMesh have gotten remarkably fast at creating geometry from a single image, but the outp...

Meta drops Llama 3.2 Vision, but EU users are blocked from the multimodal models — Inblix summary
Research

Meta drops Llama 3.2 Vision, but EU users are blocked from the multimodal models

Hugging Face Blog · Sep 25, 2024 · 3 min read

Meta just shipped Llama 3.2, and the headliner is a pair of vision models that give the open-source LLM a set of eyes....

How 1.9M YouTube videos became FineVideo's 44K highly-annotated training gem — Inblix summary
Research

How 1.9M YouTube videos became FineVideo's 44K highly-annotated training gem

Hugging Face Blog · Sep 23, 2024 · 3 min read

Anyone can scrape a million YouTube videos. The hard part—the part that separates a research toy from a genuine trainin...

Hugging Face's Daily Papers hides 7 features even power users miss — Inblix summary
Research

Hugging Face's Daily Papers hides 7 features even power users miss

Hugging Face Blog · Sep 23, 2024 · 2 min read

Hugging Face's Daily Papers page looks like a straightforward research feed, and that's exactly why most people scroll...

Hugging Face and Intel squeeze Llama 3.1 into 4-bit for edge devices with OpenVINO — Inblix summary
Research

Hugging Face and Intel squeeze Llama 3.1 into 4-bit for edge devices with OpenVINO

Hugging Face Blog · Sep 20, 2024 · 2 min read

Running a model like Meta-Llama-3.1-8B on a laptop or edge device is moving from 'technically possible' to genuinely pr...

Llama 8B shrank to 1.58 bits per parameter and still beat a full-size model — Inblix summary
Research

Llama 8B shrank to 1.58 bits per parameter and still beat a full-size model

Hugging Face Blog · Sep 18, 2024 · 2 min read

Microsoft Research’s BitNet architecture promised a world where LLM parameters are just -1, 0, or 1, slashing memory an...

Hugging Face now lets you query 12.6M-row datasets with SQL in your browser — Inblix summary
Research

Hugging Face now lets you query 12.6M-row datasets with SQL in your browser

Hugging Face Blog · Sep 17, 2024 · 2 min read

Hugging Face just flipped a switch that makes exploring massive datasets feel almost frictionless. They've rolled out a...

HuggingChat now lets you build custom AI tools from any public Space — Inblix summary
Research

HuggingChat now lets you build custom AI tools from any public Space

Hugging Face Blog · Sep 16, 2024 · 2 min read

HuggingChat just got a lot more interesting. The team has introduced Community Tools, a feature that turns any public H...

Hugging Face's Accelerate hits 1.0 after 3.5 years, baking in FP8 and bigger model support — Inblix summary
Research

Hugging Face's Accelerate hits 1.0 after 3.5 years, baking in FP8 and bigger model support

Hugging Face Blog · Sep 13, 2024 · 2 min read

Three and a half years after it started as a simple multi-GPU helper, Hugging Face's Accelerate library just dropped it...

Hugging Face taps TruffleHog to catch leaked API keys in every commit — Inblix summary
Research

Hugging Face taps TruffleHog to catch leaked API keys in every commit

Hugging Face Blog · Sep 4, 2024 · 2 min read

Hugging Face is making a serious push to stop developers from accidentally broadcasting their passwords and API keys to...

Hugging Face's LeRobot shrinks robot datasets 86% with video codecs — Inblix summary
Research

Hugging Face's LeRobot shrinks robot datasets 86% with video codecs

Hugging Face Blog · Aug 27, 2024 · 2 min read

Robot datasets are monstrously bloated. The standard practice of storing visual data as individual PNG frames creates f...

Hugging Face's 5 hidden tools let you build semantic search pipelines for free — Inblix summary
Research

Hugging Face's 5 hidden tools let you build semantic search pipelines for free

Hugging Face Blog · Aug 22, 2024 · 2 min read

Most developers know Hugging Face for models and datasets. But the platform has quietly shipped infrastructure features...

Hugging Face packs 2x training speed bump into Flash Attention 2 with a single collator swap — Inblix summary
Research

Hugging Face packs 2x training speed bump into Flash Attention 2 with a single collator swap

Hugging Face Blog · Aug 21, 2024 · 2 min read

Padding tokens have always been the silent performance killer in LLM training. You batch your examples, stuff them with...

Google Cloud A3 nodes can now run Meta's 405B Llama 3.1, but you'll need 8 H100s — Inblix summary
Research

Google Cloud A3 nodes can now run Meta's 405B Llama 3.1, but you'll need 8 H100s

Hugging Face Blog · Aug 19, 2024 · 2 min read

If you want to run Meta's monster 405-billion-parameter Llama 3.1 model without it melting your hardware, Google Cloud'...

Google's Infini-Attention flops at 1M tokens, but the memory math still compels — Inblix summary
Research

Google's Infini-Attention flops at 1M tokens, but the memory math still compels

Hugging Face Blog · Aug 14, 2024 · 2 min read

The brutal reality of AI research is that most ideas that work beautifully on a whiteboard shatter against the practica...

ggml packs PyTorch power into 1MB: here's why devs are switching — Inblix summary
Research

ggml packs PyTorch power into 1MB: here's why devs are switching

Hugging Face Blog · Aug 13, 2024 · 2 min read

The library that quietly powers llama.cpp and Ollama is worth a close look for any developer who's ever winced at a PyT...

Falcon Mamba: The First 7B Model to Ditch Attention Without Performance Loss — Inblix summary
Research

Falcon Mamba: The First 7B Model to Ditch Attention Without Performance Loss

Hugging Face Blog · Aug 12, 2024 · 2 min read

The team behind Falcon has released Falcon Mamba, a 7-billion-parameter model that stands as the first general-purpose,...

Hugging Face just fixed the tool-use nightmare that's been plaguing LLM developers — Inblix summary
Research

Hugging Face just fixed the tool-use nightmare that's been plaguing LLM developers

Hugging Face Blog · Aug 12, 2024 · 2 min read

Tool use in LLMs sounds great on paper — let your model call a calculator, search the web, or query a database. In prac...

Hugging Face buys XetHub to kill Git LFS on its 12PB model hub — Inblix summary
Research

Hugging Face buys XetHub to kill Git LFS on its 12PB model hub

Hugging Face Blog · Aug 8, 2024 · 3 min read

Hugging Face is acquiring XetHub, a Seattle startup founded by three ex-Apple engineers who built Apple's internal ML i...

Albumentations adds text-aware augmentation that rewrites document images without wrecking OCR — Inblix summary
Research

Albumentations adds text-aware augmentation that rewrites document images without wrecking OCR

Hugging Face Blog · Aug 6, 2024 · 2 min read

Fine-tuning vision language models on document images has always been a headache. You need the model to actually read t...

Hugging Face's 5-layer security stack: what actually stops token leaks — Inblix summary
Research

Hugging Face's 5-layer security stack: what actually stops token leaks

Hugging Face Blog · Aug 6, 2024 · 2 min read

Hugging Face rolled out a comprehensive security rundown for 2024, and the details matter more than the headlines. The...

Google drops Gemma 2 2B: a 2.6B on-device model that punches above its weight — Inblix summary
Research

Google drops Gemma 2 2B: a 2.6B on-device model that punches above its weight

Hugging Face Blog · Jul 31, 2024 · 2 min read

Google has expanded its Gemma 2 family with a 2.6 billion parameter model designed specifically for on-device deploymen...

Quanto FP8 Quantization Cuts SD3 Memory Use by 33% on a Single GPU — Inblix summary
Research

Quanto FP8 Quantization Cuts SD3 Memory Use by 33% on a Single GPU

Hugging Face Blog · Jul 30, 2024 · 2 min read

Running Stable Diffusion 3 in FP16 eats 18.765 GB of GPU memory. That's a problem for anyone without a datacenter card....

Hugging Face Kills NVIDIA NIM Serverless Inference, Points Users to New Service — Inblix summary
Research

Hugging Face Kills NVIDIA NIM Serverless Inference, Points Users to New Service

Hugging Face Blog · Jul 29, 2024 · 2 min read

Hugging Face has quietly pulled the plug on its NVIDIA NIM API, the serverless inference service it launched with consi...