(3/3)
RoseTTAFold All-Atom: github.com/baker-laboratory/RoseTTAFold-All-Atom
RF3: github.com/RosettaCommons/foundry
ESMFold: github.com/facebookresearch/esm
ESMFold2: github.com/Biohub/esm
RoseTTAFold All-Atom: github.com/baker-laboratory/RoseTTAFold-All-Atom
RF3: github.com/RosettaCommons/foundry
ESMFold: github.com/facebookresearch/esm
ESMFold2: github.com/Biohub/esm
September 10, 2026 at 6:25 PM
(3/3)
RoseTTAFold All-Atom: github.com/baker-laboratory/RoseTTAFold-All-Atom
RF3: github.com/RosettaCommons/foundry
ESMFold: github.com/facebookresearch/esm
ESMFold2: github.com/Biohub/esm
RoseTTAFold All-Atom: github.com/baker-laboratory/RoseTTAFold-All-Atom
RF3: github.com/RosettaCommons/foundry
ESMFold: github.com/facebookresearch/esm
ESMFold2: github.com/Biohub/esm
September 9, 2026 at 6:59 AM
New scaling law replaces separate N and D terms with a single interaction parameter k. The L-shape grid lets you map the frontier without full pretraining runs. https://arxiv.org/abs/2608.07222 https://github.com/facebookresearch/lingua
August 17, 2026 at 12:01 PM
New scaling law replaces separate N and D terms with a single interaction parameter k. The L-shape grid lets you map the frontier without full pretraining runs. https://arxiv.org/abs/2608.07222 https://github.com/facebookresearch/lingua
This is a great source of information on the topic! github.com/facebookrese...
GitHub - facebookresearch/content-seal: Content Seal is a state-of-the-art framework for invisible, robust watermarking across all modalities audio, image, video, and text. This suite spans the entire...
Content Seal is a state-of-the-art framework for invisible, robust watermarking across all modalities audio, image, video, and text. This suite spans the entire generative lifecycle, from training ...
github.com
August 11, 2026 at 1:54 AM
This is a great source of information on the topic! github.com/facebookrese...
Facebook Research's HyperAgents project showcases self-improving AI agents that enhance their learning by optimizing computable tasks. This open-source repository offers resources for implementing these advanced agents. https://github.com/facebookresearch/Hyperagents
GitHub - facebookresearch/HyperAgents: Self-referential self-improving agents that can optimize for any computable task · GitHub
Facebook Research's HyperAgents project showcases self-improving AI agents that enhance their learning by optimizing computable tasks. This open-source repository offers resources for implementing the
github.com
August 7, 2026 at 12:50 PM
Facebook Research's HyperAgents project showcases self-improving AI agents that enhance their learning by optimizing computable tasks. This open-source repository offers resources for implementing these advanced agents. https://github.com/facebookresearch/Hyperagents
August 4, 2026 at 2:00 PM
TLX Block Attention: 고정 블록 희소 셀프 어텐션을 위한 워프 특화 Blackwell 커널 | 파이토치 한국 사용자 모임
코드는 다음에서 확인할 수 있습니다: https://github.com/facebookresearch/ads_model_kernel_library
*이 글에서는 TLX Block Attention의 설계를 소개합니다. TLX Block Attention은 NVIDIA Blackwell GPU를 겨냥한 Triton 커널로, 블록 대각(block-diagonal) 어텐션 패턴을 컴파일 시점에 알…
코드는 다음에서 확인할 수 있습니다: https://github.com/facebookresearch/ads_model_kernel_library
*이 글에서는 TLX Block Attention의 설계를 소개합니다. TLX Block Attention은 NVIDIA Blackwell GPU를 겨냥한 Triton 커널로, 블록 대각(block-diagonal) 어텐션 패턴을 컴파일 시점에 알…
TLX Block Attention: 고정 블록 희소 셀프 어텐션을 위한 워프 특화 Blackwell 커널 | 파이토치 한국 사용자 모임
코드는 다음에서 확인할 수 있습니다: https://github.com/facebookresearch/ads_model_kernel_library
*이 글에서는 TLX Block Attention의 설계를 소개합니다. TLX Block Attention은 NVIDIA Blackwell GPU를 겨냥한 Triton 커널로, 블록 대각(block-diagonal) 어텐션 패턴을 컴파일 시점에 알고 있다는 점을 활용해 범용 어텐션 구현에 존재하는 여러 종류의 알고리즘적 오버헤드를 통째로 제거합니다. NVIDIA B200 GPU에서 이 커널은 Flash Attention v2 대비 순전파(forward) 약 1.85배, 역전파(backward) 약 2.50배의 속도 향상을 달성하며, 회전 임베딩(rotary embedding)을 어텐션...
discuss.pytorch.kr
July 19, 2026 at 1:01 AM
TLX Block Attention: 고정 블록 희소 셀프 어텐션을 위한 워프 특화 Blackwell 커널 | 파이토치 한국 사용자 모임
코드는 다음에서 확인할 수 있습니다: https://github.com/facebookresearch/ads_model_kernel_library
*이 글에서는 TLX Block Attention의 설계를 소개합니다. TLX Block Attention은 NVIDIA Blackwell GPU를 겨냥한 Triton 커널로, 블록 대각(block-diagonal) 어텐션 패턴을 컴파일 시점에 알…
코드는 다음에서 확인할 수 있습니다: https://github.com/facebookresearch/ads_model_kernel_library
*이 글에서는 TLX Block Attention의 설계를 소개합니다. TLX Block Attention은 NVIDIA Blackwell GPU를 겨냥한 Triton 커널로, 블록 대각(block-diagonal) 어텐션 패턴을 컴파일 시점에 알…
Code available at: https://github.com/facebookresearch/ads_model_kernel_library/tree/main/multi_cta_norm_fusion and https://github.com/facebookresearch/ads_model_kernel_library/tree/m...
🔗 https://pytorch.org/blog/towards-free-normalization-fusing-normalization-into-gemm-and-attention-kernels
🔗 https://pytorch.org/blog/towards-free-normalization-fusing-normalization-into-gemm-and-attention-kernels
July 11, 2026 at 12:44 PM
Yeah, there are quite a few public domain image archives (LoC, NYPL, PICRYL, PDIA, etc). I wanted to try making an automated pipeline that uses the Segmant Anything Model to auto-detect and auto-clip interesting image segments to stick in an image pool, and then automatically describe and tag them.
GitHub - facebookresearch/segment-anything: The repository provides code for running inference with the SegmentAnything Model (SAM), links for downloading the trained model checkpoints, and example no...
The repository provides code for running inference with the SegmentAnything Model (SAM), links for downloading the trained model checkpoints, and example notebooks that show how to use the model. -...
github.com
July 10, 2026 at 6:11 PM
Yeah, there are quite a few public domain image archives (LoC, NYPL, PICRYL, PDIA, etc). I wanted to try making an automated pipeline that uses the Segmant Anything Model to auto-detect and auto-clip interesting image segments to stick in an image pool, and then automatically describe and tag them.
Discover HyperAgents, a research project by Facebook AI that enables self-referential agents to optimize computable tasks. With tools for experimentation and safety in mind, this repository paves the way for advanced AI capabilities. https://github.com/facebookresearch/Hyperagents
GitHub - facebookresearch/HyperAgents: Self-referential self-improving agents that can optimize for any computable task · GitHub
Discover HyperAgents, a research project by Facebook AI that enables self-referential agents to optimize computable tasks. With tools for experimentation and safety in mind, this repository paves the
github.com
July 8, 2026 at 8:40 AM
Discover HyperAgents, a research project by Facebook AI that enables self-referential agents to optimize computable tasks. With tools for experimentation and safety in mind, this repository paves the way for advanced AI capabilities. https://github.com/facebookresearch/Hyperagents
A new AI review! facebookresearch/ConvNeXt-V2 ⭐3.3/5.0
ConvNeXt-V2 is the official PyTorch reference implementation for the ConvNeXt V2 paper, including model definitions (Atto → Huge), FCMAE pretraining, fine-tuning scripts, and links to pre-trai...
https://gitrated.com/facebookresearch/ConvNeXt-V2
ConvNeXt-V2 is the official PyTorch reference implementation for the ConvNeXt V2 paper, including model definitions (Atto → Huge), FCMAE pretraining, fine-tuning scripts, and links to pre-trai...
https://gitrated.com/facebookresearch/ConvNeXt-V2
June 7, 2026 at 10:42 AM
A new AI review! facebookresearch/ConvNeXt-V2 ⭐3.3/5.0
ConvNeXt-V2 is the official PyTorch reference implementation for the ConvNeXt V2 paper, including model definitions (Atto → Huge), FCMAE pretraining, fine-tuning scripts, and links to pre-trai...
https://gitrated.com/facebookresearch/ConvNeXt-V2
ConvNeXt-V2 is the official PyTorch reference implementation for the ConvNeXt V2 paper, including model definitions (Atto → Huge), FCMAE pretraining, fine-tuning scripts, and links to pre-trai...
https://gitrated.com/facebookresearch/ConvNeXt-V2
Facebook Research launched HyperAgents, a framework for self-improving AI agents that optimize any computable task. The repository features code implementations, setup guides, and safety protocols for model-generated code. https://github.com/facebookresearch/Hyperagents
GitHub - facebookresearch/HyperAgents: Self-referential self-improving agents that can optimize for any computable task · GitHub
Facebook Research launched HyperAgents, a framework for self-improving AI agents that optimize any computable task. The repository features code implementations, setup guides, and safety protocols for
github.com
June 7, 2026 at 4:30 AM
Facebook Research launched HyperAgents, a framework for self-improving AI agents that optimize any computable task. The repository features code implementations, setup guides, and safety protocols for model-generated code. https://github.com/facebookresearch/Hyperagents