Log inSign up
user avatar
PyTorch
@PyTorch
Tensors and neural networks in Python with strong hardware acceleration. PyTorch is an open source project at the Linux Foundation. #PyTorchFoundation
pytorch.org
Joined September 2016
87
Following
504.7K
Followers
RepliesMedia
  • user avatar
    PyTorch
    @PyTorch
    Jul 31
    PyTorch 2.13 brings FlexAttention to Apple Silicon, cuts peak memory by up to 4× for large-vocabulary models with nn.LinearCrossEntropyLoss, and updates distributed training, compilation, profiling, and on-device inference. Our live Q&A examined CUDA version support, CuTeDSL in
    00:00
  • user avatar
    PyTorch
    @PyTorch
    Jul 31
    Congratulations to SMG on graduating from @lightseekorg and moving to an independent home. As part of the PyTorch ecosystem, SMG will continue working with maintainers from @vllm_project and TokenSpeed on shared frontend and gateway infrastructure for LLM inference.
    user avatar
    LightSeek Foundation
    @lightseekorg
    Jul 31
    Today SMG (Shepherd Model Gateway), an engine-agnostic gateway for LLM inference, has officially graduated from LightSeek Foundation and moves to a new independent home: github.com/smg-project/smg SMG was the first project incubated by the LightSeek Foundation and was recently
  • user avatar
    PyTorch
    @PyTorch
    Jul 31
    Great to see! Later this year, RadixArk will roll out SGL-torchtpu as an additional @radixark PyTorch-native TPU backend, featuring eager execution, @PyTorch ecosystem compatibility, and MPMD support. It will let any AI researcher or engineer in the industry run LLMs with
    user avatar
    RadixArk
    @radixark
    Jul 30
    RadixArk and Google Cloud are joining forces with the SGLang community to make TPU a drop-in, cost-efficient path to frontier inference. SGL-JAX already serves the major open model families on the latest TPU generations: Gemma, Qwen, DeepSeek, GLM, Kimi, Ling, MiniMax, MiMo,
  • user avatar
    PyTorch
    @PyTorch
    Jul 30
    FBTriton is the Triton repo where Meta develops its experimental GPU optimization solutions (including TLX/torchTLX and autoWS). It has served as an optimization runway for Meta engineers and external partners, including NVIDIA, AMD, and academic collaborators, to co-design