Google open-sources TorchTPU, a native PyTorch back end for TPUs(github.com)
366 points by ilreb | 90 comments
GitHub repository for TorchTPU, Google's open-source native PyTorch backend for running models on Cloud TPUs.
#pytorch on Hacker News
31 stories tagged pytorch, newest first.
PyTorch-native TPU back end for vLLM(github.com)
4 points by zniturah | 0 comments
GitHub repository for vllm-torchtpu, a PyTorch-native TPU backend for the vLLM LLM inference engine.
0/1 Knapsack Problem in PyTorch(jedrzej.maczan.pl)
41 points by yu3zhou4 | 6 comments
Blog post introducing dp_knapsack_sliding_hirschberg, a PyTorch activation memory budget solver that cuts peak RAM during memory planning by about 20x.
Sub-1-Bit LLM Compression via Latent Factorization(github.com)
70 points by brainless | 12 comments
GitHub repository from SamsungLabs with the official implementation of LittleBit and LittleBit-2, sub-1-bit LLM compression via latent factorization.
Show HN: Rgpu – a PyTorch device whose tensors live on a remote GPU(github.com)
46 points by boxstream | 2 comments
GitHub repository for rgpu, a PyTorch device that runs tensor operations on a remote NVIDIA GPU while Python code stays on the local machine, including Macs.
Show HN: TurboGPT: train 22KiB transformer in 13s(github.com)
41 points by lostmsu | 5 comments
GitHub repository for TurboGPT, a CUDA-only project for training a tiny GPT transformer in under a minute, inspired by minGPT.
Dynamic Abliteration: Non-Destructive Refusal Suppression via Engram Steering(blog.madhukaraphatak.in)
96 points by phatak-dev | 32 comments
Blog post describing dynamic abliteration, using PyTorch forward hooks to inject multi-layer residual steering in Qwen3-4B to suppress refusals without modifying weights.
Show HN: I built a tiny LLM to demystify how language models work(github.com)
827 points by armanified | 125 comments
GitHub repository for GuppyLM, a ~9M parameter transformer language model built from scratch in PyTorch that talks like a small fish.
Visual introduction to PyTorch(0byte.io)
372 points by 0bytematt | 25 comments
A visual, beginner-friendly tutorial introducing PyTorch, covering tensors, autograd, and building a simple neural network.
LLM from scratch, part 28 – training a base model from scratch on an RTX 3090(gilesthomas.com)
482 points by gpjt | 101 comments
Blog post, part 28 of a series, describing an experiment training a ~163M-parameter base LLM from scratch on a single RTX 3090 at home.
Leaving Meta and PyTorch(soumith.ch)
665 points by saikatsg | 162 comments
Blog post by PyTorch co-creator Soumith Chintala announcing his departure from Meta and the PyTorch project, dated November 6, 2025.
A bug that taught me more about PyTorch than years of using it(elanapearl.github.io)
426 points by bblcla | 78 comments
Blog post recounting how a training loss plateau turned out to be a PyTorch bug, traced through optimizer internals down to GPU kernels.
PyTorch Monarch(pytorch.org)
360 points by jarbus | 42 comments
PyTorch blog post from Meta introducing Monarch, a framework moving beyond multi-controller SPMD for heterogeneous, asynchronous, fault-tolerant ML workflows.
Gemma 3 270M re-implemented in pure PyTorch for local tinkering(github.com)
399 points by ModelForge | 55 comments
GitHub directory in Sebastian Raschka's LLMs-from-scratch repo with notebooks re-implementing Google's Gemma 3 270M model in pure PyTorch, including a KV cache variant.
I have reimplemented Stable Diffusion 3.5 from scratch in pure PyTorch(github.com)
454 points by yousef_g | 73 comments
GitHub repository miniDiffusion, a from-scratch reimplementation of Stable Diffusion 3.5 in pure PyTorch, MIT licensed.
I don't like NumPy(dynomight.net)
399 points by MinimalAction | 164 comments
A Dynomight blog essay critiquing NumPy's design, arguing its array broadcasting and indexing make complex computations confusing.
PyTorch Internals: Ezyang's Blog(blog.ezyang.com)
418 points by Anon84 | 24 comments
Edward Yang's long-form blog essay on PyTorch internals, based on his talk at the PyTorch NYC meetup on May 14, 2019.
Introduction to CUDA programming for Python developers(pyspur.dev)
324 points by t55 | 87 comments
A PySpur blog post introducing CUDA GPU programming concepts for Python developers, showing how ML frameworks build on CUDA and its libraries.
SmolGPT: A minimal PyTorch implementation for training a small LLM from scratch(github.com)
415 points by amrrs | 52 comments
GitHub repository for smolGPT, a minimal PyTorch implementation for training a small GPT-style language model from scratch.
Show HN: Llama 3.2 Interpretability with Sparse Autoencoders(github.com)
542 points by PaulPauls | 80 comments
GitHub repository with an end-to-end PyTorch pipeline for LLM interpretability using sparse autoencoders on Llama 3.2, replicating Anthropic/OpenAI/DeepMind research.
Building LLMs from the Ground Up: A 3-Hour Coding Workshop(magazine.sebastianraschka.com)
910 points by mdp2021 | 110 comments
Sebastian Raschka's Ahead of AI newsletter post presenting a 3-hour video coding workshop on implementing, training, and using LLMs from scratch.
SAM 2: Segment Anything in Images and Videos(github.com)
674 points by xenova | 122 comments
GitHub repository from Meta's facebookresearch with inference code, model checkpoints, and example notebooks for Segment Anything Model 2 (SAM 2).
Llama3 implemented from scratch(github.com)
987 points by Hadi7546 | 246 comments
GitHub repository with a Jupyter notebook implementing Meta's Llama 3 model from scratch, one matrix multiplication at a time.
Implementing a ChatGPT-like LLM from scratch, step by step(github.com)
600 points by rasbt | 83 comments
GitHub repository by Sebastian Raschka with code for building a ChatGPT-like LLM from scratch in PyTorch, accompanying a Manning book.
WhisperSpeech – An open source text-to-speech system built by inverting Whisper(github.com)
405 points by nickmcc | 102 comments
GitHub repository for WhisperSpeech, Collabora's open source text-to-speech system built by inverting OpenAI's Whisper model.
A supply chain attack on PyTorch(johnstawinski.com)
444 points by roblabla | 92 comments
Blog post by John Stawinski describing how his team executed a critical supply chain attack on PyTorch via its GitHub Actions CI infrastructure.
Implementation of Mamba in one file of PyTorch(github.com)
391 points by johnma2006 | 107 comments
GitHub repository with a simple, minimal single-file PyTorch implementation of the Mamba state-space model architecture.
StyleTTS2 – open-source Eleven-Labs-quality Text To Speech(github.com)
668 points by sandslides | 220 comments
GitHub repository for StyleTTS 2, an open-source text-to-speech model using style diffusion and adversarial training with large speech language models.
Deep Learning Course(fleuret.org)
376 points by Tomte | 43 comments
François Fleuret's University of Geneva deep learning course page offering slides, recordings, and a virtual machine, with examples in PyTorch.
Llama from scratch, or how to implement a paper without crying(blog.briankitano.com)
444 points by bkitano19 | 39 comments
A blog tutorial by Brian Kitano on implementing a scaled-down Llama language model from scratch trained on TinyShakespeare, with tips for implementing research papers.
Open source Background Remover: Remove Background from images and video using AI(github.com)
325 points by thunderbong | 87 comments
GitHub repository for Background Remover, an open-source command-line tool that removes backgrounds from images and video using AI.