GitHub
Differentiable sign/round/floor for PyTorch
What it does
PyTorch's sign, round, floor, and argmax break gradients. Existing fixes (STE, Gumbel-Softmax) require rewriting your model. SLL-Core is zero-intrusive: with sll.linearize(): y = torch.sign(x) # now differentiable! loss.backward() Only linearizes inside an ε-band around decision boundaries. Everywhere else stays exactly hard. Zero deployment overhead. pip install sll-core
Does the same job
all alternatives →- ISI stripped DALL·E Mini to its bare essentials and converted it to Torch2022 · github.com · ▲663
- TLTorch Lens Maker – Differentiable Geometric Optics in PyTorch2025 · victorpoughon.github.io · ▲179
Hello HN! For the past 6 months I've been working on an open source python library that implements differentiable geometric optics in PyTorch. It's very experimental still, but eventually the goal is to use it to design optical systems with a state of the art optimization framework and a beautiful code based API. Think OpenSCAD, but for optical systems. Not only is PyTorch's autograd an amazing general purpose optimizer, but torch.nn (the neural network building blocks) can be used pretty much out of the box to model an optical system. This is because there is a strong analogy to be made…
- BGBoldly go where Gradient Descent has never gone before with DiscoGrad2024 · github.com · ▲232
Trying to do gradient descent using automatic differentiation over branchy programs? Or to combine them with neural networks for end-to-end training? Then this might be interesting to you. We develped DiscoGrad, a tool for automatic differentiation through C++ programs involving input-dependent control flow (e.g., "if (f(x) < c) { ... }", differentiating wrt. x) and randomness. Our initial motivation was to enable the use of gradient descent with simulations, which often rely heavily on such discrete branching. The latter makes plain autodiff mostly useless, since it can only account for the…
- LFLeanRL: Fast PyTorch RL with Torch.compile and CUDA Graphs2024 · github.com · ▲53
We're excited to announce that we've open-sourced LeanRL, a lightweight PyTorch reinforcement learning library that provides recipes for fast RL training using torch.compile and CUDA graphs. By leveraging these tools, we've achieved significant speed-ups compared to the original CleanRL implementations - up to 6x faster! Reinforcement learning is notoriously CPU-bound due to the high frequency of small CPU operations. PyTorch's powerful compiler can help alleviate these issues, but comes with its own costs. LeanRL addresses this challenge by providing simple recipes to accelerate your…
Pythagora - automated integration tests2023 · ▲141Save 30% of dev time - generate tests without writing code
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 27d ago · cactuscompute.com
Launched alongside, May 2026
the whole month →
Parallel agents, diff reviewer, and multi-model comparisons
Dev tools · May 2026 · kilo.ai
- NW
Hey HN, Henry here from Cactus. We open-sourced Needle, a 26M parameter function-calling (tool use) model. It runs at 6000 tok/s prefill and 1200 tok/s decode on consumer devices. We were always frustrated by the little effort made towards building agentic models that run on budget phones, so we conducted investigations that led to an observation: agentic experiences are built upon tool calling, and massive models are overkill for it. Tool calling is fundamentally retrieval-and-assembly (match query to tool name, extract argument values, emit JSON), not reasoning. Cross-attention…
Life & fun · May 2026 · github.com
- FM
Dev tools · May 2026 · github.com