Releases·Tool Release
PINNStudio: Free open-source no-code GUI for setting up, training, and visualizing PINNs, built on DeepXDE
A developer launched PINNStudio, a free, open-source, no-code GUI for setting up, training, and visualizing physics-informed neural networks (PINNs). It was built out of frustration that each new PINN problem previously required rewriting PDE definitions, boundary conditions, architectures, training schedules, and plotting code by hand. The tool targets scientific machine learning practitioners who know the physics better than the coding.
8 engagement1 sourceSun, Sep 6, 2026, 10:19 PM
AI Overview
- PINNStudio (pip install pinnstudio) generates DeepXDE-based code, streams training logs, and shows live loss and solution plots in-app
- The GUI supports PDE definitions, coupled multi-output systems, 1D/2D domains with boundary/initial conditions, custom architectures and training schedules
- It handles forward and inverse problems and ships templates for Heat, Allen-Cahn, and Cahn-Hilliard equations
- Creator Asfandyar Khan built it for researchers with limited coding experience and released it on GitHub as open source
Related
Developments·Paper
EvoPINN: LLM-driven agentic framework automates PINN design for PDEs
EvoPINN is an agentic framework that uses LLMs to automatically discover executable algorithms for physics-informed neural networks (PINNs), replacing manual trial-and-error engineering. It addresses the issue of unconstrained LLM code generation producing invalid or unstable solutions by enforcing mathematical validity and numerical stability.
EvoPINN is an agentic framework that uses LLMs to automatically discover executable algorithms for physics-informed neural networks (PINNs), replacing manual trial-and-error engineering. It addresses the issue of unconstrained LLM code generation producing invalid or unstable solutions by enforcing mathematical validity and numerical stability.
0 engagement1 sourcearxivWed, Jul 29, 2026, 05:36 AM
Products·building
Visual editor that compiles drag-and-drop graphs into PyTorch code
NeuroBranch is a visual editor for designing neural network architectures by dragging typed cards onto a canvas and wiring them together. It compiles the graph into runnable PyTorch code, supports two-way sync between graph and code, and includes an AI copilot. It solves the problem of unreadable neural net code for developers who want to see and edit models visually.
NeuroBranch is a visual editor for designing neural network architectures by dragging typed cards onto a canvas and wiring them together. It compiles the graph into runnable PyTorch code, supports two-way sync between graph and code, and includes an AI copilot. It solves the problem of unreadable neural net code for developers who want to see and edit models visually.
1 engagement1 sourceredditSat, Aug 1, 2026, 07:01 AM
Releases·Tool Release
FlashPDE: A Drop-in Fused Triton Operator Library for Neural PDE Solvers
Researchers introduce FlashPDE, a fused operator library that replaces PyTorch finite-difference execution with differentiable Triton kernels for grid-based scientific machine learning. It integrates fused stencil evaluation, analytic discrete-adjoint backward pass, and boundary-gradient correction to overcome memory and efficiency bottlenecks in large-scale PINN training.
Researchers introduce FlashPDE, a fused operator library that replaces PyTorch finite-difference execution with differentiable Triton kernels for grid-based scientific machine learning. It integrates fused stencil evaluation, analytic discrete-adjoint backward pass, and boundary-gradient correction to overcome memory and efficiency bottlenecks in large-scale PINN training.
0 engagement1 sourcearxivMon, Jul 20, 2026, 02:54 PM
Developments·Paper
Survey examines LLM-assisted PDE workflows across discovery, solvers, and diagnostics
A new arXiv survey reviews how large language models support partial differential equation workflows, spanning model discovery, solver generation, and diagnostics. It maps the emerging role of LLMs in linking natural language, symbolic math, code, and solver feedback for science and engineering.
A new arXiv survey reviews how large language models support partial differential equation workflows, spanning model discovery, solver generation, and diagnostics. It maps the emerging role of LLMs in linking natural language, symbolic math, code, and solver feedback for science and engineering.
0 engagement1 sourcearxivTue, Aug 4, 2026, 12:51 PM
Releases·Tool Release
UnFlow open-source tool tracks ML experiments as related runs
A developer released UnFlow, an open-source ML experiment tracking tool that models runs as a graph of related changes rather than independent table rows. It aims to answer what changed between experiments and how runs relate.
A developer released UnFlow, an open-source ML experiment tracking tool that models runs as a graph of related changes rather than independent table rows. It aims to answer what changed between experiments and how runs relate.
1 engagement1 sourceredditMon, Aug 17, 2026, 01:00 PM