PyTorch

Track PyTorch development. The machine learning framework powering AI research.

About this developer podcast

PyTorch turns GitHub activity into short audio updates.

Episodes cover commits, pull requests, issues, and release work.

Use this page to listen, subscribe, and find related developer podcasts.

This show page is the public hub for the podcast. It has one stable listen URL.

Use it to find RSS feeds, recent episodes, related shows, and older episode links.

If the project is quiet, new episodes may pause until there is useful activity to summarize.

Recent episodes from PyTorch

  1. Memory Safety and Lifetime Fixes Across CUDA and Distributed

    Today's activity centers on lifetime and memory-safety bugs in CUDA graph memory management, NCCL teardown, and layer norm backward kernels, alongside a broad push to make test suites and diagnostics device-agnostic.

  2. Precompile Lands, and the Correctness Sweep Continues

    Bo Ren's twelve-part precompile stack is the day's biggest structural change, hardening how compiled artifacts capture, guard, and reload across processes. Alongside it, a wave of correctness fixes across MPS, scaled matmul, and pooling…

  3. Weekly Recap - Precompile Lands, and Correctness Sweeps Across MPS and CUDA

    This week's headline is the arrival of a large, hardened stack for ahead-of-time compilation under torch.compiler.precompile, alongside a broad wave of correctness fixes spanning Apple Metal, CUDA graphs, and core tensor operations.…

  4. Batching Rules Get Unified, Precompile Gets Sharper

    The control-flow team shipped a five-part stack unifying vmap batching rules across cond, while-loop, scan, and switch, while the precompile and dynamo teams landed a companion stack refining guard provenance and serialization. Several…

  5. Undefined Contracts and the MPS Coverage Push

    Today's activity centers on making implicit assumptions explicit—undefined gradient strides, silent NaN inconsistencies, and moved-from argument stacks all got documented or fixed—while a single contributor kept closing MPS backend gaps…

  6. A New CuteDSL Reduction Engine and the Great Test Untangling

    The biggest story today is a fourteen-part stacked rollout from slayton58 building a full CuteDSL-based reduction kernel system from the ground up, while a separate wave of PRs continues splitting monolithic test files into…

  7. ROCm Ten Point Oh and the Precompile Stack

    PyTorch's ROCm builds jumped to version ten point oh across wheels, docker images, and trunk CI, while a large ghstack from Bob Ren reworked precompile and guard internals, and a cluster of relanded fixes addressed correctness bugs in…

  8. Hardening the Merge Pipeline and the Attention Stack

    August 26 brought a wave of correctness fixes to cuDNN varlen attention and Inductor's CSE logic, alongside a new multi-part "greenlight" merge guard designed to stop mismatched-commit landings. Several PRs were duplicated across GitHub…

  9. FlyDSL Lands and the GEMM Compiler Stack Grows
  10. AOT Autograd Tightens Its Grip on Backward Tangents
  11. Weekly Recap - Compiler Correctness and the Long Tail of Cleanup
  12. Precompile Guards Get a Rewrite
  13. Precompile Hardening and the Great Cleanup
  14. Precompile's Guard Overhaul
  15. Compiler Stacks, Symmetric Kernels, and Build Hygiene
  16. Precompile Grows Up, and the Great Typo Sweep
  17. Correctness Cleanup and a Suspicious Cluster
  18. Numerical Correctness Sweep
  19. Weekly Recap - Release Stabilization and Dynamo's Object Model Overhaul
  20. Dynamo's Great Untangling
  21. Attention Backend Correctness and CI Coverage Gaps
  22. The Great Hardware Decoupling
  23. The Great Device-Agnostic Test Rewrite
  24. Release Branch Cleanup and the Great Test Decoupling
  25. The Great Test Reclassification
  26. The Great Test Suite Decoupling
  27. Weekly Recap - Test Infrastructure Overhaul and Correctness Fixes
  28. The Great NaN and Negative Zero Cleanup
  29. Closing the GPU Test Coverage Gap
  30. The Great Device-Agnostic Test Migration
  31. Hardening Distributed Communication and Cutting Legacy Support
  32. Test Debt, GPU Correctness, and the Long Tail of Cleanup
  33. Distributed Backend Correctness Overhaul
  34. Boundary Checks and the Hardware-Agnostic Test Push
  35. Weekly Recap - Silent Correctness Fixes Take Center Stage
  36. Silent Corruption Sweep
  37. Compiler Internals Get Hardened Ahead of Bigger Changes
  38. Distributed Backends Get Serious About Parity
  39. A GEMM Epilogue Megastack and Distributed Hardening
  40. Cleaning Up the Compiler's Noise
  41. Crash Fixes and NCCL's Next Chapter
  42. Compiler Correctness Meets Test Portability
  43. Weekly Recap - Performance Fixes and Toolchain Modernization
  44. Toolchain Upgrades and Correctness Fixes Across the Stack
  45. Native Kernel Push and Distributed Backend Parity
  46. Correctness Hardening Across Inductor and AOTI
  47. The NVGEMM Epilogue Fusion Marathon
  48. Distributed Correctness Push
  49. CUDA Graphs Get a Lifecycle, Distributed Gets Cleaned Up
  50. One Engineer, Many Kernels
  51. Weekly Recap - Cleanup, Correctness, and a Random Number Overhaul
  52. Cleanup Sweeps and a New RNG Foundation
  53. Hardening Memory and Lifetime Management Across GPU Backends
  54. Correctness Fixes and Reverts Take Center Stage
  55. Distributed Backends and Overflow Fixes Take Center Stage
  56. Packaging Fixes and Bounds-Check Sweep
  57. Autograd Overhead Cuts and a New GEMM Backend Migration
  58. Hardening Boundary Checks and MPS Correctness
  59. Weekly Recap - Metal Kernels, Compile Reliability, and Dynamo's Long Tail
  60. Apple Silicon Gets a Metal Overhaul
  61. Dynamo's Great Type-System Cleanup
  62. Dynamo's CPython Test Push and a Rocky Exception-Handling Rewrite
  63. Distributed Backend Overhaul and a Wave of Cleanup
  64. The Great Runner Migration, and a Distributed Rollback
  65. Distributed Cleanup and CI Infrastructure Overhaul
  66. Correctness Fixes Sweep the Compiler Stack
  67. Weekly Recap - Release Hardening and Correctness Cleanup
  68. Determinism and Correctness Sweep
  69. Silent Correctness Bugs Take Center Stage
  70. FlexAttention's Rough Week
  71. Accelerator Backends and Memory Management
  72. Weekly Recap - Release Stabilization & Core Improvements
  73. GEMM Optimization and Core Stability Fixes
  74. C++20 Migration and Infrastructure Improvements
  75. Release Infrastructure and Compiler Stability Fixes
  76. Header Reorganization and Dynamo Improvements
  77. Infrastructure Organization and Core API Cleanup
  78. Dynamo Correctness and Build Infrastructure
  79. Performance Optimizations and Backend Improvements
  80. Weekly Recap - Compiler Optimization & Reliability
  81. Dynamo Compilation Stability and Performance
  82. Dynamo Stability and Inductor Optimizations
  83. Compiler Performance and Distributed Computing Infrastructure
  84. Distributed Naming Overhaul and Compiler Fixes
  85. Dynamo Stability and Compilation Robustness
  86. Compiler Numerical Accuracy and Index Expression Fixes
  87. Compiler Error Handling and Stability Fixes
  88. Weekly Recap - Stability and Error Handling Improvements
  89. Weekly Recap - Code Quality & Infrastructure Improvements
  90. Weekly Recap - Dynamo Enhancements & Critical Fixes
  91. Optimization Engine Revamp and Hash System Breakthrough
  92. Operator Improvements and Build System Updates
  93. Weekly Recap - Reverts and Infrastructure Updates
  94. Inductor Improvements and Bug Fixes
  95. Optimization Improvements and Cache Unification
  96. Weekly Recap - Performance Optimizations and Code Cleanup
  97. Weekly Recap - Infrastructure Stability & Mixed Precision
  98. Dynamo Improvements and Distributed Computing Fixes
  99. Weekly Recap - Performance Optimization and Infrastructure
  100. The Debugging Detective Story

Episode archive for PyTorch

  1. The Great Rollback and Recovery
  2. Exception Handling Revolution & Stateless RNG Arrives
  3. Cross-Platform Expansion and Developer Experience Wins
  4. Building Stronger Foundations
  5. Spring Cleaning & Major Infrastructure Upgrades
  6. AOT AutoGrad Fixes and Cross-Platform Polish
  7. Building Bridges - Distributed Computing Gets a Major Upgrade
  8. Profiling Power-Ups and Infrastructure Smoothing
  9. Fixes, Reverts, and Moving Forward
  10. The Infrastructure Acceleration Edition
  11. Lanczos Interpolation Breakthrough
  12. Stream Management Mastery & RNG Fixes
  13. Matrix Math Gets a Speed Boost
  14. Under the Hood Improvements and Future-Proofing
  15. Complex Math Gets Smarter & Build Improvements
  16. Memory Optimization Revolution
  17. Testing Gets Smarter and Graphs Go Universal
  18. Polish & Performance Day
  19. Distributed Computing Gets Real - Compilation, Clustering, and Convolutions
  20. Performance Revolution and Developer Experience Upgrades
  21. Windows Testing Gets Flexible & Dynamic Shapes Take Flight
  22. Metal Shaders Get a Precision Fix
  23. The Testing & Error Handling Polish Episode
  24. Stream Safety and Performance Wins
  25. Subclass Evolution and Memory Management Improvements
  26. Performance Tuning and Code Health Day
  27. Variable-Length Attention Gets Supercharged
  28. Spring Cleaning and Performance Boosts
  29. Stream Wizardry and Symbolic Shapes Magic
  30. CI Optimizations and Cross-Platform Fixes
  31. Spring Cleaning and Precision Fixes
  32. Memory Safety Fixes and Development Velocity
  33. Speed Wins and Better Error Messages
  34. Distributed Computing Gets Smarter
  35. Distributed Computing Gets Smarter & Vision Models Get Lightning Fast
  36. Release Dance and Rapid Recovery
  37. Performance Wins and Stability Fixes
  38. Distributed Computing Gets Smarter
  39. Valentine's Day Cleanup and Distributed Computing Love
  40. The Day of Rollbacks and Second Chances
  41. TPU Integration and the Dance of Reverts
  42. The Performance Optimization Sprint
  43. The Great Performance Revolution - Tests Run 70% Faster!
  44. Bug Fixes and Performance Wins
  45. The Great Test Speed Revolution
  46. Cleanup and Optimization Day
  47. Testing Cleanup and Pattern Matching Progress
  48. Type Safety Revolution and Infrastructure Cleanup
  49. Backend Flexibility Revolution
  50. The Great Configuration Cleanup & XPU Expansion
  51. Hardware Expansion and Developer Experience Polish
  52. Backend Harmony and Memory Magic
  53. Spring Cleaning and Building Blocks
  54. Bytecode Magic and Buffer Management Mastery
  55. Kernel Optimization and Clean Code Victory
  56. FMA Optimization Focus and Debugging Improvements
  57. Developer Tooling Revolution
  58. Deep Dive into PyTorch's Core - Opaque Objects and Performance Wins

More public developer podcasts

Browse other public Podlog show hubs with crawler-readable RSS and episode links.

  1. Next Js Daily 53 episodes
  2. Headroom Daily 62 episodes
  3. Agent Of Empires Daily 53 episodes
  4. Frigate NVR Updates 196 episodes
  5. Linux Kernel 187 episodes
  6. Homebrew 208 episodes
  7. TypeScript 97 episodes
  8. Ruby on Rails 201 episodes
  9. Redis 176 episodes
  10. Next.js 210 episodes
  11. Vue.js 149 episodes
  12. Rust 180 episodes
  13. Ruby Core Updates 203 episodes
  14. Python 193 episodes
  15. Go 193 episodes
  16. VS Code 211 episodes
  17. Node.js 205 episodes
  18. Django 197 episodes
  19. React Native 154 episodes
  20. PostgreSQL 197 episodes
  21. Kubernetes 206 episodes
  22. LangChain 191 episodes
  23. Ollama 195 episodes
  24. Openharness Daily 25 episodes
  25. Maestro Daily 52 episodes
  26. Videorc Daily 53 episodes
  27. Navidrome Daily 71 episodes
  28. Next.js Daily 212 episodes
  29. Pi Mono 92 episodes
  30. OpenClaw 169 episodes
  31. RuView 130 episodes
  32. Shannon 83 episodes
  33. Godot Daily 157 episodes
  34. Agora Next Updates 227 episodes
  35. Rails Daily 211 episodes
  36. Linux Kernel Daily 192 episodes
  37. Home Assistant Daily 216 episodes
  38. React Daily 108 episodes
  39. BlocksBeyondTheStars Daily 45 episodes
  40. Jabref Daily 63 episodes
  41. Jabref Daily 64 episodes
  42. iCloud Photos Downloader 27 episodes
  43. Onlook Design Updates 34 episodes
  44. Buzz Transcription 80 episodes
  45. The Algorithm Daily 15 episodes
  46. TailwindCSS 119 episodes
  47. AtomVM Daily 15 episodes
  48. NativeScript iOS Daily 26 episodes
  49. Agora Next Daily 12 episodes
  50. Energy Inc Ace Mate Learning Daily 2 episodes
  51. tiny-gpu Daily 20 episodes
  52. OpenAI Skills 52 episodes
  53. Crush Daily 1 episode
  54. ControlNet Daily 0 episodes
  55. FinceptTerminal Daily 0 episodes