Issues

• prepare_softmax_blockwide3 loops forever when V < blockDim.x * x128::size (signed/unsigned comparison)

  karpathy/llm.c#868 · Filter by author nuxyel opened on Oct 20, 2026
  1

• Formal Refutation of Continuous Backpropagation Constraints: Truncating the ZFC Infinite Regress via j=1/j

  karpathy/llm.c#865 · Filter by author frintroper opened on Oct 3, 2026

• train_gpt2cu not compiling

  karpathy/llm.c#863 · Filter by author AliHaider20 opened on Sep 30, 2026

• memory-safety findings in llm.c CUDA kernels

  karpathy/llm.c#855 · Filter by author TarekIbnZiad opened on Jun 5, 2026

• Bare-metal AVX2/FMA SGEMM kernel for edge inference — 1.65ms @ 256×256

  karpathy/llm.c#848 · Filter by author gbatista-esalq opened on Apr 22, 2026

• Can we use Flash Attention 2?

  karpathy/llm.c#836 · Filter by author rajhlinux opened on Dec 22, 2025

• Is there any pretrain weights for this project?

  karpathy/llm.c#833 · Filter by author theodoruszq opened on Nov 6, 2025

• Why do I encounter a "Bad magic model file" error when running train_gpt2fp32cu?

  karpathy/llm.c#826 · Filter by author GGGrin8 opened on Sep 9, 2025

• Error nvcc

  karpathy/llm.c#824 · Filter by author mix127 opened on Jul 14, 2025
  1

• Some things that might be optimizations

  karpathy/llm.c#823 · Filter by author Swight1423 opened on Jul 6, 2025

• PMLL.c/.h/.CUDA GPU TWINS

  karpathy/llm.c#818 · Filter by author drQedwards opened on Jun 21, 2025
  1

• The PMLL

  karpathy/llm.c#817 · Filter by author drQedwards opened on Jun 21, 2025
  2

• The persistent memory logic loop by Dr. Edwards

  karpathy/llm.c#816 · Filter by author drQedwards opened on Jun 21, 2025

• cu compile fail

  karpathy/llm.c#813 · Filter by author huangruiteng opened on May 22, 2025

• MPI run with upto 3xH100 GPU works but starts giving an "illegal memory access" error when run using "mpirun -np 4 ./train_gpt2cu"

  karpathy/llm.c#807 · Filter by author dm1976-git opened on Apr 17, 2025
  1

• [Feature Req]: Any plan on parallel training like deepseeed(zero3, offload nvme) and megatron(tensor parallem, layer pileline)?

  karpathy/llm.c#806 · Filter by author SeekPoint opened on Apr 16, 2025
  1

• Runs under Windows

  karpathy/llm.c#804 · Filter by author code101015 opened on Apr 14, 2025
  1

• Kernel out-of-bounds access in normalization_kernel

  karpathy/llm.c#794 · Filter by author dndnda opened on Feb 6, 2025

• Dataloader shuffle_rng logic bug under multi-gpu settings?

  karpathy/llm.c#792 · Filter by author codedeft opened on Jan 6, 2025

• Token out of vocabulary at train_gpt2.cu:675

  karpathy/llm.c#786 · Filter by author aidando73 opened on Nov 20, 2024
  1

• Online Softmax is wrong

  karpathy/llm.c#776 · Filter by author NoSavedDATA opened on Oct 4, 2024

• Makefile incorrectly finds that nccl is installed for Linux systems with libvncclclient

  karpathy/llm.c#774 · Filter by author leiDnedyA opened on Oct 2, 2024
  1

• Question concerning float4

  karpathy/llm.c#767 · Filter by author dongrixinyu opened on Sep 25, 2024
  1

• [cudnn_frontend] Error: No execution plans support the graph.

  karpathy/llm.c#761 · Filter by author Necktwi opened on Sep 19, 2024
  5

• Will this repo update new documentation later?

  karpathy/llm.c#759 · Filter by author dongrixinyu opened on Sep 19, 2024