tcgen05

Here are 3 public repositories matching this topic...

RishiShah99 / lethe

Native Blackwell (sm_100) tcgen05 training backward for the gated-linear-recurrence family (GDN/KDA/GLA/SSD/LA), plus a contract-grade verifier that audits LLM-generated GPU kernels. Six Mamba-3 Triton kernels; GRPO trainer graded on contract correctness. arXiv:2608.12700

cuda triton cutlass mamba gpu-kernels state-space-models blackwell mamba-3 tcgen05 kernel-verification

Updated Oct 18, 2026
Python

qdox-labs / thor-hacks

Star

Measured sm_110 / sm_110a facts for NVIDIA Jetson AGX Thor: tensor-core capability matrix (2:4 sparsity works on plain sm_110; tcgen05 + NVFP4 block-scale is sm_110a-only), CUDA 11/12->13 migration breaks, CPU ISA, and the bandwidth roofline. With probes.

Updated Aug 16, 2026
Shell

idrees2516 / deepgemm-rust

Star

DeepGEMM in pure Rust: FP8/BF16 fine-grained-scaled GEMM + MoE grouped GEMM for NVIDIA SM90a (TMA+WGMMA warp-specialized kernels, NVRTC runtime JIT, zero torch/nvcc dependency)

rust gpu cuda moe hopper gemm b200 nvrtc blackwell fp8 deepseek fp4 wgmma deepgemm tcgen05

Updated Oct 4, 2026
Rust

Add this topic to your repo

To associate your repository with the tcgen05 topic, visit your repo's landing page and select "manage topics."

Learn more