Issues

• What is the motivation behind replacing the fp16 Neumann inverse with an 8x8 fp32 forward substitution

  MoonshotAI/FlashKDA#34 · Filter by author zilanlia opened on Sep 11, 2026
  1

• Add a differentiable CPU backend for KDA research and prototyping

  MoonshotAI/FlashKDA#26 · Filter by author xy200303 opened on Jul 29, 2026

• Add Pythonic API wrapper with automatic workspace management

  MoonshotAI/FlashKDA#23 · Filter by author haix716 opened on Jul 28, 2026

• Add random seed to bench_fwd.py for reproducible benchmarks

  MoonshotAI/FlashKDA#22 · Filter by author haix716 opened on Jul 28, 2026

• Add standalone usage example and inference-only note

  MoonshotAI/FlashKDA#21 · Filter by author haix716 opened on Jul 28, 2026

• Centralize CHUNK and D constants as single source of truth

  MoonshotAI/FlashKDA#20 · Filter by author haix716 opened on Jul 28, 2026

• Surface CUTLASS initialization and CUDA launch-configuration failures

  MoonshotAI/FlashKDA#15 · Filter by author morluto opened on Jul 27, 2026

• Partial output tiles serialize stores through one lane

  MoonshotAI/FlashKDA#14 · Filter by author morluto opened on Jul 27, 2026

• [Bug] test_fwd failed with CUDA 13

  MoonshotAI/FlashKDA#12 · Filter by author qianyp18 opened on Jul 17, 2026

• Publish FlashKDA kernels on Hugging Face Kernels?

  MoonshotAI/FlashKDA#11 · Filter by author NielsRogge opened on Jul 15, 2026
  6

• [Feature] Intra-Card Context Parallelism to improve B=1 long-sequence performance

  MoonshotAI/FlashKDA#8 · Filter by author yyq0210 opened on May 29, 2026

• [Question] Roadmap for supporting additional K/V head dimensions and backward kernels

  MoonshotAI/FlashKDA#4 · Filter by author dmm19941210 opened on Apr 24, 2026
  3

• gate mode diff

  MoonshotAI/FlashKDA#3 · Filter by author theNiemand opened on Apr 23, 2026
  4

• intermidiate ssm state store support

  MoonshotAI/FlashKDA#2 · Filter by author theNiemand opened on Apr 23, 2026