Search results

check_cuda_arch() rejects SM 12.0 (Blackwell) due to string comparison

needs-triage

Status: Open.

#3945 In flashinfer-ai/flashinfer;

· 1morello opened on Jul 13, 2026

[bug]b300 test_groupwise_scaled_gemm_fp8 Mismatched elements: 124 / 8192 (1.5%)

needs-triage

Status: Open.

#3944 In flashinfer-ai/flashinfer;

· nvamyt opened on Jul 13, 2026

[Bug]Flaky issue on B300 and GB200: test_trtllm_gen_moe_autotune_tactics.py timeout with no results

op: moe

Status: Open.

#3936 In flashinfer-ai/flashinfer;

· nvamyt opened on Jul 12, 2026

[Bug] ~11% long-prefill regression on SM120 (RTX 5090) between 0.6.12 and 0.6.13, hybrid GDN model via vLLM

needs-triage

Status: Open.

#3935 In flashinfer-ai/flashinfer;

· jhsmith409 opened on Jul 12, 2026

[RFC] Should FlashInfer explore an E0M3-based NVINT4 W4A4 path on Blackwell?

needs-triage

Status: Open.

#3934 In flashinfer-ai/flashinfer;

· im0qianqian opened on Jul 12, 2026

trtllm_batch_decode_with_kv_cache: runFmhaReduction IMA on multi-bs CUDA graph capture (Gemma-4 hd=512, q_len=6)

needs-triage

Status: Open.

#3929 In flashinfer-ai/flashinfer;

· kpham-sgl opened on Jul 11, 2026

test(moe): cover do_finalize=False expert_weights dtype in the unified MoE fuzzer

op: moe

unified_apiUnified API task tracking - promoting good API practices, streamlining kernel access and robustnessUnified API task tracking - promoting good API practices, streamlining kernel access and robustness

Status: Open.

#3926 In flashinfer-ai/flashinfer;

· aleozlx opened on Jul 11, 2026

[RFC] Autotuner v2: FlashInfer-managed persistent cache and deployment-matched measurement

autotuner

RFCRequest for CommentsRequest for Comments

Status: Open.

#3920 In flashinfer-ai/flashinfer;

· YangXu1990uiuc opened on Jul 10, 2026

[Bug] TRTLLM-GEN MLA rejects valid native-width block tables and forces padding

needs-triage

Status: Open.

#3915 In flashinfer-ai/flashinfer;

· yihwang-nv opened on Jul 10, 2026

[bug] test_trtllm_fused_moe_autotuner_integration.py : 'ProfilingCacheKey' object is not subscriptable

op: moe

Status: Open.

#3913 In flashinfer-ai/flashinfer;

· nvamyt opened on Jul 10, 2026

Adjust vectorized load vs. TMA boundary for flashinfer.nvfp4_quantize

op: misc

performance

Status: Open.

#3905 In flashinfer-ai/flashinfer;

· kahyunnam opened on Jul 10, 2026

[Feature] Warn or raise when wrapper run() is captured in a CUDA graph without use_cuda_graph=True

needs-triage

Status: Open.

#3904 In flashinfer-ai/flashinfer;

· mgoin opened on Jul 9, 2026