Search results

[Feature Request] Support NVFP4 KV Cache Kernel on SM120 through XQA

needs-triage

Status: Open.

#2294 In flashinfer-ai/flashinfer;

· samuellees opened on Jan 6, 2026

Feature Request: Support NVFP4 KV Cache Attention Kernel on SM100 through TRTLLM-GEN Backend

needs-triage

Status: Open.

#2293 In flashinfer-ai/flashinfer;

· samuellees opened on Jan 6, 2026

Feature Request: Move fmha_v2 cubins into flashinfer

needs-triage

Status: Open.

#2292 In flashinfer-ai/flashinfer;

· qsang-nv opened on Jan 6, 2026

[Feature Request] Update TRTLLM GEN Fmha Kernel to the Latest Version

needs-triage

Status: Open.

#2291 In flashinfer-ai/flashinfer;

· samuellees opened on Jan 6, 2026

Feature Request: Add bias input in gemm interfaces

needs-triage

Status: Open.

#2290 In flashinfer-ai/flashinfer;

· qsang-nv opened on Jan 6, 2026

sampling_from_probs always returns the same prediction for the same input probs

needs-triage

Status: Open.

#2284 In flashinfer-ai/flashinfer;

· WeiXiaoSummer opened on Jan 2, 2026

Is the FA3 backend for prefill batch-invariant when the number of tokens is fixed?

needs-triage

Status: Open.

#2283 In flashinfer-ai/flashinfer;

· rajagond opened on Jan 2, 2026

trtllm_alltoall sm120

needs-triage

Status: Open.

#2282 In flashinfer-ai/flashinfer;

· voipmonitor opened on Jan 2, 2026

[BUG] Failed to initialize the TMA descriptor due to an illegal memory access.

needs-triage

Status: Open.

#2274 In flashinfer-ai/flashinfer;

· xiaoxiaosuaxuan opened on Dec 31, 2025

norm.cuh fails to compile without ENABLE_FP8

needs-triage

Status: Open.

#2271 In flashinfer-ai/flashinfer;

· KevinL10 opened on Dec 29, 2025

Batch-invariant chunked prefill with FlashInfer Backend

needs-triage

Status: Open.

#2270 In flashinfer-ai/flashinfer;

· rajagond opened on Dec 27, 2025

trtllm_moe_finalize_allreduce_fusion - slower for batch sizes >= 16 on DeepSeek V3

needs-triage

Status: Open.

#2269 In flashinfer-ai/flashinfer;

· aidando73 opened on Dec 26, 2025