All issues

[Perf] a8w8_blockscale tuned table for gfx950 appears to be generated without --splitK: re-tuning gains 13-61% on low-M shapes

#5277

· stefanskiasan opened

on Sep 4, 2026

Untuned-shape logging exists only for bf16 (AITER_TUNE_GEMM); fp8/blockscale/a4w4 have none

#5267

· ThomasNing opened

on Sep 3, 2026

JIT baton dead-holder check is ambiguous under --network host (pid + hostname collide across containers)

#5266

· ThomasNing opened

on Sep 3, 2026

One infeasible tune candidate fails every kernel in the module (ninja aborts the whole build)

#5265

· ThomasNing opened

on Sep 3, 2026

CK Tile tune livelocks: the JIT build counts against the per-task --timeout

#5264

· ThomasNing opened

on Sep 3, 2026

GEMM tuners: an interrupted run loses everything (results are only written at the end)

#5263

· ThomasNing opened

on Sep 3, 2026

fused_moe_2stages passes num_rows to per_tensor_quant_hip, which asserts it is None

#5256

· i-kosarev opened

on Sep 3, 2026

[Perf] 0.1.21 slower than 0.1.19 for GLM-5.3 MXFP4 on gfx950 (−18 % prefill, −10 % throughput); FlyDSL MoE tuning?

#5243

· stefanskiasan opened

on Sep 3, 2026

MXFP4 MoE: CK-Tile stage-2 mis-indexes E8M0 scales when inter_dim % 256 != 0, returning silently wrong results

#5233

· xiaofei-zheng opened

on Sep 3, 2026

[Issue]: gfx1201 (R9700) — Triton unified_attention overflows RDNA's LDS budget, and the ROCm 10.0 vLLM container ships no gfx1201 code objects

#5229

· hanchi-gao opened

on Sep 3, 2026

[Bug]: MLA get_meta_param returns an lru_cache'd device tensor, corrupting the split indptr under CUDA graph replay

#5193

· peizhang56 opened

on Sep 2, 2026

[Bug]: Memory access fault due to top_k_renorm_probs

#5192

· peizhang56 opened

on Sep 1, 2026