Close matches. Each row lists what differs.
Differs from the request. dtype: requested bf16, observed fp32/fp8_e4m3/int8. The source's designated baseline implementation. Reported by source; not independently reproduced.
Differs from the request. dtype: requested bf16, observed fp32/fp8_e4m3/int8; m: requested 128, observed 256. The source's designated baseline implementation. Reported by source; not independently reproduced.
Differs from the request. dtype: requested bf16, observed fp32/fp8_e4m3/int8; m: requested 128, observed 1024. The source's designated baseline implementation. Reported by source; not independently reproduced.
Missing a kernel here? Evidence submissions open with the contribution beta.
FlashInfer-Bench · last observed 2026-06-06