ltqin
bf07a0150e
[rocm-libraries] ROCm/rocm-libraries#7723 (commit 4ed6c51)
[CK Tile] Enable LSE output for fp8bf16 V3 FMHA kernels
(#7723)
### Motivation
The V3 pipeline (qr_async_trload_v3) for fp8bf16 FMHA kernels did not
support LSE (Log-Sum-Exp) output. This PR enables LSE output support for
fp8bf16 V3 FMHA kernels, allowing users to retrieve attention statistics
alongside attention outputs.
### Technical Details
- StandardAttention: lse = softmax_scale * m + log(l)
- LogitsSoftCap: lse = (m / log2(e)) + log(l)
### Test Plan
Run FMHA forward example with fp8bf16 precision and LSE output enabled:
- Test 1: Basic LSE functionality
./build/bin/tile_example_fmha_fwd -v=1 -b=1 -h=8 -s=1024 -d=128
-prec=fp8bf16 -init=3 -qscale=1 -lse=1
- Test 2: LSE with LogitsSoftCap (CMakeList should remove Logits filter)
./build/bin/tile_example_fmha_fwd -v=1 -b=1 -h=8 -s=1024 -d=128
-prec=fp8bf16 -init=3 -qscale=1 -lse=1 -logits_soft_cap=30.0
2026-05-28 15:58:54 +00:00
..
2026-05-22 02:43:50 +00:00
2026-05-15 06:46:51 -07:00
2025-11-28 13:49:54 -08:00
2026-05-15 06:46:51 -07:00
2026-05-15 06:46:51 -07:00
2026-05-15 06:46:51 -07:00
2025-10-16 03:10:57 -07:00
2026-03-06 09:27:27 -07:00
2026-01-07 16:30:57 +01:00
2026-03-12 09:47:41 +01:00
2026-05-27 06:56:58 -07:00
2026-05-15 06:46:51 -07:00
2025-12-18 07:59:45 +01:00
2025-11-28 13:49:54 -08:00
2026-05-15 06:46:51 -07:00
2026-05-18 17:46:01 +02:00
2026-05-15 06:46:51 -07:00
2026-05-15 06:46:51 -07:00
2026-01-07 16:30:57 +01:00
2026-05-15 06:46:51 -07:00
2026-01-17 08:30:27 +01:00
2026-05-22 02:43:50 +00:00
2025-11-28 13:49:54 -08:00
2026-05-27 06:56:58 -07:00
2026-05-27 06:56:58 -07:00
2026-05-15 06:46:51 -07:00
2026-05-15 06:46:51 -07:00
2026-05-22 02:43:50 +00:00
2026-01-07 16:30:57 +01:00
2025-11-28 13:49:54 -08:00
2026-05-15 06:46:51 -07:00
2026-05-27 06:56:58 -07:00
2026-01-20 13:06:59 -08:00
2026-05-27 09:59:14 +03:00
2026-05-27 06:56:58 -07:00
2026-05-08 07:14:14 -07:00
2026-05-15 06:46:51 -07:00
2026-01-07 16:30:57 +01:00
2026-05-27 06:56:58 -07:00
2026-01-07 16:30:57 +01:00
2026-01-07 16:30:57 +01:00
2025-11-28 13:49:54 -08:00
2025-11-28 13:49:54 -08:00
2025-11-28 13:49:54 -08:00
2025-11-28 13:49:54 -08:00
2025-11-28 13:49:54 -08:00
2025-11-28 13:49:54 -08:00
2025-11-28 13:49:54 -08:00
2025-11-28 13:49:54 -08:00
2025-11-28 13:49:54 -08:00
2026-02-25 05:16:07 +00:00
2026-05-15 06:46:51 -07:00
2025-11-28 13:49:54 -08:00
2026-05-22 02:43:50 +00:00
2025-11-28 13:49:54 -08:00
2026-05-22 02:43:50 +00:00
2026-05-27 18:55:15 +00:00
2026-05-22 02:43:50 +00:00
2026-05-15 06:46:51 -07:00
2026-01-15 16:43:02 +01:00
2026-01-15 16:43:02 +01:00
2026-05-28 15:58:54 +00:00
2026-05-25 11:26:26 +02:00
2024-12-04 00:46:47 +01:00