[Kernel] Gemma-4 FA4 FP8 Kernel (#48666)

Signed-off-by: Qixiang Lin <qixiangl@nvidia.com>
Signed-off-by: Jhao-Ting Chen <jhaotingc@nvidia.com>
Co-authored-by: Qixiang Lin <qixiangl@nvidia.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Yongye Zhu <zyy1102000@gmail.com>
This commit is contained in:
Jhao-Ting Chen
2026-08-13 17:42:00 -07:00
committed by GitHub
co-authored by Qixiang Lin Claude Opus 4.8 Yongye Zhu
parent 38f097fab8
commit 827a2af806
8 changed files with 152 additions and 29 deletions
+3 -3
View File
@@ -241,9 +241,9 @@ def flash_attn_supports_kv_cache_dtype(
head_size_v=head_size_v,
has_sinks=has_sinks,
)
return (fa_version == 3 and current_platform.is_device_capability_family(90)) or (
fa_version == 4 and current_platform.is_device_capability_family(100)
)
return (
fa_version in (3, 4) and current_platform.is_device_capability_family(90)
) or (fa_version == 4 and current_platform.is_device_capability_family(100))
def flash_attn_supports_quant_query_input() -> bool: