[XPU][CI] Make test_mamba_prefix_cache block-size agnostic (#58269)

Signed-off-by: Lin, Fanli <fanli.lin@intel.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
This commit is contained in:
Fanli Lin
2026-09-29 07:12:43 +08:00
committed by GitHub
co-authored by Kunshang Ji
parent 3bd7302396
commit 6a6c2952fa
2 changed files with 110 additions and 103 deletions
+7 -4
View File
@@ -464,13 +464,16 @@ class XPUPlatform(Platform):
)
cache_config.block_size = new_block_size
logger.info(
"[XPU]Setting attention block size to %d tokens to ensure multiple of %d, "
"set mamba_page_size_padded to %d bytes accordingly, before was %d bytes.",
"[XPU]Setting attention block size to %d tokens to ensure multiple of %d.",
new_block_size,
kernel_block_size,
cache_config.mamba_page_size_padded,
original_mamba_page_size_padded,
)
if original_mamba_page_size_padded is not None:
logger.info(
"[XPU]Scaled mamba_page_size_padded from %d to %d bytes accordingly.",
original_mamba_page_size_padded,
cache_config.mamba_page_size_padded,
)
@classmethod
def support_hybrid_kv_cache(cls) -> bool: