[Bugfix][ROCm] Fix memory access fault in AITER MLA backend for DPA+FP8 KV (#47276)
Signed-off-by: simondanielsson <simon.danielsson99@hotmail.com> Co-authored-by: nnyrhila <niko.nyrhila@amd.com>
This commit is contained in:
co-authored by
nnyrhila
parent
567a78432d
commit
482e5524fe
@@ -196,6 +196,7 @@ class AiterMLAMetadataBuilder(MLACommonMetadataBuilder[AiterMLAMetadata]):
|
||||
kv_cache_dtype_str = getattr(vllm_config.cache_config, "cache_dtype", "auto")
|
||||
if kv_cache_dtype_str in ("fp8", "fp8_e4m3", "fp8_e5m2"):
|
||||
kv_cache_dtype_str = "fp8"
|
||||
q_dtype = dtypes.fp8
|
||||
else:
|
||||
kv_cache_dtype_str = "bf16"
|
||||
kv_dtype = dtypes.d_dtypes.get(kv_cache_dtype_str, dtypes.bf16)
|
||||
|
||||
Reference in New Issue
Block a user