A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 11:00 UTC
[BUG] [ROCm] [MLA] Fix variable name bug due to change in variable name in PR #17483 (#17961)
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>
TJian committed
May 12, 2025 at 00:14 UTC
7de18d541b0da661685d481d7306cbe5e9f7960b
1 file changed
+3
-3
vllm/v1/attention/backends/mla/rocm_aiter_mla.py
+3
-3
@@ -98,17 +98,17 @@ class AiterMLAMetadataBuilder(MLACommonMetadataBuilder[AiterMLAMetadata]):
98
paged_kv_last_page_len,
99
)
100
101
- def _build_decode(self, block_table: torch.Tensor,
101
+ def _build_decode(self, block_table_tensor: torch.Tensor,
102
seq_lens: torch.Tensor) -> AiterMLADecodeMetadata:
103
104
(
105
paged_kv_indices,
106
paged_kv_indptr,
107
paged_last_page_len,
108
- ) = self._get_paged_kv_tensors(block_table, seq_lens)
108
+ ) = self._get_paged_kv_tensors(block_table_tensor, seq_lens)
109
110
attn_metadata = AiterMLADecodeMetadata(
111
- block_table=block_table,
111
+ block_table=block_table_tensor,
112
seq_lens=seq_lens,
113
paged_kv_indptr=paged_kv_indptr,
114
paged_kv_indices=paged_kv_indices,