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 14, 05:30 UTC
[Mamba] - refactor: Renamed mamba_attn to mamba2_attn (#22818)
Signed-off-by: asafg <asafg@ai21.com> Co-authored-by: asafg <asafg@ai21.com>
Asaf Joseph Gardin committed
Aug 15, 2025 at 09:38 UTC
3d232dbd19f9c9d782a47b579f4a3c5a2f996499
6 files changed
+5
-5
tests/kernels/mamba/test_mamba_ssm_ssd.py
+1
-1
@@ -9,7 +9,7 @@ from einops import rearrange, repeat
9
from vllm.model_executor.layers.mamba.ops.ssd_combined import (
10
mamba_chunk_scan_combined)
11
from vllm.platforms import current_platform
12
-from vllm.v1.attention.backends.mamba_attn import (
12
+from vllm.v1.attention.backends.mamba2_attn import (
13
_query_start_loc_to_chunk_indices_offsets)
14
15
# Added by the IBM Team, 2024
tests/v1/attention/test_mamba_selectors.py
+1
-1
@@ -4,7 +4,7 @@
4
5
import pytest
6
7
-from vllm.v1.attention.backends.mamba_attn import Mamba2AttentionBackend
7
+from vllm.v1.attention.backends.mamba2_attn import Mamba2AttentionBackend
8
from vllm.v1.attention.backends.mamba_selectors import get_mamba_attn_backend
9
10
vllm/model_executor/layers/mamba/mamba2_metadata.py
+1
-1
@@ -11,7 +11,7 @@ from vllm.attention.backends.placeholder_attn import (
11
PlaceholderAttentionMetadata)
12
from vllm.attention.backends.utils import PAD_SLOT_ID
13
from vllm.platforms import current_platform
14
-from vllm.v1.attention.backends.mamba_attn import (
14
+from vllm.v1.attention.backends.mamba2_attn import (
15
Mamba2AttentionMetadata, _query_start_loc_to_chunk_indices_offsets)
16
17
vllm/model_executor/layers/mamba/mamba_mixer2.py
+1
-1
@@ -36,7 +36,7 @@ from vllm.model_executor.models.mamba_cache import MambaCacheParams
36
from vllm.model_executor.utils import set_weight_attrs
37
from vllm.platforms import current_platform
38
from vllm.utils import direct_register_custom_op
39
-from vllm.v1.attention.backends.mamba_attn import Mamba2AttentionMetadata
39
+from vllm.v1.attention.backends.mamba2_attn import Mamba2AttentionMetadata
40
41
# Added by the IBM Team, 2024
42
vllm/v1/attention/backends/mamba2_attn.py
renamed
vllm/v1/attention/backends/mamba_selectors.py
+1
-1
@@ -3,7 +3,7 @@
3
from vllm.attention.backends.abstract import AttentionBackend
4
from vllm.v1.attention.backends.linear_attn import LinearAttentionBackend
5
from vllm.v1.attention.backends.mamba1_attn import Mamba1AttentionBackend
6
-from vllm.v1.attention.backends.mamba_attn import Mamba2AttentionBackend
6
+from vllm.v1.attention.backends.mamba2_attn import Mamba2AttentionBackend
7
8
9
def get_mamba_attn_backend(mamba_type: str) -> type[AttentionBackend]: