Hugging Face's logo

Kernels:
kernels-community
/
flash-mla
Trusted publisher

Kernel card Files
xet
 Community
danieldk HF Staff commited on
Commit
f10ba3f
·
1 Parent(s): 745c77f

Remove builds incompatible with kernels >= 0.14

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. build/torch25-cxx11-cu118-x86_64-linux/flash_mla/__init__.py +0 -33
  2. build/torch25-cxx11-cu118-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
  3. build/torch25-cxx11-cu118-x86_64-linux/flash_mla/_ops.py +0 -9
  4. build/torch25-cxx11-cu121-x86_64-linux/flash_mla/__init__.py +0 -33
  5. build/torch25-cxx11-cu121-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
  6. build/torch25-cxx11-cu121-x86_64-linux/flash_mla/_ops.py +0 -9
  7. build/torch25-cxx11-cu124-x86_64-linux/flash_mla/__init__.py +0 -33
  8. build/torch25-cxx11-cu124-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
  9. build/torch25-cxx11-cu124-x86_64-linux/flash_mla/_ops.py +0 -9
  10. build/torch25-cxx98-cu118-x86_64-linux/flash_mla/__init__.py +0 -33
  11. build/torch25-cxx98-cu118-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
  12. build/torch25-cxx98-cu118-x86_64-linux/flash_mla/_ops.py +0 -9
  13. build/torch25-cxx98-cu121-x86_64-linux/flash_mla/__init__.py +0 -33
  14. build/torch25-cxx98-cu121-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
  15. build/torch25-cxx98-cu121-x86_64-linux/flash_mla/_ops.py +0 -9
  16. build/torch25-cxx98-cu124-x86_64-linux/flash_mla/__init__.py +0 -33
  17. build/torch25-cxx98-cu124-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
  18. build/torch25-cxx98-cu124-x86_64-linux/flash_mla/_ops.py +0 -9
  19. build/torch26-cxx11-cu118-x86_64-linux/flash_mla/__init__.py +0 -33
  20. build/torch26-cxx11-cu118-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
  21. build/torch26-cxx11-cu118-x86_64-linux/flash_mla/_ops.py +0 -9
  22. build/torch26-cxx11-cu124-x86_64-linux/flash_mla/__init__.py +0 -33
  23. build/torch26-cxx11-cu124-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
  24. build/torch26-cxx11-cu124-x86_64-linux/flash_mla/_ops.py +0 -9
  25. build/torch26-cxx11-cu126-aarch64-linux/flash_mla/__init__.py +0 -33
  26. build/torch26-cxx11-cu126-aarch64-linux/flash_mla/_flash_mla_341ab77.abi3.so +0 -3
  27. build/torch26-cxx11-cu126-aarch64-linux/flash_mla/_ops.py +0 -9
  28. build/torch26-cxx11-cu126-x86_64-linux/flash_mla/__init__.py +0 -33
  29. build/torch26-cxx11-cu126-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
  30. build/torch26-cxx11-cu126-x86_64-linux/flash_mla/_ops.py +0 -9
  31. build/torch26-cxx98-cu118-x86_64-linux/flash_mla/__init__.py +0 -33
  32. build/torch26-cxx98-cu118-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
  33. build/torch26-cxx98-cu118-x86_64-linux/flash_mla/_ops.py +0 -9
  34. build/torch26-cxx98-cu124-x86_64-linux/flash_mla/__init__.py +0 -33
  35. build/torch26-cxx98-cu124-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
  36. build/torch26-cxx98-cu124-x86_64-linux/flash_mla/_ops.py +0 -9
  37. build/torch26-cxx98-cu126-aarch64-linux/flash_mla/__init__.py +0 -33
  38. build/torch26-cxx98-cu126-aarch64-linux/flash_mla/_flash_mla_341ab77.abi3.so +0 -3
  39. build/torch26-cxx98-cu126-aarch64-linux/flash_mla/_ops.py +0 -9
  40. build/torch26-cxx98-cu126-x86_64-linux/flash_mla/__init__.py +0 -33
  41. build/torch26-cxx98-cu126-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
  42. build/torch26-cxx98-cu126-x86_64-linux/flash_mla/_ops.py +0 -9
  43. build/torch27-cxx11-cu118-x86_64-linux/flash_mla/__init__.py +0 -33
  44. build/torch27-cxx11-cu118-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
  45. build/torch27-cxx11-cu118-x86_64-linux/flash_mla/_ops.py +0 -9
  46. build/torch27-cxx11-cu126-aarch64-linux/flash_mla/__init__.py +0 -33
  47. build/torch27-cxx11-cu126-aarch64-linux/flash_mla/_flash_mla_341ab77.abi3.so +0 -3
  48. build/torch27-cxx11-cu126-aarch64-linux/flash_mla/_ops.py +0 -9
  49. build/torch27-cxx11-cu126-x86_64-linux/flash_mla/__init__.py +0 -33
  50. build/torch27-cxx11-cu126-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so +0 -3
build/torch25-cxx11-cu118-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch25-cxx11-cu118-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:a9a2b99b276b5aa714b27d1f54cc5da2d451e65a9ed385c583daf528f2c030a9
3
- size 2564144
 
 
 
 
build/torch25-cxx11-cu118-x86_64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_d4f4195
3
- ops = torch.ops._flash_mla_d4f4195
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_d4f4195::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch25-cxx11-cu121-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch25-cxx11-cu121-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:77e241f633fa5b103f379ba6ac58d2cc068e0c3fc4d4f20ac1e1c679fc19614f
3
- size 2595176
 
 
 
 
build/torch25-cxx11-cu121-x86_64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_d4f4195
3
- ops = torch.ops._flash_mla_d4f4195
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_d4f4195::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch25-cxx11-cu124-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch25-cxx11-cu124-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:31aa895a57efbd29aeff693b65b02842926bf1788d6f98022c32470a60265f9e
3
- size 2580248
 
 
 
 
build/torch25-cxx11-cu124-x86_64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_d4f4195
3
- ops = torch.ops._flash_mla_d4f4195
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_d4f4195::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch25-cxx98-cu118-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch25-cxx98-cu118-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:7454c10a3b29128e035bdb3fa18d5fc3706f7970542a0bcb55d9714f0999d42f
3
- size 2556792
 
 
 
 
build/torch25-cxx98-cu118-x86_64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_d4f4195
3
- ops = torch.ops._flash_mla_d4f4195
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_d4f4195::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch25-cxx98-cu121-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch25-cxx98-cu121-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:c8cb9402f3091420227cbccf1ec4938a444765e26f5d34c356c76bf7c85630d0
3
- size 2587896
 
 
 
 
build/torch25-cxx98-cu121-x86_64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_d4f4195
3
- ops = torch.ops._flash_mla_d4f4195
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_d4f4195::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch25-cxx98-cu124-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch25-cxx98-cu124-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:bb4be09cbde1979c1aa17e3bc93c1538f129b438d305bee0fe96f3c08efeee04
3
- size 2572968
 
 
 
 
build/torch25-cxx98-cu124-x86_64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_d4f4195
3
- ops = torch.ops._flash_mla_d4f4195
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_d4f4195::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx11-cu118-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx11-cu118-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:35c67c788220d8988e47cd4ad976495450b71cd682bd8ab08af3db066d625126
3
- size 2564496
 
 
 
 
build/torch26-cxx11-cu118-x86_64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_d4f4195
3
- ops = torch.ops._flash_mla_d4f4195
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_d4f4195::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx11-cu124-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx11-cu124-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:069fb3e3a051c91e73390245c7463218829b8decf0f60bd6fc9a0ba8127b5bd2
3
- size 2580592
 
 
 
 
build/torch26-cxx11-cu124-x86_64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_d4f4195
3
- ops = torch.ops._flash_mla_d4f4195
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_d4f4195::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx11-cu126-aarch64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx11-cu126-aarch64-linux/flash_mla/_flash_mla_341ab77.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:e1e97fef62f5ebbe6b19b0d5fbe700fcdf6b9acd7a54cba6f0b1d23665188fa9
3
- size 2643848
 
 
 
 
build/torch26-cxx11-cu126-aarch64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_341ab77
3
- ops = torch.ops._flash_mla_341ab77
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_341ab77::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx11-cu126-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx11-cu126-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:50fad86fa7bc15096c2a1feadf8091b20e188e32b8c0633423ec26e4e8e8e7ce
3
- size 2560552
 
 
 
 
build/torch26-cxx11-cu126-x86_64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_d4f4195
3
- ops = torch.ops._flash_mla_d4f4195
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_d4f4195::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx98-cu118-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx98-cu118-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:ae937ddfbc3e6097b2fdd9197f2ddb5b9f66c65146a4de30ccab59dab6e18dd4
3
- size 2557136
 
 
 
 
build/torch26-cxx98-cu118-x86_64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_d4f4195
3
- ops = torch.ops._flash_mla_d4f4195
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_d4f4195::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx98-cu124-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx98-cu124-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:503910324475f8bd9dab47687339005f58e5b623bf0c9e4234fabf099c08da33
3
- size 2573312
 
 
 
 
build/torch26-cxx98-cu124-x86_64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_d4f4195
3
- ops = torch.ops._flash_mla_d4f4195
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_d4f4195::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx98-cu126-aarch64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx98-cu126-aarch64-linux/flash_mla/_flash_mla_341ab77.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:6f15b3b0bd0bee56760bd6500175ca5a1fd17f2742ef9496c28ea3720d038c66
3
- size 2640208
 
 
 
 
build/torch26-cxx98-cu126-aarch64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_341ab77
3
- ops = torch.ops._flash_mla_341ab77
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_341ab77::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx98-cu126-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch26-cxx98-cu126-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:2c41fa4058ee2bb5d3d90458a7f92f0ef1c10e8bc854329cf7c208025bb244b2
3
- size 2553280
 
 
 
 
build/torch26-cxx98-cu126-x86_64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_d4f4195
3
- ops = torch.ops._flash_mla_d4f4195
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_d4f4195::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch27-cxx11-cu118-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch27-cxx11-cu118-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:59c4034880f4482b06e447a2c4810aaf8009b7d4c86a4fd71356f169df986535
3
- size 2564632
 
 
 
 
build/torch27-cxx11-cu118-x86_64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_d4f4195
3
- ops = torch.ops._flash_mla_d4f4195
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_d4f4195::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch27-cxx11-cu126-aarch64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch27-cxx11-cu126-aarch64-linux/flash_mla/_flash_mla_341ab77.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:fb925b062d31034672a45d925a3767d953e97a3c6c483467e6b81833d42b5a27
3
- size 2644048
 
 
 
 
build/torch27-cxx11-cu126-aarch64-linux/flash_mla/_ops.py DELETED
@@ -1,9 +0,0 @@
1
- import torch
2
- from . import _flash_mla_341ab77
3
- ops = torch.ops._flash_mla_341ab77
4
-
5
- def add_op_namespace_prefix(op_name: str):
6
- """
7
- Prefix op by namespace.
8
- """
9
- return f"_flash_mla_341ab77::{op_name}"
 
 
 
 
 
 
 
 
 
 
build/torch27-cxx11-cu126-x86_64-linux/flash_mla/__init__.py DELETED
@@ -1,33 +0,0 @@
1
- import torch
2
-
3
- from ._ops import ops
4
-
5
-
6
- def get_mla_metadata(seqlens_k: torch.Tensor, s_q: int, h_kv: int):
7
- return ops.get_mla_metadata(seqlens_k, s_q, h_kv)
8
-
9
-
10
- def mha_fwd_kvcache_mla(
11
- q: torch.Tensor,
12
- kcache: torch.Tensor,
13
- vcache_: torch.Tensor,
14
- head_size_v: int,
15
- seqlens_k: torch.Tensor,
16
- block_table: torch.Tensor,
17
- softmax_scale: float,
18
- is_causal_: bool,
19
- tile_scheduler_metadata: torch.Tensor,
20
- num_splits: torch.Tensor,
21
- ) -> torch.Tensor:
22
- return ops.mha_fwd_kvcache_mla(
23
- q,
24
- kcache,
25
- vcache_,
26
- head_size_v,
27
- seqlens_k,
28
- block_table,
29
- softmax_scale,
30
- is_causal_,
31
- tile_scheduler_metadata,
32
- num_splits
33
- )
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
build/torch27-cxx11-cu126-x86_64-linux/flash_mla/_flash_mla_d4f4195.abi3.so DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:5db69ef4975e2eee001e6a9b7466c1fe40bc2228ed64eb8c24caf3e0fb6ed0b2
3
- size 2560584