executorch/backends/cuda/triton/kernels/__init__.py at cb76e2eaa7c374aa98284bc9bf8e404d7b239336 · pytorch/executorch · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the BSD-style license found in the
# LICENSE file in the root directory of this source tree.

from executorch.backends.cuda.triton.kernels.fused_moe import (
    fused_moe,
    fused_moe_batched,
    fused_moe_batched_gemm,
    moe_align_block_size,
)
from executorch.backends.cuda.triton.kernels.sdpa import sdpa, sdpa_decode_splitk
from executorch.backends.cuda.triton.kernels.topk import topk

__all__ = [
    "fused_moe",
    "fused_moe_batched",
    "fused_moe_batched_gemm",
    "moe_align_block_size",
    "sdpa",
    "sdpa_decode_splitk",
    "topk",
]

try:
    from executorch.backends.cuda.triton.kernels.chunk_gated_delta_rule import (  # noqa: F401
        chunk_gated_delta_rule,
    )

    __all__.append("chunk_gated_delta_rule")
except ImportError:
    pass

try:
    from executorch.backends.cuda.triton.kernels.tq4_sdpa import tq4_sdpa  # noqa: F401

    __all__.append("tq4_sdpa")
except ImportError:
    pass

try:
    from executorch.backends.cuda.triton.kernels.fused_deltanet_decode import (  # noqa: F401
        fused_deltanet_decode,
    )

    __all__.append("fused_deltanet_decode")
except ImportError:
    pass