Uploaded using `kernel-builder`.
Browse files
build/torch211-cxx11-cu130-x86_64-linux/__init__.py
CHANGED
|
@@ -7,6 +7,37 @@ import torch
|
|
| 7 |
from ._ops import add_op_namespace_prefix, ops
|
| 8 |
|
| 9 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
def sfa_size_bytes(rows: int, dim: int) -> int:
|
| 11 |
if rows <= 0 or dim <= 0 or dim % 16 != 0:
|
| 12 |
raise ValueError("rows must be positive and dim must be positive/divisible by 16")
|
|
@@ -76,6 +107,34 @@ def _gemv_warpsplit_fake(a_packed, b_packed, sfa, sfb, out, alpha: float = 1.0,
|
|
| 76 |
return None
|
| 77 |
|
| 78 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 79 |
@torch.library.register_fake(add_op_namespace_prefix("fp4_w4a16_linear_bf16"))
|
| 80 |
def _legacy_linear_fake(
|
| 81 |
a_packed: torch.Tensor,
|
|
@@ -559,6 +618,89 @@ def fp4_w4a4_gemv_warpsplit_bf16(
|
|
| 559 |
return out
|
| 560 |
|
| 561 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 562 |
def fp4_w4a16_linear_bf16(
|
| 563 |
a_packed: torch.Tensor,
|
| 564 |
b_packed: torch.Tensor,
|
|
@@ -714,16 +856,26 @@ def nvfp4_gemm_streamk_bias_bf16(
|
|
| 714 |
|
| 715 |
|
| 716 |
__all__ = [
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 717 |
"aligned_fp4_dim",
|
| 718 |
"cutlass_fp4_gemm_geglu_il_hw_v10",
|
| 719 |
"dequantize_fp4_sfa_fp16",
|
| 720 |
"e0m3_weight_gemm_fp16",
|
| 721 |
"fp4_w4a16_linear_bf16",
|
|
|
|
| 722 |
"fp4_w4a4_gemv_warpsplit_bf16",
|
|
|
|
| 723 |
"nvfp4_gemm_bf16",
|
| 724 |
"nvfp4_gemm_fp16",
|
| 725 |
"nvfp4_gemm_variant_bf16",
|
| 726 |
"nvfp4_gemm_nvfp4",
|
|
|
|
|
|
|
| 727 |
"nvfp4_gemm_geglu_nvfp4_fp16",
|
| 728 |
"nvfp4_gemm_bias_gelu_nvfp4_fp16",
|
| 729 |
"nvfp4_gemm_bias_residual_fp16",
|
|
|
|
| 7 |
from ._ops import add_op_namespace_prefix, ops
|
| 8 |
|
| 9 |
|
| 10 |
+
NVFP4_BLOCK_SIZE = 16
|
| 11 |
+
NVFP4_SCALE_TILE_ROWS = 128
|
| 12 |
+
NVFP4_SCALE_TILE_BLOCKS = 4
|
| 13 |
+
SUPPORTED_LAYOUTS = ("row-major-packed-e2m1", "cutlass-sm1xx-blockscaled")
|
| 14 |
+
SUPPORTED_CUDA_CAPABILITIES = ("11.0a", "12.0a")
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
def capabilities() -> dict[str, object]:
|
| 18 |
+
"""Return the public Tensor API contract used by runtime dispatchers."""
|
| 19 |
+
return {
|
| 20 |
+
"quantization": "W4A4 NVFP4 E2M1",
|
| 21 |
+
"block_size": NVFP4_BLOCK_SIZE,
|
| 22 |
+
"scale_layout": "pad128(rows) x pad4(dim/16), 512 bytes per tile",
|
| 23 |
+
"scale_size": "ceil(rows/128) * ceil((dim/16)/4) * 512 bytes",
|
| 24 |
+
"layouts": SUPPORTED_LAYOUTS,
|
| 25 |
+
"cuda_capabilities": SUPPORTED_CUDA_CAPABILITIES,
|
| 26 |
+
"public_m_alignment": 1,
|
| 27 |
+
"raw_sm120_tile_m": 128,
|
| 28 |
+
"sm120_m256_min_m": 512,
|
| 29 |
+
"sm120_m256_qualified_nk": (
|
| 30 |
+
(17408, 5120),
|
| 31 |
+
(5120, 17408),
|
| 32 |
+
(12288, 5120),
|
| 33 |
+
),
|
| 34 |
+
"sm120_m256_diagnostic_nk": ((16384, 5120),),
|
| 35 |
+
"sm120_interleaved_gemv_m": 1,
|
| 36 |
+
"sm120_interleaved_weight_layout": "groups-of-8 x K/64 x 8 x 32B",
|
| 37 |
+
"errors": "exceptions",
|
| 38 |
+
}
|
| 39 |
+
|
| 40 |
+
|
| 41 |
def sfa_size_bytes(rows: int, dim: int) -> int:
|
| 42 |
if rows <= 0 or dim <= 0 or dim % 16 != 0:
|
| 43 |
raise ValueError("rows must be positive and dim must be positive/divisible by 16")
|
|
|
|
| 107 |
return None
|
| 108 |
|
| 109 |
|
| 110 |
+
@torch.library.register_fake(add_op_namespace_prefix("fp4_repack_b_interleaved_sm120"))
|
| 111 |
+
def _repack_b_interleaved_fake(b_packed, b_interleaved) -> None:
|
| 112 |
+
if b_interleaved.shape != b_packed.shape:
|
| 113 |
+
raise RuntimeError("b_interleaved must have the same shape as b_packed")
|
| 114 |
+
return None
|
| 115 |
+
|
| 116 |
+
|
| 117 |
+
@torch.library.register_fake(add_op_namespace_prefix("fp4_w4a4_gemv_warpsplit_interleaved_bf16"))
|
| 118 |
+
def _gemv_warpsplit_interleaved_fake(
|
| 119 |
+
a_packed, b_interleaved, sfa, sfb, out,
|
| 120 |
+
alpha: float = 1.0, warps: int = 4, stages: int = 4,
|
| 121 |
+
) -> None:
|
| 122 |
+
if a_packed.shape[0] != 1:
|
| 123 |
+
raise RuntimeError("interleaved warp-split GEMV serves M=1 only")
|
| 124 |
+
if out.shape != (1, b_interleaved.shape[0]):
|
| 125 |
+
raise RuntimeError("out must have shape (1, N)")
|
| 126 |
+
return None
|
| 127 |
+
|
| 128 |
+
|
| 129 |
+
@torch.library.register_fake(add_op_namespace_prefix("nvfp4_gemm_m256_bf16"))
|
| 130 |
+
def _m256_fake(a_packed, b_packed, sfa, sfb, workspace, out, alpha: float = 1.0) -> None:
|
| 131 |
+
if a_packed.shape[0] < 512:
|
| 132 |
+
raise RuntimeError("the M256 tier requires M >= 512")
|
| 133 |
+
if out.shape != (a_packed.shape[0], b_packed.shape[0]):
|
| 134 |
+
raise RuntimeError("out must have shape (M, N)")
|
| 135 |
+
return None
|
| 136 |
+
|
| 137 |
+
|
| 138 |
@torch.library.register_fake(add_op_namespace_prefix("fp4_w4a16_linear_bf16"))
|
| 139 |
def _legacy_linear_fake(
|
| 140 |
a_packed: torch.Tensor,
|
|
|
|
| 618 |
return out
|
| 619 |
|
| 620 |
|
| 621 |
+
def fp4_repack_b_interleaved_sm120(
|
| 622 |
+
b_packed: torch.Tensor,
|
| 623 |
+
*,
|
| 624 |
+
out: Optional[torch.Tensor] = None,
|
| 625 |
+
) -> torch.Tensor:
|
| 626 |
+
"""Repack dense row-major packed FP4 weights for the SM120 M=1 GEMV.
|
| 627 |
+
|
| 628 |
+
This is a bind-time operation. Cache the returned tensor with the packed
|
| 629 |
+
weight; do not execute it in the decode hot path.
|
| 630 |
+
"""
|
| 631 |
+
if out is None:
|
| 632 |
+
out = torch.empty_like(b_packed)
|
| 633 |
+
ops.fp4_repack_b_interleaved_sm120(b_packed, out)
|
| 634 |
+
return out
|
| 635 |
+
|
| 636 |
+
|
| 637 |
+
def fp4_w4a4_gemv_warpsplit_interleaved_bf16(
|
| 638 |
+
a_packed: torch.Tensor,
|
| 639 |
+
b_interleaved: torch.Tensor,
|
| 640 |
+
sfa: torch.Tensor,
|
| 641 |
+
sfb: torch.Tensor,
|
| 642 |
+
*,
|
| 643 |
+
alpha: float = 1.0,
|
| 644 |
+
warps: int = 8,
|
| 645 |
+
stages: int = 3,
|
| 646 |
+
out: Optional[torch.Tensor] = None,
|
| 647 |
+
) -> torch.Tensor:
|
| 648 |
+
"""M=1 SM120 W4A4 GEMV using bind-time interleaved FP4 weights."""
|
| 649 |
+
if out is None:
|
| 650 |
+
out = torch.empty(
|
| 651 |
+
(1, b_interleaved.shape[0]),
|
| 652 |
+
device=a_packed.device,
|
| 653 |
+
dtype=torch.bfloat16,
|
| 654 |
+
)
|
| 655 |
+
ops.fp4_w4a4_gemv_warpsplit_interleaved_bf16(
|
| 656 |
+
a_packed, b_interleaved, sfa, sfb, out,
|
| 657 |
+
float(alpha), int(warps), int(stages),
|
| 658 |
+
)
|
| 659 |
+
return out
|
| 660 |
+
|
| 661 |
+
|
| 662 |
+
def nvfp4_gemm_m256_workspace_size(
|
| 663 |
+
a_packed: torch.Tensor,
|
| 664 |
+
b_packed: torch.Tensor,
|
| 665 |
+
sfa: torch.Tensor,
|
| 666 |
+
sfb: torch.Tensor,
|
| 667 |
+
) -> int:
|
| 668 |
+
"""Return workspace bytes for the SM120 large-M tier.
|
| 669 |
+
|
| 670 |
+
Query and allocate once before CUDA Graph capture.
|
| 671 |
+
"""
|
| 672 |
+
return int(ops.nvfp4_gemm_m256_workspace_size(a_packed, b_packed, sfa, sfb))
|
| 673 |
+
|
| 674 |
+
|
| 675 |
+
def nvfp4_gemm_m256_bf16(
|
| 676 |
+
a_packed: torch.Tensor,
|
| 677 |
+
b_packed: torch.Tensor,
|
| 678 |
+
sfa: torch.Tensor,
|
| 679 |
+
sfb: torch.Tensor,
|
| 680 |
+
*,
|
| 681 |
+
workspace: Optional[torch.Tensor] = None,
|
| 682 |
+
alpha: float = 1.0,
|
| 683 |
+
out: Optional[torch.Tensor] = None,
|
| 684 |
+
) -> torch.Tensor:
|
| 685 |
+
"""SM120 M>=512 NVFP4 GEMM with a caller-owned graph-stable workspace."""
|
| 686 |
+
if workspace is None:
|
| 687 |
+
workspace = torch.empty(
|
| 688 |
+
nvfp4_gemm_m256_workspace_size(a_packed, b_packed, sfa, sfb),
|
| 689 |
+
device=a_packed.device,
|
| 690 |
+
dtype=torch.uint8,
|
| 691 |
+
)
|
| 692 |
+
if out is None:
|
| 693 |
+
out = torch.empty(
|
| 694 |
+
(a_packed.shape[0], b_packed.shape[0]),
|
| 695 |
+
device=a_packed.device,
|
| 696 |
+
dtype=torch.bfloat16,
|
| 697 |
+
)
|
| 698 |
+
ops.nvfp4_gemm_m256_bf16(
|
| 699 |
+
a_packed, b_packed, sfa, sfb, workspace, out, float(alpha)
|
| 700 |
+
)
|
| 701 |
+
return out
|
| 702 |
+
|
| 703 |
+
|
| 704 |
def fp4_w4a16_linear_bf16(
|
| 705 |
a_packed: torch.Tensor,
|
| 706 |
b_packed: torch.Tensor,
|
|
|
|
| 856 |
|
| 857 |
|
| 858 |
__all__ = [
|
| 859 |
+
"NVFP4_BLOCK_SIZE",
|
| 860 |
+
"NVFP4_SCALE_TILE_ROWS",
|
| 861 |
+
"NVFP4_SCALE_TILE_BLOCKS",
|
| 862 |
+
"SUPPORTED_CUDA_CAPABILITIES",
|
| 863 |
+
"SUPPORTED_LAYOUTS",
|
| 864 |
+
"capabilities",
|
| 865 |
"aligned_fp4_dim",
|
| 866 |
"cutlass_fp4_gemm_geglu_il_hw_v10",
|
| 867 |
"dequantize_fp4_sfa_fp16",
|
| 868 |
"e0m3_weight_gemm_fp16",
|
| 869 |
"fp4_w4a16_linear_bf16",
|
| 870 |
+
"fp4_repack_b_interleaved_sm120",
|
| 871 |
"fp4_w4a4_gemv_warpsplit_bf16",
|
| 872 |
+
"fp4_w4a4_gemv_warpsplit_interleaved_bf16",
|
| 873 |
"nvfp4_gemm_bf16",
|
| 874 |
"nvfp4_gemm_fp16",
|
| 875 |
"nvfp4_gemm_variant_bf16",
|
| 876 |
"nvfp4_gemm_nvfp4",
|
| 877 |
+
"nvfp4_gemm_m256_bf16",
|
| 878 |
+
"nvfp4_gemm_m256_workspace_size",
|
| 879 |
"nvfp4_gemm_geglu_nvfp4_fp16",
|
| 880 |
"nvfp4_gemm_bias_gelu_nvfp4_fp16",
|
| 881 |
"nvfp4_gemm_bias_residual_fp16",
|
build/torch211-cxx11-cu130-x86_64-linux/{_fp4_gemm_cuda_e79264b.abi3.so → _fp4_gemm_cuda_014dcce.abi3.so}
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0d0f55ad6312f014e86e534fce4f0758b2f6f353a1db23c1e0fbe430d0f67921
|
| 3 |
+
size 8296848
|
build/torch211-cxx11-cu130-x86_64-linux/_ops.py
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
import torch
|
| 2 |
-
from . import
|
| 3 |
-
ops = torch.ops.
|
| 4 |
|
| 5 |
def add_op_namespace_prefix(op_name: str):
|
| 6 |
"""
|
| 7 |
Prefix op by namespace.
|
| 8 |
"""
|
| 9 |
-
return f"
|
|
|
|
| 1 |
import torch
|
| 2 |
+
from . import _fp4_gemm_cuda_014dcce
|
| 3 |
+
ops = torch.ops._fp4_gemm_cuda_014dcce
|
| 4 |
|
| 5 |
def add_op_namespace_prefix(op_name: str):
|
| 6 |
"""
|
| 7 |
Prefix op by namespace.
|
| 8 |
"""
|
| 9 |
+
return f"_fp4_gemm_cuda_014dcce::{op_name}"
|
build/torch211-cxx11-cu130-x86_64-linux/metadata.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
{
|
| 2 |
"name": "fp4-gemm",
|
| 3 |
-
"id": "
|
| 4 |
"version": 1,
|
| 5 |
"license": "Apache-2.0",
|
| 6 |
"python-depends": [],
|
|
@@ -14,10 +14,21 @@
|
|
| 14 |
"digest": {
|
| 15 |
"algorithm": "sha256",
|
| 16 |
"files": {
|
| 17 |
-
"__init__.py": "
|
| 18 |
-
"
|
| 19 |
-
"_ops.py": "
|
| 20 |
"fp4_gemm/__init__.py": "DFYPlrhXwYjEqCl/8n0SmWGZV8NFml5DPhMjKfv98GY="
|
| 21 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
}
|
| 23 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"name": "fp4-gemm",
|
| 3 |
+
"id": "_fp4_gemm_cuda_014dcce",
|
| 4 |
"version": 1,
|
| 5 |
"license": "Apache-2.0",
|
| 6 |
"python-depends": [],
|
|
|
|
| 14 |
"digest": {
|
| 15 |
"algorithm": "sha256",
|
| 16 |
"files": {
|
| 17 |
+
"__init__.py": "nTluIQWuLs7LotHRt43tTkaA+FufNMoFyUhMfJfbB+c=",
|
| 18 |
+
"_fp4_gemm_cuda_014dcce.abi3.so": "DQ9VrWMS8BToblNPzk8HWLL281Oh2yPB4PvkMND2eSE=",
|
| 19 |
+
"_ops.py": "3SSvweMyei7LivOTJNIPAdX+/MPOur59X7Qf5gA8m0Q=",
|
| 20 |
"fp4_gemm/__init__.py": "DFYPlrhXwYjEqCl/8n0SmWGZV8NFml5DPhMjKfv98GY="
|
| 21 |
}
|
| 22 |
+
},
|
| 23 |
+
"provenance": {
|
| 24 |
+
"kernel-builder": {
|
| 25 |
+
"version": "0.17.0-dev0",
|
| 26 |
+
"sha": "b39ca23f1b36383df00b27b3ffe1276cd5dbea85",
|
| 27 |
+
"dirty": false
|
| 28 |
+
},
|
| 29 |
+
"kernel": {
|
| 30 |
+
"sha": "014dcce673c8660d9e8932be3926f3b85340aac8",
|
| 31 |
+
"dirty": false
|
| 32 |
+
}
|
| 33 |
}
|
| 34 |
}
|