liangsu9988 commited on
Commit
b5b6101
·
verified ·
1 Parent(s): 61331bb

Uploaded using `kernel-builder`.

Browse files
build/torch213-cxx11-cu132-x86_64-linux/__init__.py CHANGED
@@ -7,6 +7,28 @@ 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")
@@ -714,6 +736,12 @@ 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",
 
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
+ "errors": "exceptions",
29
+ }
30
+
31
+
32
  def sfa_size_bytes(rows: int, dim: int) -> int:
33
  if rows <= 0 or dim <= 0 or dim % 16 != 0:
34
  raise ValueError("rows must be positive and dim must be positive/divisible by 16")
 
736
 
737
 
738
  __all__ = [
739
+ "NVFP4_BLOCK_SIZE",
740
+ "NVFP4_SCALE_TILE_ROWS",
741
+ "NVFP4_SCALE_TILE_BLOCKS",
742
+ "SUPPORTED_CUDA_CAPABILITIES",
743
+ "SUPPORTED_LAYOUTS",
744
+ "capabilities",
745
  "aligned_fp4_dim",
746
  "cutlass_fp4_gemm_geglu_il_hw_v10",
747
  "dequantize_fp4_sfa_fp16",
build/torch213-cxx11-cu132-x86_64-linux/{_fp4_gemm_cuda_8fba1d7.abi3.so → _fp4_gemm_cuda_8b8a2cd.abi3.so} RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:43c925ab9d40cf7693ae0d3b3c37f30d86911509f48c1765bf7be06191ad650e
3
- size 7887496
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ffad9d86c8ccabb083312b5fd81c11973bd0bde4fc5f89d5cb60cda9ab8779e
3
+ size 7891968
build/torch213-cxx11-cu132-x86_64-linux/_ops.py CHANGED
@@ -1,9 +1,9 @@
1
  import torch
2
- from . import _fp4_gemm_cuda_8fba1d7
3
- ops = torch.ops._fp4_gemm_cuda_8fba1d7
4
 
5
  def add_op_namespace_prefix(op_name: str):
6
  """
7
  Prefix op by namespace.
8
  """
9
- return f"_fp4_gemm_cuda_8fba1d7::{op_name}"
 
1
  import torch
2
+ from . import _fp4_gemm_cuda_8b8a2cd
3
+ ops = torch.ops._fp4_gemm_cuda_8b8a2cd
4
 
5
  def add_op_namespace_prefix(op_name: str):
6
  """
7
  Prefix op by namespace.
8
  """
9
+ return f"_fp4_gemm_cuda_8b8a2cd::{op_name}"
build/torch213-cxx11-cu132-x86_64-linux/metadata.json CHANGED
@@ -1,6 +1,6 @@
1
  {
2
  "name": "fp4-gemm",
3
- "id": "_fp4_gemm_cuda_8fba1d7",
4
  "version": 1,
5
  "license": "Apache-2.0",
6
  "python-depends": [],
@@ -14,9 +14,9 @@
14
  "digest": {
15
  "algorithm": "sha256",
16
  "files": {
17
- "__init__.py": "Mr1/bAUzs+af68D+bRXX6iONDof//dzKgn6QU5pKdu8=",
18
- "_fp4_gemm_cuda_8fba1d7.abi3.so": "Q8klq51Az3aTrg07PDfzDYaRFQn0jBdlv3vgYZGtZQ4=",
19
- "_ops.py": "I0EBGA7zB2SbeU365ZN8iiP8skVN4omShZNmKTtnfJU="
20
  }
21
  },
22
  "provenance": {
@@ -26,7 +26,7 @@
26
  "dirty": false
27
  },
28
  "kernel": {
29
- "sha": "8fba1d7f2e4e627b1a2c3c8018f01c0ca4f430ee",
30
  "dirty": false
31
  }
32
  }
 
1
  {
2
  "name": "fp4-gemm",
3
+ "id": "_fp4_gemm_cuda_8b8a2cd",
4
  "version": 1,
5
  "license": "Apache-2.0",
6
  "python-depends": [],
 
14
  "digest": {
15
  "algorithm": "sha256",
16
  "files": {
17
+ "__init__.py": "lJLCqFZi1vLuBzLeaNpAAfy/5/sHKv/E70REf7nu6ds=",
18
+ "_fp4_gemm_cuda_8b8a2cd.abi3.so": "f/rZ2GyMyrsIMxK1/YHBGXO9C95PxfidXLYM2pq4d54=",
19
+ "_ops.py": "/ey6vBILo+/4gY2quK0S+K1K6803F0rDBFL60eQ0VdE="
20
  }
21
  },
22
  "provenance": {
 
26
  "dirty": false
27
  },
28
  "kernel": {
29
+ "sha": "8b8a2cd59fce4f00274d194853953d1cd8eb16d8",
30
  "dirty": false
31
  }
32
  }