Update for build.toml changes
Browse files- build.toml +3 -6
- torch-ext/registration.h +0 -27
build.toml
CHANGED
@@ -1,17 +1,14 @@
|
|
1 |
[general]
|
2 |
-
|
3 |
|
4 |
[torch]
|
5 |
-
name = "quantization_eetq"
|
6 |
src = [
|
7 |
-
"torch-ext/registration.h",
|
8 |
"torch-ext/torch_binding.cpp",
|
9 |
"torch-ext/torch_binding.h"
|
10 |
]
|
11 |
-
pyroot = "torch-ext"
|
12 |
|
13 |
[kernel.cutlass_kernels]
|
14 |
-
capabilities = [ "7.0", "7.5", "8.0", "8.6", "8.7", "8.9", "9.0" ]
|
15 |
src = [
|
16 |
"cutlass_extensions/include/cutlass_extensions/arch/mma.h",
|
17 |
"cutlass_extensions/include/cutlass_extensions/compute_occupancy.h",
|
@@ -61,7 +58,7 @@ depends = [ "cutlass_2_10", "torch" ]
|
|
61 |
include = [ ".", "utils", "cutlass_extensions/include" ]
|
62 |
|
63 |
[kernel.weight_only_batched_gemv]
|
64 |
-
capabilities = [ "7.0", "7.5", "8.0", "8.6", "8.7", "8.9", "9.0" ]
|
65 |
src = [
|
66 |
"cutlass_extensions/include/cutlass_extensions/interleaved_numeric_conversion.h",
|
67 |
"cutlass_extensions/include/cutlass_extensions/gemm/kernel/mixed_gemm_B_layout.h",
|
|
|
1 |
[general]
|
2 |
+
name = "quantization_eetq"
|
3 |
|
4 |
[torch]
|
|
|
5 |
src = [
|
|
|
6 |
"torch-ext/torch_binding.cpp",
|
7 |
"torch-ext/torch_binding.h"
|
8 |
]
|
|
|
9 |
|
10 |
[kernel.cutlass_kernels]
|
11 |
+
cuda-capabilities = [ "7.0", "7.5", "8.0", "8.6", "8.7", "8.9", "9.0" ]
|
12 |
src = [
|
13 |
"cutlass_extensions/include/cutlass_extensions/arch/mma.h",
|
14 |
"cutlass_extensions/include/cutlass_extensions/compute_occupancy.h",
|
|
|
58 |
include = [ ".", "utils", "cutlass_extensions/include" ]
|
59 |
|
60 |
[kernel.weight_only_batched_gemv]
|
61 |
+
cuda-capabilities = [ "7.0", "7.5", "8.0", "8.6", "8.7", "8.9", "9.0" ]
|
62 |
src = [
|
63 |
"cutlass_extensions/include/cutlass_extensions/interleaved_numeric_conversion.h",
|
64 |
"cutlass_extensions/include/cutlass_extensions/gemm/kernel/mixed_gemm_B_layout.h",
|
torch-ext/registration.h
DELETED
@@ -1,27 +0,0 @@
|
|
1 |
-
#pragma once
|
2 |
-
|
3 |
-
#include <Python.h>
|
4 |
-
|
5 |
-
#define _CONCAT(A, B) A##B
|
6 |
-
#define CONCAT(A, B) _CONCAT(A, B)
|
7 |
-
|
8 |
-
#define _STRINGIFY(A) #A
|
9 |
-
#define STRINGIFY(A) _STRINGIFY(A)
|
10 |
-
|
11 |
-
// A version of the TORCH_LIBRARY macro that expands the NAME, i.e. so NAME
|
12 |
-
// could be a macro instead of a literal token.
|
13 |
-
#define TORCH_LIBRARY_EXPAND(NAME, MODULE) TORCH_LIBRARY(NAME, MODULE)
|
14 |
-
|
15 |
-
// A version of the TORCH_LIBRARY_IMPL macro that expands the NAME, i.e. so NAME
|
16 |
-
// could be a macro instead of a literal token.
|
17 |
-
#define TORCH_LIBRARY_IMPL_EXPAND(NAME, DEVICE, MODULE) \
|
18 |
-
TORCH_LIBRARY_IMPL(NAME, DEVICE, MODULE)
|
19 |
-
|
20 |
-
// REGISTER_EXTENSION allows the shared library to be loaded and initialized
|
21 |
-
// via python's import statement.
|
22 |
-
#define REGISTER_EXTENSION(NAME) \
|
23 |
-
PyMODINIT_FUNC CONCAT(PyInit_, NAME)() { \
|
24 |
-
static struct PyModuleDef module = {PyModuleDef_HEAD_INIT, \
|
25 |
-
STRINGIFY(NAME), nullptr, 0, nullptr}; \
|
26 |
-
return PyModule_Create(&module); \
|
27 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|