Hugging Face's logo

kernels-community
/
quantization

kernel
Model card Files Community
1
danieldk HF staff commited on
Commit
116e562
·
1 Parent(s): 64f4463

Update for build.toml changes

Browse files
Files changed (16) hide show
  1. build.toml +10 -12
  2. flake.lock +0 -95
  3. {ext-torch → torch-ext}/quantization/__init__.py +0 -0
  4. {ext-torch → torch-ext}/quantization/compressed_tensors.py +0 -0
  5. {ext-torch → torch-ext}/quantization/cutlass.py +0 -0
  6. {ext-torch → torch-ext}/quantization/marlin.py +0 -0
  7. {ext-torch → torch-ext}/quantization/scalar_type.py +0 -0
  8. {ext-torch → torch-ext}/quantization/utils/__init__.py +0 -0
  9. {ext-torch → torch-ext}/quantization/utils/marlin_utils.py +0 -0
  10. {ext-torch → torch-ext}/quantization/utils/marlin_utils_fp8.py +0 -0
  11. {ext-torch → torch-ext}/quantization/utils/marlin_utils_test.py +0 -0
  12. {ext-torch → torch-ext}/quantization/utils/marlin_utils_test_24.py +0 -0
  13. {ext-torch → torch-ext}/quantization/utils/marlin_utils_test_qqq.py +0 -0
  14. {ext-torch → torch-ext}/quantization/utils/quant_utils.py +0 -0
  15. {ext-torch → torch-ext}/torch_binding.cpp +0 -0
  16. {ext-torch → torch-ext}/torch_binding.h +0 -0
build.toml CHANGED
@@ -1,19 +1,17 @@
1
  [general]
2
- version = "0.0.1"
3
 
4
  [torch]
5
- name = "quantization"
6
  src = [
7
  "core/registration.h",
8
  "core/scalar_type.hpp",
9
- "ext-torch/torch_binding.cpp",
10
- "ext-torch/torch_binding.h"
11
  ]
12
  include = [ "." ]
13
- pyroot = "ext-torch"
14
 
15
  [kernel.cutlass_w8a8]
16
- capabilities = [ "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "9.0a" ]
17
  src = [
18
  "core/math.hpp",
19
  "cutlass_w8a8/common.hpp",
@@ -31,7 +29,7 @@ include = [ "." ]
31
  depends = [ "cutlass_3_6", "torch" ]
32
 
33
  [kernel.cutlass_w8a8_hopper]
34
- capabilities = [ "9.0", "9.0a" ]
35
  src = [
36
  "core/math.hpp",
37
  "cutlass_w8a8/common.hpp",
@@ -48,7 +46,7 @@ include = [ "." ]
48
  depends = [ "cutlass_3_6", "torch" ]
49
 
50
  [kernel.fp8_common]
51
- capabilities = [ "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "9.0a" ]
52
  src = [
53
  "fp8/common.cu",
54
  "fp8/common.cuh",
@@ -59,7 +57,7 @@ include = [ "." ]
59
  depends = [ "torch" ]
60
 
61
  [kernel.fp8_marlin]
62
- capabilities = [ "8.0", "8.6", "8.7", "8.9", "9.0", "9.0a" ]
63
  src = [
64
  "fp8/fp8_marlin.cu",
65
  "gptq_marlin/marlin.cuh",
@@ -68,7 +66,7 @@ src = [
68
  depends = [ "torch" ]
69
 
70
  [kernel.int8_common]
71
- capabilities = [ "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "9.0a" ]
72
  src = [
73
  "compressed_tensors/int8_quant_kernels.cu",
74
  "dispatch_utils.h"
@@ -77,7 +75,7 @@ include = [ "." ]
77
  depends = [ "torch" ]
78
 
79
  [kernel.gptq_marlin]
80
- capabilities = [ "8.0", "8.6", "8.7", "8.9", "9.0", "9.0a" ]
81
  src = [
82
  "core/scalar_type.hpp",
83
  "gptq_marlin/awq_marlin_repack.cu",
@@ -90,7 +88,7 @@ include = [ "." ]
90
  depends = [ "torch" ]
91
 
92
  [kernel.marlin]
93
- capabilities = [ "8.0", "8.6", "8.7", "8.9", "9.0", "9.0a" ]
94
  src = [
95
  "core/scalar_type.hpp",
96
  "marlin/dense/common/base.h",
 
1
  [general]
2
+ name = "quantization"
3
 
4
  [torch]
 
5
  src = [
6
  "core/registration.h",
7
  "core/scalar_type.hpp",
8
+ "torch-ext/torch_binding.cpp",
9
+ "torch-ext/torch_binding.h"
10
  ]
11
  include = [ "." ]
 
12
 
13
  [kernel.cutlass_w8a8]
14
+ cuda-capabilities = [ "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "9.0a" ]
15
  src = [
16
  "core/math.hpp",
17
  "cutlass_w8a8/common.hpp",
 
29
  depends = [ "cutlass_3_6", "torch" ]
30
 
31
  [kernel.cutlass_w8a8_hopper]
32
+ cuda-capabilities = [ "9.0", "9.0a" ]
33
  src = [
34
  "core/math.hpp",
35
  "cutlass_w8a8/common.hpp",
 
46
  depends = [ "cutlass_3_6", "torch" ]
47
 
48
  [kernel.fp8_common]
49
+ cuda-capabilities = [ "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "9.0a" ]
50
  src = [
51
  "fp8/common.cu",
52
  "fp8/common.cuh",
 
57
  depends = [ "torch" ]
58
 
59
  [kernel.fp8_marlin]
60
+ cuda-capabilities = [ "8.0", "8.6", "8.7", "8.9", "9.0", "9.0a" ]
61
  src = [
62
  "fp8/fp8_marlin.cu",
63
  "gptq_marlin/marlin.cuh",
 
66
  depends = [ "torch" ]
67
 
68
  [kernel.int8_common]
69
+ cuda-capabilities = [ "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "9.0a" ]
70
  src = [
71
  "compressed_tensors/int8_quant_kernels.cu",
72
  "dispatch_utils.h"
 
75
  depends = [ "torch" ]
76
 
77
  [kernel.gptq_marlin]
78
+ cuda-capabilities = [ "8.0", "8.6", "8.7", "8.9", "9.0", "9.0a" ]
79
  src = [
80
  "core/scalar_type.hpp",
81
  "gptq_marlin/awq_marlin_repack.cu",
 
88
  depends = [ "torch" ]
89
 
90
  [kernel.marlin]
91
+ cuda-capabilities = [ "8.0", "8.6", "8.7", "8.9", "9.0", "9.0a" ]
92
  src = [
93
  "core/scalar_type.hpp",
94
  "marlin/dense/common/base.h",
flake.lock DELETED
@@ -1,95 +0,0 @@
1
- {
2
- "nodes": {
3
- "flake-compat": {
4
- "locked": {
5
- "lastModified": 1732722421,
6
- "narHash": "sha256-HRJ/18p+WoXpWJkcdsk9St5ZiukCqSDgbOGFa8Okehg=",
7
- "owner": "edolstra",
8
- "repo": "flake-compat",
9
- "rev": "9ed2ac151eada2306ca8c418ebd97807bb08f6ac",
10
- "type": "github"
11
- },
12
- "original": {
13
- "owner": "edolstra",
14
- "repo": "flake-compat",
15
- "type": "github"
16
- }
17
- },
18
- "flake-utils": {
19
- "inputs": {
20
- "systems": "systems"
21
- },
22
- "locked": {
23
- "lastModified": 1731533236,
24
- "narHash": "sha256-l0KFg5HjrsfsO/JpG+r7fRrqm12kzFHyUHqHCVpMMbI=",
25
- "owner": "numtide",
26
- "repo": "flake-utils",
27
- "rev": "11707dc2f618dd54ca8739b309ec4fc024de578b",
28
- "type": "github"
29
- },
30
- "original": {
31
- "owner": "numtide",
32
- "repo": "flake-utils",
33
- "type": "github"
34
- }
35
- },
36
- "kernel-builder": {
37
- "inputs": {
38
- "flake-compat": "flake-compat",
39
- "flake-utils": "flake-utils",
40
- "nixpkgs": "nixpkgs"
41
- },
42
- "locked": {
43
- "lastModified": 1737048117,
44
- "narHash": "sha256-WEsPsP3RU2XdUWSm5i23gonDxitA/IwxsroMX0OowTw=",
45
- "ref": "refs/heads/main",
46
- "rev": "9767f35a39472900a0957b2b028da9068431e53f",
47
- "revCount": 71,
48
- "type": "git",
49
- "url": "ssh://[email protected]/huggingface/kernel-builder"
50
- },
51
- "original": {
52
- "type": "git",
53
- "url": "ssh://[email protected]/huggingface/kernel-builder"
54
- }
55
- },
56
- "nixpkgs": {
57
- "locked": {
58
- "lastModified": 1732866476,
59
- "narHash": "sha256-fQPhGk0rBifqOtcq07Que6MWuWU08AmiaOVZ6TTAHUw=",
60
- "owner": "nixos",
61
- "repo": "nixpkgs",
62
- "rev": "564e219aa5ca9306ab22fb803c16391aacb4417b",
63
- "type": "github"
64
- },
65
- "original": {
66
- "owner": "nixos",
67
- "ref": "nixos-unstable-small",
68
- "repo": "nixpkgs",
69
- "type": "github"
70
- }
71
- },
72
- "root": {
73
- "inputs": {
74
- "kernel-builder": "kernel-builder"
75
- }
76
- },
77
- "systems": {
78
- "locked": {
79
- "lastModified": 1681028828,
80
- "narHash": "sha256-Vy1rq5AaRuLzOxct8nz4T6wlgyUR7zLU309k9mBC768=",
81
- "owner": "nix-systems",
82
- "repo": "default",
83
- "rev": "da67096a3b9bf56a91d16901293e51ba5b49a27e",
84
- "type": "github"
85
- },
86
- "original": {
87
- "owner": "nix-systems",
88
- "repo": "default",
89
- "type": "github"
90
- }
91
- }
92
- },
93
- "root": "root",
94
- "version": 7
95
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
{ext-torch → torch-ext}/quantization/__init__.py RENAMED
File without changes
{ext-torch → torch-ext}/quantization/compressed_tensors.py RENAMED
File without changes
{ext-torch → torch-ext}/quantization/cutlass.py RENAMED
File without changes
{ext-torch → torch-ext}/quantization/marlin.py RENAMED
File without changes
{ext-torch → torch-ext}/quantization/scalar_type.py RENAMED
File without changes
{ext-torch → torch-ext}/quantization/utils/__init__.py RENAMED
File without changes
{ext-torch → torch-ext}/quantization/utils/marlin_utils.py RENAMED
File without changes
{ext-torch → torch-ext}/quantization/utils/marlin_utils_fp8.py RENAMED
File without changes
{ext-torch → torch-ext}/quantization/utils/marlin_utils_test.py RENAMED
File without changes
{ext-torch → torch-ext}/quantization/utils/marlin_utils_test_24.py RENAMED
File without changes
{ext-torch → torch-ext}/quantization/utils/marlin_utils_test_qqq.py RENAMED
File without changes
{ext-torch → torch-ext}/quantization/utils/quant_utils.py RENAMED
File without changes
{ext-torch → torch-ext}/torch_binding.cpp RENAMED
File without changes
{ext-torch → torch-ext}/torch_binding.h RENAMED
File without changes