Hugging Face's logo

Kernels:
kernels-community
/
quantization-eetq

Kernel card Files
xet
 Community
drbh commited on
Commit
aeb4e54
·
unverified ·
0 Parent(s):

Migrated from kernels-community/quantization-eetq

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +36 -0
  2. README.md +14 -0
  3. build/torch210-cxx11-cu126-aarch64-linux/__init__.py +3 -0
  4. build/torch210-cxx11-cu126-aarch64-linux/_ops.py +9 -0
  5. build/torch210-cxx11-cu126-aarch64-linux/_quantization_eetq_cuda_4bc0051.abi3.so +3 -0
  6. build/torch210-cxx11-cu126-aarch64-linux/custom_ops.py +36 -0
  7. build/torch210-cxx11-cu126-aarch64-linux/metadata.json +18 -0
  8. build/torch210-cxx11-cu126-aarch64-linux/quantization_eetq/__init__.py +26 -0
  9. build/torch210-cxx11-cu126-x86_64-linux/__init__.py +3 -0
  10. build/torch210-cxx11-cu126-x86_64-linux/_ops.py +9 -0
  11. build/torch210-cxx11-cu126-x86_64-linux/_quantization_eetq_cuda_4bc0051.abi3.so +3 -0
  12. build/torch210-cxx11-cu126-x86_64-linux/custom_ops.py +36 -0
  13. build/torch210-cxx11-cu126-x86_64-linux/metadata.json +18 -0
  14. build/torch210-cxx11-cu126-x86_64-linux/quantization_eetq/__init__.py +26 -0
  15. build/torch210-cxx11-cu128-aarch64-linux/__init__.py +3 -0
  16. build/torch210-cxx11-cu128-aarch64-linux/_ops.py +9 -0
  17. build/torch210-cxx11-cu128-aarch64-linux/_quantization_eetq_cuda_4bc0051.abi3.so +3 -0
  18. build/torch210-cxx11-cu128-aarch64-linux/custom_ops.py +36 -0
  19. build/torch210-cxx11-cu128-aarch64-linux/metadata.json +21 -0
  20. build/torch210-cxx11-cu128-aarch64-linux/quantization_eetq/__init__.py +26 -0
  21. build/torch210-cxx11-cu128-x86_64-linux/__init__.py +3 -0
  22. build/torch210-cxx11-cu128-x86_64-linux/_ops.py +9 -0
  23. build/torch210-cxx11-cu128-x86_64-linux/_quantization_eetq_cuda_4bc0051.abi3.so +3 -0
  24. build/torch210-cxx11-cu128-x86_64-linux/custom_ops.py +36 -0
  25. build/torch210-cxx11-cu128-x86_64-linux/metadata.json +21 -0
  26. build/torch210-cxx11-cu128-x86_64-linux/quantization_eetq/__init__.py +26 -0
  27. build/torch211-cxx11-cu126-aarch64-linux/__init__.py +3 -0
  28. build/torch211-cxx11-cu126-aarch64-linux/_ops.py +9 -0
  29. build/torch211-cxx11-cu126-aarch64-linux/_quantization_eetq_cuda_4bc0051.abi3.so +3 -0
  30. build/torch211-cxx11-cu126-aarch64-linux/custom_ops.py +36 -0
  31. build/torch211-cxx11-cu126-aarch64-linux/metadata.json +18 -0
  32. build/torch211-cxx11-cu126-aarch64-linux/quantization_eetq/__init__.py +26 -0
  33. build/torch211-cxx11-cu126-x86_64-linux/__init__.py +3 -0
  34. build/torch211-cxx11-cu126-x86_64-linux/_ops.py +9 -0
  35. build/torch211-cxx11-cu126-x86_64-linux/_quantization_eetq_cuda_4bc0051.abi3.so +3 -0
  36. build/torch211-cxx11-cu126-x86_64-linux/custom_ops.py +36 -0
  37. build/torch211-cxx11-cu126-x86_64-linux/metadata.json +18 -0
  38. build/torch211-cxx11-cu126-x86_64-linux/quantization_eetq/__init__.py +26 -0
  39. build/torch211-cxx11-cu128-aarch64-linux/__init__.py +3 -0
  40. build/torch211-cxx11-cu128-aarch64-linux/_ops.py +9 -0
  41. build/torch211-cxx11-cu128-aarch64-linux/_quantization_eetq_cuda_4bc0051.abi3.so +3 -0
  42. build/torch211-cxx11-cu128-aarch64-linux/custom_ops.py +36 -0
  43. build/torch211-cxx11-cu128-aarch64-linux/metadata.json +21 -0
  44. build/torch211-cxx11-cu128-aarch64-linux/quantization_eetq/__init__.py +26 -0
  45. build/torch211-cxx11-cu128-x86_64-linux/__init__.py +3 -0
  46. build/torch211-cxx11-cu128-x86_64-linux/_ops.py +9 -0
  47. build/torch211-cxx11-cu128-x86_64-linux/_quantization_eetq_cuda_4bc0051.abi3.so +3 -0
  48. build/torch211-cxx11-cu128-x86_64-linux/custom_ops.py +36 -0
  49. build/torch211-cxx11-cu128-x86_64-linux/metadata.json +21 -0
  50. build/torch211-cxx11-cu128-x86_64-linux/quantization_eetq/__init__.py +26 -0
.gitattributes ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
5
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
6
+ *.ftz filter=lfs diff=lfs merge=lfs -text
7
+ *.gz filter=lfs diff=lfs merge=lfs -text
8
+ *.h5 filter=lfs diff=lfs merge=lfs -text
9
+ *.joblib filter=lfs diff=lfs merge=lfs -text
10
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
12
+ *.model filter=lfs diff=lfs merge=lfs -text
13
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
14
+ *.npy filter=lfs diff=lfs merge=lfs -text
15
+ *.npz filter=lfs diff=lfs merge=lfs -text
16
+ *.onnx filter=lfs diff=lfs merge=lfs -text
17
+ *.ot filter=lfs diff=lfs merge=lfs -text
18
+ *.parquet filter=lfs diff=lfs merge=lfs -text
19
+ *.pb filter=lfs diff=lfs merge=lfs -text
20
+ *.pickle filter=lfs diff=lfs merge=lfs -text
21
+ *.pkl filter=lfs diff=lfs merge=lfs -text
22
+ *.pt filter=lfs diff=lfs merge=lfs -text
23
+ *.pth filter=lfs diff=lfs merge=lfs -text
24
+ *.rar filter=lfs diff=lfs merge=lfs -text
25
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
26
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
27
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
28
+ *.tar filter=lfs diff=lfs merge=lfs -text
29
+ *.tflite filter=lfs diff=lfs merge=lfs -text
30
+ *.tgz filter=lfs diff=lfs merge=lfs -text
31
+ *.wasm filter=lfs diff=lfs merge=lfs -text
32
+ *.xz filter=lfs diff=lfs merge=lfs -text
33
+ *.zip filter=lfs diff=lfs merge=lfs -text
34
+ *.zst filter=lfs diff=lfs merge=lfs -text
35
+ *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ *.so filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ tags:
4
+ - kernels
5
+ ---
6
+
7
+ ![Status](https://hubwebhook.dholtz.com/shield?repo=kernels-community/quantization-eetq)
8
+
9
+ ## eetq
10
+
11
+ EETQ kernels from [NetEase-FuXi/EETQ](https://github.com/NetEase-FuXi/EETQ).
12
+
13
+ Kernel source: https://github.com/huggingface/kernels-community/tree/main/quantization-eetq
14
+
build/torch210-cxx11-cu126-aarch64-linux/__init__.py ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ from .custom_ops import w8_a16_gemm, w8_a16_gemm_, preprocess_weights, quant_weights
2
+
3
+ __all__ = ["w8_a16_gemm", "w8_a16_gemm_", "preprocess_weights", "quant_weights"]
build/torch210-cxx11-cu126-aarch64-linux/_ops.py ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ import torch
2
+ from . import _quantization_eetq_cuda_4bc0051
3
+ ops = torch.ops._quantization_eetq_cuda_4bc0051
4
+
5
+ def add_op_namespace_prefix(op_name: str):
6
+ """
7
+ Prefix op by namespace.
8
+ """
9
+ return f"_quantization_eetq_cuda_4bc0051::{op_name}"
build/torch210-cxx11-cu126-aarch64-linux/_quantization_eetq_cuda_4bc0051.abi3.so ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d7c4c1ddd8d31cee535f23012aa490749a38b28dc3592473eea939e3611b3b1
3
+ size 39010048
build/torch210-cxx11-cu126-aarch64-linux/custom_ops.py ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from typing import List
2
+ import torch
3
+
4
+ from ._ops import ops
5
+
6
+
7
+ def w8_a16_gemm(
8
+ input: torch.Tensor, weight: torch.Tensor, scale: torch.Tensor
9
+ ) -> torch.Tensor:
10
+ return ops.w8_a16_gemm(input, weight, scale)
11
+
12
+
13
+ def w8_a16_gemm_(
14
+ input: torch.Tensor,
15
+ weight: torch.Tensor,
16
+ scale: torch.Tensor,
17
+ output: torch.Tensor,
18
+ m: int,
19
+ n: int,
20
+ k: int,
21
+ ) -> torch.Tensor:
22
+ return ops.w8_a16_gemm_(input, weight, scale, output, m, n, k)
23
+
24
+
25
+ def preprocess_weights(origin_weight: torch.Tensor, is_int4: bool) -> torch.Tensor:
26
+ return ops.preprocess_weights(origin_weight, is_int4)
27
+
28
+
29
+ def quant_weights(
30
+ origin_weight: torch.Tensor,
31
+ quant_type: torch.dtype,
32
+ return_unprocessed_quantized_tensor: bool,
33
+ ) -> List[torch.Tensor]:
34
+ return ops.quant_weights(
35
+ origin_weight, quant_type, return_unprocessed_quantized_tensor
36
+ )
build/torch210-cxx11-cu126-aarch64-linux/metadata.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "version": 1,
3
+ "license": "Apache-2.0",
4
+ "python-depends": [],
5
+ "backend": {
6
+ "type": "cuda",
7
+ "archs": [
8
+ "7.0",
9
+ "7.2",
10
+ "7.5",
11
+ "8.0",
12
+ "8.6",
13
+ "8.7",
14
+ "8.9",
15
+ "9.0+PTX"
16
+ ]
17
+ }
18
+ }
build/torch210-cxx11-cu126-aarch64-linux/quantization_eetq/__init__.py ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import ctypes
2
+ import importlib.util
3
+ import sys
4
+ from pathlib import Path
5
+ from types import ModuleType
6
+
7
+
8
+ def _import_from_path(file_path: Path) -> ModuleType:
9
+ # We cannot use the module name as-is, after adding it to `sys.modules`,
10
+ # it would also be used for other imports. So, we make a module name that
11
+ # depends on the path for it to be unique using the hex-encoded hash of
12
+ # the path.
13
+ path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
14
+ module_name = path_hash
15
+ spec = importlib.util.spec_from_file_location(module_name, file_path)
16
+ if spec is None:
17
+ raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
18
+ module = importlib.util.module_from_spec(spec)
19
+ if module is None:
20
+ raise ImportError(f"Cannot load module {module_name} from spec")
21
+ sys.modules[module_name] = module
22
+ spec.loader.exec_module(module) # type: ignore
23
+ return module
24
+
25
+
26
+ globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))
build/torch210-cxx11-cu126-x86_64-linux/__init__.py ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ from .custom_ops import w8_a16_gemm, w8_a16_gemm_, preprocess_weights, quant_weights
2
+
3
+ __all__ = ["w8_a16_gemm", "w8_a16_gemm_", "preprocess_weights", "quant_weights"]
build/torch210-cxx11-cu126-x86_64-linux/_ops.py ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ import torch
2
+ from . import _quantization_eetq_cuda_4bc0051
3
+ ops = torch.ops._quantization_eetq_cuda_4bc0051
4
+
5
+ def add_op_namespace_prefix(op_name: str):
6
+ """
7
+ Prefix op by namespace.
8
+ """
9
+ return f"_quantization_eetq_cuda_4bc0051::{op_name}"
build/torch210-cxx11-cu126-x86_64-linux/_quantization_eetq_cuda_4bc0051.abi3.so ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d60d2faac38ccbfeca46a6f9b025cdce2590bb5291da8ae42f0e2235a29ec066
3
+ size 39060416
build/torch210-cxx11-cu126-x86_64-linux/custom_ops.py ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from typing import List
2
+ import torch
3
+
4
+ from ._ops import ops
5
+
6
+
7
+ def w8_a16_gemm(
8
+ input: torch.Tensor, weight: torch.Tensor, scale: torch.Tensor
9
+ ) -> torch.Tensor:
10
+ return ops.w8_a16_gemm(input, weight, scale)
11
+
12
+
13
+ def w8_a16_gemm_(
14
+ input: torch.Tensor,
15
+ weight: torch.Tensor,
16
+ scale: torch.Tensor,
17
+ output: torch.Tensor,
18
+ m: int,
19
+ n: int,
20
+ k: int,
21
+ ) -> torch.Tensor:
22
+ return ops.w8_a16_gemm_(input, weight, scale, output, m, n, k)
23
+
24
+
25
+ def preprocess_weights(origin_weight: torch.Tensor, is_int4: bool) -> torch.Tensor:
26
+ return ops.preprocess_weights(origin_weight, is_int4)
27
+
28
+
29
+ def quant_weights(
30
+ origin_weight: torch.Tensor,
31
+ quant_type: torch.dtype,
32
+ return_unprocessed_quantized_tensor: bool,
33
+ ) -> List[torch.Tensor]:
34
+ return ops.quant_weights(
35
+ origin_weight, quant_type, return_unprocessed_quantized_tensor
36
+ )
build/torch210-cxx11-cu126-x86_64-linux/metadata.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "version": 1,
3
+ "license": "Apache-2.0",
4
+ "python-depends": [],
5
+ "backend": {
6
+ "type": "cuda",
7
+ "archs": [
8
+ "7.0",
9
+ "7.2",
10
+ "7.5",
11
+ "8.0",
12
+ "8.6",
13
+ "8.7",
14
+ "8.9",
15
+ "9.0+PTX"
16
+ ]
17
+ }
18
+ }
build/torch210-cxx11-cu126-x86_64-linux/quantization_eetq/__init__.py ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import ctypes
2
+ import importlib.util
3
+ import sys
4
+ from pathlib import Path
5
+ from types import ModuleType
6
+
7
+
8
+ def _import_from_path(file_path: Path) -> ModuleType:
9
+ # We cannot use the module name as-is, after adding it to `sys.modules`,
10
+ # it would also be used for other imports. So, we make a module name that
11
+ # depends on the path for it to be unique using the hex-encoded hash of
12
+ # the path.
13
+ path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
14
+ module_name = path_hash
15
+ spec = importlib.util.spec_from_file_location(module_name, file_path)
16
+ if spec is None:
17
+ raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
18
+ module = importlib.util.module_from_spec(spec)
19
+ if module is None:
20
+ raise ImportError(f"Cannot load module {module_name} from spec")
21
+ sys.modules[module_name] = module
22
+ spec.loader.exec_module(module) # type: ignore
23
+ return module
24
+
25
+
26
+ globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))
build/torch210-cxx11-cu128-aarch64-linux/__init__.py ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ from .custom_ops import w8_a16_gemm, w8_a16_gemm_, preprocess_weights, quant_weights
2
+
3
+ __all__ = ["w8_a16_gemm", "w8_a16_gemm_", "preprocess_weights", "quant_weights"]
build/torch210-cxx11-cu128-aarch64-linux/_ops.py ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ import torch
2
+ from . import _quantization_eetq_cuda_4bc0051
3
+ ops = torch.ops._quantization_eetq_cuda_4bc0051
4
+
5
+ def add_op_namespace_prefix(op_name: str):
6
+ """
7
+ Prefix op by namespace.
8
+ """
9
+ return f"_quantization_eetq_cuda_4bc0051::{op_name}"
build/torch210-cxx11-cu128-aarch64-linux/_quantization_eetq_cuda_4bc0051.abi3.so ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0aa8d32b1d6f07f1b095c7228dbe14325aeeba864aefcdc614654261babce162
3
+ size 45366048
build/torch210-cxx11-cu128-aarch64-linux/custom_ops.py ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from typing import List
2
+ import torch
3
+
4
+ from ._ops import ops
5
+
6
+
7
+ def w8_a16_gemm(
8
+ input: torch.Tensor, weight: torch.Tensor, scale: torch.Tensor
9
+ ) -> torch.Tensor:
10
+ return ops.w8_a16_gemm(input, weight, scale)
11
+
12
+
13
+ def w8_a16_gemm_(
14
+ input: torch.Tensor,
15
+ weight: torch.Tensor,
16
+ scale: torch.Tensor,
17
+ output: torch.Tensor,
18
+ m: int,
19
+ n: int,
20
+ k: int,
21
+ ) -> torch.Tensor:
22
+ return ops.w8_a16_gemm_(input, weight, scale, output, m, n, k)
23
+
24
+
25
+ def preprocess_weights(origin_weight: torch.Tensor, is_int4: bool) -> torch.Tensor:
26
+ return ops.preprocess_weights(origin_weight, is_int4)
27
+
28
+
29
+ def quant_weights(
30
+ origin_weight: torch.Tensor,
31
+ quant_type: torch.dtype,
32
+ return_unprocessed_quantized_tensor: bool,
33
+ ) -> List[torch.Tensor]:
34
+ return ops.quant_weights(
35
+ origin_weight, quant_type, return_unprocessed_quantized_tensor
36
+ )
build/torch210-cxx11-cu128-aarch64-linux/metadata.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "version": 1,
3
+ "license": "Apache-2.0",
4
+ "python-depends": [],
5
+ "backend": {
6
+ "type": "cuda",
7
+ "archs": [
8
+ "10.0",
9
+ "10.1",
10
+ "12.0+PTX",
11
+ "7.0",
12
+ "7.2",
13
+ "7.5",
14
+ "8.0",
15
+ "8.6",
16
+ "8.7",
17
+ "8.9",
18
+ "9.0"
19
+ ]
20
+ }
21
+ }
build/torch210-cxx11-cu128-aarch64-linux/quantization_eetq/__init__.py ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import ctypes
2
+ import importlib.util
3
+ import sys
4
+ from pathlib import Path
5
+ from types import ModuleType
6
+
7
+
8
+ def _import_from_path(file_path: Path) -> ModuleType:
9
+ # We cannot use the module name as-is, after adding it to `sys.modules`,
10
+ # it would also be used for other imports. So, we make a module name that
11
+ # depends on the path for it to be unique using the hex-encoded hash of
12
+ # the path.
13
+ path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
14
+ module_name = path_hash
15
+ spec = importlib.util.spec_from_file_location(module_name, file_path)
16
+ if spec is None:
17
+ raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
18
+ module = importlib.util.module_from_spec(spec)
19
+ if module is None:
20
+ raise ImportError(f"Cannot load module {module_name} from spec")
21
+ sys.modules[module_name] = module
22
+ spec.loader.exec_module(module) # type: ignore
23
+ return module
24
+
25
+
26
+ globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))
build/torch210-cxx11-cu128-x86_64-linux/__init__.py ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ from .custom_ops import w8_a16_gemm, w8_a16_gemm_, preprocess_weights, quant_weights
2
+
3
+ __all__ = ["w8_a16_gemm", "w8_a16_gemm_", "preprocess_weights", "quant_weights"]
build/torch210-cxx11-cu128-x86_64-linux/_ops.py ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ import torch
2
+ from . import _quantization_eetq_cuda_4bc0051
3
+ ops = torch.ops._quantization_eetq_cuda_4bc0051
4
+
5
+ def add_op_namespace_prefix(op_name: str):
6
+ """
7
+ Prefix op by namespace.
8
+ """
9
+ return f"_quantization_eetq_cuda_4bc0051::{op_name}"
build/torch210-cxx11-cu128-x86_64-linux/_quantization_eetq_cuda_4bc0051.abi3.so ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4735c25762548fcf265184a9c06da1bddc95f663f183b15e53b09f98c35ad321
3
+ size 45394000
build/torch210-cxx11-cu128-x86_64-linux/custom_ops.py ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from typing import List
2
+ import torch
3
+
4
+ from ._ops import ops
5
+
6
+
7
+ def w8_a16_gemm(
8
+ input: torch.Tensor, weight: torch.Tensor, scale: torch.Tensor
9
+ ) -> torch.Tensor:
10
+ return ops.w8_a16_gemm(input, weight, scale)
11
+
12
+
13
+ def w8_a16_gemm_(
14
+ input: torch.Tensor,
15
+ weight: torch.Tensor,
16
+ scale: torch.Tensor,
17
+ output: torch.Tensor,
18
+ m: int,
19
+ n: int,
20
+ k: int,
21
+ ) -> torch.Tensor:
22
+ return ops.w8_a16_gemm_(input, weight, scale, output, m, n, k)
23
+
24
+
25
+ def preprocess_weights(origin_weight: torch.Tensor, is_int4: bool) -> torch.Tensor:
26
+ return ops.preprocess_weights(origin_weight, is_int4)
27
+
28
+
29
+ def quant_weights(
30
+ origin_weight: torch.Tensor,
31
+ quant_type: torch.dtype,
32
+ return_unprocessed_quantized_tensor: bool,
33
+ ) -> List[torch.Tensor]:
34
+ return ops.quant_weights(
35
+ origin_weight, quant_type, return_unprocessed_quantized_tensor
36
+ )
build/torch210-cxx11-cu128-x86_64-linux/metadata.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "version": 1,
3
+ "license": "Apache-2.0",
4
+ "python-depends": [],
5
+ "backend": {
6
+ "type": "cuda",
7
+ "archs": [
8
+ "10.0",
9
+ "10.1",
10
+ "12.0+PTX",
11
+ "7.0",
12
+ "7.2",
13
+ "7.5",
14
+ "8.0",
15
+ "8.6",
16
+ "8.7",
17
+ "8.9",
18
+ "9.0"
19
+ ]
20
+ }
21
+ }
build/torch210-cxx11-cu128-x86_64-linux/quantization_eetq/__init__.py ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import ctypes
2
+ import importlib.util
3
+ import sys
4
+ from pathlib import Path
5
+ from types import ModuleType
6
+
7
+
8
+ def _import_from_path(file_path: Path) -> ModuleType:
9
+ # We cannot use the module name as-is, after adding it to `sys.modules`,
10
+ # it would also be used for other imports. So, we make a module name that
11
+ # depends on the path for it to be unique using the hex-encoded hash of
12
+ # the path.
13
+ path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
14
+ module_name = path_hash
15
+ spec = importlib.util.spec_from_file_location(module_name, file_path)
16
+ if spec is None:
17
+ raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
18
+ module = importlib.util.module_from_spec(spec)
19
+ if module is None:
20
+ raise ImportError(f"Cannot load module {module_name} from spec")
21
+ sys.modules[module_name] = module
22
+ spec.loader.exec_module(module) # type: ignore
23
+ return module
24
+
25
+
26
+ globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))
build/torch211-cxx11-cu126-aarch64-linux/__init__.py ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ from .custom_ops import w8_a16_gemm, w8_a16_gemm_, preprocess_weights, quant_weights
2
+
3
+ __all__ = ["w8_a16_gemm", "w8_a16_gemm_", "preprocess_weights", "quant_weights"]
build/torch211-cxx11-cu126-aarch64-linux/_ops.py ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ import torch
2
+ from . import _quantization_eetq_cuda_4bc0051
3
+ ops = torch.ops._quantization_eetq_cuda_4bc0051
4
+
5
+ def add_op_namespace_prefix(op_name: str):
6
+ """
7
+ Prefix op by namespace.
8
+ """
9
+ return f"_quantization_eetq_cuda_4bc0051::{op_name}"
build/torch211-cxx11-cu126-aarch64-linux/_quantization_eetq_cuda_4bc0051.abi3.so ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b1d6d36599980e86a95f70c00d05c9aacef4a4d24188a84c3b258a2a06133fd
3
+ size 39006256
build/torch211-cxx11-cu126-aarch64-linux/custom_ops.py ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from typing import List
2
+ import torch
3
+
4
+ from ._ops import ops
5
+
6
+
7
+ def w8_a16_gemm(
8
+ input: torch.Tensor, weight: torch.Tensor, scale: torch.Tensor
9
+ ) -> torch.Tensor:
10
+ return ops.w8_a16_gemm(input, weight, scale)
11
+
12
+
13
+ def w8_a16_gemm_(
14
+ input: torch.Tensor,
15
+ weight: torch.Tensor,
16
+ scale: torch.Tensor,
17
+ output: torch.Tensor,
18
+ m: int,
19
+ n: int,
20
+ k: int,
21
+ ) -> torch.Tensor:
22
+ return ops.w8_a16_gemm_(input, weight, scale, output, m, n, k)
23
+
24
+
25
+ def preprocess_weights(origin_weight: torch.Tensor, is_int4: bool) -> torch.Tensor:
26
+ return ops.preprocess_weights(origin_weight, is_int4)
27
+
28
+
29
+ def quant_weights(
30
+ origin_weight: torch.Tensor,
31
+ quant_type: torch.dtype,
32
+ return_unprocessed_quantized_tensor: bool,
33
+ ) -> List[torch.Tensor]:
34
+ return ops.quant_weights(
35
+ origin_weight, quant_type, return_unprocessed_quantized_tensor
36
+ )
build/torch211-cxx11-cu126-aarch64-linux/metadata.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "version": 1,
3
+ "license": "Apache-2.0",
4
+ "python-depends": [],
5
+ "backend": {
6
+ "type": "cuda",
7
+ "archs": [
8
+ "7.0",
9
+ "7.2",
10
+ "7.5",
11
+ "8.0",
12
+ "8.6",
13
+ "8.7",
14
+ "8.9",
15
+ "9.0+PTX"
16
+ ]
17
+ }
18
+ }
build/torch211-cxx11-cu126-aarch64-linux/quantization_eetq/__init__.py ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import ctypes
2
+ import importlib.util
3
+ import sys
4
+ from pathlib import Path
5
+ from types import ModuleType
6
+
7
+
8
+ def _import_from_path(file_path: Path) -> ModuleType:
9
+ # We cannot use the module name as-is, after adding it to `sys.modules`,
10
+ # it would also be used for other imports. So, we make a module name that
11
+ # depends on the path for it to be unique using the hex-encoded hash of
12
+ # the path.
13
+ path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
14
+ module_name = path_hash
15
+ spec = importlib.util.spec_from_file_location(module_name, file_path)
16
+ if spec is None:
17
+ raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
18
+ module = importlib.util.module_from_spec(spec)
19
+ if module is None:
20
+ raise ImportError(f"Cannot load module {module_name} from spec")
21
+ sys.modules[module_name] = module
22
+ spec.loader.exec_module(module) # type: ignore
23
+ return module
24
+
25
+
26
+ globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))
build/torch211-cxx11-cu126-x86_64-linux/__init__.py ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ from .custom_ops import w8_a16_gemm, w8_a16_gemm_, preprocess_weights, quant_weights
2
+
3
+ __all__ = ["w8_a16_gemm", "w8_a16_gemm_", "preprocess_weights", "quant_weights"]
build/torch211-cxx11-cu126-x86_64-linux/_ops.py ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ import torch
2
+ from . import _quantization_eetq_cuda_4bc0051
3
+ ops = torch.ops._quantization_eetq_cuda_4bc0051
4
+
5
+ def add_op_namespace_prefix(op_name: str):
6
+ """
7
+ Prefix op by namespace.
8
+ """
9
+ return f"_quantization_eetq_cuda_4bc0051::{op_name}"
build/torch211-cxx11-cu126-x86_64-linux/_quantization_eetq_cuda_4bc0051.abi3.so ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1264b126ed4d9c9d80758671f9965e830bfa916631934b81784e194c74b293a0
3
+ size 39053336
build/torch211-cxx11-cu126-x86_64-linux/custom_ops.py ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from typing import List
2
+ import torch
3
+
4
+ from ._ops import ops
5
+
6
+
7
+ def w8_a16_gemm(
8
+ input: torch.Tensor, weight: torch.Tensor, scale: torch.Tensor
9
+ ) -> torch.Tensor:
10
+ return ops.w8_a16_gemm(input, weight, scale)
11
+
12
+
13
+ def w8_a16_gemm_(
14
+ input: torch.Tensor,
15
+ weight: torch.Tensor,
16
+ scale: torch.Tensor,
17
+ output: torch.Tensor,
18
+ m: int,
19
+ n: int,
20
+ k: int,
21
+ ) -> torch.Tensor:
22
+ return ops.w8_a16_gemm_(input, weight, scale, output, m, n, k)
23
+
24
+
25
+ def preprocess_weights(origin_weight: torch.Tensor, is_int4: bool) -> torch.Tensor:
26
+ return ops.preprocess_weights(origin_weight, is_int4)
27
+
28
+
29
+ def quant_weights(
30
+ origin_weight: torch.Tensor,
31
+ quant_type: torch.dtype,
32
+ return_unprocessed_quantized_tensor: bool,
33
+ ) -> List[torch.Tensor]:
34
+ return ops.quant_weights(
35
+ origin_weight, quant_type, return_unprocessed_quantized_tensor
36
+ )
build/torch211-cxx11-cu126-x86_64-linux/metadata.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "version": 1,
3
+ "license": "Apache-2.0",
4
+ "python-depends": [],
5
+ "backend": {
6
+ "type": "cuda",
7
+ "archs": [
8
+ "7.0",
9
+ "7.2",
10
+ "7.5",
11
+ "8.0",
12
+ "8.6",
13
+ "8.7",
14
+ "8.9",
15
+ "9.0+PTX"
16
+ ]
17
+ }
18
+ }
build/torch211-cxx11-cu126-x86_64-linux/quantization_eetq/__init__.py ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import ctypes
2
+ import importlib.util
3
+ import sys
4
+ from pathlib import Path
5
+ from types import ModuleType
6
+
7
+
8
+ def _import_from_path(file_path: Path) -> ModuleType:
9
+ # We cannot use the module name as-is, after adding it to `sys.modules`,
10
+ # it would also be used for other imports. So, we make a module name that
11
+ # depends on the path for it to be unique using the hex-encoded hash of
12
+ # the path.
13
+ path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
14
+ module_name = path_hash
15
+ spec = importlib.util.spec_from_file_location(module_name, file_path)
16
+ if spec is None:
17
+ raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
18
+ module = importlib.util.module_from_spec(spec)
19
+ if module is None:
20
+ raise ImportError(f"Cannot load module {module_name} from spec")
21
+ sys.modules[module_name] = module
22
+ spec.loader.exec_module(module) # type: ignore
23
+ return module
24
+
25
+
26
+ globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))
build/torch211-cxx11-cu128-aarch64-linux/__init__.py ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ from .custom_ops import w8_a16_gemm, w8_a16_gemm_, preprocess_weights, quant_weights
2
+
3
+ __all__ = ["w8_a16_gemm", "w8_a16_gemm_", "preprocess_weights", "quant_weights"]
build/torch211-cxx11-cu128-aarch64-linux/_ops.py ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ import torch
2
+ from . import _quantization_eetq_cuda_4bc0051
3
+ ops = torch.ops._quantization_eetq_cuda_4bc0051
4
+
5
+ def add_op_namespace_prefix(op_name: str):
6
+ """
7
+ Prefix op by namespace.
8
+ """
9
+ return f"_quantization_eetq_cuda_4bc0051::{op_name}"
build/torch211-cxx11-cu128-aarch64-linux/_quantization_eetq_cuda_4bc0051.abi3.so ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:894bebb496c1327c2855c6b1c4b362ebde395fbdc2a5050127b20ebd2faa943d
3
+ size 45296712
build/torch211-cxx11-cu128-aarch64-linux/custom_ops.py ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from typing import List
2
+ import torch
3
+
4
+ from ._ops import ops
5
+
6
+
7
+ def w8_a16_gemm(
8
+ input: torch.Tensor, weight: torch.Tensor, scale: torch.Tensor
9
+ ) -> torch.Tensor:
10
+ return ops.w8_a16_gemm(input, weight, scale)
11
+
12
+
13
+ def w8_a16_gemm_(
14
+ input: torch.Tensor,
15
+ weight: torch.Tensor,
16
+ scale: torch.Tensor,
17
+ output: torch.Tensor,
18
+ m: int,
19
+ n: int,
20
+ k: int,
21
+ ) -> torch.Tensor:
22
+ return ops.w8_a16_gemm_(input, weight, scale, output, m, n, k)
23
+
24
+
25
+ def preprocess_weights(origin_weight: torch.Tensor, is_int4: bool) -> torch.Tensor:
26
+ return ops.preprocess_weights(origin_weight, is_int4)
27
+
28
+
29
+ def quant_weights(
30
+ origin_weight: torch.Tensor,
31
+ quant_type: torch.dtype,
32
+ return_unprocessed_quantized_tensor: bool,
33
+ ) -> List[torch.Tensor]:
34
+ return ops.quant_weights(
35
+ origin_weight, quant_type, return_unprocessed_quantized_tensor
36
+ )
build/torch211-cxx11-cu128-aarch64-linux/metadata.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "version": 1,
3
+ "license": "Apache-2.0",
4
+ "python-depends": [],
5
+ "backend": {
6
+ "type": "cuda",
7
+ "archs": [
8
+ "10.0",
9
+ "10.1",
10
+ "12.0+PTX",
11
+ "7.0",
12
+ "7.2",
13
+ "7.5",
14
+ "8.0",
15
+ "8.6",
16
+ "8.7",
17
+ "8.9",
18
+ "9.0"
19
+ ]
20
+ }
21
+ }
build/torch211-cxx11-cu128-aarch64-linux/quantization_eetq/__init__.py ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import ctypes
2
+ import importlib.util
3
+ import sys
4
+ from pathlib import Path
5
+ from types import ModuleType
6
+
7
+
8
+ def _import_from_path(file_path: Path) -> ModuleType:
9
+ # We cannot use the module name as-is, after adding it to `sys.modules`,
10
+ # it would also be used for other imports. So, we make a module name that
11
+ # depends on the path for it to be unique using the hex-encoded hash of
12
+ # the path.
13
+ path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
14
+ module_name = path_hash
15
+ spec = importlib.util.spec_from_file_location(module_name, file_path)
16
+ if spec is None:
17
+ raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
18
+ module = importlib.util.module_from_spec(spec)
19
+ if module is None:
20
+ raise ImportError(f"Cannot load module {module_name} from spec")
21
+ sys.modules[module_name] = module
22
+ spec.loader.exec_module(module) # type: ignore
23
+ return module
24
+
25
+
26
+ globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))
build/torch211-cxx11-cu128-x86_64-linux/__init__.py ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ from .custom_ops import w8_a16_gemm, w8_a16_gemm_, preprocess_weights, quant_weights
2
+
3
+ __all__ = ["w8_a16_gemm", "w8_a16_gemm_", "preprocess_weights", "quant_weights"]
build/torch211-cxx11-cu128-x86_64-linux/_ops.py ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ import torch
2
+ from . import _quantization_eetq_cuda_4bc0051
3
+ ops = torch.ops._quantization_eetq_cuda_4bc0051
4
+
5
+ def add_op_namespace_prefix(op_name: str):
6
+ """
7
+ Prefix op by namespace.
8
+ """
9
+ return f"_quantization_eetq_cuda_4bc0051::{op_name}"
build/torch211-cxx11-cu128-x86_64-linux/_quantization_eetq_cuda_4bc0051.abi3.so ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5782cd3f5577334deb176bd5e7502850d2302ade7b7df9ff9491c011ccba469
3
+ size 45382832
build/torch211-cxx11-cu128-x86_64-linux/custom_ops.py ADDED
@@ -0,0 +1,36 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ from typing import List
2
+ import torch
3
+
4
+ from ._ops import ops
5
+
6
+
7
+ def w8_a16_gemm(
8
+ input: torch.Tensor, weight: torch.Tensor, scale: torch.Tensor
9
+ ) -> torch.Tensor:
10
+ return ops.w8_a16_gemm(input, weight, scale)
11
+
12
+
13
+ def w8_a16_gemm_(
14
+ input: torch.Tensor,
15
+ weight: torch.Tensor,
16
+ scale: torch.Tensor,
17
+ output: torch.Tensor,
18
+ m: int,
19
+ n: int,
20
+ k: int,
21
+ ) -> torch.Tensor:
22
+ return ops.w8_a16_gemm_(input, weight, scale, output, m, n, k)
23
+
24
+
25
+ def preprocess_weights(origin_weight: torch.Tensor, is_int4: bool) -> torch.Tensor:
26
+ return ops.preprocess_weights(origin_weight, is_int4)
27
+
28
+
29
+ def quant_weights(
30
+ origin_weight: torch.Tensor,
31
+ quant_type: torch.dtype,
32
+ return_unprocessed_quantized_tensor: bool,
33
+ ) -> List[torch.Tensor]:
34
+ return ops.quant_weights(
35
+ origin_weight, quant_type, return_unprocessed_quantized_tensor
36
+ )
build/torch211-cxx11-cu128-x86_64-linux/metadata.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "version": 1,
3
+ "license": "Apache-2.0",
4
+ "python-depends": [],
5
+ "backend": {
6
+ "type": "cuda",
7
+ "archs": [
8
+ "10.0",
9
+ "10.1",
10
+ "12.0+PTX",
11
+ "7.0",
12
+ "7.2",
13
+ "7.5",
14
+ "8.0",
15
+ "8.6",
16
+ "8.7",
17
+ "8.9",
18
+ "9.0"
19
+ ]
20
+ }
21
+ }
build/torch211-cxx11-cu128-x86_64-linux/quantization_eetq/__init__.py ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import ctypes
2
+ import importlib.util
3
+ import sys
4
+ from pathlib import Path
5
+ from types import ModuleType
6
+
7
+
8
+ def _import_from_path(file_path: Path) -> ModuleType:
9
+ # We cannot use the module name as-is, after adding it to `sys.modules`,
10
+ # it would also be used for other imports. So, we make a module name that
11
+ # depends on the path for it to be unique using the hex-encoded hash of
12
+ # the path.
13
+ path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
14
+ module_name = path_hash
15
+ spec = importlib.util.spec_from_file_location(module_name, file_path)
16
+ if spec is None:
17
+ raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
18
+ module = importlib.util.module_from_spec(spec)
19
+ if module is None:
20
+ raise ImportError(f"Cannot load module {module_name} from spec")
21
+ sys.modules[module_name] = module
22
+ spec.loader.exec_module(module) # type: ignore
23
+ return module
24
+
25
+
26
+ globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))