Build for Torch 2.8 (x86_64-linux)

build/torch28-cxx11-cu126-x86_64-linux/quantization_eetq/__init__.py

@@ -0,0 +1,3 @@
+from .custom_ops import w8_a16_gemm, w8_a16_gemm_, preprocess_weights, quant_weights
+
+__all__ = ["w8_a16_gemm", "w8_a16_gemm_", "preprocess_weights", "quant_weights"]

build/torch28-cxx11-cu126-x86_64-linux/quantization_eetq/_ops.py

@@ -0,0 +1,9 @@
+import torch
+from . import _quantization_eetq_0c60fb4
+ops = torch.ops._quantization_eetq_0c60fb4
+
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_quantization_eetq_0c60fb4::{op_name}"

build/torch28-cxx11-cu126-x86_64-linux/quantization_eetq/_quantization_eetq_0c60fb4.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:854740b7499076f0f7c54be036dc9a66bdf4aa0bab322f8f2ef7e92723f76f93
+size 30881144

build/torch28-cxx11-cu126-x86_64-linux/quantization_eetq/custom_ops.py

@@ -0,0 +1,36 @@
+from typing import List
+import torch
+
+from ._ops import ops
+
+
+def w8_a16_gemm(
+    input: torch.Tensor, weight: torch.Tensor, scale: torch.Tensor
+) -> torch.Tensor:
+    return ops.w8_a16_gemm(input, weight, scale)
+
+
+def w8_a16_gemm_(
+    input: torch.Tensor,
+    weight: torch.Tensor,
+    scale: torch.Tensor,
+    output: torch.Tensor,
+    m: int,
+    n: int,
+    k: int,
+) -> torch.Tensor:
+    return ops.w8_a16_gemm_(input, weight, scale, output, m, n, k)
+
+
+def preprocess_weights(origin_weight: torch.Tensor, is_int4: bool) -> torch.Tensor:
+    return ops.preprocess_weights(origin_weight, is_int4)
+
+
+def quant_weights(
+    origin_weight: torch.Tensor,
+    quant_type: torch.dtype,
+    return_unprocessed_quantized_tensor: bool,
+) -> List[torch.Tensor]:
+    return ops.quant_weights(
+        origin_weight, quant_type, return_unprocessed_quantized_tensor
+    )

build/torch28-cxx11-cu128-x86_64-linux/quantization_eetq/__init__.py

@@ -0,0 +1,3 @@
+from .custom_ops import w8_a16_gemm, w8_a16_gemm_, preprocess_weights, quant_weights
+
+__all__ = ["w8_a16_gemm", "w8_a16_gemm_", "preprocess_weights", "quant_weights"]

build/torch28-cxx11-cu128-x86_64-linux/quantization_eetq/_ops.py

@@ -0,0 +1,9 @@
+import torch
+from . import _quantization_eetq_0c60fb4
+ops = torch.ops._quantization_eetq_0c60fb4
+
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_quantization_eetq_0c60fb4::{op_name}"

build/torch28-cxx11-cu128-x86_64-linux/quantization_eetq/_quantization_eetq_0c60fb4.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:213e1c9f8e5db3152b870ed377f10e7acd1f3c1fd584bce8721c1b88c612a5e9
+size 37476304

build/torch28-cxx11-cu128-x86_64-linux/quantization_eetq/custom_ops.py

@@ -0,0 +1,36 @@
+from typing import List
+import torch
+
+from ._ops import ops
+
+
+def w8_a16_gemm(
+    input: torch.Tensor, weight: torch.Tensor, scale: torch.Tensor
+) -> torch.Tensor:
+    return ops.w8_a16_gemm(input, weight, scale)
+
+
+def w8_a16_gemm_(
+    input: torch.Tensor,
+    weight: torch.Tensor,
+    scale: torch.Tensor,
+    output: torch.Tensor,
+    m: int,
+    n: int,
+    k: int,
+) -> torch.Tensor:
+    return ops.w8_a16_gemm_(input, weight, scale, output, m, n, k)
+
+
+def preprocess_weights(origin_weight: torch.Tensor, is_int4: bool) -> torch.Tensor:
+    return ops.preprocess_weights(origin_weight, is_int4)
+
+
+def quant_weights(
+    origin_weight: torch.Tensor,
+    quant_type: torch.dtype,
+    return_unprocessed_quantized_tensor: bool,
+) -> List[torch.Tensor]:
+    return ops.quant_weights(
+        origin_weight, quant_type, return_unprocessed_quantized_tensor
+    )

build/torch28-cxx11-cu129-x86_64-linux/quantization_eetq/__init__.py

@@ -0,0 +1,3 @@
+from .custom_ops import w8_a16_gemm, w8_a16_gemm_, preprocess_weights, quant_weights
+
+__all__ = ["w8_a16_gemm", "w8_a16_gemm_", "preprocess_weights", "quant_weights"]

build/torch28-cxx11-cu129-x86_64-linux/quantization_eetq/_ops.py

@@ -0,0 +1,9 @@
+import torch
+from . import _quantization_eetq_0c60fb4
+ops = torch.ops._quantization_eetq_0c60fb4
+
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_quantization_eetq_0c60fb4::{op_name}"

build/torch28-cxx11-cu129-x86_64-linux/quantization_eetq/_quantization_eetq_0c60fb4.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91d74be2ae273fc4826b1d9f6c725e9462217e45705df2dd3aa4574479ae4e1b
+size 38205656

build/torch28-cxx11-cu129-x86_64-linux/quantization_eetq/custom_ops.py

@@ -0,0 +1,36 @@
+from typing import List
+import torch
+
+from ._ops import ops
+
+
+def w8_a16_gemm(
+    input: torch.Tensor, weight: torch.Tensor, scale: torch.Tensor
+) -> torch.Tensor:
+    return ops.w8_a16_gemm(input, weight, scale)
+
+
+def w8_a16_gemm_(
+    input: torch.Tensor,
+    weight: torch.Tensor,
+    scale: torch.Tensor,
+    output: torch.Tensor,
+    m: int,
+    n: int,
+    k: int,
+) -> torch.Tensor:
+    return ops.w8_a16_gemm_(input, weight, scale, output, m, n, k)
+
+
+def preprocess_weights(origin_weight: torch.Tensor, is_int4: bool) -> torch.Tensor:
+    return ops.preprocess_weights(origin_weight, is_int4)
+
+
+def quant_weights(
+    origin_weight: torch.Tensor,
+    quant_type: torch.dtype,
+    return_unprocessed_quantized_tensor: bool,
+) -> List[torch.Tensor]:
+    return ops.quant_weights(
+        origin_weight, quant_type, return_unprocessed_quantized_tensor
+    )