Build uploaded using `kernels`.

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+build/torch28-cxx11-cpu-x86_64-linux/_quantization_gptq_8c16cd6.abi3.so filter=lfs diff=lfs merge=lfs -text
+build/torch29-cxx11-cpu-x86_64-linux/_quantization_gptq_8c16cd6.abi3.so filter=lfs diff=lfs merge=lfs -text

build/torch28-cxx11-cpu-x86_64-linux/__init__.py

@@ -0,0 +1,3 @@
+from .custom_ops import gemm_int4_forward
+
+__all__ = ["gemm_int4_forward"]

build/torch28-cxx11-cpu-x86_64-linux/_ops.py

@@ -0,0 +1,9 @@
+import torch
+from . import _quantization_gptq_8c16cd6
+ops = torch.ops._quantization_gptq_8c16cd6
+
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_quantization_gptq_8c16cd6::{op_name}"

build/torch28-cxx11-cpu-x86_64-linux/_quantization_gptq_8c16cd6.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b092ba64e57468f1ef86d4003017afe6dcc44b792300543492c0463903d498f3
+size 101912

build/torch28-cxx11-cpu-x86_64-linux/custom_ops.py

@@ -0,0 +1,19 @@
+import torch
+from ._ops import ops
+
+def gemm_int4_forward(
+    input: torch.Tensor,
+    weight: torch.Tensor,
+    zeros: torch.Tensor,
+    absmax: torch.Tensor,
+    blocksize: int,
+) -> torch.Tensor:
+    original_dtype = input.dtype
+    if original_dtype != torch.bfloat16:
+        input = input.to(torch.bfloat16)
+
+    output = ops.gemm_int4_forward(input, weight, zeros, absmax, blocksize)
+    if original_dtype != torch.bfloat16:
+        output = output.to(original_dtype)
+
+    return output

build/torch28-cxx11-cpu-x86_64-linux/metadata.json

@@ -0,0 +1 @@
+{"python-depends":[]}

build/torch28-cxx11-cpu-x86_64-linux/quantization_gptq/__init__.py

@@ -0,0 +1,26 @@
+import ctypes
+import sys
+
+import importlib
+from pathlib import Path
+from types import ModuleType
+
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+
+
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))

build/torch29-cxx11-cpu-x86_64-linux/__init__.py

@@ -0,0 +1,3 @@
+from .custom_ops import gemm_int4_forward
+
+__all__ = ["gemm_int4_forward"]

build/torch29-cxx11-cpu-x86_64-linux/_ops.py

@@ -0,0 +1,9 @@
+import torch
+from . import _quantization_gptq_8c16cd6
+ops = torch.ops._quantization_gptq_8c16cd6
+
+def add_op_namespace_prefix(op_name: str):
+    """
+    Prefix op by namespace.
+    """
+    return f"_quantization_gptq_8c16cd6::{op_name}"

build/torch29-cxx11-cpu-x86_64-linux/_quantization_gptq_8c16cd6.abi3.so

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6036c7d9889872a5de65029db37fb88ee23a3efd4dfc62ec78c7ccebc2cd0c4d
+size 105960

build/torch29-cxx11-cpu-x86_64-linux/custom_ops.py

@@ -0,0 +1,19 @@
+import torch
+from ._ops import ops
+
+def gemm_int4_forward(
+    input: torch.Tensor,
+    weight: torch.Tensor,
+    zeros: torch.Tensor,
+    absmax: torch.Tensor,
+    blocksize: int,
+) -> torch.Tensor:
+    original_dtype = input.dtype
+    if original_dtype != torch.bfloat16:
+        input = input.to(torch.bfloat16)
+
+    output = ops.gemm_int4_forward(input, weight, zeros, absmax, blocksize)
+    if original_dtype != torch.bfloat16:
+        output = output.to(original_dtype)
+
+    return output

build/torch29-cxx11-cpu-x86_64-linux/metadata.json

@@ -0,0 +1 @@
+{"python-depends":[]}

build/torch29-cxx11-cpu-x86_64-linux/quantization_gptq/__init__.py

@@ -0,0 +1,26 @@
+import ctypes
+import sys
+
+import importlib
+from pathlib import Path
+from types import ModuleType
+
+def _import_from_path(file_path: Path) -> ModuleType:
+    # We cannot use the module name as-is, after adding it to `sys.modules`,
+    # it would also be used for other imports. So, we make a module name that
+    # depends on the path for it to be unique using the hex-encoded hash of
+    # the path.
+    path_hash = "{:x}".format(ctypes.c_size_t(hash(file_path.absolute())).value)
+    module_name = path_hash
+    spec = importlib.util.spec_from_file_location(module_name, file_path)
+    if spec is None:
+        raise ImportError(f"Cannot load spec for {module_name} from {file_path}")
+    module = importlib.util.module_from_spec(spec)
+    if module is None:
+        raise ImportError(f"Cannot load module {module_name} from spec")
+    sys.modules[module_name] = module
+    spec.loader.exec_module(module)  # type: ignore
+    return module
+
+
+globals().update(vars(_import_from_path(Path(__file__).parent.parent / "__init__.py")))