[Transforms] Transform Registry Support

src/compressed_tensors/transforms/__init__.py

@@ -0,0 +1,18 @@
+# Copyright (c) 2021 - present / Neuralmagic, Inc. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from .base import Transforms
+from .hadamard import Hadamard
+from .matrix_multiply import MatrixMultiply
+from .random_hadamard import RandomHadamard

src/compressed_tensors/transforms/base.py

@@ -0,0 +1,87 @@
+# Copyright (c) 2021 - present / Neuralmagic, Inc. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Any, Optional, Union
+
+import torch
+from compressed_tensors.registry.registry import RegistryMixin
+from compressed_tensors.transforms.utils import apply_matrix_transform
+
+
+__all__ = ["Transforms"]
+
+
+# TODO: We don't need to save all the __call__ args for serialization or even have
+# them defined by a recipe. Some of them, such as if the transformation should be the
+# first or second matirx in torch.matmul depending on dimensions, can be inferred
+# by the layer time likely.
+
+MATIRX_TRANSFORMS = ["matrix-mul", "hadamard", "random-hadamard"]
+
+
+class Transforms(RegistryMixin):
+    def __new__(
+        cls,
+        transform: torch.Tensor,
+        device: Optional[Union[str, torch.device]] = "cuda",
+        dtype: Optional[torch.dtype] = torch.bfloat16,
+        *args,
+        **kwargs,
+    ):
+        """
+        Base class for setting up transforms. The registry creates transforms
+        as parameters which can be attached to modules.
+
+        import torch
+
+        size = 1024
+        dtype = torch.bfloat16
+        module = torch.nn.Linear(size, size)
+
+        hadamard_transform = Transforms.load_from_registry(
+            "random_hadamard", size=size, dtype=dtype
+        )
+        hadamard_apply = Transforms.fetch_apply("random_hadamard")
+        module.weight_transform = hadamard_transform
+
+        transformed_output = hadamard_apply(input_tensor=module.weight,
+            transform=moduel.weight_transform)
+
+        hadamard_inverse = Transforms.fetch_inverse_apply("random_hadamard")
+        original_weight = hadamard_inverse(input_tensor=transformed_output,
+            transform=model.weight_trainsform,
+            transpose=True)
+
+        :param transform: transform (e.g. torch.Tensor, scalar) to be applied
+        """
+        return torch.nn.Parameter(transform.to(device).to(dtype), requires_grad=False)
+
+    @classmethod
+    def fetch_apply(cls, name: str):
+        if name in MATIRX_TRANSFORMS:
+            return apply_matrix_transform
+        raise NotImplementedError("Only matrix transforms are supported")
+
+    @classmethod
+    def fetch_inverse_apply(cls, name: str):
+        return cls.get_value_from_registry(name=name).inverse_apply
+
+    @staticmethod
+    def inverse_apply(
+        transform: torch.Tensor, input_tensor: torch.Tensor, *args, **kwargs
+    ) -> torch.Tensor:
+        """
+        Apply the inverse operation applied by the apply method
+        """
+        raise NotImplementedError()

src/compressed_tensors/transforms/hadamard.py

@@ -0,0 +1,82 @@
+# Copyright (c) 2021 - present / Neuralmagic, Inc. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from typing import Optional, Union
+
+import torch
+from compressed_tensors.transforms import Transforms
+from compressed_tensors.transforms.hadamard_utils import deterministic_hadamard_matrix
+from compressed_tensors.transforms.utils import apply_matrix_transform
+
+
+@Transforms.register("hadamard")
+class Hadamard(Transforms):
+    def __new__(
+        cls,
+        size: int,
+        empty: Optional[bool] = False,
+        device: Optional[Union[str, torch.device]] = "cuda",
+        dtype: Optional[torch.dtype] = torch.bfloat16,
+    ):
+        """
+        Produces a hadamard matrix with dims (size, size), with values
+        -1 and 1, and the property HH.T == nI i.e the transformation
+        matrix when multiplied by its transpose is a multiple of the identity.
+        All rows and columns are orthonormal. The matrix returned
+        is not normalized and will be deterministic.
+
+        :param size: size of the matrix, if generating a new Hadamard matrix.
+            The generated matrix will have dimensions (size, size)
+        :param transform: if loading in a previously generated matrix, will
+            use that through this transformation, as opposed to creating a new
+            one
+        :param dtype: type to cast the rotation matrix to
+
+        """
+        if not empty:
+            # TODO: this is deterministic; we should just serialize the size
+            transform = torch.Tensor(deterministic_hadamard_matrix(size=size))
+        else:
+            transform = torch.empty((size, size))
+
+        return super().__new__(cls, transform=transform, device=device, dtype=dtype)
+
+    @staticmethod
+    def inverse_apply(
+        transform: torch.Tensor,
+        input_tensor: torch.Tensor,
+        transpose: bool = False,
+        first: bool = True,
+    ) -> torch.Tensor:
+        """
+        Apply the inverse operation of `apply`
+
+        :param transform: hadamard tensor
+        :param input_tensor: tensor to which the transform matrix is applied
+        :param transpose: whether or not the transform matrix is transposed before
+            being applied.
+        :param first: if the transform matrix will be the first or second matrix to be
+            multiplied
+        """
+        transpose = not transpose
+        # need to normalize before sending back
+        return (
+            apply_matrix_transform(
+                transform=transform,
+                input_tensor=input_tensor,
+                transpose=transpose,
+                first=first,
+            )
+            / transform.shape[0]
+        )

src/compressed_tensors/transforms/hadamard_utils.py

@@ -0,0 +1,166 @@
+# Copyright (c) 2021 - present / Neuralmagic, Inc. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import math
+
+import numpy
+import torch
+
+
+__all__ = ["random_hadamard_matrix", "deterministic_hadamard_matrix"]
+
+# adapted from:
+# https://github.com/scipy/scipy/blob/v1.15.2/scipy/linalg/_special_matrices.py
+def deterministic_hadamard_matrix(size: int):
+    """
+    Construct an Hadamard matrix.
+
+    Constructs an n-by-n Hadamard matrix, using Sylvester's
+    construction. `n` must be a power of 2.
+
+    :param size: order of the matrix; must be a power of 2
+
+    returns a (size, size) hadamard matrix
+    """
+
+    dtype = int
+    if size < 1:
+        lg2 = 0
+    else:
+        lg2 = int(math.log(size, 2))
+    if 2**lg2 != size:
+        raise ValueError("size must be an positive integer and a power of 2")
+
+    H = numpy.array([[1]], dtype=dtype)
+
+    # Sylvester's construction
+    for i in range(0, lg2):
+        H = numpy.vstack((numpy.hstack((H, H)), numpy.hstack((H, -H))))
+
+    return H
+
+
+# adapted from:
+# https://github.com/facebookresearch/SpinQuant/blob/main/utils/hadamard_utils.py
+
+# TODO: the following library exists for online rotations and should be considered
+# in the future:
+# https://github.com/Dao-AILab/fast-hadamard-transform/tree/master
+
+
+def random_hadamard_matrix(size: int) -> torch.Tensor:
+    """
+    Produces a randomly generated Hadamard matrix.
+    See https://cornell-relaxml.github.io/quip-sharp/ ,
+    Section "Randomized Hadamard Transformation"
+
+    :param size: The dimension of the matrix. Matrix generated will have dimensions
+        (size, size)
+
+    """
+    # TODO: potentially update to add "seed" as an arugment, to allow
+    # the matrix generated to be reproducible
+
+    # Benefits: support other shapes / non powers of 2, support randomization
+    Q = torch.randint(low=0, high=2, size=(size,)).to(torch.float64)
+    Q = Q * 2 - 1
+    Q = torch.diag(Q)
+    return _matmul_hadU(Q)
+
+
+def _get_hadK(n, transpose=False):
+    # NOTE: we can easily extend the list of supported shapes/sizes
+    # by adding to these methods
+    hadK, K = None, None
+    if n % 20 == 0:
+        assert _is_pow2(n // 20)
+        K = 20
+        hadK = _get_had20().T if transpose else _get_had20()
+    elif n % 12 == 0:
+        assert _is_pow2(n // 12)
+        K = 12
+        hadK = _get_had12().T if transpose else _get_had12()
+    else:
+        assert _is_pow2(n)
+        K = 1
+
+    return hadK, K
+
+
+def _matmul_hadU(X, transpose=False):
+    n = X.shape[-1]
+    # Check if we have the determined hadamard matrix
+    hadK, K = _get_hadK(n, transpose)
+    # Reshape diag matrix with randomized -1/+1
+    input = X.clone().view(-1, n, 1)
+    output = input.clone()
+
+    # for cases when hadK is not predetermined, determine hadamard matrix
+    while input.shape[1] > K:
+        input = input.view(input.shape[0], input.shape[1] // 2, 2, input.shape[2])
+        output = output.view(input.shape)
+        output[:, :, 0, :] = input[:, :, 0, :] + input[:, :, 1, :]
+        output[:, :, 1, :] = input[:, :, 0, :] - input[:, :, 1, :]
+        output = output.view(input.shape[0], input.shape[1], -1)
+        (input, output) = (output, input)
+    del output
+
+    # K == 1 when hadK is None; this happens when the size dim (n)
+    # is not comaptible with any of the maintained hadamard matrices
+
+    if K > 1:
+        # Do not explicitly repeat - OOM
+        # input = torch.bmm(
+        #     hadK.repeat(len(input), 1, 1).to(input.device).to(input.dtype), input)
+        # Use bcast instead
+
+        # for cases when hadK is pre-determined
+        input = hadK.view(1, K, K).to(input) @ input
+
+    # normalize
+    return input.view(X.shape) / torch.tensor(n).sqrt()
+
+
+def _is_pow2(n):
+    return (n & (n - 1) == 0) and (n > 0)
+
+
+def _reshape_bits(packed_bits, original_size):
+    had_unpacked = numpy.unpackbits(packed_bits)
+    had_unpacked = [1 if x == 1 else -1 for x in had_unpacked]
+    had_unpacked = numpy.array(had_unpacked).reshape((original_size, original_size))
+    return had_unpacked
+
+
+# http://www.neilsloane.com/hadamard/index.html
+def _get_had12():
+    # fmt: off
+    had_12 = numpy.array([128,  13,  29, 232, 235,  71, 218,  
+        62, 209, 246, 139, 180, 157, 168, 237, 199, 106,  59], dtype=numpy.uint8)
+    # fmt: on
+    # TODO: just unpack during apply
+    had_12_unpacked = _reshape_bits(had_12, original_size=12)
+    return torch.FloatTensor(had_12_unpacked)
+
+
+def _get_had20():
+    # fmt: off
+    had_20 = numpy.array([128, 0,  13, 133, 121, 236,  43, 203,  97,  94, 155,  10, 252, 
+        216, 87, 230, 194, 191,  54,  21, 249, 176, 171, 205, 133, 222, 108,  42, 243,  
+        97, 215, 155,  10, 188, 216, 149, 230, 200, 175, 54, 133, 121, 188,  43, 
+        205, 225,  94, 107,  10, 243], dtype=numpy.uint8)
+    # fmt: on
+    # TODO: just unpack during apply
+    had_20_unpacked = _reshape_bits(had_20, original_size=20)
+    return torch.FloatTensor(had_20_unpacked)