Add Int4PlainInt32Tensor

test/quantization/quantize_/workflows/int4/test_int4_plain_int32_tensor.py

@@ -0,0 +1,86 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD 3-Clause license found in the
+# LICENSE file in the root directory of this source tree.
+
+import tempfile
+import unittest
+
+import torch
+from torch.testing._internal.common_utils import (
+    TestCase,
+    instantiate_parametrized_tests,
+    parametrize,
+    run_tests,
+)
+
+from torchao.quantization import (
+    Int4WeightOnlyConfig,
+    quantize_,
+)
+from torchao.quantization.utils import compute_error
+from torchao.utils import (
+    torch_version_at_least,
+)
+
+
+def get_config(group_size):
+    return Int4WeightOnlyConfig(
+        group_size=group_size,
+        packing_format="plain_int32",
+        version=2,
+    )
+
+
+@unittest.skipIf(not torch_version_at_least("2.8.0"), "Need pytorch 2.8+")
+@unittest.skipIf(not torch.xpu.is_available(), "XPU not available")
+class Int4PlainInt32Tensor(TestCase):
+    @parametrize(
+        "sizes",
+        [
+            ((128,), 256, 128),
+            ((32, 128), 512, 128),
+            ((2, 32, 128), 256, 12),
+        ],
+    )
+    @parametrize("dtype", [torch.bfloat16, torch.half])
+    @parametrize("group_size", [32, 64, 128])
+    def test_linear(self, sizes, dtype, group_size):
+        device = "xpu"
+        M, N, K = sizes
+        input = torch.randn(*M, K, dtype=dtype, device=device)
+        linear = torch.nn.Linear(K, N, dtype=dtype, device=device)
+        original = linear(input)
+        quantize_(linear, get_config(group_size))
+        quantized = linear(input)
+        self.assertTrue(compute_error(original, quantized) > 20)
+
+        compiled_linear = torch.compile(linear)
+        quantized_and_compiled = compiled_linear(input)
+        self.assertTrue(compute_error(original, quantized_and_compiled) > 20)
+
+    @parametrize("dtype", [torch.bfloat16, torch.half])
+    def test_module_path(self, dtype):
+        linear = torch.nn.Linear(128, 256, dtype=dtype, device="xpu")
+        quantize_(linear, get_config(group_size=128))
+        self.assertEqual(
+            str(type(linear.weight)),
+            "<class 'torchao.quantization.Int4PlainInt32Tensor'>",
+        )
+
+        with tempfile.NamedTemporaryFile() as f:
+            torch.save(linear.state_dict(), f)
+            f.seek(0)
+            state_dict = torch.load(f)
+            self.assertEqual(
+                str(type(state_dict["weight"])),
+                "<class 'torchao.quantization.Int4PlainInt32Tensor'>",
+            )
+
+
+instantiate_parametrized_tests(Int4PlainInt32Tensor)
+
+
+if __name__ == "__main__":
+    run_tests()

torchao/quantization/__init__.py

@@ -92,6 +92,7 @@
     Float8Tensor,
     Int4MarlinSparseTensor,
     Int4OpaqueTensor,
+    Int4PlainInt32Tensor,
     Int4PreshuffledTensor,
     Int4Tensor,
     Int4TilePackedTo4dTensor,
@@ -163,6 +164,7 @@
     "FbgemmConfig",
     # tensor subclasses
     "Int4Tensor",
+    "Int4PlainInt32Tensor",
     "Int4PreshuffledTensor",
     "Int4MarlinSparseTensor",
     "IntxOpaqueTensor",

torchao/quantization/quant_api.py

@@ -74,6 +74,7 @@
     Float8Tensor,
     Int4MarlinSparseTensor,
     Int4OpaqueTensor,
+    Int4PlainInt32Tensor,
     Int4PreshuffledTensor,
     Int4Tensor,
     Int4TilePackedTo4dTensor,
@@ -522,7 +523,6 @@ def quantize_(
     torch._C._log_api_usage_once("torchao.quantization.quantize_")
 
     filter_fn = _is_linear if filter_fn is None else filter_fn
-
     if isinstance(config, ModuleFqnToConfig):
         _replace_with_custom_fn_if_matches_filter_with_name(
             model,
@@ -1131,6 +1131,12 @@ def _int4_weight_only_quantize_tensor(weight, config):
                 block_size,
             )
             return new_weight
+        elif packing_format == PackingFormat.PLAIN_INT32:
+            new_weight = Int4PlainInt32Tensor.from_hp(
+                weight,
+                block_size,
+            )
+            return new_weight
         elif packing_format == PackingFormat.MARLIN_SPARSE:
             new_weight = Int4MarlinSparseTensor.from_hp(
                 weight,

torchao/quantization/quantize_/common/packing_format.py

@@ -41,6 +41,12 @@ class PackingFormat(str, Enum):
     """
     UNPACKED_TO_INT8 = "unpacked_to_int8"
 
+    """
+    plain_int32 is referring to the format used by int4 weight-only quantization.
+    which is a groupwise quantization format 2*int4 is store in a byte and 4*(int4*2) is stored in a int32.
+    """
+    PLAIN_INT32 = "plain_int32"
+
     """
     tile_packed_to_4d is referring to the format used by tinygemm kernels for int4 quantization
     """

torchao/quantization/quantize_/workflows/__init__.py

@@ -8,6 +8,9 @@
 from .int4.int4_opaque_tensor import (
     Int4OpaqueTensor,
 )
+from .int4.int4_plain_int32_tensor import (
+    Int4PlainInt32Tensor,
+)
 from .int4.int4_preshuffled_tensor import (
     Int4PreshuffledTensor,
 )
@@ -26,6 +29,7 @@
     "Int4Tensor",
     "Int4PreshuffledTensor",
     "Int4MarlinSparseTensor",
+    "Int4PlainInt32Tensor",
     "Int4TilePackedTo4dTensor",
     "Float8Tensor",
     "QuantizeTensorToFloat8Kwargs",

torchao/quantization/quantize_/workflows/int4/int4_plain_int32_tensor.py

@@ -0,0 +1,182 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD 3-Clause license found in the
+# LICENSE file in the root directory of this source tree.
+
+
+from typing import List
+
+import torch
+
+from torchao.quantization.quant_primitives import (
+    MappingType,
+    choose_qparams_affine,
+    quantize_affine,
+)
+from torchao.utils import (
+    TorchAOBaseTensor,
+)
+
+__all__ = [
+    "Int4PlainInt32Tensor",
+]
+
+aten = torch.ops.aten
+
+
+class Int4PlainInt32Tensor(TorchAOBaseTensor):
+    """
+    int4 weight-only quantization on XPU with oneDNN as backend (groupwise quantization only)
+
+    Tensor Attributes:
+        qdata: (N, K/8), packed int4 weight, the data type is int32 here with 4*(int4*2), the original data type can be half and bfloat16
+        scale: (K/group_size, N), dtype is the same as the original Tensor dtype
+        zero_point: (K/group_size, N), dtype is int8
+
+    Non-Tensor Attributes:
+        block_size: the block size for quantization, representing the granularity.
+        shape: shape of the original Tensor
+
+    """
+
+    tensor_data_names = ["qdata", "scale", "zero_point"]
+    tensor_attribute_names = ["block_size", "shape"]
+
+    def __new__(
+        cls,
+        qdata,
+        scale,
+        zero_point,
+        block_size,
+        shape,
+    ):
+        kwargs = {}
+        kwargs["device"] = qdata.device
+        kwargs["dtype"] = scale.dtype
+        kwargs["requires_grad"] = False
+        return torch.Tensor._make_wrapper_subclass(cls, shape, **kwargs)  # type: ignore[attr-defined]
+
+    def __init__(self, qdata, scale, zero_point, block_size, shape):
+        self.qdata = qdata
+        self.scale = scale
+        self.zero_point = zero_point
+        self.block_size = block_size
+
+    def _quantization_type(self):
+        return f"shape={self.shape}, block_size={self.block_size}, device={self.device}"
+
+    @classmethod
+    def from_hp(
+        cls,
+        w: torch.Tensor,
+        block_size: List[int],
+    ):
+        assert w.ndim == 2 and w.device.type == "xpu", (
+            f"Expecting 2D tensor on XPU, but got: {w.shape} on {w.device.type}"
+        )
+        assert len(block_size) == w.ndim
+        assert w.dtype in [torch.float16, torch.bfloat16], (
+            f"Expecting float16 or bfloat16 weight tensor, but got: {w.dtype}"
+        )
+        original_shape = w.shape
+        mapping_type = MappingType.ASYMMETRIC
+        target_dtype = torch.int32
+        quant_min = 0
+        quant_max = 15
+        eps = 1e-6
+        scale_dtype = None
+        zero_point_dtype = torch.int32
+        scale, zero_point = choose_qparams_affine(
+            w,
+            mapping_type,
+            block_size,
+            target_dtype,
+            quant_min,
+            quant_max,
+            eps,
+            scale_dtype,
+            zero_point_dtype,
+        )
+        int_data = quantize_affine(
+            w,
+            block_size,
+            scale,
+            zero_point,
+            target_dtype,
+            quant_min,
+            quant_max,
+        )
+        assert int_data.dtype == torch.int32, (
+            "torch.ops.aten._convert_weight_to_int4pack expects `int32` dtype"
+        )
+        packed_weight = (int_data[::, 1::2] << 4 | int_data[::, ::2]).to(torch.uint8)
+        packed_weight = torch.ops.aten._convert_weight_to_int4pack(
+            packed_weight.contiguous(), 8
+        )
+        scale = scale.reshape(int_data.shape[0], -1)
+        zero_point = zero_point.reshape(int_data.shape[0], -1)
+        return Int4PlainInt32Tensor(
+            packed_weight,
+            scale.transpose(0, 1).contiguous(),
+            zero_point.transpose(0, 1).contiguous().to(torch.int8),
+            block_size,
+            original_shape,
+        )
+
+
+implements = Int4PlainInt32Tensor.implements
+
+
+@implements([torch.nn.functional.linear, aten.linear.default])
+def _(func, types, args, kwargs):
+    input_tensor, weight_tensor, bias = (
+        args[0],
+        args[1],
+        args[2] if len(args) > 2 else None,
+    )
+    assert input_tensor.device.type == "xpu", (
+        f"For XPU device only but got: {input_tensor.device}"
+    )
+    assert isinstance(weight_tensor, Int4PlainInt32Tensor), (
+        f"Expected weight_tensor to be Int4PlainInt32Tensor, got: {type(weight_tensor)}"
+    )
+    assert weight_tensor.block_size[0] == 1, (
+        f"Requires groupwise quantization, got block_size: {weight_tensor.block_size}"
+    )
+    assert input_tensor.shape[-1] == weight_tensor.shape[1], (
+        f"Shapes of input and weight do not match, input:{input_tensor.shape}, weight: {weight_tensor.shape}"
+    )
+
+    act_mat = input_tensor
+    packed_weight = weight_tensor.qdata
+    scale = weight_tensor.scale
+    zero_point = weight_tensor.zero_point
+
+    orig_act_size = act_mat.size()
+    orig_dtype = act_mat.dtype
+
+    # reshape to 2D
+    act_mat = act_mat.reshape(-1, act_mat.shape[-1])
+
+    # groupwise int4 quantization
+    groupsize = weight_tensor.block_size[1]
+    y = torch.ops.aten._weight_int4pack_mm_with_scales_and_zeros(
+        act_mat, packed_weight, groupsize, scale, zero_point
+    )
+
+    # remove out_feature padding
+    assert weight_tensor.ndim == 2
+    orig_out_features = weight_tensor.shape[-2]
+    y = y[:, :orig_out_features]
+    y = y.reshape(*orig_act_size[:-1], orig_out_features)
+
+    if bias is not None:
+        y += bias
+    return y.to(orig_dtype)
+
+
+Int4PlainInt32Tensor.__module__ = "torchao.quantization"
+
+# Allow a model with Int4PlainInt32Tensor weights to be loaded with `weights_only=True`
+torch.serialization.add_safe_globals([Int4PlainInt32Tensor])