[CPU] add Float8OpaqueTensor for dynamic float8 act float8 weight

test/quantization/quantize_/workflows/float8/test_float8_opaque_tensor.py

@@ -0,0 +1,164 @@
+# 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 import common_utils
+from torch.testing._internal.common_utils import (
+    TestCase,
+    run_tests,
+)
+
+from torchao import quantize_
+from torchao.quantization import PerGroup, PerRow, PerTensor
+from torchao.quantization.quant_api import (
+    Float8DynamicActivationFloat8WeightConfig,
+)
+from torchao.quantization.utils import compute_error
+from torchao.utils import (
+    torch_version_at_least,
+)
+
+
+def get_config(granularity):
+    return Float8DynamicActivationFloat8WeightConfig(
+        activation_dtype=torch.float8_e4m3fn,
+        granularity=granularity,
+        float8_packing_format="opaque",
+    )
+
+
+class ToyLinearModel(torch.nn.Module):
+    def __init__(self, K=64, N=32, bias=False):
+        super().__init__()
+        self.linear1 = torch.nn.Linear(K, N, bias=bias).to(torch.float)
+        self.linear2 = torch.nn.Linear(N, K, bias=bias).to(torch.float)
+
+    def example_inputs(self, batch_size=1, dtype=torch.float, device="cpu"):
+        return (
+            torch.rand(batch_size, self.linear1.in_features, dtype=dtype, device=device)
+            * 0.1,
+        )
+
+    def forward(self, x):
+        x = self.linear1(x)
+        x = self.linear2(x)
+        return x
+
+
+class TestFloat8OpaqueTensor(TestCase):
+    """Test cases for Float8OpaqueTensor on CPU"""
+
+    @unittest.skipIf(
+        "CPU" not in torch._C._dispatch_dump("torchao::float8_linear_cpu"),
+        reason="cpp kernels not built",
+    )
+    @unittest.skipIf(not torch_version_at_least("2.6.0"), "Test only enabled for 2.6+")
+    @common_utils.parametrize("dtype", [torch.float, torch.bfloat16, torch.half])
+    @common_utils.parametrize("x_dim", [2, 3])
+    @common_utils.parametrize("bias", [True, False])
+    @common_utils.parametrize("bs", [1, 160])
+    @common_utils.parametrize(
+        "x_granularity",
+        [PerTensor(), PerRow(), PerGroup(32), PerGroup(64), PerGroup(128)],
+    )
+    @common_utils.parametrize(
+        "w_granularity",
+        [PerTensor(), PerRow(), PerGroup(32), PerGroup(64), PerGroup(128)],
+    )
+    def test_dynamic_float8_linear(
+        self, dtype, x_dim, bias, bs, x_granularity, w_granularity
+    ):
+        if isinstance(x_granularity, PerGroup):
+            if not isinstance(w_granularity, PerGroup):
+                return
+            if w_granularity.group_size != x_granularity.group_size:
+                return
+        device = "cpu"
+        m = ToyLinearModel(256, 256, bias=bias).eval().to(dtype).to(device)
+        example_inputs = m.example_inputs(batch_size=bs, dtype=dtype, device=device)
+        if x_dim == 3:
+            example_inputs = (example_inputs[0].unsqueeze(0),)
+        y = m(*example_inputs)
+
+        with torch.no_grad():
+            quantize_(
+                m,
+                get_config([x_granularity, w_granularity]),
+            )
+            y1 = m(*example_inputs)
+            assert compute_error(y, y1) > 20
+            y2, code = torch._inductor.utils.run_and_get_code(
+                torch.compile(m, fullgraph=True, dynamic=True),
+                *example_inputs,
+            )
+            # ensure the expected op is in the code
+            assert "torch.ops.torchao.float8_linear_cpu.default" in code[0]
+            assert compute_error(y, y2) > 20
+
+    @unittest.skipIf(
+        "CPU" not in torch._C._dispatch_dump("torchao::float8_linear_cpu"),
+        reason="cpp kernels not built",
+    )
+    @unittest.skipIf(not torch_version_at_least("2.6.0"), "Test only enabled for 2.6+")
+    @common_utils.parametrize("dtype", [torch.float, torch.bfloat16, torch.half])
+    @common_utils.parametrize("x_dim", [2, 3])
+    @common_utils.parametrize("bias", [True, False])
+    @common_utils.parametrize("bs", [4, 128])
+    def test_dynamic_float8_linear_ref(self, dtype, x_dim, bias, bs):
+        device = "cpu"
+        # the shape is not supported by cpp kernel, so the ref path will be used.
+        m = ToyLinearModel(120, 120, bias=bias).eval().to(dtype).to(device)
+        example_inputs = m.example_inputs(batch_size=bs, dtype=dtype, device=device)
+        if x_dim == 3:
+            example_inputs = (example_inputs[0].unsqueeze(0),)
+        y = m(*example_inputs)
+
+        with torch.no_grad():
+            quantize_(
+                m,
+                get_config(PerRow()),
+            )
+            y1 = m(*example_inputs)
+            assert compute_error(y, y1) > 20
+            y2, code = torch._inductor.utils.run_and_get_code(
+                torch.compile(m, fullgraph=True, dynamic=True),
+                *example_inputs,
+            )
+            # ensure the expected op is in the code
+            assert "torch.ops.torchao.float8_linear_cpu.default" in code[0]
+            assert compute_error(y, y2) > 20
+
+    @unittest.skipIf(
+        "CPU" not in torch._C._dispatch_dump("torchao::float8_linear_cpu"),
+        reason="cpp kernels not built",
+    )
+    @common_utils.parametrize("dtype", [torch.float32, torch.bfloat16, torch.float16])
+    def test_module_path(self, dtype):
+        linear = torch.nn.Linear(128, 256, dtype=dtype)
+        quantize_(linear, get_config(PerRow()))
+        self.assertEqual(
+            str(type(linear.weight)),
+            "<class 'torchao.quantization.Float8OpaqueTensor'>",
+        )
+
+        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.Float8OpaqueTensor'>",
+            )
+
+
+common_utils.instantiate_parametrized_tests(TestFloat8OpaqueTensor)
+
+
+if __name__ == "__main__":
+    run_tests()

torchao/float8/inference.py

@@ -14,6 +14,7 @@
 from torchao.float8.float8_utils import is_row_major, pad_tensor_for_matmul
 from torchao.float8.types import FP8Granularity
 from torchao.quantization.granularity import (
+    PerGroup,
     PerRow,
     PerTensor,
 )
@@ -204,28 +205,41 @@ def _normalize_granularity(
             list[FP8Granularity],
         ]
     ],
+    supported_granularities: tuple[FP8Granularity] = (PerTensor, PerRow),
+    support_different_granularities: bool = False,
 ) -> Tuple[FP8Granularity, FP8Granularity]:
     processed_granularity = None
     if granularity is None:
         processed_granularity = (PerTensor(), PerTensor())
-    elif isinstance(granularity, (PerTensor, PerRow)):
+    elif isinstance(granularity, supported_granularities):
         processed_granularity = (granularity, granularity)
     elif isinstance(granularity, (tuple, list)) and len(granularity) == 2:
         if not (
-            isinstance(granularity[0], (PerTensor, PerRow))
-            and isinstance(granularity[1], (PerTensor, PerRow))
+            isinstance(granularity[0], supported_granularities)
+            and isinstance(granularity[1], supported_granularities)
         ):
             raise ValueError(
-                f"Invalid granularity types: {granularity}, only PerTensor or PerRow are supported."
+                f"Invalid granularity types: {granularity}, only {supported_granularities} are supported."
             )
-        if not isinstance(granularity[0], type(granularity[1])):
+        if not support_different_granularities and not isinstance(
+            granularity[0], type(granularity[1])
+        ):
             raise ValueError(
-                f"Different granularities for activation and weight are not supported: {granularity}, only PerTensor or PerRow are supported."
+                f"Different granularities for activation and weight are not supported: {granularity}, only {supported_granularities} are supported."
             )
+        if isinstance(granularity[0], PerGroup):
+            if not isinstance(granularity[1], PerGroup):
+                raise ValueError(
+                    "When granularity for activation is PerGroup, granularity for weight must be PerGroup, too."
+                )
+            if granularity[0].group_size != granularity[1].group_size:
+                raise ValueError(
+                    f"Group sizes for activation and weight must be the same, got {granularity[0].group_size} and {granularity[1].group_size}."
+                )
         processed_granularity = tuple(granularity)
     else:
         raise ValueError(
-            f"Invalid granularity specification: {granularity}, only PerTensor or PerRow are supported."
+            f"Invalid granularity specification: {granularity}, only {supported_granularities} are supported."
         )
     return processed_granularity
 

torchao/float8/types.py

@@ -12,8 +12,8 @@
 from typing import TYPE_CHECKING, Union
 
 if TYPE_CHECKING:
-    from torchao.quantization.granularity import PerRow, PerTensor
+    from torchao.quantization.granularity import PerGroup, PerRow, PerTensor
 
 
 # Define FP8Granularity type alias to break circular import dependencies
-FP8Granularity = Union["PerTensor", "PerRow"]
+FP8Granularity = Union["PerTensor", "PerRow", "PerGroup"]

torchao/quantization/__init__.py

@@ -78,6 +78,7 @@
     quantize_affine,
 )
 from .quantize_.workflows import (
+    Float8OpaqueTensor,
     Float8Tensor,
     Int4MarlinSparseTensor,
     Int4OpaqueTensor,
@@ -148,6 +149,7 @@
     "Int4TilePackedTo4dTensor",
     "Float8Tensor",
     "Int4OpaqueTensor",
+    "Float8OpaqueTensor",
     # smooth quant - subject to change
     "get_scale",
     "SmoothFakeDynQuantMixin",

torchao/quantization/quant_api.py

@@ -72,6 +72,8 @@
     KernelPreference,
 )
 from torchao.quantization.quantize_.workflows import (
+    Float8OpaqueTensor,
+    Float8PackingFormat,
     Float8Tensor,
     Int4ChooseQParamsAlgorithm,
     Int4MarlinSparseTensor,
@@ -1676,6 +1678,22 @@ def _input_activation_quant_func_fp8(
     return activation
 
 
+def _input_activation_quant_cpu_fp8(
+    x: torch.Tensor,
+    activation_granularity: FP8Granularity,
+    activation_dtype: torch.dtype,
+):
+    """Dynamic quantize activation to fp8 for CPU."""
+    block_size = get_block_size(x.shape, activation_granularity)
+    return to_affine_quantized_floatx(
+        input_float=x,
+        block_size=block_size,
+        target_dtype=activation_dtype,
+        scale_dtype=torch.float32,
+        _layout=PlainLayout(),
+    )
+
+
 def _fp8_mm_compat(weight: torch.Tensor) -> bool:
     """
     Check if a weight tensor meets float8 quantization requirements.
@@ -1734,15 +1752,24 @@ class Float8DynamicActivationFloat8WeightConfig(AOBaseConfig):
     kernel_preference: KernelPreference = KernelPreference.AUTO
     set_inductor_config: bool = True
     version: int = 2
+    float8_packing_format: Float8PackingFormat = Float8PackingFormat.PLAIN
 
     def __post_init__(self):
         torch._C._log_api_usage_once(
             "torchao.quantization.Float8DynamicActivationFloat8WeightConfig"
         )
         if self.mm_config is None:
             self.mm_config = Float8MMConfig(use_fast_accum=True)
+        supported_granularities = ()
+        if self.float8_packing_format == Float8PackingFormat.PLAIN:
+            supported_granularities = (PerTensor, PerRow)
+        elif self.float8_packing_format == Float8PackingFormat.OPAQUE:
+            supported_granularities = (PerTensor, PerRow, PerGroup)
+        support_different_granularities = (
+            self.float8_packing_format == Float8PackingFormat.OPAQUE
+        )
         activation_granularity, weight_granularity = _normalize_granularity(
-            self.granularity
+            self.granularity, supported_granularities, support_different_granularities
         )
         self.granularity = [activation_granularity, weight_granularity]
 
@@ -1755,17 +1782,12 @@ def _float8_dynamic_activation_float8_weight_quantize_tensor(weight, config):
     activation_value_lb = config.activation_value_lb
     activation_value_ub = config.activation_value_ub
     kernel_preference = config.kernel_preference
+    float8_packing_format = config.float8_packing_format
 
     # Ensure works on device
-    _check_hardware_support(granularity)
     activation_granularity, weight_granularity = granularity
 
-    if not _fp8_mm_compat(weight):
-        # TODO(future PR): this should really throw an exception instead of silently
-        # not doing what the user asked
-        return weight
-
-    if isinstance(weight_granularity, PerRow):
+    if weight.device.type != "cpu" and isinstance(weight_granularity, PerRow):
         assert weight.dtype == torch.bfloat16, (
             "PerRow quantization only works for bfloat16 precision input weight"
         )
@@ -1775,6 +1797,12 @@ def _float8_dynamic_activation_float8_weight_quantize_tensor(weight, config):
             "Config Deprecation: version 1 of Float8DynamicActivationFloat8WeightConfig is deprecated and will no longer be supported in a future release, please use version 2, see https://github.com/pytorch/ao/issues/2649 for more details"
         )
 
+        _check_hardware_support(granularity)
+        if not _fp8_mm_compat(weight):
+            # TODO(future PR): this should really throw an exception instead of silently
+            # not doing what the user asked
+            return weight
+
         block_size = get_block_size(weight.shape[-2:], weight_granularity)
         if weight.dim() == 3:
             block_size = tuple([1] + list(block_size))
@@ -1805,14 +1833,26 @@ def _float8_dynamic_activation_float8_weight_quantize_tensor(weight, config):
             kernel_preference=kernel_preference,
         )
 
-        quantized_weight = Float8Tensor.from_hp(
-            weight,
-            float8_dtype=weight_dtype,
-            granularity=weight_granularity,
-            mm_config=mm_config,
-            kernel_preference=kernel_preference,
-            act_quant_kwargs=act_quant_kwargs,
-        )
+        if float8_packing_format == Float8PackingFormat.PLAIN:
+            quantized_weight = Float8Tensor.from_hp(
+                weight,
+                float8_dtype=weight_dtype,
+                granularity=weight_granularity,
+                mm_config=mm_config,
+                kernel_preference=kernel_preference,
+                act_quant_kwargs=act_quant_kwargs,
+            )
+        elif float8_packing_format == Float8PackingFormat.OPAQUE:
+            block_size = get_block_size(weight.shape, weight_granularity)
+            quantized_weight = Float8OpaqueTensor.from_hp(
+                weight,
+                block_size=block_size,
+                act_quant_kwargs=act_quant_kwargs,
+            )
+        else:
+            raise ValueError(
+                f"Unsupported float8 packing format: {float8_packing_format}"
+            )
 
     return quantized_weight
 
@@ -1821,12 +1861,6 @@ def _float8_dynamic_activation_float8_weight_quantize_tensor(weight, config):
 def _float8_dynamic_activation_float8_weight_transform(
     module: torch.nn.Module, config: Float8DynamicActivationFloat8WeightConfig
 ):
-    assert is_sm_at_least_89() or is_MI300(), (
-        "Float8 dynamic activation quantization is only supported on CUDA>=8.9 and MI300+"
-    )
-    if config.set_inductor_config:
-        torchao.quantization.utils.recommended_inductor_config_setter()
-
     assert hasattr(module, "weight"), (
         "applying float8 dynamic activation quant requires module to have weight attribute"
         + f"but {module} does not have one"