wip

Author: jcaip

test/sparsity/test_sparse_api.py

@@ -7,6 +7,7 @@
 
 from torchao.sparsity import (
     apply_fake_sparsity,
+    apply_fake_block_sparsity,
     sparsify_,
     semi_sparse_weight,
 )
@@ -96,5 +97,67 @@ def test_sparse_marlin(self):
 
         assert torch.allclose(dense_result, sparse_result, atol=3e-1), "Results are not close"
 
+class TestBlockSparseWeight(TestCase):
+    @unittest.skipIf(not TORCH_VERSION_AT_LEAST_2_3, "pytorch 2.3+ feature")
+    @unittest.skipIf(not torch.cuda.is_available(), "Need CUDA available")
+    def test_sparse(self):
+        input = torch.rand((1024, 1024)).half().cuda()
+        model = (
+            nn.Sequential(
+                nn.Linear(1024, 2048),
+                nn.Linear(2048, 1024),
+            )
+            .half()
+            .cuda()
+        )
+
+        from torchao.sparsity.utils import create_block_sparse_tensor
+        M, N = model[0].weight.shape
+        model[0].weight.data = create_block_sparse_tensor(M, N, 64, 0.5, torch.float16)
+        M, N = model[1].weight.shape
+        model[1].weight.data = create_block_sparse_tensor(M, N, 64, 0.5, torch.float16)
+        dense_result = model(input)
+
+        from torchao.sparsity.prototype.superblock.blocksparse import block_sparse_weight
+        sparsify_(model, block_sparse_weight())
+        sparse_result = model(input)
+
+        assert torch.allclose(dense_result, sparse_result, rtol=1e-3, atol=1e-3)
+
+class TestQuantBlockSparseWeight(TestCase):
+    @unittest.skipIf(not TORCH_VERSION_AT_LEAST_2_3, "pytorch 2.3+ feature")
+    @unittest.skipIf(not torch.cuda.is_available(), "Need CUDA available")
+    def test_sparse(self):
+        input = torch.rand((128, 128)).to(torch.bfloat16).cuda()
+        model = (
+            nn.Sequential(
+                nn.Linear(128, 256),
+                nn.Linear(256, 128),
+            )
+            .to(torch.bfloat16)
+            .cuda()
+        )
+
+        from torchao.sparsity.utils import create_block_sparse_tensor
+        M, N = model[0].weight.shape
+        model[0].weight.data = create_block_sparse_tensor(M, N, 64, 0.5, torch.bfloat16) * torch.rand(M, N, dtype=torch.bfloat16).cuda()
+        print(model[0].weight)
+        M, N = model[1].weight.shape
+        model[1].weight.data = create_block_sparse_tensor(M, N, 64, 0.5, torch.bfloat16)
+        print(model[1].weight)  
+
+        model_copy = copy.deepcopy(model)
+
+        quantize_(model_copy, int8_dynamic_activation_int8_weight())
+        reference = model_copy(input)
+
+        from torchao.dtypes.affine_quantized_tensor import BlockSparseLayoutType
+        quantize_(model, int8_dynamic_activation_int8_weight(layout_type=BlockSparseLayoutType(), ))
+        sparse_result = model(input)
+
+        print(reference)
+        print(sparse_result)
+        assert torch.allclose(reference, sparse_result, rtol=1e-2, atol=1e-2)
+
 if __name__ == "__main__":
     unittest.main()

torchao/dtypes/affine_quantized_tensor.py

@@ -47,7 +47,6 @@
 from torchao.float8.inference import Float8MMConfig
 aten = torch.ops.aten
 
-
 ###############################
 # Base Layout Tensor Subclass #
 ###############################
@@ -472,6 +471,13 @@ def pre_process(self, input: torch.Tensor) -> torch.Tensor:
         temp.view(-1, 4).scatter_(1, pruning_inds, value=0)
         return temp
 
+@dataclass(frozen=True)
+class BlockSparseLayoutType(LayoutType):
+
+    def pre_process(self, input: torch.Tensor) -> torch.Tensor:
+        return input
+
+
 
 @dataclass(frozen=True)
 class TensorCoreTiledLayoutType(LayoutType):
@@ -669,6 +675,162 @@ def from_plain(
         int_data_compressed = torch._cslt_compress(int_data)
         return cls(int_data_compressed, scale, zero_point, layout_type)
 
+@register_layout_cls(BlockSparseLayoutType)
+class BlockSparseAQTLayout(PlainAQTLayout):
+    quantized_linear_impl = "block"
+    bsr_crow_indices: Optional[torch.Tensor]
+    bsr_col_indices: Optional[torch.Tensor]
+    bsr_values: Optional[torch.Tensor]
+    scale: Optional[torch.Tensor]
+    zero_point: Optional[torch.Tensor]
+
+    __slots__ = ["bsr_crow_indices", "bsr_col_indices", "bsr_values", "scale", "zero_point"] 
+
+    @staticmethod
+    def __new__(  # noqa: PYI034
+        cls,
+        shape: torch.Size,
+        bsr_crow_indices: Optional[torch.Tensor],
+        bsr_col_indices: Optional[torch.Tensor],
+        bsr_values: Optional[torch.Tensor],
+        scale: Optional[torch.Tensor], 
+        zero_point: Optional[torch.Tensor],
+        layout_type: LayoutType,
+        requires_grad: bool = False,
+    ):
+        if bsr_values is None:
+            raise ValueError("bsr values must be provided!")
+        else:
+            previous_tensor = bsr_values
+
+        kwargs = {
+            "device": previous_tensor.device,
+            "dtype": previous_tensor.dtype,
+            "layout": previous_tensor.layout,
+            "requires_grad": requires_grad,
+        }
+        return torch.Tensor._make_wrapper_subclass(cls, shape, **kwargs)  # type: ignore[attr-defined]
+
+    def __init__(  # noqa: PYI034
+        self,
+        shape: torch.Size,
+        bsr_crow_indices: Optional[torch.Tensor],
+        bsr_col_indices: Optional[torch.Tensor],
+        bsr_values: Optional[torch.Tensor],
+        scale: Optional[torch.Tensor], 
+        zero_point: Optional[torch.Tensor],
+        layout_type: LayoutType,
+        requires_grad: bool = False,
+    ):
+        self.bsr_crow_indices = bsr_crow_indices
+        self.bsr_col_indices = bsr_col_indices
+        self.bsr_values = bsr_values
+        self.scale = scale
+        self.zero_point = zero_point
+        self.layout_type = layout_type
+
+    def __repr__(self) -> str:  # type: ignore[override]
+        assert hasattr(self, "shape")
+        return f"{self.__class__.__name__}(shape={self.shape})"
+
+    def __tensor_flatten__(self): 
+        inner_tensors = list(
+            filter(lambda x: getattr(self, x) is not None, self.__slots__)
+        )
+        tensor_meta = (self.shape, self.layout_type, self.requires_grad)
+        return inner_tensors, tensor_meta
+
+    @classmethod
+    def __tensor_unflatten__(
+        cls,
+        inner_tensors,
+        tensor_meta: Tuple[torch.Size, bool],
+        outer_size,
+        outer_stride,
+    ) -> torch.Tensor:
+        shape, layout_type, requires_grad = tensor_meta
+        return cls(
+            shape=shape,
+            bsr_crow_indices=inner_tensors.get("bsr_crow_indices", None),
+            bsr_col_indices=inner_tensors.get("bsr_col_indices", None),
+            bsr_values=inner_tensors.get("bsr_values", None),
+            scale=inner_tensors.get("scale", None),
+            zero_point=inner_tensors.get("zero_point", None),
+            layout_type=layout_type,
+            requires_grad=requires_grad,
+        )
+
+    @classmethod
+    def from_plain(cls, int_data, scale, zero_point, layout_type):
+        bsr_tensor = int_data.to_sparse_bsr(64)
+        return cls(
+            shape=int_data.shape,
+            bsr_crow_indices=bsr_tensor.crow_indices(),
+            bsr_col_indices=bsr_tensor.col_indices(),
+            bsr_values=bsr_tensor.values(),
+            scale=scale, 
+            zero_point=zero_point,
+            layout_type = layout_type,
+            requires_grad=False,
+        )
+
+    @torch._dynamo.disable
+    def get_plain(self):
+        # asdf = torch.eye(self.shape[1]).to(self.device)
+        # self_bsr = torch.sparse_bsr_tensor(
+        #             self.crow_indices().to(self.device),
+        #             self.col_indices().to(self.device),
+        #             self.values().to(self.device),
+        #             size=(self.shape[0], self.shape[1])).to(self.dtype)
+        # int_data_bsr = bsr_dense_mm(self_bsr, asdf)
+        return torch.zeros(self.shape, device=self.device).to(self.dtype), self.scale, self.zero_point
+
+    def _apply_fn_to_data(self, func):
+        return self.__class__(
+            shape = self.shape,
+            bsr_crow_indices=func(self.bsr_crow_indices),
+            bsr_col_indices=func(self.bsr_col_indices),
+            bsr_values=func(self.bsr_values),
+            scale=self.scale,
+            zero_point=self.zero_point,
+            layout_type=self.layout_type,
+            requires_grad=self.requires_grad,
+        )
+
+    @classmethod
+    def __torch_dispatch__(cls, func, types, args, kwargs):
+        kwargs = {} if kwargs is None else kwargs
+
+        if func is aten.detach.default:
+            return return_and_correct_aliasing(
+                func, args, kwargs, args[0]._apply_fn_to_data(torch.detach)
+            )
+        if func is aten.clone.default:
+            return return_and_correct_aliasing(
+                func, args, kwargs, args[0]._apply_fn_to_data(torch.clone)
+            )
+        if func is aten.t.default:
+            """we don't need to repack the weight and just rely on external
+            shape being changed and record the status of transpose/no-transpose
+            """
+            args[0].transposed = not args[0].transposed
+            return return_and_correct_aliasing(func, args, kwargs, args[0])
+
+        if func is aten.crow_indices.default:
+            return args[0].bsr_crow_indices.detach()
+
+        if func is aten.col_indices.default:
+            return args[0].bsr_col_indices.detach()
+
+        if func is aten.values.default:
+            return args[0].bsr_values.detach()
+
+        if func is aten._nnz.default:
+            return args[0].bsr_values.shape[0]  
+
+        raise NotImplementedError(
+            f"BlockSparseAQTLayout dispatch: attempting to run {func}, this is not supported"
+        )
 
 @register_layout_cls(MarlinSparseLayoutType)
 class MarlinSparseAQTLayout(AQTLayout):
@@ -1221,6 +1383,42 @@ def _linear_int8_act_int8_weight_semi_structured_sparse_impl(input_tensor, weigh
         y += bias
     return y
 
+def _linear_int8_act_int8_weight_block_sparse_check(input_tensor, weight_tensor, bias):
+    return (
+        isinstance(input_tensor, AffineQuantizedTensor) and
+        _aqt_is_int8_reduced_range(input_tensor) and
+        isinstance(weight_tensor, AffineQuantizedTensor) and
+        weight_tensor.is_cuda and
+        input_tensor.dtype == weight_tensor.dtype and
+        isinstance(input_tensor.layout_type, PlainLayoutType) and
+        isinstance(weight_tensor.layout_type, BlockSparseLayoutType) and
+        weight_tensor.layout_tensor.quantized_linear_impl == "block"
+    )
+
+
+def _linear_int8_act_int8_weight_block_sparse_impl(input_tensor, weight_tensor, bias):
+    x_vals_int8 = input_tensor.layout_tensor.int_data
+    x_scales = input_tensor.layout_tensor.scale
+    w_vals = weight_tensor.layout_tensor
+    w_scales = weight_tensor.layout_tensor.scale
+    tmp = x_vals_int8.reshape(-1, x_vals_int8.shape[-1])
+    tmp_t = tmp.t()
+
+    y = torch.ops.blocksparse.int_addmm(w_vals.crow_indices(),
+                                        w_vals.col_indices(),
+                                        w_vals.values(),
+                                        tmp_t,
+                                        w_scales,
+                                        x_scales.reshape(-1))
+    y_shape = (*x_vals_int8.shape[:-1], w_scales.shape[-1])
+    y = y.reshape(*y_shape)
+
+    # can downcast only at the very end
+    output_dtype = input_tensor.dtype
+    y = y.to(output_dtype)
+    if bias is not None:
+        y += bias
+    return y
 def _linear_bf16_act_uint4_weight_check(input_tensor, weight_tensor, bias):
     return (
         # input is native bfloat16 tensor
@@ -1473,6 +1671,7 @@ def _register_aqt_quantized_linear_dispatches():
     for dispatch_condition, impl in [
         (_linear_int8_act_int8_weight_check, _linear_int8_act_int8_weight_impl),
         (_linear_int8_act_int8_weight_semi_structured_sparse_check, _linear_int8_act_int8_weight_semi_structured_sparse_impl),
+        (_linear_int8_act_int8_weight_block_sparse_check, _linear_int8_act_int8_weight_block_sparse_impl),
         (_linear_fp8_act_fp8_weight_check, _linear_fp8_act_fp8_weight_impl),
         (_linear_bf16_act_uint4_weight_check, _linear_bf16_act_uint4_weight_impl),
         (_linear_fp_act_int8_weight_check, _linear_fp_act_int8_weight_impl),

torchao/sparsity/__init__.py

@@ -8,6 +8,7 @@
 from .utils import PerChannelNormObserver  # noqa: F403
 from .sparse_api import (
     apply_fake_sparsity,
+    apply_fake_block_sparsity,
     sparsify_,
     semi_sparse_weight,
     int8_dynamic_activation_int8_semi_sparse_weight
@@ -17,6 +18,7 @@
     "WandaSparsifier",
     "PerChannelNormObserver",
     "apply_fake_sparsity",
+    "apply_fake_block_sparsity",
     "sparsify_"
     "semi_sparse_weight",
     "int8_dynamic_activation_int8_semi_sparse_weight"

torchao/sparsity/prototype/superblock/benchmark.py

@@ -13,6 +13,7 @@
 import utils
 from torch import nn
 from torch.sparse._triton_ops_meta import optimize_bsr_dense_addmm
+from torch.sparse._triton_ops_meta import dump as store_tuned_kernel_params
 from torchao.sparsity.prototype.superblock.utils import accelerate_with_sparsity, simulate_sparsity
 from torchao.utils import benchmark_model, profiler_runner
 
@@ -34,15 +35,30 @@ def main(args):
     # BSR kernel tuning
     if args.bsr and args.tune_kernel_params:
         print("Tuning kernel params")
+        kwargs = dict(
+            dtype=torch.int8 if args.quantization else dtype,
+            sparsity=args.sparsity_linear, verbose=True,
+            # per blocksparse_int_addmm:
+            alpha=1, beta=0, use_left_alpha=True, use_right_alpha=True,
+            # force tuning because existing tuning parameters are
+            # computed for use_left/right_alpha=False, however, it
+            # turns out that re-tuning for use_left/right_alpha=False
+            # leads to the same set of tuning parametes:
+            # force=True
+        )
         if args.model == "vit_b_16":
-            optimize_bsr_dense_addmm(3072, 768, 50432, args.bsr, args.bsr, dtype=dtype, sparsity=args.sparsity_linear, verbose=True)
-            optimize_bsr_dense_addmm(768, 3072, 50432, args.bsr, args.bsr, dtype=dtype, sparsity=args.sparsity_linear, verbose=True)
+            optimize_bsr_dense_addmm(3072, 768, 50432, args.bsr, args.bsr, **kwargs)
+            optimize_bsr_dense_addmm(768, 3072, 50432, args.bsr, args.bsr, **kwargs)
         elif args.model == "vit_h_14":
-            optimize_bsr_dense_addmm(5120, 1280, 65792, args.bsr, args.bsr, dtype=dtype, sparsity=args.sparsity_linear, verbose=True)
-            optimize_bsr_dense_addmm(1280, 5120, 65792, args.bsr, args.bsr, dtype=dtype, sparsity=args.sparsity_linear, verbose=True)
+            optimize_bsr_dense_addmm(5120, 1280, 65792, args.bsr, args.bsr, **kwargs)
+            optimize_bsr_dense_addmm(1280, 5120, 65792, args.bsr, args.bsr, **kwargs)
         else:
             raise NotImplementedError("Tuning kernel params for this model is not supported yet.")
-
+        # Warning: the following call will overwrite the source code
+        # of torch.sparse._triton_ops_meta (hence it is commented out
+        # by default) but when used, it'll enables reusing the tuned
+        # parameters in subsequent runs of this script:
+        # store_tuned_kernel_params()
     print("Creating model")
     model = torchvision.models.get_model(args.model, weights=args.weights, num_classes=num_classes)