[mxfp8 moe training] add compile support

benchmarks/prototype/moe_training/benchmark_scaled_grouped_mm_dq.py

@@ -28,6 +28,9 @@
 # Needed since changing args to function causes recompiles
 torch._dynamo.config.cache_size_limit = 1000
 
+# Workaround for https://github.com/pytorch/ao/pull/2990#issuecomment-3285762681
+torch._dynamo.config.automatic_dynamic_shapes = False
+
 
 @dataclass(frozen=True)
 class ExperimentConfig:

test/prototype/moe_training/test_training.py

@@ -34,10 +34,7 @@
 
 @pytest.mark.parametrize(
     "target_fqns",
-    [
-        ["experts"],
-        ["does.not.exist"],
-    ],
+    [["experts"], ["experts,shared_expert"], ["invalid.fqns"]],
 )
 @pytest.mark.parametrize("compile", [False, True])
 @pytest.mark.parametrize(

torchao/prototype/moe_training/kernels/mxfp8_blocked_scales.py

@@ -4,6 +4,7 @@
 import triton
 import triton.language as tl
 from torch import Tensor
+from torch.library import triton_op, wrap_triton
 
 from torchao.prototype.mx_formats.utils import to_blocked
 from torchao.utils import ceil_div
@@ -192,6 +193,7 @@ def compute_blocked_scale_offsets_for_K_groups(
     return group_sizes, starting_col_after_padding
 
 
+@triton_op("torchao::triton_mx_block_rearrange_2d_M_groups", mutates_args={})
 def triton_mx_block_rearrange_2d_M_groups(
     scales_tensor: torch.Tensor,
     input_group_end_offsets: torch.Tensor,
@@ -216,10 +218,10 @@ def triton_mx_block_rearrange_2d_M_groups(
         "Expected element size to be 1 byte (8 bits)"
     )
     rows, cols = scales_tensor.shape
-    num_groups = input_group_end_offsets.numel()
+    num_groups = input_group_end_offsets.shape[0]
 
     # Final offset is the total number of rows in the tensor
-    padded_rows = output_group_start_offsets[-1]
+    padded_rows = rows + num_groups * 128  # output_group_start_offsets[-1]
 
     num_col_blocks = ceil_div(cols, 4)
     padded_cols = num_col_blocks * 4
@@ -238,7 +240,7 @@ def triton_mx_block_rearrange_2d_M_groups(
         num_groups,
         num_col_blocks,
     )
-    triton_scale_swizzle_M_groups[grid](
+    wrap_triton(triton_scale_swizzle_M_groups)[grid](
         # Input scales
         scales_tensor.view(torch.uint8),
         scales_tensor.stride(0),
@@ -336,6 +338,7 @@ def triton_scale_swizzle_M_groups(
         current_start_row += BLOCK_ROWS
 
 
+@triton_op("torchao::triton_mx_block_rearrange_per_group_3d", mutates_args={})
 def triton_mx_block_rearrange_per_group_3d(scale_tensor: torch.Tensor) -> torch.Tensor:
     """
     Rearranges an E8M0 tensor scale to block-scaled swizzle format.
@@ -379,7 +382,7 @@ def triton_mx_block_rearrange_per_group_3d(scale_tensor: torch.Tensor) -> torch.
         num_col_blocks,
     )
 
-    triton_scale_swizzle_per_group_3d[grid](
+    wrap_triton(triton_scale_swizzle_per_group_3d)[grid](
         scale_tensor.view(torch.uint8),
         input_stride_dim0,
         input_stride_dim1,
@@ -454,6 +457,7 @@ def triton_scale_swizzle_per_group_3d(
     )
 
 
+@triton_op("torchao::triton_mx_block_rearrange_2d_K_groups", mutates_args={})
 def triton_mx_block_rearrange_2d_K_groups(
     scales_tensor: torch.Tensor,
     input_group_end_offsets: torch.Tensor,
@@ -479,12 +483,12 @@ def triton_mx_block_rearrange_2d_K_groups(
     )
     rows, cols = scales_tensor.shape
     # Calculate blocks needed
-    num_groups = input_group_end_offsets.numel()
+    num_groups = input_group_end_offsets.shape[0]
     num_row_blocks = ceil_div(rows, 128)
     padded_rows = num_row_blocks * 128
 
     # output_group_start_offsets always starts with 0 and ends with the total number of cols
-    padded_cols = output_group_start_offsets[-1]
+    padded_cols = cols + num_groups * 4  # output_group_start_offsets[-1]
     output = scales_tensor.new_empty((padded_rows, padded_cols))
 
     # Output block stride for the rearranged format
@@ -497,7 +501,7 @@ def triton_mx_block_rearrange_2d_K_groups(
         num_groups,
         num_row_blocks,
     )
-    triton_scale_swizzle_2d_K_groups[grid](
+    wrap_triton(triton_scale_swizzle_2d_K_groups)[grid](
         # Input scales
         scales_tensor.view(torch.uint8),
         scales_tensor.stride(0),

torchao/prototype/moe_training/scaled_grouped_mm.py

@@ -58,15 +58,15 @@ def _scaled_grouped_mm(
     """
     # TODO: Remove logging once prototype is more mature. This is currently very useful for development and debugging.
     if scaling_type == MoEScalingType.FP8_ROWWISE:
-        logger.info("Using fp8 rowwise for _scaled_grouped_mm")
+        logger.debug("Using fp8 rowwise for _scaled_grouped_mm")
         return _Float8GroupedMM.apply(
             A,
             B_t,
             offs,
             out_dtype,
         )
     elif scaling_type == MoEScalingType.MXFP8:
-        logger.info("Using mxfp8 for _scaled_grouped_mm")
+        logger.debug("Using mxfp8 for _scaled_grouped_mm")
         block_size = 32  # TODO: should we make this configurable? plumb it through in a config somehow?
         return _MXFP8GroupedMM.apply(
             A,

torchao/prototype/moe_training/tensor.py

@@ -27,7 +27,7 @@
     torch.ops.aten.copy_.default,
     torch.ops.aten.view.default,
     torch.ops.aten.as_strided.default,
-    torch.ops.aten._to_copy.default,
+    torch.ops.aten._to_copy.default,  # for *.to(dtype)
     torch.ops.aten._pin_memory.default,
     torch.ops.aten.split.Tensor,
     torch.ops.aten.clone.default,
@@ -94,11 +94,11 @@ def __torch_function__(cls, func, types, args, kwargs={}):
                 "B should be a ScaledGroupedMMTensor"
             )
             scaling_type = B.scaling_type
-            A_is_2d = A.dim() == 2
-            B_is_3d = B.dim() == 3
+            A_is_2d = A.ndim == 2
+            B_is_2d_or_3d = B.ndim == 2 or B.ndim == 3
             has_offs = kwargs.get(cls.offs_arg_name) is not None
             other_args = args[2:]
-            if A_is_2d and B_is_3d and has_offs:
+            if A_is_2d and B_is_2d_or_3d and has_offs:
                 return _scaled_grouped_mm(
                     A,
                     B,
@@ -125,28 +125,31 @@ def unwrap(t):
                 assert t.scaling_type == scaling_type
             return t._data
 
-        args, kwargs = pytree.tree_map_only(
+        args_unwrapped, kwargs_unwrapped = pytree.tree_map_only(
             ScaledGroupedMMTensor, unwrap, (args, kwargs or {})
         )
-        assert scaling_type is not None
+        assert scaling_type is not None, (
+            f"__torch_dispatch__ called on {func.__name__} without any ScaledGroupedMMTensor arguments"
+        )
 
         # detach is special case
         if func == torch.ops.aten.detach.default:
             return ScaledGroupedMMTensor(args[0], scaling_type)
 
         # perform op
-        out = func(*args, **kwargs)
+        out = func(*args_unwrapped, **kwargs_unwrapped)
 
         # return regular tensors for ops that don't preserve subclass
         if func not in _ops_to_preserve_subclass:
             return out
 
-        # wrap outputs back into ScaledGroupedMMTensor for ops that do preserve subclass
-        return pytree.tree_map_only(
-            torch.Tensor,
-            lambda x: ScaledGroupedMMTensor(x, scaling_type),
-            out,
-        )
+        # Only wrap tensor outputs, prevent double wrapping
+        def selective_wrap(x):
+            if isinstance(x, torch.Tensor) and type(x) == torch.Tensor:
+                return ScaledGroupedMMTensor(x, scaling_type)
+            return x
+
+        return pytree.tree_map(selective_wrap, out)
 
     def __repr__(self):
         return f"ScaledGroupedMMTensor(data={self._data}, scaling_type={self.scaling_type})"