[mlir][linalg] Add linalg.conv_2d_ngchw_gfchw_q to named ops #92136
Conversation
zjgarvey
requested review from
ftynse,
makslevental and
stellaraccident
as code owners
|
Thank you for submitting a Pull Request (PR) to the LLVM Project! This PR will be automatically labeled and the relevant teams will be If you wish to, you can add reviewers by using the "Reviewers" section on this page. If this is not working for you, it is probably because you do not have write If you have received no comments on your PR for a week, you can request a review If you have further questions, they may be answered by the LLVM GitHub User Guide. You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums. |
|
@llvm/pr-subscribers-mlir-linalg @llvm/pr-subscribers-mlir Author: None (zjgarvey) ChangesI'm not sure what kinds of unit tests would be appropriate for simply adding a new op like this. Any suggestions? Full diff: https://github.com/llvm/llvm-project/pull/92136.diff 2 Files Affected:
diff --git a/mlir/include/mlir/Dialect/Linalg/IR/LinalgNamedStructuredOps.yaml b/mlir/include/mlir/Dialect/Linalg/IR/LinalgNamedStructuredOps.yaml
index 584bfcd8b59dc..98f20809a60fa 100644
--- a/mlir/include/mlir/Dialect/Linalg/IR/LinalgNamedStructuredOps.yaml
+++ b/mlir/include/mlir/Dialect/Linalg/IR/LinalgNamedStructuredOps.yaml
@@ -304,41 +304,6 @@ structured_op: !LinalgStructuredOpConfig
- !ScalarExpression
scalar_arg: I
--- !LinalgOpConfig
-metadata: !LinalgOpMetadata
- name: reciprocal
- cpp_class_name: ReciprocalOp
- doc: |-
- Applies reciprocal(x) elementwise.
-
- No numeric casting is performed on the input operand.
-structured_op: !LinalgStructuredOpConfig
- args:
- - !LinalgOperandDefConfig
- name: I
- kind: input_tensor
- type_var: T1
- shape_map: affine_map<() -> ()>
- - !LinalgOperandDefConfig
- name: O
- kind: output_tensor
- type_var: T1
- shape_map: affine_map<() -> ()>
- indexing_maps: !LinalgIndexingMapsConfig
- static_indexing_maps:
- - affine_map<() -> ()>
- - affine_map<() -> ()>
- iterator_types: []
- assignments:
- - !ScalarAssign
- arg: O
- value: !ScalarExpression
- scalar_fn:
- kind: unary
- fn_name: reciprocal
- operands:
- - !ScalarExpression
- scalar_arg: I
---- !LinalgOpConfig
metadata: !LinalgOpMetadata
name: round
cpp_class_name: RoundOp
@@ -516,7 +481,7 @@ structured_op: !LinalgStructuredOpConfig
--- !LinalgOpConfig
metadata: !LinalgOpMetadata
name: erf
- cpp_class_name: erfOp
+ cpp_class_name: ErfOp
doc: |-
Applies erf(x) elementwise.
@@ -959,7 +924,7 @@ structured_op: !LinalgStructuredOpConfig
--- !LinalgOpConfig
metadata: !LinalgOpMetadata
name: powf
- cpp_class_name: PowFOp
+ cpp_class_name: PowfOp
doc: |-
Takes the powf(lhs, rhs) between two inputs, elementwise. For powf(arg, 2) use `linalg.square`.
@@ -3421,6 +3386,144 @@ structured_op: !LinalgStructuredOpConfig
- !ScalarExpression
scalar_arg: K
--- !LinalgOpConfig
+metadata: !LinalgOpMetadata
+ name: conv_2d_ngchw_gfchw_q
+ cpp_class_name: Conv2DNgchwGfchwQOp
+ doc: |-
+ Performs 2-D grouped convolution with zero-point offsets.
+
+ Layout:
+ * Input: NGCHW.
+ * Kernel: GFCHW.
+
+ Numeric casting is performed on the operands to the inner multiply, promoting
+ them to the same data type as the accumulator/output. This includes the zero
+ point offsets common to quantized operations.
+ implements:
+ - LinalgConvolutionOpInterface
+structured_op: !LinalgStructuredOpConfig
+ args:
+ - !LinalgOperandDefConfig
+ name: I
+ kind: input_tensor
+ type_var: T1
+ shape_map: affine_map<()[s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11] ->
+ (s0, s1, s2, s3 * s4 + s5 * s6, s7 * s8 + s9 * s10)>
+ - !LinalgOperandDefConfig
+ name: K
+ kind: input_tensor
+ type_var: T2
+ shape_map: affine_map<()[s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11] ->
+ (s1, s11, s2, s5, s9)>
+ - !LinalgOperandDefConfig
+ name: IZp
+ kind: scalar
+ type_var: I32
+ - !LinalgOperandDefConfig
+ name: KZp
+ kind: scalar
+ type_var: I32
+ - !LinalgOperandDefConfig
+ name: O
+ kind: output_tensor
+ type_var: U
+ shape_map: affine_map<()[s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11] ->
+ (s0, s1, s11, s3, s7)>
+ - !LinalgOperandDefConfig
+ name: strides
+ kind: index_attr
+ index_attr_map: affine_map<()[s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11]
+ -> (s4, s8)>
+ default_indices:
+ - 1
+ - 1
+ - !LinalgOperandDefConfig
+ name: dilations
+ kind: index_attr
+ index_attr_map: affine_map<()[s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11]
+ -> (s6, s10)>
+ default_indices:
+ - 1
+ - 1
+ indexing_maps: !LinalgIndexingMapsConfig
+ static_indexing_maps:
+ - affine_map<(d0, d1, d2, d3, d4, d5, d6, d7)[s0, s1, s2, s3, s4, s5, s6, s7,
+ s8, s9, s10, s11] -> (d0, d1, d5, d3 * s4 + d6 * s6, d4 * s8 + d7 * s10)>
+ - affine_map<(d0, d1, d2, d3, d4, d5, d6, d7)[s0, s1, s2, s3, s4, s5, s6, s7,
+ s8, s9, s10, s11] -> (d1, d2, d5, d6, d7)>
+ - affine_map<(d0, d1, d2, d3, d4, d5, d6, d7)[s0, s1, s2, s3, s4, s5, s6, s7,
+ s8, s9, s10, s11] -> ()>
+ - affine_map<(d0, d1, d2, d3, d4, d5, d6, d7)[s0, s1, s2, s3, s4, s5, s6, s7,
+ s8, s9, s10, s11] -> ()>
+ - affine_map<(d0, d1, d2, d3, d4, d5, d6, d7)[s0, s1, s2, s3, s4, s5, s6, s7,
+ s8, s9, s10, s11] -> (d0, d1, d2, d3, d4)>
+ iterator_types:
+ - parallel
+ - parallel
+ - parallel
+ - parallel
+ - parallel
+ - reduction
+ - reduction
+ - reduction
+ assignments:
+ - !ScalarAssign
+ arg: O
+ value: !ScalarExpression
+ scalar_fn:
+ kind: binary
+ fn_name: add
+ operands:
+ - !ScalarExpression
+ scalar_arg: O
+ - !ScalarExpression
+ scalar_fn:
+ kind: binary
+ fn_name: mul
+ operands:
+ - !ScalarExpression
+ scalar_fn:
+ kind: binary
+ fn_name: sub
+ operands:
+ - !ScalarExpression
+ scalar_fn:
+ kind: type
+ fn_name: cast_signed
+ type_var: U
+ operands:
+ - !ScalarExpression
+ scalar_arg: I
+ - !ScalarExpression
+ scalar_fn:
+ kind: type
+ fn_name: cast_signed
+ type_var: U
+ operands:
+ - !ScalarExpression
+ scalar_arg: IZp
+ - !ScalarExpression
+ scalar_fn:
+ kind: binary
+ fn_name: sub
+ operands:
+ - !ScalarExpression
+ scalar_fn:
+ kind: type
+ fn_name: cast_signed
+ type_var: U
+ operands:
+ - !ScalarExpression
+ scalar_arg: K
+ - !ScalarExpression
+ scalar_fn:
+ kind: type
+ fn_name: cast_signed
+ type_var: U
+ operands:
+ - !ScalarExpression
+ scalar_arg: KZp
+--- !LinalgOpConfig
metadata: !LinalgOpMetadata
name: conv_3d_ndhwc_dhwcf
cpp_class_name: Conv3DNdhwcDhwcfOp
diff --git a/mlir/python/mlir/dialects/linalg/opdsl/ops/core_named_ops.py b/mlir/python/mlir/dialects/linalg/opdsl/ops/core_named_ops.py
index ca2bb0c5f7f8a..f1790b1fa2893 100644
--- a/mlir/python/mlir/dialects/linalg/opdsl/ops/core_named_ops.py
+++ b/mlir/python/mlir/dialects/linalg/opdsl/ops/core_named_ops.py
@@ -937,6 +937,40 @@ def conv_2d_ngchw_gfchw(
U, I[D.n, D.g, D.c, D.oh * S.SH + D.kh * S.DH, D.ow * S.SW + D.kw * S.DW]
) * TypeFn.cast_signed(U, K[D.g, D.fg, D.c, D.kh, D.kw])
+@linalg_structured_op
+def conv_2d_ngchw_gfchw_q(
+ I=TensorDef(
+ T1, S.N, S.G, S.C, S.OH * S.SH + S.KH * S.DH, S.OW * S.SW + S.KW * S.DW
+ ),
+ K=TensorDef(T2, S.G, S.FG, S.C, S.KH, S.KW),
+ IZp=ScalarDef(I32),
+ KZp=ScalarDef(I32),
+ O=TensorDef(U, S.N, S.G, S.FG, S.OH, S.OW, output=True),
+ strides=IndexAttrDef(S.SH, S.SW, default=[1, 1]),
+ dilations=IndexAttrDef(S.DH, S.DW, default=[1, 1]),
+):
+ """Performs 2-D grouped convolution with zero-point offsets.
+
+ Layout:
+ * Input: NGCHW.
+ * Kernel: GFCHW.
+
+ Numeric casting is performed on the operands to the inner multiply, promoting
+ them to the same data type as the accumulator/output. This includes the zero
+ point offsets common to quantized operations.
+ """
+ implements(ConvolutionOpInterface)
+ domain(D.n, D.g, D.fg, D.oh, D.ow, D.c, D.kh, D.kw)
+ O[D.n, D.g, D.fg, D.oh, D.ow] += (
+ TypeFn.cast_signed(
+ U, I[D.n, D.g, D.c, D.oh * S.SH + D.kh * S.DH, D.ow * S.SW + D.kw * S.DW]
+ )
+ - TypeFn.cast_signed(U, IZp)
+ ) * (
+ TypeFn.cast_signed(U, K[D.g, D.fg, D.c, D.kh, D.kw])
+ - TypeFn.cast_signed(U, KZp)
+ )
+
@linalg_structured_op
def conv_3d_ndhwc_dhwcf(
|
zjgarvey
force-pushed
the
conv_2d_ngchw_gfchw_q
branch
from
fe5039d to
d5f918b
Compare
zjgarvey
force-pushed
the
conv_2d_ngchw_gfchw_q
branch
from
d5f918b to
1e7d870
Compare
|
For future reference, is bin/update_core_linalg_named_ops.sh supposed to regenerate the current yaml file exactly? If not, is the general practice to run this script and revert any unrelated diffs manually? |
It should generate the current file. If there are discrepancies, please consider sending a separate patch so we can examine what they are and advise.
A roundtripping test showing that the op exists can can be parsed and printed back, and a test that converts this op to a generic op to check the correctness of the model and guard against regression in the specification. Both are FileCheck tests. Also please note that the first message of the PR is being picked up as commit description and update it to something more meaningful. |
Will do.
I think I added the types of lit tests you were talking about. Please let me know if those look appropriate.
Changed, thanks! |
|
@nicolasvasilache mind taking a quick look to make sure you approve |
|
✅ With the latest revision this PR passed the Python code formatter. |
|
@ftynse I haven't had any luck getting in touch with Nico yet. Would you mind revisiting this PR and merging it if you think it is acceptable? I don't have merge permissions myself. |
|
Sorry just saw this will look tomorrow. |
Thanks! |
|
@zjgarvey Congratulations on having your first Pull Request (PR) merged into the LLVM Project! Your changes will be combined with recent changes from other authors, then tested Please check whether problems have been caused by your change specifically, as How to do this, and the rest of the post-merge process, is covered in detail here. If your change does cause a problem, it may be reverted, or you can revert it yourself. If you don't get any reports, no action is required from you. Your changes are working as expected, well done! |
This addresses 7 of the model failures I'm seeing in the test suite. See [Shark-Turbine issue #566](nod-ai/SHARK-ModelDev#566). Need the op ```linalg.conv_2d_ngchw_gfchw_q``` to be added upstream before merging this. See [llvm-project PR #92136 ](llvm/llvm-project#92136). A small additional expansion to operand quantization is included in this patch to address a model failure that occurs when unblocking the quantized group convolutions in one of these onnx models.
This addresses 7 of the model failures I'm seeing in the test suite. See [Shark-Turbine issue llvm#566](nod-ai/SHARK-ModelDev#566). Need the op ```linalg.conv_2d_ngchw_gfchw_q``` to be added upstream before merging this. See [llvm-project PR #92136 ](llvm/llvm-project#92136). A small additional expansion to operand quantization is included in this patch to address a model failure that occurs when unblocking the quantized group convolutions in one of these onnx models.
Adds a named op: linalg.conv_2d_ngchw_gfchw_q. This op is similar to linalg.conv_2d_ngchw_gfchw, but additionally incorporates zero point offset corrections.