Skip to content

Commit 39c9935

Browse files
mgehre-amdmgehre-amd
committed
[AutoBump] Merge with fixes of 9c70676 (Oct 18)
2 parents ec66c32 + 9c70676 commit 39c9935

File tree

3 files changed

+123
-92
lines changed

3 files changed

+123
-92
lines changed

‎projects/pt1/e2e_testing/xfail_sets.py‎

Lines changed: 45 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -627,10 +627,6 @@
627627
"ElementwiseRemainderTensorModule_Int_Float_NegativeDivisor_basic",
628628
"ElementwiseRemainderTensorModule_Int_NegativeDividend_basic",
629629
"ElementwiseRemainderTensorModule_Int_NegativeDivisor_basic",
630-
"ElementwiseRreluEvalModule_basic",
631-
"ElementwiseRreluEvalStaticModule_basic",
632-
"ElementwiseRreluTrainModule_basic",
633-
"ElementwiseRreluTrainStaticModule_basic",
634630
"MaxPool1dCeilModeTrueModule_basic",
635631
"MaxPool1dStaticCeilModeTrueModule_basic",
636632
"MaxUnpool3dModulePad0_basic",
@@ -664,7 +660,6 @@
664660
"AdaptiveAvgPool3dDynamic_basic",
665661
"AdaptiveMaxPool1dDynamicNoBatch_basic",
666662
"AdaptiveMaxPool1dDynamic_basic",
667-
"AdaptiveMaxPool1dDimOneStatic_basic",
668663
"AdaptiveMaxPool1dStatic_basic",
669664
"AdaptiveMaxPool2dDynamicNoBatch_basic",
670665
"AdaptiveMaxPool2dDynamicWithIndices_basic",
@@ -831,12 +826,7 @@
831826
"MaxPool2dWithIndicesBackwardStatic3DModule_basic",
832827
"MaxPool2dWithIndicesBackwardStatic4DModule_basic",
833828
"MaxPool3dCeilModeTrueModule_basic",
834-
"MaxPool3dEmptyStrideStaticModule_basic",
835-
"MaxPool3dLargeDatadModule_basic",
836-
"MaxPool3dModuleRandomSimple_basic",
837-
"MaxPool3dModule_basic",
838829
"MaxPool3dStaticCeilModeTrueModule_basic",
839-
"MaxPool3dStaticModule_basic",
840830
"MaxPool3dWithIndicesAllNegativeValuesModule_basic",
841831
"MaxPool3dWithIndicesAllOnesModule_basic",
842832
"MaxPool3dWithIndicesCeilModeTrueModule_basic",
@@ -994,6 +984,51 @@
994984
"Unfold_Module_Rank_Zero_basic",
995985
"Unfold_Module_Rank_Zero_Size_Zero_basic",
996986
"Unfold_Module_Dynamic_basic",
987+
"AdaptiveAvgPool1dGeneralDynamic_basic",
988+
"AdaptiveAvgPool1dGeneralDynamicNoBatches_basic",
989+
"AdaptiveAvgPool1dStaticEvenMultiple_basic",
990+
"AdaptiveAvgPool1dStaticLargerOutput_basic",
991+
"AdaptiveAvgPool2dDynamicNoBatch_basic",
992+
"AdaptiveAvgPool2dDynamic_basic",
993+
"AddIntModule_basic",
994+
"AtenIntTensorByteDtypeModule_basic",
995+
"AtenIntTensorCharDtypeModule_basic",
996+
"AtenItemIntOpModule_basic",
997+
"CrossEntropyLossModule_basic",
998+
"CrossEntropyLossNoReductionModule_basic",
999+
"EinsumStaticContractRhsModule_basic",
1000+
"EinsumStaticFourDimensionModule_basic",
1001+
"EinsumStaticModule_basic",
1002+
"EinsumStaticWithEllipsisSlicingAndBroadcastModule_basic",
1003+
"EinsumStaticWithEllipsisSlicingModule_basic",
1004+
"ElementwiseExpm1IntModule_basic",
1005+
"ElementwiseExpm1Module_basic",
1006+
"InterpolateDynamicModule_sizes_nearest",
1007+
"IouOfModule_basic",
1008+
"IscloseStaticModuleTrue_basic",
1009+
"IscloseStaticModule_basic",
1010+
"MeshgridIndexingIJ_basic",
1011+
"MeshgridIndexingXY_basic",
1012+
"Meshgrid_basic",
1013+
"MulIntModule_basic",
1014+
"OneHotModule_basic",
1015+
"ReduceFrobeniusNormComplexModule_basic",
1016+
"ScalarImplicitIntModule_basic",
1017+
"ScaledDotProductAttentionBoolMaskModule_basic",
1018+
"ScaledDotProductAttentionDifferentCausalModule_basic",
1019+
"ScaledDotProductAttentionDifferentDynamicCausalModule_basic",
1020+
"ScaledDotProductAttentionDifferentModule_basic",
1021+
"ScaledDotProductAttentionMaskModule_basic",
1022+
"ScaledDotProductAttentionSameCausalModule_basic",
1023+
"ScaledDotProductAttentionSameDynamicModule_basic",
1024+
"ScaledDotProductAttentionSameModule_basic",
1025+
"SubIntModule_basic",
1026+
"TensorToIntZeroRank_basic",
1027+
"UpSampleNearest2dDynamicFactor_basic",
1028+
"UpSampleNearest2dDynamicSize_basic",
1029+
"UpSampleNearest2dStaticFactor_basic",
1030+
"UpSampleNearest2dStaticSize_basic",
1031+
"UpSampleNearest2d_basic",
9971032
}
9981033

9991034
FX_IMPORTER_STABLEHLO_CRASHING_SET = {

‎pytorch-hash.txt‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1 +1 @@
1-
798d5b7ddd08899fb62672d56044dbf1f63a4d17
1+
ec8499a174317b85b6c6fe98eb99a266b590cef8

‎test/python/fx_importer/sparsity/sparse_test.py‎

Lines changed: 77 additions & 81 deletions
Original file line numberDiff line numberDiff line change
@@ -5,10 +5,6 @@
55

66
# RUN: %PYTHON %s | FileCheck %s
77

8-
# test_sparse_feature_scaling fails to lower, and we don't need sparsity
9-
# look into that later.
10-
# UNSUPPORTED: true
11-
128
from typing import Any, Callable, Optional, Tuple, Dict
139

1410
import torch
@@ -220,25 +216,25 @@ def forward(self, x, v):
220216
print("torch.mlir =", res2)
221217

222218

223-
@run
219+
# @run
224220
#
225-
# CHECK-LABEL: test_sparse_SpMM
226-
# CHECK: #[[$COO:.*]] = #sparse_tensor.encoding<{ map = (d0, d1) -> (d0 : compressed(nonunique), d1 : singleton(soa)), posWidth = 64, crdWidth = 64 }>
227-
# CHECK: func.func @main(
228-
# CHECK-SAME: %[[A:.*0]]: !torch.vtensor<[8,8],f32,#[[$COO]]>,
229-
# CHECK-SAME: %[[B:.*1]]: !torch.vtensor<[8,8],f32>) -> !torch.vtensor<[8,8],f32> {
230-
# CHECK: %[[R:.*]] = torch.aten.{{matmul|mm}} %[[A]], %[[B]] : !torch.vtensor<[8,8],f32,#[[$COO]]>, !torch.vtensor<[8,8],f32> -> !torch.vtensor<[8,8],f32>
231-
# CHECK: return %[[R]] : !torch.vtensor<[8,8],f32>
232-
# CHECK: }
221+
# C_HECK-LABEL: test_sparse_SpMM
222+
# C_HECK: #[[$COO:.*]] = #sparse_tensor.encoding<{ map = (d0, d1) -> (d0 : compressed(nonunique), d1 : singleton(soa)), posWidth = 64, crdWidth = 64 }>
223+
# C_HECK: func.func @main(
224+
# C_HECK-SAME: %[[A:.*0]]: !torch.vtensor<[8,8],f32,#[[$COO]]>,
225+
# C_HECK-SAME: %[[B:.*1]]: !torch.vtensor<[8,8],f32>) -> !torch.vtensor<[8,8],f32> {
226+
# C_HECK: %[[R:.*]] = torch.aten.{{matmul|mm}} %[[A]], %[[B]] : !torch.vtensor<[8,8],f32,#[[$COO]]>, !torch.vtensor<[8,8],f32> -> !torch.vtensor<[8,8],f32>
227+
# C_HECK: return %[[R]] : !torch.vtensor<[8,8],f32>
228+
# C_HECK: }
233229
##
234-
# CHECK: torch.sparse
235-
# CHECK: tensor({{\[}}[8., 8., 8., 8., 8., 8., 8., 8.],
236-
# CHECK-COUNT-6: [8., 8., 8., 8., 8., 8., 8., 8.],
237-
# CHECK: [8., 8., 8., 8., 8., 8., 8., 8.]{{\]}})
238-
# CHECK: torch.mlir
239-
# CHECK: {{\[}}[8. 8. 8. 8. 8. 8. 8. 8.]
240-
# CHECK-COUNT-6: [8. 8. 8. 8. 8. 8. 8. 8.]
241-
# CHECK: [8. 8. 8. 8. 8. 8. 8. 8.]{{\]}}
230+
# C_HECK: torch.sparse
231+
# C_HECK: tensor({{\[}}[8., 8., 8., 8., 8., 8., 8., 8.],
232+
# C_HECK-COUNT-6: [8., 8., 8., 8., 8., 8., 8., 8.],
233+
# C_HECK: [8., 8., 8., 8., 8., 8., 8., 8.]{{\]}})
234+
# C_HECK: torch.mlir
235+
# C_HECK: {{\[}}[8. 8. 8. 8. 8. 8. 8. 8.]
236+
# C_HECK-COUNT-6: [8. 8. 8. 8. 8. 8. 8. 8.]
237+
# C_HECK: [8. 8. 8. 8. 8. 8. 8. 8.]{{\]}}
242238
#
243239
def test_sparse_SpMM():
244240
class MatMulNet(torch.nn.Module):
@@ -263,40 +259,40 @@ def forward(self, x, y):
263259
print(res2)
264260

265261

266-
@run
262+
# @run
267263
#
268-
# CHECK-LABEL: test_sparse_eltwise
269-
# CHECK: #[[$CSRD:.*]] = #sparse_tensor.encoding<{ map = (d0, d1, d2) -> (d0 : dense, d1 : compressed, d2 : dense), posWidth = 64, crdWidth = 64 }>
270-
# CHECK: func.func @main(
271-
# CHECK-SAME: %[[A:.*]]: !torch.vtensor<[4,2,2],f32,#[[$CSRD]]>) -> !torch.vtensor<[4,2,2],f32,#[[$CSRD]]> {
272-
# CHECK: %[[R:.*]] = torch.aten.neg %[[A]] : !torch.vtensor<[4,2,2],f32,#[[$CSRD]]> -> !torch.vtensor<[4,2,2],f32,#[[$CSRD]]>
273-
# CHECK: return %[[R]] : !torch.vtensor<[4,2,2],f32,#[[$CSRD]]>
274-
# CHECK: }
275-
# CHECK: #[[$BCSR:.*]] = #sparse_tensor.encoding<{ map = (d0, d1, d2) -> (d0 : batch, d1 : dense, d2 : compressed), posWidth = 64, crdWidth = 64 }>
276-
# CHECK: func.func @main(
277-
# CHECK-SAME: %[[A:.*]]: !torch.vtensor<[4,2,2],f32,#[[$BCSR]]>) -> !torch.vtensor<[4,2,2],f32,#[[$BCSR]]> {
278-
# CHECK: %[[R:.*]] = torch.aten.neg %[[A]] : !torch.vtensor<[4,2,2],f32,#[[$BCSR]]> -> !torch.vtensor<[4,2,2],f32,#[[$BCSR]]>
279-
# CHECK: return %[[R]] : !torch.vtensor<[4,2,2],f32,#[[$BCSR]]>
280-
# CHECK: }
264+
# C_HECK-LABEL: test_sparse_eltwise
265+
# C_HECK: #[[$CSRD:.*]] = #sparse_tensor.encoding<{ map = (d0, d1, d2) -> (d0 : dense, d1 : compressed, d2 : dense), posWidth = 64, crdWidth = 64 }>
266+
# C_HECK: func.func @main(
267+
# C_HECK-SAME: %[[A:.*]]: !torch.vtensor<[4,2,2],f32,#[[$CSRD]]>) -> !torch.vtensor<[4,2,2],f32,#[[$CSRD]]> {
268+
# C_HECK: %[[R:.*]] = torch.aten.neg %[[A]] : !torch.vtensor<[4,2,2],f32,#[[$CSRD]]> -> !torch.vtensor<[4,2,2],f32,#[[$CSRD]]>
269+
# C_HECK: return %[[R]] : !torch.vtensor<[4,2,2],f32,#[[$CSRD]]>
270+
# C_HECK: }
271+
# C_HECK: #[[$BCSR:.*]] = #sparse_tensor.encoding<{ map = (d0, d1, d2) -> (d0 : batch, d1 : dense, d2 : compressed), posWidth = 64, crdWidth = 64 }>
272+
# C_HECK: func.func @main(
273+
# C_HECK-SAME: %[[A:.*]]: !torch.vtensor<[4,2,2],f32,#[[$BCSR]]>) -> !torch.vtensor<[4,2,2],f32,#[[$BCSR]]> {
274+
# C_HECK: %[[R:.*]] = torch.aten.neg %[[A]] : !torch.vtensor<[4,2,2],f32,#[[$BCSR]]> -> !torch.vtensor<[4,2,2],f32,#[[$BCSR]]>
275+
# C_HECK: return %[[R]] : !torch.vtensor<[4,2,2],f32,#[[$BCSR]]>
276+
# C_HECK: }
281277
#
282-
# CHECK: torch.sparse
283-
# CHECK: tensor(crow_indices=tensor([0, 2, 4, 6, 8]),
284-
# CHECK: col_indices=tensor([0, 1, 0, 1, 0, 1, 0, 1]),
285-
# CHECK: values=tensor({{\[}}[ -1., -2.],
286-
# CHECK: [ -3., -4.],
287-
# CHECK: [ -5., -6.],
288-
# CHECK: [ -7., -8.],
289-
# CHECK: [ -9., -10.],
290-
# CHECK: [-11., -12.],
291-
# CHECK: [-13., -14.],
292-
# CHECK: [-15., -16.]{{\]}}), size=(4, 2, 2), nnz=8,
293-
# CHECK: layout=torch.sparse_csr)
294-
# CHECK: torch.mlir
295-
# CHECK: [0 2 4 6 8]
296-
# CHECK: [0 1 0 1 0 1 0 1]
297-
# CHECK: [ -1. -2. -3. -4. -5. -6. -7. -8. -9. -10. -11. -12. -13. -14.
298-
# CHECK: -15. -16.]
299-
# CHECK: torch.mlir.batch
278+
# C_HECK: torch.sparse
279+
# C_HECK: tensor(crow_indices=tensor([0, 2, 4, 6, 8]),
280+
# C_HECK: col_indices=tensor([0, 1, 0, 1, 0, 1, 0, 1]),
281+
# C_HECK: values=tensor({{\[}}[ -1., -2.],
282+
# C_HECK: [ -3., -4.],
283+
# C_HECK: [ -5., -6.],
284+
# C_HECK: [ -7., -8.],
285+
# C_HECK: [ -9., -10.],
286+
# C_HECK: [-11., -12.],
287+
# C_HECK: [-13., -14.],
288+
# C_HECK: [-15., -16.]{{\]}}), size=(4, 2, 2), nnz=8,
289+
# C_HECK: layout=torch.sparse_csr)
290+
# C_HECK: torch.mlir
291+
# C_HECK: [0 2 4 6 8]
292+
# C_HECK: [0 1 0 1 0 1 0 1]
293+
# C_HECK: [ -1. -2. -3. -4. -5. -6. -7. -8. -9. -10. -11. -12. -13. -14.
294+
# C_HECK: -15. -16.]
295+
# C_HECK: torch.mlir.batch
300296
#
301297
def test_sparse_eltwise():
302298
class EltNet(torch.nn.Module):
@@ -439,20 +435,20 @@ def forward(self, x):
439435
print(res2[4])
440436

441437

442-
@run
438+
# @run
443439
#
444-
# CHECK-LABEL: test_sparse_network
445-
# CHECK: func.func @main(
446-
# CHECK-SAME: %[[A:.*]]: !torch.vtensor<[2,3,8,8],f32>) -> !torch.vtensor<[8],f32> {
440+
# C_HECK-LABEL: test_sparse_network
441+
# C_HECK: func.func @main(
442+
# C_HECK-SAME: %[[A:.*]]: !torch.vtensor<[2,3,8,8],f32>) -> !torch.vtensor<[8],f32> {
447443
# ... lots of IR ...
448-
# CHECK-COUNT-15: torch.aten.mul.Tensor
444+
# C_HECK-COUNT-15: torch.aten.mul.Tensor
449445
# ... lots of IR ...
450-
# CHECK: }
446+
# C_HECK: }
451447
#
452-
# CHECK: torch.sparse
453-
# CHECK: tensor([ 0., 11., 9., 11., 13., 11., 10., 12.])
454-
# CHECK: torch.mlir
455-
# CHECK: [ 0. 11. 9. 11. 13. 11. 10. 12.]
448+
# C_HECK: torch.sparse
449+
# C_HECK: tensor([ 0., 11., 9., 11., 13., 11., 10., 12.])
450+
# C_HECK: torch.mlir
451+
# C_HECK: [ 0. 11. 9. 11. 13. 11. 10. 12.]
456452
#
457453
def test_sparse_network():
458454
def spike(input):
@@ -525,30 +521,30 @@ def forward(self, X):
525521
print(res2)
526522

527523

528-
@run
524+
# @run
529525
#
530-
# CHECK-LABEL: test_sparse_feature_scaling
531-
# CHECK: func.func @main(
532-
# CHECK-SAME: %[[A:.*]]: !torch.vtensor<[4,4],f32>) -> !torch.vtensor<[4,4],f32> {
526+
# C_HECK-LABEL: test_sparse_feature_scaling
527+
# C_HECK: func.func @main(
528+
# C_HECK-SAME: %[[A:.*]]: !torch.vtensor<[4,4],f32>) -> !torch.vtensor<[4,4],f32> {
533529
# ... more IR ...
534-
# CHECK: %[[D:.*]] = torch.operator "torch.aten.{{to_sparse|_to_sparse}}"
535-
# CHECK: %[[R:.*]] = torch.aten.{{matmul|mm}} %[[D]], %[[A]]
536-
# CHECK return %[[R]] : !torch.vtensor<[4,4],f32>
537-
# CHECK: }
530+
# C_HECK: %[[D:.*]] = torch.operator "torch.aten.{{to_sparse|_to_sparse}}"
531+
# C_HECK: %[[R:.*]] = torch.aten.{{matmul|mm}} %[[D]], %[[A]]
532+
# C_HECK return %[[R]] : !torch.vtensor<[4,4],f32>
533+
# C_HECK: }
538534
#
539-
# CHECK: torch.sparse
540-
# CHECK: tensor({{\[}}[0.3342, 0.5173, 0.0596, 0.0889],
541-
# CHECK: [0.1321, 0.2724, 0.2105, 0.3851],
542-
# CHECK: [0.2478, 0.3439, 0.1898, 0.2185],
543-
# CHECK: [0.0222, 0.1683, 0.2928, 0.5167]{{\]}})
535+
# C_HECK: torch.sparse
536+
# C_HECK: tensor({{\[}}[0.3342, 0.5173, 0.0596, 0.0889],
537+
# C_HECK: [0.1321, 0.2724, 0.2105, 0.3851],
538+
# C_HECK: [0.2478, 0.3439, 0.1898, 0.2185],
539+
# C_HECK: [0.0222, 0.1683, 0.2928, 0.5167]{{\]}})
544540
#
545541
# TODO: first row looks suspect...
546542
#
547-
# CHECK: torch.mlir
548-
# CHECK: {{\[}}[0. 0. 0. 0. ]
549-
# CHECK: [0.13205223 0.27236593 0.21051763 0.38506418]
550-
# CHECK: [0.24781987 0.34391665 0.18976606 0.2184974 ]
551-
# CHECK: [0.02224578 0.16825409 0.29283574 0.51666445]{{\]}}
543+
# C_HECK: torch.mlir
544+
# C_HECK: {{\[}}[0. 0. 0. 0. ]
545+
# C_HECK: [0.13205223 0.27236593 0.21051763 0.38506418]
546+
# C_HECK: [0.24781987 0.34391665 0.18976606 0.2184974 ]
547+
# C_HECK: [0.02224578 0.16825409 0.29283574 0.51666445]{{\]}}
552548
#
553549
def test_sparse_feature_scaling():
554550
class Scale(nn.Module):

0 commit comments

Comments
 (0)