Functionalization integration #4158
alanwaketan
merged
130 commits into
master
from
functionalization
alanwaketan
commented
on 2022-12-16
alanwaketan
commented
on 2022-12-16
alanwaketan
commented
on 2022-12-20
POC of functionalization integration
110d3275
Add debugging print lines
d228571d
Run linter
08a6e088
Run linter again
2f3dc479
Update torch pin
b75adde5
Add more debugging lines
ca92a1cc
Lower lift and lift_copy ops
a3462c54
Update XLATensor namespace calls to tensor_methods
f96fa13b
Update pytorch_test_base.py
8c87e51c
Disable failing tests for now
e39d8255
Update lift_fresh ops
ff5d44f7
PoC for detach_copy
f05dcb97
Refine the comment
1a0c55f2
POC of functionalization integration
f082f286
Add debugging print lines
511cd7d7
Run linter
a999a144
Run linter again
cbffcacd
Update torch pin
2ed5ec24
Add more debugging lines
c37ff8b5
Update XLATensor namespace calls to tensor_methods
4f1c8bd6
Disable more failing tests
c29450b2
Run linter
fc227d35
Disable more tests
bc57c8f1
Clean up some comments
945ce35c
Skip test_empty_strided
a27bfd07
Fix detach_copy
4ad080b1
Skip some test_operations.py tests
558c8ada
Enable test_advancedindex_mixed_cpu_devices_xla
4834c26c
Remove some comments
4bc241b7
Enable test_basic_advanced_combined_xla
b0512b06
Skip test_squeeze_nonzero
70484f49
Enable tests in test_embedding and test_pooling
e70093db
error on view op cpu fallback
09c841e9
Fix crash on test_empty_strided
1910a524
Fix index ops
2f38ae0b
Unskip fixed tests
0c76f095
POC of functionalization integration
110d3275
Add debugging print lines
d228571d
Run linter again
2f3dc479
Update torch pin
b75adde5
Add more debugging lines
ca92a1cc
Lower lift and lift_copy ops
a3462c54
Update XLATensor namespace calls to tensor_methods
f96fa13b
Update pytorch_test_base.py
8c87e51c
Disable failing tests for now
e39d8255
Update lift_fresh ops
ff5d44f7
PoC for detach_copy
f05dcb97
Refine the comment
1a0c55f2
POC of functionalization integration
f082f286
Add debugging print lines
511cd7d7
Run linter
a999a144
Run linter again
cbffcacd
Update torch pin
2ed5ec24
Add more debugging lines
c37ff8b5
Update XLATensor namespace calls to tensor_methods
4f1c8bd6
Disable more failing tests
c29450b2
Run linter
fc227d35
Disable more tests
bc57c8f1
Clean up some comments
945ce35c
Skip test_empty_strided
a27bfd07
Fix detach_copy
4ad080b1
Skip some test_operations.py tests
558c8ada
Enable test_advancedindex_mixed_cpu_devices_xla
4834c26c
Remove some comments
4bc241b7
Enable test_basic_advanced_combined_xla
b0512b06
Skip test_squeeze_nonzero
70484f49
Enable tests in test_embedding and test_pooling
e70093db
Fix crash on test_empty_strided
1910a524
Fix index ops
2f38ae0b
Lower _to_copy
21cf6d5b
unskip test_squeeze_nonzero
bffe20cc
Skip test_squeeze_nonzero as it fails with dynamic shapes
c73c31e6
Fix test_affine tests by lowering affine_grid_generator
fc92fff3
Move affine_grid_generator to functionalization ops
b1e831eb
Delete debugging cout statements
4943c5c8
Run linter
d1ebfb3e
Lower _euclidean_dist
c68bf5b3
lower copy
ea0ae0b8
Route embeeding to functionalize pass
c66f5854
Lower reshape
2701b654
Introduce xla_expand_outplace
e3a360d7
Route _trilinear to functionalize pass
0cc2e051
lower mvlgamma
382f4dc2
lower masked_scatter
25e21c6c
Lower zeros_symint
29b3c5eb
Move xla_expand_outplace to tensor_util.h
177f0286
Add another xla_expand_outplace function to tensor_util.h to fix inde…
b4dacce0
Lower slice.Tensor
fd351730
Lower permute.Tensor
79841746
Lower select.int
811a4e4a
Lower t
2d2d1530
Skip test_nonzero_cast
a41d68f7
Disable some more failing tests
6222b03f
Disable failing XLA tests
d5642013
Enable test_mesh_service.py tests again to expose regression errors
37aecd4e
Skip even more tests
9cea4763
Revert masked_scatter
75c8d0bf
Revert "lower copy"
a4448a35
Revert "Lower zeros_symint"
8c40f3e5
Skip fsdp
138fd9a5
Skip more tests for GPU
7095c540
Adds some counters
398366a4
Fix linters
c2f8a126
Skip test_sizeAdd
8104de9a
Skip test_narrow_copy_non_contiguous
9e10f9f9
Run linter
1a69e3cb
Disable new dynamo tests that fail with functionalization
872badc2
Skip more spmd tests
c8533fe9
Set CONTINUE_ON_ERROR=true for testing
9734fcf3
Disable cpp tests
a6a9be9c
Disable Keep Going
d3466c37
[Functionalization] Enable cpp tests (#4462)
7f93346f
Re-enable dynamo tests (#4454)
edc4cf72
[Functionalization] Fix test_simple_expand_on_2d_tensor (#4452)
25538575
Lower _prelu_kernel to fix prelu cpp tests
462f1aa2
Lower diag_embed to fix diagflat cpp tests
e4c4e893
Lower copy_ and copy ops to fix instance_norm op tests
1b02433d
Update DiagFlat cpp unit tests expected counter changes
6d0405c1
Rename prelu to _prelu_kernel and update diag_embed op to functioanli…
abc373db
Run linter
b4da21b2
Update copy_ op
456a38ca
Enable TestEinsumPyTorchLowerRepeatedAxisBackward cpp test
eaea4c77
remove skip test on "test_simple_expand". (#4481)
3b52048a
[Functionalization] Fix torch.distributed tests (#4477)
24b1f651
[Functionalization] Mark test_logcumsumexp NTBF (#4506)
8e0835cc
Skip test_exponential
7f237a56
Turn on keep going
834f3f50
Skip dynamic shape tests again
bd0c9b6b
Revert "Turn on keep going"
ec3a6963
[Functionalization] Mark test_narrow_copy_non_contiguous won't fix (#…
3d60a23e
[Functionalization] Re-enable some of the dynamic shape test (#4518)
43536e24
[Functionalization] Fix more cpp tests that require new lowerings (#4…
b6e81372
skip dynamic shape tests again
82a89b4d
Turn on keep going
e0f5dfb1
Revert "skip dynamic shape tests again"
100c4485
[Functionalization] Fix ScatterReduce (#4576)
135e0e4f
Add debugging print lines
d228571d
Run linter again
2f3dc479
Update torch pin
b75adde5
Add more debugging lines
ca92a1cc
Lower lift and lift_copy ops
a3462c54
Update XLATensor namespace calls to tensor_methods
f96fa13b
Update pytorch_test_base.py
8c87e51c
Disable failing tests for now
e39d8255
Update lift_fresh ops
ff5d44f7
PoC for detach_copy
f05dcb97
Refine the comment
1a0c55f2
POC of functionalization integration
f082f286
Add debugging print lines
511cd7d7
Run linter
a999a144
Run linter again
cbffcacd
Update torch pin
2ed5ec24
Add more debugging lines
c37ff8b5
Update XLATensor namespace calls to tensor_methods
4f1c8bd6
Disable more failing tests
c29450b2
Run linter
fc227d35
Disable more tests
bc57c8f1
Clean up some comments
945ce35c
Skip test_empty_strided
a27bfd07
Fix detach_copy
4ad080b1
Skip some test_operations.py tests
558c8ada
Enable test_advancedindex_mixed_cpu_devices_xla
4834c26c
Remove some comments
4bc241b7
Enable test_basic_advanced_combined_xla
b0512b06
Skip test_squeeze_nonzero
70484f49
Enable tests in test_embedding and test_pooling
e70093db
Fix crash on test_empty_strided
1910a524
Fix index ops
2f38ae0b
Lower _to_copy
21cf6d5b
unskip test_squeeze_nonzero
bffe20cc
Skip test_squeeze_nonzero as it fails with dynamic shapes
c73c31e6
Fix test_affine tests by lowering affine_grid_generator
fc92fff3
Move affine_grid_generator to functionalization ops
b1e831eb
Delete debugging cout statements
4943c5c8
Run linter
d1ebfb3e
Lower _euclidean_dist
c68bf5b3
lower copy
ea0ae0b8
Route embeeding to functionalize pass
c66f5854
Lower reshape
2701b654
Introduce xla_expand_outplace
e3a360d7
lower mvlgamma
382f4dc2
lower masked_scatter
25e21c6c
Lower zeros_symint
29b3c5eb
Move xla_expand_outplace to tensor_util.h
177f0286
Add another xla_expand_outplace function to tensor_util.h to fix inde…
b4dacce0
Lower slice.Tensor
fd351730
Lower permute.Tensor
79841746
Lower select.int
811a4e4a
Lower t
2d2d1530
Skip test_nonzero_cast
a41d68f7
Disable some more failing tests
6222b03f
Disable failing XLA tests
d5642013
Enable test_mesh_service.py tests again to expose regression errors
37aecd4e
Skip even more tests
9cea4763
Revert masked_scatter
75c8d0bf
Revert "lower copy"
a4448a35
Revert "Lower zeros_symint"
8c40f3e5
Skip fsdp
138fd9a5
Skip more tests for GPU
7095c540
Adds some counters
398366a4
Fix linters
c2f8a126
Skip test_sizeAdd
8104de9a
Skip test_narrow_copy_non_contiguous
9e10f9f9
Run linter
1a69e3cb
Disable new dynamo tests that fail with functionalization
872badc2
Skip more spmd tests
c8533fe9
Set CONTINUE_ON_ERROR=true for testing
9734fcf3
Disable cpp tests
a6a9be9c
Disable Keep Going
d3466c37
[Functionalization] Enable cpp tests (#4462)
7f93346f
Re-enable dynamo tests (#4454)
edc4cf72
[Functionalization] Fix test_simple_expand_on_2d_tensor (#4452)
25538575
Lower _prelu_kernel to fix prelu cpp tests
462f1aa2
Lower diag_embed to fix diagflat cpp tests
e4c4e893
Lower copy_ and copy ops to fix instance_norm op tests
1b02433d
Update DiagFlat cpp unit tests expected counter changes
6d0405c1
Rename prelu to _prelu_kernel and update diag_embed op to functioanli…
abc373db
Run linter
b4da21b2
Update copy_ op
456a38ca
Enable TestEinsumPyTorchLowerRepeatedAxisBackward cpp test
eaea4c77
remove skip test on "test_simple_expand". (#4481)
3b52048a
[Functionalization] Fix torch.distributed tests (#4477)
24b1f651
[Functionalization] Mark test_logcumsumexp NTBF (#4506)
8e0835cc
Skip test_exponential
7f237a56
Turn on keep going
834f3f50
Skip dynamic shape tests again
bd0c9b6b
Revert "Turn on keep going"
ec3a6963
[Functionalization] Mark test_narrow_copy_non_contiguous won't fix (#…
3d60a23e
[Functionalization] Re-enable some of the dynamic shape test (#4518)
43536e24
[Functionalization] Fix more cpp tests that require new lowerings (#4…
b6e81372
skip dynamic shape tests again
82a89b4d
Turn on keep going
e0f5dfb1
Lower masked_scatter to fix masked_scatter op test in xla/test/test_o…
389ffe0c
Revert "Turn on keep going"
ccbe24f4
Update torch_pin to latest PyTorch PR
7468484f
re-enable all ds failing tests in functionalization branch (#4598)
d0b84e38
Add debugging print lines
d228571d
Update XLATensor namespace calls to tensor_methods
f96fa13b
Disable failing tests for now
e39d8255
Update lift_fresh ops
ff5d44f7
PoC for detach_copy
f05dcb97
POC of functionalization integration
f082f286
Run linter
fc227d35
Enable tests in test_embedding and test_pooling
e70093db
Fix test_affine tests by lowering affine_grid_generator
fc92fff3
Run linter
d1ebfb3e
lower copy
ea0ae0b8
Introduce xla_expand_outplace
e3a360d7
Lower slice.Tensor
fd351730
Disable failing XLA tests
d5642013
Disable Keep Going
d3466c37
Re-enable dynamo tests (#4454)
edc4cf72
Lower diag_embed to fix diagflat cpp tests
e4c4e893
Enable TestEinsumPyTorchLowerRepeatedAxisBackward cpp test
eaea4c77
Revert "Turn on keep going"
ec3a6963
Turn on keep going
e0f5dfb1
Skip test_squeeze_nonzero
e4d1456f
[Functionalization] Enable FSDP (#4691)
f82f7180
Properly skip tests that will be fixed later
4eed8452
alanwaketan
changed the title
vanbasten23
approved these changes
on 2023-03-02
Skip DynamoTrainingOptimizerTest.test_resnet18
f9963f6c
JackCaoG
approved these changes
on 2023-03-02
Reviewers
Assignees
No one assigned
Labels
REMOVE_TORCH_PIN
Milestone
No milestone
Login to write a write a comment.
Login via GitHub