[ONNX] Add linspace symbolic #58854

shubhambhokare1 · 2021-05-24T16:16:43Z

Adds support for linspace op
Modifies arange symbolic in opset 9 to replicate the same behavior in which dtype is determined (similar to opset 11) as in https://pytorch.org/docs/stable/generated/torch.arange.html
Enabled some arange unit tests which were disabled for opset 9

facebook-github-bot · 2021-05-24T16:16:49Z

💊 CI failures summary and remediations

As of commit 2a45070 (more details on the Dr. CI page):

3/3 failures possibly* introduced in this PR
- 1/3 non-scanned failure(s)

🕵️ 2 new failures recognized by patterns

The following CI failures do not appear to be due to upstream breakages:

pytorch_linux_backward_compatibility_check_test (1/2)

Step: "Run tests" (full log | diagnosis details | 🔁 rerun)

Jun 10 19:41:54 The PR is introducing backward ...m to confirm whether this change is wanted or not.

Jun 10 19:41:54 processing existing schema:  alltoall_base(__torch__.torch.classes.dist_c10d.ProcessGroup _0, Tensor _1, Tensor _2, int[] _3, int[] _4) -> (__torch__.torch.classes.dist_c10d.Work _0)
Jun 10 19:41:54 processing existing schema:  alltoall(__torch__.torch.classes.dist_c10d.ProcessGroup _0, Tensor[] _1, Tensor[] _2) -> (__torch__.torch.classes.dist_c10d.Work _0)
Jun 10 19:41:54 processing existing schema:  send(__torch__.torch.classes.dist_c10d.ProcessGroup _0, Tensor[] _1, int _2, int _3) -> (__torch__.torch.classes.dist_c10d.Work _0)
Jun 10 19:41:54 processing existing schema:  recv(__torch__.torch.classes.dist_c10d.ProcessGroup _0, Tensor[] _1, int _2, int _3) -> (__torch__.torch.classes.dist_c10d.Work _0)
Jun 10 19:41:54 processing existing schema:  recv_anysource(__torch__.torch.classes.dist_c10d.ProcessGroup _0, Tensor[] _1, int _2) -> (__torch__.torch.classes.dist_c10d.Work _0)
Jun 10 19:41:54 processing existing schema:  barrier(__torch__.torch.classes.dist_c10d.ProcessGroup _0) -> (__torch__.torch.classes.dist_c10d.Work _0)
Jun 10 19:41:54 processing existing schema:  __init__(__torch__.torch.classes.dist_c10d.frontend _0) -> (NoneType _0)
Jun 10 19:41:54 processing existing schema:  new_process_group_helper(__torch__.torch.classes.dist_c10d.frontend _0, int _1, int _2, int[] _3, str _4, __torch__.torch.classes.dist_c10d.Store _5, str? _6, int _7) -> (__torch__.torch.classes.dist_c10d.ProcessGroup _0)
Jun 10 19:41:54 processing existing schema:  get_process_group_by_name(__torch__.torch.classes.dist_c10d.frontend _0, str _1) -> (__torch__.torch.classes.dist_c10d.ProcessGroup _0)
Jun 10 19:41:54 processing existing schema:  get_name_of_process_group(__torch__.torch.classes.dist_c10d.frontend _0, __torch__.torch.classes.dist_c10d.ProcessGroup _1) -> (str _0)
Jun 10 19:41:54 The PR is introducing backward incompatible changes to the operator library. Please contact PyTorch team to confirm whether this change is wanted or not. 
Jun 10 19:41:54 
Jun 10 19:41:54 Broken ops: [
Jun 10 19:41:54 	aten::scatter_add.out(Tensor self, int dim, Tensor index, Tensor src, *, Tensor(a!) out) -> (Tensor(a!))
Jun 10 19:41:54 	aten::retains_grad(Tensor self) -> (bool)
Jun 10 19:41:54 	aten::scatter.src_out(Tensor self, int dim, Tensor index, Tensor src, *, Tensor(a!) out) -> (Tensor(a!))
Jun 10 19:41:54 	aten::scatter.value_out(Tensor self, int dim, Tensor index, Scalar value, *, Tensor(a!) out) -> (Tensor(a!))
Jun 10 19:41:54 	aten::scatter.reduce(Tensor self, int dim, Tensor index, Tensor src, *, str reduce) -> (Tensor)
Jun 10 19:41:54 	aten::scatter.reduce_out(Tensor self, int dim, Tensor index, Tensor src, *, str reduce, Tensor(a!) out) -> (Tensor(a!))
Jun 10 19:41:54 	aten::scatter.value_reduce(Tensor self, int dim, Tensor index, Scalar value, *, str reduce) -> (Tensor)
Jun 10 19:41:54 	aten::scatter.value_reduce_out(Tensor self, int dim, Tensor index, Scalar value, *, str reduce, Tensor(a!) out) -> (Tensor(a!))

pytorch_xla_linux_bionic_py3_6_clang9_build (2/2)

Step: "Build" (full log | diagnosis details | 🔁 rerun)

Jun 10 19:31:05 AssertionError: Found an invalid operator name: scatter.src_out

Jun 10 19:31:05   File "/var/lib/jenkins/workspace/tools/codegen/gen_backend_stubs.py", line 240, in <module>
Jun 10 19:31:05     main()
Jun 10 19:31:05   File "/var/lib/jenkins/workspace/tools/codegen/gen_backend_stubs.py", line 133, in main
Jun 10 19:31:05     run(options.source_yaml, options.output_dir, options.dry_run)
Jun 10 19:31:05   File "/var/lib/jenkins/workspace/tools/codegen/gen_backend_stubs.py", line 150, in run
Jun 10 19:31:05     parsed_backend_yaml = parse_backend_yaml(source_yaml, grouped_native_functions, backend_indices)
Jun 10 19:31:05   File "/var/lib/jenkins/workspace/tools/codegen/gen_backend_stubs.py", line 84, in parse_backend_yaml
Jun 10 19:31:05     backend_idx = create_backend_index(supported, backend_key)
Jun 10 19:31:05   File "/var/lib/jenkins/workspace/tools/codegen/gen_backend_stubs.py", line 65, in create_backend_index
Jun 10 19:31:05     assert op_name in native_functions_map, f"Found an invalid operator name: {op_name}"
Jun 10 19:31:05 AssertionError: Found an invalid operator name: scatter.src_out
Jun 10 19:31:05 ~/workspace/xla
Jun 10 19:31:05 + OPTS=()
Jun 10 19:31:05 + getopts O: OPTION
Jun 10 19:31:05 + case $OPTION in
Jun 10 19:31:05 + for i in ${OPTARG}
Jun 10 19:31:05 + OPTS+=("--cxxopt=${i}")
Jun 10 19:31:05 + getopts O: OPTION
Jun 10 19:31:05 + shift 2
Jun 10 19:31:05 + CMD=install
Jun 10 19:31:05 ++ dirname /var/lib/jenkins/workspace/xla/build_torch_xla_libs.sh

ci.pytorch.org: 1 failed

Failed: pr/pytorch-linux-bionic-rocm4.2-py3.6

This comment was automatically generated by Dr. CI (expand for details).

Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

test/onnx/test_pytorch_onnx_onnxruntime.py

torch/onnx/symbolic_opset9.py

BowenBao

looks good overall, please rebase

torch/onnx/symbolic_opset11.py

* Adds support for linspace op * Modifies arange symbolic in opset 9 to replicate the same behavior in which dtype is determined (similar to opset 11) as in https://pytorch.org/docs/stable/generated/torch.arange.html * Enabled some arange unit tests which were disabled for opset 9 Co-authored-by: Shubham Bhokare <shubhambhokare@gmail.com>

* Adds support for linspace op * Modifies arange symbolic in opset 9 to replicate the same behavior in which dtype is determined (similar to opset 11) as in https://pytorch.org/docs/stable/generated/torch.arange.html * Enabled some arange unit tests which were disabled for opset 9 Co-authored-by: Shubham Bhokare <shubhambhokare@gmail.com> [ghstack-poisoned]

* Adds support for linspace op * Modifies arange symbolic in opset 9 to replicate the same behavior in which dtype is determined (similar to opset 11) as in https://pytorch.org/docs/stable/generated/torch.arange.html * Enabled some arange unit tests which were disabled for opset 9 Co-authored-by: Shubham Bhokare <shubhambhokare@gmail.com> Differential Revision: [D29494911](https://our.internmc.facebook.com/intern/diff/D29494911) [ghstack-poisoned]

Summary: Pull Request resolved: #60246 * Adds support for linspace op * Modifies arange symbolic in opset 9 to replicate the same behavior in which dtype is determined (similar to opset 11) as in https://pytorch.org/docs/stable/generated/torch.arange.html * Enabled some arange unit tests which were disabled for opset 9 Test Plan: Imported from OSS Reviewed By: zou3519, ZolotukhinM Differential Revision: D29494911 Pulled By: SplitInfinity fbshipit-source-id: bddff18a90f8a78121c8ecdd1dafc15c69962d66 Co-authored-by: Shubham Bhokare <shubhambhokare@gmail.com>

shubhambhokare1 requested review from BowenBao, neginraoof and spandantiwari as code owners May 24, 2021 16:16

facebook-github-bot added the cla signed label May 24, 2021

pytorchbot added the open source label May 24, 2021

shubhambhokare1 mentioned this pull request May 24, 2021

RuntimeError: Exporting the operator linspace to ONNX opset version 11 is not supported. #58472

Closed

BowenBao force-pushed the onnx_ms_1 branch from f83cc6f to 58f1429 Compare May 27, 2021 19:40

shubhambhokare1 force-pushed the sbhokare/linspace-op branch from 86b3584 to ce3a454 Compare June 1, 2021 18:37

BowenBao force-pushed the onnx_ms_1 branch from bd3e156 to 3320a78 Compare June 7, 2021 03:53

BowenBao reviewed Jun 9, 2021

View reviewed changes

test/onnx/test_pytorch_onnx_onnxruntime.py Show resolved Hide resolved

BowenBao reviewed Jun 9, 2021

View reviewed changes

torch/onnx/symbolic_opset9.py Show resolved Hide resolved

BowenBao reviewed Jun 9, 2021

View reviewed changes

Shubham Bhokare added 4 commits June 9, 2021 18:55

Added tests

1fe4899

Added logic

81d2692

Added float support for arange in opset 9

ac4b9b4

Fix failures

365c150

shubhambhokare1 force-pushed the sbhokare/linspace-op branch from ce3a454 to 365c150 Compare June 9, 2021 18:56

Updated tests

6a89774

BowenBao reviewed Jun 9, 2021

View reviewed changes

torch/onnx/symbolic_opset11.py Outdated Show resolved Hide resolved

shubhambhokare1 added 4 commits June 9, 2021 23:20

Avoid opset11 duplication

6ac9636

Added helper function

4e2f407

Update helper linspace->arange

32d715e

Fix mypy lint

2a45070

BowenBao approved these changes Jun 11, 2021

View reviewed changes

BowenBao merged commit e2a1236 into pytorch:onnx_ms_1 Jun 11, 2021

This was referenced Jun 18, 2021

[ONNX] Handle onnx::Size in ComputeConstant folding (#59122) #60243

Closed

[ONNX] Fix shape inference for large model (#59320) #60244

Closed

garymm mentioned this pull request Oct 8, 2021

[v.1.10.0] Release Tracker #65438

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[ONNX] Add linspace symbolic #58854

[ONNX] Add linspace symbolic #58854

[ONNX] Add linspace symbolic #58854

[ONNX] Add linspace symbolic #58854

Conversation

💊 CI failures summary and remediations

🕵️ 2 new failures recognized by patterns

pytorch_linux_backward_compatibility_check_test (1/2)

pytorch_xla_linux_bionic_py3_6_clang9_build (2/2)

ci.pytorch.org: 1 failed

Choose a reason for hiding this comment