make tests pass with enable_first_class_module() enabled. #21565
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
pytorchbot
added
oncall: jit
This was referenced Jun 8, 2019
…bled." [WIP] make tests pass with enable_first_class_module() enabled. gh-metadata: pytorch pytorch 21565 gh/zdevito/60/head
pytorchbot
added
the
module: internals
…bled." [WIP] make tests pass with enable_first_class_module() enabled. gh-metadata: pytorch pytorch 21565 gh/zdevito/60/head
…bled." [WIP] make tests pass with enable_first_class_module() enabled. gh-metadata: pytorch pytorch 21565 gh/zdevito/60/head
…bled." [WIP] make tests pass with enable_first_class_module() enabled. gh-metadata: pytorch pytorch 21565 gh/zdevito/60/head
…bled." [WIP] make tests pass with enable_first_class_module() enabled. gh-metadata: pytorch pytorch 21565 gh/zdevito/60/head
…bled." [WIP] make tests pass with enable_first_class_module() enabled. gh-metadata: pytorch pytorch 21565 gh/zdevito/60/head
zdevito
changed the title
make tests pass with enable_first_class_module() enabled. gh-metadata: pytorch pytorch 21565 gh/zdevito/60/head
ZolotukhinM
approved these changes
Jun 11, 2019
There was a problem hiding this comment.
I'd prefer this to be splitted into several PRs, since some of the changes are completely independent - e.g. quantization and requires_grad changes. And you already proposed a good separation in the description. Apart from that, the changes look good with just some minor comments (inline).
| @@ -10083,7 +10079,7 @@ def traced_fn(x): | |||
| FileCheck().check("aten::neg").check("aten::add").run(str(traced_fn.graph)) | |||
|
|
|||
| def test_call_script_mod_from_tracing_fn(self): | |||
| with self.disableEmitHook(): | |||
| with self.assertRaisesRegex(RuntimeError, "must be registered as submodules"): | |||
There was a problem hiding this comment.
FileCheck call on line 10101 should probably be removed with this change.
| Value* definition; | ||
| IValue value; | ||
| }; | ||
| static void gatherParamsFirstClass(script::Module& module, Value* module_value, std::vector<ParamValue>& params) { |
torch/csrc/jit/python_tracer.cpp
Outdated
| @@ -79,7 +79,7 @@ std::shared_ptr<torch::jit::Graph> createGraphByTracing( | |||
| "captured in traces, so it would be a no-op."); | |||
| } | |||
| tracer::exit({toIValue(out)}); | |||
| if (!script::getFirstClassMode()) { | |||
| if (true || !script::getFirstClassMode()) { | |||
There was a problem hiding this comment.
yeah, Ill remove it
torch/csrc/jit/script/compiler.cpp
Outdated
| @@ -3073,7 +3073,7 @@ void runCleanupPasses(std::shared_ptr<Graph>& to_clean, bool convert_ssa) { | |||
| // be inlined into forked closures | |||
| liftClosures(to_clean); | |||
| inlineForkedClosures(to_clean); | |||
| if (!script::getFirstClassMode()) { | |||
| if (!script::getFirstClassMode() || true) { | |||
| @@ -99,6 +99,8 @@ struct TORCH_API Method { | |||
| return *function_; | |||
| } | |||
|
|
|||
| std::pair<std::shared_ptr<Graph>, std::vector<at::Tensor>> _lowered_graph(); | |||
Grabbag of fixes required to make it possible to pass the test suite using first-class module execution: * Fixed a bug where Objects were not considered requires_grad when one of their members might be. The default requires_grad now just checks contained types for requires grad, which should be more robust. * Edited some tests to be less fragile by removing checks for implementation details and putting some tests on more standard interfaces. * Provided `_lower_graph()` call that can be used by ONNX to still get the lowered exportable graph. gh-metadata: pytorch pytorch 21565 gh/zdevito/60/head
Grabbag of fixes required to make it possible to pass the test suite using first-class module execution: * Fixed a bug where Objects were not considered requires_grad when one of their members might be. The default requires_grad now just checks contained types for requires grad, which should be more robust. * Edited some tests to be less fragile by removing checks for implementation details and putting some tests on more standard interfaces. * Provided `_lower_graph()` call that can be used by ONNX to still get the lowered exportable graph. gh-metadata: pytorch pytorch 21565 gh/zdevito/60/head
Grabbag of fixes required to make it possible to pass the test suite using first-class module execution: * Fixed a bug where Objects were not considered requires_grad when one of their members might be. The default requires_grad now just checks contained types for requires grad, which should be more robust. * Edited some tests to be less fragile by removing checks for implementation details and putting some tests on more standard interfaces. * Provided `_lower_graph()` call that can be used by ONNX to still get the lowered exportable graph. gh-metadata: pytorch pytorch 21565 gh/zdevito/60/head
This was referenced Jun 12, 2019
Grabbag of fixes required to make it possible to pass the test suite using first-class module execution: * Fixed a bug where Objects were not considered requires_grad when one of their members might be. The default requires_grad now just checks contained types for requires grad, which should be more robust. * Edited some tests to be less fragile by removing checks for implementation details and putting some tests on more standard interfaces. * Provided `_lower_graph()` call that can be used by ONNX to still get the lowered exportable graph. gh-metadata: pytorch pytorch 21565 gh/zdevito/60/head
Grabbag of fixes required to make it possible to pass the test suite using first-class module execution: * Fixed a bug where Objects were not considered requires_grad when one of their members might be. The default requires_grad now just checks contained types for requires grad, which should be more robust. * Edited some tests to be less fragile by removing checks for implementation details and putting some tests on more standard interfaces. * Provided `_lower_graph()` call that can be used by ONNX to still get the lowered exportable graph. gh-metadata: pytorch pytorch 21565 gh/zdevito/60/head
Grabbag of fixes required to make it possible to pass the test suite using first-class module execution: * Fixed a bug where Objects were not considered requires_grad when one of their members might be. The default requires_grad now just checks contained types for requires grad, which should be more robust. * Edited some tests to be less fragile by removing checks for implementation details and putting some tests on more standard interfaces. * Provided `_lower_graph()` call that can be used by ONNX to still get the lowered exportable graph. gh-metadata: pytorch pytorch 21565 gh/zdevito/60/head
zdevito
added a commit
to zdevito/ATen
that referenced
this pull request
Jun 13, 2019
Summary: Pull Request resolved: pytorch/pytorch#21565 ghimport-source-id: d1fe735fb7821eadc59116fb921d8fe39a49f818 Reviewed By: driazati Differential Revision: D15729503 Pulled By: zdevito fbshipit-source-id: fabb678f040d21fae7545e3b2be1d098e24c544e
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Grabbag of fixes required to make it possible to pass the test suite using first-class module execution:
_lower_graph()call that can be used by ONNX to still get the lowered exportable graph.Stack from ghstack:
Differential Revision: D15729503