Towards supporting quantized structured kernels #74560
Closed
Conversation
This file contains hidden or 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
This PR add support for quantized tensors with "unknown quantizer", which means that we can use standard APIs like torch.empty to allocate quantized tensors, with the understanding that we will set the quantizer later. This makes meta functions applicable to quantized tensors (they will allocate with unknown quantizer and the kernel will set the quantizer later) and fixes a bug David Dang reported where structured kernels give a weird error message when you call them with quantized inputs. This is not a complete support for quantized structured kernels because I haven't actually tried porting any of the quantized implementations to structured; qadd is probably a good choice to try first as it does its broadcasting implementation using TensorIterator. My goal here is just to show that the error message is better. See also #52680 Signed-off-by: Edward Z. Yang <ezyang@fb.com> [ghstack-poisoned]
ezyang
added a commit
that referenced
this pull request
Mar 22, 2022
This PR add support for quantized tensors with "unknown quantizer", which means that we can use standard APIs like torch.empty to allocate quantized tensors, with the understanding that we will set the quantizer later. This makes meta functions applicable to quantized tensors (they will allocate with unknown quantizer and the kernel will set the quantizer later) and fixes a bug David Dang reported where structured kernels give a weird error message when you call them with quantized inputs. This is not a complete support for quantized structured kernels because I haven't actually tried porting any of the quantized implementations to structured; qadd is probably a good choice to try first as it does its broadcasting implementation using TensorIterator. My goal here is just to show that the error message is better. See also #52680 Signed-off-by: Edward Z. Yang <ezyangfb.com> ghstack-source-id: aa7272d Pull Request resolved: #74560
Contributor
🔗 Helpful links
💊 CI failures summary and remediationsAs of commit 5a934fd (more details on the Dr. CI page):
🕵️♀️ 18 failures not recognized by patterns:The following CI failures may be due to changes from the PR
This comment was automatically generated by Dr. CI (expand for details).Please report bugs/suggestions to the (internal) Dr. CI Users group. |
bdhirsh
reviewed
Mar 22, 2022
| runtime_empty_supported_check = "" | ||
| elif backend_index.dispatch_key == DispatchKey.CompositeExplicitAutograd: | ||
| elif backend_index.dispatch_key in ( | ||
| DispatchKey.CompositeExplicitAutograd, DispatchKey.QuantizedCPU, DispatchKey.QuantizedCUDA): |
Contributor
There was a problem hiding this comment.
tiny nit if we care about the perf (probably not necessary, just calling it out): we could avoid the dispatcher hop here from calling at::empty, but we'd have to abide by the naming convention and name the native kernels at::native::empty_quantizedcpu/cuda
Contributor
Author
There was a problem hiding this comment.
oh this used to not be easy to do but now it is easy
bdhirsh
reviewed
Mar 22, 2022
| empty_impl = "at::empty" | ||
| empty_strided_impl = "at::empty_strided" | ||
| runtime_empty_supported_check = """\ | ||
| if (!c10::detail::backend_supports_empty_operator(options)) {{ |
Contributor
There was a problem hiding this comment.
thanks for killing this :')
jerryzh168
reviewed
Mar 22, 2022
| r"Registration to both CompositeImplicitAutograd and CompositeExplicitAutograd is not allowed"): | ||
| dispatcher.register(["CompositeExplicitAutograd", "CompositeImplicitAutograd"]) | ||
|
|
||
| def test_quantized_structured_not_implemented(self): |
Contributor
There was a problem hiding this comment.
can we add a test in https://github.com/pytorch/pytorch/blob/master/test/quantization/core/test_quantized_tensor.py#L142 for calling some methods (e.g. qscheme) on an unknown tensor as well?
jerryzh168
reviewed
Mar 22, 2022
jerryzh168
approved these changes
Mar 31, 2022
Contributor
jerryzh168
left a comment
jerryzh168
left a comment
There was a problem hiding this comment.
followups are non-blocking, will accept first
Contributor
Author
|
@pytorchbot merge this |
Collaborator
|
Merge failed due to Matched rule superuser, but it was not reviewed yet by any of:pbelevich,H-Huang,albanD,hlu1,jamesr66a, ... |
This PR add support for quantized tensors with "unknown quantizer", which means that we can use standard APIs like torch.empty to allocate quantized tensors, with the understanding that we will set the quantizer later. This makes meta functions applicable to quantized tensors (they will allocate with unknown quantizer and the kernel will set the quantizer later) and fixes a bug David Dang reported where structured kernels give a weird error message when you call them with quantized inputs. This is not a complete support for quantized structured kernels because I haven't actually tried porting any of the quantized implementations to structured; qadd is probably a good choice to try first as it does its broadcasting implementation using TensorIterator. My goal here is just to show that the error message is better. See also #52680 Signed-off-by: Edward Z. Yang <ezyangfb.com> [ghstack-poisoned]
Contributor
|
@dzdang has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
facebook-github-bot
pushed a commit
that referenced
this pull request
Apr 5, 2022
Summary: Pull Request resolved: #74560 This PR add support for quantized tensors with "unknown quantizer", which means that we can use standard APIs like torch.empty to allocate quantized tensors, with the understanding that we will set the quantizer later. This makes meta functions applicable to quantized tensors (they will allocate with unknown quantizer and the kernel will set the quantizer later) and fixes a bug David Dang reported where structured kernels give a weird error message when you call them with quantized inputs. This is not a complete support for quantized structured kernels because I haven't actually tried porting any of the quantized implementations to structured; qadd is probably a good choice to try first as it does its broadcasting implementation using TensorIterator. My goal here is just to show that the error message is better. See also #52680 Signed-off-by: Edward Z. Yang <ezyangfb.com> Test Plan: Imported from OSS Reviewed By: mruberry Differential Revision: D35317441 Pulled By: dzdang fbshipit-source-id: ffb85b0e06ccbcc2b01052ca6760517684048b39
Contributor
|
Hey @ezyang. |
ezyang
added
release notes: composability
dzdang
added a commit
that referenced
this pull request
Apr 5, 2022
…tion_pad1d_quantized_cpu and" Summary: With the introduction of structured kernel support for quantized tensors in #74560, we are able to remove the dimension and output resizing code in reflection_pad1d_out_template (this code already present in reflection_pad1d), as well as the implementation for reflection_pad1d_quantized_cpu. This PR should introduce no functional changes. Test plan: ``` python run_test.py ``` Differential Revision: [D35148152](https://our.internmc.facebook.com/intern/diff/D35148152) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 5, 2022
…ized_cpu and" Summary: With the introduction of structured kernel support for quantized tensors in #74560, we are able to remove the dimension and output resizing code in reflection_pad1d_out_template (this code already present in reflection_pad1d), as well as the implementation for reflection_pad1d_quantized_cpu. This PR should introduce no functional changes. Test plan: ``` python run_test.py ``` Differential Revision: [D35148152](https://our.internmc.facebook.com/intern/diff/D35148152) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 5, 2022
…tion_pad1d_quantized_cpu, dimension and output resizing code in reflection_pad1d_out_template and implemented reflection_pad1d_out_quantized_cpu" Summary: With the introduction of structured kernel support for quantized tensors in #74560, we are able to remove the dimension and output resizing code in reflection_pad1d_out_template. This code is already present in reflection_pad1d. reflection_pad1d_quantized_cpu has also been removed as quantized tensors can now use reflection_pad1d after the changes in the linked PR. reflection_pad1d_out_quantized_cpu was implemented for quantized tensors. This PR should introduce no functional changes. Test plan: ``` python run_test.py ``` Differential Revision: [D35148152](https://our.internmc.facebook.com/intern/diff/D35148152) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 5, 2022
…ized_cpu, dimension and output resizing code in reflection_pad1d_out_template and implemented reflection_pad1d_out_quantized_cpu" Summary: With the introduction of structured kernel support for quantized tensors in #74560, we are able to remove the dimension and output resizing code in reflection_pad1d_out_template. This code is already present in reflection_pad1d. reflection_pad1d_quantized_cpu has also been removed as quantized tensors can now use reflection_pad1d after the changes in the linked PR. reflection_pad1d_out_quantized_cpu was implemented for quantized tensors. This PR should introduce no functional changes. Test plan: ``` python run_test.py ``` Differential Revision: [D35148152](https://our.internmc.facebook.com/intern/diff/D35148152) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 12, 2022
…tch registration for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 12, 2022
… for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 12, 2022
…tch registration for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 12, 2022
… for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 12, 2022
…tch registration for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 12, 2022
… for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 12, 2022
…tch registration for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 12, 2022
… for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 14, 2022
…tch registration for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 14, 2022
…tch registration for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 14, 2022
… for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 14, 2022
… for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 14, 2022
…tch registration for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 14, 2022
… for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 14, 2022
…tch registration for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 14, 2022
… for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 15, 2022
…tch registration for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 15, 2022
…tch registration for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 15, 2022
… for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 15, 2022
… for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 15, 2022
…tch registration for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 15, 2022
…tch registration for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 15, 2022
… for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 15, 2022
… for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 15, 2022
…tch registration for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 15, 2022
…tch registration for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 15, 2022
… for max_pool2d & quantized_max_pool2d and implemented max_pool2d_with_indices_out_quantized_cpu" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool2d, and implements a quantized kernel for max_pool2d_with_indices. This PR also introduces isnan() support for vectorized int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors. Test plan: ``` python test/test_quantization.py -k test_max_pool2d ``` Differential Revision: [D35420901](https://our.internmc.facebook.com/intern/diff/D35420901) [ghstack-poisoned]
dzdang
added a commit
that referenced
this pull request
Apr 15, 2022
… for max_pool1d & quantized_max_pool1d" Summary: This PR is part of a series of PRs addressing #54150, related to using dispatcher for calls to quantized backends as opposed to if/else conditionals. This particular PR removes the is_quantized check from max_pool1d and modifies max_pool1d_impl to be compatible with int tensors. This PR relies on #74560, which introduces structured kernel support for quantized tensors and #72353. Test plan: ``` python test/test_quantization.py -k test_max_pool1d ``` Differential Revision: [D35431831](https://our.internmc.facebook.com/intern/diff/D35431831) [ghstack-poisoned]
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.
Stack from ghstack (oldest at bottom):
This PR add support for quantized tensors with "unknown quantizer",
which means that we can use standard APIs like torch.empty to allocate
quantized tensors, with the understanding that we will set the
quantizer later. This makes meta functions applicable to quantized
tensors (they will allocate with unknown quantizer and the kernel
will set the quantizer later) and fixes a bug David Dang reported
where structured kernels give a weird error message when you call them
with quantized inputs.
This is not a complete support for quantized structured kernels because
I haven't actually tried porting any of the quantized implementations
to structured; qadd is probably a good choice to try first as it
does its broadcasting implementation using TensorIterator. My goal
here is just to show that the error message is better.
See also #52680
Signed-off-by: Edward Z. Yang <ezyangfb.com>
Differential Revision: D35317441