Port all unary float functions to structured #56082

ezyang · 2021-04-14T21:52:01Z

Stack from ghstack:

Port all non-float unary operators to structured (and rsqrt) #56151 Port all non-float unary operators to structured (and rsqrt)
Port all unary float functions to structured #56082 Port all unary float functions to structured
Change all unary functions stubs to use TensorIteratorBase& #56078 Change all unary functions stubs to use TensorIteratorBase&

The native_functions.yaml changes were done by codemod using the
following script:

import ruamel.yaml
from ruamel.yaml.tokens import CommentToken
from ruamel.yaml.error import CommentMark
from tools.codegen.model import *  # noqa: F403

with open("aten/src/ATen/native/native_functions.yaml", "r") as f:
    contents = f.read()

yaml = ruamel.yaml.YAML()
yaml.preserve_quotes = True
yaml.width = 1000
yaml.boolean_representation = ['False', 'True']
r = yaml.load(contents)

convert = '''\
acos
acosh
asin
asinh
atan
atanh
cos
cosh
digamma
erf
erfc
erfinv
exp
expm1
exp2
lgamma
log
log10
log1p
log2
reciprocal
sigmoid
sin
sinc
sinh
special_entr
sqrt
tan
tanh'''.split()

for e in r:
    f = NativeFunction.from_yaml(e, Location("", 0))
    if f.structured or f.structured_delegate is not None:
        continue
    n = f.func.name.name.base
    if n not in convert:
        continue
    # mutate e to make changes
    if f.func.kind() == SchemaKind.out:
        e.insert(1, 'structured', True)
        e.insert(2, 'structured_inherits', 'TensorIteratorBase')
    else:
        # TODO: The .out overload assumption is not sound in general
        e.insert(1, 'structured_delegate', f'{n}.out')

        e['dispatch'].pop('CPU', None)
        e['dispatch'].pop('CUDA', None)
        e['dispatch'].pop('CPU, CUDA', None)
        e['dispatch'].pop('CompositeExplicitAutograd', None)

        *_, last_k = e.keys()
        needs_fixup = False

        if not e['dispatch']:
            if last_k == 'dispatch':
                needs_fixup = True
            del e['dispatch']

        # Manually fix up newlines at the end, because ruamel
        # made some bad life choices about where to associate trailing
        # whitespace for nested dicts; see
        # https://stackoverflow.com/questions/42172399/modifying-yaml-using-ruamel-yaml-adds-extra-new-lines
        if needs_fixup:
            *_, last_k = e.keys()
            # post_key, pre_key, post_value, pre_value
            e.ca.items[last_k] = [None, None, CommentToken('\n\n', CommentMark(0), None), None]

with open("aten/src/ATen/native/native_functions.yaml.new", "w") as f:
    yaml.dump(r, f)

Signed-off-by: Edward Z. Yang [email protected]

Differential Revision: D27777769

The native_functions.yaml changes were done by codemod using the following script: ``` import ruamel.yaml from ruamel.yaml.tokens import CommentToken from ruamel.yaml.error import CommentMark from tools.codegen.model import * # noqa: F403 with open("aten/src/ATen/native/native_functions.yaml", "r") as f: contents = f.read() yaml = ruamel.yaml.YAML() yaml.preserve_quotes = True yaml.width = 1000 yaml.boolean_representation = ['False', 'True'] r = yaml.load(contents) convert = '''\ acos acosh asin asinh atan atanh cos cosh digamma erf erfc erfinv exp expm1 exp2 lgamma log log10 log1p log2 reciprocal sigmoid sin sinc sinh special_entr sqrt tan tanh'''.split() for e in r: f = NativeFunction.from_yaml(e, Location("", 0)) if f.structured or f.structured_delegate is not None: continue n = f.func.name.name.base if n not in convert: continue # mutate e to make changes if f.func.kind() == SchemaKind.out: e.insert(1, 'structured', True) e.insert(2, 'structured_inherits', 'TensorIteratorBase') else: # TODO: The .out overload assumption is not sound in general e.insert(1, 'structured_delegate', f'{n}.out') e['dispatch'].pop('CPU', None) e['dispatch'].pop('CUDA', None) e['dispatch'].pop('CPU, CUDA', None) e['dispatch'].pop('CompositeExplicitAutograd', None) *_, last_k = e.keys() needs_fixup = False if not e['dispatch']: if last_k == 'dispatch': needs_fixup = True del e['dispatch'] # Manually fix up newlines at the end, because ruamel # made some bad life choices about where to associate trailing # whitespace for nested dicts; see # https://stackoverflow.com/questions/42172399/modifying-yaml-using-ruamel-yaml-adds-extra-new-lines if needs_fixup: *_, last_k = e.keys() # post_key, pre_key, post_value, pre_value e.ca.items[last_k] = [None, None, CommentToken('\n\n', CommentMark(0), None), None] with open("aten/src/ATen/native/native_functions.yaml.new", "w") as f: yaml.dump(r, f) ``` Signed-off-by: Edward Z. Yang <[email protected]> [ghstack-poisoned]

facebook-github-bot · 2021-04-14T21:52:07Z

💊 CI failures summary and remediations

As of commit fe413dc (more details on the Dr. CI page):

3/5 failures possibly* introduced in this PR
- 1/3 non-scanned failure(s)
2/5 tentatively recognized as flaky ❄️
- Click here to rerun these jobs

🕵️ 2 new failures recognized by patterns

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

pytorch_doc_test (1/2)

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

Apr 15 20:27:51 sccache: error: couldn't connect to server

Apr 15 20:27:50 ++++ eval 'extract_trap_cmd '
Apr 15 20:27:50 +++++ extract_trap_cmd
Apr 15 20:27:50 +++++ printf '%s\n' ''
Apr 15 20:27:50 ++++ printf '%s\n' cleanup
Apr 15 20:27:50 +++ trap -- '
Apr 15 20:27:50 cleanup' EXIT
Apr 15 20:27:50 +++ [[ pytorch-linux-xenial-py3.6-gcc5.4-build != *pytorch-win-* ]]
Apr 15 20:27:50 +++ which sccache
Apr 15 20:27:50 +++ sccache --stop-server
Apr 15 20:27:51 Stopping sccache server...
Apr 15 20:27:51 sccache: error: couldn't connect to server
Apr 15 20:27:51 sccache: caused by: Connection refused (os error 111)
Apr 15 20:27:51 +++ true
Apr 15 20:27:51 +++ rm /var/lib/jenkins/sccache_error.log
Apr 15 20:27:51 +++ [[ pytorch-linux-xenial-py3.6-gcc5.4-build == *rocm* ]]
Apr 15 20:27:51 +++ SCCACHE_ERROR_LOG=/var/lib/jenkins/sccache_error.log
Apr 15 20:27:51 +++ SCCACHE_IDLE_TIMEOUT=1200
Apr 15 20:27:51 +++ RUST_LOG=sccache::server=error
Apr 15 20:27:51 +++ sccache --start-server
Apr 15 20:27:51 sccache: Starting the server...
Apr 15 20:27:51 +++ sccache --zero-stats

pytorch_python_doc_build (2/2)

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

Apr 15 20:36:06 Makefile:38: recipe for target 'html' failed

Apr 15 20:36:04 
Apr 15 20:36:04 copying static files... ... done
Apr 15 20:36:04 copying extra files... done
Apr 15 20:36:05 dumping search index in English (code: en)... done
Apr 15 20:36:05 dumping object inventory... done
Apr 15 20:36:05 build finished with problems, 1 warning.
Apr 15 20:36:05 /var/lib/jenkins/workspace/docs/src/pytorch-sphinx-theme/pytorch_sphinx_theme/search.html:21: RemovedInSphinx30Warning: To modify script_files in the theme is deprecated. Please insert a <script> tag directly in your theme instead.
Apr 15 20:36:05   <p class="last">
Apr 15 20:36:05 /var/lib/jenkins/workspace/docs/src/pytorch-sphinx-theme/pytorch_sphinx_theme/search.html:24: RemovedInSphinx30Warning: To modify script_files in the theme is deprecated. Please insert a <script> tag directly in your theme instead.
Apr 15 20:36:05   </p>
Apr 15 20:36:06 Makefile:38: recipe for target 'html' failed
Apr 15 20:36:06 make: *** [html] Error 1
Apr 15 20:36:06 ++ code=2
Apr 15 20:36:06 ++ '[' 2 -ne 0 ']'
Apr 15 20:36:06 ++ set +x
Apr 15 20:36:06 =========================
Apr 15 20:36:06 WARNING: autodoc: failed to import module 'launch' from module 'torch.distributed'; the following exception was raised:
Apr 15 20:36:06 =========================
Apr 15 20:36:06 Docs build failed. If the failure is not clear, scan back in the log
Apr 15 20:36:06 for any WARNINGS or for the line build finished with problems
Apr 15 20:36:06 (tried to echo the WARNINGS above the ==== line)

❄️ 2 failures tentatively classified as flaky

but reruns have not yet been triggered to confirm:

binary_linux_libtorch_3_7m_cpu_devtoolset7_shared-with-deps_build (1/2)

Step: "Spin up environment" (full log | diagnosis details | 🔁 rerun) ❄️

Error response from daemon: received unexpected HTTP status: 503 Service Unavailable

  Error pulling image pytorch/manylinux-cuda102: Error response from daemon: received unexpected HTTP status: 503 Service Unavailable... retrying
  image cache not found on this host, downloading pytorch/manylinux-cuda102

  Error pulling image pytorch/manylinux-cuda102: Error response from daemon: received unexpected HTTP status: 503 Service Unavailable... retrying
  image cache not found on this host, downloading pytorch/manylinux-cuda102

  Error pulling image pytorch/manylinux-cuda102: Error response from daemon: received unexpected HTTP status: 503 Service Unavailable... retrying
  image cache not found on this host, downloading pytorch/manylinux-cuda102


Error response from daemon: received unexpected HTTP status: 503 Service Unavailable

binary_linux_libtorch_3_7m_cpu_gcc5_4_cxx11-abi_shared-with-deps_build (2/2)

Step: "Spin up environment" (full log | diagnosis details | 🔁 rerun) ❄️

Error response from daemon: received unexpected HTTP status: 503 Service Unavailable

  Error pulling image pytorch/pytorch-binary-docker-image-ubuntu16.04:latest: Error response from daemon: received unexpected HTTP status: 503 Service Unavailable... retrying
  image cache not found on this host, downloading pytorch/pytorch-binary-docker-image-ubuntu16.04:latest

  Error pulling image pytorch/pytorch-binary-docker-image-ubuntu16.04:latest: Error response from daemon: received unexpected HTTP status: 503 Service Unavailable... retrying
  image cache not found on this host, downloading pytorch/pytorch-binary-docker-image-ubuntu16.04:latest

  Error pulling image pytorch/pytorch-binary-docker-image-ubuntu16.04:latest: Error response from daemon: received unexpected HTTP status: 503 Service Unavailable... retrying
  image cache not found on this host, downloading pytorch/pytorch-binary-docker-image-ubuntu16.04:latest


Error response from daemon: received unexpected HTTP status: 503 Service Unavailable

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.

bhosmer · 2021-04-14T22:25:21Z

aten/src/ATen/native/cpu/UnaryOpsKernel.cpp

  }                                                                                              \
  REGISTER_DISPATCH(op##_stub, &op##_kernel)

-  #define IMPLEMENT_COMPLEX_STRUCTURED_KERNEL(op)                                                \


why does this have to go?

It doesn't, it just turns out it's identical to the macro above.

ah is the iter.cast_outputs() on 689 redundant

Hmm, actually, no, that might be a bug...

Ah yeah, that was added in this PR: #52277. I think it's fine to remove the old macro now that all unary ops have been ported. The old un-structured kernels were written to call cast_outputs() directly for some reason, which is why it was fine for it not to be in the macro for un-structured kernels.

wenleix · 2021-04-14T23:41:40Z

aten/src/ATen/native/native_functions.yaml


 - func: asinh.out(Tensor self, *, Tensor(a!) out) -> Tensor(a!)
+  structured: True
+  structured_inherits: TensorIteratorBase


curious: what does this mean?

https://github.com/pytorch/rfcs/blob/rfc-0005/RFC-0005-structured-kernel-definitions.md#handling-tensoriterator

The native_functions.yaml changes were done by codemod using the following script: ``` import ruamel.yaml from ruamel.yaml.tokens import CommentToken from ruamel.yaml.error import CommentMark from tools.codegen.model import * # noqa: F403 with open("aten/src/ATen/native/native_functions.yaml", "r") as f: contents = f.read() yaml = ruamel.yaml.YAML() yaml.preserve_quotes = True yaml.width = 1000 yaml.boolean_representation = ['False', 'True'] r = yaml.load(contents) convert = '''\ acos acosh asin asinh atan atanh cos cosh digamma erf erfc erfinv exp expm1 exp2 lgamma log log10 log1p log2 reciprocal sigmoid sin sinc sinh special_entr sqrt tan tanh'''.split() for e in r: f = NativeFunction.from_yaml(e, Location("", 0)) if f.structured or f.structured_delegate is not None: continue n = f.func.name.name.base if n not in convert: continue # mutate e to make changes if f.func.kind() == SchemaKind.out: e.insert(1, 'structured', True) e.insert(2, 'structured_inherits', 'TensorIteratorBase') else: # TODO: The .out overload assumption is not sound in general e.insert(1, 'structured_delegate', f'{n}.out') e['dispatch'].pop('CPU', None) e['dispatch'].pop('CUDA', None) e['dispatch'].pop('CPU, CUDA', None) e['dispatch'].pop('CompositeExplicitAutograd', None) *_, last_k = e.keys() needs_fixup = False if not e['dispatch']: if last_k == 'dispatch': needs_fixup = True del e['dispatch'] # Manually fix up newlines at the end, because ruamel # made some bad life choices about where to associate trailing # whitespace for nested dicts; see # https://stackoverflow.com/questions/42172399/modifying-yaml-using-ruamel-yaml-adds-extra-new-lines if needs_fixup: *_, last_k = e.keys() # post_key, pre_key, post_value, pre_value e.ca.items[last_k] = [None, None, CommentToken('\n\n', CommentMark(0), None), None] with open("aten/src/ATen/native/native_functions.yaml.new", "w") as f: yaml.dump(r, f) ``` Signed-off-by: Edward Z. Yang <[email protected]> Differential Revision: [D27777769](https://our.internmc.facebook.com/intern/diff/D27777769) [ghstack-poisoned]

The native_functions.yaml changes were done by codemod using the following script: ``` import ruamel.yaml from ruamel.yaml.tokens import CommentToken from ruamel.yaml.error import CommentMark from tools.codegen.model import * # noqa: F403 with open("aten/src/ATen/native/native_functions.yaml", "r") as f: contents = f.read() yaml = ruamel.yaml.YAML() yaml.preserve_quotes = True yaml.width = 1000 yaml.boolean_representation = ['False', 'True'] r = yaml.load(contents) convert = '''\ acos acosh asin asinh atan atanh cos cosh digamma erf erfc erfinv exp expm1 exp2 lgamma log log10 log1p log2 reciprocal sigmoid sin sinc sinh special_entr sqrt tan tanh'''.split() for e in r: f = NativeFunction.from_yaml(e, Location("", 0)) if f.structured or f.structured_delegate is not None: continue n = f.func.name.name.base if n not in convert: continue # mutate e to make changes if f.func.kind() == SchemaKind.out: e.insert(1, 'structured', True) e.insert(2, 'structured_inherits', 'TensorIteratorBase') else: # TODO: The .out overload assumption is not sound in general e.insert(1, 'structured_delegate', f'{n}.out') e['dispatch'].pop('CPU', None) e['dispatch'].pop('CUDA', None) e['dispatch'].pop('CPU, CUDA', None) e['dispatch'].pop('CompositeExplicitAutograd', None) *_, last_k = e.keys() needs_fixup = False if not e['dispatch']: if last_k == 'dispatch': needs_fixup = True del e['dispatch'] # Manually fix up newlines at the end, because ruamel # made some bad life choices about where to associate trailing # whitespace for nested dicts; see # https://stackoverflow.com/questions/42172399/modifying-yaml-using-ruamel-yaml-adds-extra-new-lines if needs_fixup: *_, last_k = e.keys() # post_key, pre_key, post_value, pre_value e.ca.items[last_k] = [None, None, CommentToken('\n\n', CommentMark(0), None), None] with open("aten/src/ATen/native/native_functions.yaml.new", "w") as f: yaml.dump(r, f) ``` Signed-off-by: Edward Z. Yang <[email protected]> ghstack-source-id: 18c531b Pull Request resolved: #56082

The native_functions.yaml changes were done by codemod using the following script: ``` import ruamel.yaml from ruamel.yaml.tokens import CommentToken from ruamel.yaml.error import CommentMark from tools.codegen.model import * # noqa: F403 with open("aten/src/ATen/native/native_functions.yaml", "r") as f: contents = f.read() yaml = ruamel.yaml.YAML() yaml.preserve_quotes = True yaml.width = 1000 yaml.boolean_representation = ['False', 'True'] r = yaml.load(contents) convert = '''\ acos acosh asin asinh atan atanh cos cosh digamma erf erfc erfinv exp expm1 exp2 lgamma log log10 log1p log2 reciprocal sigmoid sin sinc sinh special_entr sqrt tan tanh'''.split() for e in r: f = NativeFunction.from_yaml(e, Location("", 0)) if f.structured or f.structured_delegate is not None: continue n = f.func.name.name.base if n not in convert: continue # mutate e to make changes if f.func.kind() == SchemaKind.out: e.insert(1, 'structured', True) e.insert(2, 'structured_inherits', 'TensorIteratorBase') else: # TODO: The .out overload assumption is not sound in general e.insert(1, 'structured_delegate', f'{n}.out') e['dispatch'].pop('CPU', None) e['dispatch'].pop('CUDA', None) e['dispatch'].pop('CPU, CUDA', None) e['dispatch'].pop('CompositeExplicitAutograd', None) *_, last_k = e.keys() needs_fixup = False if not e['dispatch']: if last_k == 'dispatch': needs_fixup = True del e['dispatch'] # Manually fix up newlines at the end, because ruamel # made some bad life choices about where to associate trailing # whitespace for nested dicts; see # https://stackoverflow.com/questions/42172399/modifying-yaml-using-ruamel-yaml-adds-extra-new-lines if needs_fixup: *_, last_k = e.keys() # post_key, pre_key, post_value, pre_value e.ca.items[last_k] = [None, None, CommentToken('\n\n', CommentMark(0), None), None] with open("aten/src/ATen/native/native_functions.yaml.new", "w") as f: yaml.dump(r, f) ``` Signed-off-by: Edward Z. Yang <[email protected]> Differential Revision: [D27777769](https://our.internmc.facebook.com/intern/diff/D27777769) [ghstack-poisoned]

The native_functions.yaml changes were done by codemod using the following script: ``` import ruamel.yaml from ruamel.yaml.tokens import CommentToken from ruamel.yaml.error import CommentMark from tools.codegen.model import * # noqa: F403 with open("aten/src/ATen/native/native_functions.yaml", "r") as f: contents = f.read() yaml = ruamel.yaml.YAML() yaml.preserve_quotes = True yaml.width = 1000 yaml.boolean_representation = ['False', 'True'] r = yaml.load(contents) convert = '''\ acos acosh asin asinh atan atanh cos cosh digamma erf erfc erfinv exp expm1 exp2 lgamma log log10 log1p log2 reciprocal sigmoid sin sinc sinh special_entr sqrt tan tanh'''.split() for e in r: f = NativeFunction.from_yaml(e, Location("", 0)) if f.structured or f.structured_delegate is not None: continue n = f.func.name.name.base if n not in convert: continue # mutate e to make changes if f.func.kind() == SchemaKind.out: e.insert(1, 'structured', True) e.insert(2, 'structured_inherits', 'TensorIteratorBase') else: # TODO: The .out overload assumption is not sound in general e.insert(1, 'structured_delegate', f'{n}.out') e['dispatch'].pop('CPU', None) e['dispatch'].pop('CUDA', None) e['dispatch'].pop('CPU, CUDA', None) e['dispatch'].pop('CompositeExplicitAutograd', None) *_, last_k = e.keys() needs_fixup = False if not e['dispatch']: if last_k == 'dispatch': needs_fixup = True del e['dispatch'] # Manually fix up newlines at the end, because ruamel # made some bad life choices about where to associate trailing # whitespace for nested dicts; see # https://stackoverflow.com/questions/42172399/modifying-yaml-using-ruamel-yaml-adds-extra-new-lines if needs_fixup: *_, last_k = e.keys() # post_key, pre_key, post_value, pre_value e.ca.items[last_k] = [None, None, CommentToken('\n\n', CommentMark(0), None), None] with open("aten/src/ATen/native/native_functions.yaml.new", "w") as f: yaml.dump(r, f) ``` Signed-off-by: Edward Z. Yang <[email protected]> ghstack-source-id: 98e3cf6 Pull Request resolved: #56082

The native_functions.yaml changes were done by codemod using the following script: ``` import ruamel.yaml from ruamel.yaml.tokens import CommentToken from ruamel.yaml.error import CommentMark from tools.codegen.model import * # noqa: F403 with open("aten/src/ATen/native/native_functions.yaml", "r") as f: contents = f.read() yaml = ruamel.yaml.YAML() yaml.preserve_quotes = True yaml.width = 1000 yaml.boolean_representation = ['False', 'True'] r = yaml.load(contents) convert = '''\ acos acosh asin asinh atan atanh cos cosh digamma erf erfc erfinv exp expm1 exp2 lgamma log log10 log1p log2 reciprocal sigmoid sin sinc sinh special_entr sqrt tan tanh'''.split() for e in r: f = NativeFunction.from_yaml(e, Location("", 0)) if f.structured or f.structured_delegate is not None: continue n = f.func.name.name.base if n not in convert: continue # mutate e to make changes if f.func.kind() == SchemaKind.out: e.insert(1, 'structured', True) e.insert(2, 'structured_inherits', 'TensorIteratorBase') else: # TODO: The .out overload assumption is not sound in general e.insert(1, 'structured_delegate', f'{n}.out') e['dispatch'].pop('CPU', None) e['dispatch'].pop('CUDA', None) e['dispatch'].pop('CPU, CUDA', None) e['dispatch'].pop('CompositeExplicitAutograd', None) *_, last_k = e.keys() needs_fixup = False if not e['dispatch']: if last_k == 'dispatch': needs_fixup = True del e['dispatch'] # Manually fix up newlines at the end, because ruamel # made some bad life choices about where to associate trailing # whitespace for nested dicts; see # https://stackoverflow.com/questions/42172399/modifying-yaml-using-ruamel-yaml-adds-extra-new-lines if needs_fixup: *_, last_k = e.keys() # post_key, pre_key, post_value, pre_value e.ca.items[last_k] = [None, None, CommentToken('\n\n', CommentMark(0), None), None] with open("aten/src/ATen/native/native_functions.yaml.new", "w") as f: yaml.dump(r, f) ``` Signed-off-by: Edward Z. Yang <[email protected]> Differential Revision: [D27777769](https://our.internmc.facebook.com/intern/diff/D27777769) [ghstack-poisoned]

facebook-github-bot · 2021-04-15T23:07:02Z

@ezyang merged this pull request in f17c9ea.

Summary: Pull Request resolved: pytorch#56082 The native_functions.yaml changes were done by codemod using the following script: ``` import ruamel.yaml from ruamel.yaml.tokens import CommentToken from ruamel.yaml.error import CommentMark from tools.codegen.model import * # noqa: F403 with open("aten/src/ATen/native/native_functions.yaml", "r") as f: contents = f.read() yaml = ruamel.yaml.YAML() yaml.preserve_quotes = True yaml.width = 1000 yaml.boolean_representation = ['False', 'True'] r = yaml.load(contents) convert = '''\ acos acosh asin asinh atan atanh cos cosh digamma erf erfc erfinv exp expm1 exp2 lgamma log log10 log1p log2 reciprocal sigmoid sin sinc sinh special_entr sqrt tan tanh'''.split() for e in r: f = NativeFunction.from_yaml(e, Location("", 0)) if f.structured or f.structured_delegate is not None: continue n = f.func.name.name.base if n not in convert: continue # mutate e to make changes if f.func.kind() == SchemaKind.out: e.insert(1, 'structured', True) e.insert(2, 'structured_inherits', 'TensorIteratorBase') else: # TODO: The .out overload assumption is not sound in general e.insert(1, 'structured_delegate', f'{n}.out') e['dispatch'].pop('CPU', None) e['dispatch'].pop('CUDA', None) e['dispatch'].pop('CPU, CUDA', None) e['dispatch'].pop('CompositeExplicitAutograd', None) *_, last_k = e.keys() needs_fixup = False if not e['dispatch']: if last_k == 'dispatch': needs_fixup = True del e['dispatch'] # Manually fix up newlines at the end, because ruamel # made some bad life choices about where to associate trailing # whitespace for nested dicts; see # https://stackoverflow.com/questions/42172399/modifying-yaml-using-ruamel-yaml-adds-extra-new-lines if needs_fixup: *_, last_k = e.keys() # post_key, pre_key, post_value, pre_value e.ca.items[last_k] = [None, None, CommentToken('\n\n', CommentMark(0), None), None] with open("aten/src/ATen/native/native_functions.yaml.new", "w") as f: yaml.dump(r, f) ``` Signed-off-by: Edward Z. Yang <[email protected]> Test Plan: Imported from OSS Reviewed By: bhosmer Differential Revision: D27777769 Pulled By: ezyang fbshipit-source-id: 1ecbac7cb3e0093167bb61c7d2b1ecb95b8ae17c

ezyang mentioned this pull request Apr 14, 2021

Change all unary functions stubs to use TensorIteratorBase& #56078

Closed

facebook-github-bot added the cla signed label Apr 14, 2021

ezyang requested review from SplitInfinity, bdhirsh, bhosmer and wenleix April 14, 2021 21:52

bhosmer approved these changes Apr 14, 2021

View reviewed changes

wenleix approved these changes Apr 14, 2021

View reviewed changes

bdhirsh approved these changes Apr 15, 2021

View reviewed changes

This was referenced Apr 15, 2021

Migrate acos to structured kernel #56139

Closed

Port kernels to be structured [tracker] #55070

Closed

Port all non-float unary operators to structured (and rsqrt) #56151

Closed

ezyang added 4 commits April 15, 2021 10:01

facebook-github-bot closed this in f17c9ea Apr 15, 2021

facebook-github-bot added the Merged label Apr 15, 2021

facebook-github-bot deleted the gh/ezyang/1002/head branch April 19, 2021 14:15

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Port all unary float functions to structured #56082

Port all unary float functions to structured #56082

Uh oh!

ezyang commented Apr 14, 2021 •

edited

Loading

Uh oh!

facebook-github-bot commented Apr 14, 2021 •

edited

Loading

Uh oh!

bhosmer Apr 14, 2021

Uh oh!

ezyang Apr 14, 2021

Uh oh!

bhosmer Apr 14, 2021

Uh oh!

ezyang Apr 15, 2021

Uh oh!

bdhirsh Apr 15, 2021 •

edited

Loading

Uh oh!

wenleix Apr 14, 2021

Uh oh!

ezyang Apr 15, 2021

Uh oh!

facebook-github-bot commented Apr 15, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

Port all unary float functions to structured #56082

Port all unary float functions to structured #56082

Uh oh!

Conversation

ezyang commented Apr 14, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

facebook-github-bot commented Apr 14, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

💊 CI failures summary and remediations

🕵️ 2 new failures recognized by patterns

pytorch_doc_test (1/2)

pytorch_python_doc_build (2/2)

❄️ 2 failures tentatively classified as flaky

binary_linux_libtorch_3_7m_cpu_devtoolset7_shared-with-deps_build (1/2)

binary_linux_libtorch_3_7m_cpu_gcc5_4_cxx11-abi_shared-with-deps_build (2/2)

Uh oh!

bhosmer Apr 14, 2021

Choose a reason for hiding this comment

Uh oh!

ezyang Apr 14, 2021

Choose a reason for hiding this comment

Uh oh!

bhosmer Apr 14, 2021

Choose a reason for hiding this comment

Uh oh!

ezyang Apr 15, 2021

Choose a reason for hiding this comment

Uh oh!

bdhirsh Apr 15, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

wenleix Apr 14, 2021

Choose a reason for hiding this comment

Uh oh!

ezyang Apr 15, 2021

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot commented Apr 15, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

ezyang commented Apr 14, 2021 •

edited

Loading

facebook-github-bot commented Apr 14, 2021 •

edited

Loading

bdhirsh Apr 15, 2021 •

edited

Loading