Pipeline decorator #2629

szalpal · 2021-01-22T14:39:30Z

Signed-off-by: Michał Szołucha mszolucha@nvidia.com

This PR adds new feature to the API: Pipeline decorator. It eases up pipeline creation and likely will help users, especially those who are new to DALI

Why we need this PR?

Pick one, remove the rest

It adds new feature decorator for DALI pipeline functions

What happened in this PR?

Fill relevant points, put NA otherwise. Replace anything inside []

What solution was applied:
added decorator
Affected modules and functionalities:
pipeline.py
Key points relevant for the review:
NA
Validation and testing:
YEs
Documentation (including examples):
Yes

JIRA TASK: [Use DALI-XXXX or NA]

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

szalpal · 2021-01-27T02:15:12Z

!build

dali-automaton · 2021-01-27T03:08:25Z

CI MESSAGE: [2018301]: BUILD STARTED

dali-automaton · 2021-01-27T05:22:53Z

CI MESSAGE: [2018301]: BUILD PASSED

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

jantonguirao · 2021-01-29T11:45:32Z

dali/python/nvidia/dali/pipeline.py

@@ -990,3 +996,62 @@ def iter_setup(self):
        For example, one can use this function to feed the input
        data from NumPy arrays."""
        pass
+
+
+def _discriminate_args(**kwargs):


Suggested change

def _discriminate_args(**kwargs):

def _separate_args(**kwargs):

or

Suggested change

def _discriminate_args(**kwargs):

def _split_args(**kwargs):

jantonguirao · 2021-01-29T11:48:10Z

dali/python/nvidia/dali/pipeline.py

+    **Usage**
+
+    First, implement a function, that defines a DALI pipeline.
+    Such function shall return DALI's DataNodes. These return


Suggested change

Such function shall return DALI's DataNodes. These return

Such function shall return one or more DALI DataNodes, representing the outputs of the pipeline.

jantonguirao · 2021-01-29T11:49:25Z

dali/python/nvidia/dali/pipeline.py

+    return ctor_args, fn_args
+
+
+def pipeline(fn=None, **pipeline_kwargs):


I think this name can be misleading because we have
nvidia.dali.pipeline.Pipeline (the pipeline class) and nvidia.dali.pipeline.pipeline the decorator.
How about renaming to pipeline_def to avoid confusion?

Yes, I agree, pipeline_def or def_pipeline, or def_pipe or make_pipeline. Whatever to no have three different nvidia.dali.pipeline.[Pp]ipeline things.

Personally, I'm fine with pipeline, because we do already have the Pipeline in the same package and place. So we do already:

from nvidia.dali.pipeline import Pipeline pipe = Pipeline(...)

Why not doing

from nvidia.dali.pipeline import pipeline @pipeline def pipe(): ...

If you make a mistake there, the interpreter will boldly tell you about it

Let's leave more sophisticated naming for the user:

import nvidia.dali.pipeline.pipeline as pipeline_decorator import nvidia.dali.pipeline.Pipeline as pipeline_object

jantonguirao · 2021-01-29T11:49:53Z

dali/python/nvidia/dali/pipeline.py

+        my_pipe = pipe(0, batch_size=32, num_threads=1, device_id=0, flip_horizontal=1)
+        my_pipe.build()
+
+    Additionally, decorator can accept Pipeline constructor


Suggested change

Additionally, decorator can accept Pipeline constructor

Additionally, the decorator can accept Pipeline constructor

jantonguirao · 2021-01-29T11:51:55Z

dali/test/python/test_pipeline_decorator.py

+    return pipeline
+
+
+@nottest


why do you need nottest? The function won't be run automatically (not prefixed with test_). If you need that, why is it not there in reference_pipeline?

To test, that pipeline decorator works with other decorators

It's not obvious, maybe add a comment?

jantonguirao · 2021-01-29T11:53:52Z

docs/examples/use_cases/pytorch/resnet50/main.py

@@ -20,7 +20,7 @@

 try:
    from nvidia.dali.plugin.pytorch import DALIClassificationIterator, LastBatchPolicy
-    from nvidia.dali.pipeline import Pipeline
+    from nvidia.dali.pipeline import pipeline


have we decided that we want to replace examples?

I'm replacing subset of examples

jantonguirao · 2021-01-29T11:54:34Z

docs/examples/use_cases/pytorch/resnet50/main.py

-        pipeline.set_outputs(images, labels)
-    return pipeline
+
+@pipeline


If we decide to go ahead with changing the examples, I'd use here the whole name:
@nvidia.dali.pipeline.pipeline_def or @nvidia.dali.pipeline.pipeline

In the examples we use Pipeline instead of nvidia.dali.pipeline.Pipeline, so I believe it's better to use pipeline_def instead of nvidia.dali.pipeline.pipeline_def

jantonguirao · 2021-01-29T11:55:11Z

docs/examples/use_cases/pytorch/single_stage_detector/src/coco_pipeline.py

@@ -14,77 +14,76 @@


 import torch
-from nvidia.dali.pipeline import Pipeline
+from nvidia.dali.pipeline import pipeline


same as in the previous file, do we want to replace our examples to use this?

I'm replacing subset of examples

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

szalpal · 2021-02-01T16:23:11Z

Attaching a screenshot with documentation prerender:

szalpal · 2021-02-01T16:36:06Z

!build

dali-automaton · 2021-02-01T16:42:10Z

CI MESSAGE: [2033757]: BUILD STARTED

dali/python/nvidia/dali/pipeline.py

klecki · 2021-02-01T16:44:36Z

docs/examples/use_cases/pytorch/resnet50/main.py

-        pipeline.set_outputs(images, labels)
-    return pipeline
+
+@pipeline


Having a more distinct name from the pipeline module and Pipeline object would remove the need for the full blown specification here.

klecki · 2021-02-01T16:48:23Z

dali/python/nvidia/dali/pipeline.py

+
+def pipeline(fn=None, **pipeline_kwargs):
+    """
+    Decorator for wrapping functions that define DALI pipelines.


Maybe add something like: Decorated functions becomes a pipeline factory? Or Decorated function will return instances of Pipeline with processing defined by the function.

The latter - or sth like:

Suggested change

Decorator for wrapping functions that define DALI pipelines.

A decorator which creates a factory of DALI pipelines whose processing graph is defined by the decorated function.

klecki · 2021-02-01T16:51:54Z

dali/python/nvidia/dali/pipeline.py

+                _ = pipe_outputs if isinstance(pipe_outputs, tuple) else pipe_outputs,
+                pipe.set_outputs(*_)


PLZ NO.
_ is for ignored values.

dali-automaton · 2021-02-01T18:27:22Z

CI MESSAGE: [2033757]: BUILD FAILED

klecki · 2021-02-01T18:33:13Z

dali/python/nvidia/dali/pipeline.py

+    **Usage**
+
+    First, implement a function, that defines a DALI pipeline.
+    Such function shall return DALI's DataNodes. These return


It can return not only DataNodes.

I'm missing the information, that this decorator will turn this function into something like Pipeline factory. And that the returns of this function become output of the Pieline (for example .run()) and the function will return Pipeline instances.

klecki · 2021-02-01T18:34:22Z

dali/python/nvidia/dali/pipeline.py

+            data = fn.external_source(source=my_generator)
+            return data
+
+        my_pipe = pipe(batch_size=128)  # batch_size=128 overwrites batch_size=32


I'm also thinking if the doc should mention that this is equivalent to using with pipe: and set_outputs?

klecki · 2021-02-01T18:35:13Z

docs/pipeline.rst

@@ -71,3 +84,9 @@ DataNode
 --------
 .. autoclass:: DataNode
   :members:
+
+
+Decorator


This subtitle is subpar.

Suits me good.

Not good at all.
Something like Pipeline factor decorator would be more appropriate.

Suggested change

Decorator

Pipeline factory decorator

or

Suggested change

Decorator

Pipeline definition decorator

would be better ideas.

mzient · 2021-02-02T13:19:30Z

dali/python/nvidia/dali/pipeline.py

+    """
+    Decorator for wrapping functions that define DALI pipelines.
+
+    **Usage**


I think that some description should go before usage/example:

``pipeline_def`` returns a function that creates a pipeline, where the processing graph is defined by the function ``fn`` passed as the first positional argument. It can be used as a decorator, in which case the decorated function becomes a pipeline factory. The function returned is roughly equivalent to:: def factory(arg1, arg2, ...) pipe = nvidia.dali.Pipeline() with pipe: pipe.set_outputs(*fn(arg1, arg2, ...)) return pipe

I believe that all this is written when you google "how to write a decorator?". So I wouldn't add it at the beginning, but I added it at the end of the paragraph, cause this note is in fact valuable.

Sorry, but.... how can you possibly google that a decorator creates a DALI pipeline, calls your function within the with-scope of the newly created pipeline and calls set_outputs with the return values of the function you supplied???
Perhaps it should be written as text, not as code.

I think the doc is fine, we shouldn't use multi compound statements, it is hard to follow.

I was thinking whether to move the equivalent code to the top, but I'm not sure if it's better than usage examples.

mzient · 2021-02-02T13:27:53Z

dali/python/nvidia/dali/pipeline.py

+def _discriminate_args(**kwargs):
+    """Split args on those applicable to Pipeline constructor and the rest."""
+    fca = inspect.getfullargspec(Pipeline.__init__)  # Full Ctor Args
+    ctor_args = dict(filter(lambda x: x[0] in fca.args or x[0] in fca.kwonlyargs, kwargs.items()))


This should be the other way round: detect arguments for this function - if not found, try Pipeline - if not found either, raise an error.
Rationale: if we add new arguments to Pipeline's constructor, we may break users' code because a named argument would be redirected to the Pipeline constructor instead.

Done with one exception: there's be a problem with the error part. I believe we'd like this to be legal:

@pipeline_def def peculiar_pipeline(operator=fn.old_color_twist, **operator_kwargs): return operator(**operator_kwargs) p = peculiar_pipeline(operator=fn.color_twist, brightness=5)

So the remaining args we still need to assign to the func_args

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

mzient · 2021-02-05T17:00:48Z

dali/python/nvidia/dali/pipeline.py

+            pipe = Pipeline(**{**pipeline_kwargs, **ctor_args})  # Merge and overwrite dict
+            with pipe:
+                pipe_outputs = func(*args, **fn_kwargs)
+                po = pipe_outputs if isinstance(pipe_outputs, tuple) else (pipe_outputs,)


Suggested change

po = pipe_outputs if isinstance(pipe_outputs, tuple) else (pipe_outputs,)

po = pipe_outputs if isinstance(pipe_outputs, tuple) else () if pipe outputs is None else (pipe_outputs,)

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

mzient · 2021-02-08T11:26:25Z

dali/python/nvidia/dali/pipeline.py

@@ -990,3 +996,93 @@ def iter_setup(self):
        For example, one can use this function to feed the input
        data from NumPy arrays."""
        pass
+
+
+def _discriminate_args(func, **func_kwargs):


I wonder how it plays with political correctness ;P

I mentioned it before but the comment got lost.

I'd use "split" or "separate", discriminate might carry a negative connotation.

There is a function def _separate_kwargs(kwargs): in ops.py. Isn't this exactly what you need? (haven't read it)

There's a discriminator in electronics, mathematics, GANs also have a discriminator. I wouldn't like to be oversensitive - should we avoid to use "gap", just because there is a wage-gap?

(sacrasm warning)
After all, that's precisely what this function does - treats function args better than ctor args ;)

Achievement unlocked: troll ;)

How about the existing function I pointed out? Is it a duplicate or not?

mzient · 2021-02-08T11:40:07Z

dali/python/nvidia/dali/pipeline.py

+        if farg[0] in func_argspec.args or farg[0] in func_argspec.kwonlyargs:
+            fn_args[farg[0]] = farg[1]
+        elif farg[0] in ctor_argspec.args or farg[0] in ctor_argspec.kwonlyargs:
+            ctor_args[farg[0]] = farg[1]


How about a warning when shadowing?

Suggested change

if farg[0] in func_argspec.args or farg[0] in func_argspec.kwonlyargs:

fn_args[farg[0]] = farg[1]

elif farg[0] in ctor_argspec.args or farg[0] in ctor_argspec.kwonlyargs:

ctor_args[farg[0]] = farg[1]

is_fn_arg = farg[0] in func_argspec.args or farg[0] in func_argspec.kwonlyargs

is_pipe_arg = farg[0] in ctor_argspec.args or farg[0] in ctor_argspec.kwonlyargs

if is_fn_arg:

if is_pipe_arg:

print("Warning: the argument `{}` shadows a Pipeline constructor argument of the same name.".format(farg[0]))

fn_args[farg[0]] = farg[1]

elif is_pipe_arg:

ctor_args[farg[0]] = farg[1]

mzient · 2021-02-08T11:48:05Z

dali/python/nvidia/dali/pipeline.py

+
+        pipe = my_pipe(batch_size=128)  # batch_size=128 overrides batch_size=32
+
+    .. warning::


Suggested change

.. warning::

.. warning::

Avoid the use of `**kwargs` in the graph definition function, since it may result in unwanted,

silent hijacking of some arguments of the same name by Pipeline constructor. Code written

this way may cease to work with future versions of DALI when new parameters are added

to the Pipeline constructor.

.. warning::

Can't we just disallow the **kwargs completely for the function that is to be decorated and raise an error. I'd say play safe rather than write lengthy manuals. If we find a better way or this is an issue we can always relax that check (and doing the other way would be hard).

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

jantonguirao · 2021-02-10T09:14:45Z

dali/python/nvidia/dali/pipeline.py

@@ -990,3 +996,93 @@ def iter_setup(self):
        For example, one can use this function to feed the input
        data from NumPy arrays."""
        pass
+
+
+def _discriminate_args(func, **func_kwargs):


I mentioned it before but the comment got lost.

I'd use "split" or "separate", discriminate might carry a negative connotation.

There is a function def _separate_kwargs(kwargs): in ops.py. Isn't this exactly what you need? (haven't read it)

jantonguirao · 2021-02-10T09:23:30Z

dali/test/python/test_pipeline_decorator.py

+    return pipeline
+
+
+@nottest


It's not obvious, maybe add a comment?

jantonguirao · 2021-02-10T09:25:13Z

dali/test/python/test_pipeline_decorator.py

+
+
+
+@nottest


in other test suites we use check_. You wouldn't have to mark it as nottest if you follow this suggestion

check_ is vague. When you see it, it doesn't answer a question, why is there a function called in the same way, just check_ instead of test_? And @nottest gives you a clear answer ;)

jantonguirao · 2021-02-10T09:26:41Z

dali/test/python/test_pipeline_decorator.py

+            yield test_pipeline_static, vert, hori
+            yield test_pipeline_runtime, vert, hori
+            yield test_pipeline_override, vert, hori, 16
+    yield test_pipeline_runtime, fn.random.coin_flip(seed=123), fn.random.coin_flip(seed=234)


does this work? can we pass data nodes from outside with pipe: ... ?

Apparently yes...

We can if they don't have side-effects.

jantonguirao · 2021-02-10T09:31:29Z

General comment: Please use the template for the PR description (answering the not relevant questions with N/A if needed)

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

mzient · 2021-02-10T12:17:22Z

dali/python/nvidia/dali/pipeline.py

+    """Split args on those applicable to Pipeline constructor and the decorated function."""
+    func_argspec = inspect.getfullargspec(func)
+    ctor_argspec = inspect.getfullargspec(Pipeline.__init__)
+


Suggested change

if func_argspec.varkw is not None:

raise ValueError("Use of variadic keyword argument `**{}` in graph definition function is not allowed.".format(func_argspec.varkw))

I agree that this is probably the simplest and cleanest solution, that we can easily document - just say that all the variadic kwargs are used for forwarding arguments to the pipeline and the original function is not supposed to use it/have it as we may steal them with changes to Pipeline class.

mzient · 2021-02-10T12:19:50Z

dali/python/nvidia/dali/pipeline.py

+            raise TypeError(
+                "Using non-explicitly declared arguments in graph-defining function is not allowed. "
+                "Please remove `{}` argument or declare it explicitly in the function signature.".format(
+                    farg[0]))


Suggested change

raise TypeError(

"Using non-explicitly declared arguments in graph-defining function is not allowed. "

"Please remove `{}` argument or declare it explicitly in the function signature.".format(

farg[0]))

raise ValueError("`{}` is neither an argument of the graph definition function nor of `Pipeline.__init__`".format(farg[0]))

We can explicitly check **kwargs (see above), but we can still produce a nice error here - but with a slightly different descirption.

mzient · 2021-02-10T12:21:32Z

dali/python/nvidia/dali/pipeline.py

+        Pipeline constructor. Code written this way may cease to work with future versions of DALI
+        when new parameters are added to the Pipeline constructor.


Suggested change

Pipeline constructor. Code written this way may cease to work with future versions of DALI

when new parameters are added to the Pipeline constructor.

Pipeline constructor. Code written this way would cease to work with future versions of DALI

when new parameters are added to the Pipeline constructor.

It's not allowed, so this explanation is describes a purely hypothetical scenario.

mzient · 2021-02-10T12:26:35Z

dali/test/python/test_pipeline_decorator.py

+def test_kwargs_exception_1():
+    pipeline_kwargs(1, arg2=2)
+
+
+def test_kwargs_exception_2():
+    pipeline_kwonlyargs(1, arg2=2)


These should raise an exception for mere presence of **kwargs.

klecki · 2021-02-10T15:39:05Z

dali/python/nvidia/dali/pipeline.py

+        elif is_ctor_arg:
+            ctor_args[farg[0]] = farg[1]
+        else:
+            fn_args[farg[0]] = farg[1]


Why not just raise here? If we disable kwargs that would save us from guessing as it would be an explicit error.

klecki

I would disallow the **kwargs from the original signature, otherwise ok.

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

mzient · 2021-02-11T10:34:19Z

docs/pipeline.rst

-#. by inheriting from Pipeline class and overriding :meth:`Pipeline.define_graph`
-#. by instantiating `Pipeline` directly, building the graph and setting the pipeline
-   outputs with :meth:`Pipeline.set_outputs`
+#. by implementing a function that uses DALI's ``Operators`` inside and decorating it with the


Suggested change

#. by implementing a function that uses DALI's ``Operators`` inside and decorating it with the

#. by implementing a function that uses DALI operators inside and decorating it with the

mzient

One nitpick, otherwise looks good.

jantonguirao · 2021-02-11T17:11:32Z

dali/python/nvidia/dali/pipeline.py

+            fn_args[farg[0]] = farg[1]
+            if is_ctor_arg:
+                print(
+                    "Warning: the argument `{}` shadows a Pipeline constructor argument of the same name.".format(


Consider

Suggested change

"Warning: the argument `{}` shadows a Pipeline constructor argument of the same name.".format(

f"Warning: the argument `{farg[0]}` shadows a Pipeline constructor argument of the same name."

jantonguirao · 2021-02-11T17:13:06Z

dali/python/nvidia/dali/pipeline.py

+        elif is_ctor_arg:
+            ctor_args[farg[0]] = farg[1]
+        else:
+            assert False, "This shouldn't happen. Please double-check the `{}` argument".format(farg[0])


Consider

Suggested change

assert False, "This shouldn't happen. Please double-check the `{}` argument".format(farg[0])

raise RuntimeError(f"This shouldn't happen. Please double-check the `{farg[0]}` argument")

jantonguirao

LGTM, except for some minor suggestions

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

szalpal · 2021-02-12T10:36:37Z

!build

dali-automaton · 2021-02-12T10:41:49Z

CI MESSAGE: [2070571]: BUILD STARTED

dali-automaton · 2021-02-12T12:05:43Z

CI MESSAGE: [2070571]: BUILD PASSED

szalpal added 4 commits January 26, 2021 16:09

pipeline_decorator

b1a4eb3

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

new idea: ctor args with pipeline args

2843e1d

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

next version

a1d7e57

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

clean version

484d370

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

szalpal force-pushed the pipeline_decorator branch from bbfea0e to 484d370 Compare January 27, 2021 02:14

add docs

7e0bd2d

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

szalpal marked this pull request as ready for review January 27, 2021 10:32

jantonguirao assigned jantonguirao and klecki Jan 27, 2021

szalpal added 2 commits January 27, 2021 12:22

remove extra check

3dfd20a

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

adding some examples

32374e7

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

jantonguirao reviewed Jan 29, 2021

View reviewed changes

adding tests for variable batch szie

22d499f

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

klecki reviewed Feb 1, 2021

View reviewed changes

dali/python/nvidia/dali/pipeline.py Outdated Show resolved Hide resolved

klecki reviewed Feb 1, 2021

View reviewed changes

mzient reviewed Feb 2, 2021

View reviewed changes

review

978f69f

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

mzient reviewed Feb 5, 2021

View reviewed changes

review

0a53386

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

mzient reviewed Feb 8, 2021

View reviewed changes

review

3fe6694

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

jantonguirao reviewed Feb 10, 2021

View reviewed changes

szalpal added 2 commits February 10, 2021 12:42

forbid kwargs

9b687e1

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

reverting variable bs test change - won't work with disallowing kwargs

12c01d8

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

mzient reviewed Feb 10, 2021

View reviewed changes

klecki reviewed Feb 10, 2021

View reviewed changes

review

a1ad829

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

mzient reviewed Feb 11, 2021

View reviewed changes

mzient approved these changes Feb 11, 2021

View reviewed changes

klecki approved these changes Feb 11, 2021

View reviewed changes

jantonguirao reviewed Feb 11, 2021

View reviewed changes

jantonguirao approved these changes Feb 11, 2021

View reviewed changes

nitpick

932f1de

Signed-off-by: Michał Szołucha <mszolucha@nvidia.com>

szalpal merged commit 5ec7c5f into NVIDIA:master Feb 12, 2021

JanuszL mentioned this pull request May 19, 2021

DALI 2021 roadmap #2978

Closed

szalpal deleted the pipeline_decorator branch February 9, 2024 00:25

	def _discriminate_args(**kwargs):
	def _separate_args(**kwargs):

	Such function shall return DALI's DataNodes. These return
	Such function shall return one or more DALI DataNodes, representing the outputs of the pipeline.

		return ctor_args, fn_args


		def pipeline(fn=None, **pipeline_kwargs):

	Additionally, decorator can accept Pipeline constructor
	Additionally, the decorator can accept Pipeline constructor

	Decorator for wrapping functions that define DALI pipelines.
	A decorator which creates a factory of DALI pipelines whose processing graph is defined by the decorated function.

		_ = pipe_outputs if isinstance(pipe_outputs, tuple) else pipe_outputs,
		pipe.set_outputs(*_)

	po = pipe_outputs if isinstance(pipe_outputs, tuple) else (pipe_outputs,)
	po = pipe_outputs if isinstance(pipe_outputs, tuple) else () if pipe outputs is None else (pipe_outputs,)


		pipe = my_pipe(batch_size=128) # batch_size=128 overrides batch_size=32

		.. warning::


	if func_argspec.varkw is not None:
	raise ValueError("Use of variadic keyword argument `**{}` in graph definition function is not allowed.".format(func_argspec.varkw))

		Pipeline constructor. Code written this way may cease to work with future versions of DALI
		when new parameters are added to the Pipeline constructor.




		@nottest

	#. by implementing a function that uses DALI's ``Operators`` inside and decorating it with the
	#. by implementing a function that uses DALI operators inside and decorating it with the

	"Warning: the argument `{}` shadows a Pipeline constructor argument of the same name.".format(
	f"Warning: the argument `{farg[0]}` shadows a Pipeline constructor argument of the same name."

	assert False, "This shouldn't happen. Please double-check the `{}` argument".format(farg[0])
	raise RuntimeError(f"This shouldn't happen. Please double-check the `{farg[0]}` argument")

Pipeline decorator #2629

Pipeline decorator #2629

Conversation

szalpal commented Jan 22, 2021 • edited Loading

Why we need this PR?

What happened in this PR?

szalpal commented Jan 27, 2021

dali-automaton commented Jan 27, 2021

dali-automaton commented Jan 27, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

szalpal Feb 1, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

szalpal commented Feb 1, 2021

szalpal commented Feb 1, 2021

dali-automaton commented Feb 1, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dali-automaton commented Feb 1, 2021

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mzient Feb 2, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mzient Feb 3, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

szalpal Feb 3, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mzient Feb 8, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jantonguirao commented Feb 10, 2021

Choose a reason for hiding this comment

klecki Feb 10, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

szalpal commented Jan 22, 2021 •

edited

Loading

szalpal Feb 1, 2021 •

edited

Loading

mzient Feb 2, 2021 •

edited

Loading

mzient Feb 3, 2021 •

edited

Loading

szalpal Feb 3, 2021 •

edited

Loading

mzient Feb 8, 2021 •

edited

Loading

klecki Feb 10, 2021 •

edited

Loading

mzient Feb 11, 2021 •

edited

Loading