add utitlity to visualize datapipe graphs #330

pmeier · 2022-03-25T10:54:38Z

Closes #299.

ejguan

Noob question: Wondering do we have a potential extensibility using graphviz to visualize the attributes for each Node?

torchdata/datapipes/utils/_visualization.py

pmeier · 2022-03-25T19:52:25Z

Wondering do we have a potential extensibility using graphviz to visualize the attributes for each Node?

Sure. Everything we put in Node.__str__ will be displayed. For example, we could add the name of the functions of a Filter or Mapper to the node. Doing this would turn

from torchdata.datapipes.iter import IterableWrapper


def no_map(x):
    return x


def no_filter(x):
    return True


dp1, dp2, dp3 = IterableWrapper([]).fork(3)
dp1 = dp1.map(no_map)
dp2 = dp2.filter(no_filter)
dp = dp1.zip(dp2, dp3)

into

This is why I suggested #284 (comment). This would reduce Node.__str__ to return str(self.dp).

pmeier · 2022-03-28T17:06:22Z

@ejguan in #237 (comment)

Wondering what would be the visualized graph with such circular reference?

from torchdata.datapipes.iter import IterDataPipe, IterableWrapper

class CustomIterDataPipe(IterDataPipe):
    def noop(self, x):
        return x + 1

    def __init__(self):
        self._dp = IterableWrapper([1, 2, 4]).map(self.noop)

    def __iter__(self):
        yield from self._dp

Due to the circular reference, we either need to use traverse(..., only_datapipe=True) as suggested in #330 (comment) or fix the traversal similar to what I proposed in #237 (comment). Otherwise, I don't think we need extra handling for "circular" datapipes.

ejguan

This looks good to me.
@NivekT What do you think? We have inline doc for to_graph. Do we need an example for users shown in README or another place?

docs/source/conf.py

NivekT

Sorry for not getting back to this earlier. Overall, it looks very good to me!

I tried playing around with it to see if there are cases where the output is not sensible. This is the only case where things may be confusing, and I think we may want to show the ChildDataPipe (perhaps renamed to something else like DemuxChild1 or something). Admittedly, I think the code snippet below may not be representative of the usual use cases of DataPipes in a pipeline:

from torchdata.datapipes.iter import IterableWrapper
from torchdata.datapipes.utils import to_graph

dp = IterableWrapper(range(10))
dp = dp.map(lambda x: x + 1)
dp = dp.filter(lambda _: True)
dp1, dp2 = dp.demux(num_instances=2, classifier_fn=lambda x: x % 0 == 0)
dp3 = dp1.zip(dp2)
g = to_graph(dp3)
g.view()

We can also apply the function on a few torchvision data pipelines and see what they look like.

pmeier · 2022-05-27T12:18:05Z

I wrote a small script to test this on the torchvision datasets. Assuming this PR is merged and you have a nightly from torchdata and torchvision installed, this is the result.

very easy

CIFAR10

easy

MNIST

medium

VOC

difficult

COCO

CelebA

torchdata/datapipes/utils/_visualization.py

NivekT

I wrote a small script to test this on the torchvision datasets. Assuming this PR is merged and you have a nightly from torchdata and torchvision installed, this is the result.

very easy

CIFAR10

easy

MNIST

medium

VOC

difficult

COCO
CelebA

Those graphs look complicated but sensible to me. Unless there are ways to simplify the pre-processing logic, I think they look good. I think the remaining part is the edge case with _ChildDataPipes combining back together immediately.

pmeier · 2022-06-03T09:30:20Z

@NivekT I've added a debug: bool = False flag as explained in #330 (comment). This gives the following results for your example in #330 (review):

`debug=False` (default)

`debug=True`

`

NivekT

This PR is good to land as it is. I will follow up with a different PR of documentations.

facebook-github-bot · 2022-06-07T23:43:23Z

@NivekT has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

Summary: Closes #299. Pull Request resolved: #330 Reviewed By: ejguan Differential Revision: D36990978 Pulled By: NivekT fbshipit-source-id: 423a52e6d59da59826d1422676d28287537dd902

…507) * add utility to visualize datapipe graphs (#330) Summary: Closes #299. Pull Request resolved: #330 Reviewed By: ejguan Differential Revision: D36990978 Pulled By: NivekT fbshipit-source-id: 423a52e6d59da59826d1422676d28287537dd902 * Update examples and tutorial after automatic sharding has landed (#505) Summary: Pull Request resolved: #505 After the `DataLoader` automatic sharding for DataPipe PR has landed in core, this PR is updating the relevant example and tutorial Test Plan: Imported from OSS Reviewed By: VitalyFedyunin, ejguan Differential Revision: D37040383 Pulled By: NivekT fbshipit-source-id: 446b9e6a0ae804b1a0908a26058680f9b3c2f080 Co-authored-by: Philip Meier <github.pmeier@posteo.de>

add utitlity to visualize datapipe graphs

6baad98

facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Mar 25, 2022

appease mypy

e93f052

pmeier requested a review from NivekT March 25, 2022 13:11

pmeier marked this pull request as ready for review March 25, 2022 13:11

ejguan reviewed Mar 25, 2022

View reviewed changes

torchdata/datapipes/utils/_visualization.py Outdated Show resolved Hide resolved

ejguan approved these changes Apr 20, 2022

View reviewed changes

docs/source/conf.py Show resolved Hide resolved

NivekT reviewed May 26, 2022

View reviewed changes

pmeier added 2 commits May 27, 2022 13:36

Merge branch 'main' into graphviz

3aa130a

set only_datapipe=True for datapipes with circular references

396bae0

pmeier commented May 27, 2022

View reviewed changes

torchdata/datapipes/utils/_visualization.py Show resolved Hide resolved

NivekT reviewed May 31, 2022

View reviewed changes

pmeier added 2 commits June 3, 2022 11:19

Merge branch 'main' into graphviz

df43bc0

add debug flag

14f5fec

NivekT approved these changes Jun 7, 2022

View reviewed changes

facebook-github-bot closed this in 2cc539b Jun 8, 2022

pmeier deleted the graphviz branch June 9, 2022 06:40

NivekT mentioned this pull request Jun 9, 2022

[v0.4.0] Release Tracker #448

Closed

NivekT mentioned this pull request Jun 9, 2022

[Release 0.4.0] Adding utility function to visualize DataPipe graphs and update sharding tutorial #507

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

add utitlity to visualize datapipe graphs #330

add utitlity to visualize datapipe graphs #330

pmeier commented Mar 25, 2022

ejguan left a comment

pmeier commented Mar 25, 2022

pmeier commented Mar 28, 2022

ejguan left a comment

NivekT left a comment

pmeier commented May 27, 2022

NivekT left a comment •

edited

Loading

pmeier commented Jun 3, 2022

NivekT left a comment

facebook-github-bot commented Jun 7, 2022

add utitlity to visualize datapipe graphs #330

add utitlity to visualize datapipe graphs #330

Conversation

pmeier commented Mar 25, 2022

ejguan left a comment

Choose a reason for hiding this comment

pmeier commented Mar 25, 2022

pmeier commented Mar 28, 2022

ejguan left a comment

Choose a reason for hiding this comment

NivekT left a comment

Choose a reason for hiding this comment

pmeier commented May 27, 2022

very easy

easy

medium

difficult

NivekT left a comment • edited Loading

Choose a reason for hiding this comment

very easy

easy

medium

difficult

very easy

easy

medium

difficult

pmeier commented Jun 3, 2022

debug=False (default)

debug=True

NivekT left a comment

Choose a reason for hiding this comment

facebook-github-bot commented Jun 7, 2022

NivekT left a comment •

edited

Loading

`debug=False` (default)

`debug=True`