Upgraded fastai from v1 to v2 (>=2.4.1) #4715

jinzhang21 · 2021-08-20T02:14:51Z

What changes are proposed in this pull request?

Deprecated fastai v1 and upgraded it to v2 with minimum version 2.4.1. Closes #4188

How is this patch tested?

Unit tests and integration tests.

Release Notes

Deprecated fastai v1 and upgraded it to v2 (minimum version 2.4.1).

Is this a user-facing change?

No. You can skip the rest of this section.
Yes. Give a description of this change to be included in the release notes for MLflow users.

(Details in 1-2 sentences. You can just refer to another PR with a description if this PR is part of a larger change.)

What component(s), interfaces, languages, and integrations does this PR affect?

Components

Interface

area/uiux: Front-end, user experience, plotting, JavaScript, JavaScript dev server
area/docker: Docker use across MLflow's components, such as MLflow Projects and MLflow Models
area/sqlalchemy: Use of SQLAlchemy in the Tracking Service or Model Registry
area/windows: Windows support

Language

language/r: R APIs and clients
language/java: Java APIs and clients
language/new: Proposals for new client languages

Integrations

integrations/azure: Azure and Azure ML integrations
integrations/sagemaker: SageMaker integrations
integrations/databricks: Databricks integrations

How should the PR be classified in the release notes? Choose one:

rn/breaking-change - The PR will be mentioned in the "Breaking Changes" section
rn/none - No description will be included. The PR will be mentioned only by the PR number in the "Small Bugfixes and Documentation Updates" section
rn/feature - A new user-facing feature worth mentioning in the release notes
rn/bug-fix - A user-facing bug fix worth mentioning in the release notes
rn/documentation - A user-facing documentation change worth mentioning in the release notes

Signed-off-by: Lluis <l.salord.quetglas@gmail.com>

github-actions · 2021-08-20T02:15:06Z

@mlflow Thanks for the contribution! The DCO check failed. Please sign off your commits by following the instructions here: https://github.com/mlflow/mlflow/runs/3377994134. See https://github.com/mlflow/mlflow/blob/master/CONTRIBUTING.rst#sign-your-work for more details.

Signed-off-by: jinzhang21 <jin.zhang@databricks.com>

mlflow/fastai/callback.py

mlflow/ml-package-versions.yml

mlflow/fastai/callback.py

Co-authored-by: Harutaka Kawamura <hkawamura0130@gmail.com>

examples/fastai/train.py

mlflow/fastai/callback.py

harupy · 2021-09-13T10:53:17Z

tests/fastai/test_fastai_autolog.py

+@pytest.fixture(scope="session")
+def mnist_data():
+    mnist = untar_data(URLs.MNIST_TINY)
+    return ImageDataLoaders.from_folder(mnist, num_workers=0)
+
+
+def fastai_visual_model(data, **kwargs):
+    return cnn_learner(data, models.resnet18, normalize=False, **kwargs)


I think training a ResNet model with the MNIST dataset takes a long time. Can we use a smaller dataset and a simpler model?

In this run, pytest tests/fastai/test_fastai_autolog.py --large took about 10 minutes:
https://github.com/mlflow/mlflow/pull/4715/checks?check_run_id=3583140216

I didn't check every model there, but resnet18 has "only" 11M params. It appears to be the smallest model among those offered: https://fastai1.fast.ai/vision.models.html

@jinzhang21
I found a way to create a fine-tunable tabular model. Can I push some commits?

@jinzhang21 Puhsed a commit to update the code. Now it takes about 3 min. to run tests for fastai autologging: https://github.com/mlflow/mlflow/pull/4715/checks?check_run_id=3605777911

Super nice! Thanks for the change, Haru!

tests/autologging/test_autologging_utils.py

tests/fastai/test_fastai_autolog.py

tests/fastai/test_fastai_model_export.py

mlflow/fastai/callback.py

Signed-off-by: jinzhang21 <jin.zhang@databricks.com>

jinzhang21 · 2021-09-13T20:09:15Z

There's some problem with sklearn versioning, causing errors mentioned here: https://stackoverflow.com/questions/62322882/load-iris-got-an-unexpected-keyword-argument-as-frame. Not sure how it's related to this PR or a general problem.

Signed-off-by: harupy <hkawamura0130@gmail.com>

harupy

LGTM!

lluissalord added 9 commits August 19, 2021 19:07

Update compatibility with fastai v2

566d2eb

Signed-off-by: Lluis <l.salord.quetglas@gmail.com>

Fix not recording on predict

f043d99

Signed-off-by: Lluis <l.salord.quetglas@gmail.com>

Compatibility with just released 2.3.0

9c24e60

Signed-off-by: Lluis <l.salord.quetglas@gmail.com>

Update merge with fastai v2

1365e4e

Signed-off-by: Lluis <l.salord.quetglas@gmail.com>

Add support to fine_tune function

417660b

Signed-off-by: Lluis <l.salord.quetglas@gmail.com>

Fix log on predictions and make opt logging robust

448734b

Signed-off-by: Lluis <l.salord.quetglas@gmail.com>

Add support SaveModelCallback

90c8389

Signed-off-by: Lluis <l.salord.quetglas@gmail.com>

Fix lint warnings

559e560

Signed-off-by: Lluis <l.salord.quetglas@gmail.com>

Fix log_model_info for all kind of TrackerCallback

07fda6c

jinzhang21 added 7 commits August 19, 2021 19:17

more conflicts resolved

d3df74b

more fix

7593809

fix conda env

b32e2a4

fix conda env1

08fa009

Fixed pickle issue

625e0d8

Merge master

cee7f5c

Minor changes

720668d

jinzhang21 changed the title ~~[Do not review] test PR~~ Upgraded fastai from v1 to v2 (>=2.3.0) Sep 12, 2021

github-actions bot added area/models MLmodel format, model serialization/deserialization, flavors rn/feature Mention under Features in Changelogs. labels Sep 12, 2021

Updated fastai requirement to 2.4.0+

083f7b6

Signed-off-by: jinzhang21 <jin.zhang@databricks.com>

jinzhang21 changed the title ~~Upgraded fastai from v1 to v2 (>=2.3.0)~~ Upgraded fastai from v1 to v2 (>=2.4.0) Sep 12, 2021

jinzhang21 added 2 commits September 11, 2021 22:30

Fixed the version

063ec58

Signed-off-by: jinzhang21 <jin.zhang@databricks.com>

Changed version to the released ones

63f1b07

Signed-off-by: jinzhang21 <jin.zhang@databricks.com>

jinzhang21 changed the title ~~Upgraded fastai from v1 to v2 (>=2.4.0)~~ Upgraded fastai from v1 to v2 (>=2.4.1) Sep 12, 2021

Fixed lint

1d0b859

Signed-off-by: jinzhang21 <jin.zhang@databricks.com>

harupy reviewed Sep 13, 2021

View reviewed changes

mlflow/fastai/callback.py Show resolved Hide resolved

harupy reviewed Sep 13, 2021

View reviewed changes

mlflow/ml-package-versions.yml Outdated Show resolved Hide resolved

harupy reviewed Sep 13, 2021

View reviewed changes

mlflow/fastai/callback.py Outdated Show resolved Hide resolved

harupy reviewed Sep 13, 2021

View reviewed changes

mlflow/fastai/callback.py Outdated Show resolved Hide resolved

jinzhang21 and others added 2 commits September 12, 2021 21:01

Update mlflow/fastai/callback.py

3005503

Co-authored-by: Harutaka Kawamura <hkawamura0130@gmail.com>

Update mlflow/ml-package-versions.yml

dd118c6

Co-authored-by: Harutaka Kawamura <hkawamura0130@gmail.com>

harupy reviewed Sep 13, 2021

View reviewed changes

examples/fastai/train.py Outdated Show resolved Hide resolved

jinzhang21 added 2 commits September 12, 2021 21:23

Add -2.x back

f2669a5

import only used mods

5b517cb

harupy reviewed Sep 13, 2021

View reviewed changes

mlflow/fastai/callback.py Show resolved Hide resolved

harupy reviewed Sep 13, 2021

View reviewed changes

WeichenXu123 reviewed Sep 13, 2021

View reviewed changes

tests/autologging/test_autologging_utils.py Show resolved Hide resolved

WeichenXu123 reviewed Sep 13, 2021

View reviewed changes

tests/fastai/test_fastai_autolog.py Outdated Show resolved Hide resolved

WeichenXu123 reviewed Sep 13, 2021

View reviewed changes

tests/fastai/test_fastai_model_export.py Outdated Show resolved Hide resolved

harupy reviewed Sep 13, 2021

View reviewed changes

mlflow/fastai/callback.py Show resolved Hide resolved

Fixed issues according to comments

7299a7c

Signed-off-by: jinzhang21 <jin.zhang@databricks.com>

jinzhang21 and others added 3 commits September 13, 2021 20:01

Empty commit

59dacef

use smaller model

340f564

Signed-off-by: harupy <hkawamura0130@gmail.com>

lint

d95029a

Signed-off-by: harupy <hkawamura0130@gmail.com>

harupy approved these changes Sep 15, 2021

View reviewed changes

harupy merged commit acdf976 into master Sep 15, 2021

harupy mentioned this pull request Dec 5, 2021

updated the _load_model to be compatible to fastai_v2 #3876

Closed

27 tasks

harupy deleted the fastai branch April 9, 2024 07:09

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Upgraded fastai from v1 to v2 (>=2.4.1) #4715

Upgraded fastai from v1 to v2 (>=2.4.1) #4715

jinzhang21 commented Aug 20, 2021 •

edited by harupy

Loading

github-actions bot commented Aug 20, 2021

harupy Sep 13, 2021

harupy Sep 13, 2021 •

edited

Loading

jinzhang21 Sep 13, 2021

harupy Sep 14, 2021

jinzhang21 Sep 14, 2021

harupy Sep 15, 2021

jinzhang21 Sep 15, 2021

jinzhang21 commented Sep 13, 2021

harupy left a comment

Upgraded fastai from v1 to v2 (>=2.4.1) #4715

Upgraded fastai from v1 to v2 (>=2.4.1) #4715

Conversation

jinzhang21 commented Aug 20, 2021 • edited by harupy Loading

What changes are proposed in this pull request?

How is this patch tested?

Release Notes

Is this a user-facing change?

What component(s), interfaces, languages, and integrations does this PR affect?

How should the PR be classified in the release notes? Choose one:

github-actions bot commented Aug 20, 2021

harupy Sep 13, 2021

Choose a reason for hiding this comment

harupy Sep 13, 2021 • edited Loading

Choose a reason for hiding this comment

jinzhang21 Sep 13, 2021

Choose a reason for hiding this comment

harupy Sep 14, 2021

Choose a reason for hiding this comment

jinzhang21 Sep 14, 2021

Choose a reason for hiding this comment

harupy Sep 15, 2021

Choose a reason for hiding this comment

jinzhang21 Sep 15, 2021

Choose a reason for hiding this comment

jinzhang21 commented Sep 13, 2021

harupy left a comment

Choose a reason for hiding this comment

jinzhang21 commented Aug 20, 2021 •

edited by harupy

Loading

harupy Sep 13, 2021 •

edited

Loading