Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix creating local tensorflow model data fails with module import and tensorflow versioning #1205

Merged
merged 7 commits into from
Apr 14, 2023
Merged

Fix creating local tensorflow model data fails with module import and tensorflow versioning #1205

Show file tree
Hide file tree
Changes from 6 commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
26a7677
init
Apr 12, 2023
b27e6f5
lint
Apr 13, 2023
e92c2cf
add todo linear task
Apr 13, 2023
a18a90d
lower bound the tensorflow package requirement
Apr 13, 2023
98d9d6a
change tf keras test
Apr 13, 2023
e0d17db
small grammar change
Apr 13, 2023
9bcc914
todo message
Apr 14, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .github/actions/setup-server/action.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,7 +62,7 @@ runs:

- name: Install packages needed for testing
shell: bash
run: python3 -m pip install nltk pytest-xdist tensorflow
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Remove tensorflow as a requirement for setup-server since we are not testing tensorflow models curently.

run: python3 -m pip install nltk pytest-xdist

- name: Wait for server again
shell: bash
Expand Down
29 changes: 20 additions & 9 deletions integration_tests/sdk/aqueduct_tests/param_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -612,6 +612,25 @@ def must_be_image(input):
assert isinstance(image_output, GenericArtifact)
assert isinstance(image_output.get(), Image.Image)

publish_flow_test(
client,
name=flow_name(),
artifacts=[
pickle_output,
bytes_output,
string_output,
tuple_output,
list_output,
image_output,
],
engine=engine,
use_local=True,
)


# TODO(ENG-2798): Unable to publish flow using TF_KERAS parameter.
@pytest.mark.skip()
def test_local_tf_keras_data(client, flow_name, engine):
from tensorflow import keras

model = keras.models.load_model("data/tf_model")
Expand All @@ -632,15 +651,7 @@ def must_be_tf_keras(input):
publish_flow_test(
client,
name=flow_name(),
artifacts=[
pickle_output,
bytes_output,
string_output,
tuple_output,
list_output,
image_output,
tf_keras_output,
],
artifacts=[tf_keras_output],
engine=engine,
use_local=True,
)
2 changes: 1 addition & 1 deletion sdk/aqueduct/utils/serialization.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -114,7 +114,7 @@ def _read_local_bytes_content(path: str) -> bytes:
def _read_local_tf_keras_model(path: str) -> Any:
from tensorflow import keras

return keras.saving.load_model(path)
return keras.models.load_model(path)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If this is backwards compatible (works with multiple TF versions), do we need to pin the TF version in the dockerfile?

Copy link
Contributor Author

@Fanjia-Yan Fanjia-Yan Apr 13, 2023 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

After I took a look at their release log:

"Moved all saving-related utilities to a new namespace, keras.saving, i.e. keras.saving.load_model, keras.saving.save_model, keras.saving.custom_object_scope, keras.saving.get_custom_objects, keras.saving.register_keras_serializable,keras.saving.get_registered_name and keras.saving.get_registered_object. The previous API locations (in keras.utils and keras.models) will stay available indefinitely, but we recommend that you update your code to point to the new API locations."

I guess we don't need to pin it to a specific version? @kenxu95 @hsubbaraj-spiral

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can we lower bound it in the dockerfile?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Lower bounded it at the latest TF release.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actually wait does lower-bounding this actually matter? @hsubbaraj-spiral Should we just rid of the constraint altogether so it always uses the latest for now.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Wait actually I just remembered - we don't want to install tensorflow in the dockerfile right since it would take forever? I'm confused as to what the point of this change is now. I think we can assume that local data doesn't work with K8s for now, and make a task for that.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we'll want to just run the parameter operators locally as a long-term solution for K8s.



# Returns a tf.keras.Model type. We don't assume that every user has it installed,
Expand Down