Updating torch version to latest stable release - 1.6.0

[shrinath-suresh]

What changes are proposed in this pull request?

Updating torch version to recent stable release - 1.6.0.

How is this patch tested?

Verified mlflow/tests/pytorch/test_pytorch_model_export.py and mlflow/examples/pytorch/mnist_tensorboard_artifact.py with pytorch 1.6.0.

Release Notes

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

• area/artifacts: Artifact stores and artifact logging
• area/build: Build and test infrastructure for MLflow
• area/docs: MLflow documentation pages
• area/examples: Example code
• area/model-registry: Model Registry service, APIs, and the fluent client calls for Model Registry
• area/models: MLmodel format, model serialization/deserialization, flavors
• area/projects: MLproject format, project running backends
• area/scoring: Local serving, model deployment tools, spark UDFs
• area/server-infra: MLflow server, JavaScript dev server
• area/tracking: Tracking Service, tracking client APIs, autologging

Interface

• area/uiux: Front-end, user experience, JavaScript, plotting
• 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

[smurching]

@shrinath-suresh thanks for the PR! We should definitely make sure we test against the latest pytorch :) (cc @apurva-koti who was also looking at bumping test dependency versions). It looks like one of the unit tests fails with torch 1.6.0 - would you have bandwidth to take a look at why?

[smurching]

LGTM if we can fix the test breakage!

[shrinath-suresh]

@shrinath-suresh thanks for the PR! We should definitely make sure we test against the latest pytorch :) (cc @apurva-koti who was also looking at bumping test dependency versions). It looks like one of the unit tests fails with torch 1.6.0 - would you have bandwidth to take a look at why?

@smurching Sure. I will debug and let you know.

[shrinath-suresh]

@smurching Shall we update the cloud pickle version too ? Eventhough we are going to remove it from main script, cloud pickle is still used for the test script (where user wants to override default pickle argument) . Moreover, I am facing error when using mlflow.pytorch.log_model with the model trained with pytorch lightning (not seeing the error with latest version of cloud pickle). Let me know your thoughts on it.

[shrinath-suresh]

@shrinath-suresh thanks for the PR! We should definitely make sure we test against the latest pytorch :) (cc @apurva-koti who was also looking at bumping test dependency versions). It looks like one of the unit tests fails with torch 1.6.0 - would you have bandwidth to take a look at why?

@smurching Sure. I will debug and let you know.

@smurching Figured out the reason why test_load_model_allows_user_to_override_pickle_module_via_keyword_argument is getting failed.

torch 1.6 saves the model as zip file and uses _load method and in turn uses pickle_module.Unpickler to load the model.

whereas in torch 1.4 , torch.load loads via _legacy_load and uses pickle_module.load to load the model.

Test case is passing locally by setting mlflow_torch_pickle_load to mlflow_pytorch_pickle_module.Unpickler.

[shrinath-suresh]

@apurva-koti @smurching Getting some random linting failure - (mlflow/mlflow/pytorch/pickle_module.py). should i need to push any dummy change to restart the linting build or is there any other way to re-run the build ?

[apurva-koti]

@shrinath-suresh , you can re-run the build by clicking on the details link in the failed check, and pressing the "re-run workflows" button on the checks page.

[shrinath-suresh]

@shrinath-suresh , you can re-run the build by clicking on the details link in the failed check, and pressing the "re-run workflows" button on the checks page.

@apurva-koti I dont see the "re-run workflows" button on the checks page (may be due to lack of permission ??). i will try command line option to add one more commit.

[apurva-koti]

Ah sorry! Probably a permissions issue. Let me take a look at this now you've addressed the lint issue.

[apurva-koti]

LGTM. Current failures are on master and unrelated to this PR.

Thanks @shrinath-suresh !

[shrinath-suresh]

@smurching Thank you very much. Can we merge the PR if the changes are fine?

[smurching]

LGTM

[smurching]

@shrinath-suresh yes! Done :)